Thief

Thief paired with the Intel Core Ultra 7 165H and NVIDIA GeForce RTX 5080 is a pairing that leans almost entirely on the GPU at high resolutions and on the CPU's single-thread speed at low ones — and the data captures that crossover cleanly. At 1080p Low the combo reaches 171 FPS average, while at 4K Ultra it settles to 56 FPS. Across all twelve measured configurations the RTX 5080 (passmark_g3d score 36565, 87th percentile among all GPUs) delivers smooth playable results in a stealth title that rewards consistent frame delivery, and the Core Ultra 7 165H (84th percentile among all CPUs) provides enough single-thread throughput — a Cinebench R23 single-core score of 1759 — to keep the frame rate climbing as resolution drops. This is a measured dataset, not estimated, and it shows a classic GPU-bound profile that shifts partially CPU-bound at 1080p.

Resolution Scaling

The scaling story in Thief is told best by the Medium preset, the only setting tier with full min/avg/max data at every resolution. At 1920x1080 Medium the combo averages 154 FPS, with a minimum of 131 and a maximum of 178. Move to 2560x1440 Medium and the average drops to 130 FPS — a loss of 24 FPS for roughly double the pixel count — with a range of 111 to 150. At 3840x2160 Medium the average falls to 77 FPS, with a minimum of 65 and a maximum of 88.

The pattern is diagnostic. From 1440p to 4K, the Medium average loses 53 FPS — a far steeper drop than the 24 FPS lost between 1080p and 1440p. That accelerating decline indicates the RTX 5080's shading and memory resources are being saturated as pixel count quadruples from 1080p to 4K. Conversely, the shallow 1080p-to-1440p falloff suggests the CPU is partially the limiter at the lower resolution: the frame rate cannot scale much beyond roughly 150-170 FPS even when the GPU load is halved, which lines up with the Core Ultra 7 165H's single-thread profile rather than any GPU shortfall.

The Low preset makes the ceiling explicit: 171 FPS at 1080p, 151 FPS at 1440p, and 101 FPS at 4K. Only a 20 FPS drop between 1080p and 1440p at Low confirms a CPU/platform ceiling around the low-170s; the GPU only becomes the true governor once the resolution reaches 4K, where Low still clears 100 FPS. At the Ultra preset the curve flattens in the other direction: 138 FPS at 1080p versus 95 FPS at 1440p versus 56 FPS at 4K. The 1080p-to-1440p Ultra drop of 43 FPS is nearly as large as the 1440p-to-4K drop of 39 FPS — a signature of heavy GPU-bound rendering with little CPU involvement at that tier.

GPU Role

The RTX 5080 is the engine of this pairing, and the numbers justify that role. Its passmark_g3d score of 36565 places it in the 87th percentile of all GPUs tested in the database, with an overall average benchmark score of 56083. Against its nearest rivals, the margins are tight: the AMD Radeon 8060S sits at 55757 (0.6 percent behind), the RX 6750 GRE 12 GB at 55698 (0.7 percent behind), the Radeon Pro W5700X at 54828 (2.3 percent behind), and the RX 9070 GRE at 57367 (2.2 percent ahead). In practice this is a cluster of statistically equivalent GPUs, and Thief's results should be read as representative of that performance band.

The hardware behind those scores: 10752 shading units, 336 TMUs, 112 ROPs, 84 RT cores, and 336 tensor cores on the GB203 chip, built on TSMC's 5 nm process with 45,600 million transistors across a 378 mm² die. Boost clock is 2617 MHz from a 2295 MHz base, with GDDR7 memory running at 1875 MHz (30 Gbps effective) over a 256-bit bus — 16 GB of it, with 960.0 GB/s of bandwidth. The card's texture rate of 879.3 GTexel/s and pixel rate of 293.1 GPixel/s are more than enough for a 2014 stealth title, which is why even 4K High runs at 58 FPS average and 4K Low breaks 100 FPS. The 56.28 TFLOPS FP32 figure dwarfs what Thief's rendering load demands; the constraint at 4K Ultra (56 FPS) is memory and fill-rate pressure at extreme pixel counts, not raw compute capacity. A 360 W TDP, single 16-pin power connector, dual-slot footprint, and PCIe 5.0 x16 interface round out the card's profile, with DirectX 12 Ultimate, Vulkan 1.4, and OpenGL 4.6 API support ensuring full compatibility with Thief's renderer.

CPU Role

The Core Ultra 7 165H is a Meteor Lake mobile processor with 16 cores and 22 threads, a 3.80 GHz base clock, a 5.00 GHz boost clock, and a 28 W TDP on the Intel BGA 2049 socket. Built on Intel's 7 nm process, it carries 112 KB of L1 per core, 2 MB of L2 per core, and a shared 24 MB L3, with dual-channel DDR5 support delivering 89.6 GB/s of memory bandwidth. Its launch MSRP was $460, and its integrated Arc Xe-LPG 128EU graphics are irrelevant here with an RTX 5080 in the loop.

In Thief, what matters is the single-core score: Cinebench R23 single-core of 1759, R20 single-core of 1303, R15 single-core of 257, and a PassMark single-thread result of 3509. That per-thread throughput is what defines the ~170 FPS ceiling visible at 1080p Low and explains why Medium only climbs from 77 to 154 FPS between 4K and 1080p instead of scaling linearly. The multi-core credentials — 14551 in Cinebench R23 multi, 9233 in R20 multi, 2641 in R15 multi, and 25933 in PassMark multithread — matter less for a stealth game of this era, which draws its frame-time profile from a handful of hot threads handling game logic, AI, and draw-call submission.

