[wccftech]Nvidia Volta Allegedly Launching In 2017 On 12nm FinFET Technology

[JoeRambo]

Given how Nvidia executed with Pascal, deploying nearly full stack in several month last year, one needs to take such rumors seriously. Some guys (you know who you are) were claiming "slow" roll out for Pascal, so here we go again.

NV could very well unleash Volta this year, they had plenty of time for architecture ( Since Pascal and Maxwell are similar), they have experience on 16nm FF and this "12nm" is probably in good shape since it is almost the same from process side of things. We could be looking at repeat of 28nm Kepler->Maxwell, but on 14nm.

Enterprise Volta stuff entry to general will be dictated by super computer contracts as they will need thousands of chips and consumer compute cards will come once that is sorted out. It could happen in 2017, no doubt about it, no tech blockers.

Consumer graphic Volta is in no rush currently, since AMD is missing from a fight, if they show up with "Fury2", expect NV to respond preemptively with GP102 consumer variant and releasing GV104 sooner than later. Can't predict product cadence since it is not driven by tech schedules, but rather by arbitrary marketing and management decisions - we got so low, that vendors are showing "unnanounced NV cards" at CES what is probably a quarter before release and they prolly sat on them for quite some time before that...

 

[moonbogg]

If Nvidia released anything faster than a TitanX this year, it would cost $2,500.

Threadcrapping is not allowed
Markfw
Anandtech Moderator

 

[crisium]

Don't worry, you'll get aftermarket 1080 Ti's later this year that will slightly edge out stock Titan XP for only ~$750-$800.

Somehow as I typed that, I realized that could be considered an optimistic view. I know some are expecting worse.

 

[Ajay]

Don't worry, you'll get aftermarket 1080 Ti's later this year that will slightly edge out stock Titan XP for only ~$750-$800.

Somehow as I typed that, I realized that could be considered an optimistic view. I know some are expecting worse.

Yeah, only if Vega bests the GTX 1080. Also, depending on where TSMC is on the yield curve, a bump in clocks could refresh the GP104/106 line.

 

[PhonakV30]

There is no news about 12nm FinFET even PDF files. Any info?

 

[tamz_msc]

There is no news about 12nm FinFET even PDF files. Any info?

Yup, TSMC's website shows 10nm->7nm->EUV. No mention of any 12nm node.
http://www.tsmc.com/english/dedicatedFoundry/technology/future_rd.htm

 

[RussianSensation]

I sincerely hope that is not the case. AMD is in good position on the CPU side to finally gain some market share back. It is definitely a missed opportunity for Vega to not launch along with Ryzen. If they could have managed their timelines better, they would have really made bank on total system update sales. Now, people who buy Ryzen and looking to build new computers are going to left wondering when Vega will release and will be tempted to pair it with an Nvidia GPU.

Ironically, your post only highlights why Vega is irrelevant for AMD in the consumer space. CPUs, APUs and pro-graphics, low-end and mid-range cards is where the $ is for AMD. Gamers haven't bought flagship AMD cards when AMD was clearly superior, so why would this change?

The presentation and architecture for Vega shows that between the lines Vega has many Pro-market architectural features for deep learning, neural networks, fast memory cache, but these features require specific game developers optimizations.

People forget that AMD is primarily a CPU/custom IP company, while NV is primarily a GPU company that's expanding into other markets like automotive and AI. Just 1 great year of profitable Ryzen sales could make AMD as much $ (profits) as 10 years of Radeon graphics sales. In fact, Radeon is a losing business. The only benefit from it is console design wins.

If you are running a CPU firm and your competitor has 99.2% market share in the enterprise/workstation space, a market worth tens of billions of dollars, where you are guaranteed to get more "easy" sales and profits, how much $/effort would you commit to another market segment where 4-5% of the entire PC gaming market makes purchases ($500-650 AMD flagship cards)? Of 4-5%, NV has 70-80% brand loyal customers. So Vega is going to sell to an irrelevant number of customers -- a total drop in the bucket for AMD. AMD is going to make more $ selling 4-core Ryzen CPUs in 2017 than probably the entire 2017 Radeon desktop graphics business.

AMD has been without a flagship card for almost a year and this has had 0 impact on their GPU market share or GPU profitability. AMD was 100% correct that for them high-end is irrelevant and their improved financial performance and GPU market share FYE 2016 back this up.

Face it, Vega is a blip on AMD's radar; and high-end AMD consumer cards are a tiny market and AMD knows it. They have to make Vega so they can target professional markets and have next-gen mid-range RX 600 cards in 2019-2020. They have to keep making next gen GCN or they'll never win PS5/XB2 consoles.

