Thief

Thief, a stealth classic from 2014, places specific demands on modern hardware. When paired with the Intel Core Ultra 5 135H and AMD Radeon 890M, the data reveals a configuration that is strictly bound by its integrated graphics. The measured FPS across resolutions and settings tells a clear story: this combo is designed for lower resolutions and moderate settings, with performance scaling that highlights the GPU as the primary bottleneck.

Resolution Scaling

The resolution scaling behavior of this configuration is definitive. At 1080p with Low settings, the system achieves an average of 77 FPS. Moving to 1440p Low drops that average to 53 FPS, and at 4K Low it falls to 31 FPS. This represents a 31% performance loss from 1080p to 1440p and a further 42% loss from 1440p to 4K. Such steep declines are characteristic of a system where the GPU is overwhelmed by pixel count.

The pattern persists across all settings presets. At High settings, the averages are 44 FPS at 1080p, 30 FPS at 1440p, and 18 FPS at 4K. The gap between 1080p and 4K at High is 26 FPS, a 59% reduction. Even at the lowest settings, the 4K average of 31 FPS is barely playable, indicating that resolution scaling is the dominant factor in performance degradation.

The data suggests that the limiting component is unequivocally the GPU. The AMD Radeon 890M, with its system-shared memory and 32 ROPs, is tasked with rendering increasingly large frames. The CPU, a 14-core Meteor Lake part with impressive synthetic scores, is not the constraint here. If the CPU were the bottleneck, we would expect to see less severe scaling between resolutions, as the workload would remain constant while the resolution changed. Instead, the dramatic FPS drops with each resolution increase point directly to the integrated graphics being saturated.

At 4K, the Ultra preset yields only 17 FPS average, making the game essentially unplayable. The Medium preset at 4K offers a slightly better 23 FPS average but with minimums of 20 FPS, which would result in noticeable stutter. The data indicates that 4K is simply not a viable target for this configuration, regardless of settings. The performance scaling is linear and severe, reinforcing the conclusion that the Radeon 890M's pixel throughput is the hard ceiling.

CPU Role

The Intel Core Ultra 5 135H is a mobile processor with 14 cores and 18 threads, based on the Meteor Lake architecture. Its benchmark results are respectable, with a Cinebench R23 multi-core score of 11875.5 and a single-core score of 1700. In PassMark, it achieves a multi-thread score of 22422 and a single-thread score of 3521. These figures place the CPU in the 81st percentile among all CPUs, indicating solid multi-threaded performance.

However, in Thief, the CPU's role appears secondary. The game is from 2014 and was designed for the hardware of that era. While the Core Ultra 5 135H's single-thread performance is strong, the measured FPS data shows that even at 1080p Low, where the CPU load is relatively higher compared to the GPU, the system only reaches 77 FPS. This is a respectable number but not exceptional for a modern processor, suggesting that the integrated GPU is already limiting the frame rate even at the lowest resolution.

The nearest rivals for the CPU in terms of average benchmark score include the AMD Ryzen 7 5800X with a delta of -0.1% and the Intel Core i5-13500HX with a delta of -0.6%. These comparisons show that the Core Ultra 5 135H is performing at a level consistent with these desktop and high-end mobile parts in synthetic tests. Yet in Thief, the bottleneck is clearly not the CPU. The 14 cores and 18 threads are more than sufficient for a game of this era, and the boost clock of 4.60 GHz provides adequate single-threaded headroom.

The data implies that the CPU is never the limiting factor in this configuration. Even at 4K Ultra, where the average FPS drops to 17, the CPU is likely idle waiting for the GPU to finish rendering frames. The fact that the 1080p Low setting produces 77 FPS, which is a high frame rate but not a triple-digit one, further suggests that the GPU is holding back performance. A more powerful discrete GPU would likely push this number much higher, given the CPU's capabilities.

GPU Role

The AMD Radeon 890M is an integrated graphics processor based on the RDNA 3.5 architecture, built on a 4 nm process at TSMC. It features 1024 shading units, 64 TMUs, and 32 ROPs, with a boost clock of 2900 MHz. Its theoretical pixel rate is 92.80 GPixel/s and its FP32 performance is 5.939 TFLOPS. The GPU has a TDP of 15 W and uses system-shared memory, meaning its bandwidth is system-dependent.

In synthetic benchmarks, the 890M scores 8076 in PassMark G3D and 40808 in Geekbench Vulkan. Its 3DMark Steel Nomad DX12 score is 590. These scores place it in the 45th percentile among all GPUs, indicating mid-range performance for an integrated solution. Its nearest rivals include the AMD Radeon Vega 8 with a delta of -0.1% and the NVIDIA GeForce GTX 960 with a delta of -0.7%. This suggests the 890M is competitive with older discrete graphics cards, but those are entry-level parts from a decade ago.

The GPU's role in Thief is the primary determinant of performance. The measured FPS data shows that the 890M struggles with higher resolutions. At 1080p Medium, it produces 58 FPS average, which is playable but not smooth. The minimum FPS of 49 at this setting indicates occasional dips that could be noticeable. At 1080p High, the average falls to 44 FPS, and at Ultra, it drops to 42 FPS.

