So Intel's bowing to the competitive pressure and finally giving mainstream desktop users something more than single digit percentage point improvements at last! Yay!
The key of the Kaby Lake refresh on mobile is offering more cores, not IPC improvement - which can actually yield quite substantial improvements in performance.
Significantly lower multi core speeds. Even with turbo mode being only about 5% higher, which nobody will notice, unless more than two cores are really being used, and pushed, by the software, I don’t see much improvement in general computing, particularly with the throttling mentioned that will occur.
I might trade-in my laptop soon then. I love 3D printing but the application I use hates sharing CPU time. Believe it or not if the cores are shared with printing load it can result in printer variations and can actually cause a failure or defect. Two additional cores would resolve that issue.
I believe you, but then again I remember when you couldn't do anything processor intensive when burning an optical disk for fear that the on-drive buffer would run out of data and the burn would be ruined.
Though I must admit my first thought was, "How does the printer connect to the computer, a DB-25 parallel port?"
The printer is USB based via the virtual COM port driver found in the Arduino driver set. The printers themselves are built piece by piece from a magazine. I print from a dedicated i5-7200U based laptop at about 40% load.
Fighting with a printer at this level takes a lot of willpower but is an amazing learning experience. The issue feels very much like a buffer underrun except the print process will continue as if nothing happened.
Oh man, do I remember those times, my first burner was an external HP unit that connected via the Parallel Port. You wouldn't dare touch the mouse when a cd was being burned. What I have now, even just a 2nd Gen i5 Desktop and still use a Core2Duo Laptop are slow by today's standards but since I am not using any software that is CPU limited, I can't justify an upgrade. If I get back into using Solidworks quite often maybe I will build again, but sadly the Sandy Lake @ 4GHz still works great for games if you have enough GPU and since I am resolution limited to 1920x1200, I don't see the CPU bottlenecking the GPU until I go UHD for a display. Sorry Intel but keep up the good work.
He specifically said UNLESS more than two cores are used, you don't see much improvement in general computing. There are workloads that will benefit hugely... but it probably won't benefit your average user's experience that much.
It's still a huge boost for many users, and hopefully more software will be able to take advantage of all these threads as they become very common. With that said KL Refresh seems like a preemptive strike against Raven Ridge since Coffee Lake isn't ready. But that's not a bad thing... with a resurgent and more competitive AMD, end users are the big winners here.
I get it. But...the systems we purchased that have the "U" CPU's in them (Lenovo X1's with 7xxxU's) they are all about being thin, light, and portable. For the users in question, more CPU power means little. Its hard to see the thermals of these systems allowing for any significant increase in performance for these types of systems, but I guess I'll await some benchmarks to see.
I own an XPS 13 so I am one of the "users in question". I spend a fair amount of time waiting on MATLAB DSP code to run. I don't want a 10 or even 5 pound laptop. I want a 3 pound laptop. If they deliver 4 cores that can sit at 2.5 GHz in a 15W package it would nearly double the speed of my 9343. That's something to be happy about.
What are you talking about lower speeds? The turbo's are the same or higher. With Speedshift 2.0 these will run dual core at the same or better clock speed as the old ones, while allowing for base or better clock speeds with all four cores utilized.
An 8 threaded CPU at 1.6-1.9GHz will still outperform a dual core/quad thread CPU at 4GHz in heavily multithreaded apps, especially encoding/archiving.
And all that at 15w. AMD is definitely turning the heat up. Intel could have done this all along but the thread wars have begun, and Intel is clearly preemptively striking a Zen-based mobile CPU.
That's part of the problem with the x86-license agreement AMD and Intel have. They get to peak at each others R&D. But Intel has been "out-fabbing" AMD for almost 15 years. The last great innovation AMD had was copper interconnects, and Intel wasn't too far behind.
Intel has had a 6-core offering waiting in the wings for a couple generations now. The actual release of 6 core CL was a direct response to Ryzen. Without Ryzen we'd have seen another 4 core offering. Intel isn't going to raise it's own costs just to satisfy consumers.
If it's a respin of an existing architecture, as the article implies, and they're moving an existing design down from the 45W stack and optimizing it for lower power... then no, it did NOT take years to develop. The deployment of 4C/8T designs at lower TDPs could very well have been in response to early glimpses of AMD's plans for higher IPC and more cores/threads (including mobile). Similar to what they've done shuffling higher-core designs down from server to HEDT - those weren't exactly ground-up redesigns just for HEDT. They used existing assets and burned off features for segmentation.
Don't care who was a preemptive strike against whom. Just care about what I can buy at the time I'm ready to buy. Who did it first is irrelevant to the consumer.
Not when the specs are lower. So single core use with turbo is a whole 5% faster, Whoopi! Everything else is a lot slower. And you don’t get 100% more processing with a second core, nor do you get 100% of each added core. It even gets worse the more cores that are added. So with the lower speeds per core, plus the throttling, it will be interesting to see just how much more performance can be gotten.
And real world use is never going to match the testing done on sites like this.
Sure, "when". But there's not a single spec being worse. The all-core turbo clock is lower, but the 2-core turbo is definitely a lot higher (as on all modern Intel CPUs), probably the similar 100 - 20 MHz higher as the single-core turbo clock. And the 4-core turbo is very probably a lot higher than the guaranteed base clock, so for heavily threaded but mathematically light workloads clocks should be fine (well, a lot better than the baseline suggests).
> plus the throttling
Turbo is a bonus. If the full turbo can't be reached, that's not throttling. That's "not going into overboost". Throttling would be going below baseline clocks, which can happen with insufficient cooling.
"In a difference between 2 and 4? There sure is a difference in even consumer usage."
Well, I can't speak for everyone. But remember these are "U" CPU's, usually destined for ultra thin and light category machines. These systems prioritize form factor over performance and the thermals will rule the day. My experience with laptops is that they hit thermal throttling pretty quickly. guess we will see when there are some real systems to look at.
For mainstream typical load, an SSD brings up the biggest difference. Any cpu with 2cores+ @2.5Ghz or more is fast enough for media playback, office apps.
A year ago (Aug 2016), leaked slides showed that Intel was planning on releasing quad-core 18W and 15W mobile Kaby Lake processors (Google "18W quad core"). This means Intel was planning this release before AMD's successful Ryzen debut disrupted some of Intel's desktop market share. As for the performance of these new chips, early Cinebench benchmarks of the 8550U indicate: - Single-threaded: equal to 7600U (Top 15W dual-core CPU) - Single-threaded: significantly better than 7300HQ (bottom 45W Kaby Lake CPU) - Multi-threaded: just shy of equal to 7300HQ - Multi-threaded: significantly better than 7600U (comparisons made using CPU benchmark list on Notebookcheck.net) Note, these early benchmarks are of the 8550U, not the 8650U, meaning there is another processor with the potential for another ~5% bump in performance (based on clocks).
These early benchmarks indicate that whether you are currently running a 3+ year old mobile 15W, 35W, or 45W CPU, there are very compelling single and/or multi-threaded performance gains to be had by upgrading to the next generation of thin & light laptops based on Intel's 15W quad-cores.
I think it would be a bit out of place in a Surface Pro. The i7 Surface Pro, with fan, uses GT3e. These new chips only use GT2. Then there's also the throttling to consider.
Currently, the Surface Book uses GT2 for both the i5 and i7 versions, which makes sense, as there's the additional dGPU in the keyboard.
Maybe not these cpu's in a Pro, but a quadcore w/Gt3e or newer should flow down the pipe soon. Maybe in time for a spring refresh of the surface. Curious as to what AMD will offer as far as APU goes. If it's performant, then I wouldn't mind seeing it in a Surface Book.
Enlighten us melgross, on the way sales are going. Do you have some insider information? We have little to no information on Surface Pro or Surface Laptop sales given they started selling in limited quantities a little over a week before the end of Microsoft's last quarter. So please, let us know as to why Microsoft will stop making Surface devices and why we won't see this in a Surface Book.
I can't speak for sales, but Consumer Reports just removed its recommendation for all Surface devices due to significant user reports of defects and returns. CR said 25% of 90,000 Surface device owners surveyed stated their device broke in under 2 years, which they also stated was significantly higher than most other brands. MS responded to the CR results, stating that their own breakage measurements were significantly different, but they failed to say whether or not CR's numbers were higher or lower than their own defect numbers. It's like Microsoft wanted to appear defensive without being capable of defending themselves. It's not a good look, but it's probably not damaging enough to deter MS from the hardware game.
