Thief

Measured benchmark data is available for Thief running on an Intel Core i3-12100F paired with an AMD Radeon RX 580, and the results show a combination that is largely limited by the GPU at higher resolutions. Across a sweep of three resolutions—1920x1080, 2560x1440, and 3840x2160—and four quality presets (Low, Medium, High, Ultra), the system delivers frame rates that range from 83 FPS at its strongest to 18 FPS at its weakest. The outcome positions this pairing near the lower end of the testing pool, ranking at 3397 of 3720 measured combos, a reflection of the RX 580's legacy architecture and the i3-12100F's compact 4-core design. The data is directly measured, not estimated, providing a concrete reference point for anyone evaluating this older desktop hardware against modern-game expectations.

FAQ

Q: What is the fastest average frame rate measured across all resolution and preset combinations?

A: The highest average observed is 83 FPS at 1920x1080 with the Low preset. The next-highest are 68 FPS at 1080p Medium, and 57 FPS at 1440p Low.

Q: Does the RX 580 + i3-12100F combo hit 60 FPS at the game's higher quality presets?

A: At 1080p Medium the average is 63 FPS, which dips above the 60 FPS line. However, at 1080p High and Ultra the averages fall to 57 and 46 FPS respectively. At 1440p, the Low preset yields 57 FPS, while Medium and above fall below 50 FPS. At 4K, every preset sits well below 35 FPS.

Q: Is the Radeon RX 580 an 8 GB card, and how does its VRAM capacity impact the results?

A: Yes, the RX 580 comes with 8 GB of GDDR5 memory on a 256-bit bus, delivering a 256.0 GB/s bandwidth. The measured FPS data shows that memory is not the visible bottleneck—even at 3840x2160 Ultra the GPU still reports 18 FPS, suggesting that raw compute rather than VRAM is what caps performance.

Q: What is the CPU's core/thread configuration, and does it allow a smooth 4K experience?

A: The Core i3-12100F has 4 Cores and 8 Threads with a base clock of 3.30 GHz and a boost clock of 4.30 GHz. At 4K Ultra the combo only reaches 18 FPS, and at 4K Low it gets to 33 FPS; the system never reaches 40 FPS at 4K, even with the lowest settings. The CPU cannot compensate for the GPU's limits when rendering the full 3840x2160 target.

Q: Which component shows a lower percentile in its category?

A: According to the performance percentile data, the CPU sits at the 68th percentile among all CPUs while the GPU is at the 54th percentile among all GPUs. Both are low-mid-range, but the graphics card has the weakest position relative to its own class.

Q: The combo rank, what is it specifically?

A: The combination occupies the 3397th place out of 3720 tested system combinations. That places it in the last-few-hundred–only about 323 entries sit below it, and nearly 3000 configurations perform better. Only a small fraction of tested combos are slower.

GPU Role

The AMD Radeon RX 580 is a Polaris-generation part built on a 14 nm process, powered by a 5,700-million-transistor Polaris 20 chip with a 232 mm² die. This is the core of the combination's graphics processing, and the measured FPS table shows how heavily it shapes Thief's output rate. At the heart of the problem is its raw shader execution: 2304 shading units, 144 texture mapping units, and 322 rows of pixel output capacity produce a pixel rate of 42.88 GPixel/s and a texture rate of 193.0 GTexel/s. The floating-point figure, 6.175 TFLOPS, across all FP32 operations remains consistent with the GPU's peak theoretical performance and is exactly where the measured 4K performance lands.

The card's VRAM is 8 GB GDDR5, operated through a 256.0 GB/s bus, a figure that does not cause a physical hardship for this game. The 8 GB capacity is far above the typical requirements of a title released in 2014; Thief was not content‑diffusion-hungry. At any of the tested settings the configuration never falls below vector's visible, no memory trin. The clock core operates at a base of 1257 MHz and a boost of 1340 MHz, while the memory is clocked at 2000 MHz with an effective 8 Gbps data rate. The performance logs in the FACTECTION table indicate the RX 580 averages about 42.88 GPixel/s per second. That memory bandwidth is sufficient for a game of Thief's generation but the low FPS when stepping to 4K suggests that a combination of the 32 ROPs and the brute pixel fill rate is the true limiting element.

The GPU's overall score in standard benchmarks is 8,813 points in the 3DMark Direct3D g-case and 3,500 in GPU compute. Its percentile stands at 53rd statistically, meaning the RX 580 sits essentially at the midpoint of available graphics cards. The three closest space in the benchmark leaderboard show NVIDIA GeForce RTX 3050 Ti Mobile (-0.1% delta), the NVIDIA GTX 1660 SUPER (-0.4% delta), and the AMD Radeon 740M (+0.5% delta). This positions the Polaris chip at the same performance ladder as much newer 50 W laptop chips, indicating that the RX 580's 1,864-bit and 2,304 ALUs provide a performance equal to those GPUs, even when they emit literally. That fact explains why at 1080p the card is still healthy; at 4K it's the chip's raw throughput that its legacy fails. The GPU has a factory launch MSRP of 229 USD, but the pricing is of no consequence to this data analysis.

