Thief

The measured Thief data pairs an Intel Core i5-9400 with an NVIDIA GeForce RTX 5070. The CPU has 6 cores and 6 threads, a 4.10 GHz boost clock, and an average benchmark score of 2305, placing it at the 50th percentile of all CPUs. The GPU has a 2512 MHz boost clock, 12 GB of GDDR7 memory, and an average benchmark score of 41687, placing it at the 83rd percentile of all GPUs. Measured average frame rates in this stealth title range from 118.1 fps at 1920x1080 Low to 64 fps at 3840x2160 High and 3840x2160 Ultra.

Settings Recommendations

The measured rows show that Ultra is the strongest candidate for the best visual experience. At 3840x2160, Ultra and High both average 64 fps, so enabling Ultra does not cost any measured frame rate relative to High. At 1920x1080, Ultra averages 98 fps, which is higher than High’s 95.4 fps. That is an unusual result, but the data is the data. At 2560x1440, Ultra averages 83.9 fps, while High averages 86.6 fps, so High has a small measured advantage at that resolution.

For players who want the highest average FPS, Low is the fastest preset at every resolution. Low produces 118.1 fps at 1080p, 105.4 fps at 1440p, and 88.3 fps at 4K. Medium is the middle option: 106.5 fps at 1080p, 95.1 fps at 1440p, and 80.1 fps at 4K. High is the least consistent preset in this dataset because it is slower than Ultra at 1080p, faster than Ultra at 1440p, and tied with Ultra at 4K.

Given those measured rows, the sensible recommendation is Ultra when image quality matters and Low when the only goal is maximum frame rate. Ultra at 4K still delivers 64 fps, which is the same as High, so there is no measured downside to selecting the highest preset at that resolution. At 1080p, Ultra is even preferable to High on average FPS. The only serious reason to use Low is the large frame-rate headroom at all three resolutions.

No minimum or maximum FPS values were recorded in this dataset, so these recommendations are based on average FPS alone.

FAQ

Q: What is the highest measured average FPS for this combination?

A: The highest measured average is 118.1 fps at 1920x1080 with Low settings. The next highest measured values are 106.5 fps at 1080p Medium and 105.4 fps at 1440p Low.

Q: What is the lowest measured average FPS?

A: The lowest measured average is 64 fps, which appears at 3840x2160 High and 3840x2160 Ultra. Those two rows are tied.

Q: How does the Intel Core i5-9400 compare with its nearest CPU rivals?

A: The i5-9400 has an average benchmark score of 2305 and sits at the 50th percentile of all CPUs. Its nearest rivals are the Intel Core i3-10305 at 2300 with a 0.2% delta, the Intel Core i3-1125G4 at 2299 with a 0.2% delta, the Intel Xeon E-2134 at 2295 with a 0.4% delta, and the Intel Core i7-8809G at 2319 with a -0.6% delta.

Q: How does the NVIDIA GeForce RTX 5070 compare with its nearest GPU rivals?

A: The RTX 5070 has an average benchmark score of 41687 and sits at the 83rd percentile of all GPUs. Its nearest rivals include the AMD Radeon Pro 5300 at 41610 with a 0.2% delta, the NVIDIA Tesla M40 at 41897 with a -0.5% delta, the NVIDIA GeForce RTX 3090 at 41441 with a 0.6% delta, and the AMD Radeon Pro 580X at 41991 with a -0.7% delta.

Q: Does the 12 GB frame buffer create any visible limit in Thief?

A: The RTX 5070 has 12 GB of GDDR7 memory on a 192-bit bus with 672.0 GB/s bandwidth. The measured 4K Ultra result is 64 fps, equal to 4K High, so the data does not show a VRAM-capacity limitation in these rows.

Q: Is the CPU or GPU the stronger component relative to their own benchmark pools?

A: The CPU sits at the 50th percentile with an average score of 2305, while the GPU sits at the 83rd percentile with an average score of 41687. The GPU is much higher in relative standing, which matches the pattern where 4K high-settings rows produce the lowest frame rates.