What a naming nightmare. A 8xxx chip can have different architectures than its siblings but the only way to tell is to look it up on Intel's labyrinthine product site.
Would have been nice and constitent for them to simply put these into the mobile core i9 7000 class. But we know for mobile it's apparently still very important to show the biggest numbers, irregardless of actual hardware.
Well as long as it keeps split between mobile and desktop its not that big of a deal. If they start mixing things more directly in the same segment, it might get more annoying.
Check out CPU world, tons of different chips not just intels, and honestly more accurate. I've found less mistakes on there than on ARK, which is funny because it's intels on site on its own chips.
It has been that way for a long long time. I have a Haswell with a 5xxx number because it is a Haswell refresh with more cores. Similar with Broadwell refresh that got a higher number.
Sandy Bridge, Kaby Lake, Coffee Lake and Cannon Lake all of them have the same IPC. The only changes is a bit more mature process and the ability to sustain turbo speed for more time (which can be PR marketed as "IPC gain") + now more cache of the 300chipset 6core cpu's.
Uhh, Sandy Bridge definitely doesn't have the same IPC as Kaby Lake, and the last two you mentioned aren't even out yet. Plenty of review sites have done clock for clock comparisons with Sandy Bridge vs the newer arch's and the new ones are definitely faster.
The naming is absolutely horrific. It was not bad enough that the i5-7360U and i7-7560U are nearly identical, now the i5-8350U is expected to offer near 2X the punch of the i7-7560U released 6 months ago. I don't understand the point of releasing so many SKU with practically the same price.
Competitive pressure from Raven Ridge - to turn around a probable loss on any and all highly threaded benchmarks to a probable win across the board, albeit at a higher price.
If not for this, Intel would have been content selling you 2 cores at higher margins for some time longer.
Agreed. If the AMD is as aggressive in the mobile space with their upcoming APUs it's very possible that Intel could start feeling additional pressure in this space and the market is huge compared to the desktop space.
I'm very curious how far down the Ryzen cores can be downclocked and still be stable. That could give us a window into what their power window would be. Yes, you'd have to add in power from the IGP, but it would still be pretty insightful.
15 watts of power budget on a quad core ryzen outta give you 2.4 - 2.5 GHz sustained clocks. Which is actually pretty good, considering this is plain desktop chip. That's just CPU cores, no GPU obviously.
What I always say is that if AMD wants to be relevant on mobile they need to release laptops under their own brand and say a huge FU to laptop OEM's or at minimum give them 2nd priority over AMD's own branded laptops.
Every time AMD has a decent APU for mobile at nice prices, OEM's simply cr@ptinated them in every possible way. What you can find in retail stores -Cheap TN screens (upgradable to IPS only visiting their online stores, with models well hidden in a store UI mess) -30-45Wh battery -single channel -5400rpm HDD -35w TPD models limited to 15-25w TDP (AMD really fooled themselves giving that kind of options to OEM's, cause they gave them yet another excuse to gimp the product)
AMD should pick one OEM and do an example device to show the rest what can be done with Ryzen, a bit like how Microsoft made the Surface tablets to get OEMs interested in making their own.
But nobody does intel chip clones, and nobody has done that for a loooong time. It is completely redundant to use "genuine". Not to mention that trademarking laws alone are enough to prevent and illicit use of a trademark.
At this point, I think it's just because that's how they have always done it. AMD uses 'Authentic AMD' -- I think the engineers just get to have fun with that string honestly. Some VIA CPU's use 'CentaurHauls' Cyrix uses 'CyrixInstead' -- Check out this page for a bunch more examples: https://en.wikipedia.org/wiki/CPUID
Not a single word about Cannon Lake....Intet had always launched new mobile cpus all together.....Core M apart. Summary: 1) Mobile 10nm Cannolake canned (likely unable to turbo high enough) 2) Desktop Coffe Lake is Cannonlake Ark on 14nm with some IPC enhancements 3) 10nm will be for Server first, like said Intel months ago.
My bet: very first 10nm product will be the new popular Xeon D with the Cannonlake core finally on 10nm
and news from AMD.... no news for mobile counter part, raven ridge (not that botched desktop ryzen put into a laptop chassis) Silent as sign of confidence or as a sign of defeat?
Except that this launch is probably a reactionary move from Intel, in anticipation of Raven Ridge. Raven Ridge is known, at least for the high end SKUs, to be one CCX = 4 cores and 8 threads.
If Intel would not have launched these , they would again have a 50% core and thread deficit vis-a-vis AMD.
The fact that we are able to get these processors, ahead of any regular release schedule, is in a large part thanks to competitive pressure.
Yes it is pretty strange do not ear something of real about Raven. No a serious leak from Chinese OEMs, not a reliable early bench with verified clock speeds. We are at one month from the launch and there is nothing around and AMD is silent, it's all about idiotic TripThreads 180W cpus. These new nodes are a damnation and they will do a lot of damage to the devasted Pc world IMO. When we have to manufacture a crap SOC for phones they are nearly decent, they fail when we have to ship a real SKU, with the result of an incredible leakage at higher voltages and stunning power density all around.
Where is the good, old and reliable 32nm (vanilla and SOI) ?
"The boxes confirm, among other things, that once these desktop parts will launch they'll have 6 cores (with HT for the i7) and require 300 series motherboards."
so that means they won't be LGA1151? Sad news, hoped to get at least another iteration on my B250 board :\
Nice to see the desktop CPU boxes mention the "UHD" IGP, it probably means we'll have hdmi 2.0 with HDCP2.2 out of the box on most 300 series motherboards.
"So despite Intel launching its 7th Generation family in January, today Intel is formally launching the 8th Generation only eight months later. To explain why Intel is breaking the usual 12-18 month cadence for the generation product, it comes down to product positioning."
Frigs sake, Intel. It's not enough that they bundle both Braswell and Broadwell into Celeron, Core Ms carry Core i* names, now this, 8th gen cores are both 8th gen cores and...7th gen cores. They're obfuscating what a processor is even more away from the consumer without doing some more digging, and it's really annoying.
Ok, so more efficient node and doubling the cores place it well above the rest of Kaby Lake in the same wattage - well then call it something else, like U series quad. Otherwise core generations mean nothing.
I personally don't see anything confusing. Most of the people here who care about tech stuff will get confused by this
The average joe does not know what kaby lake refresh or coffee lake even mean
Also, it does make sense to use 8th genetation name because of the huge improvement. This thing will be 1.5x faster in multi-thread over 7th gen U. Thats the biggest improvement we have seen per gen in multicore performance.
If they use the 7th gen name then it will be more confusing because even the i5 model have quad with HT (but smaller 6MB L3 cache), and that would be more confusing when you have new i5 quad with 7th gen name while the older one have i7 with dual core with same 7th gen model name. !!! It is better to use new 8th gen
The coffee lake model won't increase the core count. It will probably just have higher clocks with maybe 5% IPC. THey could call it i7-8750U (which sound better than 8650U). It is basically just faster quad core model, and it not worth calling it 9th gen
Cannon lake will probably released for 4.5w model, and dual core only. So, if this true then it is not aimed to relplace kaby lake refresh and coffee lake-U model
"Also, it does make sense to use 8th genetation name because of the huge improvement. This thing will be 1.5x faster in multi-thread over 7th gen U. Thats the biggest improvement we have seen per gen in multicore performance"
By your logic, if a "huge improvement" was needed no name a new generation of Intel CPU's as "new", we'll still call all Intel CPU's 2nd Gen Core.
They could have used a prefix like UQ to denote quad cores, like they do on the 45W quads (ix-7xxHQ), and also a higher part number like i5-7400UQ or something similar.
There's no IPC gain in any intel cpu until Ice Lake.
The "PR IPC" intel is giving this days is basically the ability to sustain Turbo Speeds for more time due to a more refined node, in the case of the 6 cores cpu's for 300 chipsets, you got extra L3 cache on top of that which also bumps the performance a bit-
You are still getting big boost from higher core count and increased clock speed.
Why IPC gain even matter ?! If future processor (hypothetically) can clock as high as 10GHz then it is better than anything we have now even if IPC is worse by 10%
As long as you have faster CPU with same price and TDP of previous generation then that an improvement. It does not matter if those improvement are clock speed, higher IPC or more cores
IPC still matters for single threaded applications, which is most software. As a computer programmer, I can tell you that a lot of code can't be multi-threaded and writing multi-threaded code can be very tricky, especially in C/C++. One of the reasons why so many programmers are switching from C, C++ and Objective-C to Swift, Go and Rust is because these languages make it so much easier to write multi-threaded code. There are a significant proportion of old-school programmers who share Donald Knuth's opinion that multi-threaded programming isn't worth the pain. So, yes, IPC still matters.
