Roblox

Roblox paired with the Intel Core Ultra 9 185H and NVIDIA GeForce RTX 5090 is a study in excess headroom. Roblox, a sandbox title released in 2006, sits near the lightweight end of the modern gaming spectrum, yet this pairing — a Meteor Lake mobile CPU sitting in the 85th percentile of all CPUs and a Blackwell flagship GPU in the 92nd percentile of all GPUs — pushes frame rates that most games never approach. The measured data, collected across three resolutions and four settings presets, shows a platform that never drops below triple digits in average frame rate at any tested configuration. The following analysis walks through the scaling behavior, settings trade-offs, component roles, and where this combination ranks among the 3711 combos tested in this game.

Resolution Scaling

The resolution ladder tells a consistent story: Roblox on this hardware leans away from being GPU-bound at the common settings levels, and the frame rate drop from 1080p to 4K is far gentler than a GPU-limited scenario would produce.

Take the Low preset first. At 1920x1080 the combo averages 528 FPS. Moving to 2560x1440 yields 353 FPS, and stepping again to 3840x2160 lands at 199 FPS. That is a drop from 1080p to 1440p of roughly a third, and a further drop to 4K that leaves the system just under 200 FPS even at minimum settings. If the RTX 5090 were the hard limiter, quadrupling the pixel count from 1080p to 4K would typically halve frame rates or worse; instead, the data shows 199 FPS at 4K Low versus 528 FPS at 1080p Low — a decline, but not a cliff. The pattern repeats at every preset.

At Medium, the averages run 481 FPS at 1080p, 321 FPS at 1440p, and 181 FPS at 4K, with the 4K Medium row also recording a minimum of 154 FPS and a maximum of 209 FPS. At High the progression is 425, 284, and 160 FPS. At Ultra — the heaviest preset tested — the numbers are 345, 230, and 130 FPS. In every case, the 4K figure sits comfortably above the 120 FPS mark that most high-refresh displays target.

What does this say about the limiting component? The fact that frame rates keep falling as resolution rises — rather than flattening onto a fixed ceiling — indicates the GPU is still doing meaningful work per frame; a purely CPU-limited configuration would show near-identical averages at all three resolutions. But the magnitude of the drop, and the extremely high absolute frame rates at 1080p, point toward engine and platform overhead becoming the practical constraint as pixel counts fall. At 528 FPS, the Core Ultra 9 185H's per-frame budget is razor thin, and a sandbox engine that runs user-generated content has its own per-frame logic costs independent of resolution. The scaling curve therefore suggests a hybrid picture: CPU- and engine-bound at 1080p, progressively GPU-bound by 4K — but never bound enough to threaten playability.

Settings Recommendations

The measured rows make the settings decision straightforward, because every preset at every resolution clears high-refresh thresholds.

For pure competitive smoothness — reaction time and input latency matter more than visual fidelity in a sandbox platform — the numbers favor Low. At 1080p Low, the average is 528 FPS, the highest single figure in the dataset. At 1440p Low, 353 FPS. Even at 4K Low, 199 FPS. If the display is a standard high-refresh panel, Low at 1080p extracts the maximum from the hardware.

The Medium preset is the data-backed sweet spot. It costs only 47 FPS at 1080p (481 versus 528), 32 FPS at 1440p (321 versus 353), and 18 FPS at 4K (181 versus 199) relative to Low — small sacrifices for the visual improvement Medium typically brings. Notably, Medium is also the only preset with recorded minimum and maximum frame rates in this dataset: 409 to 553 FPS at 1080p, 273 to 369 FPS at 1440p, and 154 to 209 FPS at 4K. The 4K Medium minimum of 154 FPS is itself above many monitors' refresh ceilings, which indicates the framerate floor stays smooth even in the more variable measurement window.

High is a viable choice at 1080p and 1440p, where the averages of 425 and 284 FPS remain exceptionally fluid. At 4K High the average is 160 FPS — still excellent, though the margin above high-refresh territory narrows. Ultra is the only preset where the data invites any caution, and even there the verdict is relative: 345 FPS at 1080p and 230 FPS at 1440p are outstanding, while 130 FPS at 4K Ultra is the lowest average in the entire dataset. For a 4K player on a high-refresh monitor, dropping Ultra to High recovers 30 FPS (160 versus 130); dropping to Medium recovers 51 FPS. For anyone on a 60-to-120-class display, every preset at every resolution is more than sufficient, and Ultra can be enabled without measurable consequence to the experience.

