Thief

CPU Role — cores, clocks, and how they relate to this game's results

The AMD Ryzen 9 5950X is a 16-core, 32-thread processor based on the Zen 3 (Vermeer) architecture, built on TSMC's 7 nm process. It carries a base clock of 3.40 GHz and a boost clock of 4.90 GHz, with a 105 W TDP. This is a high-end desktop part from the 5000 series, released in November 2020, and it remains an active production product. Its benchmark profile shows strong multi-threaded performance: Cinebench R23 multicore scores 26017, 3DMark max threads scores 11597, and PassMark multithread scores 45424. Single-thread performance is also respectable, with Cinebench R23 singlecore at 1614 and Geekbench singlecore at 2083.

The game's measured FPS results, however, suggest that the CPU is not the primary constraint. For a stealth title from 2014, the Ryzen 9 5950X provides far more computational headroom than the workload demands. The CPU's 91st percentile ranking among all CPUs, and an average benchmark score of 51947, place it in the top tier of processors. Its nearest rivals include the Intel Core Ultra 5 235HX (average score 52073, deltaPct -0.2), the AMD EPYC 8124P (52121, -0.3), and the Intel Core i9-13900F (51730, 0.4). The 5950X is essentially performance-equivalent to these parts, with deltas under one percent in either direction.

In the context of Thief, the CPU's role is to feed the GPU with draw calls and maintain frame pacing. The data shows that at 1080p Low, the combo achieves 78 FPS average, which is the highest measured result across all settings and resolutions. The progression from Low to Ultra at each resolution shows a consistent FPS decline that closely tracks GPU load rather than CPU capability. The CPU's 64 MB of L3 cache and dual-channel DDR4 memory support (51.2 GB/s bandwidth) are more than adequate for a 2014 game. The 3DMark 16-thread score of 10810 and 8-thread score of 6598 indicate that even a fraction of the CPU's threads could handle this game's logic without breaking a sweat.

The single-thread performance, measured at 3DMark single thread 944 and PassMark single thread 3470, is particularly relevant for older game engines that rely on a primary simulation thread. Thief's engine, while not modern, still benefits from strong single-core throughput. The 5950X delivers this in abundance. The lack of integrated graphics on this CPU is also notable — the system relies entirely on the discrete GPU for rendering, which is the Intel Arc B370 in this configuration.

The CPU's role in this specific combo, therefore, is to ensure that the GPU is never starved of data. Given that the GPU is an integrated-class part (more on that below), the CPU is vastly overqualified for the task. The benchmark data confirms this: the FPS numbers are low across the board, but they scale with resolution and settings in a way that points to GPU limitations, not CPU bottlenecks.

GPU Role — VRAM, clocks, and how they relate to this game's results

The Intel Arc B370 is an integrated GPU based on the Xe3-LPG architecture, part of the Arc Graphics-M (Panther Lake) generation. It is built on Intel's 3 nm process and has a base clock of 300 MHz with a boost clock of 2400 MHz. The GPU features 1280 shading units, 40 texture mapping units, and 20 raster output units, along with 10 ray tracing cores. Its pixel rate is 48.00 GPixel/s and texture rate is 96.00 GTexel/s, with FP32 performance of 6.144 TFLOPS.

Critically, the Arc B370 uses System Shared memory — there is no dedicated VRAM. The memory type, bus width, and bandwidth are all listed as "System Shared" or "System Dependent." This is a fundamental limitation for gaming performance, as the GPU must contend with the CPU for memory bandwidth. The TDP is just 25 W, and the slot width is IGP, confirming its integrated nature. The GPU's benchmark results are sparse — only a single 3DMark Steel Nomad DX12 score of 1184.

This score places the Arc B370 in the 5th percentile of all GPUs, with an average benchmark score of 1184. Its nearest rivals are all older, low-end parts: the ATI Mobility Radeon HD 5570 (1186, deltaPct -0.2), ATI Radeon HD 5770 (1190, -0.5), AMD Radeon HD 7650M (1192, -0.7), and AMD FirePro M2000 (1168, 1.4). The Arc B370 is essentially on par with a decade-old discrete GPU, which explains the modest FPS figures in Thief.

The GPU's role in this combo is to render frames, and the data shows it struggles even at 1080p. At 1080p Low, the combo hits 78 FPS, but dropping to High reduces that to 45 FPS, and Ultra to 43 FPS. At 1440p, the numbers fall further, and at 4K, they become largely unplayable. The shared memory architecture means that the GPU's performance is directly tied to system memory bandwidth, which is system-dependent and not a fixed spec. The lack of dedicated VRAM is particularly problematic for higher resolutions, where memory pressure increases.

