I hope Anandtech publishes long review on ryzen. im a video editor and im pretty stoked about this if the 8c/16t version is affordable, would make my video editing workflow a breeze. Also i hope ryan and ian does youtube videos for normal people!
Agreed. I'm more interested in the hex-core version as that would probably be a bit more of the performance/price sweet spot. And to just highlight this (again): pricing will be crucial. I get that they need to make money, I get that they can't give it away for nothing, I get that if it's close to Intel performance they can charge close to Intel pricing, but they NEED market share.
Yeah, I would say AMD needs to focus on increasing market share so that people start seeing AMD devices around and consider them viable. Plus being 10% behind Intel in performance but offering a 10% discount on price so they offer the same value just won't be enough, because most people would still pick Intel in that scenario, if the value is the same. So their offers need to look very appealing in terms of value.
They need profits, too. I find it silly to demand AMD to catch up with Intel and once it looks like the might, people don’t want to pay for it. Sure, it has to be good value for money, but it’s starting to sound like everyone will be dissapointed if AMD doesn’t make those Ryzen super cheap.
If the CPUs perform well, people should pay for it.
If VW wanted to, they could probably launch a car tomorrow that would have the look and feel of a BMW, but they would struggle if they expected to be paid the same amount of money. Heck, if somebody simply replaced the BMW badge with VW badge, the value of the car would instantly be reduced. Or vice-versa: BMW could put their own badge on a VW, and instantly increase its value.
The badge carries value in itself, and AMD is simply not where Intel is in terms of name recognition.
Humans are dumb. Which explains why the advertising business is as big as it is.
If the product is good enough, then they don't have to be cheap. I can remember back when the Athlon 64 X2s hit, they where fast chips and had some serious prices on them. It was north of $500 for the X2 chip I purchased and was worth it compared to the P4s of the time.
Considering Intel is currently charging $1000 for their 8core 16thread chips, even $500 for a 8c16t Ryzen would be a complete steal!
Yeah, my guess is you will see something like: $500 -- 8 - core $300-350 -- 6 - core $200-250 and below -- 4 - core Probably wont see dual core -- maybe as mobile APU's but I doubt it in the non IGP chips. I also bet AMD will leave hyper threading on all but they very lowest end or possibly not even disable it on anything.
One thing to remember is.. Intel can charge what it wants with no competition. Don't expect Intel pricing from AMD in the short-term even if they have a good chip..
I had previously estimated any time between February 21 and February 28. The silicon gas been finalized and already is being manufactured. It takes 6 weeks to fill the channel. It will launch before February 27th.
Naturally, no self-respecting professional would line up to be an early adopter of any tech. I myself would gladly get a much needed break from intel chips, providing performance is good and price is competitive, but not before the retail product and platform go through at least 6 months of testing.
This is a processor, not the first model year of an "all-new" or massively revised car model. By the time the first retail chips are in consumer hands, they usually have a pretty good handle on reliability. If there's issues, they can be corrected by microcode. If you're concerned there's going to be a major bug that when fixed affects performance, just wait a couple of months for the community to vet the chip. There's no need to entirely skip the first revision on every product, if that revision turns out to be good.
AMD could not possibly test the chip under every real world situation it could call under. They don't run people's apps, they don't crunch people's data. Specific flaws will always slip through factory testing. Phenom's TLB bug, skylake's fft bug, the original pentium's fdiv bug and several others have all slipped through testing.
Also the fact that it is a processor and not a car makes it significantly more prone to bugs. Cars are primitively simple compared to cpus. There are billions of things in a cpu that can fail, whereas a car has only few thousand components.
I've rarely been burnt, and it was never anything I couldn't work around. Early adopters are often those who want to tinker with the latest tech and know the risks. If you know what you're getting into, more power to you. With that being said I don't need the latest and greatest anymore, so I probably won't be looking too deeply at Ryzen for a while. But those who leap in headfirst are just what you call "enthusiasts", many of whom are professionals and new tech is their hobby. Enthusiasts are a major staple of sites like these. Otherwise they'd change the site name to "Stabletech" and only review something after it's been on the market for a few years (OK OK phone reviews do NOT count, strictly talking PC components here).
Are you serious? Professionals will pick up promising tech EARLY to properly evaluate the products. Professionals will not suggest tech before they can properly evaluate the products. It is called the bleeding edge of technology, and WE are expected to know about this stuff before the general public.
Normal people can buy perfectly reasonable 12 core Intel workstations off ebay for $600. They aren't the latest architecture but the encoding FPS per dollar cant be matched.
I'm mostly a gamer. As long as it's competitive the differences in CPU performance will be negligible. So price is probably the most important factor for me at this stage.
One further thing to consider: Even if Ryzen is at the same price/performance, we should buy it. Whenever possible support the underdog. We've all seen the crap Intel pulls when there is no competition.
I'd really like to support AMD to at least help maintain some amount of competition in the market, but as Anandtech articles have already pointed out in the past, many AMD processor-equipped laptops are stuck on single channel memory even though the CPU can handle dual channel. It's not a big deal until the system has to turn to RAM to support the graphics card. That really kills the whole AMD thing for me...well there was that and the anemic performance of the Athlon x4 860K I purchased last year that was at best, only a little faster than the Q6600 it replaced.
Maybe Ryzen can turn things around. I'm hopeful for that and for improved Linux support for AMD GPUs. I haven't use an AMD GPU in quite a few years due to their weak showing in Linux even though something like a C-70 or e-450 was definately on my to-buy list back when I was netbook shopping. Ultimately, I ended up with an Atom n450 because its GMA 3150 had better driver support at the time.
They could easily get me if they make it FULLY win7 compatible also. I can't see buying either side if it forces me to win10, telemetry crap, Desktop trying to act like a mobile device etc etc (jeez list of hate I have here is long). I'm almost certain many forms of *nix will have full support so maybe they'll get me that way and my current desktop could be my last WINTEL PC finally. Bring on the Vulkan games so I can finally make a move without the need for Dx or Wintel.
When you write the review, please include a discussion of what is MISSING if you choose to install on something OTHER than windows 10. I'm guessing that will be very important to MANY people.
