Phoronix article – Spectre V4 performance (2018-05-22)
Phoronix posted an article measuring Spectre V4 changes. These tests happen with kernel patches for the Spectre 3a (rogue system register read) and Spectre 4 (speculative store bypass) fixes are applied to the kernel. There is an AMD whitepaper describing the vulnerability and fix.
The primary places these might affect performance are in the kernel and hence several I/O related benchmarks are measured. The I/O writes and forking microtests show differences and more real-world apps do not.
Overall, I haven’t dug too deeply into these since more an effect of the kernel changes than a reflection of the processors. However list it here so I can find the article later.
| Benchmark | Phoronix observations | My observations | Analysis |
|---|---|---|---|
| iozone | Depends on system and the underlying I/O configuration | File system benchmark (not analyzed further) | |
| postmark | Not much difference | File system benchmark (not analyzed further) | |
| build-linux-kernel | Extremely slight effect | Real world example, 82% On_CPU; mixing disk and actual compile. | Analysis |
| pgbench | No measurable effect | Driver process of database, ~50% On_CPU | Analysis |
| stress-ng: forking | Small decreases | Kernel tests for fork operation | Analysis |
Comments
Phoronix article – Spectre V4 performance (2018-05-22) — No Comments
HTML tags allowed in your comment: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>