How This Combo Ranks

When placed against the full comparative sample of tested combos, the Intel Core i3-12100F + RX 580 setup holds a competitive position of 3397th out of 3720 total combinations. That translates to the bottom half; only 323 combos below it, so 3,023 configurations produce higher average performance. In practical terms this rig will often rival two-generation-old laptops or budget desktops that were high-end some years ago. But the frame data at 4K—between 18 and 33 FPS—is essentially unplayable by modern standards, while the 1080p Medium 63 FPS is a plausible playable frame rate. The biggest near-model on the CPU side is the Intel Core i5-9500, which is only 0.3% faster on average standard scores, yet those scores do not necessarily reflect gaming behavior in Thief; the deltaPct can also be used as a data result.

The EXACT measured frame rates from 18‰ to 83 FPS span a huge margin of 65 FPS, and the slope is not always predictable. The max the card at 1080p Low gives 83 FPS, while 4K Ultra drops to 18 FPS—that's less than a fourth of the speed. Given that the CPU is a 4-core part with a 68% percentile, the primary conclusion is that the GPU does most of the work in the lower resolution domain; at High 850 the gap in the adjacent water rapidly grows.

Measured FPS Breakdown

The measured frames data lists no min/max for most presets, but average fps provides a clean imaging of the system's peak at each config. The full table is segmented as follows:

• 1920x1080 (1080p): At Low the average is 83 FPS. Medium is 63 FPS, with a minimum of 54 and maximum 72 FPS. High reads 57 FPS; Ultra has an average of 46 FPS.
• 2560x1440 (1440p): Low is 57 FPS, Medium 43 FPS (min–max 37–50), High 33 FPS, and Ultra 31 FPS.
• 3840x2160 (4K): Low 33 FPS, Medium 25 FPS (min 22, max 31), High 19 FPS, Ultra 18 FPS (ultra min not logged, max not logged).

Looking at the slope between Low and Ultra at each resolution: at 1080p Low drops by 36 FPS (from 89 to 46) when the preset is turned up. At 1440p the span is 31 FPS (57 to 31). At 4K the span is only 16 FPS (33 to 18). The effect of settings is thereby much more visible at lower resolutions because the non-CPU GPU load is still within its reach. On the other hand, at higher resolutions the GPU is already saturated and raising quality reduces the output less dramatically, while the absolute output is always too low for smooth gameplay.

CPU Role

The Intel Core i3-12100F is a current-gen Alder Lake-S part manufactured on 10 nm with four physical cores and eight threads. Its unscaled base clock of 3.30 GHz and a boost of 4.30 GHz gives it solid single-thread capability, as shown by its Cinebench R23 single-core score of 1674 and Geekbench single-core of 1954. On the multi-thread dimensions, the 8-thread Cinebench R23 multicore number reaches 11860, while the older R20 tests produce 4981 and single-strand score 720–the CPU obviously serves enough parallel throughput for a 2014-era stealth game.

From the pool of nearest rivals—each with deltaPct of respectively one—the CPU is in close range of Core 3 N356, Intel Core i5-9500, Intel Core 5 120UL, and Core i7-1250U. Its average benchmark score is 13494 out of a 100 scale, the 68th percentile of all CPUs. The strongest rivals in this group all show extremely close scores: the i5-9500 at 13452 (0.3% delta), the i5-1250U 13594 (-0.7%), and i7-1250U 13351 (+1.1%). Only 3.0 points separate them, meaning the i3’s improvement is entirely counterbalanced by similarly capable chips.

In a stealth title like Thief the game relies on the CPU for mostly AI positioning, game state, occlusion logic, and a variable light function. The quad core i3-12100F has minimum 80 KB of L1 and 1.25 MB of L2 per core, plus a 12 MB shared L3. The 800 MB cache is likely sufficient for the level-of-detail and scripted activity typical of the 2014 build. With its base at 3.30 GHz and boost of 4.30, it appears that at all resolutions the frame pacing is limited by the GPU head to zero. Evidence: when operating the 1080p Low, the system’s GPU has to fill fewer pixels, pushing the frame to 83 FPS, and in the 4K row, it drops to 33 FPS. LEX—the CPU itself is not overwhelmed as the game's minimum at 3840k components is the graphics work. The processor does enforce a small scaling at the 1080p High preset—that's 48 FPS—still larger than those lower core versions would.

