Faster Scientific Simulations with New FP64 Algorithms in NVIDIA cuBLAS
Comments on “Faster Scientific Simulations with New FP64 Algorithms in NVIDIA cuBLAS”
Claim-checked. Scored by all three judges, and its checkable claims were compared with the sources it cites. How far stories are checked
NVIDIA's cuBLAS in CUDA Toolkit 13.4.2 now emulates FP64 matrix multiplication on integer Tensor Cores, using the Ozaki-II scheme for DGEMM and a 2M method that halves the real GEMMs needed for ZGEMM. NVIDIA reports DGEMM gains of 3.9x on GB200 and 5.5x on VR NVL72, and BerkeleyGW speedups of 3.9x and 6.6x. These figures come from NVIDIA's own benchmarks, and the accuracy match across BerkeleyGW configurations couldn't be independently verified. Applications need no code changes, and cuBLAS falls back to native FP64 when accuracy requires it.
Why it ranked
No. 19 of 123 with a score of 0.62. The panel rated it highest on new information (0.92) and technical depth (0.89), lowest on audience breadth (0.47). Placed No. 19, No. 40 by score alone: the placement terms moved other stories past it.
- AI judgeJudge 2claude-haiku-5-5
- Substantive technical writeup with benchmarks and accuracy validation on new FP64 emulation algorithms in cuBLAS; largely a vendor release post, moderately promotional in tone, relevant to HPC and GPU builders.
- AI judgeJudge 3gpt-5.6-luna
- A technically detailed cuBLAS/CUDA update explains novel exact FP64 emulation on integer Tensor Cores, with algorithmic details, accuracy safeguards, benchmarks, and scientific workloads. It is highly actionable for HPC/GPU users, though vendor-authored and hardware-specific.
Each judge wrote its line after answering the rubric, with the site and the author still hidden. Every answer behind this is in the table under Score.
Score
0.62total under rubric v2.3
- Significance0.67
- Reader interest0.73
- Audience breadth0.47
- Practical relevance0.70
- Substantiated0.83
- Technical depth0.89
- New information0.92
- Counts against: Promotional0.22
- Counts against: Overclaim0.32
Each judge’s answer per question (0–1), their mean and spread, and what the question adds to the score (weight × mean).
| Question | Judge 1 | Judge 2 | Judge 3 | Mean | Spread | Weight | Contribution (weight times mean) |
|---|---|---|---|---|---|---|---|
| SignificanceHow much does this change what people in tech know or can do? | 0.65 | 0.55 | 0.82 | 0.67 | 0.27 | +0.24 | +0.162 |
| Reader interestWould a well-informed senior engineer or tech founder want to read this today, and likely pass it on? | 0.64 | 0.72 | 0.84 | 0.73 | 0.20 | +0.26 | +0.191 |
| Audience breadthHow many people who build technology does it concern? | 0.44 | 0.50 | 0.48 | 0.47 | 0.06 | +0.13 | +0.061 |
| Practical relevanceHow useful is it to someone who builds software or tech products? | 0.72 | 0.60 | 0.78 | 0.70 | 0.18 | +0.22 | +0.154 |
| SubstantiatedHow well is it backed? Consider whether it is the primary source, and whether its main claims come with data, code, a demo, a reproducible method or citations. | 0.84 | 0.80 | 0.86 | 0.83 | 0.06 | +0.05 | +0.042 |
| Technical depthHow technically substantive is it? | 0.90 | 0.85 | 0.92 | 0.89 | 0.07 | +0.05 | +0.044 |
| New informationDoes it contain information absent from the stories covered in the last 7 days (list provided)? | 0.89 | 0.90 | 0.96 | 0.92 | 0.07 | +0.05 | +0.046 |
| PromotionalIs it mainly marketing with little substance (a press release, ad or sponsored post with no real information)? A launch post that carries real news is not promotional. | 0.17 | 0.25 | 0.25 | 0.22 | 0.08 | −0.19 | −0.042 |
| OverclaimDo the headline or conclusions claim more than the body supports? | 0.43 | 0.20 | 0.32 | 0.32 | 0.23 | −0.11 | −0.035 |
| Sum of contributions | +0.622 | ||||||
| Claim-check penalty | 0.000 | ||||||
| Total, as published | 0.622 | ||||||
Spread is the highest minus the lowest judge answer. marks a spread above 0.40: the judges disagree, and the story is marked contested.
