Thief

The AMD Ryzen 7 5800X and NVIDIA GeForce RTX 3080 Ti form a high-end pairing for Thief (2014), landing at a measured combination rank of 967 out of 3,720 tested configurations. The CPU is a desktop Zen 3 chip with 8 cores, 16 threads, a base clock of 3.80 GHz, and a boost clock of 4.70 GHz; its launch MSRP is $449. The GPU is an Ampere-generation GeForce 30 series card with 12 GB of GDDR6X memory and a launch MSRP of 1,199 USD. The measured results show a system that easily handles smooth play at 1080p and 1440p across all presets, while 4K pushes the average frame rate below playable levels at the highest settings, though the drop is more consistent than dramatic.

How This Combo Ranks

In the benchmark database this exact CPU and GPU pairing sits at rank 967 out of 3,720 tested combos in Thief. That puts the combo in the upper portion of the list, though not in the very top tier. The CPU alone holds a percentile of 81 against all CPUs in the database, and the GPU holds an 83rd percentile against all GPUs. These independent metrics align with the FPS observed here – the system performs well but is not extreme relative to the wider pie.

Looking at the CPU’s synthetic averages, the Ryzen 7 5800X has an avgBenchmarkScore of 29,123. That sits in a very tight cluster among nearest rivals: the Intel Core Ultra 5 135H scores 29,093, which is a deltaPct of 0.1; the Intel Xeon Phi 7210 scores 29,245 (-0.4 deltaPct); the Intel Core i5-13500HX scores 29,270 (-0.5); and the AMD Ryzen 5 5600XT scores 28,940 (0.6). In other words, the 5800X is effectively tied with all four rivals in aggregate synthetic compute, within a band of about 0.6%. That means the CPU is not the factor separating this combo from its neighbors — the engine is predictable across similar hardware.

The GPU, on the other hand, averages 41,187 points. Its nearest rivals are all significantly separated: the AMD Radeon Pro 5300 sits at 40,870 (0.8 deltaPct), the NVIDIA Tesla M40 24 GB at 41,707 (-1.2), the NVIDIA Tesla M40 at 41,897 (-1.7), and the newer NVIDIA GeForce RTX 5070 at 40,377 (2.0). The 3080 Ti’s score is comfortably in the middle, not wildly faster than a floating-class workstation GPU but also well ahead of a current midrange card in raw bandwidth. For Thief, a title from 2014 the data shows the combo rank of 967 is better than what 95% of the test pool would achieve, but it is not a plea.

GPU Role

The NVIDIA GeForce RTX 3080 Ti is not a modest GPU. It carries 12 GB of GDDR6X memory on a 384-bit bus, delivering a bandwidth of 912.4 GB/s. Its GPU runs at a base clock of 1365 MHz and a boost clock of 1665 MHz, with a memory clock of 1188 MHz and 19 Gbps effective. The silicon contains 10,240 shading units, 320 texture mapping units, 112 render output units, 80 ray-tracing cores, and 320 tensor cores, capable of press out 34.10 TFLOPS of FP32 performance. The pixel rate is 186.5 GPixel/s and the texture rate 532.8 GTexel/s.

That spec sheet matters because Thief is an older DirectX 12-ultimate supported title, and the GPU is well past its needs. At 1080p the card has wad of headroom. Even at High settings, which use more deferred lighting and textures, the average FPS reaches 139, while a pure Low preset barely saves the system, with 169 FPS. The GPU's power is far from maxed out at 1080p; the bottleneck shifts to the CPU sometimes, but the 5800X with 8 cores and high IPC handles it. The 12 GB frame buffer is generous for Thief, which was released in 2014 and is not a VRAM hunger beast. Even with 4K textures, the 12 GB at 912.4 GB/s would be unlikely to cause frame pops, but the data here is limited to these measured rows.

At 3840x2160 Ultra, the average FPS sits at 56, with High at 59, Medium at 77, and Low at 102. The fact that Ultra and High are nearly identical (56 vs 59) suggests the GPU is approaching its fill-rate, and that is confirmed by the memory bandwidth — 12 GB at 80 FPS, with shadows that 3840 (2560x1440) the frame numbers show the GPU has more than enough to accompany a 144 Hz display at medium settings.

