Most active commenters
  • BMSmnqXAE4yfe1(4)

←back to thread

1080 points antipaul | 11 comments | | HN request time: 1.135s | source | bottom
1. BMSmnqXAE4yfe1 ◴[] No.25067305[source]
Looking at Geekbench charts, I see that my Samsung phone has a faster processor than my $3000 gaming rig, and it sure as hell draws less power ... so why aren't datacenters just use phones instead of Xeon and Epyc servers, what am I missing?
replies(6): >>25067324 #>>25067327 #>>25067332 #>>25067335 #>>25067352 #>>25067577 #
2. gre ◴[] No.25067324[source]
Cell phone plans cost an arm and a leg!
3. akuma73 ◴[] No.25067327[source]
Nuvia is a startup that is planning to do just that https://nuviainc.com/blog

You need to be competitive on single thread performance to have a chance at datacenter. Amdahl's law is still very relevant. Up until very very recently, the CPUs were not up to par.

replies(1): >>25067425 #
4. Tiberium ◴[] No.25067332[source]
Are you sure that you compared results from the same major Geekbench version? (v4 and v5 have quite different results).
replies(1): >>25067387 #
5. dt3ft ◴[] No.25067335[source]
Probably due to overheating. Your phone can do bursts of processing every now and then, but if you let it execute heavy compute operations, the phone would probably overheat and shut itself off.
replies(1): >>25067366 #
6. crb002 ◴[] No.25067352[source]
Apple shit when they saw M1 benchmarks and delayed their data center for years to fill it with their own server chips. https://businessrecord.com/Content/Default/All-Latest-News/A...
7. regularfry ◴[] No.25067366[source]
So would a Xeon if you pulled the heatsink off.
replies(1): >>25067400 #
8. BMSmnqXAE4yfe1 ◴[] No.25067387[source]
I'm just looking at the charts, v5
9. BMSmnqXAE4yfe1 ◴[] No.25067400{3}[source]
So pull the insides of the phone out, attach a heatsink ... heatsinks cost less than LCD screens I reckon.
10. BMSmnqXAE4yfe1 ◴[] No.25067425[source]
Nice charts ... so performance per watt is indeed 10x better for S865 vs the Ryzen.
11. altmind ◴[] No.25067577[source]
Because geekbench results are not comparable across the platforms/OS. The scale is not consistent.