Monday, January 8th 2018

AMD Reveals CPU, Graphics 2018-2020 Roadmap at CES

AMD at CES shed some light on its 2018 roadmap, while taking the opportunity to further shed some light on its graphics and CPU projects up to 2020. Part of their 2018 roadmap was the company's already announced, across the board price-cuts for their first generation Ryzen processors. This move aims to increase competitiveness of its CPU offerings against rival Intel - thus taking advantage of the blue giant's currently weakened position due to the exploit saga we've been covering. This move should also enable inventory clearings of first-gen Ryzen processors - soon to be supplanted by the new Zen+ 12 nm offerings, which are expected to receive a 10% boost to power efficiency from the process shrink alone, while also including some specific improvements in optimizing their performance per watt profile. These are further bound to see their market introduction in March, and are already in the process of sampling.

On the CPU side, AMD's 2018 roadmap further points towards a Threadripper and Ryzen Pro refresh in the 2H 2018, likely in the same vein as their consumer CPUs that we just talked about. On the graphics side of their 2018 roadmap, AMD focused user's attention in the introduction of premium Vega offerings in the mobile space (with HBM2 memory integration on interposer, as well), which should enable the company to compete against NVIDIA in the discrete graphics space for mobile computers. Another very interesting tidbit announced by AMD is that they would be skipping the 12 nm process for their graphics products entirely; the company announced that it will begin sampling of 7 nm Vega products to its partners, but only on the Instinct product line of machine learning accelerators. We consumers will likely have to wait a little while longer until we see some 7 nm graphics cards from AMD.
Jumping to AMD's 2020 plans, there are some very interesting developments. At its Tech Day presentation at CES, AMD has confirmed that their Zen 2 design on the 7 nm process has been already completed. The lack of Zen 2 mentions in the 2018 roadmap means this particular piece of silicon will only be launched after 2019, but with the company having already announced that the base processor design is complete - with some deeper changes than those introduced by Zen+ - means AMD will have some time now to further tune and update its final design according to the 7 nm process' characteristics. AMD has also announced that their next update to the Zen architecture, Zen 3 on the 7 nm+ process, is on-track for 2020 (likely 4Q, or beyond), which at this point, likely means the company has planned out how and where they want to go with further improvements to the Zen architecture.
On the graphics side, the 2020 roadmap includes the aforementioned cut of the 12 nm process from AMD's graphics architecture. Instead, the company will jump from the 14 nm process currently used on its Vega dies straight to the 7 nm process, beginning with the lower volume, higher margin Instinct accelerator line. Considering that these are expected to be sampled in late 2018, it's likely AMD will introduce it to market in 2019 already. The shrinkage of Vega to 7 nm is likely a way for AMD to learn the ropes of developing for that process, without also having to design a new graphics architecture at the same time, thus saving them some valuable engineering resources that can be fully deployed in the shrinking process, and then on the design/redesign of some Navi features for the 7 nm manufacturing process.
After that, the roadmap only mentions the introduction of Navi (if it follows the other 2020 roadmaps by AMD, this also points towards a 2019 introduction), and an (expected) late 2020 next-gen, 7 nm+ post-Navi graphics architecture.
Source: AnandTech
Add your own comment

41 Comments on AMD Reveals CPU, Graphics 2018-2020 Roadmap at CES

#26
dicktracy
Officials are saying 10% performance bump at best case scenario. Meh just wait for Zen 2 and Icelake.-
Posted on Reply
#27
Casecutter
So not even a re-vamp of Polaris or "more-or less" a re-version with some Vega tweaks and then on the 12nm process? Which to me is nothing more than 14nm with a little re-spin? Then there's the whole GloFo has only so much capacity (tools, machines a clean areas) to produce whatever fab, so even if they hold to Polaris/Vega 14nm for 2018 they still appear to be "well off the mark" that Tahiti/Hawaii was delivered. They've got to find a way to prop-up mainstream and Vega or it's going to be another tough year for the graphics side.
Posted on Reply
#28
TheoneandonlyMrK
CasecutterSo not even a re-vamp of Polaris or "more-or less" a re-version with some Vega tweaks and then on the 12nm process? Which to me is nothing more than 14nm with a little re-spin? Then there's the whole GloFo has only so much capacity (tools, machines a clean areas) to produce whatever fab, so even if they hold to Polaris/Vega 14nm for 2018 they still appear to be "well off the mark" that Tahiti/Hawaii was delivered. They've got to find a way to prop-up mainstream and Vega or it's going to be another tough year for the graphics side.
They're being vague enough with Vega that you may get a mid range card with /2560 shaders , id say they're waiting on a few balls still in the air their (ddr6) , but regardless given the crypto gpu craze they are not going to be too worried about selling enough GPUs ,for them it's got to be the right balance of output v demand in gpu asic production, because they can't afford to be holding inventory whereas I think others play it a bit different.
All pretty much as i expected or knew tbh bar no Vega on 12(14++?)nm+ which is a surprise to me, but to me indicates they are using up their wafers elsewhere ,ie the Ryzen lines output is getting upped(no wafer's spare for GPU) devs.


