Thief

# CPU Role — cores, clocks, and how they relate to this game's results

The AMD Ryzen 9 8945HX is a 16-core, 32-thread mobile processor built on the Zen 4 architecture (Dragon Range codename) using TSMC's 5 nm process. It carries a base clock of 2.50 GHz and a boost clock of 5.40 GHz, with a 55 W TDP. The chip features 64 KB of L1 cache per core, 1 MB of L2 per core, and a 64 MB L3 cache. Its memory support is DDR5 over a dual-channel bus, providing 83.2 GB/s of bandwidth, and it connects via PCIe Gen 5 with 28 CPU lanes. The integrated graphics is a Radeon 610M.

In the context of Thief, a stealth title from 2014, the CPU's role is primarily to feed the GPU with draw calls and maintain frame pacing. The benchmark data shows this processor is not a bottleneck at any tested resolution. At 1080p with Low settings, the system reaches 184 FPS average; at 4K with Low settings, it still manages 74 FPS. The scaling from 1080p to 4K at Low settings shows a 110 FPS drop, which indicates the GPU is taking over as the limiting factor as resolution increases. If the CPU were the constraint, we would expect similar FPS across resolutions at the same settings — that is not the case here.

The processor's multi-threaded capabilities are substantial. In Cinebench R23, it scores 42713 multi-core and 6030 single-core. PassMark multi-thread score is 51405, with a single-thread score of 3907. The average benchmark score across all tests is 76212, placing it in the 95th percentile of all CPUs. Compared to its nearest rivals, the Intel Core Ultra 9 285HX scores 76155 (0.1% lower), the AMD EPYC Embedded 8224P scores 76492 (0.4% higher), the AMD Ryzen Threadripper PRO 9945WX scores 76513 (0.4% higher), and the AMD Ryzen 9 9950X3D scores 75779 (0.6% lower). The Ryzen 9 8945HX sits essentially at parity with these high-end parts, meaning its 16 cores and 32 threads are more than adequate for a game like Thief, which was designed for much older and weaker hardware.

Thief is not a heavily threaded title by modern standards. The game's engine was built around the era of quad-core CPUs, so the Ryzen 9 8945HX's extra cores are largely idle during gameplay. However, the high single-thread performance — a Cinebench R23 single-core score of 6030 — ensures that the game's primary simulation thread runs without stutter. The boost clock of 5.40 GHz is particularly relevant here, as Thief's frame generation relies on rapid, low-latency CPU responses. The data confirms this: even at 4K Ultra, the system maintains a playable 41 FPS, and at 1080p Ultra, it hits 101 FPS. These numbers would not be possible if the CPU were struggling to keep pace.

The CPU's memory bandwidth of 83.2 GB/s over dual-channel DDR5 is also sufficient for Thief's asset streaming. The game loads level geometry and textures into memory, and with 32 threads available, any background tasks are handled without competing with game threads. The 64 MB L3 cache further helps by keeping frequently accessed data close to the cores, reducing latency. In benchmark terms, the PassMark data compression score of 677755 and floating-point math score of 117453 show strong computational throughput, but these are general workloads — the gaming relevance is that the CPU never becomes a limiting factor in the measured FPS rows.

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

The NVIDIA GeForce RTX 4070 is built on the Ada Lovelace architecture with the AD104 chip on TSMC's 5 nm process. It has 5888 shading units, 184 texture mapping units, and 64 raster output units. The GPU operates at a base clock of 1920 MHz and boosts to 2475 MHz. Memory consists of 12 GB of GDDR6X on a 192-bit bus, delivering 504.2 GB/s of bandwidth. The card supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. Its pixel rate is 158.4 GPixel/s, texture rate is 455.4 GTexel/s, and FP32 performance is 29.15 TFLOPS. The TDP is 200 W.

For Thief, the RTX 4070 is the dominant performance driver. The game's measured FPS scales directly with GPU load as settings and resolution increase. At 1080p Low, the card produces 184 FPS; at 4K Ultra, it drops to 41 FPS. This 143 FPS range across the tested configurations demonstrates that the GPU is the primary component determining frame rates. The 12 GB VRAM is ample for Thief, as the game's textures are modest by 2025 standards — the VRAM never becomes a constraint, even at 4K Ultra.

The GPU's benchmark scores position it in the 81st percentile of all GPUs, with an average benchmark score of 37648. Its nearest rivals are the NVIDIA Tesla P4 (37628, 0.1% lower), AMD Radeon RX Vega 56 (37507, 0.4% lower), NVIDIA GeForce RTX 4080 Mobile (38135, 1.3% higher), and AMD Radeon PRO W6400 (37157, 1.3% lower). The RTX 4070 sits close to the RTX 4080 Mobile, which is notable because the desktop 4070 is being compared against a high-end laptop part. In 3DMark Steel Nomad DX12, the card scores 3854, and in Geekbench Vulkan it reaches 174152. The PassMark G3D score is 26927, with a GPU compute score of 14720.

