Overclock.net › Forums › Industry News › Video Game News › [computerbase.de] DOOM + Vulkan Benchmarked.
New Posts  All Forums:Forum Nav:

[computerbase.de] DOOM + Vulkan Benchmarked. - Page 17

post #161 of 632
Quote:
Originally Posted by Glottis View Post

how long was computerbase benchmark run, and why can it be considered, because it's convenient for your argument?

BY THE WAY, using TSSAA only enables Async for AMD but not yet for Nvidia.

"Currently asynchronous compute is only supported on AMD GPUs and requires DOOM Vulkan supported drivers to run. We are working with NVIDIA to enable asynchronous compute in Vulkan on NVIDIA GPUs. We hope to have an update soon."

"As for nVIDIAs mythical driver... where is the Maxwell driver that nVIDIA said would enabled Async Compute? We are to believe that Pascal will get such a driver now as well?"

Oh and Pascal does not support Asynchronous Compute + Graphics. What Pascal supports is improved pre-emption coupled with Dynamic Load Balancing. Basically... attempting to run async compute + graphics will cut the Pascal GPU into little pieces (figuratively). What I mean is that the Pascal GPU is separated into GPC clusters. One cluster will handle the Graphics and an adjacent cluster will handle the compute. While one cluster is handling graphics... it cannot handle compute in parallel and vice versa. So if the GPCs are already busy then you will not get a performance boost (as seen under AotS at higher resolutions). At lower resolutions you might get a slight boost (tiny).

I will not hold my breath waiting on an nVIDIA driver to enable this feature (because you need a driver and code to enable it).

By this time next year... the FuryX may in fact be faster than a GTX 1070 in most of the newer titles... as has been the case for AMD GPUs for quite some time now. Even an R9 390x will be giving the GTX 1070 a run for its money as more and more titles optimize for these new APIs (DX12 and Vulkan).

This is sort of the trend since Tahiti.
Edited by Mahigan - 7/14/16 at 7:24am
Kn0wledge
(20 items)
 
Pati3nce
(14 items)
 
Wisd0m
(10 items)
 
Reply
Kn0wledge
(20 items)
 
Pati3nce
(14 items)
 
Wisd0m
(10 items)
 
Reply
post #162 of 632
Is there a hardware reason why Kepler is doing so poor? Or is it just that nVidia doesn't care anymore?
post #163 of 632
both biggrin.gif
FX
(7 items)
 
  
CPUMotherboardGraphicsRAM
FX-8320@4.4Ghz M5A99FX PRO R2.0 AMD Radeon R9 290X Patriot Memory  
Hard DriveCoolingMonitor
Samsung 840 Raijintek Ereboss iiyama X4071UHSU (4K) 
  hide details  
Reply
FX
(7 items)
 
  
CPUMotherboardGraphicsRAM
FX-8320@4.4Ghz M5A99FX PRO R2.0 AMD Radeon R9 290X Patriot Memory  
Hard DriveCoolingMonitor
Samsung 840 Raijintek Ereboss iiyama X4071UHSU (4K) 
  hide details  
Reply
post #164 of 632
this forum is funny. when some benchmarks didn't use Async enabling TSSAA setting for AMD poeple cried fault. but apparently for Nvidia Async doesn't work yet in DOOM, no one cares to even mention that until I just found that out. doesn't matter if it gives only little performance for nvidia, benchmarks should still be invalidated and recalculated when GPUs from both brands run under same settings!
post #165 of 632
problem is that nvidia cant run async wink.gif because their hardware lacks parts that AMD have as standart 4 years ago
FX
(7 items)
 
  
CPUMotherboardGraphicsRAM
FX-8320@4.4Ghz M5A99FX PRO R2.0 AMD Radeon R9 290X Patriot Memory  
Hard DriveCoolingMonitor
Samsung 840 Raijintek Ereboss iiyama X4071UHSU (4K) 
  hide details  
Reply
FX
(7 items)
 
  
CPUMotherboardGraphicsRAM
FX-8320@4.4Ghz M5A99FX PRO R2.0 AMD Radeon R9 290X Patriot Memory  
Hard DriveCoolingMonitor
Samsung 840 Raijintek Ereboss iiyama X4071UHSU (4K) 
  hide details  
Reply
post #166 of 632
Quote:
Originally Posted by OneB1t View Post

problem is that nvidia cant run async wink.gif because their hardware lacks parts that AMD have as standart 4 years ago
except where Pascal has pretty good gains in DX12 with Async. funny how you accept that benchmarks not run under same identical circumstances are OK. if it was other way around I bet you would have different stance.
post #167 of 632
Quote:
Originally Posted by Glottis View Post

this forum is funny. when some benchmarks didn't use Async enabling TSSAA setting for AMD poeple cried fault. but apparently for Nvidia Async doesn't work yet in DOOM, no one cares to even mention that until I just found that out. doesn't matter if it gives only little performance for nvidia, benchmarks should still be invalidated and recalculated when GPUs from both brands run under same settings!