Did anyone else note more than one intel/amd gpu Combo chip mentioned, Kaby G with Gh gpu for their Vr Nuc but a higher spec H gpu part got listed but no info other then a placeholder , can't remember where i saw that little list ,probs wccf.

Anyway I am sure AMD will be making a few more variants of Vega available, Vega for laptops is unlikely to be the Vega 64 after all so they still have mobile descrete Vega , small Form factor Vega and mid to bottom range Vega to release imho and they definitely have designs that could fit these applications.
Posted on Reply
#29
evernessince
Vayra86If you consider trailing the top end by 35-40% performance is 'doing just fine', I'm not sure what to tell you. These ain't CPUs where the competitor decides to relax for a decade because its design is superior; Nvidia has already shown that and the latest release (GV100) sits lonely at the top, with a price tag of 3K and nothing exists that can even remotely touch its complexity/die size/processing power.

Shrinking Vega won't close that gap at all, its quite worrying.
I don't think branging about how much Nvidia is bum fucking you for cash is a good strategy.
Posted on Reply
#31
Vayra86
evernessinceI don't think branging about how much Nvidia is bum fucking you for cash is a good strategy.
Do you really consider me or this forum that influential in the business, that this is 'strategy'? I'm just calling things as I see them... It would be great if AMD did a better job, however for the past two years not a single release has shown that they are capable of it. Its all too little, too late.
Posted on Reply
#32
TheoneandonlyMrK
Vayra86Do you really consider me or this forum that influential in the business, that this is 'strategy'? I'm just calling things as I see them... It would be great if AMD did a better job, however for the past two years not a single release has shown that they are capable of it. Its all too little, too late.
I think you and others should then relent on your opinion since Amd are not a football team nor Nvidia, they are not really after some mythical top card trophy, they are trying to ship units and are doing so successfully day in day out regardless of who some think hold the Title.
So where do you think trailing the top 3k card by 30-40% puts Amd, because they're still selling gpus fast enough no one can apparently buy one easily for the budget price they seem to expect.
Posted on Reply
#33
Vayra86
theoneandonlymrkI think you and others should then relent on your opinion since Amd are not a football team nor Nvidia, they are not really after some mythical top card trophy, they are trying to ship units and are doing so successfully day in day out regardless of who some think hold the Title.
So where do you think trailing the top 3k card by 30-40% puts Amd, because they're still selling gpus fast enough no one can apparently buy one easily for the budget price they seem to expect.
So far the market share says otherwise. Moving units has never been AMDs problem, making profit has.
Posted on Reply
#34
deu
AquinusFind me a single Vega 56 for MSRP and we'll talk. I can literally go almost anywhere and buy a 1070 right now. I can't just go buy a Vega 56 because they're literally nowhere to be found and the ones you do find have an insanely inflated price.
That should tell you that there is higher demand for the vega 56 relative to its production quantities. Its like saying: I can litteraly go buy a VW up RIGHT now, but the a Bugatti Veyron i cant get right now! Performance is i relevant in this case. Arguebly a 1070 is easier to come by and the price of a vega 56 is too high due to demand. But at its original price (seen in denmark for 1-2 months) the card holds its ground compared to the 1070. But we all know that there a thing called miners that what to trap all the vega 56 card in a big pile and have them burn for fictive ore!
Posted on Reply
#35
Vayra86
deuThat should tell you that there is higher demand for the vega 56 relative to its production quantities. Its like saying: I can litteraly go buy a VW up RIGHT now, but the a Bugatti Veyron i cant get right now! Performance is i relevant in this case. Arguebly a 1070 is easier to come by and the price of a vega 56 is too high due to demand. But at its original price (seen in denmark for 1-2 months) the card holds its ground compared to the 1070. But we all know that there a thing called miners that what to trap all the vega 56 card in a big pile and have them burn for fictive ore!
Vega with HBM as a miner card? Hardly the most effective choice

