Thief

The measured performance of Thief, the 2014 stealth game, on an Intel Core i5-13400F coupled with an NVIDIA GeForce RTX 4090 shows a system with substantial headroom. The dataset covers three resolutions (1920x1080, 2560x1440, 3840x2160) and four quality presets (Low, Medium, High, Ultra), with frame rates ranging from a low of 80 FPS at 4K Ultra to a high of 178 FPS at 1080p Low. This combination ranks in the top 10% in a per-kernel comparison, landing at position 379 out of a total of 3,720 tested combinations. Both components are high-end offerings from their respective generations, and the measured results reveal how they interact under different graphic loads.

FAQ

Q: What is the average FPS at 1080p with High settings?

A: The system delivers an average of 146 FPS in Thief at the High preset on a 1920x1080 resolution.

Q: How large is the VRAM of the RTX 4090 in this configuration?

A: The GPU comes with 24 GB of GDDR6X memory, which is paired with a 384-bit bus and a bandwidth of 1.01 TB/s.

Q: Which CPU is the closest rival to the i5-13400F in average benchmark score?

A: The AMD Ryzen 9 6900HS matches the i5-13400F exactly, with an average score of 25,284 versus the i5-13400F’s 25,292 and a deltaPct of 0. The i5-13400F’s score is within 0.3% of the Intel Core 5 120 and AMD Ryzen 5 5600X3D as well.

Q: Where does this CPU+GPU combo rank among all tested combinations in Thief?

A: It ranks 379th out of 3,720 combinations, placing it in the top fifth of the leaderboard.

Q: What is the minimum FPS ever recorded in the measured data?

A: The lowest minimum FPS listed is 94 FPS, which appears in the 4K Medium preset. No other presets carry a min FPS value in the dataset.

Q: Does the RTX 4090 use a different architecture than the i5-13400F?

A: Yes, the RTX 4090 uses the Ada Lovelace architecture (chip AD102) manufactured on a 5 nm node, while the CPU leverages the Raptor Lake architecture (Raptor Lake-S) on a 10 nm process.

GPU Role

The NVIDIA GeForce RTX 4090 is the flagship of the GeForce 40-series, featuring 16,384 shading units, 128 ray-tracing cores, and 512 tensor cores. Its base clock is 2,235 MHz, boosting to 2,520 MHz. The GPU’s memory configuration includes 24 GB of GDDR6X on a 384-bit bus, yielding a bandwidth of 1.01 TB/s. The fixed FP32 throughput is 82.58 TFLOPS, and the texture rate is 1,290.2 GTexel/s. These caps surpass the needs of a 2014-era game like Thief, as shown by the frame rates: running High at high resolutions, the system still retains 84 FPS at 4K. The GPU’s overall performance percentile is 88, with an average benchmark score of 60,347. Its nearest rivals in the synthetic benchmark are the Intel Arc Pro A60 (average score 60,326, deltaPct 0.0) and the AMD Radeon Pro Vega 48 (60,140, deltaPct +0.3), which is effectively identical. The large rasterization and compute headroom of the RTX 4090 makes it the primary scaling factor as resolution climbs from 1080p to 4K; the absolute FPS numbers show that in Thief, the GPU is far from being a limiting concern except at the most demanding settings.

When looking at the measured frame rates, the GPU’s raw negotiation is evident. At Ultra settings, the system delivers 80 FPS at 4K, a number already above the typical 60 FPS target, and 110 FPS at Medium. Even at Low, 4K hits 132 FPS. The presence of such high frame rates at a game as graphical as Thief indicates that the game itself is not pushing the GPU's limits; rather, it is the combination of resolution and CPU that determines the final output. The GPU's high VRAM capacity (24 GB) and bandwidth (1.01 TB/s) are far beyond Thief’s requirement, which explains why the resolution scaling does not cause collapsed performance at 4K. The performance variability across settings—fluctuations of up to 90 FPS between Low and Ultra—shows that the quality presets exert a significant load on the shader and texture fetch pipelines, but the RTX 4090 has enough thermal headroom to keep frame rates comfortably high in all tested conditions.

Memory Performance

Memory subsystem performance in a combination like this comes from two sides: the CPU’s cache hierarchy and the GPU’s VRAM and bandwidth. The i5-13400F has 80 KB of L1 cache per core, 1.25 MB of L2 per core, and a shared 20 MB L3 pool. It supports dual-channel DDR5 (and DDR4), which is a configuration that provides sufficient memory bandwidth for a 65 W processor with a 4.6 GHz boost clock. In the synthetic benchmarks, the CPU’s single-thread 3Dmark score is 960, and its multithreaded score climbs to 7,307. The cache sizes are adequate for a game like Thief, which does not require an excessive level of data reuse across many threads. The combination of L3 sized 20 MB pairs well with the game’s level of object and texture streaming, preventing worst-case pose.

On the GPU side, the RTX 4090's memory is a 24 GB GDDR6X array with an effective clock of 21 Gbps (memory clock 1,313 MHz) and a 384-bit interface, giving a peak bandwidth of 1.01 TB/s. This memory bandwidth is a major reason the system can sustain over 84 FPS at 4K Ultra, as the high resolution and detail settings demand heavy texture and shader transfers from VRAM. The data shows that frame rates drop sharply between 1440p and 4K for most presets, a decrease that aligns with a memory bandwidth attach. For example, in Ultra, the average FPS falls from 127 at 1440p to 80 at 4K, a 37% drop, while the bandwidth-intensive aspects of the 4K render (4x more pixels) impose a heavier load on the GPU's memory controller. The positive side is that the ~24 GB VRAM means there is never a case of falling short; the system cannot be considered memory-constrained.