In Thief specifically, the GPU's clock behavior matters. The boost clock of 2475 MHz allows the card to push high frame rates at lower settings. At 1080p Low, the 184 FPS average indicates the GPU is running at or near its boost clock with minimal thermal throttling. At 4K Ultra, the 41 FPS average shows the card is fully saturated, with the memory bandwidth of 504.2 GB/s becoming a factor. Thief's engine was not designed for such high resolutions, and the RTX 4070's 192-bit memory bus is sufficient but not overkill for 4K.

The RTX 4070's architecture includes 46 RT cores and 184 tensor cores, though Thief is a 2014 title that does not use ray tracing or DLSS. This means those specialized units are idle, and the card relies purely on its rasterization performance. The 64 ROPs handle pixel output, and at 4K Ultra the pixel rate of 158.4 GPixel/s is the theoretical ceiling — the 41 FPS average is well below that ceiling, indicating that shader work and memory latency are the limiting factors rather than raw pixel throughput.

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

The measured FPS data across resolutions reveals a clear pattern of GPU limitation. At Low settings, the average FPS drops from 184 at 1080p to 126 at 1440p (a 31.5% reduction), then to 74 at 4K (a 41.3% reduction from 1440p). This is a classic GPU-bound scaling curve — as pixel count increases, the GPU's fill rate and memory bandwidth become the bottlenecks. If the CPU were the constraint, we would see flat FPS across resolutions.

At High settings, the drop from 106 FPS at 1080p to 73 FPS at 1440p is a 31.1% reduction, and then to 43 FPS at 4K is a 41.1% reduction from 1440p. The percentages are remarkably consistent across Low and High settings, suggesting the GPU is operating in a similar efficiency envelope. At Medium settings, the drop from 140 FPS at 1080p to 96 FPS at 1440p is 31.4%, and then to 56 FPS at 4K is a 41.7% reduction. Ultra settings show 101 FPS at 1080p, 70 FPS at 1440p (30.7% less), and 41 FPS at 4K (41.4% less from 1440p).

The consistency of these percentage drops — roughly 31% from 1080p to 1440p and 41% from 1440p to 4K — is the signature of a GPU-bound scenario. The RTX 4070's performance scales predictably with pixel count, and the CPU keeps up without introducing frame time spikes. The fact that the 4K to 1440p drop is larger than the 1440p to 1080p drop reflects the exponential nature of pixel counts: 4K has 2.25 times the pixels of 1440p, and 1440p has 1.78 times the pixels of 1080p. The FPS drops are not proportional to pixel count, which indicates that memory bandwidth and shading complexity are the limiting factors rather than pure pixel throughput.

At 4K, the gap between Low (74 FPS) and Ultra (41 FPS) is 33 FPS. At 1080p, the same settings gap is 83 FPS (184 vs 101). This narrowing of the settings gap at higher resolutions is another GPU-bound indicator — at 4K, the GPU is so saturated that additional settings have less relative impact. At 1080p, the GPU has headroom, so increasing settings from Low to Ultra costs more absolute FPS.

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

1080p (1920x1080):

• Low: 184 FPS average
• Medium: 140 FPS average, 119 min, 161 max
• High: 106 FPS average
• Ultra: 101 FPS average

At 1080p, the system delivers exceptionally high frame rates. The jump from Low to Medium costs 44 FPS, while Medium to High costs 34 FPS, and High to Ultra only 5 FPS. This suggests that Ultra settings at 1080p are still GPU-bound but with diminishing returns — the difference between High and Ultra is minimal, making Ultra the better choice for visual quality without sacrificing much performance.

1440p (2560x1440):

• Low: 126 FPS average
• Medium: 96 FPS average, 82 min, 110 max
• High: 73 FPS average
• Ultra: 70 FPS average

At 1440p, the frame rates remain high. Low to Medium costs 30 FPS, Medium to High costs 23 FPS, and High to Ultra costs only 3 FPS. The High to Ultra gap is negligible, indicating that Ultra is effectively free at this resolution. The Medium settings show a 28 FPS range between min and max, which is the widest of any configuration, suggesting some scene complexity variation.

