4 ms·
Pretty sure the op was asking for a citation specifically about aws nodes. Should be simple enough to take a test workload on Intel vs amd nodes and measure thi
by eyegor 7y ago
Pretty sure the op was asking for a citation specifically about aws nodes. Should be simple enough to take a test workload on Intel vs amd nodes and measure things (besides "safety" which is too nebulous to measure). Consumer chips aren't entirely relevant.
- Ygg2 7y agoI'm pretty sure I linked server chip comparison. And the same guys that design consumer chips, design server, so there will be a lot of "cross-pollination" between architectures. We definitely see that a lot of Intel consumer and server grade is hit with similar bugs. Safety can be roughly estimated by a number of critical vulnerabilities. Most have hit Intel, some being Intel specific bugs.
- wahern 7y agoBut the mitigations in AWS' VM monitor might be a fixed cost--i.e. not disabled on AMD architectures or otherwise not disadvantaging Intel. For example, they might use techniques like page coloring as an alternative to cache flushing and prediction suppression, which wouldn't disadvantage Intel, might not impose any significant overhead at all, and might even improve performance in shared hosting environments. Alternatively, maybe their support for AMD is nascent and optimizations for Intel absent on AMD. There's really no substitute for benchmarking AWS' actual offerings.