SYSTEM ANALYZER

Rate My PC: Intel Core Ultra 5 235T + NVIDIA GeForce RTX 3090

Get a comprehensive performance analysis of your gaming rig with detailed benchmarks, bottleneck detection, and upgrade recommendations

92 / 100
ULTIMATE READY

Apex Performer

Top 8% of systems. Capable of 4K Ultra gaming and advanced rendering.

4K 60+ FPSVR ReadyRay Tracing

System Balance Analysis

CPU vs GPU performance ratio
Well Balanced
CPU
92%
VS
GPU
93%
PROCESSOR

Intel Core Ultra 5 235T

38,561 Benchmark Score
Top 8% Market Ranking
View Full Specs →
GRAPHICS CARD

NVIDIA GeForce RTX 3090

27,565 Benchmark Score
Top 7% Market Ranking
View Full Specs →

Market Position

How your build compares to others
Budget
0-30
Mid-Range
30-60
High-End
60-85
Enthusiast
85-100
Your Build

Game Performance Benchmarks

Real-world 4K FPS in popular titles
View All Games →

Performance Insights

Tips to maximize your system

Optimal Performance

Your system is in the top tier. You can run any modern game at maximum settings.

4K Gaming Ready

Consider a 4K 144Hz monitor to fully utilize your hardware capabilities.

Compatible Games See what you can play Compare CPUs Find upgrades Compare GPUs Find upgrades

Performance Tiers Explained

90-100

Ultimate

4K Ultra gaming, VR ready, ray tracing enabled, professional workloads

4K 60+ FPS VR Ready
70-89

High-End

1440p Ultra or 4K High settings, excellent for modern AAA titles

1440p Ultra 4K High
50-69

Mid-Range

1080p Ultra or 1440p Medium, great value for most gamers

1080p Ultra 1440p Med
30-49

Entry Level

1080p Medium settings, suitable for eSports and older titles

1080p Med eSports
0-29

Legacy

Basic gaming, older titles, consider upgrading for modern games

720p-1080p Low Older Games

The pairing of an Intel Core Ultra 5 235T with an NVIDIA GeForce RTX 3090 is an unusual one: a recent, efficiency-focused desktop CPU from the Core Ultra Series 2 lineup matched with a former flagship GPU that is now end-of-life. The data shows a combined percentile of 80, meaning this desktop build sits comfortably in the upper fifth of all systems in the database, though the two components occupy very different positions within their own product families. One important caveat applies throughout: no measured FPS rows exist for this exact combination in the database, so all frame-rate discussion below is estimated from benchmark scores rather than direct game testing.

FAQ

Q: What CPU and GPU are in this build?

A: The CPU is an Intel Core Ultra 5 235T, a 14-core, 14-thread Arrow Lake-S desktop processor on the Intel Socket 1851, with a launch MSRP of $247. The GPU is an NVIDIA GeForce RTX 3090, an Ampere-generation card with 24 GB of GDDR6X memory, built on the GA102 chip.

Q: How powerful is this combination overall?

A: The combined percentile is 80, placing the build above roughly four-fifths of systems in the database. The CPU sits at the 86th percentile versus all CPUs, while the GPU sits at the 73rd percentile versus all GPUs.

Q: Does the database contain measured game FPS for this pairing?

A: No. There are no measured FPS rows for this exact CPU and GPU combination, so all frame-rate figures discussed here are estimates derived from the benchmark scores rather than direct measurements.

Q: How does the CPU compare to its nearest rivals?

A: Its average benchmark score of 38561 puts it fractionally ahead of the Intel Core i5-14500 (0.1% behind), the Intel Xeon w3-2525 (0.4% behind), the Intel Core 9 270H (0.6% behind), and the Intel Core Ultra 9 285H (0.7% behind) — an extraordinarily tight cluster of competitors.

Q: How does the GPU compare to its nearest rivals?

A: With an average score of 27565, the RTX 3090 is 0.5% ahead of the NVIDIA GeForce RTX 4070 Mobile, 0.5% ahead of the AMD Radeon RX 6700 XT, 1% behind the AMD Radeon Pro Vega 20, and 1.1% behind the AMD Radeon RX 7800M.

Q: What power supply headroom does this build need?

A: The CPU has a TDP of 65 W and the GPU a TDP of 350 W, with the GPU's suggested PSU rating at 750 W. That suggested rating already accounts for a full system, so a 750 W unit is the reference point for safe headroom.