Probably not important to many people, as the subsection of people whose tinfoil hats cover internet browsing but not Windows 10 usage are quite small. I'm surprised and disappointed your Anandtech comment section part hasn't burned through yet from all of those dangerous transmissions.
Thus launch is completely irrelevant for gamers. All we care about is single-threaded performance, and pretty much every gamer today is all set for many, many years to come with Intel CPU's.
The only thing that is of importance to gamers today is GPU's, and VEGA will come in late June and the 1080TI in Q3 or Q4 2017.
Plenty of gamers are still running on Sandy Bridge or Ivy Bridge because there has been little reason to upgrade. More cores could be a reason, provided performance is also improved.
Thank you wise gamer for speaking on behalf of all gamers. Maybe I should stop consider myself one because I'm not set with an Intel CPU and I'd love to get an 8 core beast in my pc, sacrificing a bit of single-threaded performance.
This launch is completely relevant for gamers. The vast majority of gaming builds fall in the $1200 or less catagory. With Intel giving it's customers nearly nothing in terms of IPC since Sandy Bridge there are a large amount of people out there still holding on to Sandy Bridge and Haswell i5's and i7's. So if you can pick up a quad core Ryzen cpu, motherboard, $30 cooler for the same price as a Kaby Lake i7 alone and then overclock it to near equal performance it's a no brainer.
With more and more games being able to efficiently utilize more than 4 large threads a Hexacore Ryzen cpu will outperform a quad core i7 hands down. This will only increase in AMD's favor as DX12 and Vulken become the standard for new games.
The most important thing about Ryzen to people who don't want/need a new AMD cpu is that it will pull down Intel's ridiculous pricing. This is very relevant to all considering a new PC purchase.
You only feel that it's irrelevant because it may not apply to personally and you've come to accept the dribble's of nothing that Intel calls a new generation of cpu "performance" as good enough.
It's a little bit awkward to note, but Intel's actually been churning out decent-sized performance gains for the past 5 years. Jumping from "worse than our Phenom X6" to "Came out this decade" performance doesn't deserve this level of loyalty, and Intel's been delivering consistent product improvements that are easily worth the upgrades. As an owner of both a Sandy Bridge 3960X and a Skylake 6770HQ, the latter is significantly faster than the former at everything except the longest prolonged workloads despite running at less than half the power consumption.
It's pretty amusing that you're calculating a quad-core Ryzen, motherboard, and cooler at i7 pricing. Given that a 3-year-old quad core (8350) and its top-end motherboard barely squeak under that number, there's no way a new quad core and its brand-new top-end motherboard squeak under. And don't think you're using much less, as the lower-end motherboard chipsets aren't going to get the type of overclocking you're dreaming of.
Now, let's talk about games. And how DX12 ironically has the opposite effect of what you're talking about. DX12 was in AMD's favor because it reduces the CPU load and makes more events non-blocking. (For reference, it's *Vulkan.) As more things become non-blocking Intel's large threads become more relevant, not less. Things that would be restricted to 6-core machines are available to 4-core machines, as the 4 cores can always be fed instead of sometimes running out of Out-of-Order Execution depth with blocked calls (this already is rare, as, again, unless you're using a Bulldozer or Piledriver, it's pretty hard to be CPU-bound).
A 6-core Ryzen literally cannot cost less or even with an Intel 4-core. Intel's manufacturing process is cheaper, more mature, more dense even at the same node (both TMSC and GloFo are stat-padding), at a smaller node, and in-house. Sorry to break your bubble.
Intel's pricing is hardly ridiculous. In fact, they've done admirable advancing technology in the effort to beat themselves. They were, and still are, their own best competitor.
The only thing you could really argue in AMD's favor is their IGPs, which will help them win their consolation prize in the low-end markets again. Sorry to break it to you.
Intel's markup is absolutely massive and their performance reign over the last (nearly) decade is a perfect example of why we need competition.
You argue in one paragraph that Intel's process is both cheaper and more mature, and then in the very next line that their pricing is "hardly ridiculous". FYI, the $1700 10 core 5960X has approximately the same die size as an i5 2300 which sold (presumably still at a healthy profit) for $187 almost 6 years ago. If Intel really does have a "cheap" and "mature" process, they're certainly not passing on those benefits to their consumers.
Let's forget die size and try it by transistor count for a second... Did you know also that the often derided as stupidly overpriced iPhone 7 starts at $650, and sports an A10 SOC with 3.4 billions transistors. That's slightly more transistors than a Broadwell-E CPU, which is $1700, $1000, ~$620 or ~$430 depending on the SKU, and the Intel CPU doesn't come with a premium phone. We don't know the exact transistor counts of Skylake processors, but the $330 i7 6700K has an estimated 1.75B transistors. That's far, far fewer than the RX 460 (~3 Billion) which can easily be found with superior cooling solutions, power delivery and GDDR5 RAM(!) for under $100.
Of course these aren't accurate apple to apple comparisons and they're just rough estimates. But they offer an indication of just how much Intel charges for tiny chips with relatively small numbers of transistors. Of course Intel is a company out to make money so they can charge whatever the market will pay.
BUT, it shows just how important competition is, which is precisely why so many people are excited, cautiously optimistic, or at least hopeful for a successful and competitive CPU from AMD.
Hahahaha xD. So much stupid in such a small post. I guess the 6-core/12-thread Ryzen 5 1600 being SIGNIFICANTLY cheaper than Intel's 4-core/4-thread i5-7600K (let ALONE the 4c/8t i7-7700K), while absolutely bitch-slapping it up & down the block in terms of performance was all just a false memory I'd manufactured in my head. Thanks for letting me know that! ;)
You don't know what you speak of. Ryzen has been shown to rival Intel in single thread performance *even handicapped*, and games are increasingly multithreaded. If the price is right it's relevant to this gamer (who has probably played computer games before you were born), and I'm upgrading regardless. I figure my Bulldozer deserves a break.
I think there is far too much being read into that little blurb. According to AMD, "RyZen" has already been "launched" at least in as much as they are showing off engineering samples.
