E2-2697 v2 on X9DRI-LN4F+ Revision 1.2: Capped at 3 ghz

[FuriousGeorge]

The E2-2697 v2 should hit 3.5 ghz turbo, but Linux reports that max frequency is 3000.

Oddly, in the BIOS on my Supermicro X9DRI-LN4F+ the speed is reported as 2700, which is the correct 'base' speed, but when I enter the "socket" submenu in advanced->cpu configuration, I see that 2700 is actually listed as the max speed, and 1200 the minimum. Those settings can't be changed, and I can't see anywhere to set a multiplier, nonetheless, it does hit 3ghz according to Linux.

Supermicro's product page for the board says it must have at least bios version 3.0 to support the V2 chips, and mine has 3.4, released in 2019, 5 years after the CPU.

I've google around for a while and can't seem to find anyone with this problem.

Thanks in advance.

EDIT: I notice that when the cpus are at or near idle, for some reason a few cores will exceed 3.0 ghz. I've seen close to 3.3 ghz on one core, at one point, for no obvious reason. However, once I start ./cpu_intensive_process, all cores go to between 2992.869 and 2992.879 mhz.

 

[MBastian]

I've observed the same on E5-2690 v2s also on a x9dri-ln4f(Bios 3.3) and also on an old Dell E5530 modded with an i7-3630QM.
Bottom line was, at least when I shrugged an moved on: Unless you start disabling enough cores you will never see any core go (much) over the all-cpu all-core turbo frequency. The test was with a single-core benchmark(burnK6) nailed to a specific core. When started, all cores on that physical CPU did go to their all-core turbo frequency. Only when I disabled all but one core on one physical cpu I was able to hit the 3.6GHz single-core turbo.
Sometimes I wonder if it has something to do with the BIOS and Linux performance settings and if it is actually more beneficial to let the CPUs
cores try to save more power in order to enable a single core to reach max-turbo if necessary.
Edit: Same happens with Windows 10 on the x9dri-ln4f. I did try some Windows power settings. Always ends with 3.3 GHz all-core turbo.

 

[MBastian]

Sometimes I wonder if it has something to do with the BIOS and Linux performance settings and if it is actually more beneficial to let the CPUs
cores try to save more power in order to enable a single core to reach max-turbo if necessary.

This! I've got this nagging feeling and switched the BIOS Power Management Configuration: I've set Power Technology from Max Performance (or what it is called) to Energy Efficiency with a Balanced Performance Bias. I don't know if this is optimal but now a single process is able to hit the E5-2690v2 single-core turbo of 3.6 Ghz.

You should use `turbostat` and `cpufreq frequency-info` to get accurate values.

EDIT

Oddly, in the BIOS on my Supermicro X9DRI-LN4F+ the speed is reported as 2700, which is the correct 'base' speed, but when I enter the "socket" submenu in advanced->cpu configuration, I see that 2700 is actually listed as the max speed, and 1200 the minimum. Those settings can't be changed, and I can't see anywhere to set a multiplier, nonetheless, it does hit 3ghz according to Linux.

This is normal. Turbo is always on top of that.