Accelerator
NVIDIA GeForce RTX 4090: 24 GiB of GDDR6X, 1008 GB/s peak bandwidth, and 82.58 TFLOP/s FP32.
The reference GeForce RTX 4090 keeps its AD102 FP32 peak, GDDR6X bandwidth, and card power in separate measurements rather than treating them as one score.
| Component | Role | Count | Clock | Capacity |
|---|---|---|---|---|
| NVIDIA Ada Lovelace (NVIDIA GeForce RTX 4090) The lower declared clock; the separately named upper clock is recorded as a peak-clock measurement. | GPU | 1 | 2235 MHz Nominal clock 2520 MHz Peak clock | — |
| NVIDIA GeForce RTX 4090 frame buffer | Video memory | 1 | — | 24 GiB |
| Metric | Value | Evidence | Basis | Method | Source |
|---|---|---|---|---|---|
| Memory bandwidth whole-system | 1008 GB/s | Confirmed | Derived | Theoretical peak Computed here from figures the vendor stated Adds the one fitted frame-buffer bandwidth. No host memory is included. |
| Metric | Value | Evidence | Basis | Method | Source |
|---|---|---|---|---|---|
| Clock frequency gpu | 2235 MHz | Confirmed | Vendor-rated | Nominal clock Stated by the vendor The lower declared clock. The separately named upper clock is a different quantity and uses peak-clock. | |
| Clock frequency gpu | 2520 MHz | Confirmed | Vendor-rated | Peak clock Stated by the vendor The vendor’s upper boost clock, distinct from the nominal clock. | |
| Peak FP32 rate gpu | 82.58 TFLOP/s | Confirmed | Theoretical | Theoretical peak Stated by the vendor General shader units at the 2520 MHz boost clock. A theoretical general-shader FP32 peak; tensor, AI and ray-tracing rates are excluded. | |
| Geekbench Compute score, opencl gpu | 255 036 | Confirmed | Measured | Benchmark chart average Published independently geekbench-compute 7 , opencl A dated capture of the benchmark publisher’s rolling average for this graphics processor, read from the 25 July 2026 OpenCL chart. Version 7 calibrates a compute result against a baseline of 100,000, so it may be compared only with another Geekbench 7 OpenCL figure from this curation round, never with a Metal figure, an earlier version or a processor score. | |
| Rated power consumption gpu | 450 W | Confirmed | Vendor-rated | Rated power Stated by the vendor The vendor’s board-power rating, not a measured system draw. |
| Metric | Value | Evidence | Basis | Method | Source |
|---|---|---|---|---|---|
| Memory bandwidth memory | 1008 GB/s | Confirmed | Theoretical | Theoretical peak Stated by the vendor Vendor-stated maximum transfer rate of this memory interface, not a benchmarked application result. | |
| Memory capacity whole-system | 24 GiB | Confirmed | Vendor-rated | Design capacity Stated by the vendor The frame buffer fitted to the reference accelerator; host memory is excluded. |
1 008 000 000 000 B/s
a + b + … sum v1 Adds 1 memory-bandwidth figure obtained by theoretical-peak. Nothing in the catalog is left out of it. A total is only as trustworthy as the decision about what belongs in it, so both the terms and the exclusions are named records.
This accelerator’s whole-system memory is its fitted frame buffer; host memory is excluded. Partner cards may differ from this reference configuration.