20% less frequency is still considered "frequencies are down slightly". But if it were 21% less frequency, then it would be "frequencies are down A LOT! Take out your pitchforks and torches!"
That just tells us that Intel's processors are still stable when running at an even lower speed/power use. It's more efficient to have the CPU slow down when not in use, and then turbo back up once it needs to. No sense in having your CPU running at 4.5ghz all day, especially on mobile parts.
Shows Intel *COULD* sell us quad core before, they just didn't want to. Don't give a fuck about +3% framerate on games, gonna buy from AMD as long as they make something good enough.
Lucky intel nvidia made MX150. They were about to lose it all with the upcoming iGPU of raven ridge. 500-800$ laptops will still be intel+nvidia combo.
Raven Ridge should destroy the GT1030 for mobile considering their GPU sits in between RX550-RX460 with dual channel DDR4 GT1030 has zero chance vs the RX550.
I'm not sure about Raven Ridge's GPU performance yet. The GT 1030/MX150 has dedicated VRAM and dedicated VRAM bandwidth. It presumably still uses some form of what was once called TurboCache back when Nvidia released the 6200 back in 2004 or 2005 so it'll also have access to system memory's dual DDR4 channels. So there's about 40-50GB/s for the GT 1030's 64-bit GDDR5 and then there's whatever else can be nabbed from the RAM if the CPU or other system devices aren't using it.
I don't think the same can be achieved easily if Raven Ridge's iGPU has to fight with the CPU cores for system memory bandwidth which is going to be higher latency than the GT 1030's GDDR5 and there's still the problem with concentrating all that heat into the CPU package. Additionally, Vega isn't much of an improvement over GCN so the 512SP count in prior generation APUs was already only barely competitive with the GT 730 (in a lot of cases, it fell behind in benchmarks due to bandwidth limitations) so the additional SPs won't be that much faster compared to prior generation SPs and they'll likely clocked lower than a dedicated RX GPU because of thermal limits so they may not offer a substantial increase over prior gen APUs or against the GT 1030/MX150.
Don't get me wrong, the A-series iGPU was a competitive product against Intel's non-Iris graphics and replacing those older/weaker CPU cores with Zen is a good thing, but I don't think you're going to see Raven Ridge end up in a position of dominance over the GT 1030. We really should take a wait-and-see approach and then look at benchmarks as they become available before predicting the demise of the 1030 and its mobile variant.
NVIDIA hasn't used "TurboCache" on their dGPUs since the 6000 series - it's long dead. I'd be very surprised to see it reintroduced for the sake of expanding a 2GB VRAM pool on a 384-core chip.
For Raven Ridge, dual-channel DDR4-2400 can do a theoretical 38.4GB/s of memory bandwidth - if you account for controller overhead, that's probably more like 32GB/s. Not that far behind the MX150, but as you mentioned it's got to share that with the CPU. Newer game engines that can take advantage of shared memory addressing on the APU might be able to offset some of the performance lost there.
For clock speeds, leaked benchmarks had the GPU clocks at 800MHz, but I don't know what the TDP was. I'm imagining the 35-45W range for the higher end ones.
tl;dr - Release something already AMD, so we can see what you've got.
I expect the top of the line apu to be: One ccx 4c/8t Interconect to the iGPU instead to another ccx RX550 performance because of downlocking for power budget and efficiency 2gb of hbm2 @ 1ghz as l4 cache/graphics memory
>that once these desktop parts will launch they'll have 6 cores (with HT for the i7) and require 300 series motherboards
Getting pretty tired of this socket/chipset change a year keeps the goyim in fear meme from Intel. Can we not just get some backwards compatibility with older motherboards for once?
Buy AMD and be happy? Why support the one who spits in your face?
AM4 support Ryzen 14nm Ryzen Optimized 14nm Ryzen II 7nm Ryzen III 7nm+ (probably with some AM5 and Ryzen III with dual DDR4/DDR5 memory controller like Phenom II)
Apple needs to simply BUY AMD and make its own Intel-compatible chips. This is unbelievable. Intel creates 8-core chips for laptops but did not give it a memory controller for more than 16 GB of LPDDR4 RAM. This is so stupid. So asinine. Intel is so SLOW in upgrading its hardware and so stupid in doing so. This is why Intel Chips today are not that much faster than chips made 5 years ago - 2+ generations of computers ago.
Do these even make sense in a sub 15W laptop, and isn't this a regression rather than a step forward?
You're basically losing single core performance and turbo speeds, in machines that are more likely to rely on single threaded performance than anything else.
Sure, multi threaded programs might be more common in the future, but guess what: These four cores will naturally be limited in performance just because of the low TDP...
Lower single threaded and no significant improvements to the integrated GPU. I'm "meh" until I see actual products at good prices. Mac mini for the same price? :)
The integrated graphics still using 620 brand. They are not calling it 720 as if it is a new GPU
Only major improvement in multi-core performance (which is bigger than anything we have see from other generation). Single-thead will be same if not better because of the high turbo boost
But 4 core is just a curiosity at this stage of the game
Hopefully 4 core fanless with 8GB Ram will be reliable enough with 10nm Canon Lake's power envelope to be the standard minimum spec for a laptop without to much throttling.....
But I doubt it They will probably choke and run too slow or hot when gimped with 2 or 4GB of Ram like the usual 64bit lineup
You have to remember that that extra cache for Iris is shared as a L4 basically, so you are getting better performance all around. And MacOS already does a lot of video processing for just the GUI, so putting anything weaksauce in there would make it feel really slow. Especially considering the capability to drive two 4k monitors with thunderbolt 3.
But they could easily put one in a Macbook. However, I doubt they will ever do that because it'll eat up their MBP sales.
Raven Ridge 4c/8t-4c/4t-2c/4t and maybe some 6-8 cores for desktop replacement laptops.
Ryzen is so efficient that the R7 1700 can easily fit on a 45w TDP with minor tweaks to turbo and base speeds (Asus releasing a full blown gaming laptop with it).
So... they're bringing their mobile parts to Haswell era 4 cores / threads and they're basically releasing the same architecture as Skylake with only added 4K video playback?
That's kind of sad, really. I mean, if anything, if the battery / performance ratio has been significantly increased, that is something worth celebrating, but otherwise, this is just a marketing ruse. It should be more like: Hey people! We finally can have Haswell era performance on laptops!
FWIW, 4c/8t has been available in laptop since Nehalem, and practical since Sandy Bridge. Nehalem 4c laptop parts were 45/55w TDP with very low clock speeds, Sandy got the clocks up quite a bit, and once we got to 22nm Ivy you could get 4c into the mid 3Ghz range in 45w. I wonder what the sustained perf will be, getting 4c/8t in 15w is pretty impressive if we can sustain 3Ghz+ clocks on all 4 cores.
"We expect that Intel’s 8th Generation will eventually contain three core designs of product on three different process design nodes: the launch today is Kaby Lake Refresh on 14+, and in the future we will see Coffee Lake on 14++ become part of the 8th Gen, as well as Cannon Lake on 10nm." So +2 cores at exactly the same process make the CPUs "a new generation", while +2 cores for desktop in a slightly tweaked desktop process also make the CPUs a new generation, and a new freaking *process node* makes Cannon Lake the *same* generation. I wonder why Intel is undermining the meaning of a generation from both sides of the concept.
You just have to laugh at the balls they have releasing a graphics GPU that is no faster than the 2013 Iris Pro 5200 it's been almost 5 years and not a single inch forward in progress on most fronts on the GPU side. Intel, asleep at the switch, once again ....
Actually they released Iris Pro Graphics 580 (72EU) in the Skylake gen which was ~2x the perf of the Broadwell era Iris Pro (48EU). No Kaby Lake 72EU part though.
Any word on whether Intel plans to release 15W quad-core parts with Iris Plus graphics? It'd be a real shame if they decided to drop the Crystawell/eDRAM SKUs - having ~100 MB/s of memory bandwidth, or the equivalent of running 6300 MHz RAM in dual-channel mode, is awesome, even within the 64 MB limit. Maybe Apple and Microsoft can put some pressure there.
It does, except it'd be only 2 GB compared to up to 32 GB, which is hardly acceptable. And what would it cost in terms of power; would it fit in 15 W without impacting the already low frequencies, or at all?
So, it is August 2017 and the top new part is 4 cores at 1.9GHz for $409. By comparison, mobile Core 2 Quad Q9000 from August was 4 core at 2GHz for $348. 9 years. It used to be that in 9 years you'd get something like 60x performance (doubling every 18 months).
