Methodology & Results

AI Performance Is a Property of the System, Not the Chip

LynxBenchAI measures AI performance as a property of the whole stack, sustained, per precision, with bounded optimisation. The results are live below: ranked by Global Score, re-rankable by workload, every number backed by its method.

Explore the leaderboard See every AI Executor

See It Applied

Ranked by Global Score

Provisional results — not the final, official release. LynxBenchAI is building in public: methodology is final, the results feed is live and growing.

#	Device Accelerator / GPU	GT Global Score	Training Category	Inference Category	Compute Category	Platform OS
1	NVIDIA B200	1264.11	1065.97	2203.38	585.14	Native Linux
2	NVIDIA A100-SXM4-80GB	431.56	411.48	652.75	207.50	Native Linux
3	NVIDIA GeForce RTX 3080	157.47	162.69	237.65	64.77	Native Linux
4	NVIDIA GeForce RTX 3080 (WSL)	156.12	156.52	228.56	72.46	WSL
5	NVIDIA GeForce RTX 5060 Ti	143.03	132.65	235.72	61.23	Native Linux
6	NVIDIA RTX 2000 Ada Generation Laptop GPU	73.67	60.47	99.69	59.72	WSL
7	AMD Radeon RX 9060 XT	73.61	37.00	119.39	110.74	Native Linux
8	Intel Graphics (Intel Core Ultra 7 165U)	8.02	6.65	9.15	8.95	Native Linux
9	NVIDIA GeForce MX550	5.87	8.47	4.88	4.07	WSL
10	AMD Ryzen 5 9600X	4.78	8.54	2.25	6.77	Native Linux

The Personal Edition submits your benchmark results to TechnoLynx's servers as part of participating in the public leaderboard. If that isn't acceptable for your use case, contact TechnoLynx about Press, Pro, or Enterprise licenses.

Contact TechnoLynx

LynxBenchAI End-User License Agreement · LynxBenchAI Privacy Policy

Every rank opens into the category and precision breakdown that produced it — the number is not the end of the story.

Reproducibility manifest and declared optimisation budget on every device page.

Re-rank by training, inference, or compute — the sort is a lens on the same board, not a different number.

Browse every device →

Why This Exists

Today's AI Hardware Benchmarks Mislead in Three Predictable Ways

Each one looks reasonable in isolation. Together they explain why the chart-topping number a buyer evaluates and the sustained throughput their workload actually sees can differ by an order of magnitude.

Spec sheets describe theoretical limits, not delivered performance.

Peak numbers are rare and brief; production workloads don't run on bursts.

Unbounded, undeclared optimisation makes results uncomparable.

Why spec-sheet benchmarking fails for AI →

What LynxBenchAI Measures

Four Principles, Applied Uniformly to Every Entrant

The methodology is engineered so that every published number means a specific, reproducible thing, and so two numbers can be compared without reading footnotes.

The AI Executor Is the Unit of Measurement

Methodology

Performance is a property of the hardware and the software together: driver, runtime, framework, kernels, all of it. The same chip under different stacks delivers materially different throughput. We measure the pair, not the silicon. Read the article →

Sustained, Not Peak

Methodology

We measure steady-state throughput under continuous, realistic load, the number the system can actually hold once thermal limits, memory bandwidth, and power budgets all assert themselves. Bursts are noted; deployments run on what's sustainable. Read the article →

Every Result Is Precision-Tagged

Methodology

FP8, FP16, BF16, and INT8 are different operating regimes, each with its own accuracy, throughput, and economic profile. We report each separately, and the Global Score that rolls them up never hides which regime it came from. Read the article →

Optimisation Is Bounded and Declared

Methodology

Every entrant is tuned within the same effort budget, recorded in the manifest, reproducible by a third party. Unbounded tuning makes results incomparable; bounded tuning makes them useful. Read the article →

Coverage

What's on the board today, and what we're engineering toward next. Inclusion in the roadmap list is intent, not a certified result.

On the Board Now

NVIDIA GeForce RTX 5060 Ti

AMD Radeon RX 9060 XT

Intel Arc Graphics

AMD Ryzen 5 9600X (CPU)

FP32

FP16

BF16

FP8

INT8

FP64

Engineering Toward

NVIDIA Blackwell

AMD MI350

Intel Gaudi

TPU / NPU / XPU classes

Edge deployment

Frequently Asked Questions

What is LynxBenchAI?

LynxBenchAI is a benchmarking methodology and a corresponding set of results for AI hardware. It measures performance as a property of the complete hardware-and-software stack, sustained under realistic load, reported per precision, with bounded optimisation. It ranks every AI Executor by a Global Score, but the ranking exists to be interrogated, not taken on faith: every rank opens into the category, precision, and method that produced it.

Why does the methodology matter as much as the results?

A number without a settled methodology means whatever the reader assumes. Publishing the method alongside the results means what each score measures, and doesn't, is unambiguous rather than implied. Why methodology defines what you can compare →

What is an "AI Executor"?

The AI Executor is the hardware-and-software pair that actually runs the workload: the GPU together with its driver, runtime, framework, and kernel implementations. Performance is a property of this pair, not of the silicon alone, and the same chip under a different stack can perform very differently. Read the article →

How does this differ from existing AI benchmarks?

Three differences at the method level. First, we publish sustained throughput, not peak. Second, every result is reported per precision, and the Global Score that rolls them up is a declared aggregate, not a hidden one, so what went into it is always visible. Third, every entrant is tuned under a declared, bounded optimisation budget, so two numbers compare on equal footing. The aim is procurement-grade comparability, expressed as a transparent rank rather than an opaque one. Why benchmarks mislead procurement →

Who is LynxBenchAI for?

Buyers and platform engineers who need defensible evidence for AI hardware decisions, procurement reviewers who need a comparable basis between systems, and the technical leadership inside vendors who want their products evaluated on the load they were designed for, with the trade-offs visible. Read the decision framework →

Where do I see the results?

On the leaderboard, ranked by Global Score with a full drill-down to category, precision, and method for every entry. Coverage keeps expanding: if there's a specific device you want measured, tell us about your hardware →.

Where to Start

Six articles that lay out the premise, the central claim, and the procurement application of LynxBenchAI, spanning benchmark methodology, stack-level reasoning, sustained performance, precision trade-offs, and hardware selection.

Why Spec-Sheet Benchmarking Fails for AI — How GPU Benchmarks Actually Work

Apr 14, 2026

GPU spec sheets describe theoretical limits. Real AI performance is an execution property shaped by workload, software, and sustained system behavior.

Performance Emerges from the Hardware × Software Stack

Apr 15, 2026

AI performance is an emergent property of hardware, software, and workload together.

How to Choose AI Hardware and GPU for AI Workloads: A Decision Framework

Apr 16, 2026

A decision framework for choosing AI hardware: define the decision, match evaluation to deployment, weigh total cost of ownership, preserve tradeoffs.

Peak Performance vs Steady-State Performance in AI

Apr 15, 2026

AI systems live in steady state, not at peak. This article explains the distinction, when each regime applies, and why peak-only evaluations mislead…

Precision Is a Design Parameter, Not a Quality Compromise

Apr 16, 2026

Numerical precision is an explicit design parameter in AI systems, not a moral downgrade in quality — a representation choice with intentional trade-offs.

Are GPU Benchmarks Accurate? What They Actually Measure vs Real-World Performance

Apr 14, 2026

A GPU benchmark measures an execution path, not the silicon. Stack, workload, and measurement window shape the number — read them or be misled.

Browse all articles See all