Thief

The Intel Core i5-14400F paired with the AMD Radeon RX 6650 XT was measured in Thief (Stealth, released 2014-02-24) across three resolutions and four quality presets. The dataset includes 12 measured FPS points, a combo rank of 2383 out of 3720 tested combinations, and separate benchmark profiles for both the 10-core CPU and the RX 6650 XT GPU. This page interprets those recordings — what the frame counts say about the limiting component, how the chip and card behave in this 2014 stealth title, which settings deliver the best measured experience, and where this pairing lands in the overall database.

Resolution Scaling

The measured frames show a steep and consistent drop as resolution climbs. At 1920x1080 on High settings the RX 6650 XT and i5-14400F record 93 frames per second average. Moving to 2560x1440 cuts that to 64, and at 3840x2160 the same High preset falls to 38 fps. Every single illustrative point on the slider loses ground: 1080p to 1440p loses 29, and 1440p to 4K loses another 26 — the drop is nearly linear, not a cliff, but the 4K finish is less than half the 1080p starting point.

Check the opposite end of the quality scale for contrast. Low settings at 1080p hit 162 frames per second; the same Low preset records 111 at 1440p and 65 at 4K. The relative of frame count at 4K versus the 1080p value is smaller at High than at Low — 38 is 41% two of 93, and 65 is 40% of the three hundred, so Low actually retains a slightly higher, still deceptive fraction. Medium shows the mid-point pattern: 123 at 1080p, 84 at 1440p, 49 at 4K. Ultra at its bottom: 89, 61, 38.

The shape of these reductions pinpoints show the resolution scaling is strongly pixel-bound rather than scene-chaos-bound. Dropping 1080p to 4K multiplies the count of pixels by a factor of four, yet the measured frame rate does not drop by exactly half in each case — 89 to 61 to 38, 93 to 64 to 38, 162 to 111 to 65. The falloff is profile-proximate to the rise in pixel count, which indicates the AMD GPU is what limits the pipeline at the supersets tested. The RX 6650 XT has 8 GB of GDDR6 and a 128-bit bus for 280.3 GB/s of bandwidth. At 3840x2160 with Ultra and High quality settings, the average slides to 36 and 38 — close enough to a hard raw value that cannot be rescued by any CPU permutation.

The margin between Low and Ultra also informs the scaling. At 1080p, raising quality from Low to Ultra lowers the frame average by slightly over roughly half (162 down to 89). At 4K the same quality jump goes from 65 down to 36 — a much narrower delta, indicating at lower resolutions the GPU has headroom to reflect settings, all at 4K the low-pulse already corners a limit. Resolution, not graphic detail, is the primary crowding factor.

CPU Role

The Intel Core i5-14400F is a Raptor Lake-R part with 10 of the 14th Gen desktop core, 16 threads, and a 20 MB shared L3 cache. Its base clock sits at 2.50 GHz with a boost of 4.70 GHz. For a 2014 Stealth title, per-frame CPU work is modest, and the measured data pattern — large scale changes flatline different resolutions — already suggests this chip is not pacing the combo.

The publication includes several CPU stress numbers that corroborate headroom. A 2135-run in Cinebench R23 (multicore) scores 21645; the single-core result is 3055. Cinebench R15 and R20 list 21815 / 307 and 9090 / 1283 for this same device. Geekbench shows 11268 multicore and 2058 single. Passmark multithread adds 25470, single-thread 3701. The CPU also has an average benchmark score of 32279 and a percentile vs all CPUs of 83 — meaning it outpaces the vast majority of the database.

The nearest CPU rivals put this in context. The AMD Ryzen 7 PRO 8840U averages 32233 with a delta of +0.1, the Intel Core i9-11900 sits at 32226 (+0.2), the Ryzen 7 6800HS records 32354 (−0.2), and the Intel Core i7-12800H hits 32121 (+0.5). The numbers differ by less than a single point percentage point — the i5-14400F is a mid-tier class that produces similar average scores to those four.

