Steamworld Build

The AMD Ryzen 9 9900X3D paired with the Intel Arc B370 lands at rank 2683 out of 2924 tested combos in Steamworld Build. That places the pairing in the bottom 8.2% of all configurations, a position driven almost entirely by the graphics side of the equation. The combo's percentile standing is a direct reflection of the GPU's own 5th percentile rank among all tested graphics cards, while the CPU sits in the 91st percentile of all processors. This is a stark mismatch: a top-tier processor dragging along a bottom-tier integrated graphics solution. The data suggests that for this specific game, the Arc B370 is the overwhelming bottleneck, and no amount of CPU headroom compensates for the GPU's fundamental throughput limits. In practical terms, this is not a combo that competitive players or enthusiasts would target for high-refresh-rate play, but it can serve as a baseline for understanding how a powerful CPU behaves when paired with a weak graphics component.

GPU Role

The Intel Arc B370 is the limiting factor in this pairing, and its specifications make that clear. It operates with a base clock of 300 MHz and a boost clock of 2400 MHz, which are modest figures for a modern graphics solution. The memory configuration is "System Shared," meaning it relies on the host system's DDR5 memory rather than dedicated VRAM, with bandwidth listed as "System Dependent." This shared-memory architecture is a significant constraint for a city builder like Steamworld Build, where large map data and numerous entities can stress memory throughput. The GPU's raw compute metrics are low: 1280 shading units, 40 texture mapping units, and 20 raster output units, yielding 6.144 TFLOPS of FP32 performance and a pixel rate of 48.00 GPixel/s. The texture rate sits at 96.00 GTexel/s.

The Arc B370's benchmark data is sparse, with a single 3DMark Steel Nomad DX12 score of 1184, which places it at the 5th percentile of all GPUs. Its nearest rivals in that benchmark are the ATI Mobility Radeon HD 5570 (scoring 1186, a 0.2% delta), the ATI Radeon HD 5770 (scoring 1190, a 0.5% delta), and the AMD Radeon HD 7650M (scoring 1192, a 0.7% delta). The fact that it trades blows with decade-old discrete mobile parts underscores its position as an integrated-class solution rather than a dedicated gaming GPU. The chip is built on Intel's 3 nm process with the Xe3-LPG architecture, but the power envelope is just 25 W, and it is classified as an IGP with no power connectors. This is a component designed for basic display output and light workloads, not for sustaining high frame rates in a simulation-heavy title.

In Steamworld Build, the GPU's role is further diminished by the game's nature. City builders typically rely more on CPU logic for simulation and AI, but the rendering load still falls on the GPU. The measured FPS shows that at 1080p Low, the combo manages 95 FPS, which is playable, but the moment settings climb, the GPU capitulates. At 1080p Ultra, the average drops to 37 FPS, a 61% reduction from Low. The pattern repeats across resolutions, with the GPU consistently failing to maintain 60 FPS at Medium settings in 1440p (46 FPS) and falling to 18 FPS at 4K High. The Arc B370 simply lacks the fill rate and shader throughput to handle the visual complexity of modern city builders at higher quality presets.

Settings Recommendations

The measured data points to a clear hierarchy for this combo: for a smooth experience, stick to lower settings. At 1920x1080 with Low settings, the combo achieves an average of 95 FPS, which is the only configuration where it comfortably exceeds 60 FPS at that resolution. Medium settings at 1080p yield 73 FPS, still playable, but the minimum FPS of 62 suggests occasional dips that could be noticeable in busy scenes. High settings at 1080p drop to 55 FPS, and Ultra falls to 37 FPS, making these presets less desirable for consistent performance.

At 2560x1440, the situation tightens. Low settings deliver 61 FPS, just crossing the playability threshold, while Medium drops to 46 FPS with a minimum of 39 FPS, which may feel stuttery during heavy simulation updates. High settings at 1440p yield 35 FPS, and Ultra is a slideshow at 23 FPS. For 3840x2160, the data is grim: Low produces 32 FPS, Medium 25 FPS, High 18 FPS, and Ultra 12 FPS. None of these are acceptable for real-time play, but Low at 4K is the least objectionable if resolution is a priority.

The best experience per the measured rows is 1080p Medium. It offers 73 FPS average with a minimum of 62 FPS, which provides a balance between visual fidelity and responsiveness. The jump from Low to Medium at 1080p costs 22 FPS but likely improves the game's visual clarity significantly, particularly for text and UI elements common in city builders. For players willing to trade some smoothness for better visuals, 1080p High at 55 FPS is serviceable, but the data suggests staying below Ultra. If 1440p is mandatory, Low settings at 61 FPS is the only viable option, and even then, the headroom is minimal. The recommendation is unambiguous: prioritize resolution over settings, and keep quality at Medium or below.

Measured FPS Breakdown

The measured FPS data reveals a consistent scaling pattern across all three resolutions. At 1920x1080, the combo starts at 95 FPS on Low, drops to 73 FPS on Medium (with a minimum of 62 and maximum of 84), falls further to 55 FPS on High, and bottoms out at 37 FPS on Ultra. The spread from Low to Ultra is 58 FPS, indicating that each settings tier costs roughly 15-20 FPS. The Medium preset's minimum FPS of 62 is notable because it stays above 60, suggesting that even during demanding moments, the frame rate holds reasonably well at that resolution.

Moving to 2560x1440, the performance drops proportionally. Low settings yield 61 FPS, which is a 35.8% decrease from 1080p Low. Medium produces 46 FPS (minimum 39, maximum 53), a 37% drop from its 1080p counterpart. High delivers 35 FPS, a 36.4% reduction, and Ultra falls to 23 FPS, a 37.8% decrease. The consistent ~36-38% drop across all settings from 1080p to 1440p indicates that the GPU is scaling linearly with pixel count, which is characteristic of a fill-rate-limited component.

