Thief

Intel Core Ultra 9 290K Plus on a GeForce RTX 3080 delivers a wide arc in Thief: the same GPU+CPU pairing spans from 5.4 to 237 frames per second. Release titles from 2014 rarely see such a spread, but the measured rows show exactly why. At 1080p Low, the system throws out 237 FPS, at 4K Ultra it falls to 54. It is a capable duo, but one that favors the CPU as leaves the GPU mostly under tension at lower resolutions and the opposite at the top.

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

Resolution scaling is a clear clue about tha the GPU's dominant role as pixel count climbs. Consider the Ultra preset column: 133 FPS at 1920x1080, 92 at 2560x1440, and 54 at 3840x2160. The three numbers show a steep drop from 1080 to 4K—about a 60% loss. But each step also interprets how the workload changes: the move from 1440p to 4K remains, but not as fast as the transition from 1080p into 1440p (134 to 92 and then to 54). More interesting is the Low preset: it yields 237 FPS at 1080p, 167 at 1440p, and 98 at 4K. Even at 4K Low it still sits at 98 FPS, which says that the GPU has enough headroom for lower-poly settings but the pixel load destroys the frame rate at higher detail levels.

The numbers tell a story about the bottleneck. At 1080p Low, the GPU probably takes a back seat, and the CPU has a high boost capacity with a 3.70 GHz base and 5.80 GHz boost. At 1080p Low, 237 is a strong indicator that the system is mostly feeding frames to a faster display rather than being limited by the Raster unit's texture rate. At the opposite extreme—4K Ultra, 54 FPS—the GPU is clearly the limiting factor. The RTX 3080's shader count (multiplication) and ROP rate become tired by full-screen workloads. When reducing settings from Ultra to High at 4K, the average goes from 54 to 57 FPS, a modest gain; dropping further to Medium yields 75 and Low 98. That progression shows a flat-ish response for first few quality steps, and about 30% gain from High to Low—typical of GPU bound scenarios.

But the important part is that High and Medium at 4K show a big difference: 57 vs 75. The extra shading load of High eats full GPU resources, resulting in a decrease. At lower 1440p, the same settings jumps from 97 (High) to 127 (Medium)—which is a 30% gain—while at 1080p, High only reaches 141 and Medium climbs to 185. So the scaling slope becomes steeper as resolution drops. These numbers reveal the 3080's 10 GB VRAM and 760.3 GB/s bandwidth are more than enough for Thief, a 2014 release. Instead, resolution scales the set of shading units and ROP.

FAQ

Q: What is the highest average FPS reported in the measured data?

The highest measured average is 237, that appears at 1920x1080 with Low settings.

Q: At 4K, which preset maintains a frame rate above 60 FPS?

The Low and Medium presets do. At 3840x2160, Low averages 98 FPS; Medium averages 75 FPS with a measured minimum of 63 FPS and a max of 86. High and Ultra do not: High delivers 58 (57) and Ultra 54.

Q: What about 1440p Ultra in this alignment?

The 1440p Ultra runs away at an average of 92; it might be playable if you accept lower than 60 at some moments, but data does not list min for Ultra. It drops to 57 on the 4K Ultra, so give this vide the caution label. 4K Medium is the best choice for a combination of visuals and margin.

Q: How close is the CPU to its nearest competitor?

The measured avg benchmark score for Core Ultra 9 290K Plus is 84,003, while the nearest iPhone product "Core Ultra 9 285K" sits at 83,807 with a deltaPct of 0.2. That is very close, only 0.2% difference, meaning the 290K Plus behaves nearly identical in synthetic tests.

Q: Which presets at 1440p keep a consistent 60?

All four do. 1440p Low averages 167, Medium 127, High 93, and Ultra 92. None of these have min values (except Medium) , but the 92 average is comfortably over 60.

Q: Where does the GPU rank against all other tested GPUs in this database?

The RTX 3080’s percentile is 68 on the GPU slider, meaning a bit more than one third of all GPUs beat it and most of them. In its rivals list, it measures 23,172 average vs the P106-100 at 23,249 (-0.3). It is a strong large portion, but not at the very top.

CPU Role — cores, clock speeds, and how they relate to this game’s results

The Intel Core Ultra 9 290K Plus falls into the Core Ultra series 2 and has 24 cores and 24 threads with a 3.70 GHz base and 5.80 GHz boost. The cache layout is 192 KB per core, 3 MB L2 per core, and a 36 MB shared L3. These are typically not needed for Thief, a game released in 2014 that won't use dozens of threads; it's the combination of a very high boost clock and single-core efficiency that helps here. In the synthetic suite, the R23 multicore result is 51,731 and the single-core 7,303 (Cinebench R23 single core) — that is a huge single-core number. The PassMark single-thread score of 4,823 confirms.

A 2014 stealth gameplay still relies on straightforward AI, physics and hiding mechanics. The dependency on single-thread performance is seen in the 1080p Low result: 237 FPS, meaning the CPU is able to feed the game at very high rate when the render load drops. The available thread count of 24 may not be used fully, but the boost to 5.80 GHz is likely the relevant spec. Looking at the CPU’s own percentile, it rests in the 96th percentile of all CPUs, meaning it is among the top 4% of processors, which helps this high frame rate against any possible CPU bottleneck at these settings.

