NVIDIA GeForce 20 Series (Volta) to be released later this year - GV100 announced

[Krteq]

It's more about the feature set than anything else. Vega 10 doesn't have DP nor any kind of DL accelerator.

Vega 10 supports DPFP with 1:16 rate.

 

[jpiniero]

Now I have to start saving for the upcoming GV104 part that's just around the corner. Crazy how fast the PC GPU market is moving! I haven't even had my 1080 for a year and it's already outdated not only by a higher tier Pascal card (that I was expecting) but by a whole new architecture and process node (this release date took me by surprise).

Well, just by judging the die size given the transistor count the 12 nm node nVidia is using isn't much denser. At this point I would say that any talk that GV102 has 5376 cores and GV104 3584 is premature.

Vega 10 have a FP64 support in 1:16 rate.

You know what I mean. GV100 is 1:2.

 

[T1beriu]

Vega 10 doesn't have DP nor any kind of DL accelerator

Vega 10 supports DPFP with 1:16 rate.

You know what I mean. GV100 is 1:2.

No. People don't know what you mean unless you write it. We can't read your mind. People only know what you write. Just admit you were wrong and move on.

 

[jpiniero]

No. People don't know what you mean. We can't read your mind. People only know what you write. Just admit you were wrong and move on.

I'm guessing you don't understand the importance of having fast DP for certain markets. It should be sort of implied.

 

[T1beriu]

Wow.

 

[iBoMbY]

Vega 10 supports DPFP with 1:16 rate.

That is pure speculation at this point. Vega architecture supports up to 1/2, according to official information, and just because one Vega 10 SKU may have been intended to be configured to 1:16 at some point, doesn't mean there won't be other SKUs with other configurations:

 

[Krteq]

My post was based on this info

Anyway, we are OT here.

 

[Carfax83]

Vega get 12.5 TFlop @ 300watt on a significantly smaller GPU (smaller dies are harder to cool), so whats with all the doom and gloom? NV's biggest advantages in gaming has nothing to do with ALU, for a long time (fermi?) they had a front end/geometry setup advantage which got fixed with Polaris. Then since maxwell they have had a very big raster advantage.

My point was that as long as AMD still lags significantly behind NVidia in the performance per watt category, they won't be able to catch up and reach parity. Take the recently released Titan Xp for example. Its' advertised Tflop rating is 12, but in reality it's closer to, or slightly above 13 due to boost. And this is in a 250w TDP envelope.

To hit 12.5 TFlops at 300w, AMD will likely require an AiO cooler a la Fury X, which just emphasizes what I'm saying. I'll give credit to AMD though, they have certainly come a long way since Fiji when it comes to performance per watt, and with Vega, I hope they take it even further. But they will need much more to be on somewhat equal footing with NVidia.

 

[Carfax83]

Zen and Vega are the first major CPU and GPU design efforts by AMD in half a decade. AMD has proved with Zen they can design a power efficient and competitive CPU core. So there is no reason to doubt that they cannot design a power efficient GPU architecture. Anyway the goal for AMD is to close the efficiency gap with Zen/Vega against Intel and Nvidia and continue to keep doing so. AMD does not need the performance crown. They need power and area efficient architectures (perf/watt and perf/sq mm). Most importantly they require consistently good execution. With competitive CPU and GPU architectures AMD has the unique capability of providing the world's best x86 APUs - something which Intel and Nvidia cannot do . There is one key missing technology which needs some time to be available in volume and at good yield/cost -HBM. HBM2 is just starting to become available but HBM3 and low cost HBM might be the tipping point when AMD's Fusion concept truly delivers to its potential. HPC server APUs, Game console APUs, Notebook / Desktop APUs are all going to need massive bandwidth at 7nm as its a massive process node shrink of almost 60% and extremely powerful GPUs are going to be packed into all these chips. Truly AMD has a bright future. They just need to keep building on the strong foundations they are building today.

I don't doubt AMD's capability to design and build power efficient architectures, whether CPU or GPU. But I do doubt their ability to match NVidia in that category and harken back to the days when they traded blows and were on much more even footing. I really want Vega to be something special and blow the doors off of everyone's expectations, but the fact that we are this close to the launch and AMD still seems so hesitant to share performance estimates makes me wary. Although I am partial to NVidia, I understand that a strong AMD means a stronger NVidia as well, and is beneficial to the consumer.

 

[SpaceBeer]

@Carfax83
Take a look at post above yours. 12TF, 225 W TDP, no AiO cooler, will be available in a month

So, users who need FP16 or FP32 will save a lot of money if they choose P40 or MI 25 (which are more or less the same in terms of computing performance and power consumption).

What V100 offers is better FP64 performance, though P100 is still great, especially if it's easer and cheaper to get at this moment. And of course Tensor cores, but not all HPC users need those. So many of them will go for P100 or other chips, even if V100 is available
http://www.cray.com/products/computing/cs-series/cs-storm

 

[raghu78]

I don't doubt AMD's capability to design and build power efficient architectures, whether CPU or GPU. But I do doubt their ability to match NVidia in that category and harken back to the days when they traded blows and were on much more even footing. I really want Vega to be something special and blow the doors off of everyone's expectations, but the fact that we are this close to the launch and AMD still seems so hesitant to share performance estimates makes me wary. Although I am partial to NVidia, I understand that a strong AMD means a stronger NVidia as well, and is beneficial to the consumer.

Generally AMD has made too much marketing noise / hype and failed to deliver on performance and efficiency in the past as we saw in Polaris or Fiji. I would rather prefer if AMD just delivers on performance and efficiency. If they do so while holding their cards close to themselves before launch thats fine. What matters is the final product on launch. As far as Nvidia is concerned they are marching on furiously at breakneck pace. Its upto AMD to catch up with them to be able to compete and make money in the GPU business otherwise Nvidia will trample them as we saw them do with Maxwell. If AMD competes everyone of us benefits and thats why I always root for the underdog.