BTW, to the Intel marketing people. I am still using Late 2008 fully aluminum 15" Macbook Pro with Core 2 Duo 2.4GHz and switchable Intel/Nvidia graphics, with upgradable RAM (which I have upgraded), easily replaceable HDD (which I have replaced with SSD) and easily replaceable battery (which I have replaced 2 or 3 times). And don't see anything worth replacing it with, especially with the insane trend of everything being soldered on in the latter models (making them obsolete much earlier). You are not going to get my money with the sht now on sale.
Maybe Kaby Lake R is what some are calling Coffee Lake. Intel did mention in slides that more cpus are coming this fall. With 8th generation cpus support 4 cores on U processor - it sounds like they have updated the process.
This is quad core on U processor, imagine what this means for normal mobile cpu's - like 6 or 8 core cpus coming and then higher performance mobile chips could be even higher - on mobile.
My thought on generations
8th - Kaby Lake R - possible previously called Coffee Lake 9th - Cannon Lake 10nm - ultra power only to test platform. 10th - Ice Lake - 10nm+ - Mainstream 10nm
With the other news about process changes - one thing that is likely coming is more system on chip for more compact systems.
if they put 10nm quad behind laptop screen (as in current Surface and other 2in1 tablet+keyboard designs) and use whole back-of-screen area for heat dissipation, they just might get away with it. Maybe they will have to dial TDP down to 10-12w. But it would make for a pretty nice fanless design. With 32GB of RAM please.
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
145 Comments
Back to Article
evilpaul666 - Monday, August 21, 2017 - link
So Intel's bowing to the competitive pressure and finally giving mainstream desktop users something more than single digit percentage point improvements at last! Yay!Chaitanya - Monday, August 21, 2017 - link
Kaby lake refresh is also in the list so that two digit performance gain claim might be placebo for investors and nothing more.nevcairiel - Monday, August 21, 2017 - link
The key of the Kaby Lake refresh on mobile is offering more cores, not IPC improvement - which can actually yield quite substantial improvements in performance.melgross - Monday, August 21, 2017 - link
Significantly lower multi core speeds. Even with turbo mode being only about 5% higher, which nobody will notice, unless more than two cores are really being used, and pushed, by the software, I don’t see much improvement in general computing, particularly with the throttling mentioned that will occur.maroon1 - Monday, August 21, 2017 - link
Multi-core performance shows gains as high as 1.5x over current U modelhttp://cdn.wccftech.com/wp-content/uploads/2017/08...
HomeworldFound - Monday, August 21, 2017 - link
I might trade-in my laptop soon then. I love 3D printing but the application I use hates sharing CPU time. Believe it or not if the cores are shared with printing load it can result in printer variations and can actually cause a failure or defect. Two additional cores would resolve that issue.Old_Fogie_Late_Bloomer - Monday, August 21, 2017 - link
I believe you, but then again I remember when you couldn't do anything processor intensive when burning an optical disk for fear that the on-drive buffer would run out of data and the burn would be ruined.Though I must admit my first thought was, "How does the printer connect to the computer, a DB-25 parallel port?"
HomeworldFound - Tuesday, August 22, 2017 - link
The printer is USB based via the virtual COM port driver found in the Arduino driver set. The printers themselves are built piece by piece from a magazine. I print from a dedicated i5-7200U based laptop at about 40% load.Fighting with a printer at this level takes a lot of willpower but is an amazing learning experience. The issue feels very much like a buffer underrun except the print process will continue as if nothing happened.
bob4432 - Tuesday, August 29, 2017 - link
Oh man, do I remember those times, my first burner was an external HP unit that connected via the Parallel Port. You wouldn't dare touch the mouse when a cd was being burned. What I have now, even just a 2nd Gen i5 Desktop and still use a Core2Duo Laptop are slow by today's standards but since I am not using any software that is CPU limited, I can't justify an upgrade. If I get back into using Solidworks quite often maybe I will build again, but sadly the Sandy Lake @ 4GHz still works great for games if you have enough GPU and since I am resolution limited to 1920x1200, I don't see the CPU bottlenecking the GPU until I go UHD for a display. Sorry Intel but keep up the good work.Alexvrb - Monday, August 21, 2017 - link
He specifically said UNLESS more than two cores are used, you don't see much improvement in general computing. There are workloads that will benefit hugely... but it probably won't benefit your average user's experience that much.It's still a huge boost for many users, and hopefully more software will be able to take advantage of all these threads as they become very common. With that said KL Refresh seems like a preemptive strike against Raven Ridge since Coffee Lake isn't ready. But that's not a bad thing... with a resurgent and more competitive AMD, end users are the big winners here.
Ratman6161 - Tuesday, August 22, 2017 - link
I get it. But...the systems we purchased that have the "U" CPU's in them (Lenovo X1's with 7xxxU's) they are all about being thin, light, and portable. For the users in question, more CPU power means little. Its hard to see the thermals of these systems allowing for any significant increase in performance for these types of systems, but I guess I'll await some benchmarks to see.willis936 - Saturday, September 2, 2017 - link
I own an XPS 13 so I am one of the "users in question". I spend a fair amount of time waiting on MATLAB DSP code to run. I don't want a 10 or even 5 pound laptop. I want a 3 pound laptop. If they deliver 4 cores that can sit at 2.5 GHz in a 15W package it would nearly double the speed of my 9343. That's something to be happy about.Samus - Tuesday, August 22, 2017 - link
What are you talking about lower speeds? The turbo's are the same or higher. With Speedshift 2.0 these will run dual core at the same or better clock speed as the old ones, while allowing for base or better clock speeds with all four cores utilized.An 8 threaded CPU at 1.6-1.9GHz will still outperform a dual core/quad thread CPU at 4GHz in heavily multithreaded apps, especially encoding/archiving.
And all that at 15w. AMD is definitely turning the heat up. Intel could have done this all along but the thread wars have begun, and Intel is clearly preemptively striking a Zen-based mobile CPU.
schizoide - Monday, August 21, 2017 - link
No, not really, because additional cores don't really speed up most mainstream desktop type of activity. IPC is much more impactful.This is nothing more (or less) than a direct response to Ryzen core counts.
III-V - Monday, August 21, 2017 - link
These things take years to develop. This has nothing to do with Ryzen.Hurr Durr - Monday, August 21, 2017 - link
And intel certainly knew for years how Zen will be structured.iranterres - Monday, August 21, 2017 - link
As AMD knew Intel's feature long ago, the problem was R&D money to invest on it.Samus - Tuesday, August 22, 2017 - link
That's part of the problem with the x86-license agreement AMD and Intel have. They get to peak at each others R&D. But Intel has been "out-fabbing" AMD for almost 15 years. The last great innovation AMD had was copper interconnects, and Intel wasn't too far behind.willis936 - Saturday, September 2, 2017 - link
I would say "the last great innovation AMD had" was multi die packages.MrSpadge - Monday, August 21, 2017 - link
They already had almost the complete design for desktops, this required only minor tweaking.tsk2k - Monday, August 21, 2017 - link
This has everything to do with Zen, it's a preemptive strike a raven ridge.uibo - Monday, August 21, 2017 - link
These are more like rebrands of the i7-7700HQeek2121 - Monday, August 21, 2017 - link
Intel has had a 6-core offering waiting in the wings for a couple generations now. The actual release of 6 core CL was a direct response to Ryzen. Without Ryzen we'd have seen another 4 core offering. Intel isn't going to raise it's own costs just to satisfy consumers.Alexvrb - Monday, August 21, 2017 - link
If it's a respin of an existing architecture, as the article implies, and they're moving an existing design down from the 45W stack and optimizing it for lower power... then no, it did NOT take years to develop. The deployment of 4C/8T designs at lower TDPs could very well have been in response to early glimpses of AMD's plans for higher IPC and more cores/threads (including mobile). Similar to what they've done shuffling higher-core designs down from server to HEDT - those weren't exactly ground-up redesigns just for HEDT. They used existing assets and burned off features for segmentation.Ratman6161 - Tuesday, August 22, 2017 - link
Don't care who was a preemptive strike against whom. Just care about what I can buy at the time I'm ready to buy. Who did it first is irrelevant to the consumer.nevcairiel - Monday, August 21, 2017 - link
In a difference between 2 and 4? There sure is a difference in even consumer usage.melgross - Monday, August 21, 2017 - link
Not when the specs are lower. So single core use with turbo is a whole 5% faster, Whoopi! Everything else is a lot slower. And you don’t get 100% more processing with a second core, nor do you get 100% of each added core. It even gets worse the more cores that are added. So with the lower speeds per core, plus the throttling, it will be interesting to see just how much more performance can be gotten.And real world use is never going to match the testing done on sites like this.
maroon1 - Monday, August 21, 2017 - link
2x core will still give you up to 1.5x boost in multi-core performance, even if they sacrifice clock speed, it still huge improvementhttp://cdn.wccftech.com/wp-content/uploads/2017/08...
