Platforms · Measurement record
Recordh200-2026-11-05Outputs validatedFixed
NVIDIA H200
Measured on 5 November 2026 on one GPU, with the published method. 26 double-precision measurements of the tuned code for this platform: five launches each, every launch recorded.
One whole GPU, on the 96 host threads of its memory node. Default power limit, ECC on, clocks not locked.
- Published method
- Run and checked by Xcelerit
- Suite version: November 2026 round
- Double precision
- Accelerator
- NVIDIA H200
- Power limit
- 700 W
- Driver
- 595.91.07
- Host CPU
- Intel Xeon 5th gen (Emerald Rapids), 2 × 48 cores, 192 threads
- Host memory
- 2,015 GiB
- Operating system
- Ubuntu 24.04.5 LTS
- Compiler
- CUDA 12.9.1, sm_90
- Libraries
- cuRAND
Results at a glance
The sizes the results rank at. Time is the mean of the middle three of five launches; throughput is the problem size over that time (Black-Scholes: options × 100 passes, the work of one call). Sizes in K and M are binary (1K = 1,024); throughput is per second, in decimal units. Tuned code: tuned for this chip under the same rules. Select a row to see the launches behind it.
| Workload | Size | Time (ms) | Throughput (per s) | Energy | Within-launch CI | Launch spread | |
|---|---|---|---|---|---|---|---|
| Swaption portfolio, LMM | 1,024K paths | 16.29 | 64,372,896 | 2.003 Wh/bn | ±0.15% | 0.01% | Inspect |
| Barrier options, Monte Carlo | 16,000 options | 10.60 | 1,509,071 | 0.08987 Wh/M | ±1.00% | 0.51% | Inspect |
| American options, binomial lattice | 512K options | 304.67 | 1,720,837 | 57.39 Wh/bn | ±0.97% | 0.82% | Inspect |
| European options, Black-Scholes formula | 128M options | 290.26 | 46,240,043,276 | 2.426 Wh/tn | ±0.51% | 0.74% | Inspect |
All 26 measurements
| Workload | Code | Size | Time (ms) | Within-launch CI | Launch spread | Average warm calls per launch |
|---|---|---|---|---|---|---|
| Swaption portfolio, LMM | Tuned code | 16K | 0.609 | ±0.78% | 0.54% | 6,574 |
| Swaption portfolio, LMM | Tuned code | 32K | 0.662 | ±0.85% | 0.44% | 4,303 |
| Swaption portfolio, LMM | Tuned code | 64K | 1.175 | ±0.50% | 0.29% | 2,350 |
| Swaption portfolio, LMM | Tuned code | 128K | 2.281 | ±0.73% | 0.28% | 985 |
| Swaption portfolio, LMM | Tuned code | 256K | 4.380 | ±0.63% | 0.64% | 982 |
| Swaption portfolio, LMM | Tuned code | 512K | 8.334 | ±0.36% | 0.20% | 241 |
| Swaption portfolio, LMM | Tuned code | 1,024K | 16.289 | ±0.15% | 0.01% | 123 |
| Swaption portfolio, LMM | Tuned code | 2,048K | 31.837 | ±0.18% | 0.03% | 63 |
| Barrier options, Monte Carlo | Tuned code | 2,000 | 1.948 | ±1.00% | 1.81% | 21,952 |
| Barrier options, Monte Carlo | Tuned code | 4,000 | 3.074 | ±1.00% | 0.62% | 13,235 |
| Barrier options, Monte Carlo | Tuned code | 8,000 | 5.562 | ±1.00% | 0.83% | 6,702 |
| Barrier options, Monte Carlo | Tuned code | 16,000 | 10.603 | ±1.00% | 0.51% | 3,571 |
| Barrier options, Monte Carlo | Tuned code | 32,000 | 20.411 | ±1.00% | 0.71% | 1,550 |
| American options, binomial lattice | Tuned code | 16K | 10.125 | ±1.00% | 0.21% | 1,823 |
| American options, binomial lattice | Tuned code | 32K | 19.776 | ±1.00% | 0.41% | 755 |
| American options, binomial lattice | Tuned code | 64K | 38.935 | ±1.00% | 0.18% | 343 |
| American options, binomial lattice | Tuned code | 128K | 76.518 | ±1.00% | 0.39% | 105 |
| American options, binomial lattice | Tuned code | 256K | 152.330 | ±0.88% | 0.49% | 32 |
| American options, binomial lattice | Tuned code | 512K | 304.670 | ±0.97% | 0.82% | 26 |
| American options, binomial lattice | Tuned code | 1,024K | 605.337 | ±0.62% | 0.38% | 10 |
| American options, binomial lattice | Tuned code | 2,048K | 1,241.156 | ±0.67% | 0.31% | 10 |
| European options, Black-Scholes formula | Tuned code, pinned host memory | 16M | 37.668 | ±1.00% | 0.67% | 964 |
| European options, Black-Scholes formula | Tuned code, pinned host memory | 32M | 73.470 | ±0.99% | 1.21% | 151 |
| European options, Black-Scholes formula | Tuned code, pinned host memory | 64M | 145.291 | ±0.99% | 0.06% | 29 |
| European options, Black-Scholes formula | Tuned code, pinned host memory | 128M | 290.263 | ±0.51% | 0.74% | 21 |
| European options, Black-Scholes formula | Tuned code, pinned host memory | 256M | 580.075 | ±0.25% | 0.09% | 10 |
Outputs validated before speed
A time only counts if the output is right. At every problem size, the option price the tuned code computes is validated against the reference implementation's on the same machine.
| Workload | Against the reference code | Result |
|---|---|---|
| Swaption portfolio, LMM | Within 0.78% at 128K paths, on this platform's own random-number generator | Pass |
| Barrier options, Monte Carlo | Within 0.18% at 32,000 options, on this platform's own random-number generator | Pass |
| American options, binomial lattice | Identical to double precision | Pass |
| European options, Black-Scholes formula | Identical to double precision | Pass |
Binomial and Black-Scholes are deterministic: the tuned code gives the reference code's prices. Monte Carlo: on GPUs the barrier and LMM codes generate their random numbers differently, so their prices differ only by sampling: within 4 standard errors of the reference implementation's, and fed the same random numbers they agree with the reference implementation's to within 1e-9 (relative).
Every launch behind one measurement
Pick a workload and a size. Each dot is one launch: the mean of its warm calls, run until the 95% confidence interval was within ±1% or until the call limit. The fastest and slowest launch are dropped and the middle three averaged into the result.
Zoom: drag a box on the chart or use the slider under it; reset brings it back.
| Launch | Warm calls | Warm mean (ms) | Range (ms) | Used |
|---|
How time grows with the problem
Time against size on log scales, for every size measured. The dashed line is perfect linear scaling from the smallest size: points under it mean the machine was not yet full at small sizes.
What the record holds
The measurement record behind this page: every launch, the machine and the output validation, and the results as a table.
Full record (JSON) Results (CSV)
Free to cite with attribution: Xcelerit Finance Benchmarks, record h200-2026-11-05.
SHA-2561039edc6461608f3bba232775c64d381089a87a7c43d4d1e03cc60cdb74e49c5
Records are fixed: once published, a record is never edited. A correction is published as a new record, linked from the old one. Checksums of every record