I disagree... because if the driver is never released (as was the case with Maxwell) then people go on faith alone thinking that a driver release from nVIDIA is imminent when in reality... it will never happen.

The benchmarks are not invalidated... where they invalidated due to the fact that the AMD cards were running an older version/path of OpenGL than the nVIDIA cards? I did not see you making that argument then.

For now... this is the reality under Doom. AMD Radeons perform as shown here. If nVIDIA release a driver and the Doom developers release a patch on their end then we can re-evaluate but until then... this is where we are at... just as prior the Radeon poor showings is what every review site showed under Doom and nobody was complaining then.

As for "good gains" under AotS for Pascal using Async-compute + graphics...

Not really...






What we are mostly seeing is the pre-emption and load balancing being used to negate the performance losses we saw with Maxwell. What we also see is that under Extreme or High preset there can be a slight performance boost due to the fact that the GPCs are not overly busy but once we move to the crazy preset (or towards 4K) the GPCs are overloaded and thus we do not get a boost.
Edited by Mahigan - 7/14/16 at 8:11am
Kn0wledge
(20 items)
 
Pati3nce
(14 items)
 
Wisd0m
(10 items)
 
Reply
Kn0wledge
(20 items)
 
Pati3nce
(14 items)
 
Wisd0m
(10 items)
 
Reply
post #168 of 632
Quote:
Originally Posted by Mahigan View Post

I disagree... because if the driver is never released (as was the case with Maxwell) then people go on faith alone thinking that a driver release from nVIDIA is imminent when in reality... it will never happen.

The benchmarks are not invalidated... where they invalidated due to the fact that the AMD cards were running an older version/path of OpenGL than the nVIDIA cards? I did not see you making that argument then.

For now... this is the reality under Doom. AMD Radeons perform as shown here. If nVIDIA release a driver and the Doom developers release a patch on their end then we can re-evaluate but until then... this is where we are at... just as prior the Radeon poor showings is what every review site showed under Doom and nobody was complaining then.
so this about some childish revenge for you. you don't actually care about fair results. and by the way, plenty people complained about OpenGL AMD, it's just there was nothing to do about it, as OpenGL 4.5 wasn't happening for AMD, but Bethesda and Nvidia ARE working on Async for Vulcan. maybe it will only give 1-2% boost, that's still important, and would mean that digital foundrys furyx 7.5% lead would shrink to 5% or even less!
post #169 of 632
Quote:
Originally Posted by Glottis View Post

so this about some childish revenge for you. you don't actually care about fair results. and by the way, plenty people complained about OpenGL AMD, it's just there was nothing to do about it, as OpenGL 4.5 wasn't happening for AMD, but Bethesda and Nvidia ARE working on Async for Vulcan. maybe it will only give 1-2% boost, that's still important, and would mean that digital foundrys furyx 7.5% lead would shrink to 5% or even less!

Nothing I said could be construed as being about "childish revenge". I accepted the Doom figures back then and I accept them now.

As for nVIDIA working with Bethesda in order to implement Async Compute + Graphics (or in their case pre-emption + Dynamic load balancing) I will believe it when I see it. I wish I could say otherwise but I am still waiting on the nVIDIA MAxwell drivers which enable Async Compute.
Edited by Mahigan - 7/14/16 at 8:18am
Kn0wledge
(20 items)
 
Pati3nce
(14 items)
 
Wisd0m
(10 items)
 
Reply
Kn0wledge
(20 items)
 
Pati3nce
(14 items)
 
Wisd0m
(10 items)
 
Reply
post #170 of 632
1. digital foundry results are not correct their fury x card is prolly not working as should for some reason (thermal throttling, power limit etc..)
2. 1-2% boost can be achieved just by optimalizing current build so thats probably what is nvidia doing now biggrin.gif no async for them as they missing HW for that
FX
(7 items)
 
  
CPUMotherboardGraphicsRAM
FX-8320@4.4Ghz M5A99FX PRO R2.0 AMD Radeon R9 290X Patriot Memory  
Hard DriveCoolingMonitor
Samsung 840 Raijintek Ereboss iiyama X4071UHSU (4K) 
  hide details  
Reply
FX
(7 items)
 
  
CPUMotherboardGraphicsRAM
FX-8320@4.4Ghz M5A99FX PRO R2.0 AMD Radeon R9 290X Patriot Memory  
Hard DriveCoolingMonitor
Samsung 840 Raijintek Ereboss iiyama X4071UHSU (4K) 
  hide details  
Reply
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Video Game News
Overclock.net › Forums › Industry News › Video Game News › [computerbase.de] DOOM + Vulkan Benchmarked.