How does all of this matter for consumer Volta? NV knows AMD doesn't care much about high-end gaming graphics anymore, which is also why they keep milking 1070/1080/Titan XP and purposely keep delaying 1080Ti. It makes little sense for NV to release GV104 and cannibalize the amazing ~60%+ gross margins on Pascal. Not even in NV's wildest dreams did they think in 2016-2017 they would be selling a mid-range 680 successor for $600-700 with no 7970 competitor in sight!

If we go back to the Kepler generation, 680 launched March 2012 but it took until May 2013 to launch GTX780. 1080Ti will be like 780, a cut-down flagship chip. NV released 780Ti almost 2 years after GTX680. Therefore, NV can still release a cut-down 1080Ti in March-April and a full 1080Ti Black in November 2017 (is they wanted to). But I have a feeling there is no need for that since AMD won't be targeting Vega at the $700-1000 market segments. NV will likely finish off this Pascal generation in 2017 with a cut-down 1080Ti, priced at 780Ti level or higher. Then in 2018, they will roll-out GV104.

It's possible they may have Volta GV107 in the $99-179 space by late 2017 to flash out 12nm. If NV has Volta for 2017, it directly contradicts their consumer roadmap.
http://www.kitguru.net/components/g...dia-changes-roadmap-volta-is-now-due-in-2018/

If NV has consumer Volta for 2017, they would have been lying to investors and shareholders for the last 2 years. Doesn't sound legitimate.

 

[Phynaz]

There is no news about 12nm FinFET even PDF files. Any info?

Consider the source.

 

[tential]

AMD was only superior during niche times....
Nvidia has had the better flagship GPU when the MAJORITY of high end users actually buy. At launch. Having a great GPU when everyone has already purchased a GPU is utterly irrelevant. Hence why AMD doesn't do well in the high end. I'm NOT buying Fury X, 290x, etc. because ALL were terrible values at launch... and came out after Nvidia. No. If I want a high end GPU, I WANT the 1080ti. Want the Volta Flagship monster. THE LAST thing I want to do is wait for AMD to roll out their high end GPU. It may eventually do better, but you're being EXTREMELY misleading when you talk about AMD perf at the high end.

 

[Bacon1]

290x, etc. because ALL were terrible values at launch... and came out after Nvidia.

290x/290 released before the 780 Ti which was priced higher and worse price/perf ($700 vs $550)

I WANT the 1080ti. Want the Volta Flagship monster.

So you want the 1080 Ti which hasn't launched yet but won't consider Vega because it also hasn't launched yet?

 

[energee]

Lols, now foundries are rebranding process nodes. Too bad, 16FF++ would have been more geek friendly.

I had no idea it was possible to rebrand a unit of measurement. Now I'm even more confused: would an Enhanced 2TB SSD offer more or less storage space than a standard model?

 

[tviceman]

I had no idea it was possible to rebrand a unit of measurement. Now I'm even more confused: would an Enhanced 2TB SSD offer more or less storage space than a standard model?

Not more, just BETTERer... Duh!

 

[PhonakV30]

AMD was only superior during niche times....
Nvidia has had the better flagship GPU when the MAJORITY of high end users actually buy. At launch. Having a great GPU when everyone has already purchased a GPU is utterly irrelevant. Hence why AMD doesn't do well in the high end. I'm NOT buying Fury X, 290x, etc. because ALL were terrible values at launch... and came out after Nvidia. No. If I want a high end GPU, I WANT the 1080ti. Want the Volta Flagship monster. THE LAST thing I want to do is wait for AMD to roll out their high end GPU. It may eventually do better, but you're being EXTREMELY misleading when you talk about AMD perf at the high end.

So if Big Vega Is faster than Pascal titan or even Volta , Would You buy it ?

 

[Arachnotronic]

So if Big Vega Is faster than Pascal titan or even Volta , Would You buy it ?

We've already seen "big Vega" in action, doesn't look like it will be faster than the Pascal Titan, IMO.

 

[tential]

We've already seen "big Vega" in action, doesn't look like it will be faster than the Pascal Titan, IMO.

If you believe the 1080Ti is going to be the fastest GPU on the market, then you "aren't considering Vega".

Even though you've said numerous times that you have cash set aside for the purchase of 2 Vega GPUs... nope... not considering it. This hilarious thing is it's the OTHER way around. I'm not buying a 1080Ti because I have a freesync monitor and Nvidia won't give me a Gsync 4K monitor in the specs I want (looks like that's changing though!).

In reality, I should be CONSIDERING a 1080Ti GPU. Nvidia has consistently given a high end GPU user better perf during the time frame the majority of high end users care about, while also releasing their gpus earlier to give high end users the most time to enjoy the highest levels of performance possible.

I'm not saying anything groundbreaking. Nvidia makes record profits from selling higher end GPUs at high prices when there is no competition for them.

Instead of blaming AMD for not having a competitive GPU at launch, people blame consumers for not buying then waiting for an AMD gpu to become competitive at the high end through drivere optimization, rather than just buying the faster Nvidia GPU at the high end right away.

