21 ms·

Geekbench? Yeah, that's nice. https://browser.geekbench.com/processors/2134 https://browser.geekbench.com/processors/2134 If you believe a tablet can beat a 2

Geekbench? Yeah, that's nice. https://browser.geekbench.com/processors/2134 https://browser.geekbench.com/processors/2134 If you believe a tablet can beat a 24-core $1500 server processor in a highly threaded workload, you may be interested in my upcoming ICO! Geekbench is only good as a punchline.

akmarinov 8y ago
Do you believe that the 24-core $1500 server processor can beat the iPad in single to minimum threaded workload?
- asdgklnionio 8y ago
  In practice I think the iPad would throttle. But let's assume it has adequate cooling. The 7401 is clocked at two GHz. It's by no means a single-threaded monster. So maybe.
Analemma_ 8y ago
Even if the absolute value of the Geekbench score isn’t useful, the fact that the delta between the MBP Geekbench score and the iPad Geekbench score has narrowed from huge to zero would still be meaningful.
- asdgklnionio 8y ago
  The delta between an iPad and a $5000 server is zero, but the iPad is $700 and draws only a dozen watts! That's why Google is throwing out all their servers and filling their datacenters with iPads. I know I'm being flippant, but I'm sick and tired of talking about Geekbench.
  - mcphage 8y ago
    What's a better metric?
altmind 8y ago
i also feeling something fishy here. i remember geekbench reporting different memory latencies than AIDA64 a while ago. maybe apple took geekbench and optimized their silicon for the test load(crypto)? or maybe geekbench for arm and geekbench for x86_64 may be doing the different tasks? geekbench web site have the list of benchmarks( http://support.primatelabs.com/kb/geekbench/interpreting-geekbench-4-scores http://support.primatelabs.com/kb/geekbench/interpreting-gee... ), but is light on testing methodology and there is no way to verify what/how benchmarks are performed. Unrelated: Linus Torvalds trashing gb3 in mailing list, a long thread https://www.realworldtech.com/forum/?threadid=136526&curpostid=136666 https://www.realworldtech.com/forum/?threadid=136526&curpost... Edit: added the link to gb4 benchmarks list

akmarinov 8y ago

Do you believe that the 24-core $1500 server processor can beat the iPad in single to minimum threaded workload?

asdgklnionio 8y ago
In practice I think the iPad would throttle. But let's assume it has adequate cooling. The 7401 is clocked at two GHz. It's by no means a single-threaded monster. So maybe.

Analemma_ 8y ago

Even if the absolute value of the Geekbench score isn’t useful, the fact that the delta between the MBP Geekbench score and the iPad Geekbench score has narrowed from huge to zero would still be meaningful.

asdgklnionio 8y ago
The delta between an iPad and a $5000 server is zero, but the iPad is $700 and draws only a dozen watts! That's why Google is throwing out all their servers and filling their datacenters with iPads. I know I'm being flippant, but I'm sick and tired of talking about Geekbench.
- mcphage 8y ago
  What's a better metric?

altmind 8y ago

i also feeling something fishy here. i remember geekbench reporting different memory latencies than AIDA64 a while ago. maybe apple took geekbench and optimized their silicon for the test load(crypto)? or maybe geekbench for arm and geekbench for x86_64 may be doing the different tasks? geekbench web site have the list of benchmarks( http://support.primatelabs.com/kb/geekbench/interpreting-geekbench-4-scores http://support.primatelabs.com/kb/geekbench/interpreting-gee... ), but is light on testing methodology and there is no way to verify what/how benchmarks are performed. Unrelated: Linus Torvalds trashing gb3 in mailing list, a long thread https://www.realworldtech.com/forum/?threadid=136526&curpostid=136666 https://www.realworldtech.com/forum/?threadid=136526&curpost... Edit: added the link to gb4 benchmarks list