Thief

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

The AMD Ryzen 5 8400F is a 6-core, 12-thread desktop processor built on the Zen 4 (Phoenix) architecture using a 4 nm process from TSMC. It runs at a base clock of 4.20 GHz and boosts to 4.70 GHz, with a 65 W TDP and support for DDR5 memory over a dual-channel bus delivering 83.2 GB/s. The chip carries 64 KB L1 cache per core, 1 MB L2 per core, and 16 MB shared L3 cache. In synthetic benchmarks, the CPU scores 951 in 3DMark single-thread, 1874 in 2-thread, 3563 in 4-thread, 5275 in 8-thread, 6091 in 16-thread, and 6165 in max-thread tests. Cinebench R23 results show 2943 single-core and 20851 multi-core points. PassMark multithread score is 24389, with single-thread at 3685.

For a game like Thief, released in February 2014, the CPU workload is far less demanding than modern titles. The Ryzen 5 8400F's 6 cores and 12 threads are more than sufficient to feed the GPU at any tested resolution. The measured frame rates across all settings and resolutions show that the bottleneck shifts to the graphics card as resolution increases, not the CPU. At 1920x1080 with Low settings, the combo reaches 171 FPS average; at 4K Ultra it drops to 38 FPS. This pattern is classic GPU-limited behavior — the CPU is not the constraint.

The processor sits at the 77th percentile among all CPUs, with an average benchmark score of 25005. Its nearest rivals are the AMD Ryzen 5 7500F (delta +0.2%), Intel Core i7-11850H (+0.3%), Intel Core i7-13620H (+0.4%), and AMD EPYC 9474F (-0.4%). In Thief's measured data, the CPU's single-thread performance — 2943 in Cinebench R23 single-core — is more than adequate for a 2014 stealth title that was designed for much older hardware. The game's engine does not require the full thread count, but the 8400F's high boost clock ensures consistent frame pacing.

The CPU's 16 MB L3 cache is shared across all cores, which helps with the game's asset streaming, though Thief's environment data is modest by modern standards. The 4 nm process and 65 W TDP indicate efficient power delivery, but no power figures are measured in this dataset. The chip was released in March 2024 with a launch MSRP of $170. It uses AMD Socket AM5, PCIe Gen 4 with 20 lanes (CPU only), and has no integrated graphics — the system relies entirely on the discrete GPU.

Benchmark results indicate that at 1080p Low, the frame rate nearly doubles from the High preset (171 vs 99 FPS), which points to the GPU being the limiting factor even at lower resolutions. The CPU's 6-core/12-thread configuration is not a limiting factor in any measured scenario; the delta between Low and Ultra at each resolution scales consistently with GPU load.

In terms of the 3DMark multi-thread scaling, the jump from 8-thread (5275) to 16-thread (6091) is modest, suggesting that the chip's performance does not degrade under heavier thread loads, but Thief rarely uses more than a few threads. The single-thread score of 951 in 3DMark and 3685 in PassMark single-thread are the more relevant metrics for this game's frame pacing. The CPU's percentile rank of 77 indicates it outperforms the majority of processors in the database, which aligns with the observed ability to maintain high FPS in a last-generation title.

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

The NVIDIA GeForce RTX 2070 SUPER is built on the Turing architecture (TU104 chip) using a 12 nm process from TSMC, with 13,600 million transistors on a 545 mm² die. It has a base clock of 1605 MHz and a boost clock of 1770 MHz, with memory running at 1750 MHz (14 Gbps effective). The card features 8 GB of GDDR6 memory on a 256-bit bus, delivering 448.0 GB/s of bandwidth. It has 2560 shading units, 160 texture mapping units, 64 ROPs, 40 RT cores, and 320 tensor cores. Pixel rate is 113.3 GPixel/s, texture rate is 283.2 GTexel/s, and FP32 compute is 9.062 TFLOPS (FP16 at 18.12 TFLOPS with 2:1 ratio). The card's TDP is 215 W, requiring a 550 W suggested PSU, with dual-slot cooling and 1x 6-pin + 1x 8-pin power connectors. It uses PCIe 3.0 x16 and was released in July 2019 with a launch MSRP of 499 USD.

For Thief, the RTX 2070 SUPER's 8 GB VRAM is more than sufficient — the game's textures are from 2014 and do not approach even 4 GB usage at 4K. The memory bandwidth of 448 GB/s ensures that texture streaming is not a bottleneck. The measured FPS shows a clear trend: at 3840x2160, the average FPS drops from 69 (Low) to 38 (Ultra), a 45% reduction. At 2560x1440, the range is 118 (Low) to 65 (Ultra), and at 1920x1080, it goes from 171 (Low) to 94 (Ultra). This scaling is typical for a GPU-bound scenario — the rendering load increases with resolution and quality settings, and the card's compute resources are the limiting factor.