None of these workload crunchers translate to Thief frame ceilings. The FPS pattern — near-identical 38 at 4K High vs 36 at 4K Ultra, and a 65 at 4K Low — is exactly what you'd see if the GPU were the constraint. The i5-14400F offers 16 threads and a 4.70 GHz boost, which is far more compute than a game from early 2014 demands. Its Raptor Lake architecture has more bandwidth headroom than the recorded frame rates can utilize.

One subtle hint that the CPU never the limiter: the Low-preset results. At 1080p Low, the combo produces 162 frames per second. This is well above monitor refresh — no single-core dependency (single: 3701 Passmark) could be holding it to that number. Even at 1440p Low, 111 frames is achieved without difficulty. The CPU is not the capacity for this game — its 4.70 GHz single-core boost and robust memory support (DDR4/DDR5) leave nothing to explain in frame data.

GPU Role

On its own, the AMD Radeon RX 6650 XT is a Navi 23 cut on TSMC 7 nm with 11,060 million transistors, 2048 shading units, 128 texture units and 64 ROPs. Core clocks read 2055 MHz base, 2410 game, 2635 boost, backed by 8 GB of GDDR6 on a 128-bit bus delivering 280.3 GB/s. A Passmark GTX-generation G3D test output 17166, and the card's percentile across all GPUs is 65 — mid-pack.

Its benchmark rivals are all older professional boards: the NVIDIA Tesla K40m averages 19885 (−0.6), the AMD FirePro D300 records 19637 (+0.7), the AMD FirePro W7000 19905 (−0.7), and NVIDIA Quadro K5200 19602 (+0.8). Strong evidence the RX 6650 XT sits in the same average-score bracket as those cards despite a much larger modern feature set.

Thief's response to this GPU is therefore resolution-limited. 8 GB VRAM is rarely the hard ceiling in the settings used, but the 280.3 GB/s bus and 128-bit interface constrain how fast 4K to 1920 can leave and be written. It's notable that at High the 1080p result is 93 moderate, but at 4K it is 38 — and the mid-quality settings stop at 4K Medium 49 while Ultra 4K falls to 36. This card does not have the pixel-destination capacity to keep 4K paired at a high preset.

Interesting singularity: the 1440p/Ultra average is 61. The 1440p/High is 64. A three‑fps gap is not enough to justify a full preset step; the user could just run Ultra at 1440p with most of the same speed. At 1080p though the Ultra 89, High 93, Medium still the border are 123 — GPU gets a headroom at the smallest pixels and settings can be turned up more before it saturates.

The intermediate effect: the texture bus and shading. Titan's 2048 shading units and 128 TMU handle the geometric work at 1080p, but the 128-bit connect to memory is the wall when the framebuffer grows. Thief has no 2014-era engineering artifact built in, where certain costs (lighting, terrain, crowds) stress the shader pipes — and the four measured tiers show that.

Settings Recommendations

Of the twelve measured rows, four configurations stand out as notably above their tier. At 1920x1080, the Medium preset records 123 fps average with a min 1% of 104 and a max 141 — versus 119 for Ultra and 156 for Low. The benchmark explains the reason: at 1080p with Medium, there is an average of 0.34 of a frame per preset unit, and every resolution works.

The most efficient measured combo is 2560x1440 Medium: 84 avg with min 72 and max 97. That is 21 fps lower than 1080p Medium but much higher than even an index 4K's Low. At the next step up, 1440p Ultra is fine at 61, but 1440p High is not far off — 64.4 — so the difference is negligible.

For 4K, the only row that stays playable is 4K Low at 65. If the user targets Ultra at 4K, the card sets a 36 fps result. Between Medium and High at 4K, the average slides from 49 to 38 — a 11 fps penalty for one quality tier, which is a big jump for a small visual gain in most stealth gameplay.

Hence the dataset recommends the following through measured results only:

• 1080p: prefer Ultra (89) or High (93); if you want fluid above 120, Medium144p at 123 is the best above-120 measured.
• 1440p: Medium at 84 is the sweet spot; it has headroom to 72 low and 97 peak.
• 4K: Low at 65 provides the only rock-solid 60+ row; Medium only makes it to 49.

