Monday, February 3rd 2020
NVIDIA's Next-Generation "Ampere" GPUs Could Have 18 TeraFLOPs of Compute Performance
NVIDIA will soon launch its next-generation lineup of graphics cards based on a new and improved "Ampere" architecture. With the first Tesla server cards that are a part of the Ampere lineup going inside Indiana University Big Red 200 supercomputer, we now have some potential specifications and information about its compute performance. Thanks to the Twitter user dylan552p(@dylan522p), who did some math about the potential compute performance of the Ampere GPUs based on NextPlatform's report, we discovered that Ampere is potentially going to feature up to 18 TeraFLOPs of FP64 compute performance.
With Big Red 200 supercomputer being based on Cray's Shasta supercomputer building block, it is being deployed in two phases. The first phase is the deployment of 672 dual-socket nodes powered by AMD's EPYC 7742 "Rome" processors. These CPUs provide 3.15 PetaFLOPs of combined FP64 performance. With a total of 8 PetaFLOPs planned to be achieved by the Big Red 200, that leaves just a bit under 5 PetaFLOPs to be had using GPU+CPU enabled system. Considering the configuration of a node that contains one next-generation AMD "Milan" 64 core CPU, and four of NVIDIA's "Ampere" GPUs alongside it. If we take for a fact that Milan boosts FP64 performance by 25% compared to Rome, then the math shows that the 256 GPUs that will be delivered in the second phase of Big Red 200 deployment will feature up to 18 TeraFLOPs of FP64 compute performance. Even if "Milan" doubles the FP64 compute power of "Rome", there will be around 17.6 TeraFLOPs of FP64 performance for the GPU.
Sources:
@dylan522p(Twitter), The Next Platform
With Big Red 200 supercomputer being based on Cray's Shasta supercomputer building block, it is being deployed in two phases. The first phase is the deployment of 672 dual-socket nodes powered by AMD's EPYC 7742 "Rome" processors. These CPUs provide 3.15 PetaFLOPs of combined FP64 performance. With a total of 8 PetaFLOPs planned to be achieved by the Big Red 200, that leaves just a bit under 5 PetaFLOPs to be had using GPU+CPU enabled system. Considering the configuration of a node that contains one next-generation AMD "Milan" 64 core CPU, and four of NVIDIA's "Ampere" GPUs alongside it. If we take for a fact that Milan boosts FP64 performance by 25% compared to Rome, then the math shows that the 256 GPUs that will be delivered in the second phase of Big Red 200 deployment will feature up to 18 TeraFLOPs of FP64 compute performance. Even if "Milan" doubles the FP64 compute power of "Rome", there will be around 17.6 TeraFLOPs of FP64 performance for the GPU.
172 Comments on NVIDIA's Next-Generation "Ampere" GPUs Could Have 18 TeraFLOPs of Compute Performance
to refresh rdna1 in 2020 ?
I don't know how your getting all this information cause so far amd mentioned rdna2 is rdna1 with rt and vrs but you claims it'srevolutionary for some reason,while nvidia said nothing about ampere and yet you know it's disappointing.
Since we are at the source. Any source what Ampere actually is?
BTW it is still sarcasm.
at this point you're pretty much trolling.
dude,stop trolling every gpu thread with false information.
At this point, no one (even at Nvidia) knows exactly how these new cards will perform.
and 1660's performance at 150$ may be????
who knows.....
With all do respect. I don't underestimate anything. The specs for RDNA2 are not there. Just as NV's aren't. I'm trying to be objective but it does not bode well with some members here.
Question. What are those specs? NAVI 2x? what is that mean? BTW I can say same thing about NV. Just as our friend Tarlson said Ampere it is something totally different from Turing in terms of arch (not RDNA2 becuse that is just a refresh) Well, The AMPERE, I see RT cores, Cuda, Tensore. Is this something to justify new? It looks same to me as Turing in every way. I'm not here to mark what you say but please use arguments. Comparison of the two at this point is crazy. Companies can do whatever they want. It is a market and they will do anything to sell stuff.
Agreed. Anything can happen.