And because of the high turbo clock, single and dual -thread performance will be same if not better
fanofanand - Monday, August 21, 2017 - link
I would advise against linking to wccftech if you want to be taken seriously around here.MrSpadge - Monday, August 21, 2017 - link
> Not when the specs are lower.Sure, "when". But there's not a single spec being worse. The all-core turbo clock is lower, but the 2-core turbo is definitely a lot higher (as on all modern Intel CPUs), probably the similar 100 - 20 MHz higher as the single-core turbo clock. And the 4-core turbo is very probably a lot higher than the guaranteed base clock, so for heavily threaded but mathematically light workloads clocks should be fine (well, a lot better than the baseline suggests).
> plus the throttling
Turbo is a bonus. If the full turbo can't be reached, that's not throttling. That's "not going into overboost". Throttling would be going below baseline clocks, which can happen with insufficient cooling.
Ratman6161 - Tuesday, August 22, 2017 - link
"In a difference between 2 and 4? There sure is a difference in even consumer usage."Well, I can't speak for everyone. But remember these are "U" CPU's, usually destined for ultra thin and light category machines. These systems prioritize form factor over performance and the thermals will rule the day. My experience with laptops is that they hit thermal throttling pretty quickly. guess we will see when there are some real systems to look at.
Lolimaster - Monday, August 21, 2017 - link
For mainstream typical load, an SSD brings up the biggest difference. Any cpu with 2cores+ @2.5Ghz or more is fast enough for media playback, office apps.artk2219 - Monday, August 21, 2017 - link
Someone hasn't tried to play an HEVC encoded video on said 2 core machine.imaheadcase - Monday, August 21, 2017 - link
Nothing to do with it.Netmsm - Monday, August 21, 2017 - link
I agree with your conclusion. It seems that Intel hadn't accurately predicted the real performance of Ryzen!Samus - Tuesday, August 22, 2017 - link
That just isn't true. Programs I use daily, such as WinRAR, benefit substantially from more physical cores, even if they are slower.Crazy1 - Wednesday, August 23, 2017 - link
A year ago (Aug 2016), leaked slides showed that Intel was planning on releasing quad-core 18W and 15W mobile Kaby Lake processors (Google "18W quad core"). This means Intel was planning this release before AMD's successful Ryzen debut disrupted some of Intel's desktop market share. As for the performance of these new chips, early Cinebench benchmarks of the 8550U indicate:- Single-threaded: equal to 7600U (Top 15W dual-core CPU)
- Single-threaded: significantly better than 7300HQ (bottom 45W Kaby Lake CPU)
- Multi-threaded: just shy of equal to 7300HQ
- Multi-threaded: significantly better than 7600U
(comparisons made using CPU benchmark list on Notebookcheck.net)
Note, these early benchmarks are of the 8550U, not the 8650U, meaning there is another processor with the potential for another ~5% bump in performance (based on clocks).
These early benchmarks indicate that whether you are currently running a 3+ year old mobile 15W, 35W, or 45W CPU, there are very compelling single and/or multi-threaded performance gains to be had by upgrading to the next generation of thin & light laptops based on Intel's 15W quad-cores.
imaheadcase - Monday, August 21, 2017 - link
Nothing to do with being competitive. Simply a refresh of current mobile offerings.Manch - Tuesday, August 22, 2017 - link
Now we know why they announced during the eclipse....North01 - Monday, August 21, 2017 - link
I think this could be an interesting match for something like a new Surface Book.damianrobertjones - Monday, August 21, 2017 - link
Or Surface Pro with fan.North01 - Monday, August 21, 2017 - link
I think it would be a bit out of place in a Surface Pro. The i7 Surface Pro, with fan, uses GT3e. These new chips only use GT2. Then there's also the throttling to consider.Currently, the Surface Book uses GT2 for both the i5 and i7 versions, which makes sense, as there's the additional dGPU in the keyboard.
Manch - Monday, August 21, 2017 - link
Maybe not these cpu's in a Pro, but a quadcore w/Gt3e or newer should flow down the pipe soon. Maybe in time for a spring refresh of the surface. Curious as to what AMD will offer as far as APU goes. If it's performant, then I wouldn't mind seeing it in a Surface Book.melgross - Monday, August 21, 2017 - link
The way sales are going, there may not be a new Surface anything.North01 - Monday, August 21, 2017 - link
Enlighten us melgross, on the way sales are going. Do you have some insider information? We have little to no information on Surface Pro or Surface Laptop sales given they started selling in limited quantities a little over a week before the end of Microsoft's last quarter. So please, let us know as to why Microsoft will stop making Surface devices and why we won't see this in a Surface Book.Crazy1 - Wednesday, August 23, 2017 - link
I can't speak for sales, but Consumer Reports just removed its recommendation for all Surface devices due to significant user reports of defects and returns. CR said 25% of 90,000 Surface device owners surveyed stated their device broke in under 2 years, which they also stated was significantly higher than most other brands. MS responded to the CR results, stating that their own breakage measurements were significantly different, but they failed to say whether or not CR's numbers were higher or lower than their own defect numbers. It's like Microsoft wanted to appear defensive without being capable of defending themselves. It's not a good look, but it's probably not damaging enough to deter MS from the hardware game.serendip - Monday, August 21, 2017 - link
What a naming nightmare. A 8xxx chip can have different architectures than its siblings but the only way to tell is to look it up on Intel's labyrinthine product site.Lord-Bryan - Monday, August 21, 2017 - link
Yeah, this is more confusing than last year's m7/i7 naming fiascoMrSpadge - Monday, August 21, 2017 - link
Would have been nice and constitent for them to simply put these into the mobile core i9 7000 class. But we know for mobile it's apparently still very important to show the biggest numbers, irregardless of actual hardware.nevcairiel - Monday, August 21, 2017 - link
Well as long as it keeps split between mobile and desktop its not that big of a deal. If they start mixing things more directly in the same segment, it might get more annoying.nonoverclock - Monday, August 21, 2017 - link
I use ark.intel.com to get the low-down on a particular CPU.artk2219 - Monday, August 21, 2017 - link
Check out CPU world, tons of different chips not just intels, and honestly more accurate. I've found less mistakes on there than on ARK, which is funny because it's intels on site on its own chips.http://www.cpu-world.com/
carewolf - Monday, August 21, 2017 - link
It has been that way for a long long time. I have a Haswell with a 5xxx number because it is a Haswell refresh with more cores. Similar with Broadwell refresh that got a higher number.Lolimaster - Monday, August 21, 2017 - link
Intel "arch" names are a joke.Sandy Bridge, Kaby Lake, Coffee Lake and Cannon Lake all of them have the same IPC. The only changes is a bit more mature process and the ability to sustain turbo speed for more time (which can be PR marketed as "IPC gain") + now more cache of the 300chipset 6core cpu's.
extide - Tuesday, August 22, 2017 - link
Uhh, Sandy Bridge definitely doesn't have the same IPC as Kaby Lake, and the last two you mentioned aren't even out yet. Plenty of review sites have done clock for clock comparisons with Sandy Bridge vs the newer arch's and the new ones are definitely faster.zodiacfml - Monday, August 21, 2017 - link
Not compelling. Not the best time to buy mobile CPUs considering it less than a year for Raven Ridge and Cannonlake.yankeeDDL - Monday, August 21, 2017 - link
The naming is absolutely horrific.It was not bad enough that the i5-7360U and i7-7560U are nearly identical, now the i5-8350U is expected to offer near 2X the punch of the i7-7560U released 6 months ago. I don't understand the point of releasing so many SKU with practically the same price.
Spoelie - Monday, August 21, 2017 - link
Competitive pressure from Raven Ridge - to turn around a probable loss on any and all highly threaded benchmarks to a probable win across the board, albeit at a higher price.If not for this, Intel would have been content selling you 2 cores at higher margins for some time longer.
bill.rookard - Monday, August 21, 2017 - link
Agreed. If the AMD is as aggressive in the mobile space with their upcoming APUs it's very possible that Intel could start feeling additional pressure in this space and the market is huge compared to the desktop space.I'm very curious how far down the Ryzen cores can be downclocked and still be stable. That could give us a window into what their power window would be. Yes, you'd have to add in power from the IGP, but it would still be pretty insightful.
ddriver - Monday, August 21, 2017 - link
According to this chart:http://image.prntscr.com/image/f6f90b7c98aa4d2dbce...
