560.94 drivers series -observational notes

[venturi]

Et al.
There isn't a reason to rehash all the benchmarking and fps reporting I did in the first 560 driver releases notes, I'm just confirming that with this version (as expected) there are no improvements on the drop in performance and stability of older titles that don't use DLSS in the 560.94

Stability issues have occurred in Vulkan titles, and DX 9-DX11 drop in performance, and Vulkan drop in performance.

While I use professional applications, I'm using gaming as a common language here as it would be apropos. Internal benchmark enabled games were used. Game titles that use DLSS m DLAA and FG, like Cyberpunk can show a slight increase (1-5 FPS). Titles or the choice mot to use DLSS, shows a drop in performance in the 5-15% range. Benchmark tools that allow for single and multi gpu and testing in both vulkan and DX12 were also used, such as GravityMark.

For those results, please search the prior threads on the 560 series.

the parameters of the observation are on 4090 FE and 3090Ti.
4k, max eye candy, max RTX, no variable resolution, no vignetting, no film grain, not motion blur, no DOF, no DLSS, no FG. no texture optimizations when possible. Max visual fidelity. Ob either DP 1,4a or HDMI 2.1, 10-bit., all monitors have freesynch pro, dynamic dimming FALD background, systems -no overclocking in any way.

Observational conclusion:
in highest visual settings mode, with no DLSS, and the above parameters, the 560 series drivers show a drop in performance compared to 556 and significantly more compared to 522 series. Milage may vary depending on the equipment used and the configuration savviness of the user.


Please look at this from a "Spock" point of view without emotional attachment. While the next part is a polarizing statement, please bare in mind the difference in concepts:

Nvidia --> please match or exceed the performance of the 522 series driver. While DLSS is important to some folks, running at a lower resolution and throwing in the Ai motif is gimmick in exchange for real frame to frame performance at max visuals settings. DLSS may be nice, but don't hobble the hardware at its core performance. Why buy an 8k or 4k monitor just to run it at 2560 or 1080? Please don't hobble the GPUs with the passage of time and or in pre-release to a new GPU release.


Thank you


Driver comparison: 522. 556.12, and 560 series
OS tests, MS Server 2022 data center, latest MS win 11
(Faster performance was consistently present on server 2022)
Non Steam games, or non steam variants, executables run directly without rent games third parties (pure mode)
Some steam games were used on the 4060Ti builds yielding similar results.
Separate benches: GravityMark, Superposition, Haven, Valley, AIDA 64, and Furmark were used as non game benchmarks
Game benchmarks: Quake2 RTX, RDR2, Strange brigade (Vulkan and DX12), Zombie Army 4, Cyberpunk, Chernobylite, Batman Arkham Knight, Total War: Warhammer DX12 (1 and 2), Ashes of the Singularity, the Talos Principle, ROTTR, SOTTR, Deus EX MKD, Different Final fantasy demo benchmarks, and others.

systems tested on:

A
2x 4090 RTX Founders Edition
2x AMD 9684X Epyc (192/384 cores) 2.2GB of L3 cache , Gigabyte MZ73-LM0 Dual socket motherboard, PCIE 5.0
1.5 TB RAM DDR5 ECC LRDIMMs 4800mhz 2050W digital power supply
4x Micron 9300 Max (15.4TB each) Sabrent Rocket 4 Plus (8TB). 2x Back up drives Micron 5300 (8TB)
Asus PA32UCG-K monitor, MS Data Center 2022, MS 11, & Ubuntu,

B
2x 3090Ti Founders Edition, Nvidia nvlink/SLI bridge
2x AMD 773X Epyc (128/256 cores) 1.5GB of L3 cache , Gigabyte MZ72-HB0 Dual socket motherboard, PCIE 4.0
1 TB RAM DDR4 ECC LRDIMMs 3200mhz 2050W digital power supply
4x Micron 9300 Max (15.4TB each) Raid, Sabrent Rocket 4 Plus (8TB). 2x Back up drives Micron 5300 (8TB)
Asus PA32UCG-K monitor, MS Data Center 2022, MS 11, & Ubuntu,

C
2x 3090Ti Founders Edition, Nvidia nvlink/SLI bridge
2x Intel 8280L (56/112 cores) Asus WS Sage Dual socket motherboard, PCIE 4.0
768 TB RAM DDR4 ECC LRDIMMs 2933mhz 2050W digital power supply
4x Micron 9300 Max (15.4TB each) Sabrent Rocket 4 Plus (8TB). 1x Back up drive Micron 5300 (8TB)
Asus PA32UCK monitor, MS Data Center 2022, MS 11, & Ubuntu.

D and E
4060Ti 16GB (2x clone systems)
Intel E-2699A, 512 GB of DDR4 2666mhz, PCIE 3.0
2x Samsung 860 Pro (4TB), raid, Sabrent Rocket 4 Plus (4TB). 1x Back up drive Samsung 860Pro (4TB)
Corsair AX1600i power supply.
Asus PA32UCK monitors, MS Data Center 2022, MS 11

 

[Ruru]

No issues so far with my 3080 10GB.

 

[iameatingjam]

So you are saying if there's some issues with older games, 522 is the way to go?

 

[venturi]

The 560 has issues with older games, my suggestion is to go with the 556.12

522 has better overall performance but may not include recent titles, 556.12 would be my recommendation until improvement occurs

 

[nomdeplume]

I've been experimenting with drivers for the reasons listed in here after noticing changes installing 560.94 with Strix 1060. Improvements to modern titles were noticeable without being as impactful as on older games meeting parameters been suggested above. I would go as far as producing instabilities in my card.

 

[chrcoluk]

The 560 has issues with older games, my suggestion is to go with the 556.12

522 has better overall performance but may not include recent titles, 556.12 would be my recommendation until improvement occurs

Do you plan to keep testing this? as I am on a 4080 super I cant go back to 522, currently I am on 551.86.

 

[iameatingjam]

I wonder if nvidia cares about fixing performance issues on old games...? Especially with them not so much investing in the games driver department.

 

[butcherbird]

Is MS Flight Sim 2020 a game known to be negatively impacted by NVIDIA drivers? 6-12 months ago, my PC was able to run MSFS 2020 well enough. Since then, though, the sim became unplayable -- extremely low frame rates. So, I uninstalled it.

I see the recommendation above to stick with 556.12.

3080ti
i5-11600K processor

Thank you.

 

[iameatingjam]

I noticed in the latest driver update it says this

"[Chaos V-Ray] Performance regression in some workloads when running R560 drivers [4766640]"

Is this.... possibly aimed at this problem? I don't know what chaos v-ray means. Or the R ( does that mean anything else other than just the 560 drivers?)..... Anyway just thought it might be relevant.