Thief

This entry covers the measured performance of Thief (2024? Actually, the game’s release date is 2014-02-24) running on an AMD Ryzen 9 7900X with an AMD Radeon RX 6700 XT. The livestream is flagged as `dataIsMeasured`, so every FPS figure comes from an actual benchmark run rather than estimation. The combination ranks 1802 out of 3720 combinations for this game. The dataset provides average FPS at three resolutions (1920x1080, 2560x1440, 3840x2160) and four quality presets (Low, Medium, High, Ultra), with a full min/max range available for the Medium settings at each resolution. The results show a wide spread: from 186 FPS at 1080p/Low down to 41 FPS at 4K/Ultra. This range, combined with the CPU’s strong benchmark profile and the GPU’s Mid ‑range standing, tells a clear story about where the system’s limits sit in this stealth classic.

FAQ

Q: What is the fastest frame rate recorded for this combination?

A: The highest average FPS in the measured set is 186 frames per second, which occurs at 1920x1080 with the Low graphics preset. The lowest average is 41 FPS, at 3840x2160 with the Ultra preset.

Q: Does the CPU have many cores?

A: Yes. The AMD Ryzen 9 7900X is a 12‑core, 24‑thread processor with a base clock of 4.70 GHz and a boost clock of 5.60 GHz. It has 13,140 million transistors and a 64 MB shared L3 cache, and its benchmark percentile places it among the top 91% of all CPUs.

Q: How much video memory does the RX 6700 XT have?

A: The card brings 12 GB of GDDR6 memory on a 192‑bit bus, giving a memory bandwidth of 384.0 GB/s. That configuration is listed in the hardware data for the card, though the frame‑rate measurements in Thief do not include VRAM usage information.

Q: Which resolution should be used for the highest average FPS on this machine?

A: According to the measured data, the best performer is 1920x1080. Even at the Ultra preset there, the average FPS is 103, while the Low preset achieves 186 FPS on that resolution. The two higher resolutions (2560x1440 and 3840x2160) are always slower.

Q: Does the game actually use the CPU’s and GPU’s compute resources?

A: The FPS scaling strongly suggests that the GPU is the bottleneck at high resolution. For example, the CPU’s multi‑core benchmark (Cinebench R23) scores 29300, while the GPU’s PassMark score is 19786; yet in Thief the frame rate falls from 186 FPS at 1080p/Low to 75 FPS at 4K/Low, a drop we attribute to raster workload because the CPU has plenty of headroom in its low‑settings slider testing.

Settings Recommendations

Deciding on the right preset depends heavily on your display’s native resolution. The measured data gives a clear device at each resolution.

At 3840x2160, the Low preset averages 75 FPS, which is the highest number available at that resolution. Medium sits at 57 FPS, and both High and Ultra fall under 50, at 43 and 41 FPS respectively. For a 4K display, the difference between High and Ultra is only 2 FPS (43 vs 41), so choosing Ultra over High does not give a measurable framerate up bonus. If smoothness is the goal, Low is the only setting that clearly distances itself from the others. Medium offers a decent single step up from Low with 12 Fps less capacity *.

At 2560x1440, the Low preset still yields a high 128 FPS. Medium drops to 97, but the min/max range for Medium shows 82 to 112 FPS. The High preset averages 74 FPS, and Ultra is just under at 70 FPS. For 1440p the most contended pick is Medium parameters: it offers a ~70 FPS improvement over High and a far narrower variation (82–112 vs the 74 average of the High). Low is there if you want to hit more than 120 FPS average.

At 1920x1080, the story turns. Every setting manages over 100 FPS. Low is 186, Medium 141, High 108, and Ultra 103. The Medium core, with a min of 120 and max of 163, shows a stable range. Because even the Ultra preset averages above 100 FPS, the resolution is essentially never limits the experience at this resolution. For a 1080p display, users can pick High or Ultra freely, sharing the 5% between those two. The only reason to go lower than High is the very high in‑order frame time at Ultra (min values only for Medium, but the difference in av res likely covers the full range).

Practically: for a 4K panel or TV, set the quality to Medium or Low to hold a playable frame rate; for 1440p, Medium is your sweet spot with High following closely; and at 1080p there is no penalty for cranking Ultra.

Resolution Scaling

The measured FPS across the three resolutions gives insight into the card’s ability to handle high resolution data about the game environment. For a given quality preset, the frame rate always decreases as the resolution goes from 1920x1080 to 2560x1440 and further to 3840x2160.

Take the High preset. Starting at 1080p with an average of 108 FPS, moving the game to 1440p reduces the average to 74 FPS. Moving to 2160p reduces it all the way to 43 FPS. The same lane appears in Low: 186 to 128 to 75. Medium: 141 to 97 to 57. Ultra: 103 to 70 to 41.

These numbers show that the percentage the GPU has to perform at higher resolution is significant. For example, the current jump from 1080p to 4K at the High preset brings the average from 108 to 43; this is a drop of 65 FPS on the average surface. The 1080p to 1440p step likewise costs about 30 FPS, not 42% as some might expect (we cannot put it). The effect is consistent across all four quality presets, which indicates that the GPU is indeed the primary limiter when resolution climbs. A CPU wearing an average Core Mat equally would not hit a wall at 4K because the engine’s threading is already saturated at 1080p.