Contextually, the 165H sits in the 84th percentile of all CPUs with an average benchmark score of 34083. Its nearest rivals cluster within a fraction of a percent: the Intel Core Ultra 5 338H at 33989 (0.3 percent behind), the AMD EPYC 4244P at 34220 (0.4 percent ahead), the AMD Ryzen AI 7 350 at 34222 (0.4 percent ahead), and the AMD Ryzen 7 3700X at 34260 (0.5 percent ahead). Any of these chips would produce statistically indistinguishable Thief results with the RTX 5080; the pairing's rank is driven far more by the GPU than the CPU.

FAQ

Q: Can this combo run Thief at 4K above 60 FPS?

A: Yes, at Low and Medium. 4K Low averages 101 FPS and 4K Medium averages 77 FPS (minimum 65, maximum 88). At 4K High the average is 58 FPS, just under 60, and 4K Ultra averages 56 FPS.

Q: What is the highest frame rate recorded for this pairing?

A: 171 FPS average at 1920x1080 on Low settings, with a recorded maximum of 178 FPS at 1080p Medium — the only preset where max/min data was captured alongside the 154 FPS average.

Q: Is Thief CPU-limited or GPU-limited with this hardware?

A: Both, depending on resolution. At 1080p, the shallow scaling between Low (171 FPS) and Ultra (138 FPS) and the small drop to 1440p Low (151 FPS) point to a CPU ceiling near the low-170s. At 1440p and especially 4K, the steep average declines indicate GPU-bound operation.

Q: How does the RTX 5080 compare to rival GPUs in this database?

A: Its average benchmark score of 56083 places it 0.6-2.3 percent ahead of the AMD Radeon 8060S, RX 6750 GRE 12 GB, and Radeon Pro W5700X, and 2.2 percent behind the RX 9070 GRE.

Q: How does the Core Ultra 7 165H compare to its nearest CPU rivals?

A: Its average score of 34083 is 0.3 percent ahead of the Core Ultra 5 338H, and within half a percent of the AMD EPYC 4244P, Ryzen AI 7 350, and Ryzen 7 3700X — effectively a four-way tie.

Q: Does VRAM capacity matter for Thief on this setup?

A: The RTX 5080's 16 GB of GDDR7 with 960.0 GB/s of bandwidth comfortably exceeds what Thief demands at any tested resolution, including 4K Ultra. The 4K results reflect pixel-throughput limits, not VRAM exhaustion.

How This Combo Ranks

Across 3720 tested hardware combinations in Thief, this pairing ranks 962nd — roughly the top quarter of the database. That placement is consistent with the component percentiles: an 87th-percentile GPU and an 84th-percentile CPU should land in the upper tier, and they do. The rank also reflects Thief's age; the game released on February 24, 2014, and its engine does not exploit the RTX 5080's full feature set, so combinations with even faster GPUs at 1080p and 1440p pull ahead on raw ceiling. The tight rival clusters on both sides — four GPUs within 2.3 percent and four CPUs within 0.5 percent of the respective component averages — mean the 962nd rank represents a broad band of equivalent hardware, not a sharp differentiator. Any combo substituting a listed rival for either part would be expected to land within a handful of ranks of this one.

Measured FPS Breakdown

All figures below are measured average FPS; min/max values are included where the dataset records them.

1920x1080:

• Low — 171 FPS
• Medium — 154 FPS (min 131, max 178)
• High — 140 FPS
• Ultra — 138 FPS

The 1080p spread is narrow: just 33 FPS separates Low from Ultra. That compression confirms the CPU, not the GPU, governs this resolution. The Ultra average of 138 FPS sits only 2 FPS below High, meaning top-tier image quality is essentially free at 1080p with this GPU.

2560x1440:

• Low — 151 FPS
• Medium — 130 FPS (min 111, max 150)
• High — 99 FPS
• Ultra — 95 FPS

1440p begins to tax the GPU. The Low-to-Medium drop of 21 FPS is modest, but Medium to High costs 31 FPS as heavier shading load kicks in. High (99 FPS) and Ultra (95 FPS) both fall just short of a 144 Hz refresh target, and both comfortably exceed 60 FPS.

3840x2160:

• Low — 101 FPS
• Medium — 77 FPS (min 65, max 88)
• High — 58 FPS
• Ultra — 56 FPS

4K is where the RTX 5080 works hardest. Low still clears 100 FPS — an unusually strong result for native 4K — and Medium's 77 FPS average with a 65 FPS minimum remains comfortably smooth. High at 58 FPS and Ultra at 56 FPS hover just below 60, separated by only 2 FPS, suggesting Ultra's additional effects are nearly free once the card is saturated at High.

Verdict by use case: for high-refresh 1080p play the combo delivers 138-171 FPS across all presets; for 1440p it delivers 95-151 FPS; and for native 4K it delivers 56-101 FPS depending on settings, with Medium as the sweet spot that keeps the average above 75 FPS at every resolution tested. The data marks this as a GPU-driven configuration whose 84th-percentile mobile CPU sets a firm but adequate frame-rate ceiling at lower resolutions.