Thief

Thief, a stealth title released in February 2014, presents an unusual case study when paired with modern mobile hardware: an Intel Core Ultra 7 155H handling compute duties alongside an NVIDIA GeForce RTX 5070 Mobile for rendering. The measured data in this database covers twelve configurations — three resolutions crossed with four quality presets — and the picture it paints is one of a decade-old game that no longer stresses contemporary mid-range components. That raises interesting questions about where the bottlenecks actually sit, and whether the CPU or the GPU deserves credit for the results.

CPU Role

The Intel Core Ultra 7 155H is a Meteor Lake mobile processor with 16 cores and 22 threads, a 3.80 GHz base clock and a 4.80 GHz boost clock, fabricated on Intel's 7 nm process. Its launch MSRP was $503, and it sits in the 83rd percentile of all CPUs in the database — a strong showing for a 28 W TDP mobile part. The hybrid design includes an Arc Xe-LPG 128EU integrated GPU, though in this combo the discrete RTX 5070 Mobile does the rendering work.

Does Thief need this much CPU? Almost certainly not. The game predates Meteor Lake by nearly a decade, and the benchmark scores indicate the 155H is far beyond what a 2014 stealth title demands. Its Cinebench R23 single-core score of 1743 and multi-core score of 15028 place it comfortably in enthusiast mobile territory, and its nearest rivals tell a story of remarkable parity: the Intel Core i5-14600T averages 32707 versus the 155H's 32697 — a 0% delta — while the AMD Ryzen AI 7 PRO 360 sits at 32662, just 0.1% behind. The AMD Ryzen 5 7400F at 32750 leads by 0.2%, and the Ryzen 7 PRO 6850H at 32812 leads by 0.4%. In other words, the 155H trades blows with desktop and premium mobile chips across a razor-thin band.

For Thief specifically, the single-threaded metrics matter most, since older engines rarely scale across 22 threads. A PassMark single-thread score of 3468 and a PassMark physics score of 1727 suggest the CPU will never be the ceiling here — a hypothesis the resolution scaling data supports later. The 24 MB shared L3 cache and 89.6 GB/s of dual-channel memory bandwidth are likewise generous headroom for a title of this vintage.

Settings Recommendations

The measured rows suggest a clear sweet spot: Medium at 2560x1440, which delivers 94 average FPS with a documented minimum of 80 and maximum of 109. That frame spread — under 30 FPS between floor and ceiling — indicates a stable, consistent experience rather than a benchmark average masking stutters.

At 1080p, the data argues for simply maxing the game out. Ultra runs at 100 average FPS, High at 105, Medium at 138 and Low at 168. The gap between High and Ultra is only 5 FPS, which is a classic signature of diminishing visual returns: the extra fidelity costs almost nothing in performance, so there is little reason to step down from Ultra on a 1080p panel.

Things change at 4K. Ultra averages 40 FPS and High 42 — close, but both below what most players would consider smooth for a stealth game where reaction timing matters. Medium at 4K recovers to 56 average FPS with a minimum of 47 and maximum of 64, making it the sensible 4K choice. Low pushes to 73 average FPS at 4K if raw fluidity is the priority, but for a game built around atmosphere and shadow detail, dropping to Low undercuts the experience the designers intended. The recommendation, per the data: Ultra or High at 1080p, Medium at 1440p and 4K.

Resolution Scaling

Comparing the High preset across resolutions tells the limiting-component story. At 1920x1080, Thief runs at 105 average FPS; at 2560x1440 it drops to 72; at 3840x2160 it falls to 42. That is roughly a 31% drop from 1080p to 1440p and about a 60% drop from 1080p to 4K. The Ultra preset follows the same curve: 100, then 68, then 40.

This steep, near-linear decline with pixel count is the signature of GPU limitation. If the Core Ultra 7 155H were the constraint, frame rates would flatten as resolution rose — the CPU workload does not change with pixel count, so a CPU-bound game shows similar FPS at all three resolutions. Instead, every preset loses a large fraction of its throughput as the render target quadruples from 1080p to 4K. The conclusion: the RTX 5070 Mobile is the gatekeeper in Thief, and the 155H has compute to spare at every tested setting.

There is one nuance worth questioning. Even at 1080p Ultra the frame rate stops at 100 — it does not climb to the 168 FPS seen at Low. Is that a GPU shader ceiling, or does the older engine hit its own synchronization limits? The data alone cannot say, but the tight clustering of High and Ultra results at every resolution (105 vs 100, 72 vs 68, 42 vs 40) suggests the top two presets are hitting the same wall, whichever component owns it.

How This Combo Ranks

Out of 3720 tested hardware combinations in Thief, the Core Ultra 7 155H + RTX 5070 Mobile pairing ranks 1975th. That places it almost exactly at the midpoint of the field — a slightly-below-median result for a game where the rankings are presumably dominated by combinations of stronger discrete GPUs.

