AMD may have Zen4 out by the end of the year, and since AMD may be doing another 19% IPC improvement plus the benefits of DDR5 and higher clock speeds, another 40% improvement in CPU performance may be in store from AMD. We shall see, but AMD isn't talking about next generation before taking care of the demand for Zen3 based chips right now.
1) There is no need to launch because Intel is at least an entire IPC generation behind Zen3 so AMD has no competitive advantage other than to embarrass Intel even further.
2) The global manufacturing backlog (aka "chip shortage") is already making it painful for AMD to deliver CPU's. Many CPU's launched early LAST year remain OEM-only due to supply chain stress (Zen3-based Ryzen 3 4300G, Ryzen 5 4600G, Ryzen 7 4700G, etc are only available from HP and Lenovo)
3) AMD can make as much selling Zen3 through the remainder of the year and well into next year as they would with Zen4. Economic realities have demonstrated AMD is still selling the Zen 2 3200G (one of their best selling CPU's for 2 years straight) and it is still selling well. Sure, part of this is because its the only Zen APU commercially available since the previously-mentioned Zen 3 APU's are OEM-only, but it isn't stopping anybody from buying it and honestly for the low-end price segment those systems fill, nobody is going to notice a difference between a 3200G and a 4300G. Those are $400-$500 PC's. Although the 4700G is a beast of an APU - it's essentially the performance of a 3700X along with decent integrated graphics.
I believe AMD doesn't see it this way. They are going to maintain their execution to capitalize on Intel misfortune. Anyway, Zen 4 is 5nm, so there is no competition for waffers anymore. It is going to give all 7nm products an increase in capacity just by moving the CPUs to 5nm.
Not that AMD can capitalize much, plain and simple AMD have not capacity in 7nm and even less in 5nm. Zen 3 is a ghost, Zen 4 will be a ghost. Intel enjoy the absolute dominance of "volume". Funny enough AMD ruined itself leveraging on tight supply processes. Instead of Intel pumps $ billions and billions supplying the channels with big quantities of manufacturable SKUs. Looking past decades AMD have always done this issue, i remember the SOI saga. AMD was unable to destroy Intel with K8 only because Intel "volume" was massive.
Really ? It will likely take a year to figure out the product stack. Core i9 has ten cores, right? Or was that 9th Gen? Was that glued together? Which Gen combusted spontaneously? Was that Rocket or Cannon? Was that on 10nm, with the AMD GPU?
With RL new uarch integration, it gives a bad impression about what Alder lake will provide. You need to remember that 10nm is not much better than 14nm... and you get frequency regression... bad yields... Intel is screwed big time, but they don't have the choice to avoid investors lawsuits with their 10nm promises, and 14nm has nothing left to extract from it. The hard reality is that they are trap where they are until 7nm, which is supposedly worst, came into play.
Same applies to all processes around the world. Even TSMC processes show a clear frequency regression. The voltages to have a given number of Ghz are crazy on an AMD SKU. All these voltages are out TSMC official specs. In short words actual top level Ryzens are SKU with a small life cause electromigration under turbo operations. The real clock speeds allowed by 7nm TSMC are only on Epyc line, and they are a lot lower to give reliability during time. 4 GHz is a clear wall for foundry 7nm, 14nm can run wel over 5 Ghz without exit out the max allowed voltage long period of time.
Already done, Zen3 is in the same IPC family of Willow Cove core. Golden Cove will be at the level of Zen 4......but a lot early .......... This time AMD is in a great dilemma, a medium volume Zen 3+ or a real ghost Zen 4.
Roadmaps suggested Zen 3 refresh before Zen 4. I wouldn't be surprised if Zen 4 came late summer or holidays 2022, with mass availability in early 2023.
I wouldn't call it shit. I mean, if they sold them for reasonable prices I'd bite just because the platform quality is traditionally more reliable with Intel. The problem is the i7-11700 is outclassed by a $200 AMD CPU so it isn't worth anywhere near $300. Intel's entire CPU stack is effectively overpriced by 30%.
Release CPUs that are actually available for sale, and at very competitive prices? 5800X - 8C16T no iGPU - $449; i7-11700F - 8C16T with no iGPU - $298. It's a massacre of the red team, at this point.
yes buy that intel cpu for sure because of the integrated graphics :) and the AMD are out of stock because nobody wants them :) they all prefer the Intel :)
Previous generations were all on 1 die, so intel would have had to make a seperate die to not have an iGPU. They did, it was called HDET, and people whined it was expensive.
Now intel Xe is on 10nm while the cores are 14nm, but I believe they are still under the same die package.
There are actually quite a lot of applications that tax CPUs heavily but don't really need much GPU (but do need some, because anything with a UI without a GPU is not fun). Also, given the GPU shortage, rolling with the iGPU until the card you want is available at a reasonable price isn't a terrible idea either.
And they didn't launch one SKU that might actually be interesting, a 4c8t i3 with the full UHD 750 graphics. Which might actually steal some sales from AMD's unobtainium APUs.
How long will they be able to sustain those "huge profits"? Not for long, semiconductors are expensive and you can blow 5 billion USD in a quarter if you get unlucky.
They just had another record year while pushing outdated 14nm skylake crap in the face of zen 3. Intel isnt going anywhere.
Yesthey are having problems, but it would take another decade+ for intel to really start suffering at this rate, and easily 2 decades to go out of business if they somehow never put out another CPU design.
Intel main problem is not AMD, it is TSMC. It will force them to update their fab strategy, because they lost their edge for good there. As long as TSMC have better nodes, Intel will always be trailing behind. And if theygo with TSMC, they will never get the capacity required for maintaining their business model. Don't lure yourself, 14nm is done and Intel doesn't have anything to turn this around. There is massive risks arising for them in the next 5 years... massive risks...
Intel needs to partner with TSMC and give them access to their US Fabs in exchange for their engineering prowess.
Also ASML needs help to increase manufacturing output. They are the sole provider for leading edge nodes and cannot keep up with industry demand. That demand is only increasing and I fear they will fall further behind. Current shortages might be our new normal unless major changes happen across multiple industries.
I don't even think TSMC is their main problem. From what I understand, reading all these articles over the past few years, is that at 10nm or 7nm your complexity starts to suffer. You get quantum leaks, or rogue electrons. AMD made a very clever move by going to chiplets and combining them, keeping the designs relatively simple and yields high enough.
Intel has a good 10nm process, just not for the incredible complexity of their monolitic Core-design. Simplified versions for laptops give good yields, though, if I am correct. So, as soon as Intel moves to chiplets, connecting them together, their 10nm should be sufficient. For now Intel is their own worst enemy. And if I understand it correctly, their Alder Lake design is a Big/Little one. Maybe even coming with their own version of chiplets, or something.
I am an absolute AMD fan, because I understand how Intel rigged the market, back in 2006. So for me personally, they deserve to go down. But that's personal. Professionally I think Intel will come back soon enough, to my regret.
And if I understand it correctly, their Alder Lake design is a Big/Little one. Maybe even coming with their own version of chiplets, or something. " alder lake, is X big cores ( current cpu cores) with X small cores ( atom, or arm cores )
guess old Intel is going to pack up shop and exit the CPU business. What a ridiculous comment. Nerds care about benchmarks. Granny buying billy a new computer at christmas doesn't know what an AMD is, but she has heard of Intel since the 1960's. That and Intel's OEM partnerships will keep them in business a long time.
It's more likely, granny will take savvy grandson to the PC retailer with her, and allow him to choose, methinks, i'll it be a nice shiny desktop, with the nice green AMD logo (or is it red?), great choice, X5950 with Nvidis 3090, 32gb ram, 1Tb of NVME m.2. Excellent choice! Er, grannies paying...!
"Just joking, gran. It'll be a Ryzen 5 3600, along with a GeForce---" and so Billy recited the rest of his components and the two leaving the shop hand in hand, went home smiling.
"Must it have that G, what do you call it?"
"GeForce. Yes, gran, it must. Makes the games go fast."
"I see, my boy," said Granny, sighing and smiling, "I see. Well, dear, I hope your games will go *very* fast."
TSMC’s present is 5nm. Intel’s present is 10-14nm. Part of Apple A9 production (from Samsung), FIVE generations of processors ago, were etched in 14nm. Intel is indeed finished unless it can jump to another node very quickly. Else, it is now a second tier company and if you have any money in it, you should look for a way out. I can safely say that with the M1 on the market, and the M1X and future generations coming, that I will no longer buy any Intel processor for potentially the rest of my life. This is a mess...
You are acting like Apple is a trusted cie when it comes to security. Even after Intel hardware flaws tsunami, I still trust mroe Intel on that aspect than a brand new architecture relying on emulation for comon softwares on X86 than a brand new uarch put together on the go with ARM cores, which are democratize and full of design flaws.
Not an Intel proponent, but they can always strike back and are far from finished. Remember AMD, who was close to bankruptcy: they struck back with force in 2017 and turned the tables round in a matter of four years.
"Cypress Cove, by most measures, is a reflex response to a widening gap in Intel’s desktop roadmap,..."
That's a pretty strange statement, since you know how many years go into these type of projects and the number of approvals they would have to go through at Intel to get funded.
It is obvious to me that 14nm processing enables higher clock speeds than have been achieved on Intel's 10nm or on TSM's 7nm AMD parts. That is apparently a market opportunity that Intel can go after... and why shouldn't they? At the time they made this decision, they likely didn't know that 10SF would be what it is, or that 10esf would be available so soon, but even now it looks like a good use of their available 14nm capacity while they are still in the process of repurposing their 14nm fabs for 10nm.
Clock speed as AMD has demonstrated with the Zen architecture is secondary to IPC. Higher clock speed tends to equate with excessive heat, thus creating further need for exotic heating solutions and more costs to the consumer. Lessons need to be learnt from the 900 lb gorilla and there's been too much monkeying around lately!
Intel can't increase IPC further until Golden Cove arrives in Alder Lake. Alder Lake is not ready NOW, but should be quite performant and efficient. Between the choice of lower IPC higher frequency Skylake, and higher IPC lower frequency Sunny/Willow Cove, Intel chose the the third route of higher IPC and frequency Cypress Cove.
Given that 10SF is operating well, Rocket Lake is an absolute power hog, and Alder Lake is ready in a few months, it is difficult to say that this was the best possible decision. But, given the state of 10nm 12-18 months ago (when Cypress Cove would have been made), it was likely the right call to make and uses existing 14nm capacity efficiently as Intel transitions to 10nm.
It's very simple: given the supply shortages plaguing the zen3 so far, Intel only needs to bring anything to the market that is somehow competitive (and be it only single core performance), produce it in quantities and call it a day.