GPU Role

The NVIDIA GeForce RTX 5090 is, on paper, an unusual partner for a 2006-era sandbox title — and the data confirms the GPU is rarely the story in Roblox. The card's raw specifications frame why: 21760 shading units, 680 TMUs, 176 ROPs, 170 RT cores, and 680 tensor cores on the GB202 chip, built on TSMC's 5 nm process with 92,200 million transistors across a 750 mm² die. Clocks run at a 2017 MHz base and 2407 MHz boost, with memory clocked at 1750 MHz (28 Gbps effective). The 32 GB of GDDR7 on a 512-bit bus delivers 1.79 TB/s of bandwidth — capacity and throughput that Roblox cannot plausibly saturate at any tested resolution.

The VRAM question, specifically, is settled by the frame data. A game that averages 130 FPS at 4K Ultra and 345 FPS at 1080p Ultra is not working against a memory wall; texture and asset streaming demands in this title simply do not approach 32 GB. The GDDR7 bandwidth is likewise idle headroom here.

Where the GPU's capability shows up is in the flattening of settings penalties. Moving from Low to Ultra costs 183 FPS at 1080p (528 to 345), 123 FPS at 1440p (353 to 230), and 69 FPS at 4K (199 to 130). The RTX 5090's pixel rate of 423.6 GPixel/s and texture rate of 1,636.8 GTexel/s, alongside 104.8 TFLOPS of FP32 compute, mean the Ultra preset's additional shading load is absorbed without the severe multiplier effects weaker cards exhibit. In the card's own benchmark record, it posts a 3DMark Steel Nomad DX12 score of 18355, a Geekbench Vulkan score of 376728, and a PassMark G3D score of 39650 — figures consistent with its 92nd-percentile standing among all GPUs. Roblox, with a maximum demand of 4K Ultra at 130 FPS average, never pulls the card out of its comfort zone.

One architectural note: the RTX 5090 supports DirectX 12 Ultimate, Vulkan 1.4, and OpenGL 4.6. A sandbox engine that originated in 2006 and evolved across API generations is fully covered by that support matrix, so there is no feature-level bottleneck in the pipeline.

Measured FPS Breakdown

The complete measured dataset for Roblox on this combination, all rows representing measured average frame rates:

3840x2160 (4K):

• Low: 199 FPS average
• Medium: 181 FPS average (min 154, max 209)
• High: 160 FPS average
• Ultra: 130 FPS average

2560x1440 (1440p):

• Low: 353 FPS average
• Medium: 321 FPS average (min 273, max 369)
• High: 284 FPS average
• Ultra: 230 FPS average

1920x1080 (1080p):

• Low: 528 FPS average
• Medium: 481 FPS average (min 409, max 553)
• High: 425 FPS average
• Ultra: 345 FPS average

Reading the grid, several patterns stand out. The spread between the best and worst configuration — 1080p Low at 528 FPS versus 4K Ultra at 130 FPS — spans a factor of roughly four across the full matrix, yet even the worst cell exceeds the average frame rates most contemporary games achieve on this class of hardware. Each settings step costs progressively more as resolution falls: Ultra versus Low at 4K is a 69 FPS gap, at 1440p it is 123 FPS, and at 1080p it is 183 FPS. This is the signature of a GPU that has surplus capacity at lower pixel counts — visual quality costs scale with the frames available to spend.

The Medium rows carry the only min/max telemetry in the set, and they reinforce the headline numbers. The 1080p Medium range of 409 to 553 FPS brackets the Low and High averages almost perfectly, indicating consistent frame delivery rather than a spiky average propped up by brief bursts. At 1440p Medium, the 273-to-369 range sits neatly between the High average of 284 and Low average of 353. At 4K Medium, the 154-to-209 range similarly straddles the High average of 160 and Low average of 199. Internal consistency of this kind strengthens confidence that the averages are representative of real play.

How This Combo Ranks

Within the Roblox benchmark database, the Core Ultra 9 185H plus RTX 5090 combination holds rank 816 out of 3711 tested combos. That places it in the top quartile of configurations measured for this game — a solid standing, though notably short of the elite tier one might expect from a GPU in the 92nd percentile overall.

The gap between the GPU's individual standing and the combo's in-game rank is the CPU. The Core Ultra 9 185H is a mobile-segment, 45 W TDP part, and while it lands in the 85th percentile versus all CPUs, its nearest rivals by average score are telling: the AMD Ryzen AI 7 PRO 350 sits just 0.1% ahead (average score 35719 versus 35670), the Intel Core 7 250H is 0.2% ahead at 35728, the AMD Ryzen 7 PRO 5845 is 0.4% ahead at 35802, and the AMD Ryzen 7 7700X is 0.7% ahead at 35909. In other words, this is a mid-pack-upper CPU surrounded by peers, not a chart-topping desktop flagship. In an engine like Roblox, where the measured frame rates reach into the multiple hundreds, per-thread CPU throughput — not GPU horsepower — determines how far up the combo leaderboard a pairing can climb. The rank of 816 is therefore best read as: the RTX 5090 supplied far more rendering throughput than Roblox can use, and the final rank was set by the processor's ability to feed frames at extreme rates.