 

[raghu78]

I also think that it will be 2560 CUDA cores, which already brings it to the level of the GTX 1080, and I'm expecting 30 percent additional improvement due to the architectural changes, which would enable it to match the GTX 1080 Ti. This has been the trend in the past - 780 Ti vs 970, 980Ti vs 1070.

Nvidia's goal with GTX 2070 would be to match GTX 1080 Ti just like they did with GTX 970 vs GTX 780 Ti. Cuda core count would be 3200 (50 SM) or 3328 (52 SM). I don't expect a Kepler->Maxwell like jump in perf/core. I would say even if they get 10% better perf/core they would have done a fantastic job given the power efficiency gains.

 

[tamz_msc]

Nvidia's goal with GTX 2070 would be to match GTX 1080 Ti just like they did with GTX 970 vs GTX 780 Ti. Cuda core count would be 3200 (50 SM) or 3328 (52 SM). I don't expect a Kepler->Maxwell like jump in perf/core. I would say even if they get 10% better perf/core they would have done a fantastic job given the power efficiency gains.

I think that the 2070 will get the same core count as the 1080, but with a modest improvement in clocks, faster GDDR6 and architectural improvements it should catch up with the 1080Ti. Kinda like what happened with GTX 680->GTX 770.

 

[exquisitechar]

I think that the 2070 will get the same core count as the 1080, but with a modest improvement in clocks, faster GDDR6 and architectural improvements it should catch up with the 1080Ti. Kinda like what happened with GTX 680->GTX 770.

Yes, this is likely.

Although I believe the 770 was just a rebrand of the 680.

 

[tamz_msc]

Yes, this is likely.

Although I believe the 770 was just a rebrand of the 680.

By that I mean that the core config would be the same as the GTX 1080, just like the 680 and 770. Besides the x70 part usually has half the number of cores of the largest fully enabled part.

 

[moonbogg]

I wonder what the 2080ti will be like. Bring it. Its probably 18-24 months out at this point at least,

 

[[DHT]Osiris]

I wonder what the 2080ti will be like. Bring it. Its probably 18-24 months out at this point at least,

I'm casting my vote for solid 90+FPS on all AAA titles @4k at release, and guaranteed 60fps @4k for titles in 2020-2023. Along for the ride with this we'll see a small flush of 5k panels, followed by 8k and 4k @ high FPS and high chroma settings which this new gen will happily play in @60-90hz.

Hopefully some no-kidding 4k VR sets will crop up too.

 

[Phynaz]

It appears that both "brand cheerleading camps" are getting way ahead of ... everything.

The hype on both brands is enormous. Cool it down guys. Interesting times ahead.

What hype? Volta exists. It can be ordered today.

Where's Vega?

 

[Phynaz]

about which we don't know anything

Then I'm afraid you aren't trying hard enough. Volta's performance is now known.

 

[Malogeek]

What hype? Volta exists. It can be ordered today.

Let's revisit this when you can actually get a consumer Volta-based GPU.

 

[Saylick]

I wonder what the 2080ti will be like. Bring it. Its probably 18-24 months out at this point at least,

I think it wouldn't be too farfetched to see GTX 2080 being GTX 1080 Ti + 20%, similar to how GTX 980 squared up against the GTX 780 Ti. GTX 2080 Ti will thus be another 30% on top of the GTX 2080, or about 50% faster than a 1080 Ti.

EDIT: I wouldn't be surprised to see Volta be a rehash of Maxwell but at 12nm/16nm, i.e. 400mm^2 for GTX 2080 (3500 CCs, +20% IPC over Pascal, same clocks as GTX 1080 Ti) with 600mm^2 for GTX 2080 Ti (5120 CCs, +20% IPC over Pascal, same clocks as GTX 1080 Ti).

 

[Glo.]

What hype? Volta exists. It can be ordered today.

Where's Vega?

Then I must have missed any gaming benchmarks of Volta based GPUs. Can you handle me any one of them?

 

[Head1985]

I think that the 2070 will get the same core count as the 1080, but with a modest improvement in clocks, faster GDDR6 and architectural improvements it should catch up with the 1080Ti. Kinda like what happened with GTX 680->GTX 770.

i expecting volta will have better delta color compresion and 2070 wont need GDDR6.

2080
3584SP
256bit 12GB GDDR6 14-16Ghz
10-20% faster than 1080TI

2070
2560sp
256bit 12GB GDDR5x 11-12Ghz
2070 will run max out with low oc headroom.10% slower than 1080TI.I dont expecting 2070 match 1080TI because there is huge gap now vs them.1080TI is 65-75% faster than 1070.They manage only 50% performance gain with 970 to 1070.So 2070 will need bring 70% performance gain vs 1070 and thats not gonna happen.

Edit:
2060
1792SP
192bit 9Ghz DDR5 or 10Ghz GDDR5x
10%faster than 1070

 

[tamz_msc]

i expecting volta will have better delta color compresion and 2070 wont need GDDR6.

2080
3584SP
256bit 12GB GDDR6 14-16Ghz
10-20% faster than 1080TI

2070
2560sp
256bit 12GB GDDR5x 11-12Ghz
2070 will run max out with low oc headroom.10% slower than 1080TI

Edit:
2060
1792SP
192bit 9Ghz DDR5 or 10Ghz GDDR5x
10%faster than 1070

I don't think GDDR5X would be used any longer. By early 2018 Hynix, Micron and Samsung should shave GDDR6 available in plentiful supply.

 

[ElFenix]

Keep it respectful, keep it Volta-centric, and don't feed the trolls.

AT Moderator ElFenix