Today's AMD GPU division is not the company that refreshed Tahiti for 5 or 6 generations. The evidence of this was Polaris that was just as fast as Tahiti with 1/2 the power draw (That was the biggest caveat to AMD cards). Vega was good but released during the mining craze that made it's price go through the roof this made it less of a proposition vs Nvidia.
The 5700XT is currently AMD's fastest card. Not that we can but if we look at Ryzen it started at 8 cores and 16 cores max with TR4 in 2017. It is 2020 and the core count has increased to 16 and 64 cores max for HEDT. That is in 3 years, much more than Intel has done in 10. I have confidence in Lisa Su and have not heard her say anything that has not come to pass. I am not saying it is definite but I do believe the propensity is high, besides we need Nvidia to bring prices back to where they were before the mining craze.
They refuse to support newer versions of it. Radeons have supported OpenCL 2.0 since the 7790.
It's like nV making an incompatible Variable Refresh tech, to the open standard in DisplayPort. Intel adopted the open spec, the HDMI group came on board, and adopted it. Suddenly nV made gSync work with Adaptive Vsync no NV port corruption necessary.
NV will block and slow down OpenCL as much as they can. I suspect you'll see their tune change if the Intel and AMD really start challenging NV.
I still hate that NV does everything they can to make PhysX all but impossible if your main GPU isn't NV. Titan V?
I was just trying to point out that AMD's ability to adapt their lineup to what Nvidia launches is very limited, and vice versa of course. Beyond price and clock speeds, there is little they can adjust over night. It's not like they can whip up a new chip design in a couple of months. We know of three Navi 2x chips so far; Navi 21, Navi 22 and Navi 23 from Linux driver patches. If there are more than these three chips coming, then will not be launching anytime soon. I don't know exactly where you see these details, to my knowledge Nvidia haven't released anything public about "Ampere" so far.
But regardless, they've been using "Cuda cores" for many generations (since Tesla?), but their performance characteristics have changed radically over time. Just from Pascal to Turing there were major efficiency gains.
I don't know if Nvidia's next gen will be a small tweak of Turing or if it's another major architecture, but deriving this from vague rumors is utterly pointless. I've seen nothing substantive about Nvidia's next gen so far, and a pro-tip; whenever you see rumors pointing in every direction, it's usually an indication that "all of them" are just BS, it's a typical symptom of the information vacuum we're in right now. This is why I have recommended to be cautious about most "Ampere" or Navi 2x rumors so far. When we are getting closer to a real launch, we will start to see believable rumors pointing in the same direction.
As for Nvidia's next gen, you have to acknowledge that even if it's just a shrunk and tweaked Turing, it's going to cause some real challenge for AMD to compete with. Nvidia currently have a huge lead thanks to their architectural superiority, even with AMD having a node advantage. With the next gen that advantage will be gone for AMD.
do I believe ? yes,strongly.I mean you have to be delusional to think otherwise when amd's last high end competitor was fury x.
with every release their goalposts move,after fury x their high end was vega 64 that went against gtx 1080,now they're gladly selling rtx 2070 (106 die) competitor at +$400 with a limited feature set.
Why cant the red fanbase accept the obvious-amd have introduced the NVIDIA model.they are gladly selling 250mm dies at $400 and making more money than before so they dont event bother with chasing nvidia in the high end when they're milking the mid range
I do acknowledge this and I have mentioned this in the previous posts. I am sure that the new NV release will be fast and great. Not sure about the price though. Actually I'm worried about the price but that's just me.
Like, freaking, for real?
GPU gap was never even remotely as big and bad as the gap we had in CPU market. Only in Raja's time, with modest R&D obviously focusing on Zen, AMD went "missing in action". But it's getting back, with entire set from 5700 to 5500 series looking good. Bigger chips are inevitably coming.
Amd has the design to beat intel in every market,if they can do the same with navi I will be happy to see it rather than hear their marketing talk about disrupting 4k gaming.
They're on 7 nm tsmc and new uarch,why dont they put up? Cause of what I wrote before.They would lose to nvidias +$500 cards cause of lack of features,so they're happy to milk mid range.
Let us see those chips,I am not buying a word their CEO or their fanbase say about high end Radeon.