So true! Intel will print money with these and Also get bigger market share, just because They have a product. It don`t have to be fast, it don`t have to be pretty. It just needs to exist!
can anyone tell me whether sapphire rapids will launch at the end of the year? techpowerup is saying that it should. their timelines seem accurate as they're claiming alder lake should be due out in sept 2021. i'm after a dual socket setup (xeon) so consumer dates aren't necessarily of interest.
Not unless someone is an Intel employee, no. They're supposed to, but no launch dates for Sapphire Rapids (or even Ice Lake SP, which is destined to be on top of Intel's product line for about as long as Rocket Lake, which is to say certainly less than a year, and maybe as little as six months) have been announced yet.
Depending on your use-case, the heat isn't much of an issue as long as it's faster. Yeah, a big tower cooler like the NH-D15 is a bit ridiculous at first glance, but if it means that the CPU is overall faster for your specific need, then it's definitely worth looking at.
That said, a Ryzen 9 5950X has a Boost clock of 4.9 GHz, while a 11900K only goes a little higher, to 5.3 GHz (with a 4.8 GHz all-core boost), and only has 100 MHz more base clock.
As usual: Benchmark your use case. For my main work PC, I need the best single threaded performance, which is why I went with a 9900K over a Ryzen 9 3900X in Mid-2019. But if I was refreshing my machine right now (and if supplies were available), I'd definitely look at benchmarks over fanboism - the Ryzen 5000 is a darn fine CPU.
There is a difference of a bit higher power draw and almost double the power draw. Ryzen chips, going all the way up to the 16 core Ryzen 9 5950x have a 105W as the official TDP, but generally goes to around 125W of CPU power draw under normal circumstances. These Intel chips are going to 250W to get that 5.3GHz speed that is needed to stay ahead of AMD.
If we were only talking 25-50W of additional power, most would agree with you, but honestly, these Intel chips are pulling a LOT of power and it isn't even to get a full lead across the board. The reality is that Intel needs 200W and above just to get a lead in one out of 8 areas.
Yeah. Intel is basically overclocking these chips to the max right out of the box to compete with Ryzen single core speeds. They loose against Ryzen multicore. reviews I've seen show rocket lake Temps up to 98°C. Intel needed to limit the number of cores to keep the thermals within the higher limits they set. I'll be interested to see if they have issues with more failures. I'd still buy Ryzen since it gives better performance for multicore and is over 4x more efficient.
200W power draw is not sexy, but it's manageable just fine with a big air cooler or AIO. (Also see Threadripper with it's 280 W TDP - perfectly manageable and worth it for a lot of use cases) The question is whether you get a speed boost, and whether the speed boost is worth the extra cost. For a personal system maybe not so much, for a work machine that runs stuff limited to single threads maybe more so.
For me, the 9900k was worth it over the Ryzen 3xxx, but I'm not sure I'd pick the 11900K over a 5950X - but I will surely take a close look at both for the actual use case I have.
You may well just run with 25-50Watts extra on your games, because those tend to stress your 300Watt GPU much more than your CPU cores.
I had a Kaby Lake i3-7350K dual core running at 4.8GHz turbo next to ab E5-2696 v3 18 core Haswell running at 4GHz turbo: Guess who was faster in games but lost big time on CPU render benchmarks?
That was replaced eventually by a Ryzen 5800X which now delivers pretty much identical multi-core results as the Haswell Xeon and incidentally both use the same 110Watt of power. Only the Haswell chip is vastly bigger and would have cost 10x originally.
So even that 22nm 18 core Haswell still uses half the peak power of an i9-11700K, because it runs its 18 cores at only 2.7GHz, but it's still not a good gaming CPU.
And the 4.9GHz 5800X really isn't significantly faster than the Tiger Lake NUC with an i7-1165G7 on its top on any single core load, which run at 4.7GHz there for practically identical IPC and per-core TDP.
If yields and capacity for 10nm were plentiful, Rocket Lake would not exist.
But I have no doubt it will find willing buyers while the real environmental damage is much less than Bitcoin/Ethereal mining.
I'd guess that by the time enough games could use all 8 cores of Rocket Lake or Ryzen 3, the systems you built today will be out of date. In the mean-time their CPU power consumption will only be a small concern to the people who buy them.
Well, my 5950X has 5 cores boost to 5 GHz out of the box, with a max power usage of 135W. Idle, it consumes the same amount of power as my i7 6800K with 5 cores clock at 3.6 GHz and one core clocks at 3.8 GHz. In every usage like web browsing and watching YouTube, it actually consumes less power than the 6800K.
This was actually demonstrated many times by both intel and amd. The last example of this was core 2 duo or sandy bridge vs bulldozer. Intel had to push these chips to 5ghz cause they really didn't have any other alternative.
JayNor, spoken like someone who either is paid by intel, or believes their PR bs. its obvious to most, that 10nm is not where intel needs its to be right now, and may never be. the only market intel can go after now is high midrange, or lower. rocket lake looks to be still a dud, even after all those pro intel fanboys that were crying about AT using an old microcode, so they updated with a new code, and it STILL didnt do what those same fanboys were saying it would. intel has no solid answer to zen 3 yet. instead, itel releases how many rocket lake skus ? 14 ???? that is crazy, and just confuses people. face it, no amount of PR from intel, is going to help, if amd was able to make enough zen 3 cpus, intel wouldnt be selling much, but intel is, only cause its all that is on the shelf. most i know, arent even looking at intel, and are waiting for zen 3.
Intel is sampling Alder Lake, Sapphire Rapids and Xe-HP 10nm parts, and has been shipping Tiger Lake, Ice Lake and P5900 family chips. It is time to lose the fantasy that Intel can't build 10nm parts.
time to loose the fantasy that intel can build anything faster then what they have now, and that they dont use gobs of power. no matter how you spin it jaynor, rocket lake is a dud, 10 nm isnt shipping in anything meaningful, and only in low volumes
I agree, Intel's 14nm clocks are edging up a little with Alder Lake. I saw a benchmark today that listed an Alder Lake running at 4.6GHz. Rocket Lake runs at up to 5.3GHz according to the release, and they have provided new overclocking capability in the memory controller.
Intel claims to have shipped 115K Ice Lake Server chips so far, and it isn't even launched.
Tiger Lake 10nm shipped in 150 products prior to q1 cc. In 1H2021 they are adding TGL-vPro, TGL-H35, TGL-H45, TGL-EVO Chromebooks. Intel says they shipped around 30% more 10nm parts in 2020 than they estimated at the first of the year, so looks like 10nm is ramping well.
Look, you can go to Dell's web site today and order a $500 Inspiron with Tiger Lake in it and get it delivered within a week. To argue they're not producing 10nm chips in volume at this point is silly.
maybe, but dell has always been intels best buddy, and gets cpus more then any one, i still cant go into any local store here and actually see any 10nm laptops IN STOCK. saying you can go to 1 company and get them, vs any store, is NOT in volume, nice try though.
Qasar. I work for a very large company and we had to put a company wide lifecycle refresh on hold due to lack of laptops. We use Intel, AMD and Apple. The entire supply chain is stressed beyond its capabilities and Intel's 10nm woes of the past have little to do with that. We simply cannot get laptops from EVERYONE fast enough.
The board of directors is there to represent the investors and they take that seriously. If Intel lied about that their investors would know. The media would know. The public would know.
I got the proof right in front of me i7-1165G7 and Ryzen 5800X on top of each other with pretty idential single core performance and comparable power efficiency.
There is zero doubt in my mind they could produce an 8 core part that would be close to identical to the 5800X in terms of power and efficiency, yet they chose otherwise.
I doubt we'll ever know the full story, but I do not doubt that they analyzed the situation carefully and actually got much luckier (TMSC capacity constaints), than they might have otherwise.
Their 10nm isn't good enough to handle desktop pet envelopes, which is why they are only offering 10nm in server and laptop parts. This may also have to do with why ADL will only have up to 8 "big cores".
"These could all be added to any other motherboard, AMD or Intel, with discrete controllers which are slightly more expensive – those controllers don’t have to be Intel either. "
So, what advantage, other than the slightly less expense, does Intel provide by supplying them? Is it that Intel is also responsible for providing the drivers? Or perhaps that they reduce the number of pcie pins that would be required to interface with external discrete implementations? Or perhaps there is a thermal advantage to having the digital portion on chip rather than running high speed signals externally?
An and, why don't you make an article about Intel's Flex IO most people don't understand how it works.
Their chipsets can only provide 16 PCI-E lanes max, usually boards will deliver less than that. But their marketing department still list their chip sets as delivering "up to 24 PCI-E 3 lanes", while their technical documents list them as Flex IO lanes.
In reality its more like 14 PCI-E lanes, 8 SATA, 6 USB 3. With certain configurations only 12 Lanes available for m.2, and that number drops further if the manufacturer wants more than 6 USB 3 ports
Intel hasn't posted datasheets for the (desktop) 500 series PCHs yet, but the HSIO lane multiplexing diagrams from any of the datasheets for previous generations explain the situation pretty clearly.
The Z490 PCH, for instance, has 30 HSIO lanes. 6 are USB3 only, the other 24 can be PCIe. Of those 24 lanes, 4 can be used instead as additional USB lanes, up to 6 can be used for SATA, and one can be used for the integrated GbE.
Yes, OEMs have to make trade-offs, but some designs have no need for more than 6 USB3 ports or any SATA ports whatsoever and can utilize 24 lanes as PCIe without issue. At least with Rocket Lake Intel finally moved to DMI 3.0 x8 so the connection to the host CPU won't be quite so hopelessly oversubscribed.
I like the story of this launch. Although the CPU performance is a bit underwhelming and the power consumption is a bit ridiculous, we can't deny AMD CPU prices have been creeping up. I think competition is good. The 8 core i9 KF part, 70$ MSRP under the 5800X? It's worth a consideration, surely.
OK, need to pay the Intel tax on motherboard, which will probably be obsolete for thext CPU generation, so the CPU pricing advantage might evaporate when considering the platform cost. Intel has its work cut out. We can only hope AMD will respond and bring back down the 6-core prices under 200 and 8-core under 300$, where they were during Ryzen 1 and 2, when they were still the underdog.
"We can only hope AMD will respond and bring back down the 6-core prices under 200 and 8-core under 300$, where they were during Ryzen 1 and 2, when they were still the underdog "
why would they ? amd has the better performing cpu, and they should charge a little more. supply and demand has raised those prices higher, but still, if anything, intel should be charging less then the amd equivelent, only then would intel be a consideration, much like how it was when the performance role was reversed.
AMD won't bring the prices down ... They want to make money like intel wants. I find it very funny how people thought amd will always keep the 200 bucks prices when amd was basically forced to have those prices in order to gain market share. Now that they have the market share they can milk it. The kf parts are very interesting indeed. I would argue the i7 kf is better cause it is just an i9 with fewer megahertz.