Q: Can the CPU be overclocked?

A: No — the multiplier is locked, so performance tuning options are limited to memory and platform-level adjustments rather than core clock manipulation.

Benchmark Performance

The Intel Core Ultra 5 235T posts a Cinebench R23 multi-core score of 26232 and a single-core score of 3703, alongside a Cinebench R20 pair of 11017 multi-core and 1555 single-core, and a Cinebench R15 result of 2644 multi-core against 373 single-core. In the PassMark suite it records 30918 in multithread and 4339 in single-thread, with satellite scores including 295100 in data compression, 23457 in encryption, 106546 in floating point math, 84244 in integer math, 35377 in string sorting, 318 in prime finding, 2157 in physics, and 23212 in extended instructions. Its average benchmark score across the suite is 38561, placing it at the 86th percentile of all CPUs.

The RTX 3090 counters with a 3DMark Steel Nomad DX12 score of 5118, a Geekbench OpenCL result of 172758, a Vulkan result of 53927, a PassMark G3D score of 26645, and a PassMark GPU compute score of 15356. Its PassMark per-API results include 268 for DirectX 9, 182 for DirectX 10, 220 for DirectX 11, and 110 for DirectX 12, plus 1063 in G2D operations. The average score of 27565 places it at the 73rd percentile of all GPUs.

What does the combined picture tell us? The CPU is actually the stronger performer within its own category than the GPU is within its — an 86th percentile CPU paired with a 73rd percentile GPU produces a system percentile of 80, almost exactly the midpoint of the two. That symmetry is interesting because it suggests a balanced build on paper, though as the bottleneck analysis will show, the balance shifts depending on workload. The CPU's rivalry cluster being within a single percentage point of four other processors also tells us the 235T's aggregate performance is representative of a dense performance band — it is neither a standout nor a laggard within its tier.

CPU Analysis

The Core Ultra 5 235T is a 14-core, 14-thread processor — notably, one thread per core with no hyperthreading-like doubling. It is built on the Arrow Lake-S architecture, manufactured by TSMC on a 3 nm process node, packing 17,800 million transistors into a 243 mm² die. The clock configuration is a 2.20 GHz base with a 5.00 GHz boost, and the TDP is a modest 65 W — a figure that signals this is the efficiency-oriented "T" variant of the Arrow Lake Ultra 5 family rather than a performance-optimized part.

The cache hierarchy consists of 192 KB of L1 per core, 3 MB of L2 per core, and 24 MB of shared L3. Memory support is DDR5 on a dual-channel bus delivering 102.4 GB/s of bandwidth; ECC is not supported. On the graphics side, integrated Arc Xe-LPG Graphics with 24 execution units are present, though largely irrelevant in a build carrying an RTX 3090 — they could, however, serve as a fallback display output in a troubleshooting scenario.

Interpreting the scores: the Cinebench R23 single-core figure of 3703, together with a PassMark single-thread score of 4339, indicates snappy per-core responsiveness — the kind of performance that governs game logic, application launch behavior, and lightly threaded productivity work. The multi-core numbers tell a more measured story. A 26232 in R23 multi-core is solid mid-tier throughput, appropriate for the 14-thread ceiling of this chip. The ratio between single- and multi-core results suggests the 235T is more single-thread-tilted than heavily parallel chips, which raises a question worth asking: in workloads that can scale past 14 threads, this CPU will hit its limit sooner than higher-core-count rivals. The PassMark physics score of 2157 and extended instructions result of 23212 reinforce the picture of a capable but not extreme all-rounder. For real workloads, that means strong interactive performance, competent mid-weight rendering and compilation, and a ceiling on heavy sustained multiprocessor tasks.

Upgrade Path and Platform

The platform foundations are Socket 1851, DDR5 memory, and PCIe Gen 5 with 20 CPU lanes. The PCIe situation deserves a closer look: the RTX 3090 uses a PCIe 4.0 x16 bus interface, so on this Gen 5 platform the card runs at its own Gen 4 speed — no bandwidth is wasted, but no Gen 5 advantage is gained either. The 20 CPU lanes leave room for a Gen 5 NVMe SSD and the GPU without compromise, and chipset lanes can absorb additional storage.

