Once again, @ErrataRob derails my brilliant plan to build an MSP430 workstation and server platform.https://twitter.com/ErrataRob/status/980907462973902849 …
You can always have same perf (assuming //izable jobs) just by adding more cores. At perf parity, ARM is ~half power consumption.
-
-
That's what everyone thinks. Not people who know CPUs, of course, but everyone else.
-
Apparently it's what
@Cloudflare thinks.https://twitter.com/eastdakota/status/976560820611031040?s=20 …
-
That's because it's a network processor that includes fixed-function accelerators for crypto and compression that would otherwise be done in software on Intel CPUs.
-
I've seen similar results on big compile jobs, no acceleration. Need to test myself sometime.
-
My semi-educated guess is that you can do way better than 1/2 power consumption (maybe as low as 10%) just by using a larger array of simpler cores rather than higher-end out-of-order ARMs.
-
Of course if Intel made simple cores they could be competitive here too, but they abandoned the Atom microarchitecture that was the closest thing they had (and likely their only hope for making chips unaffected by Spectre).
-
I want to see a server using a Xeon Phi (72 Atom cores) do a Linux compile. The massive floating point logic may make things inefficient, but it'd still be fun to try.
-
Wow interesting. I had no idea Larrabee was non-OOOE. I might have to pick one up as a better Spectre-free option than the S1260 I've got now.
- 3 more replies
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.