I used to read every word of the review for each new Intel generation launch. Now I can't be bothered. It's all more of the same and probably a 5% performance increase.
AMD released Ryzen 1 in Feb 2017. 4 years later and Intel are still vomiting more or less the same crap out. Yawn....cue Anandtech to post ANOTHER article about the super amazing world changing performance blitzing CPU's that Intel are going to release in 5 years.
Twice the cores, higher clock speeds AND 20% more IPC, 25% more PCI-E lanes from CPU, twice the bandwidth between the CPU and chipset, much better integrated graphics. If you put the 7700K and 11700K side-by-side, for around the same price, the 7700K would get murdered. So it's kind of bad faith to say that they're spewing more or less the same stuff out.
Ummmm Intel released their first 8 core CPU in 2014. The highest number of cores they are releasing now is........8. Who cares about the graphics? People who care about graphics will buy a discrete GPU. Yawn run along Intel puppy dog.
I'm comparing the best of mainstream platform of feb 2017 (socket 1151) vs best of mainstream platform 2021 (socket 1200), and that's at around the same price bracket of 350-400 USD. To get more than 4 cores from Intel in feb. 2017 you needed to buy into the HEDT or server platforms (socket 2011).
Pneumothorax, not according to the intel fanboys. intel was nice by sticking us at quad core all those years so amd wouldnt be forced out of business, cause they couldnt compete then.
yes, because Skylake is 2015 and Ice Lake is technically a 2019 part. just because Intel kept making Skylake until 10th gen doesn't automatically make the 19% ipc uplift amazing - keeping on mind what you are comparing against.
also keeping on mind Zen 3 was 19% ipc from Zen 2, in one generation - and AMD was playing catch-up this whole time. that's more impressive than what Intel's done, quick is slapping a 2019 arch onto an old node and calling it something "new" because it's faster than an arch from 2015. no duh.
20% more IPC? Since when? IPC has been stagnant since skylake. And all that IPC does jack shit int hebenchmarks where the rocketlake CPUs occasionally lose to comet lake and consistently lose to zen 3.
Integrated graphics? Intel fanbois constantly whine about those.
You can connect a SSD to your CPU direct now. Congrats on getting to where AMD has been for years.
What's almost more interesting than the technology (which seems to have very few really unkowns left), are the supply chain struggles behind the scenes, consumers never really cared about in this CPU domain before.
Judging by my Tiger Lake i7 NUC and my Ryzen 5800X, both CPU core designs and manufacturing processes seem so very close to each other, few could probably tell a double sized Tiger Lake APU and a Ryzen 5800U apart.
But it's the foundry capacity constraints of TMSC and the yield issues with Intel's 10nm which are really the decisive factors in this battle with it's long term implications which come simply from market share or the size of the eco-system, not the technology details.
And this battle is fought on so many distinct fields, even if the manufacturing capacities are largely global (especially with the AMD CCDs). We hear about 10nm big Xeons and notebook SKUs being produced en masse, while today we reminded how Intel would rather resort to backporting a whole design than yield the desktop and we mostly wonder if it's done because the psychological impact of being a loser on the desktop was considered a risk big enough to do this or if Rocket Lake quite simply is commercially viable the way it is.
Got NUC8 (48EU Iris 655+), NUC10 (24EU HD620) and NUC11 (96EU Xe) and an Iris 550 Plus (48EU) on a Skylake i5 here.
From the various benchmarks I've run I have good reason to believe that the 32EU Xe will perform very close to the 550 Iris Plus with 48EU, which is certainly better than all HDxxx GPUs since Skylake, but certainly not twice as fast as 48 vs 24 EU would imply (the 550 GT3e with 64MB eDRAM wasn't either).
That the 96EU on Tiger Lake without eDRAM manages to achieve a significant boost vs even the 655 Iris plus with 128MB eDRAM, is quite remarkable, but in absolute terms still might only satisfy the seriously bored without access to a gaming rig.
Expect perfectly reasonable 2D and media performance at 2x 4k resolution, good-enough Google-Earth 3D with Chrome and nothing but disappointment in gaming for these desktop SKUs.
It is obvious that this is just a stopgap for intel. A stupid decision that was made a few years back and now..it is too late to be cancelled. I also think that the current situation is favorable because intel has its own fabs and can made as many as the market asks compared to amd which is at the mercy of tsmc. So I think these will actually sell well with the 11700k looking like a good option.
Of course it's a stopgap. Pretty much everything Intel's launched on the desktop since Skylake has been a stopgap (though pre-Ryzen 3, they were more than Good Enough stopgaps).
But that doesn't mean it was a stupid decision (though if Adler Lake desktop parts launch this year and availability is good -- and that may very well happen -- it may have been a wrong one) given the state of Intel 10nm at the time they would have decided to make it, and would have been a better decision if they'd done it sooner.
Thanks Ian! From my POV, the most attractive offerings of Rocket Lake are some of the i5s, assuming they're available and sell at (or below) MSRP. Best/Most bang for the buck in the whole lineup, IMO. Also, These 6 core chips should have enough extra oomph to convincingly separate from the Comet Lake refreshed i3, which otherwise are currently "it" for entry-level desktop CPUs.
Question about the Rocket Lake i5s: do the highest-priced i5s with iGPU really have the same 32 EU Xe configuration as the i7 and i9? The table suggests they do, but your article says otherwise; can you clarify? Thanks!
Lastly, about the lower 400 series Chipsets and why they don't support PCIe 4.0: I believe that has less to do with them being fabbed in 22 nm, and a lot more with those being basically rebranded 300 series chipsets. And those were old years ago.
What would make me consider Intel over AMD currently is the integration of both a 5Gbps (or better 10Gbps) NIC and TB3/TB4. X570 mobo with integrated Thunderbolt are scarce and therefore expensive. On the other hand you can get an AM4 with working ECC RAM.
I wish someone finally sold a 10Gbit NIC operating off a single PCIe 4.0 lane: I don't like sacrificing 4 lanes with an Aquantia AQC107 designed for PCIe 2.0.
But for point-to-point TB3 peer-to-peer networking is quite ok and potentially much cheaper, bridging for bigger networks a bit messy and I was able to get 1Gbyte/s throughput only with 32k MTUs.
i3-10325 looks like a great cheap chip for audio streaming where single core performance is the goal.
I will just wait for an AMD APU or the next Intel as long as I don’t need water cooling.
Have to say I’m disappointed at Intel’s lack of a fierce response. Thought for sure they had designs they could reply with. This is just another 14nm with slightly hipper marketing jargon. Hope they don’t build too many.
Jesus. Why doesn anyone even bother covering this pathetic crap? (Yes, obviously because it’s still Intel, the biggest player in the sector) Intel shuffles around a couple 100MHz spec bumps and calls it an entirely different product segment? Seriously? Tell me one single difference between their i7 and i9 lineup that isn’t a 100MHz base clock or 2 millisecond turbo boost peak that will immediately throttle because it took 250 Watts to get there? This is dumb and at some point, people need to stop giving Intel free press coverage for this bull$#!+.
Intel has announced the end of its Performance Tuning Protection Plan (PTPP). An end-user who previously bought a PTPP from Intel was guaranteed a one-time replacement CPU if they fried their chip by overclocking it, provided the chip was still within warranty.
The program has existed since the Sandy Bridge era, but Intel is bringing it to an end, effective immediately.'
Hruska positions this move as the natural result of overlocking becoming less relevant since CPUs are more optimized but I think it's hardly coincidental that Intel's CPUs are now shipping with AVX-512 and edgy power usage even when that's not running.
So, these hot and loud things that are slower than the competition (especially at the same PPD, performance per decibel aka performance per watt) — also have the dubious glory of no longer being covered by Intel's overclocking warranty.
Intel should ditch integrated gpu for desktop processors and move it to the chipset. That several hundred million transistors reduction will help them to increase yields hence reduce price
Rocket Lake is on 14+++, not 10SF. How could yields possibly get better? Intel 14nm is nothing if not a mature process at this point. It's practically geriatric.
The USB section in the chipset comparison table should be updated. As indicated in the official Intel slide included in the article:
up to 3 USB 3.2 Gen 2x2 ports up to 10 USB 3.2 Gen 2x1 ports up to 10 USB 3.2 Gen 1x1 ports
Really, that should be one row in the table—Max USB 3.2 (Gen 2x2/Gen 2x1/Gen 1x1).
The fact that OEMs can choose to add a third-party discrete USB 3.2 Gen 2x2 controller to their products is not a relevant marketing point for Intel's platform controller hub. It's not integrated into nor a feature of the 400 series chipsets in any way. At least with things like discrete Thunderbolt controllers, Intel actually produces those chips themselves and validates them specifically for use with the platform.
'276-292 W peak power draw during AVX-512, and motherboard manufacturers often default to unlimited turbo in any event. Users going after any Core i7 or Core i9 product should look to a good air cooler or liquid cooler to get the best thermal performance here.'
How many 'good' air coolers can handle that kind of wattage, and at what decibel level?
This is a very serious issue. Not only can people find themselves with inadequate cooling, to get adequate cooling they can end up with hearing loss and/or tinnitus aggravation and/or very unwelcome noise pollution.
Noise pollution and hearing damage aren't trifles.
I've never heard a desktop CPU cooler produce that kind of sound. These will not be using passive heatsinks with fan modules blowing air across them like you see in rack-mount servers. HP installed some pre-built blade racks in my data center before that were so loud they physically hurt your ears. They had to correct the issue via a firmware update.
I think you are exaggerating the sound levels desktops can produce.
'I think you are exaggerating the sound levels desktops can produce.'
An Anandtech review of the 'Snow Silent' power supply by Seasonic said it could be heard 'from rooms away'.
There is plenty of exaggeration and it's calling things like that 'silent'.
Please show some data to prove that 292 watts of power isn't going to require very significant noise. You also have to remember that many will have a GPU producing a ton of heat, too.
'Second is the comparative AMD processor, the Ryzen 7 5800X, which has 8 cores and has a $449 SEP. If both processors were found at these prices, then the comparison is a good one – the Ryzen 7 5800X in our testing scored +8% in CPU tests and +1% in gaming tests (1080p Max).'
Apples to apples. How much faster is the Ryzen at the same wattage, the same power usage? Performance per watt.
'The Ryzen is very much the more efficient processor,'
I believe I recently read an Extremetech article that said Intel has dropped its overclocking warranty program or something to that effect. That is something that is relevant enough to mention with all this overclocking information you posted.
Given how demanding these things are, what kind of warranty Intel is offering is much more pressing than in the days of quad Skylakes being the standard.
To follow up with my point about the noise... Why not write an article that shows two systems with comparable gaming performance — one using the quietest 'available' (in quotes since GPUs aren't available) parts (i.e. AMD CPU and quietest high-performance GPU or GPUs) and another using Intel CPUs + the loudest GPU or GPUs of comparable performance to the quietest options?
