- Joined
- Apr 28, 2011
- Messages
- 267 (0.06/day)
| System Name | VENTURI |
|---|---|
| Processor | 2x AMD 7773x Epyc (128/256 cores) |
| Motherboard | Gigabyte MZ72-HB0 Dual socket motherboard |
| Cooling | Air, noctua, heatsinks, silent/low noise |
| Memory | 1.TB 2 LRDIMM ECC REG |
| Video Card(s) | 2x 4090 FE RTX |
| Storage | Raid 0 Micron 9300 Max (15.4TB each / 77TB array - overprovisioned to 64TB) & 8TB OS nvme |
| Display(s) | Asus ProArt PAU32UCG-K |
| Case | TT miniITX P1 (SFF) |
| Audio Device(s) | harmon Kardon speakers / apple |
| Power Supply | 2050w 2050r |
| Mouse | Mad Catz pro X |
| Keyboard | KeyChron Q6 Pro |
| Software | MS 2022 Data Center Server, Ubuntu |
| Benchmark Scores | Gravity mark 144,742 (high score) |
Some performance observations comparing this driver to the prior, I tried to be methodical
I noticed some strange behaviors in other imaging applications, and as these would not have common ground in this forum, I decided to use games with built in benchmarks as common ground :
Not earth shattering, I just checked the driver with games that have a built in benchmarks.
I dusted off some game folders and ran their executables. No steam implementation.
Total War, Warhammer 1 and 2, running in DX12, has a 21% performance drop with this driver
Chernobylite, with DX12, shows 13% performance drop.
Cyberpunk shows a 5 fps lift, with frame generation, without FG, there's is a drop of 2 fps
Batman, Arkham Knight. 7% drop
ROTTR - 7% drop
SOTTR - within a margin of error, unchanged.
Borderlands 3, 3% drop
Zombie Army 4, 6% drop, stutter introduced*
Quake 2 RTX, 5% drop, ver 1.7
Doom 3 (not BFG) and Quake 4 - 7% drop
Unreal Tournament 3, 4% drop
RDR2. within a margin of error, unchanged. but stutter introduced*
Ashes of the Singularity 3% drop (multi-gpu)
Strange Brigade, 5% drop in dx12, unchanged in vulkan (multi gpu) stutter introduced*
Quake 2 RTX , 11% drop (multi gpu, early version)
Gravity Mark 18% drop (multi-gpu)
some considerations. I don't DLSS resolution reduction. I may use DLAA and I use frame generation, but not the DLSS resolution reduction stuff (never understood buying a nice monitor and running one or two resolution sizes lower using DLSS, I wonder if folks get it?) I run at max quality, as in max visuals and effects, but NO vignette, NO film grain (another I don't understand) NO motion blur, and no depth of field. Ray Tracing on maximum. No steam, all games are run on local executables. GSynch / Freesynch on, using HDMI 2.1, no VRR, TAA or FXAA when TAA not available. No path tracing (too much noise) only RTX ultra/max, hopefully surface noise will get better on path tracing eventually. No overclocking
DDU used as well. No steam for any of the games, local executables only.
All PCS use Resizable Bar and HAGS
*When I install the LunarG vulkan latest runtime, behind the driver install, it solved the drop in vulkan the stutter issue.
Caveats:
Systems observed on:
mine, my wife's, and my kids' systems. My rig has dual 4090 (hence the multi gpu, non SLI entries, 4k, 10 bit) wife's is dual 3090 FE Ti nvlink/SLI, 4k, 10 bit, kids are 4060ti 16gb each, 1920-1080, 10 bit,
Observational conclusion.
When no DLSS resolution reduction is used, using games with built in benchmarks, across dx12, vulkan, and max settings.
This driver exhibits a reduction in performance in most titles.
Cyberpunk exhibited an increase of 5ps, max settings, FG, nor DLSS resolution reduction. So from these titles, only Cyberpunk seemed to benefit and I suspect that's from the 3.7 dlss and FG from the driver.