The GPU is end-of-life per its production status, while the CPU is an active part — a status mismatch that hints at the natural upgrade direction. A future GPU swap on this same platform is the cleanest path, and the CPU's 65 W TDP means the platform's power budget is dominated entirely by graphics. The suggested PSU of 750 W already accommodates the 350 W GPU, so any replacement card with similar or lower power draw drops in without a PSU change; something substantially hungrier would warrant reassessing. For the CPU itself, Socket 1851 offers headroom within the Core Ultra Series 2 family. The locked multiplier on the 235T removes overclocking as a tuning avenue, so performance gains come from component replacement rather than adjustment. DDR5 on dual channels at 102.4 GB/s of bandwidth is a sound foundation that will not be the weak link in any realistic upgrade scenario.

Usage Scenarios

High-refresh gaming: With a GPU at the 73rd percentile and a CPU at the 86th, the data supports high-refresh play at 1440p and strong 4K performance, though — again — no measured FPS exists for this pairing, so these are estimates from scores. The RTX 3090's PassMark G3D of 26645 and Steel Nomad result of 5118 sit firmly in the range where high frame rates at 1440p are realistic, while the CPU's single-thread strength of 4339 keeps frame pacing intact in CPU-bound titles.

Streaming: Dual-encoding headroom comes from the GPU's 328 tensor cores and dedicated encode hardware class, and the CPU's 14 threads handle game plus encoder duty. The PassMark multithread score of 30918 suggests streaming while gaming is comfortably within reach, though streamers running heavy x264 CPU encodes will feel the 14-thread ceiling before owners of larger chips would.

Video editing: The RTX 3090's 24 GB of VRAM and 936.2 GB/s of memory bandwidth are the stars here — timeline scrubbing and effects preview at 4K are memory-capacity-bound before they are compute-bound, and 24 GB is ample. The Geekbench OpenCL score of 172758 confirms strong GPU compute for accelerated effects, while the CPU's R23 multi-core of 26232 handles export-side encode work capably.

3D rendering: This is where the split matters. GPU-path rendering leverages the 35.58 TFLOPS of FP32 throughput and Vulkan compute score of 53927 productively, but CPU-only rendering is capped by 14 threads. The 3090 carries viewport work; the 235T is the limiting factor for pure CPU renders.

Software development: Compilation benefits from both multi-core throughput and single-core speed, and the 235T scores well on both axes — a 4339 single-thread result governs incremental builds while 30918 multithread covers clean builds. The 102.4 GB/s memory bandwidth supports containerized and virtualized workflows well.

Student and office work: Frankly overkill, but pleasantly so. The 65 W TDP makes this a quiet, efficient machine for everyday productivity, with the integrated Arc graphics as a safety net should the discrete card ever be removed.

Who Should Build It

The data points to three strong candidate profiles. First, 1440p and 4K gamers: the GPU's 73rd percentile position and its rivalry cluster around the RX 6700 XT and RTX 4070 Mobile define exactly the class of play this build delivers, with the CPU providing more than enough single-thread headroom to feed it. Second, content creators whose software leans on GPU acceleration — the 24 GB VRAM and OpenCL result of 172758 make this an attractive machine for 4K video editing and GPU-accelerated 3D work, where the memory capacity matters more than raw CPU core count. Third, developers and small-business workstation users: the 14-thread CPU with strong single-core scores, ECC absence notwithstanding, handles compilation, data compression (295100 in PassMark), and everyday workload with efficiency thanks to the 65 W TDP. Students in technical fields would find it equally serviceable. What this build is not for is anyone whose primary workload is heavily multi-threaded CPU rendering — the 86th percentile CPU ranking overstates its usefulness there because sustained all-core throughput is capped by its thread count and thermal envelope.

Gaming Performance

The database contains no measured FPS rows for this exact combination — dataIsMeasured is false and measuredFpsUltraByGame is empty. Every frame-rate statement in this section is therefore an estimate extrapolated from benchmark scores, not a measured figure.