The near-absence of any CPU constraint is also implied by the difference between 1080p Low (237) and 4K Low (98). If the CPU were the limiting partner, the FPS would not vary as much with resolution; the change from to 98 shows the GPU is the prevailing limit. This is a typical profile for an high-end desktop CPU paired with a middle-of-the-pack GPU.

Settings Recommendations — which preset gives the best experience per the measured rows

The data offers six presets across three resolutions, and every combination is playable in some sense. The real question is what the user wants out of Thief. On 4K, the medium preset at 75 average and 63 minimum is the most reliable way to keep above 60, while the high set average of 57 will occasionally dip. So at 4K, Medium is the clearest choice if the target is constant fluidity. On 1440p, there is plenty of headroom. The medium setting reports a 127 average with a low-min (108 and high 146), while High is 97 and Ultra 92. All are above 60. If the display is 144 Hz, the 127 average still does not reach that refresh, but it is a 45% jump from High, so Medium is visually faster.

1080p is a free run: the lowest combo of Low renders 237, Medium in 185, High 143, and Ultra 134. So for smooth high-refresh play, 1080p High or Medium is fine. The actual default recommendation would be 1440p Medium it offers the best balance of aesthetic response and gives an acceptable minimum 108. If the user owns a 4K screen and can accept 128, there is 60 Hz capable, but if you want a strict 60 final point, Medium is the only choice.

How This Combo Ranks — use comboRankInGame vs other tested combos

The combination places rank rank 670 out of 3,720 combos in this game. That means a better position than over 81% of the pool (not % value used, just derived). The CPU sits at the 96th percentile for all CPUs and the GPU at 68th percentile. So while the GPU is rather modest in its overall score, the CPU pushes this into the top end. A lower CPU (like the 285K) has almost identical benchmark score (0.2%), but combined with a 3080 the overall impact is basically the same. What pulls the result down is mostly the GPU, which has a percentage of 68. The combo is not in the very top 20% based on rank.

Given the game released in 2014, any modern CPU can far exceed it; this rank is relative to other GPU-CPU pairs. The GPU's average benchmark 23,000 range and the CPU ~80,000, so the full system is a strong one.

Measured FPS Breakdown — resolution by resolution, settings by settings, exact numbers

Below is the full list of measured avg FPS for all rows in the FACT PACK. The Medium row even includes min and max.

| Resolution | Settings | Avg FPS | Min FPS | Max FPS |

|-----------:|:-------|:---------|---------|---------|

| 3840x2160 | High | 57 | — | — |

| 3840x2160 | Low | 98 | — | — |

| 3840x2160 | Medium | 75 | 63 | 86 |

| 3840x2160 | Ultra | 54 | — | — |

| 2560x1440 | High | 97 | — | — |

| 2560x1440 | Low | 167 | — | — |

| 2560x1440 | Medium | 127 | 108 | 146 |

| 2560x1440 | Ultra | 92 | — | — |

| 1920x1080 | High | 141 | — | — |

| 1920x1080 | Low | 237 | — | — |

| 1920x1080 | Medium | 185 | 157 | 213 |

| 1920x1080 | Ultra | 134 | — | — |

Reading it more precisely: At Medium, 4K gives 63 as the lowest frame; 1440p 63 gives 108; 1080p gives 157. So if the user absolutely wants no dips, 1440p Medium still has 108 minimum. At 1080p Medium, the 7 HR min is above any 60 mark. These mins and maxes are the only ones recorded; other presets did not have them.

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

The NVIDIA GeForce RTX 3080 uses a GA102, Ampere on an 8 nm node, with 10 GB GDDR6X of 320-bit bus and bandwidth of 760.3 GB/s. Base of 1440 MHz, boost 1710 MHz, and a TDP of 320 W. It has 87.4 shading units, 96 ROPs and 272 tensor cores. The native PCIe 4.0 interface adds none to the frame count.

Thief was release near and does not need much VRAM, but the card's 10 GB capacity never hits a wall. What the data shows is that the RTX 3080 begins as the limit at 4K Ultra: 54. It computes five pixel-rate 164.2 GPixel/s and texture 465.1 GTexel/s. These are the only render metrics that alone produce most of being bottlenecked. At 4k, Hi vs Ultra difference is only 3 fps (57 vs 54) — indicating the raster fill-rate ceiling. In contrast, lowering from Ultra to Medium at 4k, texture work persists and F jumps to 75, showing that the ROPs are not the sole limit; shaders matter as well.

The RTX3080 also carries tensor cores and ray tracing cores, but Thief (2014) has no ray tracing, so their potential is not being used. That matches with the relative performance: GPU percentile 68, yet in this game the combo reaches a rank of 670, because the CPU is very, no.

In the list of nearest GPU rivals, the 3080 rails against the P106-100 (23249, 0.3%), Radeon Pro Vega 16 (23250, 0.3%), and RX 6600M (23273, 0.6%) and R9 M290X (23276, 0.4%). The differences are all under fraction a percent, suggesting the 3D index is not the best predictor for Thief.

Final note: the RTX 3080 has a launch MSRP of 699. The number stands in the database only, but it does not adjust the measured-driven conclusion.

That sets the Combo Rank story firmly: the 3080's mediocre standing (68th percentile) is not a weaknesses when the game runs at modest settings because the CPU can catch up. It is a balanced combination for an older title, unusual for 2025.