3 ms·
I don't think so. I'm guessing that the xeons feature set is not used at all and this specific payload is just much better on that arm. There is a big chance,
by sigi45 9y ago
I don't think so. I'm guessing that the xeons feature set is not used at all and this specific payload is just much better on that arm.
There is a big chance, that whatever cloudflare is doing on that machine, is massive parallel and simple.
- Arnt 9y agoThese nine specific payloads, if you take the tweet's words literally.
- thecompilr 9y agoThose are indeed workloads, specific for a webserver and even more specific for CF. Many other workloads that we don't care about that Intel can handily win of course. We don't have use for most Intel extensions, like AVX512 and barely use any AVX2. Intel is the king of HPC and single threaded performance for a few more years. Although ARM can potentially leapfrog Intel in the HPC space, if someone implements the Salable Vector Instruction extension in 1024 or 2048 bit.
- pdpi 9y agoMy understanding is that some large parts of the feature set aren't usable anyhow. E.g. there's enough of a ramp up/ramp down power cost for the vector units that you'd never want to compile your binaries with SSE/AVX turned on unless your workload makes near constant use of those units.
- sigi45 9y agoI have never heard that one before. Do you have any sources for that?
- Jenya_ 9y agohttps://www.quora.com/Why-are-AVX-instructions-so-power-hungry https://www.quora.com/Why-are-AVX-instructions-so-power-hung... https://news.ycombinator.com/item?id=13925039 https://news.ycombinator.com/item?id=13925039
- pdpi 9y agoThanks for the quotes. Apparently I slightly misremembered the problem. From the second one, this is what I was thinking about: https://news.ycombinator.com/item?id=13926839 https://news.ycombinator.com/item?id=13926839