Context matters when interpreting that rank. The RTX 5070 Mobile sits in the 75th percentile of all GPUs with an average benchmark score of 29928, and its nearest rivals include the RTX 3070 Ti (29945, a 0.1% gap), the RTX 2080 Ti (29783, 0.5% behind), the Radeon RX 6800 (30095, 0.6% ahead) and the RX 6700 (30433, 1.7% ahead). Those are solidly upper-mainstream desktop-class competitors, and the mobile 5070 holds its own against all of them within a couple of percentage points. Against a 2014 title, a mid-field rank of 1975 of 3720 therefore reads less as a weakness of this combo and more as evidence that Thief's rankings are compressed — the game is old enough that most modern GPUs clear 60 FPS, so rank ordering reflects fine margins rather than capability gaps.

GPU Role

The RTX 5070 Mobile is a Blackwell 2.0 part built on TSMC's 5 nm process, with the GB206 chip measuring 181 mm² and packing 21,900 million transistors at a density of 121.0M per mm². It runs a 907 MHz base clock and a 1425 MHz boost clock, with 8 GB of GDDR7 on a 128-bit bus delivering 384.0 GB/s of bandwidth at 1500 MHz memory clock (24 Gbps effective).

On paper, Thief cannot stretch this hardware. The card offers 4608 shading units, 144 TMUs, 48 ROPs, 36 RT cores and 144 tensor cores, with 13.13 TFLOPS of FP32 throughput, a 68.40 GPixel/s pixel rate and a 205.2 GTexel/s texture rate. A 2014 DirectX-era title — the GPU supports DirectX 12 Ultimate (12_2), Vulkan 1.4 and OpenGL 4.6, all backwards-compatible — simply does not push shader or texture loads that approach these figures, which explains why even 4K Ultra only dents performance to 40 FPS rather than collapsing it.

VRAM is the one speculative angle. At 8 GB, the card has finite capacity, but a game from 2014 was designed around far smaller framebuffers, so even 4K textures and High presets should fit comfortably. The measured data shows no cliff suggesting VRAM saturation: performance degrades smoothly with resolution and settings, not abruptly. Benchmark scores reinforce the card's standing — a PassMark G3D score of 20355, Geekbench OpenCL at 122238 and Vulkan at 116960, and GPU compute at 8279. The 50 W TDP is also notable: this is an efficient design, and Thief will never approach its power or thermal limits.

Measured FPS Breakdown

All twelve measured rows, exact figures:

1920x1080:

• Low: 168 average FPS
• Medium: 138 average FPS (min 117, max 158)
• High: 105 average FPS
• Ultra: 100 average FPS

2560x1440:

• Low: 124 average FPS
• Medium: 94 average FPS (min 80, max 109)
• High: 72 average FPS
• Ultra: 68 average FPS

3840x2160:

• Low: 73 average FPS
• Medium: 56 average FPS (min 47, max 64)
• High: 42 average FPS
• Ultra: 40 average FPS

A few patterns deserve attention. First, the Low-to-Ultra spread at 1080p is 68 FPS (168 down to 100), but at 4K it shrinks to 33 FPS (73 down to 40) — settings matter less as resolution climbs, because the pixel workload dominates. Second, Medium is the only preset with recorded minimum and maximum values, which means it is the only configuration where frame consistency can actually be judged; the others report averages only. Third, Medium's 1080p minimum of 117 is higher than Ultra's 1080p average of 100, a striking illustration of how much performance the top presets cost relative to the visual step up.

The 60 FPS threshold — a common fluidity bar — is met or exceeded in eight of twelve configurations. Only 4K High (42), 4K Ultra (40) and, marginally, 4K Medium (56 average, 47 minimum) fall short. Everything at 1080p and 1440p clears it with room to spare.

FAQ

Q: Can this setup run Thief at 4K?

A: Yes, with settings adjustments. 4K Medium averages 56 FPS (min 47, max 64) and 4K Low averages 73 FPS, while 4K High (42) and Ultra (40) fall short of 60.

Q: What is the best settings choice for high refresh rates?

A: 1080p Low reaches 168 average FPS, and 1080p Medium reaches 138 with a 158 maximum — the highest frame rates in the dataset.

Q: Is the CPU or GPU the bottleneck in Thief?

A: The data points to the GPU. Frame rates fall sharply as resolution rises (105 to 72 to 42 at High), the classic pattern of GPU limitation, while the Core Ultra 7 155H sits in the 83rd CPU percentile with far more capability than a 2014 title requires.

Q: How does this hardware combination rank in Thief?

A: It ranks 1975 out of 3720 tested combinations — essentially mid-field, despite the GPU sitting in the 75th percentile of all GPUs and trading blows with the RTX 3070 Ti and RX 6800.

Q: Is 8 GB of VRAM enough for Thief?

A: The measured data shows no sign of VRAM trouble — performance degrades smoothly from 1080p to 4K rather than falling off a cliff, and a 2014 game was built for far smaller framebuffers than the RTX 5070 Mobile's 8 GB of GDDR7.

Q: Does dropping from Ultra to High gain much performance?

A: Only marginally: 5 FPS at 1080p (100 to 105), 4 FPS at 1440p (68 to 72) and 2 FPS at 4K (40 to 42). The bigger gains come from stepping down to Medium.