I'm not going to pretend to full understand the technical side, while also not wanting to parrot what I have brought up in the other thread, but what sort of major gains would people be seeing? A rough % and an example of a game/benchmark that gets this sort of performance increase with async enabled. Looking at scores from AMD hardware, this implementation of async compute is the fastest so far. Why are people complaining?
I would not worry about Maxwell and its async capabilities. nVidia may enable it once they are completely EOL and people would look at their performance/watt or performance/$ as so bad that they would not have resale value. They may as well enable Async compute on GeForce 8800 GTX which for today's standard is actually just space heater. What I mean by that is that enabling performance increasing feature at time people no longer care about given HW is pointless. So nVidia should act soon. I do not even remember how long it is since they promised that magical driver enablement.
Maxwell Overclock 40-50%, so it does not need Async shaders Did a run with 2x 1080 gaming x with HB Bridge: http://www.3dmark.com/spy/32439 13 187 with NVIDIA GeForce GTX 1080(2x) and Intel Core i7-6900K Graphics Score 13 955 CPU Score 10 055
I have August in my head, so thinking it must be getting close to a year. If they couldn't do it when Maxwell was their top card, the incentive must be pretty much gone now. nizzen, surely another 10% on top of that would be even better? Saying that, I would rather get 10% more performance from an overclock than from something like async compute. Edit: BTW, I'm using 10% as its the average for async, not what can be achieved with an oc.
10896 at default settings. GPU's with mild overclock +171, memory +34, core voltage +5, 84C max temperature allowed. CPU all cores at 4.0 GHz.
Intel HD530 GPU Time Spy (Default/Async Compute Enabled) score: 423 Graphics Test 1: 2.39 fps Graphics Test 2: 2.09 fps "Better than 0% of all results" lol I came last - is there a prize for that?
I lost about 200 points in total score and 1000 points in CPU Score..Damn 6950x's are beasts.. TimeSpy Result:
Futuremark released official statement regarding AC implementation in Time Spy. Very interesting reading for people wanting to know what is going on under the hood. The reading shows that despite the fact DX12 allows programmers for more control, drivers are still crucial -> please look at GTX970 example. We can now stop blaming developers for AC implementation in games causing lower performance on Maxwell GPU's (eg. Ashes of the Singularity). http://www.futuremark.com/pressrele...ime-spy?_ga=1.136205553.1110393204.1468547815
KingP|n testing the GTX 1080 Classified ... http://www.3dmark.com/hall-of-fame-2/timespy+3dmark+score+performance+preset/version+1.0/1+gpu
Got 6284 on my single GTX1070 at +120 core +500 memory but had to back off the overclock as I could see artifacts in the benchmark even though it passed. Settled for +100 core/+600 memory (i thought initially memory was the issue but it was the core causing the issue). This gets me 6260 so not much of a loss in performance. What voltage settings are you guys running for your overclocks on the new pascal cards?
no kidding , they are insane , also requiers tons of voltage if you're going higher than 4.3ghz , i think 4.4ghz needs 1.4v
Is this a response to my question about how much voltage you guys are running? I have it currently at 100% but this is because I have my GTX1070 under water cooling and get no higher than 47C on the GPU. Is this kinda like Chuck Norris doesn't walk on land, he swims through it?
I tried this bench finally and expected ~5000marks but i got 6465 instead on my system http://www.3dmark.com/3dm/13570282 6868 marks: http://www.3dmark.com/3dm/13575975
My best Time Spy so far: http://www.3dmark.com/spy/141742 Funny enough, Time Spy is way more OC friendly then Fire Strike for me. I can't get away with anything above +85 on core there.