15 watts of power budget on a quad core ryzen outta give you 2.4 - 2.5 GHz sustained clocks. Which is actually pretty good, considering this is plain desktop chip. That's just CPU cores, no GPU obviously.
Lolimaster - Monday, August 21, 2017 - link
What I always say is that if AMD wants to be relevant on mobile they need to release laptops under their own brand and say a huge FU to laptop OEM's or at minimum give them 2nd priority over AMD's own branded laptops.Every time AMD has a decent APU for mobile at nice prices, OEM's simply cr@ptinated them in every possible way.
What you can find in retail stores
-Cheap TN screens (upgradable to IPS only visiting their online stores, with models well hidden in a store UI mess)
-30-45Wh battery
-single channel
-5400rpm HDD
-35w TPD models limited to 15-25w TDP (AMD really fooled themselves giving that kind of options to OEM's, cause they gave them yet another excuse to gimp the product)
Dug - Monday, August 21, 2017 - link
Yep, you are right on your comment. Hopefully someone at AMD realizes this.If they eat their own dogfood, maybe they will figure this out.
serendip - Tuesday, August 22, 2017 - link
AMD should pick one OEM and do an example device to show the rest what can be done with Ryzen, a bit like how Microsoft made the Surface tablets to get OEMs interested in making their own.jjj - Monday, August 21, 2017 - link
GenuineIntel Family 6 Model 142 Stepping 10So yeah it is a new stepping.
ddriver - Monday, August 21, 2017 - link
Wonder what's with the "genuine". Do they have lots of knockoffs?Hurr Durr - Monday, August 21, 2017 - link
AMD began its very history as such knock-off producer because IBM got cold feet on the processor deal.ddriver - Monday, August 21, 2017 - link
But nobody does intel chip clones, and nobody has done that for a loooong time. It is completely redundant to use "genuine". Not to mention that trademarking laws alone are enough to prevent and illicit use of a trademark.extide - Tuesday, August 22, 2017 - link
At this point, I think it's just because that's how they have always done it. AMD uses 'Authentic AMD' -- I think the engineers just get to have fun with that string honestly. Some VIA CPU's use 'CentaurHauls' Cyrix uses 'CyrixInstead' -- Check out this page for a bunch more examples: https://en.wikipedia.org/wiki/CPUIDddriver - Tuesday, August 22, 2017 - link
What fun indeed. Poor guys...Gondalf - Monday, August 21, 2017 - link
Not a single word about Cannon Lake....Intet had always launched new mobile cpus all together.....Core M apart.Summary:
1) Mobile 10nm Cannolake canned (likely unable to turbo high enough)
2) Desktop Coffe Lake is Cannonlake Ark on 14nm with some IPC enhancements
3) 10nm will be for Server first, like said Intel months ago.
My bet: very first 10nm product will be the new popular Xeon D with the Cannonlake core finally on 10nm
But i am sure wrong
WorldWithoutMadness - Monday, August 21, 2017 - link
and news from AMD.... no news for mobile counter part, raven ridge (not that botched desktop ryzen put into a laptop chassis)Silent as sign of confidence or as a sign of defeat?
Spoelie - Monday, August 21, 2017 - link
Except that this launch is probably a reactionary move from Intel, in anticipation of Raven Ridge. Raven Ridge is known, at least for the high end SKUs, to be one CCX = 4 cores and 8 threads.If Intel would not have launched these , they would again have a 50% core and thread deficit vis-a-vis AMD.
The fact that we are able to get these processors, ahead of any regular release schedule, is in a large part thanks to competitive pressure.
zepi - Monday, August 21, 2017 - link
Intel has had quadcores with hyperthreading for laptops since 2009. Where is the 50% deficit coming from?Manch - Monday, August 21, 2017 - link
4C/8T is new to the 15W U series. RTFAGondalf - Monday, August 21, 2017 - link
Yes it is pretty strange do not ear something of real about Raven. No a serious leak from Chinese OEMs, not a reliable early bench with verified clock speeds.We are at one month from the launch and there is nothing around and AMD is silent, it's all about idiotic TripThreads 180W cpus.
These new nodes are a damnation and they will do a lot of damage to the devasted Pc world IMO.
When we have to manufacture a crap SOC for phones they are nearly decent, they fail when we have to ship a real SKU, with the result of an incredible leakage at higher voltages and stunning power density all around.
Where is the good, old and reliable 32nm (vanilla and SOI) ?
The Benjamins - Monday, August 21, 2017 - link
What about the leaked benchmarks of the R5 2500U and R7 2700Ualfatekpt - Monday, August 21, 2017 - link
"The boxes confirm, among other things, that once these desktop parts will launch they'll have 6 cores (with HT for the i7) and require 300 series motherboards."so that means they won't be LGA1151? Sad news, hoped to get at least another iteration on my B250 board :\
A5 - Monday, August 21, 2017 - link
Same physical socket, but the pinouts are different on the board side so it isn't compatible.Ian Cutress - Monday, August 21, 2017 - link
LGA1151 could be for different chipset compatibility now, and no longer an easy identifier of what CPUs will actually work in what boardsBlueScreenJunky - Monday, August 21, 2017 - link
Nice to see the desktop CPU boxes mention the "UHD" IGP, it probably means we'll have hdmi 2.0 with HDCP2.2 out of the box on most 300 series motherboards.Ian Cutress - Monday, August 21, 2017 - link
We confirm the CPU natively supports HDCP2.2 out of the box.mooninite - Monday, August 21, 2017 - link
Yes, these are FINALLY going to have HDMI 2.0. Intel is "innovating" by bringing us this "brand-new" feature...zepi - Monday, August 21, 2017 - link
Any wort on the iris-equipped CPU's or 28w models?mooninite - Monday, August 21, 2017 - link
I hope we see Iris... The last desktop CPU that had Iris was Broadwell. :(extide - Tuesday, August 22, 2017 - link
On the desktop, yes, but in mobile no, Sky Lake and Kaby Lake on mobile both had Iris and Iris Pro offerings in mobile.limitedaccess - Monday, August 21, 2017 - link
"So despite Intel launching its 7th Generation family in January, today Intel is formally launching the 8th Generation only eight months later. To explain why Intel is breaking the usual 12-18 month cadence for the generation product, it comes down to product positioning."Is this completely accurate given the mobile context of today's launch? 7th Gen Kabylake announcement article was dated August 30th, 2016 on Anandtech - http://www.anandtech.com/show/10610/intel-announce...
tipoo - Monday, August 21, 2017 - link
Frigs sake, Intel. It's not enough that they bundle both Braswell and Broadwell into Celeron, Core Ms carry Core i* names, now this, 8th gen cores are both 8th gen cores and...7th gen cores. They're obfuscating what a processor is even more away from the consumer without doing some more digging, and it's really annoying.Ok, so more efficient node and doubling the cores place it well above the rest of Kaby Lake in the same wattage - well then call it something else, like U series quad. Otherwise core generations mean nothing.
maroon1 - Monday, August 21, 2017 - link
I personally don't see anything confusing. Most of the people here who care about tech stuff will get confused by thisThe average joe does not know what kaby lake refresh or coffee lake even mean
Also, it does make sense to use 8th genetation name because of the huge improvement. This thing will be 1.5x faster in multi-thread over 7th gen U. Thats the biggest improvement we have seen per gen in multicore performance.
If they use the 7th gen name then it will be more confusing because even the i5 model have quad with HT (but smaller 6MB L3 cache), and that would be more confusing when you have new i5 quad with 7th gen name while the older one have i7 with dual core with same 7th gen model name. !!! It is better to use new 8th gen
The coffee lake model won't increase the core count. It will probably just have higher clocks with maybe 5% IPC. THey could call it i7-8750U (which sound better than 8650U). It is basically just faster quad core model, and it not worth calling it 9th gen
Cannon lake will probably released for 4.5w model, and dual core only. So, if this true then it is not aimed to relplace kaby lake refresh and coffee lake-U model
maroon1 - Monday, August 21, 2017 - link
Sorry I meant" Most of the people here who care about tech stuff will NOT get confused by this"
Glock24 - Monday, August 21, 2017 - link
"Also, it does make sense to use 8th genetation name because of the huge improvement. This thing will be 1.5x faster in multi-thread over 7th gen U. Thats the biggest improvement we have seen per gen in multicore performance"By your logic, if a "huge improvement" was needed no name a new generation of Intel CPU's as "new", we'll still call all Intel CPU's 2nd Gen Core.