People tend to forget that it takes usually 2 - 3 *months* to go from the start of real mass-production to when you can actually buy a real chip at retail.
If AMD was actually going to have these chips on store shelves in the next 6 weeks, they would not have been showing off "engineering samples" at CES. They would have been showing off 100% finished retail chips.
They can meet it using Corporate Speak that people around here don't truly understand.
They can "ship for revenue" in Q1 meaning they book the payments from OEMs and retailers for the first batch of RyZen parts. They can even say that they've "shipped" them. That doesn't mean that you can buy one though.
This isn't some theoretical exercise, you can't throw a rock on Google without running into reviews of fully operational 7700K Kaby Lake retail processors that were posted last November. You still couldn't buy one until last week though. In this case, nobody.. NOBODY.. has seen a full retail sample of Zen or "RyZen" floating around anywhere. Do the math, the laws of economics apply to AMD just like they do to everybody else.
Initially I thought you’re being too harsh, but the point about Zen not leaking out is a valid one. Either security is super tight—or there aren’t any CPU production lines ramping up yet. In the latter case, the launch might be Q1 on paper, with widespread availability in Q2 or even Q3 (which is what people need to care about anyway).
Also, GDC peeps, if interesting to AMD, get working machines easier than any of us.
ISTR that when AMD released Bulldozer they were available within two weeks. I suspect AMD's launch is different from Intel's. (Heck, they do a lot fewer of them for starters. :D)
Well,this i how this reality in 3d works... production of any goods takes time and shipment takes time\ plus...the human quality of service is very bad while the technological side is very good many hosting sites prefer to let me run into big minuses with my 8 websites rather than bother with answering my questions because support made by intelligent creatures not robots...is very expensive and very stupid
They have been showing off engineering samples for months really. Also, Q1 ends on March 31st, and that's a bit over 11 weeks away. I figure they'll have no trouble getting to the market in time.
True that, but do you really disagree with Krysto's contention that buyers would still go with Intel if the AMD offering is 10% less efficient and 10% less expensive?
It would be utterly idiotic to launch it during big events. The point is to get more attention not less. My bet is Feb 16. 16 threads, they had 16 mobos when announcing the AM4 the other week,maybe they got a thing for 16.
Based on binning rates and such, I speculate this:
8c/16t ($699-899+), very low numbers/least availability, low yields. 6c/12t ($399-499, $599 if 4GHz+), slightly defective 8c/16t bins Mainstream: 4c/8t ($250-299) highest yields, most availability 4c/4t ($150-199) same, just no SMT
Clocks will determine pricing within each CPU group. Highest clocks, highest price. If you were expecting 8c/16t for Intel's 4c/8t prices, what were you smoking?
Since they're all multiplier unlocked, they compete with Intel's K line of Kaby Lake chips.
I think your pricing estimates are a bit too high. I see AMD slotting their CPU pricing as this:
4c/8t = i5 Kaby Lake = $179-239 retail price 6c/8t = i7 Kaby Lake = $279-349 retail price 8c/16t = $379 - $499, maybe more for a very limited quantity, highly binned, Super FX Black VIP Edition...or whatever they call it.
That's adorable. You really think they're just going to charge $30 more than their current 4c/8t 8350 for a 50% node shrink and multiple years of engineering?
If that's the case, they're going to miss all the benchmarks they're claiming they hit too.
The leap from 32nm to 14nm means that the new chips will be much cheaper to produce. And the 8350, whatever its limitations, has always been marketed as an 8c/8t chip.
Also, they need these chips to actually sell, and despite some decent benchmarks against Broadwell-E they will likely still have somewhat lower clocks and IPC than Kaby Lake.
They need to position the 4c/8t chip to undercut the i5, the 6c/12t to undercut the i7, and the 8c/16t to undercut the 6c/12t Broadwell-E.
That means 4c below $200, 6 core below $300, and 8 core below $400. The top-of-the-line 8c may be closer to $500.
I don't know what they will do, of course, but that's what they need to do.
The FX-8350 was marketed as a true 8-core chip, but it only had 4 modules (2 int per module, 1 shared float). So, in reality, it's a 4c/8t chip.
Your pricing estimates, and those of SquarePeg, are extremely optimistic. They may slightly undercut Intel, but with the performance Ryzen has been showing, it'll be priced accordingly.
8c/16t is a low-volume, ultra high-end flagship processor. That alone demands flagship pricing.
But it's simply not as high end a part as Broadwell-E. Doesn't have quad-channel RAM or 40 PCIe lanes for high-end workstations. And AMD have to rebuild sales from basically nothing.
They have to undercut Intel at each price point, and that means AMD 4c/8t will compete on pricing with Intel's 4c/4t, AMD 6c/12t will compete with Intel 4c/8t, and AMD 8c/16t will compete with the low-end Broadwell-E, which is a ~$400 part.
They price their video cards to compete; don't know why you expect them to commit suicide on their new CPUs.
Also don't know why you think yields will be so bad that most 8c dies will end up sold as 4c parts. They've been shipping 14nm GPUs in volume for months now.
I think AMD scaled Ryzen in such a way that 8c/16t parts will be cheaper than Intel's offerings because of missing on-die features you mention relative to Intel. That's why I put it under $1000 bracket. It can be anywhere from $500-799, while Naples will be over $1000.
So, if 8c/16t Summit Ridge is 1/2 scale of 16c/32t Naples server part, then it should have 1/2 PCIe lanes and memory controllers. That means the 4c/8t versions share the same die if they retain 20 PCIe lanes and dual-channel memory, but the other cores were either defective bins or intentionally cut. That's getting really speculative though. It makes sense though, as Intel stuffs an iGPU in the Skylake and Kaby Lake parts to fill the die.
So, if Naples has 40 PCIe lanes (32+8), then SR has 20 (16+4). If Naples has quad-channel memory, SR has dual. It's clever, I think.
I don't think the lack of PCIe lanes will be a big deal, as most gamers have moved to single card setups. Motherboard manufacturers can add SKUs with more dedicated lanes for Crossfire/SLI (2 x16, rather than 2 x8) via PLX chips for VR and others wanting it.