This would be a very useful article — to give people an idea of how much they will need to spend to greatly reduce the noise pollution. Gaming is an experience. Unless one is deaf it's a big drawback to be gaming in what sounds like a tornado.
Then, of course, there is also the power usage issue — both in terms of the power bill from the machine and the cooling bill. And, sometimes the added cooling demand means more noise from that as well. Not everyone has a fancy HVAC system. Some have to make due to window ACs, which are noisy even on the lowest speed.
There is no good reason to gloss over these things. PPD (performance per decibel) should be a featured comparison.
nice one from gigabyte... call your motherboard VISION with the latest EOL Intel chipset and EOL CPU generation :) By the end of year (lets hope for intel but probably 2022) the next cpu generation is yet again another socket and the 11th series is not worth buying vs 10th or AMD zen3 .... VISION :)
tbh my real question is not regarding intels silicon but truly there instruction set, how much money will intel be loosing when or if Nvidia buys arm and it starts to become the mainstream IS "instruction set" and my other question actually relating to there silicon is one why the actual F do we need "power efficiency cores" on a desktop system the only thing i can understadn for that is better idle power consumption or web browsing shit, part two, WHEN WILL THERE ACTUALLY BE A COMPETITIVE ARCHITECTURE, intel can totally make a decent design so where is it. AMD has made improvements on navi to navi 2nd gen just with architecture so it can TOTALLY be done "mad face emoji" :..(
I suppose Intel Haifa is working on the successor to Sunny Cove as we speak. Hopefully, they've started from scratch, rather than just iterating on SC. Copying AMD's split-scheduler design might be useful, too.
As for power efficiency cores, lower power always pays off at a later time or in other ways. Having said that, I'm sceptical of Intel using this fast/slow approach.
Updated TDP versus CPU frequency estimates based on AVX2 all core with no PL1, PL2 or Tau limits ...
I am currently guessing that the i9-11900K at full 5.3GHz all core OC (if even possible) will pull close to 400W in AVX2 type benchmarks (with no AVX type offsets applied), similarly the i7-11700K at full 5.1GHz all core OC (if even possible) will pull close to 300W in AVX2 type benchmarks (with no AVX type offsets applied) ...
View: https://i.imgur.com/4QvuhpH.png Notes: Power guesstimates apply to the total CPU package only (e. g. not total system power consumption at the wall). More than adequate cooling solutions to keep maximum temperatures at or below 100C (e. g. Tjunction) for either 300W/400W CPU power draws during AVX2 operations.
How many years before Intel strikes back? Lucky for them old i7’s and Xeon’s were great CPUs and lasted me into 2021. They need a new design this year before they lose more customers to the innovative designs from AMD. Lucky for them AMD APUs aren’t readily available.
I have serious question...... I just ordered a Alienware Aurora r10 /w GTX 3080 and ryzen 7 5800x (over the i7-11700f) Benchmarks were mostly identical. I went with the ryzen mainly because their supposed to be more efficient. But the TDP of the 11700F is 65 watts and the 5800x is 105 watts. The AWr-10 case is known for heating issues (I will be replacing the fans with the more efficient quieter noctuas) but still want to have a processor that makes the lowest heat. I will be gaming in 1440p ultra wide and for hours on end usually. I'm only worried about heat. I'm not a fanboy so don't care if I have red or blue, and have just come over to PC gaming after 15 years of Xbox. I just want to keep my machine as cool as possible. Any help is apreciated thank you Mack
TDP means nothing nowadays. AMD probably keeps closest to the rated TDP, while Intel is the worst offender for breaching the rated TDP when it comes to real world usage. Those competitive numbers you see in reviews, they usually results in the processor pulling in at least 2.5x the rated TDP for an i9 processor.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
156 Comments
Back to Article
meacupla - Tuesday, March 16, 2021 - link
So AMD is kicking Intel's ass at this point and what does Intel do?Launch 30 SKUs, of which at least 50% are redundant.
Arsenica - Tuesday, March 16, 2021 - link
12th Gen (launching by year's end) will have even more SKUs.Targon - Tuesday, March 16, 2021 - link
AMD may have Zen4 out by the end of the year, and since AMD may be doing another 19% IPC improvement plus the benefits of DDR5 and higher clock speeds, another 40% improvement in CPU performance may be in store from AMD. We shall see, but AMD isn't talking about next generation before taking care of the demand for Zen3 based chips right now.Makaveli - Tuesday, March 16, 2021 - link
I don't think you will see Zen 4 this year. Probably a Q1 2022 launch.Samus - Tuesday, March 16, 2021 - link
Definitely next year for Zen4.1) There is no need to launch because Intel is at least an entire IPC generation behind Zen3 so AMD has no competitive advantage other than to embarrass Intel even further.
2) The global manufacturing backlog (aka "chip shortage") is already making it painful for AMD to deliver CPU's. Many CPU's launched early LAST year remain OEM-only due to supply chain stress (Zen3-based Ryzen 3 4300G, Ryzen 5 4600G, Ryzen 7 4700G, etc are only available from HP and Lenovo)
3) AMD can make as much selling Zen3 through the remainder of the year and well into next year as they would with Zen4. Economic realities have demonstrated AMD is still selling the Zen 2 3200G (one of their best selling CPU's for 2 years straight) and it is still selling well. Sure, part of this is because its the only Zen APU commercially available since the previously-mentioned Zen 3 APU's are OEM-only, but it isn't stopping anybody from buying it and honestly for the low-end price segment those systems fill, nobody is going to notice a difference between a 3200G and a 4300G. Those are $400-$500 PC's. Although the 4700G is a beast of an APU - it's essentially the performance of a 3700X along with decent integrated graphics.
eva02langley - Wednesday, March 17, 2021 - link
I believe AMD doesn't see it this way. They are going to maintain their execution to capitalize on Intel misfortune. Anyway, Zen 4 is 5nm, so there is no competition for waffers anymore. It is going to give all 7nm products an increase in capacity just by moving the CPUs to 5nm.Gondalf - Wednesday, March 17, 2021 - link
Not that AMD can capitalize much, plain and simple AMD have not capacity in 7nm and even less in 5nm. Zen 3 is a ghost, Zen 4 will be a ghost. Intel enjoy the absolute dominance of "volume".Funny enough AMD ruined itself leveraging on tight supply processes.
Instead of Intel pumps $ billions and billions supplying the channels with big quantities of manufacturable SKUs.
Looking past decades AMD have always done this issue, i remember the SOI saga. AMD was unable to destroy Intel with K8 only because Intel "volume" was massive.
Qasar - Wednesday, March 17, 2021 - link
ahh gondalf, love your pro intel bs.how do you like the fact that rocket lake, looks to be a dud ??
"AMD was unable to destroy Intel with K8 only because Intel " kickbacks " was massive." there fixed it for you
scineram - Saturday, March 20, 2021 - link
Cope.eva02langley - Wednesday, March 17, 2021 - link
Forest kind of admit that Zen 4 is coming in 2022. It was in the QA interview posted this week.Gondalf - Wednesday, March 17, 2021 - link
Likely 2023 real volumescineram - Saturday, March 20, 2021 - link
Cope.yeeeeman - Tuesday, March 16, 2021 - link
Zen 4 will come next year. Q2 at best. So intel, if it launches alder lake by end of this year, has some time to actually catch up with zen 3Smell This - Tuesday, March 16, 2021 - link
Really ?
It will likely take a year to figure out the product stack. Core i9 has ten cores, right? Or was that 9th Gen? Was that glued together? Which Gen combusted spontaneously? Was that Rocket or Cannon? Was that on 10nm, with the AMD GPU?
eva02langley - Wednesday, March 17, 2021 - link
With RL new uarch integration, it gives a bad impression about what Alder lake will provide. You need to remember that 10nm is not much better than 14nm... and you get frequency regression... bad yields... Intel is screwed big time, but they don't have the choice to avoid investors lawsuits with their 10nm promises, and 14nm has nothing left to extract from it. The hard reality is that they are trap where they are until 7nm, which is supposedly worst, came into play.Gondalf - Wednesday, March 17, 2021 - link
Same applies to all processes around the world. Even TSMC processes show a clear frequency regression. The voltages to have a given number of Ghz are crazy on an AMD SKU.All these voltages are out TSMC official specs.
In short words actual top level Ryzens are SKU with a small life cause electromigration under turbo operations.
The real clock speeds allowed by 7nm TSMC are only on Epyc line, and they are a lot lower to give reliability during time. 4 GHz is a clear wall for foundry 7nm, 14nm can run wel over 5 Ghz without exit out the max allowed voltage long period of time.
Gondalf - Wednesday, March 17, 2021 - link
Already done, Zen3 is in the same IPC family of Willow Cove core.Golden Cove will be at the level of Zen 4......but a lot early ..........
This time AMD is in a great dilemma, a medium volume Zen 3+ or a real ghost Zen 4.
Qasar - Thursday, March 18, 2021 - link
too bad you cant prove any of this BS, gondalf. wishful thinking, and hopes for your idol, intel ?Holliday75 - Thursday, March 18, 2021 - link
He is a fanboy that spits out fancy terms hoping to sound like he knows what he is talking about.geoxile - Wednesday, March 17, 2021 - link
Roadmaps suggested Zen 3 refresh before Zen 4. I wouldn't be surprised if Zen 4 came late summer or holidays 2022, with mass availability in early 2023.Holliday75 - Thursday, March 18, 2021 - link
AMD officially added Zen3+ to their roadmaps? I missed that one.name99 - Tuesday, March 16, 2021 - link
QC: #5G #5G #5GINTC: Screw 5G, we're launching a new G every nine months! Whoever has the highest G wins!
artifex - Thursday, March 18, 2021 - link
How many new sockets?Ashinjuka - Tuesday, March 16, 2021 - link
The flood-the-zone-with-shit strategy.Samus - Tuesday, March 16, 2021 - link
I wouldn't call it shit. I mean, if they sold them for reasonable prices I'd bite just because the platform quality is traditionally more reliable with Intel. The problem is the i7-11700 is outclassed by a $200 AMD CPU so it isn't worth anywhere near $300. Intel's entire CPU stack is effectively overpriced by 30%.heickelrrx - Tuesday, March 16, 2021 - link
you realize that those product stack are mean for SI?and AMD isn't really exist on Big SI, because well... STOCKS
Retycint - Tuesday, March 16, 2021 - link
Release CPUs that are actually available for sale, and at very competitive prices? 5800X - 8C16T no iGPU - $449; i7-11700F - 8C16T with no iGPU - $298. It's a massacre of the red team, at this point.Sharma_Ji - Wednesday, March 17, 2021 - link
Is there even logic to pair such high end CPUs with weakling integrated graphics ?duploxxx - Wednesday, March 17, 2021 - link
yes buy that intel cpu for sure because of the integrated graphics :)and the AMD are out of stock because nobody wants them :) they all prefer the Intel :)
TheinsanegamerN - Wednesday, March 17, 2021 - link
Previous generations were all on 1 die, so intel would have had to make a seperate die to not have an iGPU. They did, it was called HDET, and people whined it was expensive.Now intel Xe is on 10nm while the cores are 14nm, but I believe they are still under the same die package.
