Ashes of the Singularity June Update will improve performance

not convinced at aots as a bench tool.. if a 290x is better than a 980ti in that game something is pretty fishy some where..
But i guess it needs to be in the mix some where but really cant be used as a definitive result for bench tests.
 
Considering Async Compute can't be used for Nvidia, it really does give AMD a huge edge due to there effectiveness when using it. It's not surprise really. Although haven't seen a 290x best a 980ti
 
well at 2160p "dx12" a 290x managed 30fps average and the 980ti managed 29, that was on a quad core with no HT.
but the 980ti managed the 29fps during dx 11 as well. where the 290x only managed 21 in the dx 11 test.
so i guess you could say that the 980ti was better. and 1fps inst actually smashing the performance bench marks.
but in reality a 290x should be more allong the lines of a 970 (if that) so to be able to go head to head with a 980ti is rather worrying about either nvidia or the game.
 
every one would "given a choice and all else being equal"
thats why im not really sure AOTS portrays a full picture of performance.

AMD architecture is better suited for parallel compute. Hence why DX12 runs very well on it. And with no CPU overhead being a factor, it let's them go further in performance. Ashes has the best implementation of dx12 atm, has the only engine built with the future APIs in mind. It's a valid source. Pascal actually gains a couple FPS, not much but better than before, that's because they improved there technique slightly. Which shows that Maxwell really isn't suited for it. Hence why they either lose performance or stay equal. And at GDC this year devs said the same thing. Good DX12 implementation means you lose no performance coming from DX11, but a better one means more performance. You see that with this game
 
every one would "given a choice and all else being equal"
thats why im not really sure AOTS portrays a full picture of performance.


As stated by NBD, its not an issue with the game. Rather it is nVidias arch itself.

Obviously its understandable that you wouldnt want it to be a formal benchmark, however used as a DX12 bench, I see no reason as to why it shouldnt be done.

AOTS has the best DX12 implementation so far. You probably already know this, however it is down to nVidia that even Pascal isnt up to snuff for DX12. They want to milk Maxwell with the shrink, so will AMD. Only difference is AMDs arch had a better foundation.
 
Last edited:
Pascal does improve it's DX12 performance. Due to there improved Pre-Emption technique and better Dynamic Loading system(instead of Static). So instead of losing or remaining the same from DX11, it so far shows us small improvements. But it's still not enough to do parallel compute, which is why GCN leap frogs ahead. As it was built from the start to be great at parallel compute. Remember when GCN first came out and the OpenCL benches? You had 7970's beating the Titans in OpenCL by A LOT. Things are much closer now, but still shows where DX12 and other APIs are heading towards. It's just simply more efficient to do both at the same time rather than switching tasks and pausing as Nvidia is forced to do.

On a note about the patch though, it went live this morning. I tested it out and the last time I ran it at 1080p, I am now getting those same scores at 1440p. So i'd say big improvement. Everything maxed but MSAA(at 0),shadows(point light and the other one) both at low, and in the Heavy avg score, my lowest FPS was 42 and think the CPU was at 50 iirc. Unless i'm confusing them. Although all maxed but AA at 0, my Heavy avg was at 38FPS. So still a good improvement there, keeping it lowered so my FPS remains in the FS range(40-144)But that's pretty good considering how heavy the bench is why units and all the projectiles.
 
Back
Top