At 3840x2160, the degradation is severe. Low settings manage 32 FPS, which is a 66.3% drop from 1080p Low. Medium hits 25 FPS (minimum 21, maximum 28), a 65.8% drop. High falls to 18 FPS, a 67.3% reduction, and Ultra plummets to 12 FPS, a 67.6% decrease. The 4K results are uniformly unplayable, with even the lowest settings failing to reach 35 FPS. The jump from 1440p to 4K represents a 2.25x increase in pixel count, and the FPS drops by roughly 48-49% from the 1440p values, which again aligns with a purely pixel-bound performance curve.

The data shows no anomalies or outliers. Every settings tier at every resolution follows the expected order: Low > Medium > High > Ultra. The minimum FPS values, where recorded, are only available for Medium settings, with 62 FPS at 1080p, 39 FPS at 1440p, and 21 FPS at 4K. These minimums are roughly 15-16% below the averages, indicating relatively stable frame pacing with no catastrophic spikes, but the overall performance ceiling is low.

Resolution Scaling

The FPS scaling from 1080p to 4K tells a clear story about the limiting component. At 1080p Low, the combo achieves 95 FPS. At 1440p Low, it drops to 61 FPS, a 35.8% reduction. At 4K Low, it falls to 32 FPS, a 66.3% reduction from 1080p. The expected scaling for a purely GPU-bound scenario is roughly proportional to the pixel count: 1440p has 1.78x the pixels of 1080p, so FPS should drop to about 56% of the original, which yields 53 FPS predicted versus 61 FPS actual. The actual result is better than predicted, suggesting slight CPU involvement at lower resolutions. However, at 4K, the pixel count is 4x that of 1080p, predicting 23.75 FPS if perfectly GPU-bound; the actual result is 32 FPS, which is better than predicted, indicating the GPU is not fully saturated even at 4K.

This pattern shifts when examining Medium settings. At 1080p Medium, the average is 73 FPS. At 1440p Medium, it drops to 46 FPS, a 37% reduction. At 4K Medium, it hits 25 FPS, a 65.8% reduction. The 4K result at Medium (25 FPS) is closer to the pixel-proportional expectation (18.25 FPS predicted from 1080p) but still exceeds it, suggesting some headroom or CPU limitation at higher settings. At Ultra, the story changes: 1080p produces 37 FPS, 1440p produces 23 FPS (37.8% reduction), and 4K produces 12 FPS (67.6% reduction). The 4K Ultra result of 12 FPS is nearly exactly the pixel-proportional prediction of 9.25 FPS, indicating that at the highest settings, the GPU is fully saturated and scaling is now purely pixel-bound.

The implication is that the Arc B370 has some spare capacity at lower settings, but it gets consumed as settings increase. The 1080p Low result of 95 FPS versus the 4K Low result of 32 FPS shows that the GPU is not completely bandwidth-limited at low settings, likely because the shared system memory provides sufficient throughput for simple rendering. However, at Ultra settings, the combination of higher shader load and memory bandwidth demands pushes the GPU to its absolute limit. The scaling analysis confirms that the CPU is not the bottleneck here; the Ryzen 9 9900X3D has ample headroom, as evidenced by its Cinebench R23 multi-core score of 47737 and single-core score of 6739, which are far above what Steamworld Build requires.

CPU Role

The AMD Ryzen 9 9900X3D is a 12-core, 24-thread processor based on the Zen 5 architecture (Granite Ridge), built on TSMC's 4 nm process. It features a base clock of 4.40 GHz and a boost clock of 5.50 GHz, with a TDP of 120 W. The cache hierarchy includes 80 KB of L1 per core, 1 MB of L2 per core, and a substantial 128 MB of L3 cache. Its benchmark scores are impressive: Cinebench R23 multi-core reaches 47737, single-core 6739, and Geekbench multi-core 22884 with single-core 3041. The 3DMark multi-thread tests show 12490 for 16 threads and 13795 for max threads, while single-thread scores 1269. In PassMark, the CPU achieves a multi-thread score of 56154 and single-thread of 4646.

In the context of Steamworld Build, this CPU has far more computational power than the game can utilize. The game's simulation logic may benefit from the 12 cores, but the measured FPS shows that the GPU is the binding constraint. The CPU's nearest rivals in average benchmark score include the Intel Core Ultra 5 245KF (55093, delta -0.6%), the Intel Core Ultra 5 245K (54053, delta 1.3%), and the Intel Xeon 6505P (53701, delta 2%). The Ryzen's average benchmark score of 54762 places it just below the Core Ultra 5 245KF, but this differential is irrelevant to gaming performance in this title.

The CPU's 128 MB of L3 cache is a notable feature, as it can help with data locality in simulation-heavy workloads, but the Arc B370's limitations prevent any such benefits from manifesting in frame rates. The CPU's single-thread performance, as measured by Cinebench R23 single-core at 6739, ensures that the game's main thread is never starved, but the GPU cannot translate that into higher FPS. The data suggests that Steamworld Build is heavily GPU-bound with this combo, and the CPU's role is reduced to feeding frames to a component that cannot keep up. The 3DMark 2-thread score of 2512 and 4-thread score of 4885 indicate strong multi-threading, but the game's rendering pipeline is the bottleneck. In short, the Ryzen 9 9900X3D is overqualified for this pairing, and its performance headroom is entirely wasted when paired with the Arc B370 in Steamworld Build.