GPUTrust Logofloprate
floprate pipelineInitialising shaders: 0%
Scroll Down

From Telemetry to Trust

When a seller lists a GPU, they run our native benchmark client directly on their machine. Floprate verifies the hardware identity, monitors real-time sensors under stress, and generates four domain-specific scores.

How We Prevent Fraud:

Every test is cryptographically signed using a Trusted Platform Module (TPM) quote, proving the exact GPU model and configuration. Telemetry logs remain tied to the hardware profile, removing technical guesswork.

GPU Telemetry Pipeline (RTX 4090)
Telemetry Metric
Raw Value
Attestation Status
TPM Signature
Quote Verified
Waiting
FP32 Performance
82.6 TFLOPS
Waiting
FP16 Performance
82.6 TFLOPS
Waiting
FP64 Performance
1.3 TFLOPS
Waiting
Tensor FP16 Matrix
162.4 TFLOPS
Waiting
Tensor BF16 Matrix
162.4 TFLOPS
Waiting
Tensor TF32 Matrix
81.2 TFLOPS
Waiting
Tensor FP8 Matrix
324.8 TFLOPS
Waiting
Memory Bandwidth
1,008 GB/s
Waiting
Avg Power Draw
345 Watts
Waiting
Max Power Draw
415 Watts
Waiting
Avg Temperature
61°C
Waiting
Max Temperature
68°C
Waiting
System ready. Click "Run attestation" to start calculations.

Awaiting Pipeline Run

Start the telemetry simulation on the left to witness how raw calculations translate into verified marketplace scores.

Built on Cryptographic Trust

We eliminate the friction and uncertainty of used hardware trading through cryptographic evidence and secure financial safeguards.

TPM Verification

Every benchmark result is validated using a Trusted Platform Module hardware quote. Spoofed specs or forged logs are instantly rejected, ensuring cryptographic trust.

Stress Telemetry

We monitor real-time sensors for temperature, power draw, and clock speeds under sustained load, identifying hardware degradation and cooling inefficiencies.

Domain Scoring

Raw performance values are mapped to verified domain scores (AI, Rendering, Gaming), giving a clear, standardized overview of capabilities for different use cases.

floprate© 2026 Floprate. All rights reserved. Peer-to-peer trust layer for high-performance compute.