← sources

NVIDIA 617.14 WHQL Driver - Performance Benchmark (Ampere GPUs)

source: reddit (reddit:nvidia)published 2026-09-30 01:05ingested 2026-10-01 02:43

retained text

Hey all,

Here's my early look at the **617.14** driver, tested on an Ampere-based GPU (RTX 3070). As always: this is not a definitive review. Every PC setup is different, so treat it as a reference point, not gospel.

**Is it worth updating your driver?** 
The general rule says "yes": newer drivers bring bug fixes, security patches, and optimizations for the newest games. But "newer" doesn't automatically mean "better". Some updates introduce subtle stutter or frame pacing regressions that don't show up if you only look at average FPS. That's the main thing this round of testing tries to catch.

*Special thanks to UL Solutions for the 3DMark press license used in these tests.*

**Quick note before anything: updated all my tools this round and re-tested every reference driver, so numbers won't match my previous posts. Details in Methodology.**

---

## Test System

- **OS:** Windows 11 Pro (v25H2, latest patches)
- **CPU:** Intel Core i7-10700K (stock)
- **GPU:** ASUS ROG Strix RTX 3070 OC (stock)
- **RAM:** 64 GB DDR4-3200 (2x32 GB)
- **Storage:** 1TB NVMe SSD
- **Display:** 4K 60 Hz monitor (8 bpc, no HDR, no G-Sync)
- **Settings:** HAGS off | ReBAR on | NVCP Global Defaults
- **Driver Install:** Clean install (with NVCP)

## Benchmark Methodology

- Benchmarks ran 4 times (first run discarded), then the remaining 3 valid runs merged into one aggregated result (outliers get automatically retested). 
- Captured frametimes via **PresentMon** (integrated into CapFrameX).
- Outliers (>5% variance) are retested.
- Alongside Average FPS, 1% Low and 0.1% Low, this round also tracks CapFrameX's **Adaptive STDEV** metric (labeled "Consistency" below). It measures how much frame times wobble compared to a moving average. Lower = smoother frame delivery, even when the average FPS looks identical between two drivers.
- Game settings: mostly **Ultra**, 4K or 1440p, no scaling.
- **Tools updated for this round:** CapFrameX 1.9.1 with PresentMon 2.6.0, plus 3DMark, Superposition and LatencyMon on their latest versions as of Sep 29, 2026. All game results were re-captured on this setup, reference drivers included, so don't compare them against my older posts. AoE2 in particular reads ~10 FPS higher than before because of the tool change, not the drivers. 3DMark and Superposition don't go through CapFrameX, so those are comparable.

> **quick legend, if these terms are new to you:**
> - **Avg FPS** -> your average frame rate.
> - **1% Low** -> the average of your worst 1% of frames. Better indicator of "does it feel smooth" than the average alone.
> - **0.1% Low** -> same idea, but for your worst 0.1% of frames - basically your nastiest stutters.
> - **Consistency** -> lower is smoother. Two drivers can share the same average FPS and still feel very different to play if one has a much higher (worse) consistency score.

---

# Game Benchmarks

### The Riftbreaker
*DX12, Ultra preset, 2560x1440, Ray Tracing off, Screen Space Reflections High, Volumetric Lighting Ultra, Dynamic Vegetation Full*

| Driver | Avg FPS | 1% Low | 0.1% Low | Consistency |
|:-|:-:|:-:|:-:|:-:|
| 581.94 (stable) | 150.9 | 133.2 | 125.7 | 9.0 |
| 596.36 (stable) | 149.9 | 133.8 | 126.8 | 7.5 |
| 610.88 | 149.9 | 133.6 | 127.7 | 5.8 |
| 616.92 | 149.8 | 132.6 | 126.2 | 6.9 |
| 617.14 | 149.9 | 133.5 | 128.1 | 6.0 |

Same ~150 avg on everything. Consistency got better since 581.94 and 617.14 keeps it. 616.92's 6.9 is just a slightly worse run, nothing more.