The scaling pattern reveals that Thief’s graphical load is dominated by pixel count and fill‑rate. At 1080p the card is less active and often gives high headroom (the 186 FPS ceiling). But as the pixel count grows, the rasterizer and shader complex become the limiting factor, and the CPU has plenty of spare core count.

This type of scaling is exactly what we expect from a 2014 stealth game running on a modern GPU. The resolution jump in the dataset demonstrates that Thief does not simply saturate the render pipeline at 1440p; it reaches a true UHD pixel count where the geometry take it from 41 FPS at Ultra — the least playable number in the table.

Measured FPS Breakdown

The full set of measured frame‑rate numbers for this combo is listed below. All are average FPS (Avg) and, where available, the minimum/maximum frame rates (Min/Max) for the Medium preset.

1920x1080

• Low: Avg 186
• Medium: Avg 141 (Min 120, Max 163)
• High: Avg 108
• Ultra: Avg 103

2560x1440

• Low: Avg 128
• Medium: Avg 97 (Min 82, Max 112)
• High: Avg 74
• Ultra: Avg 70

3840x2160

• Low: Avg 75
• Medium: Avg 57 (Min 48, Max 66)
• High: Avg 43
• Ultra: Avg 41

The pattern is monotonic at every resolution, confirming that more render quality is purchased with frame rate. The Medium preset is the only one for which the full frame‑rate minus (min and max) is recorded; for the other presets, only the average is provided in the data set.

The 4K/Ultra number (41) is the only instance where the average goes below 50 FPS. All other settings all keep at least 57 FPS or more, which some players may prefer for a smoother run. The difference between High and Ultra at 4K with a PC is just 2 FPS (43 vs 41), so presetting beyond the High hardly matters even in the highest‑load state.

At 1440p, the Ultra preset acts at an average 70 FPS, over the 60‑positive threshold that most gamers consider. This means the GPU can handle 1440p Ultra in the game without very deliberat. At 1080p, the Ultra preset reaches 103 FPS – more than enough for any high‑refresh 1080p monitor.

GPU Role

The Radeon RX 6700 XT uses the Navi 22 chip with 2,560 shading units and an RDNA 2.0 architecture. The active clock rise from 2321 MHz base to 2581 MHz boost speed, and the memory side runs at effective 16 Gbps across a 192‑bit bus, producing 384.0 GB/s of bandwidth.

This GPU sits in the 72nd percentile of all GPUs, and its PassMark G3D score is 19,786. For comparison, the near‑neighboring tables for this GPU list an RTX 3090, which has a deltaPct of -0.5 in the dataset – meaning the 6700 XT is rated 0.5% slower on average score. NVIDIA’s RTX 4070 Mobile shows a deltaPct of 0.0, essentially a running match.

In the Thief benchmark, the GPU shows a 6‑ms, and the effect of the 4K resolution was described above. The card’s performance is not limited by cache capacity; 12 GB of GDDR6 is generous for a 2014 game. The data shows that the GPU’s ability to handle the increased raster work at 4K is the primary bottleneck. At 1440p and 1080p, the card retains high averages; at 2160p, all settings drop below 60 FPS, and only the Low preset goes over 70. The framerate scaling is a uniform shift between quality levels, indicating that the rendering pipeline uses a standard increase in shader complexity and fill‑rate pressure.

The 12 GB buffer and 384.0 GB/s bandwidth are present in the measured results, as the FPS does not tumble due to memory bandwidth at medium and high resolutions, but further as resolution increases – which is typical of an GPU with a wide rasterizer. The GPU’s compute scores (it achieves 10 TT) are secondary, as Thief’s engine is not 2018 ray‑tracing code but plain rasterization secure.

Since the game scales so heavily with resolution, the RX 6700 XT's primary role is to deliver high temporal smoothness at 1440p and below, and a playable experience at 4K only when settings to Low or Medium. It is exactly the kind of card that gives excellent returns for 1080p/1440p gaming but start to twist under UHD load.

CPU Role

The system’s brain is the AMD Ryzen 9 7900X, a 12‑core/24‑thread Zen 4 part. It runs at a base frequency of 4.70 GHz and fixes to a boost clock of 5.60 GHz. The L2 cache is 1 MB per core, the L3 is a ?em??64 shared MB. The measured average benchmark score is 53,288, an 91st percentile ranking among all CPUs.

Its CPU performance is substantial in comparison: Cinebench R23 multi‑core score 29,300, single‑core 2016.5 points, and the Geekbench multi‑core 19,267. The nearest rivals in the record for this CPU are the AMD EPYC 7313P (average score 53,206, deltaPct 0.2) and the Intel Xeon 4228? Actually the data lists Intel Xeon 634 (average 52,974, deltaPct 0.6) and Intel Core i7‑14700F (average 53,620, deltaPct -0.6). The deltas show all are within 0.6% of the 7900X’s result.

In Thief, the CPU role appears to be modest. At 1080p/Low, the game reaches another 186 FPS – a figure that far exceeds any reasonable frame‑rate target and indicates that the game is nowhere near CPU‑limited at the standard definition setting. Even 1440p – all variants remain above 70 FPS, and the only single sub‑50 numbers appear at 4K. This behavior is typical: the game mainly pushes the GPU, not the deep many‑thread cores. A 8‑core/16‑thread CPU could offer a similar result, but the 12‑core does not get in the way to dominate feed the run, especially since the engine (2014) does not massively parallelise such scene.

The 7900X’s strong single‑thread result (2016.5 in Cinebench R23) ensures that scene signal is never hobbled by low per‑core