Thief

This build pairs a current-generation 24-core desktop processor with a last-generation mid-range graphics card. The Intel Core Ultra 7 270K Plus runs Arrow Lake Refresh silicon at a 3.70 GHz base and a 5.50 GHz boost, while the GeForce RTX 2060 brings Turing-architecture silicon with 6 GB of GDDR6 memory and a 1680 MHz boost clock. In the 2014 stealth title Thief, the measured result is a usable but not exceptional pairing; the CPU sits in the 96th percentile of all tested processors, while the GPU lands at the 58th percentile, which predicts exactly what the frame data shows: a CPU that is never pushed, and a GPU that sets the playable ceiling.

FAQ

*Q: Can this system run Thief at 60+ FPS at 1080p?*

A: Yes. At 1920x1080, every measured setting row is over 60. Low averages 142 FPS, Medium 108 FPS, High 82 FPS, and Ultra 78 FPS. Even the average 60 FPS mark is only missed at 4K.

Q: What about 1440p performance?

A: At 2560x1440, Low hits 97 FPS, Medium reaches 74 FPS, High falls to 56 FPS, and Ultra stays at 54 FPS. Medium and Low clear 60; High and Ultra do not.

Q: Is 4K playable?

A: Only conditionally. At 3840x2160, Low produces 57 FPS and Medium produces 43 FPS, which is under 60. High sits at 33 FPS and Ultra at 31 FPS. For smooth 4K, the Low preset is the only likely sustained 60, while Medium has a measured peak of 50 max when best.

Q: What makes the CPU appear so strong here?

A: The Core Ultra 7 270K has a 44253 Cinebench R23 multi-core score and a 2439 single-core. Its nearest competitive alternative in the data is the Intel Xeon 6520P with 93786 average, showing zero percentile delta—Good CPU headroom. In this game, higher frame rates at 1080p Low (142 FPS) confirm the CPU is nowhere near saturated.

Q: Is the RTX 2060 the limiting component?

A: Yes. The GPU’s 6 GB VRAM, 192-bit bus, and 14 Gbps effective memory produce 336 GB/s bandwidth. The drop from 1080p Low at 142 FPS to 4K Low at 57 FPS does not match that, 4K demands scaling; the GPU at 4K Ultra only averages 31 FPS. All evidence points to the GPU as the constraint when resolution increases.

Q: Do Min/Max values show frame pacing issues?

A: The only measured min/max rows are on Medium settings. At 1080p Medium the run ranges from 91 to 124 FPS; at 1440p Medium 63 to 85 FPS; and at 4K Medium 37 to 50 FPS. The spread at 4K is about a 44% variation, a weaker result than the average would suggest.

---

GPU Role

The GeForce RTX 2060 uses the TU106 chip, a 445mm² die with 10,800 million transistors. It has a base clock of 1365 MHz, a boost of 1680 MHz, and memory that runs 1750 MHz with a rating of 14 Gbps effective. The card has 1920 shading units, 120 texture units, 48 ROPs, 30 ray tracing cores, and 240 tensor cores, producing 6.451 TFLOPS of FP32 performance. Memory is 6 GB of GDDR6 on a 192-bit bus, giving 336 GB/s. Pixel rate is 80.64 GPixel/s and texture rate is 201.6 GTexel/s.

These GPUs tend to be the functional limit in this test. At 4K Ultra and 4K High, the card can average 33 and 31 FPS—so the same three-dimensional performance demands of Thief at 2160p simply scale beyond what the TU106 can supply. The gap between 4K Medium (43 FPS) and 4K Ultra (31 FPS) is only 12 FPS. That is a 23% reduction from scene to scene.

The GPU benchmark percentile shows a 58th-percentile standing versus all tested. In the data, it scores 14111 on Passmark G3D, 14111 on Geekbench. The card’s nearest GPU rivals are the GeForce GTX 580 (average 15283, delta 0%), the AMD Radeon 680M (15270, +0.1% above), the NVIDIA RTX 3050 OEM (15199, +0.6% above), and the AMD RX 7600 (15171, +0.8%). That bracket is a midrange slot, which is exactly how it performs in this game.

RTX 2060's compute scores are lower: 5488 in Passmark GPU compute and 1242 in 3DMark Steel Nomad DX12. With DirectX 12 support it can still get into modern APIs, but the result below will be limited by raw fill at 4K.