The GPU's boost clock of 2400 MHz is respectable for an integrated part, but the 1280 shading units and 20 ROPs are simply too few to handle the pixel fill requirements of 4K. The pixel rate of 48.00 GPixel/s sounds adequate on paper but proves insufficient in practice, as the measured 4K results show. The GPU is the clear bottleneck in this system, as evidenced by the FPS scaling across resolutions and settings.

Resolution Scaling — how FPS drops from 1080p to 4K and what it says about the limiting component

The measured FPS data reveals a clear pattern of GPU-limited performance. Moving from 1080p to 4K at each settings level shows substantial FPS drops, which is the hallmark of a GPU bottleneck. The CPU, as established, is more than capable of handling the workload; the GPU is the limiting factor.

At Low settings, the average FPS drops from 78 at 1080p to 53 at 1440p, then to 31 at 4K. That is a 32% drop from 1080p to 1440p, and a further 41% drop from 1440p to 4K. The total drop from 1080p to 4K is 60%, from 78 FPS to 31 FPS. This steep decline indicates that the GPU is being overwhelmed by the increasing pixel count.

At Medium settings, the pattern is similar: 59 FPS at 1080p, 41 FPS at 1440p, and 24 FPS at 4K. The drop from 1080p to 1440p is 31%, and from 1440p to 4K is 41%, closely mirroring the Low settings scaling. This consistency suggests that the GPU's fill rate is the primary constraint, rather than memory bandwidth or texture fetch.

At High settings, the numbers are 45 FPS at 1080p, 31 FPS at 1440p, and 18 FPS at 4K. The percentage drops are 31% and 42%, respectively. At Ultra settings, the figures are 43 FPS at 1080p, 29 FPS at 1440p, and 17 FPS at 4K, with drops of 33% and 41%.

The near-uniform scaling across all settings levels is telling. If the CPU were the bottleneck, we would expect the FPS to remain relatively flat across resolutions, especially at lower settings where the CPU workload is proportionally higher. Instead, the FPS drops consistently by roughly one-third per resolution step, which is classic GPU-bound behavior. The gap between Low and Ultra at each resolution is also informative: at 1080p, the difference between Low (78 FPS) and Ultra (43 FPS) is 35 FPS, or 45%. At 4K, the difference between Low (31 FPS) and Ultra (17 FPS) is 14 FPS, or 45%. This consistent relative gap indicates that the GPU's shader and texture units are the limiting factor, not the CPU or memory subsystem.

The data also shows that even at 1080p Low, the combo cannot reach 80 FPS, which is a modest target for a 2014 game. This is a direct consequence of the Arc B370's performance tier, sitting at the 5th percentile of all GPUs.

FAQ

Q: Why does the combo achieve such low FPS in Thief despite having a high-end CPU?

A: The Intel Arc B370 is an integrated GPU with System Shared memory, placing it in the 5th percentile of all GPUs. Its 3DMark Steel Nomad score of 1184 is comparable to decade-old discrete GPUs like the ATI Radeon HD 5770 (1190). The CPU is not the constraint; the GPU's limited shading units (1280) and ROPs (20) cannot keep up with the rendering load.

Q: Is the Ryzen 9 5950X a good match for the Arc B370?

A: The Ryzen 9 5950X is vastly overqualified for this GPU. The CPU's average benchmark score of 51947 places it in the 91st percentile, while the GPU's score of 1184 places it in the 5th percentile. The CPU is not a bottleneck in any of the measured results; the GPU holds the system back.

Q: What is the best resolution to play Thief with this combo?

A: The only resolution that yields playable FPS is 1080p. At Low settings, the combo averages 78 FPS, and at Medium, it averages 59 FPS. At 1440p, the best result is 53 FPS on Low, and at 4K, the best result is 31 FPS on Low, which is below typical playability thresholds.

Q: How does the Arc B370 compare to its nearest GPU rivals?

A: The Arc B370's average benchmark score of 1184 is nearly identical to the ATI Mobility Radeon HD 5570 (1186, deltaPct -0.2), the ATI Radeon HD 5770 (1190, -0.5), and the AMD Radeon HD 7650M (1192, -0.7). It is slightly ahead of the AMD FirePro M2000 (1168, 1.4). All of these are older, low-end GPUs.

Q: Does the CPU's single-thread performance matter for Thief?