### Age of Empires II: Definitive Edition
*Ultra preset, 3840x2160 fullscreen, 100% render scale*

| Driver | Avg FPS | 1% Low | 0.1% Low | Consistency |
|:-|:-:|:-:|:-:|:-:|
| 581.94 (stable) | 70.0 | 51.5 | 43.8 | 7.4 |
| 596.36 (stable) | 71.2 | 51.9 | 40.9 | 7.5 |
| 610.88 | 70.4 | 50.5 | 41.5 | 7.6 |
| 616.92 | 70.4 | 51.5 | 41.1 | 7.7 |
| 617.14 | 70.3 | 50.9 | 42.1 | 7.7 |

### Age of Mythology: Retold
*DX12, everything Ultra/Mythic (Texture Quality High, Texture Filtering 16x, Terrain Quality Mythic, Terrain Detail Quality Mythic, Water Quality Mythic, Foliage Quality Full, Shadow Quality Mythic, Dynamic Reflections SSR, Ambient Occlusion Quality SSAO, Effect Quality High, Dynamic Lights enabled, Physics Quality High), 3840x2160 fullscreen, FXAA*

| Driver | Avg FPS | 1% Low | 0.1% Low | Consistency |
|:-|:-:|:-:|:-:|:-:|
| 581.94 (stable) | 48.5 | 36.2 | 30.6 | 2.4 |
| 596.36 (stable) | 48.3 | 35.3 | 29.4 | 2.8 |
| 610.88 | 47.9 | 34.9 | 24.1 | 2.7 |
| 616.92 | 48.8 | 36.8 | 34.2 | 2.2 |
| 617.14 | 48.0 | 36.7 | 34.3 | 2.3 |

Only thing that moved in this whole round: 616.92 and 617.14 both have better lows than 610.88 and older. Small, and 0.1% low at 48 FPS is jumpy, but two drivers in a row showing it isn't nothing.

---

# Non-Synthetic Benchmark

### Unigine Superposition
*Direct3D11, 3840x2160 fullscreen, Custom preset, 4K Optimized shaders, High textures, Depth of Field off, Motion Blur off*

| Driver | Avg FPS | Score | Min FPS | Max FPS |
|:-|:-:|:-:|:-:|:-:|
| 596.36 | 94.63 | 12651 | 78.84 | 117.64 |
| 610.88 | 94.87 | 12684 | 79.20 | 117.87 |
| 616.92 | 94.29 | 12607 | 78.72 | 116.68 |
| 617.14 | 94.38 | 12619 | 78.62 | 117.00 |

---

# 3DMark Benchmarks

### Numbers in parentheses are % difference vs 596.36

| Benchmark | 596.36 | 610.88 | 616.92 | 617.14 |
|:-|:-:|:-:|:-:|:-:|
| Speed Way (DX12U, 1440p)  | 3572 | 3588 (+0.4%) | 3607 (+1.0%) | 3588 (+0.4%) |
| Steel Nomad (DX12, 4K)    | 3186 | 3232 (+1.4%) | 3201 (+0.5%) | 3208 (+0.7%) |
| Steel Nomad (Vulkan, 4K)  | 3536 | 3542 (+0.2%) | 3549 (+0.4%) | 3543 (+0.2%) |
| Fire Strike (DX11, 1080p) | 25235 | 25572 (+1.3%) | 25634 (+1.6%) | 25476 (+1.0%) |

All noise. Same driver re-run moved Speed Way 0.3%, so nothing here is a real difference.

---

##  Sys. Stability & Desktop

- **Hardware Acceleration (Browsers / Discord):** **Solid.** Zero flickering, black screens, or visual anomalies detected after 30 minutes of mixed use (including 4K60 YouTube playback in Firefox with zero dropped frames and fluid page scrolling).
- **DPC Latency (LatencyMon):** **PASSED.** (System remained perfectly in the green. The highest DPC routine execution time was a safe 466 µs from `nvlddmkm.sys`).
- I checked Event Viewer and got zero errors or warnings.

---

## Verdict

617.14 performs the same as 610.88. Nothing I tested moved outside noise, except AoM's lows looking a bit better.

If you're on 616.56, update. I had the black screen flickering on that one (8 bpc) and haven't seen it since the 616.86 hotfix. It also fixes Judgement / Lost Judgement / VF5 R.E.V.O. not launching.

If you're on 610.88 or older, no rush unless you want Game Ready for the new games.

document fetched 2026-10-01 00:39 · canonical https://reddit.com/r/nvidia/comments/1wtrm3f/nvidia_61714_whql_driver_performance_benchmark