The CPU also runs in dual-slot memory DDR4 or DDR5 over dual-channel as quick as 58 W TDP; it provides a memory access that does not bottleneck Thief, which by the era was not ultra-taxing. The TR multi-thread score 11860 suggests the i3 has enough multi-threaded skill outsource for real engine tasks. But it also needs lifting: in the 3690 and 1550 a dedicated CPU: when the GPU is set to low and 1980 resulted in 63, the processor does not easily saturate. With 8 threads at 3.3 base (many rises to 4.3 under single thread, 4.0–4.2 in two-thread loads), the data shows no individual core spiking abnormally. The in the Thief workload does not appear to escalate beyond a pair of cores. This matches the Cinebench R15 single- spec of 168 and Cinebench R20 single of 703, which very good at per-core basis, 87a 45. So the choice of the lower segment CPU doesn't get blamed for not seeing 60 FPS at 4K.

Settings Recommendations

The measured values offer a clear view of which preset is the best trade-off between visual and smoothness.

At 1920x1080, the Medium preset strikes the most usable outcome of all measured rows. It yields an average of 63 FPS with a reported minimum of 54 FPS and maximum 72 FPS, producing a fluid scene even if not flat over 60. The High preset is 48 FPS and Ultra at 46 FPS — both below the smooth marker, though for a slower-paced stealth game you might survive, yet there’s little reason to favor them given the gap from Medium to High only costs a small margin of visuals. The Low setting shoots 83 FPS, but the Lumber scene usually far; no common sense attractions except it's if you have a 144Hz display, in that case Low obviously superior makes frame*playable. The dataset says the sweet spot is Medium for most 60 Hz monitors.

At 2560x1440, the balanced option changes. The Low setting's 57 FPS is magnified and close; sometimes a speed within margin of 60, whereas Medium arrives only 43 FPS. For a clean maneuver, a user would use either Low at 1440p, or sacrifice to Medium—43 FPS is generally tolerable in a solitary stealth action game in which reaction speed matters less than timing. The High is 33 FPS and Ultra 31 FPS, presumably visually rich but too stutter-prone for active gameplay.

At 3840x2160, the data reveals no truly playable preset. Even the best case (Low) is only 33 FPS, and the rest sit at 18–25 FPS. It is obvious from this section that 4K is not feasible, regardless of the quality slider, at any of the four configurations. If 4K is the target, one would need to lower the resolution or accept sub-30 FPS in certain static moments. Without a V-sync indicator, the output implies far worse.

Therefore the recommendation is clear: use Medium at 1080p to get an average 63 FPS and a comfortable frame window, or Low at 1440p for the highest smoothness at 57 FPS (almost 60), whichever visual preference you prioritize.

Resolution Scaling

As the resolution increases, average FPS falls consistently within each preset group. The data shows a clear relationship between the number of filled rows and the RX 580's performance. Like from 1920x1080 to 2560x1440, on the Low preset, the frame rate falls from 83 FPS to 57 FPS—a drop of 26 FPS. For Medium, it dries from 63 FPS to 43–44 FPS. Moving from 1440p to 2160p, the Low preset goes from 57 to 48. Actually the 1440p High is 33 FPS and 4K High 19 FPS, a 14 FPS decline. The Ultra similarly goes from 46 to 31 to 18.

This scaling pattern is characteristic of a GPU-bound segment: the dropping framerate is roughly proportional to the number of pixels, and the CPU single-core speeds have no upward help. For example, the 4K Low (3 33) is worse than 1080p Low (83) by a factor of ⅓. The 4K Ultra (18) is roughly one-fourth (20) of the 1080p Low. That strong dependence on output resolution is a classic sign that the limiting component is the Radeon 580's pixel and texture fill rates. The CPU plays a secondary role: it can deliver a good average at low resolutions, such as 83 FPS at low, but it cannot do anything to rescue higher resolution files.

It also appears that the difference between 1440p and 4K is consistently large: the medium preset for instance reports 43 FPS at 1440p but 25 FPS at 4K, a difference close to half the original output. Meanwhile the 120-FPS ceiling is not reached anywhere by this CUMULATIVE root. This conclusion is consistent with the GPU's bandwidth / compute numbers: the 256.0 GB/s memory bus provides enough bandwidth at 1080p and partially at 1440p, but at 2160 pulls the 33 ROPs into near-bottleneck perfection. The 660 TFLOPS peak compute is simply not capable of shoving those many shading factions at 4K ultra rates.

In the end, the data shows a system that will handle Thief smoothly at 1080p low—well beyond 60 FPS—and at 1080p medium folding very close to 60 fps. Higher resolutions to receive (custom) from the GPU to keep a constant frame rate. The same GPU’s positive percentile and the CPU’s impressive single-thread results converge: this pair appears the absolute ratio for a 1080p game, not for 1440p or higher at the targeted frame time benchmark.