A: Yes, but the Ryzen 9 5950X has ample single-thread performance, with a Cinebench R23 singlecore score of 1614 and a 3DMark single thread score of 944. Even at 1080p Low, where CPU load is relatively highest, the FPS is limited by the GPU, not the CPU's single-thread capability.

Q: What settings are playable at 1440p?

A: At 1440p Low, the combo averages 53 FPS, which is marginally playable. At Medium, it drops to 41 FPS, and at High, it falls to 31 FPS. Ultra is 29 FPS. For smoother gameplay, 1080p Medium (59 FPS) is a better choice than 1440p Low (53 FPS).

How This Combo Ranks

The combination of the AMD Ryzen 9 5950X and Intel Arc B370 ranks 2634 out of 2851 tested combos in Thief. This places it in the bottom 8% of all configurations tested. The ranking reflects the GPU's severe performance limitation, not the CPU's capability. The CPU alone ranks in the 91st percentile, but the GPU's 5th percentile ranking drags the entire combo down.

The rank of 2634 out of 2851 means that 2633 other CPU-GPU combinations delivered higher average FPS in Thief. This is a poor result, but it is entirely attributable to the integrated GPU. The combo's position in the rankings would be dramatically different if paired with a discrete GPU that has dedicated VRAM and higher throughput. The data shows a stark contrast between the CPU's top-tier standing and the GPU's bottom-tier standing, and the overall combo rank reflects the weaker component.

The delta between the CPU's nearest rivals (all within 0.7% of its average score) and the GPU's nearest rivals (all within 1.4% of its average score) shows that both components are at the extremes of their respective performance spectra. The CPU is at the top, and the GPU is at the bottom. This mismatch is the defining characteristic of the combo's performance profile.

Settings Recommendations

Given the measured data, the optimal settings for this combo depend on the target resolution and acceptable frame rate. The data provides clear guidance.

At 1080p, the best experience is achieved at Low settings, which yields 78 FPS average. This is the only setting that approaches high refresh rates. Medium settings at 1080p deliver 59 FPS, which is still smooth for most players and offers better visual quality. High settings drop to 45 FPS, and Ultra to 43 FPS. The difference between High and Ultra is minimal (2 FPS), so Ultra is not worth the potential visual improvement over High.

At 1440p, Low settings are the only option that maintains playability, with 53 FPS. Medium drops to 41 FPS, which may be acceptable for some, but it is a 23% reduction from Low. High (31 FPS) and Ultra (29 FPS) are not recommended for a smooth experience.

At 4K, no settings level achieves playable frame rates. The best result is 31 FPS on Low, and the worst is 17 FPS on Ultra. 4K is not a viable resolution for this combo in Thief.

The data suggests that 1080p Low is the sweet spot for maximum smoothness, while 1080p Medium offers a balanced trade-off between visual quality and performance. For those who prioritize visual fidelity over frame rate, 1080p High at 45 FPS is the highest settings level that remains semi-playable. The measured FPS at 1080p Medium (59 FPS) with min FPS of 50 and max of 68 indicates consistent frame delivery, making it the recommended choice for a good experience.

Measured FPS Breakdown

The following is a complete breakdown of the measured average FPS (and where available, min and max FPS) for each resolution and settings level.

1920x1080 (1080p):

• Low: 78 FPS average
• Medium: 59 FPS average (min 50, max 68)
• High: 45 FPS average
• Ultra: 43 FPS average

2560x1440 (1440p):

• Low: 53 FPS average
• Medium: 41 FPS average (min 35, max 47)
• High: 31 FPS average
• Ultra: 29 FPS average

3840x2160 (4K):

• Low: 31 FPS average
• Medium: 24 FPS average (min 20, max 27)
• High: 18 FPS average
• Ultra: 17 FPS average

The data shows a monotonic decrease in FPS as both resolution and settings increase. The only settings levels with min/max data are Medium at each resolution, indicating that the benchmark captured frame variability for those runs. At 1080p Medium, the frame time consistency is reasonable (50-68 FPS range). At 1440p Medium, the range is 35-47, and at 4K Medium, it is 20-27. The lack of min/max data for Low, High, and Ultra means those runs may have been less variable, but the average scores alone are sufficient to draw conclusions.

The gap between Low and Ultra at each resolution is consistent: 35 FPS at 1080p, 24 FPS at 1440p, and 14 FPS at 4K. This consistent delta suggests that the GPU's settings-dependent workload scales predictably. The 4K Ultra result of 17 FPS is the lowest measured, and the 1080p Low result of 78 FPS is the highest. The full range of measured averages spans from 17 to 78 FPS, a 61 FPS spread across all tested configurations.