NVIDIA · DRIVER PERFORMANCE

NVIDIA driver results.

Every driver we test, in one place. See the FPS and frame times recorded for each version.

Counter-Strike 2PresentMonResults by test date

Controlled test conditions: Windows settings, game CFG files and user-configured NVIDIA settings are kept the same across driver tests. We change the driver version, then repeat the same benchmark on the same hardware. Any driver-default changes, game updates or unavoidable differences are documented in the session notes.

NVIDIA 621.01

Tested 2026-10-23

Counter-Strike 2 · FPSHeaven Ancient / de_ancient

Average FPS

309.03FPS

Higher is betterTest range 303.43–312.58

1% low FPS

114.13FPS

Higher is betterTest range 98.58–125.49

p99 frame time

5.83ms

Lower is betterTest range 5.69–6.09

Median of 4 measured tests1440 × 1080 · 144 HzNVIDIA GeForce RTX 4060

Individual measurements

PresentMon · fixed 60-second windows · warmup excluded
Test	Average FPS	1% low FPS	p99 ms
1	312.58	125.49	5.69
2	307.46	105.80	5.83
3	310.61	122.46	5.83
4	303.43	98.58	6.09

Separate VProf medians: 485.55 average FPS / 192.90 P1 FPS. VProf covers the full game benchmark interval and uses different definitions from PresentMon.

Test setup & session notes

GPU: NVIDIA GeForce RTX 4060
CPU: AMD Ryzen 5 7600X3D
Display: 1440 × 1080 · 144 Hz
Game: Counter-Strike 2 · build 26144732
Scene: FPSHeaven Ancient / de_ancient
Protocol: fixed60-offset10-v1

One warmup per driver was excluded; four measured tests per driver. All baseline tests preceded the new-driver tests; this was not an interleaved comparison.
The new driver was installed without a system restart.
Saved CFG hashes, Steam overlay state (off), game build and captured Windows conditions matched. Driver-defined defaults changed; this is a driver-package comparison, not a claim that every internal driver setting was identical.
A new predefined DRS entry (548384728 = 1) was left untouched. NVIDIA Ansel policy 271965065 was 0 in both groups.
Display alias changed from DISPLAY1 to DISPLAY6; resolution, refresh rate, position and orientation matched.
The second group used a bounded NVML shutdown fix; sensor cadence and the measured PresentMon window were unchanged.
Both sessions completed with restoration verified and no cleanup errors. Four tests on one system do not establish statistical certainty, live-match performance or physical input latency.

NVIDIA 617.42

Tested 2026-10-23

Counter-Strike 2 · FPSHeaven Ancient / de_ancient

Average FPS

308.12FPS

Higher is betterTest range 306.86–308.81

1% low FPS

118.77FPS

Higher is betterTest range 107.06–138.40

p99 frame time

5.74ms

Lower is betterTest range 5.64–5.89

Median of 4 measured tests1440 × 1080 · 144 HzNVIDIA GeForce RTX 4060

Individual measurements

PresentMon · fixed 60-second windows · warmup excluded
Test	Average FPS	1% low FPS	p99 ms
1	308.81	138.40	5.64
2	307.76	113.58	5.83
3	308.48	123.97	5.65
4	306.86	107.06	5.89

Separate VProf medians: 484.00 average FPS / 194.25 P1 FPS. VProf covers the full game benchmark interval and uses different definitions from PresentMon.

Test setup & session notes

GPU: NVIDIA GeForce RTX 4060
CPU: AMD Ryzen 5 7600X3D
Display: 1440 × 1080 · 144 Hz
Game: Counter-Strike 2 · build 26144732
Scene: FPSHeaven Ancient / de_ancient
Protocol: fixed60-offset10-v1

One warmup per driver was excluded; four measured tests per driver. All baseline tests preceded the new-driver tests; this was not an interleaved comparison.
The new driver was installed without a system restart.
Saved CFG hashes, Steam overlay state (off), game build and captured Windows conditions matched. Driver-defined defaults changed; this is a driver-package comparison, not a claim that every internal driver setting was identical.
A new predefined DRS entry (548384728 = 1) was left untouched. NVIDIA Ansel policy 271965065 was 0 in both groups.
Display alias changed from DISPLAY1 to DISPLAY6; resolution, refresh rate, position and orientation matched.
The second group used a bounded NVML shutdown fix; sensor cadence and the measured PresentMon window were unchanged.
Both sessions completed with restoration verified and no cleanup errors. Four tests on one system do not establish statistical certainty, live-match performance or physical input latency.

How to read these results

Download data JSON ↓

Each headline value is the median of the measured tests for that driver. The range shows the lowest and highest observed result. Higher FPS is better; lower p99 frame time is better. Small differences within overlapping ranges do not establish a reliable performance change.

PresentMon captures a fixed 60-second window starting 10 seconds after the observed VProf start. Average FPS = 1,000 ÷ mean consecutive CPU frame-start interval in ms. 1% low = 1,000 ÷ mean of the slowest ceil(1%) intervals. p99 uses the nearest-rank 99th percentile interval.

FPS and frame times do not measure physical input latency. Results on one system and a workshop scene do not guarantee the same performance in live matches or on other PCs. Downloads contain derived results and method notes, without private paths, account identifiers or raw configuration files.