Measured performance · local AI
Benchmarks
Loading benchmark evidence…
Measured performance · local AI
Loading benchmark evidence…
Methodology · UOS Benchmark v1
UOS Benchmark Intelligence derives workload indices from globally sourced, validated measured evidence. Each observation retains exact hardware identity, source provenance, test context, and version history.
Scores on rankings, hardware pages, comparisons and upgrade checks are UOS workload indices derived from measured results published by independent sources. A product gets a score in a workload only when its result is ready at the current reference for that workload domain and cleared for publication (confirmed source rights, no evidence still under review); otherwise no score is shown, never zero.
The UOS lab benchmark suites (method v1.0.0) have 26 frozen workload definitions and 0 measured reference sets, so they do not produce any public score yet. A frozen workload defines how a test is run; it does not mean a product has been measured.
Suite version 1.0.0. Missing measurements and references remain unknown; no values are estimated.
Step 1
A result belongs to one registry identity and scope. Sibling models, board revisions, and system configurations do not inherit scores.
Step 2
Deterministic parsers or Qwen structure public benchmark results; estimates and specifications do not enter performance indices.
Step 3
Code validates exact identity and context, then compares products within compatible source cohorts.
Step 4
Cohort-relative evidence connects through a benchmark graph to frozen research-derived references; raw observations stay unchanged.
Workload indices, platform capability, and memory impacts answer different questions and are never folded into one universal score.
Native gaming, ray tracing, upscaling, frame generation, AI, rendering, and GPU video encoding stay in separate workloads.
UOS-derived native gaming performance index from validated measured evidence.
Average FPS and 1% low FPS at native resolution, with ray tracing, upscaling and frame generation off.
UOS-derived AI workload performance index from validated measured evidence.
LLM prompt and generation speed, time to first token, image generation and embedding throughput on pinned models.
UOS-derived rendering performance index from validated measured evidence.
Render throughput on pinned scenes; CPU and GPU rendering are measured separately.
UOS-derived Blender Cycles rendering performance for the HIP backend.
Blender Cycles render throughput using HIP. Scores are comparable only within this HIP workload reference.
UOS-derived Blender Cycles rendering performance for the ONEAPI backend.
Blender Cycles render throughput using ONEAPI. Scores are comparable only within this ONEAPI workload reference.
UOS-derived productivity index from validated measured evidence.
Compression, video encoding and code compilation on pinned inputs, covering lightly and heavily threaded work.
Gaming performance with ray tracing on, at native resolution.
Average FPS and 1% low with ray tracing enabled; upscaling and frame generation off. Separate from the UOS Gaming Index.
Gaming performance with the hardware's upscaler enabled.
Average FPS with upscaling at a pinned quality mode; frame generation off. Separate from native gaming.
Displayed frame rate with frame generation enabled.
Displayed FPS including generated frames. Generated frames never count toward the UOS Gaming Index.
Gaming uses a fixed GPU and platform context. CPU rendering, AI, and productivity have separate measures.
UOS-derived native gaming performance index from validated measured evidence.
Average FPS and 1% low FPS at native resolution, with ray tracing, upscaling and frame generation off.
UOS-derived AI workload performance index from validated measured evidence.
LLM prompt and generation speed, time to first token, image generation and embedding throughput on pinned models.
UOS-derived rendering performance index from validated measured evidence.
Render throughput on pinned scenes; CPU and GPU rendering are measured separately.
UOS-derived productivity index from validated measured evidence.
Compression, video encoding and code compilation on pinned inputs, covering lightly and heavily threaded work.
UOS-derived compression performance index for one explicitly versioned 7-Zip test domain.
7-Zip compression throughput, normalized only against the same PTS profile, application version, options, unit and frozen comparison cohort.
UOS-derived CPU compilation performance for one explicitly versioned build workload.
Elapsed seconds on the exact pinned compilation profile, application version, options and frozen CPU reference.
UOS-derived CPU encoding performance for one explicitly versioned video encoding workload.
Frames per second for the exact pinned encoder profile, application version, input, options and frozen CPU reference.
Memory reports measured workload impact versus a frozen comparison platform; it has no standalone memory speed index.
Gaming change from this memory kit versus the UOS comparison platform.
Native gaming FPS change with only the memory configuration changed.
Local AI change from this memory kit versus the UOS comparison platform.
CPU-side AI throughput change with only the memory configuration changed.
Rendering change from this memory kit versus the UOS comparison platform.
CPU rendering throughput change with only the memory configuration changed.
Productivity change from this memory kit versus the UOS comparison platform.
Compression, encoding and compile time change with only the memory configuration changed.
Measured platform behavior and sustained performance are separate from sourced manufacturer capability.
Measured platform behavior of this motherboard.
Measured storage, PCIe and I/O behavior on a pinned test platform.
How well this motherboard sustains CPU performance under long load.
Sustained CPU throughput and VRM thermal behavior measured over a long run; never derived from VRM phase counts.
Measured memory stability and capability of this motherboard.
Highest stable memory configuration reached on pinned kits and CPU, verified by a stability test.
How much room this motherboard leaves for future upgrades.
Manufacturer-specified support for CPUs, memory, PCIe and storage; labeled as specification, not measurement.
Quality, electrical behavior, acoustics, and upgrade headroom use recorded lab conditions and measurements.
Overall measured quality of this power supply.
Combines measured electrical, efficiency and acoustic results; never brand, wattage or 80 PLUS tier alone.
Measured electrical performance of this power supply.
Voltage regulation, ripple, transient response, hold-up time and protection behavior at stated input voltage and load.
Measured noise of this power supply.
Noise at pinned load points and ambient temperature.
How much room this power supply leaves for a more demanding upgrade.
Measured transient and over-power behavior together with connector specification; never wattage alone.
Compute results belong to the exact system or configuration. Battery, sustained, storage, and thermal measures stay separate.
UOS-derived native gaming performance index from validated measured evidence.
Average FPS and 1% low FPS at native resolution, with ray tracing, upscaling and frame generation off.
UOS-derived AI workload performance index from validated measured evidence.
LLM prompt and generation speed, time to first token, image generation and embedding throughput on pinned models.
UOS-derived rendering performance index from validated measured evidence.
Render throughput on pinned scenes; CPU and GPU rendering are measured separately.
UOS-derived productivity index from validated measured evidence.
Compression, video encoding and code compilation on pinned inputs, covering lightly and heavily threaded work.
Battery runtime of this exact system configuration.
Runtime on a pinned workload at fixed display brightness, measured on the exact configuration; reported separately from performance.
How well this system holds performance under a long load.
Rendering throughput at the end of a 30-minute continuous load in this exact chassis; the same chip can differ between systems.
Measured internal storage speed of this configuration.
Sequential and 4K random read on the built-in SSD with a pinned fio job.
How hot this system gets under sustained load.
Peak surface and internal temperature above ambient under the sustained workload; lower is better.