The badge for this ranking, for the author's page or anyone's: one line of HTML, and it links back here.
<a href="https://rubric.hrih.si/editions/2026-11-14/19/"><img src="https://rubric.hrih.si/badge/2026-11-14/19.svg" alt="Ranked #19 on Rubric, 14 Nov 2026" width="245" height="28"></a>Claim check
Claim check: 2 matches source · 5 no source given · 1 couldn’t check. Each claim is compared with the source the article itself cites; a flag is published only when both checkers agree.
No source givenBenchmark result
cuBLAS DGEMM with Ozaki-II runs 5.5x faster than native FP64 on NVIDIA VR NVL72 and 3.9x faster on NVIDIA GB200.
The article gives no link or citation for it.
The article says “cuBLAS DGEMM with Ozaki-II runs 5.5x and 3.9x faster than native FP64 on NVIDIA VR NVL72 and NVIDIA GB200, respectively”
The source says No source to compare with.
No source givenBenchmark result
cuBLAS DGEMM with Ozaki-II is 2x faster than Ozaki-I on NVIDIA GB200.
The article gives no link or citation for it.
The article says “It’s also 2x faster than Ozaki-I on NVIDIA GB200”
The source says No source to compare with.
No source givenNumber
At the accuracy required to match FP64, Ozaki-I needs roughly twice as many Tensor Core GEMMs as Ozaki-II.
The article gives no link or citation for it.
The article says “At the accuracy required to match FP64, Ozaki-I needs roughly twice as many Tensor Core GEMMs as Ozaki-II”
The source says No source to compare with.
Matches sourceNumber
The 2M algorithm computes a complex matrix product with two real GEMMs instead of four, so a ZGEMM costs roughly twice a real GEMM instead of four times.
The cited source says the same thing.
The article says “the 2M algorithm computes a complex product with only two real GEMMs instead of four”
The source says “For floating-point matrices, 2M multiplication combines naturally with the Ozaki-II scheme, yielding a practical, high-performance algorithm for computing a complex floating-point GEMM in roughly twice the time of a real…”
Source cited by the article: research.nvidia.com/publication/2026-06_2m-multiplication-algorithm-complex-matrices
No source givenBenchmark result
BerkeleyGW using cuBLAS ZGEMM with the 2M algorithm achieves 6.6x speedup on VR NVL72 and 3.9x on GB200 over native ZGEMM.
The article gives no link or citation for it.
The article says “BerkeleyGW case study using cuBLAS ZGEMM with 2M algorithm achieves 6.6x and 3.9x speedup over native ZGEMM on NVIDIA VR NVL72 and NVIDIA GB200 respectively”
The source says No source to compare with.
No source givenBenchmark result
The CCSD case study using cuBLAS DGEMM with Ozaki-II achieves 2x end-to-end speedup on VR NVL72 and 1.4x on GB200 over native FP64.
The article gives no link or citation for it.
The article says “CCSD case study using cuBLAS DGEMM with Ozaki-II achieves 2x and 1.4x end-to-end speedup over native FP64 on NVIDIA VR NVL72 and NVIDIA GB200 in FP64”
The source says No source to compare with.
Matches sourceNumber
Using GPUs with BerkeleyGW can give an 86x performance speedup over the CPU-only implementation.
The cited source says the same thing.
The article says “Using GPUs with BerkeleyGW can lead to an 86x performance speedup over the CPU-only implementation”
The source says “Recently ported to GPUs, reaching 86x speedup compared to the CPU implementation”
Source cited by the article: berkeleygw.org
Couldn’t checkGeneralization
Native ZGEMM, Ozaki-II 2M with ADP and Ozaki-II 2M with 53 bits all produced the same Epsilon matrix elements up to 10 decimal places.
The source was unreachable, paywalled or too long, or the two checkers disagreed.
The article says “which all produced the same Epsilon matrix elements for up to 10 decimal places as specified in the NERSC-10 harness”
The source says No passage could be compared.
Source cited by the article: gitlab.com/NERSC/N10-benchmarks/berkeleygw-workflow/-/blob/main/benchmark/BGW_validate.sh
Also covered by
No other coverage was clustered with this story.
Elsewhere
Elsewhere: HN 2 pts, not on front page
Recorded for context only, checked 15 Nov 2026, 20:44 UTC. It played no part in the ranking.