I would LOVE to see a cut Naples die (8c/16t or 6c/12t) with a high-end iGPU and HBM. But that's just fantasy for now.
Yes that's what I think. Obviously if they only undercut Intel's Kaby Lake prices by 10% to 15% all Intel will do is bump down Skylake and Broadwell prices.
I don't expect AMD to ship "no SMT" (or at least in a condition you can't easily turn it on). I can't think of many cases that AMD disables working parts, at least not on CPUs (and I doubt on GPUs once nvidia pulled conclusively ahead).
Intel can pull those stunts as long as they are the clear leader. AMD usually ships all the value they can. I know I ran a Semperon32 in 64 bit at one point (didn't "officially" have it, but you didn't even need to enable it in bios. Just feed it some 64 bit code and it ran). Locked cores on Phenon allowed unlock (you might have had to pay a buck for the black edition, but it wasn't much and probably harder to find non-black than black) even if they didn't always work.
I saw this already yesterday in the forums. Has anyone actually thought that this text snippet was sent to GDC organizers maybe already weeks or month ago (such conferences are obviously planned way ahead) and since then the actual release schedule has changed due to some issue(s)?
Sounds believable. Just remember that AMD is in a position where management might drag the thing to market well before engineering is ready to sign off.
If Ryzen demonstrates the IPC gain they are talking about, the great thing will be that Intel will have to put significant RnD $ into architecture development. They have been too comfortable with the core architecture for too long.
Like, what? There was an interesting bit with TSX instructions, but that was pretty much only interesting to database guys. The Knights Landing "tons of cores on a chip" are also pretty cool, but in practice I suspect that most code parallizable code can be convinced to run on a GPU much better.
if anyone fromn anand is is listening... can you do an indepth test if HYPER THREADING is worth it with the latest CPU´s?
i did my own limitedt test and i don´t see any benefits in real world application i use (mostly video apps, premiere, after effects and lightroom, photoshop).
Has anyone noticed that they pulled the pertinent couple of words from the sentence over at the GDC website? It now says, "Join AMD Game Engineering team members for an introduction to the AMD Ryzen CPU followed by advanced optimization topics." I guess they caught their mistake. ;)
I am really excited about how Ryzen will shake things up. I agree that Intel has been needing to kick CPU development into gear and that AMD needs this to survive. We all need actually. We have already seen what happens when there is no strong competition. I can appreciate Intel's improvements to their iGPUs, especially for laptop owners, and I can't honestly say I personally need more CPU power than I already have (3210M in a 13" 2012 MBP) but I know that some people do. (my brother's brother-in-law does his own youtube videos and has several insane editing rigs).
If this is the competition it takes to get Intel back into the big R&D it takes to come up with something completely new and amazing then I say bring it on. That said, even once Intel has come out with the next big thing to try and jump ahead again I expect I'll still be looking to AMD anyway since they have much better iGPUs, and they are promoting HSA. I'd consider taking a drop in CPU power for a better iGPU and HSA support from AMD, since HSA has the potential to absolute destroy CPU only performance, once us programmers can get our heads around how to optimize for it...
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
72 Comments
Back to Article
wiineeth - Wednesday, January 11, 2017 - link
I hope Anandtech publishes long review on ryzen. im a video editor and im pretty stoked about this if the 8c/16t version is affordable, would make my video editing workflow a breeze. Also i hope ryan and ian does youtube videos for normal people!bill.rookard - Wednesday, January 11, 2017 - link
Agreed. I'm more interested in the hex-core version as that would probably be a bit more of the performance/price sweet spot. And to just highlight this (again): pricing will be crucial. I get that they need to make money, I get that they can't give it away for nothing, I get that if it's close to Intel performance they can charge close to Intel pricing, but they NEED market share.Krysto - Wednesday, January 11, 2017 - link
Yeah, I would say AMD needs to focus on increasing market share so that people start seeing AMD devices around and consider them viable. Plus being 10% behind Intel in performance but offering a 10% discount on price so they offer the same value just won't be enough, because most people would still pick Intel in that scenario, if the value is the same. So their offers need to look very appealing in terms of value.xype - Wednesday, January 11, 2017 - link
They need profits, too. I find it silly to demand AMD to catch up with Intel and once it looks like the might, people don’t want to pay for it. Sure, it has to be good value for money, but it’s starting to sound like everyone will be dissapointed if AMD doesn’t make those Ryzen super cheap.If the CPUs perform well, people should pay for it.
BikeDude - Thursday, February 16, 2017 - link
If VW wanted to, they could probably launch a car tomorrow that would have the look and feel of a BMW, but they would struggle if they expected to be paid the same amount of money. Heck, if somebody simply replaced the BMW badge with VW badge, the value of the car would instantly be reduced. Or vice-versa: BMW could put their own badge on a VW, and instantly increase its value.The badge carries value in itself, and AMD is simply not where Intel is in terms of name recognition.
Humans are dumb. Which explains why the advertising business is as big as it is.
Mr Perfect - Wednesday, January 11, 2017 - link
If the product is good enough, then they don't have to be cheap. I can remember back when the Athlon 64 X2s hit, they where fast chips and had some serious prices on them. It was north of $500 for the X2 chip I purchased and was worth it compared to the P4s of the time.Considering Intel is currently charging $1000 for their 8core 16thread chips, even $500 for a 8c16t Ryzen would be a complete steal!
extide - Thursday, January 12, 2017 - link
Yeah, my guess is you will see something like:$500 -- 8 - core
$300-350 -- 6 - core
$200-250 and below -- 4 - core
Probably wont see dual core -- maybe as mobile APU's but I doubt it in the non IGP chips. I also bet AMD will leave hyper threading on all but they very lowest end or possibly not even disable it on anything.