GPU Role

The RTX 5070 is built on Blackwell 2.0 and uses a 5 nm process from TSMC with 31,100 million transistors on a 263 mm² die. The GPU’s memory subsystem is substantial: 12 GB of GDDR7, a 192-bit bus, 672.0 GB/s bandwidth, and a memory clock of 1750 MHz with 28 Gbps effective speed. Its base clock is 2325 MHz and its boost clock is 2512 MHz. The card also posts 30.87 TFLOPS of FP32 throughput, a texture rate of 482.3 GTexel/s, and a pixel rate of 201.0 GPixel/s.

In the measured Thief rows, the GPU’s workload becomes most visible at 4K. At 3840x2160, Low averages 88.3 fps, Medium averages 80.1 fps, and High and Ultra both average 64 fps. The identical High and Ultra results show that the highest settings are not adding measured frame-time pressure beyond High, or at least that the average lands on the same number. At 1080p, the GPU has far less pixel work, and the measured results climb to 98 fps on Ultra and 118.1 fps on Low.

The GPU’s relative benchmark position supports the idea that it is the more capable half of this pairing. The RTX 5070 averages 0.6% higher than the RTX 3090 in average benchmark score, and its 83rd-percentile standing is well above the CPU’s 50th-percentile standing. The 12 GB VRAM capacity is not evidently stressed in this 2014 title because the 4K High and 4K Ultra rows both reach 64 fps, and the Low preset at 4K reaches 88.3 fps.

Measured FPS Breakdown

The dataset contains average FPS values only; no minimum or maximum FPS fields are populated. The complete measured table is below.

| Resolution | Settings | Avg FPS |

| --- | --- | --- |

| 3840x2160 | High | 64 |

| 3840x2160 | Low | 88.3 |

| 3840x2160 | Medium | 80.1 |

| 3840x2160 | Ultra | 64 |

| 2560x1440 | High | 86.6 |

| 2560x1440 | Low | 105.4 |

| 2560x1440 | Medium | 95.1 |

| 2560x1440 | Ultra | 83.9 |

| 1920x1080 | High | 95.4 |

| 1920x1080 | Low | 118.1 |

| 1920x1080 | Medium | 106.5 |

| 1920x1080 | Ultra | 98 |

Several patterns stand out. Low is the fastest preset at every resolution. Medium is the second-fastest preset at every resolution. The ordering of High and Ultra shifts: High beats Ultra at 1440p, Ultra beats High at 1080p, and they tie at 4K. At 1080p, Ultra is measured at 98 fps, while High is measured at 95.4 fps, meaning the higher visual preset is actually faster in this dataset. At 4K, High and Ultra both land on 64 fps, creating a clear plateau at the top of the settings list.

Resolution Scaling

As resolution increases from 1920x1080 to 2560x1440 to 3840x2160, every preset loses measured average FPS. Low goes from 118.1 to 105.4 to 88.3. Medium goes from 106.5 to 95.1 to 80.1. High goes from 95.4 to 86.6 to 64. Ultra goes from 98 to 83.9 to 64.

The 4K results tell the clearest scaling story. At Low, 4K still averages 88.3 fps. At Medium, 4K averages 80.1 fps. At High and Ultra, 4K drops to 64 fps. The fact that High and Ultra converge at the same 64 fps value suggests the GPU is doing the heavy lifting at high-settings 4K and reaching a measured ceiling. Meanwhile, the 1080p Low result of 118.1 fps is the highest number in the entire dataset, which indicates that the CPU is not preventing high frame rates at lower resolution and lower settings.

The scaling pattern is not uniform across presets. High drops from 95.4 at 1080p to 86.6 at 1440p to 64 at 4K, while Low remains at 88.3 even at 4K. That uneven response points toward the limiting component shifting with workload. At 1080p Low, the i5-9400’s 4.10 GHz boost clock and 50th-percentile average score are sufficient to keep the frame rate high. At 4K Ultra, the RTX 5070’s boost clock and memory bandwidth are being pushed harder, and the average settles at 64 fps.

The CPU and GPU percentiles reinforce this reading. The CPU is at the 50th percentile with an average score of 2305; the GPU is at the 83rd percentile with an average score of 41687. A configuration with a stronger GPU relative to its pool is expected to scale poorly into very high resolution when settings are maximized, and the measured Thief data shows exactly that: 1080p Ultra is 98 fps, while 4K Ultra is 64 fps. The resolution jump costs the top preset a clear slice of performance, while the Low preset keeps a much smaller gap between 1080p and 4K.