They could have used a prefix like UQ to denote quad cores, like they do on the 45W quads (ix-7xxHQ), and also a higher part number like i5-7400UQ or something similar.
Lolimaster - Monday, August 21, 2017 - link
There's no IPC gain in any intel cpu until Ice Lake.The "PR IPC" intel is giving this days is basically the ability to sustain Turbo Speeds for more time due to a more refined node, in the case of the 6 cores cpu's for 300 chipsets, you got extra L3 cache on top of that which also bumps the performance a bit-
maroon1 - Monday, August 21, 2017 - link
So what ?You are still getting big boost from higher core count and increased clock speed.
Why IPC gain even matter ?! If future processor (hypothetically) can clock as high as 10GHz then it is better than anything we have now even if IPC is worse by 10%
As long as you have faster CPU with same price and TDP of previous generation then that an improvement. It does not matter if those improvement are clock speed, higher IPC or more cores
amosbatto - Sunday, October 29, 2017 - link
IPC still matters for single threaded applications, which is most software. As a computer programmer, I can tell you that a lot of code can't be multi-threaded and writing multi-threaded code can be very tricky, especially in C/C++. One of the reasons why so many programmers are switching from C, C++ and Objective-C to Swift, Go and Rust is because these languages make it so much easier to write multi-threaded code. There are a significant proportion of old-school programmers who share Donald Knuth's opinion that multi-threaded programming isn't worth the pain. So, yes, IPC still matters.Silma - Monday, August 21, 2017 - link
I thought Intel promised that Thunderbolt would be integrated to all of its next gen processors, ie Gen 8. That doesn't see to be the case.Hurr Durr - Monday, August 21, 2017 - link
This generation is being launched in two stages, all new connectivity will come later, somewhere in 1Q 2018 probably.SaturnusDK - Monday, August 21, 2017 - link
And will likely require an entirely new motherboard, not compatible with z370.duploxxx - Monday, August 21, 2017 - link
lol: the base frequencies are down slightly.in most cases they are down 20% i would not call that slightly
Glock24 - Monday, August 21, 2017 - link
20% less frequency is still considered "frequencies are down slightly". But if it were 21% less frequency, then it would be "frequencies are down A LOT! Take out your pitchforks and torches!"LOL!
carewolf - Monday, August 21, 2017 - link
If they still have the same boost maximum, that shouldn't really matter much though.maroon1 - Monday, August 21, 2017 - link
http://cdn.wccftech.com/wp-content/uploads/2017/08...8550U is still getting around 50% boost in cinebench
Single-thread will be probably same if not better because of the high turbo core.
Kakti - Thursday, August 31, 2017 - link
That just tells us that Intel's processors are still stable when running at an even lower speed/power use. It's more efficient to have the CPU slow down when not in use, and then turbo back up once it needs to. No sense in having your CPU running at 4.5ghz all day, especially on mobile parts.Da W - Monday, August 21, 2017 - link
Shows Intel *COULD* sell us quad core before, they just didn't want to. Don't give a fuck about +3% framerate on games, gonna buy from AMD as long as they make something good enough.YoloPascual - Monday, August 21, 2017 - link
Lucky intel nvidia made MX150. They were about to lose it all with the upcoming iGPU of raven ridge. 500-800$ laptops will still be intel+nvidia combo.Lolimaster - Monday, August 21, 2017 - link
Raven Ridge should destroy the GT1030 for mobile considering their GPU sits in between RX550-RX460 with dual channel DDR4 GT1030 has zero chance vs the RX550.RX550 512SP Polaris
Raven Ridge APU 704SP Vega
RX460 896SP Polaris.
BrokenCrayons - Monday, August 21, 2017 - link
I'm not sure about Raven Ridge's GPU performance yet. The GT 1030/MX150 has dedicated VRAM and dedicated VRAM bandwidth. It presumably still uses some form of what was once called TurboCache back when Nvidia released the 6200 back in 2004 or 2005 so it'll also have access to system memory's dual DDR4 channels. So there's about 40-50GB/s for the GT 1030's 64-bit GDDR5 and then there's whatever else can be nabbed from the RAM if the CPU or other system devices aren't using it.I don't think the same can be achieved easily if Raven Ridge's iGPU has to fight with the CPU cores for system memory bandwidth which is going to be higher latency than the GT 1030's GDDR5 and there's still the problem with concentrating all that heat into the CPU package. Additionally, Vega isn't much of an improvement over GCN so the 512SP count in prior generation APUs was already only barely competitive with the GT 730 (in a lot of cases, it fell behind in benchmarks due to bandwidth limitations) so the additional SPs won't be that much faster compared to prior generation SPs and they'll likely clocked lower than a dedicated RX GPU because of thermal limits so they may not offer a substantial increase over prior gen APUs or against the GT 1030/MX150.
Don't get me wrong, the A-series iGPU was a competitive product against Intel's non-Iris graphics and replacing those older/weaker CPU cores with Zen is a good thing, but I don't think you're going to see Raven Ridge end up in a position of dominance over the GT 1030. We really should take a wait-and-see approach and then look at benchmarks as they become available before predicting the demise of the 1030 and its mobile variant.
Anonymous Blowhard - Monday, August 21, 2017 - link
NVIDIA hasn't used "TurboCache" on their dGPUs since the 6000 series - it's long dead. I'd be very surprised to see it reintroduced for the sake of expanding a 2GB VRAM pool on a 384-core chip.For Raven Ridge, dual-channel DDR4-2400 can do a theoretical 38.4GB/s of memory bandwidth - if you account for controller overhead, that's probably more like 32GB/s. Not that far behind the MX150, but as you mentioned it's got to share that with the CPU. Newer game engines that can take advantage of shared memory addressing on the APU might be able to offset some of the performance lost there.
For clock speeds, leaked benchmarks had the GPU clocks at 800MHz, but I don't know what the TDP was. I'm imagining the 35-45W range for the higher end ones.
tl;dr - Release something already AMD, so we can see what you've got.
AleXopf - Tuesday, August 22, 2017 - link
I expect the top of the line apu to be:One ccx 4c/8t
Interconect to the iGPU instead to another ccx
RX550 performance because of downlocking for power budget and efficiency
2gb of hbm2 @ 1ghz as l4 cache/graphics memory
JoeyJoJo123 - Monday, August 21, 2017 - link
>that once these desktop parts will launch they'll have 6 cores (with HT for the i7) and require 300 series motherboardsGetting pretty tired of this socket/chipset change a year keeps the goyim in fear meme from Intel. Can we not just get some backwards compatibility with older motherboards for once?
Lolimaster - Monday, August 21, 2017 - link
Buy AMD and be happy?Why support the one who spits in your face?
AM4 support
Ryzen 14nm
Ryzen Optimized 14nm
Ryzen II 7nm
Ryzen III 7nm+ (probably with some AM5 and Ryzen III with dual DDR4/DDR5 memory controller like Phenom II)
Hurr Durr - Monday, August 21, 2017 - link
Ten years of nonperformance is the real spit in the face here.JoeyJoJo123 - Monday, August 21, 2017 - link
I plan to. But I would also prefer if Intel dropped its unsavory business practices, too.jameskatt - Monday, August 21, 2017 - link
Apple needs to simply BUY AMD and make its own Intel-compatible chips.This is unbelievable. Intel creates 8-core chips for laptops but did not give it a memory controller for more than 16 GB of LPDDR4 RAM.
This is so stupid. So asinine.
Intel is so SLOW in upgrading its hardware and so stupid in doing so.
This is why Intel Chips today are not that much faster than chips made 5 years ago - 2+ generations of computers ago.
Teknobug - Monday, August 21, 2017 - link
Apple would make AMD very unaffordable.Ktracho - Monday, August 21, 2017 - link
As in unobtainable (except in their own products).V900 - Monday, August 21, 2017 - link
Do these even make sense in a sub 15W laptop, and isn't this a regression rather than a step forward?You're basically losing single core performance and turbo speeds, in machines that are more likely to rely on single threaded performance than anything else.
Sure, multi threaded programs might be more common in the future, but guess what: These four cores will naturally be limited in performance just because of the low TDP...