On the GPU side, the RTX 5090's own rival list — Tesla P100 PCIe 16 GB at 0.3% behind (average 79605), Tesla P100 PCIe 12 GB at 0.6% behind, AMD Radeon RX 6850M XT at 1.1% behind, and Radeon Pro Vega 64X at 1.4% ahead — confirms it as an aggregate-performance leader in the database's composite scoring. None of that surplus changes Roblox outcomes; the game's demands are met many times over.

CPU Role

The Intel Core Ultra 9 185H brings 16 cores and 22 threads on the Meteor Lake architecture, built on Intel's 7 nm process, with a 3.90 GHz base clock and 5.10 GHz boost. Cache comes in at 112 KB of L1 per core, 2 MB of L2 per core, and a 24 MB shared L3. Memory support is DDR5 on a dual-channel bus with 89.6 GB/s of bandwidth, and the chip carries integrated Arc Xe-LPG Graphics with 128 EU. It released in December 2023 with a launch MSRP of $640.

Roblox does not thread deeply across 22 threads; sandbox engines of this lineage are dominated by single-thread and low-thread-count logic. That makes the single-core figures the relevant ones. The 185H scores 264.5 in Cinebench R15 single-core, 1462 in R20 single-core, 1810.5 in R23 single-core, 1794 in Geekbench single-core, and 3697 in PassMark single-thread. Its multi-core results — 2801.5 in R15, 10356 in R20, 17984.5 in R23, 11657 in Geekbench multicore, and 29376 in PassMark multithread — show the chip scales well across its full core count, but that capacity goes largely unused here.

The interplay with the measured FPS data is direct. At 1080p Low, 528 FPS means the CPU must complete its share of every frame in under two milliseconds; at that cadence, per-clock efficiency and boost behavior matter more than core count. The 5.10 GHz boost clock is what sustains the mid-500s at 1080p and the 481 FPS Medium average. As resolution rises, the frame budget loosens — 199 FPS at 4K Low allows over five milliseconds per frame — and the CPU recedes from the critical path, letting the GPU's load become the defining variable. This dynamic explains the resolution scaling curve described earlier: the flatter-than-expected drop from 1080p to 4K reflects a system transitioning from CPU-constrained to GPU-constrained territory rather than remaining pinned to one limiter.

FAQ

Q: Can this combo run Roblox at 4K on maximum settings?

A: Yes. At 3840x2160 with the Ultra preset, the measured average is 130 FPS. Dropping to High at the same resolution raises the average to 160 FPS, and Medium reaches 181 FPS with a recorded minimum of 154 FPS.

Q: What is the highest frame rate recorded for this combination?

A: 528 FPS, at 1920x1080 on the Low preset. The best 4K figure is 199 FPS at Low, and the best 1440p figure is 353 FPS at Low.

Q: Does the CPU or the GPU limit Roblox performance on this pairing?

A: The data indicates both, at different points. At 1080p, where averages reach 528 FPS, the Core Ultra 9 185H's per-frame budget becomes extremely tight and the system behaves CPU- and engine-bound. At 4K, frame rates fall with rising pixel counts and rising settings, showing the RTX 5090 doing the limiting work — though never severely enough to threaten playability.

Q: How does this combination rank against other tested configs in Roblox?

A: Rank 816 out of 3711 tested combos — within the top quartile but not the top tier, reflecting that the mobile-segment CPU, not the RTX 5090, caps performance at the extreme frame rates Roblox enables.

Q: Is 32 GB of VRAM needed for Roblox?

A: No evidence in the data supports that. Frame rates scale predictably with resolution and settings with no VRAM-related collapse; the RTX 5090's 32 GB of GDDR7 and 1.79 TB/s of bandwidth represent idle headroom for this title.

Q: Is the Intel Core Ultra 9 185H a strong CPU overall?

A: It sits in the 85th percentile versus all CPUs in the database, with an average benchmark score of 35670. Its closest rivals — the Ryzen AI 7 PRO 350, Core 7 250H, Ryzen 7 PRO 5845, and Ryzen 7 7700X — all fall within 0.7% of its average score, placing it in a tightly packed upper-mid cluster.