The VRAM situation is critical. The GPU uses system-shared memory, so its bandwidth is entirely dependent on the system's memory configuration. For a game like Thief, which was released in 2014, the memory requirements are moderate, but the shared memory architecture can still cause performance issues if the system memory bandwidth is insufficient. The data suggests that the GPU's compute and pixel throughput are the limiting factors, not the memory, as the scaling between settings at a fixed resolution is relatively modest compared to the scaling between resolutions.

FAQ

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

A: The data shows that 1920x1080 is the only resolution where the game remains comfortably playable. At 1080p, even the Ultra preset achieves an average of 42 FPS, while at 1440p, the High preset drops to 30 FPS and at 4K, the Low preset only manages 31 FPS.

Q: How does the Intel Core Ultra 5 135H compare to its nearest rivals?

A: The CPU's average benchmark score of 29093 is nearly identical to the AMD Ryzen 7 5800X, which scores 29123, a delta of -0.1%. It also compares closely to the Intel Core i5-13500HX, which scores 29270, a delta of -0.6%.

Q: Is the AMD Radeon 890M better than a dedicated GPU?

A: Based on its PassMark G3D score of 8076, the 890M is on par with the NVIDIA GeForce GTX 960, which scores 9273, a delta of -0.7%. It is also very close to the AMD Radeon Vega 8, which scores 9221, a delta of -0.1%.

Q: What is the highest playable setting at 1080p?

A: The data indicates that the Ultra preset at 1080p produces an average of 42 FPS, which is technically playable. However, the Medium preset at 58 FPS provides a more comfortable margin for smooth gameplay.

Q: Why does performance drop so sharply at 4K?

A: The Radeon 890M has 32 ROPs and a pixel rate of 92.80 GPixel/s. Rendering at 3840x2160 requires processing over 8 million pixels per frame, which is four times the workload of 1080p. The GPU's pixel throughput is insufficient for this task, leading to averages as low as 17 FPS at Ultra.

Q: Does the CPU ever become a bottleneck in Thief?

A: No. The CPU's Cinebench R23 multi-core score of 11875.5 and single-core score of 1700 indicate strong performance. The fact that FPS scales linearly with resolution changes, rather than remaining constant, proves the GPU is the limiting component.

Measured FPS Breakdown

The measured data provides a complete picture of performance across all combinations. At 1920x1080, the system delivers 77 FPS average on Low, 58 FPS on Medium (with a range of 49-67 FPS), 44 FPS on High, and 42 FPS on Ultra. This shows a clear degradation as settings increase, but all presets remain at least marginally playable.

At 2560x1440, the averages are 53 FPS on Low, 40 FPS on Medium (with a range of 34-46 FPS), 30 FPS on High, and 29 FPS on Ultra. The drop from Low to Ultra at this resolution is 24 FPS, a 45% reduction. The Medium preset at 1440p is the only one that maintains a consistent experience above 30 FPS.

At 3840x2160, the performance becomes marginal. The Low preset yields 31 FPS average, Medium yields 23 FPS (with a range of 20-27 FPS), High yields 18 FPS, and Ultra yields 17 FPS. None of these are suitable for a smooth gaming experience. The step from Low to Medium at 4K represents a 26% performance loss, and from Medium to High, another 22% loss.

The data also shows that the gap between High and Ultra presets is narrow across all resolutions. At 1080p, the difference is only 2 FPS (44 vs 42). At 1440p, it is 1 FPS (30 vs 29). At 4K, it is 1 FPS (18 vs 17). This suggests that the Ultra preset offers minimal visual benefit over High while costing the same performance, making it an inefficient choice for this hardware.

Settings Recommendations

Based on the measured FPS data, the optimal experience for this configuration is at 1080p with the Medium preset. This combination yields an average of 58 FPS with a minimum of 49 FPS, providing a consistently smooth frame rate that stays above the critical 30 FPS threshold. The Low preset at 1080p offers 77 FPS, but the visual quality reduction is substantial, and the Medium preset provides a better balance.

For users who prioritize visual fidelity over frame rate, the 1080p High preset at 44 FPS is a viable option. It is playable, though it lacks the headroom of the Medium preset. The Ultra preset at 1080p, at 42 FPS, offers no meaningful improvement over High and is therefore not recommended.

At 1440p, the only acceptable preset is Low, which produces 53 FPS. The Medium preset at 40 FPS is borderline, but its minimum of 34 FPS could lead to noticeable stutters in demanding scenes. The High and Ultra presets at 1440p, with averages of 30 and 29 FPS respectively, are not recommended for consistent gameplay.

At 4K, no preset provides a playable experience. The Low preset at 31 FPS is technically above 30 FPS, but the lack of a minimum FPS measurement suggests potential instability. The Medium preset at 23 FPS average with 20 FPS minimums is too inconsistent for a stealth game that requires precision. Users should avoid 4K entirely with this hardware.

The data strongly indicates that the 1080p Medium preset is the sweet spot. It offers a 58 FPS average, which is 15 FPS higher than the High preset and only 19 FPS lower than Low. This provides the best balance of visual quality and performance, ensuring that the game remains responsive and enjoyable. The combination of the Intel Core Ultra 5 135H and AMD Radeon 890M is best suited for 1080p gaming, and the Medium preset maximizes the potential of this integrated graphics solution.