drothgery - Thursday, March 18, 2021 - link
There are actually quite a lot of applications that tax CPUs heavily but don't really need much GPU (but do need some, because anything with a UI without a GPU is not fun). Also, given the GPU shortage, rolling with the iGPU until the card you want is available at a reasonable price isn't a terrible idea either.futurepastnow - Wednesday, March 17, 2021 - link
And they didn't launch one SKU that might actually be interesting, a 4c8t i3 with the full UHD 750 graphics. Which might actually steal some sales from AMD's unobtainium APUs.WaltC - Thursday, March 18, 2021 - link
Rocket Lake appears to have blown up on the launch pad...;)baka_toroi - Tuesday, March 16, 2021 - link
>In order to enable Cypress Cove, Intel has taken its Sunny Cove core design built on 10nm, and re-architected it for 14nmAHAHAHAH INTEL IS FINISHED!
plonk420 - Tuesday, March 16, 2021 - link
might want to realize there's more than just the DIY market...baka_toroi - Tuesday, March 16, 2021 - link
I'm aware. Doesn't change the fact that Intel is finished.I'm sure all CTOs are lining up to buy worse technology than what they already have.
RanFodar - Tuesday, March 16, 2021 - link
Intel is finished?A company that is larger than AMD's and has huge profits that can make up their lawsuits, is finished?
baka_toroi - Wednesday, March 17, 2021 - link
How long will they be able to sustain those "huge profits"? Not for long, semiconductors are expensive and you can blow 5 billion USD in a quarter if you get unlucky.TheinsanegamerN - Wednesday, March 17, 2021 - link
They just had another record year while pushing outdated 14nm skylake crap in the face of zen 3. Intel isnt going anywhere.Yesthey are having problems, but it would take another decade+ for intel to really start suffering at this rate, and easily 2 decades to go out of business if they somehow never put out another CPU design.
eva02langley - Wednesday, March 17, 2021 - link
Intel main problem is not AMD, it is TSMC. It will force them to update their fab strategy, because they lost their edge for good there. As long as TSMC have better nodes, Intel will always be trailing behind. And if theygo with TSMC, they will never get the capacity required for maintaining their business model. Don't lure yourself, 14nm is done and Intel doesn't have anything to turn this around. There is massive risks arising for them in the next 5 years... massive risks...Holliday75 - Thursday, March 18, 2021 - link
Intel needs to partner with TSMC and give them access to their US Fabs in exchange for their engineering prowess.Also ASML needs help to increase manufacturing output. They are the sole provider for leading edge nodes and cannot keep up with industry demand. That demand is only increasing and I fear they will fall further behind. Current shortages might be our new normal unless major changes happen across multiple industries.
Timoo - Saturday, March 20, 2021 - link
I don't even think TSMC is their main problem.From what I understand, reading all these articles over the past few years, is that at 10nm or 7nm your complexity starts to suffer. You get quantum leaks, or rogue electrons. AMD made a very clever move by going to chiplets and combining them, keeping the designs relatively simple and yields high enough.
Intel has a good 10nm process, just not for the incredible complexity of their monolitic Core-design. Simplified versions for laptops give good yields, though, if I am correct. So, as soon as Intel moves to chiplets, connecting them together, their 10nm should be sufficient. For now Intel is their own worst enemy. And if I understand it correctly, their Alder Lake design is a Big/Little one. Maybe even coming with their own version of chiplets, or something.
I am an absolute AMD fan, because I understand how Intel rigged the market, back in 2006. So for me personally, they deserve to go down. But that's personal. Professionally I think Intel will come back soon enough, to my regret.
Qasar - Saturday, March 20, 2021 - link
And if I understand it correctly, their Alder Lake design is a Big/Little one. Maybe even coming with their own version of chiplets, or something. "alder lake, is X big cores ( current cpu cores) with X small cores ( atom, or arm cores )
geoxile - Tuesday, March 16, 2021 - link
Because they can actually supply their business partners? What a dumb commentDetJohnKimble - Tuesday, March 16, 2021 - link
guess old Intel is going to pack up shop and exit the CPU business. What a ridiculous comment. Nerds care about benchmarks. Granny buying billy a new computer at christmas doesn't know what an AMD is, but she has heard of Intel since the 1960's. That and Intel's OEM partnerships will keep them in business a long time.Jasonovich - Wednesday, March 17, 2021 - link
It's more likely, granny will take savvy grandson to the PC retailer with her, and allow him to choose, methinks, i'll it be a nice shiny desktop, with the nice green AMD logo (or is it red?), great choice, X5950 with Nvidis 3090, 32gb ram, 1Tb of NVME m.2.Excellent choice!
Er, grannies paying...!
Jasonovich - Wednesday, March 17, 2021 - link
Nvidis? Of course I meant Nvidia...a deliberate choice just to give balance to the argument :)GeoffreyA - Wednesday, March 17, 2021 - link
Granny, arching her eyebrow, cleared her throat."Just joking, gran. It'll be a Ryzen 5 3600, along with a GeForce---" and so Billy recited the rest of his components and the two leaving the shop hand in hand, went home smiling.
"Must it have that G, what do you call it?"
"GeForce. Yes, gran, it must. Makes the games go fast."
"I see, my boy," said Granny, sighing and smiling, "I see. Well, dear, I hope your games will go *very* fast."
GeoffreyA - Wednesday, March 17, 2021 - link
* games goJasonovich - Wednesday, March 17, 2021 - link
LOL!jutre - Wednesday, March 17, 2021 - link
TSMC’s present is 5nm. Intel’s present is 10-14nm. Part of Apple A9 production (from Samsung), FIVE generations of processors ago, were etched in 14nm. Intel is indeed finished unless it can jump to another node very quickly. Else, it is now a second tier company and if you have any money in it, you should look for a way out. I can safely say that with the M1 on the market, and the M1X and future generations coming, that I will no longer buy any Intel processor for potentially the rest of my life. This is a mess...29a - Wednesday, March 17, 2021 - link
Intel is truly fucked if you're not going to buy their processors any more.eva02langley - Wednesday, March 17, 2021 - link
You are acting like Apple is a trusted cie when it comes to security. Even after Intel hardware flaws tsunami, I still trust mroe Intel on that aspect than a brand new architecture relying on emulation for comon softwares on X86 than a brand new uarch put together on the go with ARM cores, which are democratize and full of design flaws.GeoffreyA - Thursday, March 18, 2021 - link
Not an Intel proponent, but they can always strike back and are far from finished. Remember AMD, who was close to bankruptcy: they struck back with force in 2017 and turned the tables round in a matter of four years.5j3rul3 - Tuesday, March 16, 2021 - link
Intel the Toothpaste:(
JayNor - Tuesday, March 16, 2021 - link
"Cypress Cove, by most measures, is a reflex response to a widening gap in Intel’s desktop roadmap,..."That's a pretty strange statement, since you know how many years go into these type of projects and the number of approvals they would have to go through at Intel to get funded.
It is obvious to me that 14nm processing enables higher clock speeds than have been achieved on Intel's 10nm or on TSM's 7nm AMD parts. That is apparently a market opportunity that Intel can go after... and why shouldn't they? At the time they made this decision, they likely didn't know that 10SF would be what it is, or that 10esf would be available so soon, but even now it looks like a good use of their available 14nm capacity while they are still in the process of repurposing their 14nm fabs for 10nm.
Jasonovich - Tuesday, March 16, 2021 - link
Clock speed as AMD has demonstrated with the Zen architecture is secondary to IPC. Higher clock speed tends to equate with excessive heat, thus creating further need for exotic heating solutions and more costs to the consumer. Lessons need to be learnt from the 900 lb gorilla and there's been too much monkeying around lately!Otritus - Tuesday, March 16, 2021 - link
Intel can't increase IPC further until Golden Cove arrives in Alder Lake. Alder Lake is not ready NOW, but should be quite performant and efficient. Between the choice of lower IPC higher frequency Skylake, and higher IPC lower frequency Sunny/Willow Cove, Intel chose the the third route of higher IPC and frequency Cypress Cove.Given that 10SF is operating well, Rocket Lake is an absolute power hog, and Alder Lake is ready in a few months, it is difficult to say that this was the best possible decision. But, given the state of 10nm 12-18 months ago (when Cypress Cove would have been made), it was likely the right call to make and uses existing 14nm capacity efficiently as Intel transitions to 10nm.
yeeeeman - Tuesday, March 16, 2021 - link
Rocket lake has almost the same level of ipc as zen 3. About 5-10% less, but more important, higher than zen 2TheinsanegamerN - Thursday, March 18, 2021 - link
>almost>5-10% less
Pick one
Qasar - Friday, March 19, 2021 - link
but still looks to be a dud.ottonis - Tuesday, March 16, 2021 - link
It's very simple: given the supply shortages plaguing the zen3 so far, Intel only needs to bring anything to the market that is somehow competitive (and be it only single core performance), produce it in quantities and call it a day.haukionkannel - Tuesday, March 16, 2021 - link
So true!Intel will print money with these and Also get bigger market share, just because They have a product. It don`t have to be fast, it don`t have to be pretty. It just needs to exist!
McCartney - Tuesday, March 16, 2021 - link
can anyone tell me whether sapphire rapids will launch at the end of the year? techpowerup is saying that it should. their timelines seem accurate as they're claiming alder lake should be due out in sept 2021. i'm after a dual socket setup (xeon) so consumer dates aren't necessarily of interest.drothgery - Thursday, March 18, 2021 - link
Not unless someone is an Intel employee, no. They're supposed to, but no launch dates for Sapphire Rapids (or even Ice Lake SP, which is destined to be on top of Intel's product line for about as long as Rocket Lake, which is to say certainly less than a year, and maybe as little as six months) have been announced yet.eva02langley - Wednesday, March 17, 2021 - link
The 5800x and 5600x are all over the place, it is not going to matter at all.MenhirMike - Tuesday, March 16, 2021 - link
Depending on your use-case, the heat isn't much of an issue as long as it's faster. Yeah, a big tower cooler like the NH-D15 is a bit ridiculous at first glance, but if it means that the CPU is overall faster for your specific need, then it's definitely worth looking at.That said, a Ryzen 9 5950X has a Boost clock of 4.9 GHz, while a 11900K only goes a little higher, to 5.3 GHz (with a 4.8 GHz all-core boost), and only has 100 MHz more base clock.