Jumangi - Monday, August 21, 2017 - link
Another minor refresh passed off as a "New Generation" of processor....Intel is getting so lazy they just don't care anymore.maroon1 - Monday, August 21, 2017 - link
So, ~1.5x speed boost in cinebench multi-thread at same 15w TDP is minor refresh ?!Alistair - Monday, August 21, 2017 - link
Lower single threaded and no significant improvements to the integrated GPU. I'm "meh" until I see actual products at good prices. Mac mini for the same price? :)maroon1 - Monday, August 21, 2017 - link
Single-threaded will be same if the better. Look at turbo boost manmaroon1 - Monday, August 21, 2017 - link
The integrated graphics still using 620 brand. They are not calling it 720 as if it is a new GPUOnly major improvement in multi-core performance (which is bigger than anything we have see from other generation). Single-thead will be same if not better because of the high turbo boost
Lolimaster - Monday, August 21, 2017 - link
Intel continues being a joke of his formers self. they can't compete at any price bracket with the Ryzen onslaught (except sub <$80 cpu's).Users who bough overpriced i7 quads are being spitted on their face, Ryzen 5 quad's were a clear message if anyone had a doubt.
Bullwinkle J Moose - Monday, August 21, 2017 - link
EXCELLENT performance improvementsBut 4 core is just a curiosity at this stage of the game
Hopefully 4 core fanless with 8GB Ram will be reliable enough with 10nm Canon Lake's power envelope to be the standard minimum spec for a laptop without to much throttling.....
But I doubt it
They will probably choke and run too slow or hot when gimped with 2 or 4GB of Ram like the usual 64bit lineup
8GB PLEASE!!!!!
iwod - Monday, August 21, 2017 - link
I cant help but wonder, Apple shipped a Macbook Pro knowingly there is a Quad Core version coming in less then three months time?* These 8th Generation can be TDP-up to 25W.
*I am pretty sure there are lots of Programmers who dont need the Iris Pro graphics.
hechacker1 - Monday, August 21, 2017 - link
You have to remember that that extra cache for Iris is shared as a L4 basically, so you are getting better performance all around. And MacOS already does a lot of video processing for just the GUI, so putting anything weaksauce in there would make it feel really slow. Especially considering the capability to drive two 4k monitors with thunderbolt 3.But they could easily put one in a Macbook. However, I doubt they will ever do that because it'll eat up their MBP sales.
Lolimaster - Monday, August 21, 2017 - link
AMD lineup for mobile will probably be:Raven Ridge 4c/8t-4c/4t-2c/4t and maybe some 6-8 cores for desktop replacement laptops.
Ryzen is so efficient that the R7 1700 can easily fit on a 45w TDP with minor tweaks to turbo and base speeds (Asus releasing a full blown gaming laptop with it).
SanX - Monday, August 21, 2017 - link
Same unanswered question: why Intel's $/mm2 is 10-20 times larger then Samsung's or Qualcomm's $/mm2 and 50x larger for server chips?extide - Tuesday, August 22, 2017 - link
Simple Answer: Because they can.SanX - Tuesday, August 22, 2017 - link
Then another question: Why no one in so called "tech media" discusses this?fanofanand - Monday, August 21, 2017 - link
This would seem to prove that Intel really has been resting on it's laurels, and AMD has pushed this launch forward at least a few months. Kudos AMD!m16 - Monday, August 21, 2017 - link
So... they're bringing their mobile parts to Haswell era 4 cores / threads and they're basically releasing the same architecture as Skylake with only added 4K video playback?That's kind of sad, really. I mean, if anything, if the battery / performance ratio has been significantly increased, that is something worth celebrating, but otherwise, this is just a marketing ruse. It should be more like: Hey people! We finally can have Haswell era performance on laptops!
extide - Tuesday, August 22, 2017 - link
FWIW, 4c/8t has been available in laptop since Nehalem, and practical since Sandy Bridge. Nehalem 4c laptop parts were 45/55w TDP with very low clock speeds, Sandy got the clocks up quite a bit, and once we got to 22nm Ivy you could get 4c into the mid 3Ghz range in 45w. I wonder what the sustained perf will be, getting 4c/8t in 15w is pretty impressive if we can sustain 3Ghz+ clocks on all 4 cores.Ro_Ja - Tuesday, August 22, 2017 - link
Not to mention, Core 2 Quad had a mobile version too and that is even more power hungry.Santoval - Monday, August 21, 2017 - link
"We expect that Intel’s 8th Generation will eventually contain three core designs of product on three different process design nodes: the launch today is Kaby Lake Refresh on 14+, and in the future we will see Coffee Lake on 14++ become part of the 8th Gen, as well as Cannon Lake on 10nm."So +2 cores at exactly the same process make the CPUs "a new generation", while +2 cores for desktop in a slightly tweaked desktop process also make the CPUs a new generation, and a new freaking *process node* makes Cannon Lake the *same* generation. I wonder why Intel is undermining the meaning of a generation from both sides of the concept.
extide - Tuesday, August 22, 2017 - link
Prior to Ryzen coming out Intel had settled into a very predictable pattern every generation, but post-Ryzen has been all over the place!serendip - Tuesday, August 22, 2017 - link
After all the Lake names, I'm sure Intel is going up Sh-t Creek.systemBuilder - Tuesday, August 22, 2017 - link
You just have to laugh at the balls they have releasing a graphics GPU that is no faster than the 2013 Iris Pro 5200 it's been almost 5 years and not a single inch forward in progress on most fronts on the GPU side. Intel, asleep at the switch, once again ....extide - Tuesday, August 22, 2017 - link
Actually they released Iris Pro Graphics 580 (72EU) in the Skylake gen which was ~2x the perf of the Broadwell era Iris Pro (48EU). No Kaby Lake 72EU part though.yhselp - Tuesday, August 22, 2017 - link
Any word on whether Intel plans to release 15W quad-core parts with Iris Plus graphics? It'd be a real shame if they decided to drop the Crystawell/eDRAM SKUs - having ~100 MB/s of memory bandwidth, or the equivalent of running 6300 MHz RAM in dual-channel mode, is awesome, even within the 64 MB limit. Maybe Apple and Microsoft can put some pressure there.AleXopf - Tuesday, August 22, 2017 - link
It makes mores sense add 2gb of hbm2 @1ghz for 128Gb/syhselp - Tuesday, August 22, 2017 - link
It does, except it'd be only 2 GB compared to up to 32 GB, which is hardly acceptable. And what would it cost in terms of power; would it fit in 15 W without impacting the already low frequencies, or at all?peevee - Tuesday, August 22, 2017 - link
So, it is August 2017 and the top new part is 4 cores at 1.9GHz for $409. By comparison, mobile Core 2 Quad Q9000 from August was 4 core at 2GHz for $348.9 years. It used to be that in 9 years you'd get something like 60x performance (doubling every 18 months).
peevee - Tuesday, August 22, 2017 - link
BTW, to the Intel marketing people.I am still using Late 2008 fully aluminum 15" Macbook Pro with Core 2 Duo 2.4GHz and switchable Intel/Nvidia graphics, with upgradable RAM (which I have upgraded), easily replaceable HDD (which I have replaced with SSD) and easily replaceable battery (which I have replaced 2 or 3 times). And don't see anything worth replacing it with, especially with the insane trend of everything being soldered on in the latter models (making them obsolete much earlier). You are not going to get my money with the sht now on sale.
HStewart - Tuesday, August 22, 2017 - link
Here is one thought to consider,Maybe Kaby Lake R is what some are calling Coffee Lake. Intel did mention in slides that more cpus are coming this fall. With 8th generation cpus support 4 cores on U processor - it sounds like they have updated the process.
This is quad core on U processor, imagine what this means for normal mobile cpu's - like 6 or 8 core cpus coming and then higher performance mobile chips could be even higher - on mobile.
My thought on generations
8th - Kaby Lake R - possible previously called Coffee Lake
9th - Cannon Lake 10nm - ultra power only to test platform.
10th - Ice Lake - 10nm+ - Mainstream 10nm
With the other news about process changes - one thing that is likely coming is more system on chip for more compact systems.
Ian Cutress - Friday, August 25, 2017 - link
Android Password Breaker hacking tutorials hacking ebooks hacking news hacking tools android technology https://myhacker.netsolnyshok - Tuesday, September 5, 2017 - link
if they put 10nm quad behind laptop screen (as in current Surface and other 2in1 tablet+keyboard designs) and use whole back-of-screen area for heat dissipation, they just might get away with it. Maybe they will have to dial TDP down to 10-12w. But it would make for a pretty nice fanless design. With 32GB of RAM please.