4K (3840x2160):

• Low: 74 FPS average
• Medium: 56 FPS average, 48 min, 65 max
• High: 43 FPS average
• Ultra: 41 FPS average

At 4K, the system becomes playable but not high-refresh-rate. Low settings achieve 74 FPS, which is smooth, but Ultra drops to 41 FPS, which is at the edge of playability for a stealth game. The Medium settings have a 17 FPS min-to-max range, showing that 4K scenes vary significantly in GPU load. The High to Ultra drop is only 2 FPS, reinforcing that Ultra is almost free over High at this resolution.

The combo rank for this CPU-GPU pair in Thief is 1407 out of 2851 combinations, placing it in the top half of all tested configurations. This mid-pack ranking reflects that while the hardware is strong, there are faster combinations available, and Thief's older engine does not fully utilize the modern hardware's capabilities.

# FAQ

Q: What is the best resolution to play Thief with this hardware?

A: The data shows 1440p as the sweet spot. At 1440p High, the system achieves 73 FPS average, and at 1440p Ultra, it hits 70 FPS. This provides high visual quality with smooth frame rates. 1080p is faster but wastes the GPU's headroom, while 4K at Ultra drops to 41 FPS, which is below the ideal smoothness threshold.

Q: How much performance is lost by going from 1080p to 4K?

A: At Low settings, the drop is from 184 FPS to 74 FPS, a 110 FPS loss. At Ultra settings, the drop is from 101 FPS to 41 FPS, a 60 FPS loss. The percentage reduction is consistent across settings at roughly 60% less FPS from 1080p to 4K, confirming the GPU is the limiting factor.

Q: Is the Ryzen 9 8945HX CPU a bottleneck in Thief?

A: No. The benchmark data shows FPS scaling with resolution and settings, which is the opposite of CPU bottleneck behavior. If the CPU were limiting, frame rates would be similar across resolutions. The CPU's 95th percentile ranking and high single-thread score of 6030 in Cinebench R23 ensure it can handle Thief's simulation load without issue.

Q: Does the RTX 4070's 12 GB VRAM cause any problems in Thief?

A: No. Thief's textures are from 2014 and are not demanding by modern standards. The measured FPS at 4K Ultra (41 FPS) shows no VRAM-related stutter or crashes. The 504.2 GB/s memory bandwidth is sufficient for the game's asset streaming.

Q: What is the difference between High and Ultra settings in Thief?

A: The performance cost is minimal. At 1080p, High gives 106 FPS and Ultra gives 101 FPS (5 FPS difference). At 1440p, the difference is 3 FPS (73 vs 70). At 4K, it is 2 FPS (43 vs 41). The visual quality increase from High to Ultra is worth the tiny performance cost.

Q: How does this CPU-GPU combo rank against other systems in Thief?

A: The combination ranks 1407 out of 2851 tested configurations, putting it in the top half. While not the fastest pairing available, the Ryzen 9 8945HX and RTX 4070 deliver playable frame rates at all tested settings and resolutions, with 1080p and 1440p being particularly strong.

# Settings Recommendations

The data provides clear guidance on the optimal settings for each resolution. At 1080p, the system has so much headroom that Ultra settings are recommended. The 101 FPS average at Ultra is still high, and the 5 FPS difference from High is negligible. Players with 1080p displays should enable Ultra settings for the best visual experience without sacrificing smoothness.

At 1440p, Ultra settings are again the best choice. The 70 FPS average is above the 60 FPS threshold for smooth gameplay, and the 3 FPS difference from High is imperceptible. The Medium settings at 1440p show a wider min-max range (82 to 110 FPS), indicating that Ultra provides more consistent frame pacing. For players who want extra headroom, High at 73 FPS is acceptable, but Ultra is the recommended preset.

At 4K, the choice depends on the player's tolerance for lower frame rates. Ultra at 41 FPS is playable for a stealth game, which relies more on careful movement than twitch reflexes. However, the Medium preset at 56 FPS provides a safer margin above 50 FPS, with a min of 48 FPS that stays above the 45 FPS threshold for most players. The High preset at 43 FPS is barely better than Ultra, so the choice is between Medium (56 FPS) and Ultra (41 FPS). For a game like Thief, where atmosphere and lighting matter, Ultra at 41 FPS may be worth it, but Medium is the safer recommendation for consistent frame times.

The overall best experience across the board is 1440p Ultra, which delivers 70 FPS average. This resolution balances image quality with performance, and the Ultra preset costs almost nothing over High. For 1080p users, Ultra is the clear winner. For 4K users, Medium provides the best balance of visual fidelity and frame rate, with Ultra as an option for those who prioritize graphics over smoothness. The measured data shows that the RTX 4070 is capable of running Thief at high quality at 1080p and 1440p, while 4K requires a settings compromise.