Resolution Scaling

The measured data contains twelve runs, and the resolution spread shows a clear downward trend. At 1920x1080, the High preset yields an average of 139 FPS. Moving up one step to 2560x1440, the same settings fall to 100 FPS. At 3840x2160, that drops further to 59 FPS — a reduction of about 41 frames from the 1440p result, and 81 frames from 1080p.

The Ultra preset follows nearly the same path: 136 FPS at 1080p, 95 at 1440p, and 56 at 4K. Low settings, which reduce the shadow quality and post-processing, see 107 FPS at 1080p, 149 at 1440p, and 102 at 4K. Medium presets are a comfortable middle ground: 153 at 1080p, 132 at 1440p, and 77 at 4K.

All four presets show a consistent pattern: scaling from 1440p to 4K hits the hardest, while the jump from 1080p to 1440p is smaller. For example, the High preset loses 39 FPS going from 1080p to 1440p, then loses 41 more going to 4K. Ultra loses 41 FPS from 1080p to 1440p, then 39 to 4K. The FPS departure from a 4K render parggests the GPU is the limiting factor at those high pixel counts. At 1080p the CPU can easily send frames fast, but at 4K the rendering load outruns the GPU, compressing the difference between settings. The largest single gap is between Low and Medium at 4K: 102 vs 77, which is a 25-point gap, while the jump from High to Ultra at 1080p is only 3 points (139 vs 136). That implicit hierarchy tells the reader which setting to pick for smooth 4K.

Settings Recommendations

The benchmark rows provide four presets per resolution, and they reveal a simple ordering: Low and Medium tend to have the highest headroom, while High and Ultra hug the lower end. The measured data is all averages, except the Medium settings which also include minimum and maximum frames.

For 4K, the recommended preset according to the measured performance is Medium. It delivers an average of 77 FPS, with a minimum seen of 66 and a maximum of 89. If consistent smoothness is a priority, that Min of 66 means it stays above 60 in semi-predictable section. In contrast, High averages 59, which is right at the 60 boundary, and Ultra drops to 56, so the Low preset provides the highest overall count at 102 average but adds a small amount of visual quality back. A reasonable choice is either Low for smoothness or Medium for the best balance. The Ultra preset at 4K needs 56 FPS and will inevitably dip, making it acceptable only for screenshots or slower gameplay.

At 1440p, the sweet spot is Medium: 132 average FPS, with a minimum of 112 and maximum of 151. High gets you 100 and Ultra gets 95, which isn't a big drop, but the gains from disabling the Ultra extras are may not justify it. Low at 149 leaves a lot of visually quality on the table, so there is no reason to use it at 1440p. In fact, at 1080p Middle is also the balanced option across the board — 153 average with a 130 minimum — while High (139) and Ultra (136) are still speedy. If you want the absolute highest average, Low reaches 169, but 136 is still genuinely smooth at 1080p. The trade-off between image quality and average frames is not huge at 1080p, with all four presets within 33 FPS of each other.

The data cover only a few settings, and each is still above 230 average at 1080p, so for a single display at that resolution load any preset. It’s only when pushing 4K that the settings decision actually changes playability: Medium is the highest preset still above 60, and High and Ultra step down into the 40–60 range.

FAQ

Q: What is the measured combo rank for this CPU-GPU pair in Thief?

A: The brand sits at 967 out of 3,720 total measured combos in the database, toggling to the top ~14% the majority of tested systems.

Q: Does the game reach 60 FPS at 4K UItra with these parts?

A: No. At 3840x2160 Ultra the average is 56 FPS, and High is 59 FPS. Both are just off the target. Medium at 4K hits 77 FPS, and Low is consistent at 102 FPS.

Q: Which preset has the highest minimum FPS in the data?

A: The Minimum FPS is only recorded for the Medium preset, not for Low, High, or Ultra. Among medium entries, the highest minimum is 130 FPS at 1920x1080, followed by 112 at 2560x1440 and 66 at 2160.

Q: Is the GPU stronger than the CPU in terms of synthetic percentiles?

