I know that. And the recent introduction of Maxwell wouldn't allow it to show up in the Green 500 in the first place. Just pointing out that it doesn't add a lot to the discussion about GCN and the chips against which it will have to compete.
AMD: Pirate Islands (R* 3** series) Speculation/Rumor Thread
- Thread starter iMacmatician
- Start date
-
- Tags
- amd
firstminion
Newcomer
firstminion
Newcomer
But it adds a lot to the discussion about GCN and the chips against which it's competing right now. I would like to add that GCN was released in very early 2012.
Yes, if you care about perf/W in an HPC context.
firstminion
Newcomer
I see your point, AMD targeted perf/W on HPC when nvidia targeted perf/W on gaming (and relevant computing workloads). Maxwell doesn't target HPC.
I don't know if AMD specifically cared about HPC perf/W. (Though I doubt it.) I'm just acknowledging that they seem to have an edge in that respect compared to Kepler. We don't know how it would compare against an FP64 enabled Maxwell because such a thing doesn't exist.
The HPC numbers are interesting by themselves if you care about perf/W in HPC, but I have no clue what to do with those numbers when talking about desktop gaming or even desktop compute. There is only a 10% difference in perf/W between the AMD based and Nvidia based supercomputer.
Is that because of silicon perf/W? If I look at perf/W at hardware.fr it shouldn't. Is it because they have a difference power architecture? We're talking megawatts here. Because the cooling requires more power? Is it because the AMD HPC is specifically tuned for perf/W (eg lower clocks) while the Nvidia one is tuned for max performance? Is AMD specifically better at FP64 than Nvidia?
10% difference is too little to make conclusions when there are this many variables that we don't know and that we don't have any expertise in.
Imagine Anandtech reviewed Kepler and GCN. But the setups had different PSUs, different methods of cooling (say water vs convection only), different CPUs with different amounts of RAM and different amounts of storage and types of storage (say SSD and mechanical HD). And one GPU overclocked and the other not. And one is SLI and the other not. And they run exactly one benchmark. And they only measure wall power of the whole system.
And then they would conclude that one GPU architecture has 10% better perf/W.
An interesting data point on its own, if you care about comparing machines you can buy from Alienware vs Northwest, but useless when discussing minute details of the GPUs only.
Last edited:
firstminion
Newcomer
It's useless to suppose how would a theoretical FP64 Maxwell fare against GCN. If nvidia didn't make one it's because they don't want to make one. Anyway, you can't disregard actual results saying "Yes, it compares HPC numbers, but it doesn't compare against maxwell" when there's nothing to compare to and then rollback and say that "we shouldn't use HPC to compare".
See the problem? At first the issue was that the 500green doesn't have Maxwell numbers, later 500green is irrelevant by itself. Either way the L-CSC team is computing away and AMD is right to use it as a proof of their capabilities.
This discussion lost it's point AFAIK. Would you care to explain what point are you trying to make besides HPC is different to desktop gaming?
See the problem? At first the issue was that the 500green doesn't have Maxwell numbers, later 500green is irrelevant by itself. Either way the L-CSC team is computing away and AMD is right to use it as a proof of their capabilities.
This discussion lost it's point AFAIK. Would you care to explain what point are you trying to make besides HPC is different to desktop gaming?
My point was that HPC shouldn't have been brought up by Dave in the first place, because it adds nothing to the discussion. We're in complete agreement.
Frenetic Pony
Veteran
Yay me! Earlier speculation confirmed, as I forgot that Tonga runs at 918mhz, so the extra 10% needed to bump 4096 "Stream Processors" to the rumored performance of Fiji is just a clockspeed bump. Not to knock GCN 1.2 too much, Tonga is a pretty decent GPU. But that would also mean AMD was indeed synthing GCN 2.0 on 20nm, and when they found yields weren't up to GPU work delayed it into next year and going on Finfet. Which isn't too hard of a transition as the new Finfet nodes have the same backend and density as 20nm.
D
Deleted member 13524
Guest
Wasn't that obvious from the first rumors?
What else were you hoping for?
The Synth is strong with you!
But I don't think anything of what you describe about actually happened. Because it doesn't make any sense.
Except they did want to make one... why they didn't is anyone's guess.
Yep, i think Nvidia and AMD was know a long time ago, they will be on 28nm,,, and then jump to 14-16nm when it will be ready.. 20nm was dead and buried a long time ago.
I've suspected it for a long time but until now, there was nothing official(-ish) to support it.