 

[IllogicalGlory]

AMD was only superior during niche times....
Nvidia has had the better flagship GPU when the MAJORITY of high end users actually buy. At launch. Having a great GPU when everyone has already purchased a GPU is utterly irrelevant. Hence why AMD doesn't do well in the high end. I'm NOT buying Fury X, 290x, etc. because ALL were terrible values at launch... and came out after Nvidia.

You're correct with the Fury X, however, you're wrong about the 290/X. 290 brought 780+ performance at $200 less, 290X brought Titan+ performance at $450 less. Really terrible value for sure. Yeah, the 780 Ti came out a few weeks after, but it was a worse value (290/X had 52%/17% better performance/$ respectively according to TPU) and it had less VRAM.

 

[tamz_msc]

I don't understand the justification for buying high-end NVIDIA at launch prices over AMD's offerings that would give you 80-90 percent of that performance at hundreds of dollars less.

You want your 1080Ti - go ahead. But then, what will you do when a cut-down GV104 gives the same performance for half the price a year later?

 

[Samwell]

You're correct with the Fury X, however, you're wrong about the 290/X. 290 brought 780+ performance at $200 less, 290X brought Titan+ performance at $450 less. Really terrible value for sure. Yeah, the 780 Ti came out a few weeks after, but it was a worse value (290/X had 52%/17% better performance/$ respectively according to TPU) and it had less VRAM.

The 290X was a great card, but AMD totally messed up the launch with the cooler. Choosing between Quiet mode in which the card lost fps quite a bit and Uber mode which was way too loud. Unfortunately this first impression stayed and because of this not even the really good custom models sold. 390X was the same card +5%, but sold much better because of its image. I don't understand why they didn't refresh it earlier.

 

[Phynaz]

You want your 1080Ti - go ahead. But then, what will you do when a cut-down GV104 gives the same performance for half the price a year later?

Enjoy the 1080Ti for half a year.

 

[Qwertilot]

Or, why buy GV104 when there's something notably better coming the year after it releases? NV are very regular now so just jump on when something hits your price/performance target.

 

[tviceman]

I don't understand the justification for buying high-end NVIDIA at launch prices over AMD's offerings that would give you 80-90 percent of that performance at hundreds of dollars less.

You want your 1080Ti - go ahead. But then, what will you do when a cut-down GV104 gives the same performance for half the price a year later?

Was Fury X hundreds of dollars less than the GTX 980 TI? And how much is Vega going to cost?

 

[tamz_msc]

Was Fury X hundreds of dollars less than the GTX 980 TI? And how much is Vega going to cost?

Yeah, because the Fury X was the first time when AMD tried to compete with the NVIDIA single-GPU flagship at the same price, while people seem to forget that AMD and NVIDIA have been competing at different pricing-tiers for a number of years.

Look at the AMD/NVIDIA of the past few generations - HD 4870/GTX 280(5), HD 5870/GTX 480, HD 6970/GTX 580 - each one of those AMD single-GPU cards were better value compared to their NVIDIA counterparts.Knowing this fact, why would anyone buy the Fury X at launch when the R9 290X/R9 390 gave you 70 percent of the performance of the stock GTX 980 Ti at half the price?

With the the launch of Kepler AMD had to cut down the price of the HD 7970(50), and something similar happened after the launch of Maxwell with the R9 290(X) which was followed by a refresh with the R9 300 series. Even so, while everybody agreed that AMD screwed up the prices on these two occasions, it is undeniable that the cards themselves have held up far better expected.

The AMD line-up proves that the upgrade path to reach a certain performance after a given time in the future costs significantly less than what you'd have to spend if you go with the NVIDIA route - yet the mindshare is such that AMD automatically means cheap and inferior while NVIDIA is superior because it costs more?

When the GTX 1070 matches the performance of the flagship GTX 980 Ti at half the price after just one year, When another flagship like the GTX 780 Ti is made obsolete by a GTX 970 in one year, nobody bats an eye - yet when AMD screws up the launch of one flagship, everybody loses their mind.

 

[tamz_msc]

Enjoy the 1080Ti for half a year.

You'd pay 900$ for a product that would last half a year? I get that its your money - you decide how you spend it, but don't you think that's unacceptable?

 

[2is]

I don't understand the justification for buying high-end NVIDIA at launch prices over AMD's offerings that would give you 80-90 percent of that performance at hundreds of dollars less.

You want your 1080Ti - go ahead. But then, what will you do when a cut-down GV104 gives the same performance for half the price a year later?

Some simply prefer Nvidia, some have gsync monitors. And why would 1080ti owners need to do anything at all when GV104 releases?

 

[tamz_msc]

And why would 1080ti owners need to do anything at all when GV104 releases?

Then why were 980Tis selling for 300$ on eBay when Pascal launched?