just4U - Saturday, January 14, 2017 - link
One thing to remember is.. Intel can charge what it wants with no competition. Don't expect Intel pricing from AMD in the short-term even if they have a good chip..redwarrior - Wednesday, January 11, 2017 - link
I had previously estimated any time between February 21 and February 28. The silicon gas been finalized and already is being manufactured. It takes 6 weeks to fill the channel. It will launch before February 27th.ddriver - Wednesday, January 11, 2017 - link
Naturally, no self-respecting professional would line up to be an early adopter of any tech. I myself would gladly get a much needed break from intel chips, providing performance is good and price is competitive, but not before the retail product and platform go through at least 6 months of testing.profquatermass - Wednesday, January 11, 2017 - link
Or as I say: "Never buy version one of anything!"Alexvrb - Thursday, January 12, 2017 - link
This is a processor, not the first model year of an "all-new" or massively revised car model. By the time the first retail chips are in consumer hands, they usually have a pretty good handle on reliability. If there's issues, they can be corrected by microcode. If you're concerned there's going to be a major bug that when fixed affects performance, just wait a couple of months for the community to vet the chip. There's no need to entirely skip the first revision on every product, if that revision turns out to be good.ddriver - Thursday, January 12, 2017 - link
AMD could not possibly test the chip under every real world situation it could call under. They don't run people's apps, they don't crunch people's data. Specific flaws will always slip through factory testing. Phenom's TLB bug, skylake's fft bug, the original pentium's fdiv bug and several others have all slipped through testing.ddriver - Thursday, January 12, 2017 - link
Also the fact that it is a processor and not a car makes it significantly more prone to bugs. Cars are primitively simple compared to cpus. There are billions of things in a cpu that can fail, whereas a car has only few thousand components.DM0407 - Tuesday, January 31, 2017 - link
AMD TLB bug ring a bell?http://www.anandtech.com/show/2477/2
Alexvrb - Thursday, January 12, 2017 - link
I've rarely been burnt, and it was never anything I couldn't work around. Early adopters are often those who want to tinker with the latest tech and know the risks. If you know what you're getting into, more power to you. With that being said I don't need the latest and greatest anymore, so I probably won't be looking too deeply at Ryzen for a while. But those who leap in headfirst are just what you call "enthusiasts", many of whom are professionals and new tech is their hobby. Enthusiasts are a major staple of sites like these. Otherwise they'd change the site name to "Stabletech" and only review something after it's been on the market for a few years (OK OK phone reviews do NOT count, strictly talking PC components here).Targon - Sunday, January 15, 2017 - link
Are you serious? Professionals will pick up promising tech EARLY to properly evaluate the products. Professionals will not suggest tech before they can properly evaluate the products. It is called the bleeding edge of technology, and WE are expected to know about this stuff before the general public.bananaforscale - Wednesday, January 11, 2017 - link
That would fit that "recently launched" in GDC slides. All I know I want it yesterday.profquatermass - Wednesday, January 11, 2017 - link
You really need to look at the top-end graphic cards for reducing rendering times on frames and M.2 SSD storage for fast retrieval of the files.Shadowmaster625 - Wednesday, January 11, 2017 - link
Normal people can buy perfectly reasonable 12 core Intel workstations off ebay for $600. They aren't the latest architecture but the encoding FPS per dollar cant be matched.pencea - Wednesday, January 11, 2017 - link
Let's hope they won't be late and publish the review right on time.zodiacfml - Thursday, January 12, 2017 - link
I agree to this. This will be the value of Ryzen.lilmoe - Wednesday, January 11, 2017 - link
I'm super excited about this announcement, and hopeful that AMD won't mess this up. I really want a performance Ryzen APU/SoC in my next laptop.Chaser - Wednesday, January 11, 2017 - link
I'm mostly a gamer. As long as it's competitive the differences in CPU performance will be negligible. So price is probably the most important factor for me at this stage.Replikant - Wednesday, January 11, 2017 - link
One further thing to consider: Even if Ryzen is at the same price/performance, we should buy it. Whenever possible support the underdog. We've all seen the crap Intel pulls when there is no competition.BrokenCrayons - Wednesday, January 11, 2017 - link
I'd really like to support AMD to at least help maintain some amount of competition in the market, but as Anandtech articles have already pointed out in the past, many AMD processor-equipped laptops are stuck on single channel memory even though the CPU can handle dual channel. It's not a big deal until the system has to turn to RAM to support the graphics card. That really kills the whole AMD thing for me...well there was that and the anemic performance of the Athlon x4 860K I purchased last year that was at best, only a little faster than the Q6600 it replaced.Maybe Ryzen can turn things around. I'm hopeful for that and for improved Linux support for AMD GPUs. I haven't use an AMD GPU in quite a few years due to their weak showing in Linux even though something like a C-70 or e-450 was definately on my to-buy list back when I was netbook shopping. Ultimately, I ended up with an Atom n450 because its GMA 3150 had better driver support at the time.
TheJian - Wednesday, January 11, 2017 - link
They could easily get me if they make it FULLY win7 compatible also. I can't see buying either side if it forces me to win10, telemetry crap, Desktop trying to act like a mobile device etc etc (jeez list of hate I have here is long). I'm almost certain many forms of *nix will have full support so maybe they'll get me that way and my current desktop could be my last WINTEL PC finally. Bring on the Vulkan games so I can finally make a move without the need for Dx or Wintel.When you write the review, please include a discussion of what is MISSING if you choose to install on something OTHER than windows 10. I'm guessing that will be very important to MANY people.
Ming3r - Wednesday, January 11, 2017 - link
Do you want them to go over this for Kaby Lake as well? Or a full article on how Microsoft is requiring it?lmcd - Thursday, January 12, 2017 - link
Probably not important to many people, as the subsection of people whose tinfoil hats cover internet browsing but not Windows 10 usage are quite small. I'm surprised and disappointed your Anandtech comment section part hasn't burned through yet from all of those dangerous transmissions.prtskg - Friday, January 13, 2017 - link
You can check sites like Phoronix.com which are linux oriented for zen's performance in linux.helvete - Wednesday, February 8, 2017 - link
Why shouldn't it be compatible with w7?Also SW has to make use of HW not vice versa.