Working from those scores: the RTX 3090's PassMark G3D of 26645, its position 0.5% ahead of the RX 6700 XT and 0.5% ahead of the RTX 4070 Mobile, and its Steel Nomad score of 5118 collectively describe a GPU firmly in the upper-mainstream performance band. At 1080p ultra settings, that class of card is typically CPU-limited in esports and fast-paced titles, and here the 235T's single-core scores — 3703 in R23, 4339 in PassMark — indicate the CPU can drive high frame rates without becoming the constraint. At 1440p ultra, estimates would place most modern titles in comfortably playable territory with high settings, since the 936.2 GB/s of memory bandwidth and 10496 shading units have headroom at that resolution. At 4K ultra, the GPU becomes the limiter, as it does for any card in this tier; the 24 GB frame buffer means texture-heavy 4K presets do not exhaust VRAM, which prevents the cliff-edge frame drops that afflict smaller-memory cards in this performance class. Ray tracing is supported via the 82 RT cores and DirectX 12 Ultimate, though enabling heavy RT at 4K would push estimated frame rates down into territory where upscaling would be advisable.

Build Overview

This is a desktop build — buildClass confirms a tower-class system, not a laptop — pairing a current-generation efficiency CPU with a previous-flagship GPU. The overall tier, expressed by the combined percentile of 80, places it in the top fifth of systems in the database: a high-performance machine, but not an extreme one. The component mix is generationally asymmetric in an interesting way. The Core Ultra 5 235T is a 2025 release built on TSMC's 3 nm process, while the RTX 3090 dates to 2020 and is built by Samsung on an 8 nm node with a 628 mm² die packing 28,300 million transistors at a density of 45.1M per mm². The pairing reads like a sensible economy of silicon: spend efficiency and platform modernity on the CPU side, and raw capacity — 24 GB of VRAM, a 384-bit bus — on the GPU side. With the GPU end-of-life and the CPU active, this configuration is more likely encountered as an upgrade or a rebuilt system than a fresh from-scratch pairing.

Balance and Bottleneck

The bottleneck question here resolves differently by workload, and the percentiles make the mechanics legible. In gaming, the GPU is the limiter: at the 73rd percentile it trails the CPU's 86th, and the CPU's PassMark physics score of 2157 alongside its single-thread result of 4339 shows more CPU headroom than the GPU can consume at graphics-heavy settings. Increasing resolution shifts load further onto the GPU, improving balance; decreasing it makes the GPU wait on the CPU far less than vice versa. In compute workloads the relationship inverts. GPU compute tasks draw on a PassMark compute score of 15356 and OpenCL score of 172758 — plenty — while sustained CPU multiprocessor work hits the 14-thread ceiling and the 65 W thermal envelope. The tightness of the CPU's rivalry cluster (every rival within 0.7%) means there is no meaningful escape from that ceiling via close alternatives; a meaningful multi-core step-up requires jumping to a materially higher tier. The upshot: this is a GPU-forward build in which the CPU rarely limits graphics but does define the multiprocessing ceiling.

GPU Analysis

The RTX 3090's specification sheet still reads impressively: 10496 shading units, 328 texture mapping units, 112 render outputs, 82 RT cores, and 328 tensor cores on the GA102 chip. Clocks run from a 1395 MHz base to a 1695 MHz boost, with memory at an effective 19.5 Gbps. The 24 GB of GDDR6X on a 384-bit bus yields 936.2 GB/s of bandwidth — a figure that remains the card's defining asset, since few workloads at its performance tier can saturate that capacity. Theoretical rates include 189.8 GPixel/s of pixel fill and 556.0 GTexel/s of texture fill, with FP32 throughput of 35.58 TFLOPS and FP16 at a 1:1 ratio.

What do the benchmark scores say this hardware delivers? The Geekbench OpenCL result of 172758 and Vulkan score of 53927 confirm strong general compute, and the PassMark compute score of 15356 reinforces it. The API-split PassMark results — 268 in DX9, 182 in DX10, 220 in DX11, 110 in DX12 — are relative-weight scores rather than absolute frame counts, and they show the card scaling across generations of graphics APIs as expected. For rendering workloads, the tensor cores enable AI-accelerated denoising and upscaling in supported renderers, and the RT cores handle hardware ray tracing under DirectX 12 Ultimate, OpenGL 4.6, and Vulkan 1.4. The practical takeaway from the data: this is a memory-rich, bandwidth-heavy compute card whose 73rd-percentile gaming rank slightly understates its usefulness in content-creation contexts, where 24 GB of VRAM and 936.2 GB/s of bandwidth punch above the card's aggregate score. Physically, it is a triple-slot card measuring 336 mm long, 140 mm tall, and 61 mm wide, drawing power through a single 12-pin connector — case clearance and connector format are the practical integration considerations alongside the 750 W suggested PSU.