The reason you can't find Vega in stock is because the stock never arrives. Don't fool yourself :)
Posted on Reply
#36
deu
Vayra86Vega with HBM as a miner card? Hardly the most effective choice

The reason you can't find Vega in stock is because the stock never arrives. Don't fool yourself :)
Well they did in Denmark. But the point is that the demand for vega 56 is there: otherwise the price would not rise.
Posted on Reply
#37
Casecutter
Do we think that there's some amount of AMD Vega 64 production still being channeled by AMD (only) to fulfill HPC, Professional contracts? While basically Vega 56 (geldings) are going to AIB's.

Then I don't understand why folks think HBM2 is a reason miners won't go for Vega? Sure they don't see any boon for such a memory being use on the card, but it's all about the hash, to power (at least those paying for their power) and how soon they see a pay-off (ROI). Those working no power cost (college students in a dorm) just want to make the hash-fast and hope not to get caught. Heck these guy are mobile and move from dorm to dorm paying friends for the inconvenience of the noise and heat. Colleges are now walking the halls with FLIR cameras checking. As long they can make money/sense they'll want them. I talked with a guy the other day who is contacting Mom & Pop stores looking for any AMD cards they come across that just Artifact, and getting them for mining. Heck an old 280X even if Artifacting during gaming, still might have value as a "shovel"... Cool!
Posted on Reply
#38
Aquinus
Resident Wat-man
If only AMD could have done a die shrink of the 390x...
deuthe price of a vega 56 is too high due to demand.
For people reselling it. You literally can't find it anywhere.
deuI can litteraly go buy a VW up RIGHT now, but the a Bugatti Veyron i cant get right now!
The difference in cost is a whole lot more than it is for GPUs. It's more accurate to say, you could buy a VW right now, but you can also buy a Subaru, or a BMW, or a Ford, or a (enter name here.) As for Bugatti, if you're even considering it, you'd be the kind of person to be complaining that you had to get a Lambo, Mazarati, or Porsche instead.
deuprice of a vega 56 is too high due to demand.
...or is it because demand is normal but, yields are crap?
deuBut we all know that there a thing called miners that what to trap all the vega 56 card in a big pile and have them burn for fictive ore!
Miners are going to grab whatever they can at this point. They're not just sitting around hoping that Vega 56 will come around. In the meantime, I've been seeing GPUs popping up in 6-packs with the sole intent to mine. That alone should tell you that Miners aren't the reason because miners want the best bang for your buck when it comes to compute and they also want to own multiple of them to mine more at once. All in all Vega's failure wasn't because of miners, it was because there are literally none of them. A die shrink of Hawaii probably would have costed less and netted more money for AMD which is what's insane and the only reason they probably didn't do it is so it wouldn't cut into Vega sales.

If there was a 590X, I bet it would sell like hot cakes.
Posted on Reply
#39
Relayer
Vayra86If you consider trailing the top end by 35-40% performance is 'doing just fine', I'm not sure what to tell you. These ain't CPUs where the competitor decides to relax for a decade because its design is superior; Nvidia has already shown that and the latest release (GV100) sits lonely at the top, with a price tag of 3K and nothing exists that can even remotely touch its complexity/die size/processing power.

Shrinking Vega won't close that gap at all, its quite worrying.
Why wait for AMD? Buy the big Volta. It seems to be awesome. :D
Posted on Reply
#40
eidairaman1
The Exiled Airman
RelayerWhy wait for AMD? Buy the big Volta. It seems to be awesome. :D
Because we have choices
Posted on Reply
#41
r9
Vayra86Wow, they are banking on 7nm to save GCN. First they banked on HBM1 which failed, then HBM2 which failed even more miserably, now they bank on heavily contested fab capacity on the smallest node.

o_Oo_Oo_O

Let's just hope their CPU keeps going as it does, because they'll need it. Desperately.
ATi always gambled with new process nodes ahead on Nvidia. They banked more times than they haven't in the past. Just bad luck in the roll of the dice.
Posted on Reply
Add your own comment
Apr 23rd, 2024 06:08 EDT change timezone

New Forum Posts

Popular Reviews

Controversial News Posts