Achaios - Wednesday, January 11, 2017 - link
Thus launch is completely irrelevant for gamers. All we care about is single-threaded performance, and pretty much every gamer today is all set for many, many years to come with Intel CPU's.The only thing that is of importance to gamers today is GPU's, and VEGA will come in late June and the 1080TI in Q3 or Q4 2017.
So yeah, *yawn*.
Nagorak - Wednesday, January 11, 2017 - link
Plenty of gamers are still running on Sandy Bridge or Ivy Bridge because there has been little reason to upgrade. More cores could be a reason, provided performance is also improved.benedict - Wednesday, January 11, 2017 - link
Thank you wise gamer for speaking on behalf of all gamers. Maybe I should stop consider myself one because I'm not set with an Intel CPU and I'd love to get an 8 core beast in my pc, sacrificing a bit of single-threaded performance.SquarePeg - Wednesday, January 11, 2017 - link
This launch is completely relevant for gamers. The vast majority of gaming builds fall in the $1200 or less catagory. With Intel giving it's customers nearly nothing in terms of IPC since Sandy Bridge there are a large amount of people out there still holding on to Sandy Bridge and Haswell i5's and i7's. So if you can pick up a quad core Ryzen cpu, motherboard, $30 cooler for the same price as a Kaby Lake i7 alone and then overclock it to near equal performance it's a no brainer.With more and more games being able to efficiently utilize more than 4 large threads a Hexacore Ryzen cpu will outperform a quad core i7 hands down. This will only increase in AMD's favor as DX12 and Vulken become the standard for new games.
The most important thing about Ryzen to people who don't want/need a new AMD cpu is that it will pull down Intel's ridiculous pricing. This is very relevant to all considering a new PC purchase.
You only feel that it's irrelevant because it may not apply to personally and you've come to accept the dribble's of nothing that Intel calls a new generation of cpu "performance" as good enough.
xype - Wednesday, January 11, 2017 - link
Geez, stop talking sense on a hardware forum, will ya?! Get the pitchforks and let’s shit on something! Whee!bananaforscale - Wednesday, January 11, 2017 - link
Watch out for flying pigs!lmcd - Thursday, January 12, 2017 - link
It's a little bit awkward to note, but Intel's actually been churning out decent-sized performance gains for the past 5 years. Jumping from "worse than our Phenom X6" to "Came out this decade" performance doesn't deserve this level of loyalty, and Intel's been delivering consistent product improvements that are easily worth the upgrades. As an owner of both a Sandy Bridge 3960X and a Skylake 6770HQ, the latter is significantly faster than the former at everything except the longest prolonged workloads despite running at less than half the power consumption.It's pretty amusing that you're calculating a quad-core Ryzen, motherboard, and cooler at i7 pricing. Given that a 3-year-old quad core (8350) and its top-end motherboard barely squeak under that number, there's no way a new quad core and its brand-new top-end motherboard squeak under. And don't think you're using much less, as the lower-end motherboard chipsets aren't going to get the type of overclocking you're dreaming of.
Now, let's talk about games. And how DX12 ironically has the opposite effect of what you're talking about. DX12 was in AMD's favor because it reduces the CPU load and makes more events non-blocking. (For reference, it's *Vulkan.) As more things become non-blocking Intel's large threads become more relevant, not less. Things that would be restricted to 6-core machines are available to 4-core machines, as the 4 cores can always be fed instead of sometimes running out of Out-of-Order Execution depth with blocked calls (this already is rare, as, again, unless you're using a Bulldozer or Piledriver, it's pretty hard to be CPU-bound).
A 6-core Ryzen literally cannot cost less or even with an Intel 4-core. Intel's manufacturing process is cheaper, more mature, more dense even at the same node (both TMSC and GloFo are stat-padding), at a smaller node, and in-house. Sorry to break your bubble.
Intel's pricing is hardly ridiculous. In fact, they've done admirable advancing technology in the effort to beat themselves. They were, and still are, their own best competitor.
The only thing you could really argue in AMD's favor is their IGPs, which will help them win their consolation prize in the low-end markets again. Sorry to break it to you.
rhysiam - Thursday, January 12, 2017 - link
Intel's markup is absolutely massive and their performance reign over the last (nearly) decade is a perfect example of why we need competition.You argue in one paragraph that Intel's process is both cheaper and more mature, and then in the very next line that their pricing is "hardly ridiculous". FYI, the $1700 10 core 5960X has approximately the same die size as an i5 2300 which sold (presumably still at a healthy profit) for $187 almost 6 years ago. If Intel really does have a "cheap" and "mature" process, they're certainly not passing on those benefits to their consumers.
Let's forget die size and try it by transistor count for a second...
Did you know also that the often derided as stupidly overpriced iPhone 7 starts at $650, and sports an A10 SOC with 3.4 billions transistors. That's slightly more transistors than a Broadwell-E CPU, which is $1700, $1000, ~$620 or ~$430 depending on the SKU, and the Intel CPU doesn't come with a premium phone.
We don't know the exact transistor counts of Skylake processors, but the $330 i7 6700K has an estimated 1.75B transistors. That's far, far fewer than the RX 460 (~3 Billion) which can easily be found with superior cooling solutions, power delivery and GDDR5 RAM(!) for under $100.
Of course these aren't accurate apple to apple comparisons and they're just rough estimates. But they offer an indication of just how much Intel charges for tiny chips with relatively small numbers of transistors.
Of course Intel is a company out to make money so they can charge whatever the market will pay.
BUT, it shows just how important competition is, which is precisely why so many people are excited, cautiously optimistic, or at least hopeful for a successful and competitive CPU from AMD.
Cooe - Sunday, February 28, 2021 - link
Hahahaha xD. So much stupid in such a small post. I guess the 6-core/12-thread Ryzen 5 1600 being SIGNIFICANTLY cheaper than Intel's 4-core/4-thread i5-7600K (let ALONE the 4c/8t i7-7700K), while absolutely bitch-slapping it up & down the block in terms of performance was all just a false memory I'd manufactured in my head. Thanks for letting me know that! ;)bananaforscale - Wednesday, January 11, 2017 - link
You don't know what you speak of. Ryzen has been shown to rival Intel in single thread performance *even handicapped*, and games are increasingly multithreaded. If the price is right it's relevant to this gamer (who has probably played computer games before you were born), and I'm upgrading regardless. I figure my Bulldozer deserves a break.lmcd - Thursday, January 12, 2017 - link
Do you mean that one random blender demo?If you bought a Bulldozer, isn't that basically an admission that you don't care about the numbers?