On my kids PCs, I did some testing using DLSS, quality and the results where demonstrative of slight improvement, cyberpunk gaining 11-16fps
Driver reversion solved the performance drop on the 4k machines except for cyberpunk which lost 5 fps it had gained. DLSS gains went away for benchmarks on the kids' machines.
For this driver, If I stand back and look at it, it seems DLSS has improvements, at the expense of titles that do not use it, disappointing.
Additionally, this driver borks our asus PA32UCG monitors as they only come on in Windows when we recycle the power on the monitors, driver reversion fixed this issue, reproducible.
I noticed some strange behaviors in other imaging applications, and as these would not have common ground in this forum, I decided to use games with built in benchmarks as common ground :
Not earth shattering, I just checked the driver with games that have a built in benchmarks.
I dusted off some game folders and ran their executables. No steam implementation.
Total War, Warhammer 1 and 2, running in DX12, has a 21% performance drop with this driver
Chernobylite, with DX12, shows 13% performance drop.
Cyberpunk shows a 5 fps lift, with frame generation, without FG, there's is a drop of 2 fps
Batman, Arkham Knight. 7% drop
ROTTR - 7% drop
SOTTR - within a margin of error, unchanged.
Borderlands 3, 3% drop
Zombie Army 4, 6% drop, stutter introduced*
Quake 2 RTX, 5% drop, ver 1.7
Doom 3 (not BFG) and Quake 4 - 7% drop
Unreal Tournament 3, 4% drop
RDR2. within a margin of error, unchanged. but stutter introduced*
Ashes of the Singularity 3% drop (multi-gpu)
Strange Brigade, 5% drop in dx12, unchanged in vulkan (multi gpu) stutter introduced*
Quake 2 RTX , 11% drop (multi gpu, early version)
Gravity Mark 18% drop (multi-gpu)
some considerations. I don't DLSS resolution reduction. I may use DLAA and I use frame generation, but not the DLSS resolution reduction stuff (never understood buying a nice monitor and running one or two resolution sizes lower using DLSS, I wonder if folks get it?) I run at max quality, as in max visuals and effects, but NO vignette, NO film grain (another I don't understand) NO motion blur, and no depth of field. Ray Tracing on maximum. No steam, all games are run on local executables. GSynch / Freesynch on, using HDMI 2.1, no VRR, TAA or FXAA when TAA not available. No path tracing (too much noise) only RTX ultra/max, hopefully surface noise will get better on path tracing eventually. No overclocking
DDU used as well. No steam for any of the games, local executables only.
All PCS use Resizable Bar and HAGS
*When I install the LunarG vulkan latest runtime, behind the driver install, it solved the drop in vulkan the stutter issue.
Caveats:
Systems observed on:
mine, my wife's, and my kids' systems. My rig has dual 4090 (hence the multi gpu, non SLI entries, 4k, 10 bit) wife's is dual 3090 FE Ti nvlink/SLI, 4k, 10 bit, kids are 4060ti 16gb each, 1920-1080, 10 bit,
Observational conclusion.
When no DLSS resolution reduction is used, using games with built in benchmarks, across dx12, vulkan, and max settings.
This driver exhibits a reduction in performance in most titles.
Cyberpunk exhibited an increase of 5ps, max settings, FG, nor DLSS resolution reduction. So from these titles, only Cyberpunk seemed to benefit and I suspect that's from the 3.7 dlss and FG from the driver.
On my kids PCs, I did some testing using DLSS, quality and the results where demonstrative of slight improvement, cyberpunk gaining 11-16fps
Driver reversion solved the performance drop on the 4k machines except for cyberpunk which lost 5 fps it had gained. DLSS gains went away for benchmarks on the kids' machines.
For this driver, If I stand back and look at it, it seems DLSS has improvements, at the expense of titles that do not use it, disappointing.
Additionally, this driver borks our asus PA32UCG monitors as they only come on in Windows when we recycle the power on the monitors, driver reversion fixed this issue, reproducible.
Last edited:
