AMD accuses SYSbench of Intel bias
January 19, 2016 | 13:08
Companies: #amd #federal-trade-commission #ftc #intel
AMD has hit out at popular benchmark SYSmark, claiming its results are biased towards Intel processors and it offers a skewed comparative between the products of the two rival companies.
Even the most ardent AMD fan would have to admit that the company has lost its way in recent years. After capitalising on Intel's NetBurst architecture folly to take the performance crown, it wandered down the same garden path itself with predictable results: rival Intel regained its performance lead when it abandoned NetBurst, and today anybody with performance in mind would always buy Intel first with AMD a distant second for those who need more powerful integrated graphics.
While AMD has admitted a few missteps, which it hopes to resolve with the impending launch of the far-more-efficient Zen microarchitecture, the company has released a video which downplays the performance differential between its products and those of Intel by attacking the SYSmark benchmark. In 'Truth or Myth: Is SYSmark a Reliable Benchmark,' the company describes results from SYSmark showing a 50 per cent performance differential between a laptop-centric Intel Core i5 processor and a comparable AMD FX chip as 'quite astonishing and not realistic in what real-life performance is like,' pointing to somewhat less synthetic benchmarks - such as PCMark 8 and a set of Microsoft Office scripts - which show the performance differential to be closer to 7 per cent in Intel's favour.
The core of AMD's complaint, though, is not that SYSbench's CPU-centric synthetic workloads are a bad indicator of real-world performance but that SYSbench itself is biased in Intel's favour. In support of its complaint, AMD refers to an FTC ruling (PDF warning) which forced Intel to include a disclaimer that 'software and workloads used in performance tests may have been optimised for performance only on Intel microprocessors,' and which specifically calls out SYSmark and MobileMark, when the company uses comparative benchmarking in promotional materials.
For AMD, however, the video may get across its distaste for SYSmark but in doing so causes the financially ailing company undeniable embarrassment: its own testing still shows that Intel chips outperform comparable AMD chips, albeit at a fifth of the differential claimed by SYSmark.
AMD's video is reproduced below.
Even the most ardent AMD fan would have to admit that the company has lost its way in recent years. After capitalising on Intel's NetBurst architecture folly to take the performance crown, it wandered down the same garden path itself with predictable results: rival Intel regained its performance lead when it abandoned NetBurst, and today anybody with performance in mind would always buy Intel first with AMD a distant second for those who need more powerful integrated graphics.
While AMD has admitted a few missteps, which it hopes to resolve with the impending launch of the far-more-efficient Zen microarchitecture, the company has released a video which downplays the performance differential between its products and those of Intel by attacking the SYSmark benchmark. In 'Truth or Myth: Is SYSmark a Reliable Benchmark,' the company describes results from SYSmark showing a 50 per cent performance differential between a laptop-centric Intel Core i5 processor and a comparable AMD FX chip as 'quite astonishing and not realistic in what real-life performance is like,' pointing to somewhat less synthetic benchmarks - such as PCMark 8 and a set of Microsoft Office scripts - which show the performance differential to be closer to 7 per cent in Intel's favour.
The core of AMD's complaint, though, is not that SYSbench's CPU-centric synthetic workloads are a bad indicator of real-world performance but that SYSbench itself is biased in Intel's favour. In support of its complaint, AMD refers to an FTC ruling (PDF warning) which forced Intel to include a disclaimer that 'software and workloads used in performance tests may have been optimised for performance only on Intel microprocessors,' and which specifically calls out SYSmark and MobileMark, when the company uses comparative benchmarking in promotional materials.
For AMD, however, the video may get across its distaste for SYSmark but in doing so causes the financially ailing company undeniable embarrassment: its own testing still shows that Intel chips outperform comparable AMD chips, albeit at a fifth of the differential claimed by SYSmark.
AMD's video is reproduced below.
MSI MPG Velox 100R Chassis Review
October 14 2021 | 15:04
Want to comment? Please log in.