The GPU's percentile rank is 65 among all GPUs, with an average benchmark score of 20282. Its nearest rivals are the NVIDIA Quadro M4000M (delta -1%), NVIDIA GeForce RTX 3070 Mobile (-1.2%), Intel Arc B570 (-1.3%), and Intel Arc A750 (-1.5%). In Thief, the card's 9.062 TFLOPS of FP32 compute is ample for a 2014 stealth game that was designed for DirectX 11-era hardware. The card supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4, but Thief's engine likely runs on older API paths, so these modern features are not fully utilized.

The 40 RT cores and 320 tensor cores are irrelevant for Thief, as the game does not use ray tracing or DLSS. The card's 160 TMUs and 64 ROPs determine texture fill and pixel output; at 4K Ultra, the 38 FPS average suggests the pixel fill rate is the limiting factor. The 113.3 GPixel/s pixel rate is the theoretical ceiling, and at 4K (8.3 million pixels), the engine needs to fill that many pixels per frame — the measured 38 FPS aligns with the card's raw fill capability.

The GPU's memory clock of 1750 MHz (14 Gbps effective) and 256-bit bus produce 448 GB/s bandwidth. For Thief, which uses moderate texture sizes, this bandwidth is not saturated, as evidenced by the fact that lowering texture quality (from High to Low) at 4K boosts FPS from 40 to 69 — a 72% improvement, which is more about shading and fill than memory.

The RTX 2070 SUPER's end-of-life production status and 12 nm process node indicate it is an older card, but for a 2014 game, it delivers playable frame rates at all tested settings except 4K Ultra, where the average dips to 38 FPS. The card's DirectX 12 Ultimate support is forward-looking, but Thief's engine likely uses DirectX 11 or earlier, so the card's full feature set is not exercised.

FAQ

Q: Can this combo run Thief at 4K with playable frame rates?

A: Yes, but only at lower settings. At 3840x2160 with Low settings, the average FPS is 69, which is smooth. Medium drops to 52 FPS (min 45, max 60), High to 40 FPS, and Ultra to 38 FPS. For a 60 FPS target, Low is the only 4K setting that consistently exceeds it; Medium is borderline with a 45 FPS minimum.

Q: How does the Ryzen 5 8400F compare to its nearest CPU rivals in overall performance?

A: The 8400F has an average benchmark score of 25005, which is 0.2% higher than the AMD Ryzen 5 7500F (24964), 0.3% higher than the Intel Core i7-11850H (24935), and 0.4% higher than the Intel Core i7-13620H (24911). It is 0.4% lower than the AMD EPYC 9474F (25103). These differences are within noise, so the 8400F is effectively on par with these chips.

Q: Is the RTX 2070 SUPER's 8 GB VRAM enough for Thief?

A: Yes, the game's textures are from 2014 and do not require more than 8 GB. The measured FPS at 4K High (40 FPS) versus 4K Medium (52 FPS) shows a 12 FPS drop when increasing quality, but this is due to GPU compute load, not VRAM capacity. The card's 448 GB/s memory bandwidth is ample for Thief's asset streaming.

Q: What is the best resolution for this combo to maintain 60+ FPS?

A: At 2560x1440, the Low preset averages 118 FPS, Medium 89 FPS (min 76), High 68 FPS, and Ultra 65 FPS. All 1440p settings exceed 60 FPS, making 1440p the sweet spot. At 1920x1080, all settings exceed 90 FPS (Low 171, Medium 130, High 99, Ultra 94). At 4K, only Low (69 FPS) and Medium (52 FPS) are close to 60 FPS.

Q: How does the RTX 2070 SUPER rank against its nearest GPU rivals?

A: The RTX 2070 SUPER has an average benchmark score of 20282, which is 1% lower than the NVIDIA Quadro M4000M (20480), 1.2% lower than the RTX 3070 Mobile (20534), 1.3% lower than the Intel Arc B570 (20556), and 1.5% lower than the Intel Arc A750 (20582). These deltas are small, indicating the 2070 SUPER is in the same performance tier.

Q: Does the CPU or GPU limit performance in Thief?

A: The GPU is the primary limiter. At 1080p Low, the combo hits 171 FPS, but at 4K Ultra, it drops to 38 FPS. The CPU's single-thread score (2943 in Cinebench R23) is more than enough for Thief's engine, and the measured FPS scales with GPU load (resolution and settings) rather than CPU thread count.

How This Combo Ranks

This combination of AMD Ryzen 5 8400F and NVIDIA GeForce RTX 2070 SUPER ranks 2091 out of 3720 tested combos in Thief, placing it in the 56th percentile (2091 / 3720 = 0.562). This means it outperforms roughly 56% of all tested system configurations in this game. The rank is driven by the GPU's mid-tier standing (65th percentile) and the CPU's higher standing (77th percentile). The combo's position suggests that while the CPU is capable of pushing higher frame rates, the GPU's compute limits cap the overall performance, especially at higher resolutions.