As usual: Benchmark your use case. For my main work PC, I need the best single threaded performance, which is why I went with a 9900K over a Ryzen 9 3900X in Mid-2019. But if I was refreshing my machine right now (and if supplies were available), I'd definitely look at benchmarks over fanboism - the Ryzen 5000 is a darn fine CPU.
Targon - Tuesday, March 16, 2021 - link
There is a difference of a bit higher power draw and almost double the power draw. Ryzen chips, going all the way up to the 16 core Ryzen 9 5950x have a 105W as the official TDP, but generally goes to around 125W of CPU power draw under normal circumstances. These Intel chips are going to 250W to get that 5.3GHz speed that is needed to stay ahead of AMD.If we were only talking 25-50W of additional power, most would agree with you, but honestly, these Intel chips are pulling a LOT of power and it isn't even to get a full lead across the board. The reality is that Intel needs 200W and above just to get a lead in one out of 8 areas.
CBeddoe - Tuesday, March 16, 2021 - link
Yeah. Intel is basically overclocking these chips to the max right out of the box to compete with Ryzen single core speeds. They loose against Ryzen multicore. reviews I've seen show rocket lake Temps up to 98°C.Intel needed to limit the number of cores to keep the thermals within the higher limits they set. I'll be interested to see if they have issues with more failures. I'd still buy Ryzen since it gives better performance for multicore and is over 4x more efficient.
MenhirMike - Tuesday, March 16, 2021 - link
Yeah, but: So what?200W power draw is not sexy, but it's manageable just fine with a big air cooler or AIO. (Also see Threadripper with it's 280 W TDP - perfectly manageable and worth it for a lot of use cases) The question is whether you get a speed boost, and whether the speed boost is worth the extra cost. For a personal system maybe not so much, for a work machine that runs stuff limited to single threads maybe more so.
For me, the 9900k was worth it over the Ryzen 3xxx, but I'm not sure I'd pick the 11900K over a 5950X - but I will surely take a close look at both for the actual use case I have.
TheinsanegamerN - Wednesday, March 17, 2021 - link
Just fine doesnt cut it, you need to keep it cool to get that 5.3 GHz, that mean sunder 70C. That means LOUD fans or an AIO to keep the temps down.abufrejoval - Tuesday, March 16, 2021 - link
You may well just run with 25-50Watts extra on your games, because those tend to stress your 300Watt GPU much more than your CPU cores.I had a Kaby Lake i3-7350K dual core running at 4.8GHz turbo next to ab E5-2696 v3 18 core Haswell running at 4GHz turbo: Guess who was faster in games but lost big time on CPU render benchmarks?
That was replaced eventually by a Ryzen 5800X which now delivers pretty much identical multi-core results as the Haswell Xeon and incidentally both use the same 110Watt of power. Only the Haswell chip is vastly bigger and would have cost 10x originally.
So even that 22nm 18 core Haswell still uses half the peak power of an i9-11700K, because it runs its 18 cores at only 2.7GHz, but it's still not a good gaming CPU.
And the 4.9GHz 5800X really isn't significantly faster than the Tiger Lake NUC with an i7-1165G7 on its top on any single core load, which run at 4.7GHz there for practically identical IPC and per-core TDP.
If yields and capacity for 10nm were plentiful, Rocket Lake would not exist.
But I have no doubt it will find willing buyers while the real environmental damage is much less than Bitcoin/Ethereal mining.
I'd guess that by the time enough games could use all 8 cores of Rocket Lake or Ryzen 3, the systems you built today will be out of date. In the mean-time their CPU power consumption will only be a small concern to the people who buy them.
Hossein - Tuesday, March 16, 2021 - link
Well, my 5950X has 5 cores boost to 5 GHz out of the box, with a max power usage of 135W. Idle, it consumes the same amount of power as my i7 6800K with 5 cores clock at 3.6 GHz and one core clocks at 3.8 GHz. In every usage like web browsing and watching YouTube, it actually consumes less power than the 6800K.yeeeeman - Tuesday, March 16, 2021 - link
This was actually demonstrated many times by both intel and amd. The last example of this was core 2 duo or sandy bridge vs bulldozer. Intel had to push these chips to 5ghz cause they really didn't have any other alternative.JayNor - Tuesday, March 16, 2021 - link
Intel also announced a couple of Cryo Cooling technology/Peltier heat sinks, if you're worried about the heat.I don't believe AMD made a choice to run at lower clock speeds ... just a matter of they didn't have a choice with the current TSM technology.
Qasar - Tuesday, March 16, 2021 - link
" Intel also announced a couple of Cryo Cooling technology/Peltier heat sinks " yea only because intel needs them to cool these power hungry hot cpus.Qasar - Tuesday, March 16, 2021 - link
JayNor, spoken like someone who either is paid by intel, or believes their PR bs.its obvious to most, that 10nm is not where intel needs its to be right now, and may never be. the only market intel can go after now is high midrange, or lower. rocket lake looks to be still a dud, even after all those pro intel fanboys that were crying about AT using an old microcode, so they updated with a new code, and it STILL didnt do what those same fanboys were saying it would. intel has no solid answer to zen 3 yet. instead, itel releases how many rocket lake skus ? 14 ???? that is crazy, and just confuses people. face it, no amount of PR from intel, is going to help, if amd was able to make enough zen 3 cpus, intel wouldnt be selling much, but intel is, only cause its all that is on the shelf. most i know, arent even looking at intel, and are waiting for zen 3.
damianrobertjones - Tuesday, March 16, 2021 - link
I like what you're saying, but capitals at the start of sentences would be amazing!JayNor - Tuesday, March 16, 2021 - link
Intel is sampling Alder Lake, Sapphire Rapids and Xe-HP 10nm parts, and has been shipping Tiger Lake, Ice Lake and P5900 family chips. It is time to lose the fantasy that Intel can't build 10nm parts.Qasar - Tuesday, March 16, 2021 - link
time to loose the fantasy that intel can build anything faster then what they have now, and that they dont use gobs of power. no matter how you spin it jaynor, rocket lake is a dud, 10 nm isnt shipping in anything meaningful, and only in low volumesJayNor - Tuesday, March 16, 2021 - link
I agree, Intel's 14nm clocks are edging up a little with Alder Lake. I saw a benchmark today that listed an Alder Lake running at 4.6GHz. Rocket Lake runs at up to 5.3GHz according to the release, and they have provided new overclocking capability in the memory controller.Intel claims to have shipped 115K Ice Lake Server chips so far, and it isn't even launched.
Tiger Lake 10nm shipped in 150 products prior to q1 cc. In 1H2021 they are adding TGL-vPro, TGL-H35, TGL-H45, TGL-EVO Chromebooks. Intel says they shipped around 30% more 10nm parts in 2020 than they estimated at the first of the year, so looks like 10nm is ramping well.
Qasar - Tuesday, March 16, 2021 - link
yea as far as intel says yea right, if they said they shipped 200, you would believe that too i bet. i dont believe intels marketing or pr, sorry..drothgery - Wednesday, March 17, 2021 - link
Look, you can go to Dell's web site today and order a $500 Inspiron with Tiger Lake in it and get it delivered within a week. To argue they're not producing 10nm chips in volume at this point is silly.Qasar - Thursday, March 18, 2021 - link
maybe, but dell has always been intels best buddy, and gets cpus more then any one, i still cant go into any local store here and actually see any 10nm laptops IN STOCK. saying you can go to 1 company and get them, vs any store, is NOT in volume, nice try though.Holliday75 - Thursday, March 18, 2021 - link
Qasar. I work for a very large company and we had to put a company wide lifecycle refresh on hold due to lack of laptops. We use Intel, AMD and Apple. The entire supply chain is stressed beyond its capabilities and Intel's 10nm woes of the past have little to do with that. We simply cannot get laptops from EVERYONE fast enough.Holliday75 - Thursday, March 18, 2021 - link
The board of directors is there to represent the investors and they take that seriously. If Intel lied about that their investors would know. The media would know. The public would know.Qasar - Thursday, March 18, 2021 - link
we already do know, how long has intel been saying 10nm is " on track " for ? 3, 4, 5 years now ? and it still isnt where intel needs it to be.abufrejoval - Tuesday, March 16, 2021 - link
I got the proof right in front of me i7-1165G7 and Ryzen 5800X on top of each other with pretty idential single core performance and comparable power efficiency.There is zero doubt in my mind they could produce an 8 core part that would be close to identical to the 5800X in terms of power and efficiency, yet they chose otherwise.
I doubt we'll ever know the full story, but I do not doubt that they analyzed the situation carefully and actually got much luckier (TMSC capacity constaints), than they might have otherwise.
Bagheera - Wednesday, March 17, 2021 - link
Their 10nm isn't good enough to handle desktop pet envelopes, which is why they are only offering 10nm in server and laptop parts. This may also have to do with why ADL will only have up to 8 "big cores".Jorgp2 - Tuesday, March 16, 2021 - link
>JayNor, spoken like someone who either is paid by intelWait you're telling me you shill for free?
arashi - Wednesday, March 17, 2021 - link
You'd know the best I guess.WaltC - Tuesday, March 16, 2021 - link
Yawn....wake me when Intel begins marketing something competitive. Z-z-z-zzzzz....JayNor - Tuesday, March 16, 2021 - link
"cannot be supported, at least at the PCIe 3.0 speeds required."you mean PCIe 4.0, I think...
JayNor - Tuesday, March 16, 2021 - link
"These could all be added to any other motherboard, AMD or Intel, with discrete controllers which are slightly more expensive – those controllers don’t have to be Intel either. "So, what advantage, other than the slightly less expense, does Intel provide by supplying them? Is it that Intel is also responsible for providing the drivers? Or perhaps that they reduce the number of pcie pins that would be required to interface with external discrete implementations? Or perhaps there is a thermal advantage to having the digital portion on chip rather than running high speed signals externally?
Jorgp2 - Tuesday, March 16, 2021 - link
An and, why don't you make an article about Intel's Flex IO most people don't understand how it works.Their chipsets can only provide 16 PCI-E lanes max, usually boards will deliver less than that.
But their marketing department still list their chip sets as delivering "up to 24 PCI-E 3 lanes", while their technical documents list them as Flex IO lanes.
In reality its more like 14 PCI-E lanes, 8 SATA, 6 USB 3.
With certain configurations only 12 Lanes available for m.2, and that number drops further if the manufacturer wants more than 6 USB 3 ports
repoman27 - Tuesday, March 16, 2021 - link
Intel hasn't posted datasheets for the (desktop) 500 series PCHs yet, but the HSIO lane multiplexing diagrams from any of the datasheets for previous generations explain the situation pretty clearly.The Z490 PCH, for instance, has 30 HSIO lanes. 6 are USB3 only, the other 24 can be PCIe. Of those 24 lanes, 4 can be used instead as additional USB lanes, up to 6 can be used for SATA, and one can be used for the integrated GbE.