The frame data shows a good 6 GB VRAM budget for a 2014 title; the GPU has 48 ROPs, which handle the pixel-rate at 80.64 GPixel/s. But at 2160p, those ROPs are overloaded by about 12% more than at 1080p. The biggest single performance drop comes from 43 to 33 that occurs when the settings move from Medium to High at 4K.

Note the GPU's 58th percentile vs. CPU's 96th percentile—the real speed differential comes from the Intel. The CPU has a passmark multi-thread of 68574 and single-thread of 5068, enough for any frame at 1920x1080. In the 1080p Low setting the system reaches 142 FPS, numbers that will put the pressure entirely onto the RTX 2060.

---

How This Combo Ranks

The combo position among all tested processor/graphics combos is 2927th out of 3721, as measured in the game. That denotes the bottom two-thirds of the batch. The CPU alone ranks in the 96th percentile, as its overhead is high; the GPU only reaches 58th percentile. So, In a game like this, the combo rank is far worse than the CPU's individual rank.

The data shows that comparisons to the GPU’s nearest rivals do not create much difference: the RTX 2060 is just 0.8% behind the RX 7600 and 0.1% ahead of GTX 580 in average. This narrow spread means the combination with the 2060 will be positioned very close to several popular cards, and all within 0.8% in synthetic terms.

The rank can also be expressed with the fact that 2927 of 3720 combos place ahead; that is 78.6% of the pool. With a CPU at the very top of the percentile list, this poor combo rank is bubbling up from the video element. In a CPU-heavy game such as Thief (released 2014), the CPU is not a bottleneck.

The average benchmark score of the CPU is 93785. Among CPU rivals, the Intel Xeon 6520P average 93786 is delta 0%, AMD EPYC 4564P at 95183 is -1.5% below, EPYC 9175F at 95615 is -1.9%, and EPYC 4565P 95764 is -2.1%. This position, at the level of a large EPYC server processors, ensures this CPU has excess capability for an old game.

In contrast, the GPU's average 15290 puts it near far older cards. Its delta to the RX 7600 is 0.8%, which is still closer than the 0.8% gap support. The combination of an elite CPU with a mid 2060 means the system ranks as 27th closest. For a builder optimizing Thief, the rank by itself says that a CPU upgrade is not necessary.

---

Measured FPS Breakdown

All measurements are measured, not estimates. Starting at 1920x1080:

• Low: 142 FPS average
• Medium: 108 FPS average, 91 min / 124 max
• High: 82 FPS average
• Ultra: 78 FPS average

The 1080p results show a range of only 64 FPS between the absolute lowest (Ultra 78) and the highest (Low 142). The Medium preset’s minimum of 91 FPS is already a full 13 FPS above the Ultra average, which means that even during dropped frames, Medium would be safer than a 60 Hz refresh.

At 2560x1440:

• Low: 97 FPS average
• Medium: 74 FPS average, 63 min / 85 max
• High: 56 FPS average
• Ultra: 54 FPS average

The 1440p results show a cliff between Medium and High: dropping from Low 37 to 74, then down to only 65. That’s a 18% reduction— a spot that points to the RTX 2060's raw throughput. Medium at 1440p still has a min of 53 FPS; High (56) is fine, but a min is not given, which leaves no floor guarantees.

At 3840x2160:

• Low: 57 FPS average
• Medium: 43 FPS average, 37 min / 50 max
• High: 33 FPS average
• Ultra: 31 FPS average

4K is the AAAKK separator. From Low (57) to Medium (43) the loss is 14 FPS, an 24% drop. Medium’s maximum of 50 FPS is still less than the Low average of 57. Bumping from Medium to High goes from 43 to 33, another 23% drop, and from High to Ultra it goes 33 to 31, which is only 6%. So the RTX 2060's limits appear to be flat at the Ultra setting; 4K Medium uses the ceiling tolerance.

If we think of the Min/Max row at 1080p Medium, the 91-frame floor means at 1080p this can run on a 75 Hz display with the GPU averaging 91+. At 1440p Medium, the 63-frame floor is marginal for 60Hz, but the 85-frame peak shows there is no headroom. At 4K Medium, 37 min/50 max is just barely playable on a 30 Hz panel.

---

Resolution Scaling

The key takeaway is the scaling factor between the settings.