The rank of 2091 is not a top-tier result — the top 10% of combos likely pair high-end CPUs with flagship GPUs. However, for a 2014 game, this combo delivers playable frame rates at 1080p and 1440p across all settings, and at 4K with reduced quality. The combo's rank reflects the GPU's age (released 2019) and the CPU's recent release (2024), but the game's age means the CPU is overqualified while the GPU is the deciding factor.

Compared to the nearest rival GPUs, the RTX 2070 SUPER's 20282 average score is within 1.5% of the Intel Arc A750 and RTX 3070 Mobile, so combos with those GPUs would likely rank similarly. The CPU's nearest rivals (Ryzen 5 7500F, Core i7-11850H, etc.) have average scores within 0.4%, so swapping the CPU would not change the rank materially. The combo's 2091 rank is therefore a function of the GPU tier, not the CPU.

Settings Recommendations

Given the measured FPS data, the recommended settings depend on the target resolution. For 1920x1080, the Ultra preset averages 94 FPS, which is above 60 FPS, so players can max out the game at 1080p without issues. The Medium preset at 130 FPS (min 111) provides a higher refresh rate headroom for 144Hz monitors. For 2560x1440, the Ultra preset averages 65 FPS, which is still above 60, but the Medium preset at 89 FPS (min 76) offers a more comfortable margin for frame drops. For 3840x2160, the Low preset averages 69 FPS, which is the only 4K setting that maintains a consistent 60+ FPS experience; Medium averages 52 FPS (min 45), which is borderline for 60 Hz displays.

The best experience per the measured rows is at 2560x1440 with Medium settings, providing 89 FPS average with a minimum of 76 FPS — this ensures smooth gameplay with no dips below 60 FPS. The Ultra preset at 1440p (65 FPS) is also playable, but the Medium preset's higher minimum (76 vs unknown for Ultra) makes it safer. At 1080p, the Ultra preset (94 FPS) is recommended for maximum visual quality, as the frame rate remains high. At 4K, the Low preset (69 FPS) is the only viable option for 60+ FPS; players prioritizing visuals over frame rate could choose Medium (52 FPS) but risk drops to 45 FPS.

The data shows that the GPU's 8 GB VRAM is not a limiting factor; the settings' impact on FPS is due to shading and fill rate, not memory. Therefore, players can enable high texture quality without worrying about VRAM overflow, but should reduce shadow and post-processing settings to maintain frame rates at higher resolutions.

Measured FPS Breakdown

At 1920x1080, the measured average FPS by preset: Low 171, Medium 130 (min 111, max 150), High 99, Ultra 94. The gap between Low and Medium is 41 FPS (24%), and between High and Ultra is only 5 FPS (5%), indicating diminishing returns beyond Medium. The minimum FPS for Medium (111) is still above the average for High (99), showing that Medium provides a smoother experience than High or Ultra.

At 2560x1440, the averages are: Low 118, Medium 89 (min 76, max 103), High 68, Ultra 65. The drop from Low to Medium is 29 FPS (25%), and from Medium to High is 21 FPS (24%). The Medium preset's minimum of 76 FPS is above the 60 FPS threshold, while High and Ultra average below 70 FPS.

At 3840x2160, the averages are: Low 69, Medium 52 (min 45, max 60), High 40, Ultra 38. The Low preset is the only one above 60 FPS. The Medium preset's maximum of 60 FPS shows it can hit 60 FPS in light scenes, but the minimum of 45 FPS indicates significant drops. High and Ultra are below 40 FPS, making them unsuitable for smooth gameplay.

Comparing resolutions at the same preset, the scaling is consistent: Low at 1080p (171) to 1440p (118) to 4K (69) shows a 31% and 42% reduction respectively. Medium at 1080p (130) to 1440p (89) to 4K (52) shows a 32% and 42% reduction. High at 1080p (99) to 1440p (68) to 4K (40) shows a 31% and 41% reduction. Ultra at 1080p (94) to 1440p (65) to 4K (38) shows a 31% and 42% reduction. This consistent ~30-40% drop per resolution step confirms the GPU is the scaling bottleneck.

The measured rows include minimum and maximum FPS only for the Medium preset at each resolution. At 1080p Medium, the range is 111-150 FPS (spread of 39); at 1440p Medium, 76-103 (spread of 27); at 4K Medium, 45-60 (spread of 15). The tighter range at higher resolutions suggests the GPU is more consistently loaded, with less variance. The absence of min/max data for other presets means only average FPS can be compared for those settings.

The data confirms that this combo delivers playable performance for Thief at 1080p and 1440p across all quality settings, with 4K requiring Low or Medium settings for acceptable frame rates. The overall ranking of 2091 out of 3720 combos indicates a mid-pack result, appropriate for a GPU from 2019 paired with a 2024 CPU.