Yes, OEMs have to make trade-offs, but some designs have no need for more than 6 USB3 ports or any SATA ports whatsoever and can utilize 24 lanes as PCIe without issue. At least with Rocket Lake Intel finally moved to DMI 3.0 x8 so the connection to the host CPU won't be quite so hopelessly oversubscribed.
dragosmp - Tuesday, March 16, 2021 - link
I like the story of this launch. Although the CPU performance is a bit underwhelming and the power consumption is a bit ridiculous, we can't deny AMD CPU prices have been creeping up. I think competition is good. The 8 core i9 KF part, 70$ MSRP under the 5800X? It's worth a consideration, surely.OK, need to pay the Intel tax on motherboard, which will probably be obsolete for thext CPU generation, so the CPU pricing advantage might evaporate when considering the platform cost. Intel has its work cut out. We can only hope AMD will respond and bring back down the 6-core prices under 200 and 8-core under 300$, where they were during Ryzen 1 and 2, when they were still the underdog.
Qasar - Tuesday, March 16, 2021 - link
"We can only hope AMD will respond and bring back down the 6-core prices under 200 and 8-core under 300$, where they were during Ryzen 1 and 2, when they were still the underdog "why would they ? amd has the better performing cpu, and they should charge a little more. supply and demand has raised those prices higher, but still, if anything, intel should be charging less then the amd equivelent, only then would intel be a consideration, much like how it was when the performance role was reversed.
yeeeeman - Tuesday, March 16, 2021 - link
AMD won't bring the prices down ... They want to make money like intel wants. I find it very funny how people thought amd will always keep the 200 bucks prices when amd was basically forced to have those prices in order to gain market share. Now that they have the market share they can milk it.The kf parts are very interesting indeed. I would argue the i7 kf is better cause it is just an i9 with fewer megahertz.
Tomatotech - Tuesday, March 16, 2021 - link
I used to read every word of the review for each new Intel generation launch. Now I can't be bothered. It's all more of the same and probably a 5% performance increase.Sub31 - Tuesday, March 16, 2021 - link
11400F and 11400 seem pretty competitive.ZoZo - Tuesday, March 16, 2021 - link
"Backporting 14nm to 10nm"Isn't it the other way around, backporting 10nm to 14nm?
shabby - Tuesday, March 16, 2021 - link
YesDannyH246 - Tuesday, March 16, 2021 - link
AMD released Ryzen 1 in Feb 2017. 4 years later and Intel are still vomiting more or less the same crap out. Yawn....cue Anandtech to post ANOTHER article about the super amazing world changing performance blitzing CPU's that Intel are going to release in 5 years.ZoZo - Tuesday, March 16, 2021 - link
Twice the cores, higher clock speeds AND 20% more IPC, 25% more PCI-E lanes from CPU, twice the bandwidth between the CPU and chipset, much better integrated graphics.If you put the 7700K and 11700K side-by-side, for around the same price, the 7700K would get murdered. So it's kind of bad faith to say that they're spewing more or less the same stuff out.
DannyH246 - Tuesday, March 16, 2021 - link
Ummmm Intel released their first 8 core CPU in 2014. The highest number of cores they are releasing now is........8. Who cares about the graphics? People who care about graphics will buy a discrete GPU. Yawn run along Intel puppy dog.ZoZo - Wednesday, March 17, 2021 - link
I'm comparing the best of mainstream platform of feb 2017 (socket 1151) vs best of mainstream platform 2021 (socket 1200), and that's at around the same price bracket of 350-400 USD.To get more than 4 cores from Intel in feb. 2017 you needed to buy into the HEDT or server platforms (socket 2011).
Pneumothorax - Wednesday, March 17, 2021 - link
And if it weren't for AMD, the 11900k would be another 4 core chip as Intel decided that 4 cores were enough for 10 years for the consumer space....Qasar - Wednesday, March 17, 2021 - link
Pneumothorax, not according to the intel fanboys. intel was nice by sticking us at quad core all those years so amd wouldnt be forced out of business, cause they couldnt compete then.Bagheera - Wednesday, March 17, 2021 - link
actually that's 19% IPC against an architecture from 2015ZoZo - Wednesday, March 17, 2021 - link
Is that supposed to invalidate my observation?Bagheera - Wednesday, March 17, 2021 - link
yes, because Skylake is 2015 and Ice Lake is technically a 2019 part. just because Intel kept making Skylake until 10th gen doesn't automatically make the 19% ipc uplift amazing - keeping on mind what you are comparing against.also keeping on mind Zen 3 was 19% ipc from Zen 2, in one generation - and AMD was playing catch-up this whole time. that's more impressive than what Intel's done, quick is slapping a 2019 arch onto an old node and calling it something "new" because it's faster than an arch from 2015. no duh.
TheinsanegamerN - Thursday, March 18, 2021 - link
20% more IPC? Since when? IPC has been stagnant since skylake. And all that IPC does jack shit int hebenchmarks where the rocketlake CPUs occasionally lose to comet lake and consistently lose to zen 3.Integrated graphics? Intel fanbois constantly whine about those.
You can connect a SSD to your CPU direct now. Congrats on getting to where AMD has been for years.
abufrejoval - Tuesday, March 16, 2021 - link
What's almost more interesting than the technology (which seems to have very few really unkowns left), are the supply chain struggles behind the scenes, consumers never really cared about in this CPU domain before.Judging by my Tiger Lake i7 NUC and my Ryzen 5800X, both CPU core designs and manufacturing processes seem so very close to each other, few could probably tell a double sized Tiger Lake APU and a Ryzen 5800U apart.
But it's the foundry capacity constraints of TMSC and the yield issues with Intel's 10nm which are really the decisive factors in this battle with it's long term implications which come simply from market share or the size of the eco-system, not the technology details.
And this battle is fought on so many distinct fields, even if the manufacturing capacities are largely global (especially with the AMD CCDs). We hear about 10nm big Xeons and notebook SKUs being produced en masse, while today we reminded how Intel would rather resort to backporting a whole design than yield the desktop and we mostly wonder if it's done because the psychological impact of being a loser on the desktop was considered a risk big enough to do this or if Rocket Lake quite simply is commercially viable the way it is.
29a - Tuesday, March 16, 2021 - link
It would be nice to see what the iGPU is capable of.shabby - Tuesday, March 16, 2021 - link
It can't mine if that's what you're wondering 😂abufrejoval - Tuesday, March 16, 2021 - link
Got NUC8 (48EU Iris 655+), NUC10 (24EU HD620) and NUC11 (96EU Xe) and an Iris 550 Plus (48EU) on a Skylake i5 here.From the various benchmarks I've run I have good reason to believe that the 32EU Xe will perform very close to the 550 Iris Plus with 48EU, which is certainly better than all HDxxx GPUs since Skylake, but certainly not twice as fast as 48 vs 24 EU would imply (the 550 GT3e with 64MB eDRAM wasn't either).
That the 96EU on Tiger Lake without eDRAM manages to achieve a significant boost vs even the 655 Iris plus with 128MB eDRAM, is quite remarkable, but in absolute terms still might only satisfy the seriously bored without access to a gaming rig.
Expect perfectly reasonable 2D and media performance at 2x 4k resolution, good-enough Google-Earth 3D with Chrome and nothing but disappointment in gaming for these desktop SKUs.
yeeeeman - Tuesday, March 16, 2021 - link
It is obvious that this is just a stopgap for intel. A stupid decision that was made a few years back and now..it is too late to be cancelled. I also think that the current situation is favorable because intel has its own fabs and can made as many as the market asks compared to amd which is at the mercy of tsmc. So I think these will actually sell well with the 11700k looking like a good option.Qasar - Tuesday, March 16, 2021 - link
yea. only if zen 3 is not instock at the time, and the person doesnt want to wait. and the 11700k would still be the bad option.drothgery - Tuesday, March 16, 2021 - link
Of course it's a stopgap. Pretty much everything Intel's launched on the desktop since Skylake has been a stopgap (though pre-Ryzen 3, they were more than Good Enough stopgaps).But that doesn't mean it was a stupid decision (though if Adler Lake desktop parts launch this year and availability is good -- and that may very well happen -- it may have been a wrong one) given the state of Intel 10nm at the time they would have decided to make it, and would have been a better decision if they'd done it sooner.
eastcoast_pete - Tuesday, March 16, 2021 - link
Thanks Ian! From my POV, the most attractive offerings of Rocket Lake are some of the i5s, assuming they're available and sell at (or below) MSRP. Best/Most bang for the buck in the whole lineup, IMO. Also, These 6 core chips should have enough extra oomph to convincingly separate from the Comet Lake refreshed i3, which otherwise are currently "it" for entry-level desktop CPUs.Question about the Rocket Lake i5s: do the highest-priced i5s with iGPU really have the same 32 EU Xe configuration as the i7 and i9? The table suggests they do, but your article says otherwise; can you clarify? Thanks!
Lastly, about the lower 400 series Chipsets and why they don't support PCIe 4.0: I believe that has less to do with them being fabbed in 22 nm, and a lot more with those being basically rebranded 300 series chipsets. And those were old years ago.
evilpaul666 - Tuesday, March 16, 2021 - link
What's the point of the new Gear Modes? Is the IMC flaky or power consumption a problem? Or is it just BS product segmentation?johanpm - Tuesday, March 16, 2021 - link
What would make me consider Intel over AMD currently is the integration of both a 5Gbps (or better 10Gbps) NIC and TB3/TB4. X570 mobo with integrated Thunderbolt are scarce and therefore expensive.On the other hand you can get an AM4 with working ECC RAM.
abufrejoval - Tuesday, March 16, 2021 - link
I wish someone finally sold a 10Gbit NIC operating off a single PCIe 4.0 lane: I don't like sacrificing 4 lanes with an Aquantia AQC107 designed for PCIe 2.0.But for point-to-point TB3 peer-to-peer networking is quite ok and potentially much cheaper, bridging for bigger networks a bit messy and I was able to get 1Gbyte/s throughput only with 32k MTUs.
Latencies are at Infiniband levles, though...
repoman27 - Tuesday, March 16, 2021 - link
The AQC107 is capable of PCIe Gen2 or Gen3 at x4, x2, or x1.https://www.marvell.com/content/dam/marvell/en/pub...
I believe Apple usually configures them as Gen3 x2.
TheinsanegamerN - Thursday, March 18, 2021 - link
Yeah but nooyd makes the cards with a x1 connection.repoman27 - Sunday, March 21, 2021 - link
That's what Kapton tape is for.dsplover - Tuesday, March 16, 2021 - link
i3-10325 looks like a great cheap chip for audio streaming where single core performance is the goal.I will just wait for an AMD APU or the next Intel as long as I don’t need water cooling.