None of the 4K presets beat 60 fps — the RX 6650 XT's measured 4K ceiling is 65 at Low, and that's the only data-threshold over 60. Buyers on 4K panels should limit expectations to Low or dropped frames to 1440p Medium.

FAQ

Q: What FPS does this combination reach in Thief at 1080p?

A: The best 1080p result is Low at 162 FPS, then Medium at 123, High at 93, and Ultra at 89 — a span of 73 FPS between the fastest and slowest of the four presets.

Q: Can this PC (i5-14400F + RX 6650 XT) handle 4K in Thief?

A: Only at Low settings does the pair’s 4K average clear 60 — 65. Medium gives 49, High 38, Ultra 36. None reach the 1440p/Ultra figure of 61.

Q: Which preset should I pick for 1440p?

A: Medium is best per the measurement, 84 average with min 72 and max 97. 1440p High scores 64, Ultra 61, Low 111 — so Low is if you need the headroom, otherwise Medium for balance.

Q: How does the CPU affect 4K performance in this game?

A: The i5-14400F has 10 cores and 16 threads (boost 4.70 GHz) and a 20 MB shared L3. The FPS drop from 1080p (Low 162) to 4K (Low 65) is dominated by the GPU’s memory bottleneck, not by the 10-core CPU.

Q: Is Core i5-14400F a high-end CPU by the time of this data?

A: Its 83rd percentile vs all CPUs, Cinebench R23 multicore 21645/single 3055, and rivals within 87.0—at least neighboring an 8400U — show it's firmly above the middle tier, though many newer parts, nearby, exceed it.

Q: What is the CPU’s GPU a failed-limiter at 4K Ultra?

A: At 4K Ultra the average is 2%? That is 36 fps. The RX 6650 XT with 128-bit bus and 2048 shaders — not the CPU issue. Raising memory bandwidth to 280.3 GB/s doesn’t break that wall.

Measured FPS Breakdown

The full measured table is reproduced exactly as recorded.

1920x1080 (Full HD):

• High: 93 FPS
• Low: 162 FPS
• Medium: 123 FPS (min 104, max 141)
• Ultra: 89 FPS

2560x1440 (QHD):

• High: 64 FPS
• Low: 111 FPS
• Medium: 84 FPS (min 72, max 97)
• Ultra: 61 FPS

3840x2160 (4K):

• High: 38 FPS
• Low: 65 FPS
• Medium: 49 FPS (min 42, min 42, max 57)
• Ultra: 36 FPS

That's twelve distinct measured points, all with a consistent ordering: Low > Medium > High > Ultra at every resolution. The only cited min/max pairs were recorded for Medium at 1080p, 1440p, and 4K; the rarer points for High, Low, and Ultra have no min/max fields in the source pack.

How This Combo Ranks

The combo (i5-14400F + RX 6650 XT) occupies rank 2383 out of 3720 fully measured combinations in Thief. That places the pairing in the upper third — it takes better hardware to break into the top ~600 slots, but a middle-tier desktop representing 65 % of overall benefit nicely. A middle-of-the-pack ranking here is consistent with the measured behavior: this pairing is very capable at 1080p and 1440p, but its 4K performance caps out below 80% of testing because the RX's high-bandwidth rating (280.3 GB/s) and its GPU percentile than his AMD tier. The average CPU frame of 3701 single-thread and the 83rd percentile GPU suggest the underuse the throughput is 83rd percentile — but at the ranks lower place the GPU. Compare the GPU to its nearest synthetic rivals: the 19765 average (vs the rival 0.5% deltas) mirrors that the actual Thief frames put the 6650 XT in with the mid-pack range.

In specific words: a 4:1 ratio of resolutions means a user should expect medium 1080p and lose half the frames at 4K. Ranked 2383 of 3720 — this is a performance-balanced build that rewards settings care more than core count. Even at modest Low the 14500 or $ 720 demonstrated compared to 4K, but you still need to watch the GPU side. There is no razor-thin CPU participation: the 14400F stands empty behind, while the Radeon thins everything out past 89% of the dataset slate were measured above their baseline.