Vodietsche - Wednesday, January 11, 2017 - link
I'll be getting one of those ZEN/Vega-APUs for sure. This is gonna sweet! Might upgrade to a Zen CPU too, but still waiting for those benchmarks...CajunArson - Wednesday, January 11, 2017 - link
I think there is far too much being read into that little blurb.According to AMD, "RyZen" has already been "launched" at least in as much as they are showing off engineering samples.
People tend to forget that it takes usually 2 - 3 *months* to go from the start of real mass-production to when you can actually buy a real chip at retail.
If AMD was actually going to have these chips on store shelves in the next 6 weeks, they would not have been showing off "engineering samples" at CES. They would have been showing off 100% finished retail chips.
TemjinGold - Wednesday, January 11, 2017 - link
So in other words, according to you, there is ZERO chance AMD will meet their Q1 target date, right?CajunArson - Wednesday, January 11, 2017 - link
They can meet it using Corporate Speak that people around here don't truly understand.They can "ship for revenue" in Q1 meaning they book the payments from OEMs and retailers for the first batch of RyZen parts. They can even say that they've "shipped" them. That doesn't mean that you can buy one though.
This isn't some theoretical exercise, you can't throw a rock on Google without running into reviews of fully operational 7700K Kaby Lake retail processors that were posted last November. You still couldn't buy one until last week though. In this case, nobody.. NOBODY.. has seen a full retail sample of Zen or "RyZen" floating around anywhere. Do the math, the laws of economics apply to AMD just like they do to everybody else.
xype - Wednesday, January 11, 2017 - link
Initially I thought you’re being too harsh, but the point about Zen not leaking out is a valid one. Either security is super tight—or there aren’t any CPU production lines ramping up yet. In the latter case, the launch might be Q1 on paper, with widespread availability in Q2 or even Q3 (which is what people need to care about anyway).Also, GDC peeps, if interesting to AMD, get working machines easier than any of us.
bananaforscale - Wednesday, January 11, 2017 - link
ISTR that when AMD released Bulldozer they were available within two weeks. I suspect AMD's launch is different from Intel's. (Heck, they do a lot fewer of them for starters. :D)TemjinGold - Wednesday, January 11, 2017 - link
Or they can you know, actually launch for real in Q1 like they said they would.Ozymankos - Monday, January 16, 2017 - link
Well,this i how this reality in 3d works...production of any goods takes time and shipment takes time\
plus...the human quality of service is very bad while the technological side is very good
many hosting sites prefer to let me run into big minuses with my 8 websites rather than bother with answering my questions
because support made by intelligent creatures not robots...is very expensive and very stupid
bananaforscale - Wednesday, January 11, 2017 - link
They have been showing off engineering samples for months really. Also, Q1 ends on March 31st, and that's a bit over 11 weeks away. I figure they'll have no trouble getting to the market in time.milleron - Wednesday, January 11, 2017 - link
True that, but do you really disagree with Krysto's contention that buyers would still go with Intel if the AMD offering is 10% less efficient and 10% less expensive?jjj - Wednesday, January 11, 2017 - link
It would be utterly idiotic to launch it during big events. The point is to get more attention not less.My bet is Feb 16.
16 threads, they had 16 mobos when announcing the AM4 the other week,maybe they got a thing for 16.
JasonMZW20 - Wednesday, January 11, 2017 - link
Based on binning rates and such, I speculate this:8c/16t ($699-899+), very low numbers/least availability, low yields.
6c/12t ($399-499, $599 if 4GHz+), slightly defective 8c/16t bins
Mainstream:
4c/8t ($250-299) highest yields, most availability
4c/4t ($150-199) same, just no SMT
Clocks will determine pricing within each CPU group. Highest clocks, highest price. If you were expecting 8c/16t for Intel's 4c/8t prices, what were you smoking?
Since they're all multiplier unlocked, they compete with Intel's K line of Kaby Lake chips.
I look forward to the Zen APUs in 2H 2017.
SquarePeg - Wednesday, January 11, 2017 - link
I think your pricing estimates are a bit too high. I see AMD slotting their CPU pricing as this:4c/8t = i5 Kaby Lake = $179-239 retail price
6c/8t = i7 Kaby Lake = $279-349 retail price
8c/16t = $379 - $499, maybe more for a very limited quantity, highly binned, Super FX Black VIP Edition...or whatever they call it.
lmcd - Thursday, January 12, 2017 - link
That's adorable. You really think they're just going to charge $30 more than their current 4c/8t 8350 for a 50% node shrink and multiple years of engineering?If that's the case, they're going to miss all the benchmarks they're claiming they hit too.
PixyMisa - Thursday, January 12, 2017 - link
The leap from 32nm to 14nm means that the new chips will be much cheaper to produce. And the 8350, whatever its limitations, has always been marketed as an 8c/8t chip.Also, they need these chips to actually sell, and despite some decent benchmarks against Broadwell-E they will likely still have somewhat lower clocks and IPC than Kaby Lake.
They need to position the 4c/8t chip to undercut the i5, the 6c/12t to undercut the i7, and the 8c/16t to undercut the 6c/12t Broadwell-E.
That means 4c below $200, 6 core below $300, and 8 core below $400. The top-of-the-line 8c may be closer to $500.
I don't know what they will do, of course, but that's what they need to do.
JasonMZW20 - Thursday, January 12, 2017 - link
The FX-8350 was marketed as a true 8-core chip, but it only had 4 modules (2 int per module, 1 shared float). So, in reality, it's a 4c/8t chip.Your pricing estimates, and those of SquarePeg, are extremely optimistic. They may slightly undercut Intel, but with the performance Ryzen has been showing, it'll be priced accordingly.