From a performance standpoint, the CPU’s memory support (DDR4/DDR5) matters less than the L3 cache hierarchy, because the CPU rarely becomes the primary force behind the observed frame rates. Despite the CPU's dual-channel limitation, Thief does not present an extreme scene complexity that would saturate the CPU’s memory interface. The real memory bottleneck in this title is the GPU's bandwidth, and the measured values show that the RTX 4090's 1.01 TB/s is more than adequate, as the jump from 1080p to 4K does not cause a catastrophic frame collapse that would be indicative of a bandwidth deficiency. All figures point to a well-balanced memory picture for the intended use case.

Resolution Scaling

Analysing the measured FPS across the three target resolutions (1920x1080, 2560x1440, 3840x2160) reveals a consistent pattern of scaling, but with notable variations depending on the quality preset.

• At High preset: the average frame count is 146 FPS at 1080p, 129 FPS at 1440p, and 84 FPS at 4K. This translates to a 11.5% drop from 1080p to 1440p, and a 47.9% drop from 1080p to 4K. The 4K/1080p ratio is 0.575, meaning the GPU pushes only about 57% of the original performance.
• At Low preset: FPS goes from 178 at 1080p to 157 at 1440p, a 11.5% drop, and then to 132 at 4K, a 25.8% drop from 1080p; the 4K/1080 fraction is 0.742. The much shallower decline shows that at Low settings, the CPU and memory are keeping up better with the GPU even at 4K.
• At Medium (the only preset with min/max values), the frame rate is 161 FPS at 1080p, 142 FPS at 1440p, and 110 FPS at 4K. The 4K/1080 ratio is 0.683, with the min FPS at 4K being 94 and max 127. The tendency here is similar to High but slightly less steep.
• At Ultra, the frame rate falls from 144 FPS at 1080p to 127 FPS at 1440p (a drop of 14.2%) and finally to 80 FPS at 4K, a 44.4% reduction from 1080p. The 4K/1080 ratio is 0.556, the lowest among all presets.

These numbers point to a scaling pattern where the GPU becomes the limiting factor at 4K. On one hand, when the quality preset is Low, the GPU is not under maximum stress, and the resolution increase only results in a moderate FPS drop. When the preset is Ultra, the GPU’s shader and memory units are fully utilized, and the jump to 4K exposes a larger bottleneck. The largest relative drop occurs from 1440p to 4K: for instance, High goes from 129 to 84 ( –45 FPS ), while Low goes from 157 to 132 (–25 FPS). This suggests a shift from a CPU-bound or occasionally engine-limited situation at lower resolutions to a purely GPU-bound regime at 4K.

Additionally, the difference between Medium and Ultra at each resolution is consistent—at 1080p, Ultra is 83% of Medium’s FPS (144/161), at 1440p 89.5% (127/142), at 4K 72.7% (80/110). This pattern again points to the heavy workload triggered by the highest geometric complexity and shadows enabled by the Ultra preset, which is especially taxing in the pixel deluge of 4K. The overall takeaway is that the i5-13400F combined with the RTX 4090 maintains very playable frame rates at all tested settings, but the ultrawide data clearly shows the pixel count is the main challenge to maintaining high FPS.

CPU Role

The Intel Core i5-13400F is a 10-core, 16-thread desktop processor built on the Raptor Lake architecture (codename Raptor Lake-S). Its base clock is 2.50 GHz and the boost clock reaches 4.60 GHz, with a 65 W TDP that reliably threads through the same benchmark suite. In CPU benchmarks, it achieves a 3DMark 16-thread score of 7,314 and a single-thread score of 960; Cinebench R23 multicore 22,604 and single-core 3,191; Geekbench multi (11,068) and single (1,996). These translate to a 77th percentile ranking among all CPUs and an average benchmark score of 25,292 (upper ratings). Its nearest rivals in average score are separated by at most 0.4%; AMD Ryzen 9 6900HS sits at exactly +0% (25284), while the Intel Core 5 120 and AMD Ryzen 5 5600X3D each have a deltaPct of -0.3%. Thus, seating-matched hardware shows that the i5-13400F is flat in the performance centric.

In Thief, the CPU’s influence is observable in the lower resolution and lower quality settings. At 1080p Low, the system achieves 178 FPS, which is the highest recorded in the dataset, and at 1080p High it pulls 146 FPS. These figures indicate that the processor can feed the GPU even when the load is CPU-bound – the frame rate is not bottlenecked by the CPU, since a more powerful GPU would not produce higher numbers if the CPU were the limit. Aggressive boost clock of 4.6 GHz across a 10-core (6+4?) hybrid design thread layout handles the game’s simulation, AI, and draw calls, resulting in high throughput. As resolution increases towards 4K, the impact of the CPU diminishes: the difference between 1440p and 4K at High preset is 129 →84, showing that the GPU takes over the responsibility. The absence of any severe min-FPS drops (in the one dataset showing min/max, Medium preset) further supports that the CPU delivers stable frame pacing.

In summary, the i5-13400F provides a strong foundation for the RTX 4090 in this 2014 stealth game. With its 16 threads and 4.6 GHz turbo, it can drive the frame rate to over 170 FPS at 1080p, and even at 4K Ultra frame rates remain in the region of 80, which is only limited by the GPU. The CPU’s ranking relative to its immediate peers (deltaPct of 0% to +0.4%) ensures that no user will notice a shortcoming in the pairing from the central processing unit. It is entirely fair to state that the processor’s role is to provide a solid floor for the extreme GPU (88th percentile in its own league) and that the overall symmetric geometry works effectively in the context of Thief’s era when every pixel matters less.