Nvidia GeForce GTX 900 Series MegaThread: FAQ and Resources

Page 18 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
I got yesterday 980ti g1. It's amazing card, easily overclocked to 1500/8000, in my case for 1080p gaming so I will have planty of room even if I upgrade to 144hz due to reason that I'm avoiding nvidia gameworks for a reason. G1 all the way 😀
 
http://www.techpowerup.com/reviews/ASUS/GTX_980_Ti_Matrix/23.html
perfrel_1920_1080.png
 


Single 980ti have 6gigs of vram, use less power and provide less heat then 970sli. Not all games support sli neither. I sold my 970 and bought 980ti and there are no regrets.
 


Cheapest at the moment should be G1. With overclock on stock bios I'm getting huge boost performance, it has highest base clock aswel if I'm correct. It's cold, quite aswel in most cases. So I would go for g1.
 


No problem, btw SLI doesnt stack vram, so that graph up there is not correct, on 970 SLI you would still have 4GB of VRAM. Not 8 as it's stated.
 


I think that Async transfer queue support is independent of the API. It's good if someone checks it out. I don't have my PC with me, nor do I have that game.

They have already released the dx12 drivers.
 
Maybe that is for pascal? But for hitman the performance advantage for radeon was the game engine itself was mainly build for radeon hardware. So even without async compute the game already faster on radeon hardware. Though when it comes to DX12 it is much more complicated because even radeon can suffer from negative percormance. As for async compute itself to be honest i'm not really sure if the feature will use by many developer unless they were sponsored by amd. As mentioned by hitman dev you need to make specific tweak for each card instead of general optimization on the architecture itself.
 
In all the hype over Async Compute, only HardOCP decided to test it specifically. While everyone was inclined to accept AMD's performance advantage in Ashes of the Singularity as a direct result of Async Compute, HardOCP noticed that there is a toggle switch, Async On and Async Off:

You can toggle Async Compute on and off in Ashes of the Singularity with the "AsyncComputeOff" flag in the Settings Initialization file. Over the last couple of weeks I have worked through both AMD and Intel CPUs toggling Asynch Compute on and off with both AMD and NVIDIA video cards. The fact is that even on the highest end of the performance spectrum, the biggest frames per second increase we saw was 3 frames per second; from 90.5 to 93.3...a whopping 3% increase, which I would suggest is within our margin of error. That said, I was able to reproduce that increase consistently, so I would suggest that the increase is "real" even inside our MoE. Generally though, we would see gains of less than 1 frame per second when toggling between Async on and Async off.


It would seem to me that while AotS is being held up as the go-to Async Compute game, it is far from it.
http://www.hardocp.com/article/2016/04/13/dx11_vs_dx12_amd_cpu_scaling_gaming_framerate/5#.VyNtCY-cGUk