Have to say I’m disappointed at Intel’s lack of a fierce response.
Thought for sure they had designs they could reply with.
This is just another 14nm with slightly hipper marketing jargon.
Hope they don’t build too many.
AMv8(1day) - Tuesday, March 16, 2021 - link
Jesus. Why doesn anyone even bother covering this pathetic crap? (Yes, obviously because it’s still Intel, the biggest player in the sector)Intel shuffles around a couple 100MHz spec bumps and calls it an entirely different product segment? Seriously? Tell me one single difference between their i7 and i9 lineup that isn’t a 100MHz base clock or 2 millisecond turbo boost peak that will immediately throttle because it took 250 Watts to get there?
This is dumb and at some point, people need to stop giving Intel free press coverage for this bull$#!+.
Oxford Guy - Wednesday, March 17, 2021 - link
'Intel Discontinues Overclocking WarrantiesIntel has announced the end of its Performance Tuning Protection Plan (PTPP). An end-user who previously bought a PTPP from Intel was guaranteed a one-time replacement CPU if they fried their chip by overclocking it, provided the chip was still within warranty.
The program has existed since the Sandy Bridge era, but Intel is bringing it to an end, effective immediately.'
Hruska positions this move as the natural result of overlocking becoming less relevant since CPUs are more optimized but I think it's hardly coincidental that Intel's CPUs are now shipping with AVX-512 and edgy power usage even when that's not running.
So, these hot and loud things that are slower than the competition (especially at the same PPD, performance per decibel aka performance per watt) — also have the dubious glory of no longer being covered by Intel's overclocking warranty.
zamroni - Tuesday, March 16, 2021 - link
It's should be called rocket lame.At least Intel correctly uses "rocket" naming because it does run really hot
zamroni - Tuesday, March 16, 2021 - link
Intel should ditch integrated gpu for desktop processors and move it to the chipset.That several hundred million transistors reduction will help them to increase yields hence reduce price
repoman27 - Tuesday, March 16, 2021 - link
Rocket Lake is on 14+++, not 10SF. How could yields possibly get better? Intel 14nm is nothing if not a mature process at this point. It's practically geriatric.repoman27 - Tuesday, March 16, 2021 - link
The USB section in the chipset comparison table should be updated. As indicated in the official Intel slide included in the article:up to 3 USB 3.2 Gen 2x2 ports
up to 10 USB 3.2 Gen 2x1 ports
up to 10 USB 3.2 Gen 1x1 ports
Really, that should be one row in the table—Max USB 3.2 (Gen 2x2/Gen 2x1/Gen 1x1).
The fact that OEMs can choose to add a third-party discrete USB 3.2 Gen 2x2 controller to their products is not a relevant marketing point for Intel's platform controller hub. It's not integrated into nor a feature of the 400 series chipsets in any way. At least with things like discrete Thunderbolt controllers, Intel actually produces those chips themselves and validates them specifically for use with the platform.
Oxford Guy - Wednesday, March 17, 2021 - link
'Backporting 14nm to 10nm'Eh?
Oxford Guy - Wednesday, March 17, 2021 - link
'276-292 W peak power draw during AVX-512, and motherboard manufacturers often default to unlimited turbo in any event. Users going after any Core i7 or Core i9 product should look to a good air cooler or liquid cooler to get the best thermal performance here.'How many 'good' air coolers can handle that kind of wattage, and at what decibel level?
This is a very serious issue. Not only can people find themselves with inadequate cooling, to get adequate cooling they can end up with hearing loss and/or tinnitus aggravation and/or very unwelcome noise pollution.
Noise pollution and hearing damage aren't trifles.
Holliday75 - Thursday, March 18, 2021 - link
I've never heard a desktop CPU cooler produce that kind of sound. These will not be using passive heatsinks with fan modules blowing air across them like you see in rack-mount servers. HP installed some pre-built blade racks in my data center before that were so loud they physically hurt your ears. They had to correct the issue via a firmware update.I think you are exaggerating the sound levels desktops can produce.
Oxford Guy - Thursday, March 18, 2021 - link
'I think you are exaggerating the sound levels desktops can produce.'An Anandtech review of the 'Snow Silent' power supply by Seasonic said it could be heard 'from rooms away'.
There is plenty of exaggeration and it's calling things like that 'silent'.
Please show some data to prove that 292 watts of power isn't going to require very significant noise. You also have to remember that many will have a GPU producing a ton of heat, too.
Oxford Guy - Wednesday, March 17, 2021 - link
'Second is the comparative AMD processor, the Ryzen 7 5800X, which has 8 cores and has a $449 SEP. If both processors were found at these prices, then the comparison is a good one – the Ryzen 7 5800X in our testing scored +8% in CPU tests and +1% in gaming tests (1080p Max).'Apples to apples. How much faster is the Ryzen at the same wattage, the same power usage? Performance per watt.
'The Ryzen is very much the more efficient processor,'
Please be specific.
Oxford Guy - Wednesday, March 17, 2021 - link
'On the Overclocking Enhancement side of things'I believe I recently read an Extremetech article that said Intel has dropped its overclocking warranty program or something to that effect. That is something that is relevant enough to mention with all this overclocking information you posted.
Given how demanding these things are, what kind of warranty Intel is offering is much more pressing than in the days of quad Skylakes being the standard.
Oxford Guy - Wednesday, March 17, 2021 - link
To follow up with my point about the noise... Why not write an article that shows two systems with comparable gaming performance — one using the quietest 'available' (in quotes since GPUs aren't available) parts (i.e. AMD CPU and quietest high-performance GPU or GPUs) and another using Intel CPUs + the loudest GPU or GPUs of comparable performance to the quietest options?This would be a very useful article — to give people an idea of how much they will need to spend to greatly reduce the noise pollution. Gaming is an experience. Unless one is deaf it's a big drawback to be gaming in what sounds like a tornado.
Then, of course, there is also the power usage issue — both in terms of the power bill from the machine and the cooling bill. And, sometimes the added cooling demand means more noise from that as well. Not everyone has a fancy HVAC system. Some have to make due to window ACs, which are noisy even on the lowest speed.
There is no good reason to gloss over these things. PPD (performance per decibel) should be a featured comparison.
RanFodar - Thursday, March 18, 2021 - link
What do you mean by this? Intel has the loudest SKUs on the market?Oxford Guy - Thursday, March 18, 2021 - link
Obviously, if a part uses 250 watts to do the same work another part uses 150 watts for, the 250 watt part is the louder one.duploxxx - Wednesday, March 17, 2021 - link
nice one from gigabyte... call your motherboard VISION with the latest EOL Intel chipset and EOL CPU generation :) By the end of year (lets hope for intel but probably 2022) the next cpu generation is yet again another socket and the 11th series is not worth buying vs 10th or AMD zen3 .... VISION :)Linustechtips12#6900xt - Thursday, March 18, 2021 - link
The packaging for the 11900k tho 😂LMAOLinustechtips12#6900xt - Thursday, March 18, 2021 - link
tbh my real question is not regarding intels silicon but truly there instruction set, how much money will intel be loosing when or if Nvidia buys arm and it starts to become the mainstream IS "instruction set" and my other question actually relating to there silicon is one why the actual F do we need "power efficiency cores" on a desktop system the only thing i can understadn for that is better idle power consumption or web browsing shit, part two, WHEN WILL THERE ACTUALLY BE A COMPETITIVE ARCHITECTURE, intel can totally make a decent design so where is it. AMD has made improvements on navi to navi 2nd gen just with architecture so it can TOTALLY be done "mad face emoji" :..(GeoffreyA - Thursday, March 18, 2021 - link
I suppose Intel Haifa is working on the successor to Sunny Cove as we speak. Hopefully, they've started from scratch, rather than just iterating on SC. Copying AMD's split-scheduler design might be useful, too.GeoffreyA - Thursday, March 18, 2021 - link
As for power efficiency cores, lower power always pays off at a later time or in other ways. Having said that, I'm sceptical of Intel using this fast/slow approach.velanapontinha - Thursday, March 18, 2021 - link
Intel's latest gen top of the line costs 399$???Thanks, AMD!
Everett F Sargent - Thursday, March 18, 2021 - link
Updated TDP versus CPU frequency estimates based on AVX2 all core with no PL1, PL2 or Tau limits ...I am currently guessing that the i9-11900K at full 5.3GHz all core OC (if even possible) will pull close to 400W in AVX2 type benchmarks (with no AVX type offsets applied), similarly the i7-11700K at full 5.1GHz all core OC (if even possible) will pull close to 300W in AVX2 type benchmarks (with no AVX type offsets applied) ...
View: https://i.imgur.com/4QvuhpH.png
Notes: Power guesstimates apply to the total CPU package only (e. g. not total system power consumption at the wall). More than adequate cooling solutions to keep maximum temperatures at or below 100C (e. g. Tjunction) for either 300W/400W CPU power draws during AVX2 operations.
Conquest - Friday, March 19, 2021 - link
TDP is confusing, as its mean something totally different from AMD´s TDP.With this generation, there will be no problem to have 300W consumptation.
Oxford Guy - Friday, March 19, 2021 - link
No problem?dsplover - Tuesday, March 23, 2021 - link
How many years before Intel strikes back? Lucky for them old i7’s and Xeon’s were great CPUs and lasted me into 2021.They need a new design this year before they lose more customers to the innovative designs from AMD.
Lucky for them AMD APUs aren’t readily available.
fabrizioVR - Friday, March 26, 2021 - link
The question is : What is better 11700 K /11900 K or 10900 K for a Z490 motherboard ?MackMurder - Sunday, March 28, 2021 - link
I have serious question......I just ordered a Alienware Aurora r10 /w GTX 3080 and ryzen 7 5800x (over the i7-11700f)
Benchmarks were mostly identical. I went with the ryzen mainly because their supposed to be more efficient. But the TDP of the 11700F is 65 watts and the 5800x is 105 watts.
The AWr-10 case is known for heating issues (I will be replacing the fans with the more efficient quieter noctuas) but still want to have a processor that makes the lowest heat. I will be gaming in 1440p ultra wide and for hours on end usually. I'm only worried about heat.
I'm not a fanboy so don't care if I have red or blue, and have just come over to PC gaming after 15 years of Xbox.
I just want to keep my machine as cool as possible.
Any help is apreciated thank you
Mack
watzupken - Thursday, April 15, 2021 - link
TDP means nothing nowadays. AMD probably keeps closest to the rated TDP, while Intel is the worst offender for breaching the rated TDP when it comes to real world usage. Those competitive numbers you see in reviews, they usually results in the processor pulling in at least 2.5x the rated TDP for an i9 processor.