A: The GPU has a percentile of 83 against all GPUs, which is slightly higher than the CPU’s 81st percentile against all CPUs. Both are in the same band of top-end hardware, but the GPU edges out slightly in relative rank.

Q: How does the CPU’s core and thread layout help in Thief?

A: Race with 8 cores and 16 threads, running a base of 3.80 GHz up to a 4.70 GHz boost on an AM4 socket. The 32 MB L3 cache and 16 threads ensure the system never runs raw compute dry, which the high 1080p FPS tables in the data confirm.

Q: Is 12 GB of GPU memory enough for Thief at 4K?

A: The 12 GB GDDR6X on a 384-bit bus at 912.4 GB/s is far more than the game's 2014 requirements. The measured results show no indication that VRAM is the limiting factor; at 4K Ultra the card falls to 56 FPS regardless of its memory, a fill limitation rather than a memory one.

CPU Role

From a computational standpoint the AMD Ryzen 7 5800X is a desktop processor built on the Zen 3 architecture, with a 7 nm process from TSMC. It packs 8 cores and 16 threads, which is enough for game logic, audio, and background tasks in a 2014 single-player title. The boost clock of 4.70 GHz is single-core headroom, and the L3 cache is a generous 32 MB, helping with cache-hungry simulation sizes even in the older engine.

Synthetic data shows the CPU achieving a Cinebench R23 multicore score of 15,476 and a single-core score of 1,574.5, both high enough to move frames without a bottleneck below 4K. At 1080p, the FPS is the highest across all settings: for example, the average at 1080p Low is 169, at High 139. These values are typical when the CPU can push games faster than the GPU can draw, especially since Thief contains a lot of small objects and shadows where single-thread speed matters.

The CPU’s nearest synthetic rivals cloud judgment on how overkill it is. Its 8-core/16-thread layout is mirrored by the Intel Core Ultra 5 125H (14C/18T) but the aggregate benchmark scores differ by less than 0.4%. The 5800X’s extremely stable average places it just behind the Core i5-13500HX by a margin of -0.5%, and slightly ahead of the Ryzen 5 5600XT by 0.6%. Those differences are tiny in chip terms, and for Thief they manifest only at the extreme low-resolution high-FPS regime where the CPU becomes the partial limiter.

At 1440p and 4K, the CPU absolutely fades into the background. The difference between the 1080p High and the 1440p High is 100 FPS (139 to 100), but the CPU is still nowhere near its limit; it’s the GPU being tasked with 70% more pixels. The design of the 5800X, with high IPC and only 8 cores, is firmly in the category of processors that deliver good frame pacing across three resolutions, and nothing in the measured data indicates a CPU stall. The single-core score of 1.574.5 in Cinebench R23 and a Geekbench 6 single score of 2,032 back up the idea that older engine logic running the main thread likes this chip.

Measured FPS Breakdown

Below are the twelve measured framerate results, grouped by resolution. All rows are average FPS, also including the minimum and maximum values for the Medium preset where available.

• 1920x1080 – High – avg 139 FPS
• 1920x1080 – Low – avg 169 FPS
• 1920x1080 – Medium – avg 153 FPS, min 130 FPS, max 176 FPS
• 1920x1080 – Ultra – avg 136 FPS
• 2560x1440 – High – avg 100 FPS
• 2560x1440 – Low – avg 149 FPS
• 2560x1440 – Medium – avg 132 FPS, min 112 FPS, max 151 FPS
• 2560x1440 – Ultra – avg 95 FPS
• 3840x2160 – High – avg 59 FPS
• 3840x2160 – Low – avg 102 FPS
• 3840x2160 – Medium – avg 77 FPS, min 66 FPS, max 89 FPS
• 3840x2160 – Ultra – avg 56 FPS

These numbers cover the full measured space for Thief. Notice that only Medium rows have min and max bands, meaning we lack direct variability data for the other presets. However, the plateaus at High and Ultra for each resolution are still informative: they sit nearer to the limit of the GPU, and the delta between High and Ultra is a small, hard-fought 3 frames at 1080p (139 vs 136), 5 frames at 1440p, and 3 frames at 4K (59 vs 56). The two low-effort presets (Low and Medium) have large headroom, making them the primary candidates for high refresh rates