8c/16t is a low-volume, ultra high-end flagship processor. That alone demands flagship pricing.
PixyMisa - Friday, January 13, 2017 - link
But it's simply not as high end a part as Broadwell-E. Doesn't have quad-channel RAM or 40 PCIe lanes for high-end workstations. And AMD have to rebuild sales from basically nothing.They have to undercut Intel at each price point, and that means AMD 4c/8t will compete on pricing with Intel's 4c/4t, AMD 6c/12t will compete with Intel 4c/8t, and AMD 8c/16t will compete with the low-end Broadwell-E, which is a ~$400 part.
They price their video cards to compete; don't know why you expect them to commit suicide on their new CPUs.
Also don't know why you think yields will be so bad that most 8c dies will end up sold as 4c parts. They've been shipping 14nm GPUs in volume for months now.
TemjinGold - Friday, January 13, 2017 - link
Keep telling yourself that. If it beats Broadwell-E consistently, I don't care of it relies on a toaster and a cardboard box for its tech.JasonMZW20 - Sunday, January 15, 2017 - link
I think AMD scaled Ryzen in such a way that 8c/16t parts will be cheaper than Intel's offerings because of missing on-die features you mention relative to Intel. That's why I put it under $1000 bracket. It can be anywhere from $500-799, while Naples will be over $1000.So, if 8c/16t Summit Ridge is 1/2 scale of 16c/32t Naples server part, then it should have 1/2 PCIe lanes and memory controllers. That means the 4c/8t versions share the same die if they retain 20 PCIe lanes and dual-channel memory, but the other cores were either defective bins or intentionally cut. That's getting really speculative though. It makes sense though, as Intel stuffs an iGPU in the Skylake and Kaby Lake parts to fill the die.
So, if Naples has 40 PCIe lanes (32+8), then SR has 20 (16+4). If Naples has quad-channel memory, SR has dual. It's clever, I think.
I don't think the lack of PCIe lanes will be a big deal, as most gamers have moved to single card setups. Motherboard manufacturers can add SKUs with more dedicated lanes for Crossfire/SLI (2 x16, rather than 2 x8) via PLX chips for VR and others wanting it.
I would LOVE to see a cut Naples die (8c/16t or 6c/12t) with a high-end iGPU and HBM. But that's just fantasy for now.
SquarePeg - Friday, January 13, 2017 - link
Yes that's what I think. Obviously if they only undercut Intel's Kaby Lake prices by 10% to 15% all Intel will do is bump down Skylake and Broadwell prices.Cooe - Sunday, February 28, 2021 - link
Hahaha. That's exactly what they did dumbass.wumpus - Friday, January 13, 2017 - link
I don't expect AMD to ship "no SMT" (or at least in a condition you can't easily turn it on). I can't think of many cases that AMD disables working parts, at least not on CPUs (and I doubt on GPUs once nvidia pulled conclusively ahead).Intel can pull those stunts as long as they are the clear leader. AMD usually ships all the value they can. I know I ran a Semperon32 in 64 bit at one point (didn't "officially" have it, but you didn't even need to enable it in bios. Just feed it some 64 bit code and it ran). Locked cores on Phenon allowed unlock (you might have had to pay a buck for the black edition, but it wasn't much and probably harder to find non-black than black) even if they didn't always work.
Shadowmaster625 - Wednesday, January 11, 2017 - link
The chip is 6 weeks from release and there isnt one single passmark baseline in the database? Yeah, pull the other one please.beginner99 - Thursday, January 12, 2017 - link
I saw this already yesterday in the forums. Has anyone actually thought that this text snippet was sent to GDC organizers maybe already weeks or month ago (such conferences are obviously planned way ahead) and since then the actual release schedule has changed due to some issue(s)?TemjinGold - Friday, January 13, 2017 - link
If anything, their CES comments seem to indicate Ryzen has been pulled earlier, rather than pushed later.wumpus - Friday, January 13, 2017 - link
Sounds believable. Just remember that AMD is in a position where management might drag the thing to market well before engineering is ready to sign off.doggface - Thursday, January 12, 2017 - link
If Ryzen demonstrates the IPC gain they are talking about, the great thing will be that Intel will have to put significant RnD $ into architecture development. They have been too comfortable with the core architecture for too long.It needs more than just media feature additions.
wumpus - Friday, January 13, 2017 - link
Like, what? There was an interesting bit with TSX instructions, but that was pretty much only interesting to database guys. The Knights Landing "tons of cores on a chip" are also pretty cool, but in practice I suspect that most code parallizable code can be convinced to run on a GPU much better.Gothmoth - Tuesday, January 17, 2017 - link
if anyone fromn anand is is listening... can you do an indepth test if HYPER THREADING is worth it with the latest CPU´s?i did my own limitedt test and i don´t see any benefits in real world application i use (mostly video apps, premiere, after effects and lightroom, photoshop).
rtharston - Monday, January 23, 2017 - link
Has anyone noticed that they pulled the pertinent couple of words from the sentence over at the GDC website? It now says, "Join AMD Game Engineering team members for an introduction to the AMD Ryzen CPU followed by advanced optimization topics."I guess they caught their mistake. ;)
I am really excited about how Ryzen will shake things up. I agree that Intel has been needing to kick CPU development into gear and that AMD needs this to survive. We all need actually. We have already seen what happens when there is no strong competition. I can appreciate Intel's improvements to their iGPUs, especially for laptop owners, and I can't honestly say I personally need more CPU power than I already have (3210M in a 13" 2012 MBP) but I know that some people do. (my brother's brother-in-law does his own youtube videos and has several insane editing rigs).
If this is the competition it takes to get Intel back into the big R&D it takes to come up with something completely new and amazing then I say bring it on. That said, even once Intel has come out with the next big thing to try and jump ahead again I expect I'll still be looking to AMD anyway since they have much better iGPUs, and they are promoting HSA. I'd consider taking a drop in CPU power for a better iGPU and HSA support from AMD, since HSA has the potential to absolute destroy CPU only performance, once us programmers can get our heads around how to optimize for it...
http://schedule.gdconf.com/session/optimizing-for-...