Greetings all,
This is my third early analysis of an nvidia driver, and as such, it will be the first reference I will have to guide us in future versions. This will be something totally voluntary, so that you can be guided in some way, do not take it as something absolute because each PC configuration is a totally different world, so I will try to make it as global as possible (referring e.g. to the configuration of a game).
Is it worth updating the driver for my current GPU? This is a recurring question from many gamers and PC users. The general rule suggests updating our GPU drivers to the latest version, as it ensures that we get the latest bug fixes, security updates, optimizations for most of the latest AAA games, and support for the newest features.
However, it does not guarantee the best raw performance (average FPS numbers) and frametime stability (1% and 0.1% average low/integral FPS or P1 and P0.2 FPS percentile numbers). Comparative performance analyses of GPU drivers are useful because they allow us to reliably validate and estimate significant improvements or regressions in graphics performance attributable to changes in the driver version.
PC specs are: custom desktop with Windows 10 v22H2 (latest Windows Update patches applied), 16 GB DDR4-3200 RAM (2x8), SSD NVME 1 TB, Intel i7-10700k, ROG Strix GeForce RTX 3070 OC, Single HDC 4K 60 Hz monitor (no HDR or G-Sync). Factory clocks on CPU and GPU. Hardware Accelerated GPU Scheduling (HAGS for short) is disabled. Resizable BAR is disabled. NVCP Global Settings (default).
Frame Times are recorded using PresentMon during the built-in benchmark run inside each game. Each benchmark is run initially four times, and the first result is discarded. Outliers results, with more than 5% variance in any of my used metrics from the average, are also discarded and repeated.
To the methodology, I have decided to add a Non-Synthetic Benchmarks, in this case Unigine Benchmark. For each installation of each subsequent version of a driver, the installation will be performed in a clean option. In case of any problems, DDU will be used for the complete cleanup. When installing each driver, the only option I will not enable will be GeForce Experience.
This initial commentary will be updated as we make more exhaustive analyses with different games.
Also, I apologize if I don't have games of recent date, since I don't have the possibility to buy them at the moment due to the economic situation in my country Argentina.
Using the default settings everything in Ultra with Dx12. Rendered at 100% in 2K resolution. No scaling. Ray Tracing Off. Screen Space Reflections high. Volumetric Lighting no.
The Riftbreaker - driver 537.58 on Win10 v22H2 (considered stable by the community):
- Avg. FPS: 166.7 / 166.5 / 166.8
- 1% Low FPS: 144.2 / 140.3 / 142
- 0.1% Low FPS: 139.1 / 106.4 / 122.7
- Frametimes (ms): Avg. 6.0 | Low 1% 7.0 | Low 0.1% 7.8
The Riftbreaker - driver 552.22 on Win10 v22H2:
- Avg. FPS: 165 / 164.1 / 164.9
- 1% Low FPS: 135.5 / 133.5 / 133.1
- 0.1% Low FPS: 76.8 / 74.8 / 74.9
- Frametimes (ms): Avg. 6.1 | Low 1% 7.5 | Low 0.1% 13.2
The Riftbreaker - driver 552.44 on Win10 v22H2:
- Avg. FPS: 166.1 / 165.1 / 166.9
- 1% Low FPS: 136.5 / 135.5 / 136.1
- 0.1% Low FPS: 77.8 / 75.8 / 75.9
- Frametimes (ms): Avg. 6.0 | Low 1% 7.1 | Low 0.1% 12.2
Using the default settings everything in Ultra. Rendered at 100% in 4K resolution.
AoE2:D.E. - driver 537.58 on Win10 v22H2 (considered stable by the community):
- Avg. FPS: 57.4 / 57.2 / 57.7
- 1% Low FPS: 37.1 / 37.1 / 37.1
- 0.1% Low FPS: 33.2 / 33.6 / 33.1
- Frametimes (ms): Avg. 17.4 | Low 1% 26.9 | Low 0.1% 30.1
AoE2:D.E. - driver 552.22 on Win10 v22H2:
- Avg. FPS: 55.6 / 54.9 / 55.4
- 1% Low FPS: 36.9 / 36.4 / 36.3
- 0.1% Low FPS: 32.5 / 33 / 32.1
- Frametimes (ms): Avg. 18.1 | Low 1% 27.4 | Low 0.1% 30.6
AoE2:D.E. - driver 552.44 on Win10 v22H2:
- Avg. FPS: 55.7 / 54.9 / 55.2
- 1% Low FPS: 37.1 / 36.3 / 36.2
- 0.1% Low FPS: 32.5 / 33.2 / 32.4
- Frametimes (ms): Avg. 18.2 | Low 1% 27.5 | Low 0.1% 30.5
Render: Direct3D11 - Mode: 3840x2160 fullscreen - Preset Custom - Quality Ultra - Tessellation: Disabled.
UHB 4.0 - driver 537.58 on Win10 v22H2 (considered stable by the community):
- Avg. FPS: 80.7
- Score: 2033
- Min FPS: 35.8
- Max FPS: 142.8
UHB 4.0 - driver 552.22 on Win10 v22H2:
- Avg. FPS: 80.1
- Score: 2018
- Min FPS: 8.8
- Max FPS: 142.4
UHB 4.0 - driver 552.44 on Win10 v22H2:
- Avg. FPS: 80.2
- Score: 2020
- Min FPS: 10.2
- Max FPS: 142.1
Render: Direct3D11 - Mode: 3840x2160 fullscreen - Preset Custom - Shaders 4K Optimized - Textures: High.
USB - driver 537.58 on Win10 v22H2 (considered stable by the community):
- Avg. FPS: 94.41
- Score: 12622
- Min FPS: 78.81
- Max FPS: 117.92
USB - driver 551.86 on Win10 v22H2:
- Avg. FPS: 92.47
- Score: 12363
- Min FPS: 55.96
- Max FPS: 119.95
USB - driver 552.22 on Win10 v22H2:
- Avg. FPS: 94.53
- Score: 12638
- Min FPS: 9.04
- Max FPS: 117.65
USB - driver 552.44 on Win10 v22H2:
- Avg. FPS: 94.48
- Score: 12635
- Min FPS: 78.04
- Max FPS: 117.55
At the latest, by tomorrow I will have some early results of the analysis of this driver, greetings to all!
UPDATE: as we can see, this new driver release has fortunately not degraded the experience in any way, but has kept it within the margins obtained in previous updates.
Unless you play very recent games, it is advisable to wait for a future version of the driver in which there may be more bugs fixed because in my opinion it is not worth upgrading at the moment, considering that the previous version is in the same margins.
CHANGES IN THE FUTURE:
- 1- it is possible that in the next few weeks I will have another workbench to test with Intel 14th (so that I can include people who have this type of more modern cpu).
- 2- I will restructure the information of the next drivers to make it more didactic for everyone and summarize some redundant information.
- 3- thanks to all who support these follow-ups, especially to m_w_h for his great work with us.
- 4- this time, I ask for help to those of you who read this, if you can leave me in the comments which games I should include in the tests, it is possible that the comment with more votes above (to avoid writing the same) will be taken into account.