Going from 1080p to 1440p (a 2560x1440 pixel area, about 1.78x the pixel count of 1920x1080), the frame rates by construct:

• Low: 142 → 97, a 31.7% drop
• Medium: 108 → 74, a 31.5% drop
• High: 82 → 56, a 31.7% drop
• Ultra: 78 → 54, a 30.8% drop

This consistent ~31% reduction in each setting is a hallmark of a GPU tied resolution-dependent fill. The drop matches the pixel count proportionally, reflecting a fill-limited scenario.

Going from 1080p to 4K (4x pixels, 1920x1080 vs 3840x2160), the frame rates drop to approximately 40% of the 1080p value:

• Low: 142 → 57, that’s 40.1% of the 1080p
• Medium: 108 → 43, 39.8%
• High: 82 → 33, 40.2%
• Ultra: 78 → 33, 39.7%

This is an almost perfect 39.9% average retention—exactly what you would expect if the GPU is fully utilized and if the 4K framerate is determined simply by 4X the pixel cost. So the data says: there is zero evidence of CPU influence in the scale. At the CPU's passmark single-thread 50 is high enough. All the resolution/resolution behavior follows a theoretical fill-rate.

The 1440p to 4K is a 2.25x pixel increase, and the measured average is 0.170% of 1440p as well: Low 57/97 = 58.8%, Medium 43/74=58.1%, High 33/56=58.9%, Ultra 31/54=57.4%. Again the ≈58% coherence of the CPU is irrelevant, the GPU is peak-limited.

The only deviation in the scaling is the tiny base effect at the Medium row where min/max data is available; at 4K, a 37 min vs 50 max means the no. That variance indicates memory pressure on the 6 GB buffering is not present, but 336 GB/s is at its upper bound.

---

Settings Recommendations

Given the four independent tests, the best "experience per the measured rows" changes with display target.

1080p at 60 Hz: Any preset—Low 4, Medium 36, High 82, Ultra 78—all win. If enjoying a high refresh, use 2K Low at 142 or Medium 108, but the High and Ultra also above 60. Selecting Medium is a smart compromise: 108 FPS average and a 91-frame minimum, leaving 60% headroom. Best pick: Medium.

1080p at 144 Hz: That is 60-top; only Low and Medium have averages above 120. Medium's 108 is below 144 but the 124 average max suggests? breaking: Regardless, use Low 142 for high frames. Best pick: Low.

1440p at 60 Hz: Medium 74 av, min 63 ensures 60 is never missed. High 56 average is below 60, but the GPU will dip. Therefore Medium is the only setting with a measured 60+ min. Best pick: Medium.

1440p at 100/144 Hz: High (56) and Ultra (54) are under expected refresh. Low pour (97) is the target at 100Hz. Best pick: Low or Medium.

4K at 56-60 Hz: no preset can guarantee 60 inclusive; the closest is Low 57. There are entirely different. If you want a 60 near, use Low. If you want visuals: Medium avg 43 with 37 min is the only acceptable; but vs 63 Hz possible. Best pick: Low for CPU often fps, Medium for quality.

4K for 30 Hz: Multiple settings work. Medium has 37 min and 50 max—so 30Hz with the GAF, or High 33. Ultra 63: 31 average. Medium has quality gain: Low 57, 43 vs 31 all above. I recommend Medium 4K for 30fps or require.

Overall, the data strongly supports recommending Medium as the value spot: it has average twice and a minimum among the best. The percentage delta from Very High to Ultra is small, except at 1080. The only setting that doubles visuals while staying in the rules is 1080p/1440p Medium.

Since the GPU results max at 78 FPS only at 1080p, that mixes completely: If you prefer high detail at 1080p, Ultra (78 FPS) is usable on a 60-75Hz screen; otherwise trade to 1440p Medium (74 FPS and min 63) for visibly higher textures, still near. The measured towers pin this as a GPU-bound system, so the lowest quality vs framerate decisions are on the RTX.

For anyone on a 200Hz screen, alternatives not plausible; 1080p Low 142 is just above 140Hz, and 1440p Low barely 97. For a standard 60Hz monitor, 1440p Medium is the best quality per row—average 74, min 63—and 4K is only a low experiment.

That's the conclusion: use Medium at 1440p for sway, Ultra/High at 1080p for looks, and only Low at 4K crosses 55.