• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

core i5 2500 on intel board throttling

A

ashetos

Senior member
I know it's ancient history but nonetheless I was debugging my sandy bridge system so here goes. Intel board DH67CL, core i5 2500. I was running linpack with the latest intel burn test utility, and the result is the CPU downclocking from 3.3GHz to 1.6GHz for about half the run duration. Maximum CPU package power was about 90 Watts but during the downclocking period it dropped to about 30 Watts.

I have been monitoring the temperature sensors with HWinfo, HWmonitor and Intel Desktop utilities. All of them don't show anything out of the ordinary, CPU at about 70 celcius, voltage regulator 41, PCH 45. The EFI menu does not show any additional sensors as well.

With stock BIOS settings, I get about 70 GFLOPS. I also tried disabling turbo and setting max clock to 2.8GHz and it still throttled to 1.6GHz. Maximum CPU package power at 65 Watts, still not low enough for some strange reason.

I also disabled 2 cores in the BIOS and ran the test again. I didn't notice any throttling.

Back to stock BIOS, 4 cores enabled. I also ran throttlestop and Clock and Chipset Modulation monitoring always showed 100%, still the CPU downclocked itself for no reason. Finally I disabled BD PROCHOT. After that, no throttling whatsoever and 90 GFLOPS.

So, one obvious explanation is that the motherboard itself is sending the PROCHOT signal to the CPU. The question is why? And how can I fix it? I fear that there is some electrical fault on the board, and some hidden sensor reports overheating or something.

Thanks!
 
It may actually be the board causing problems.

Have you tried any other intensive AVX workloads? How about y-cruncher?
 
I'm using Arctic Cooling Freezer 7 PRO Rev. 2 in a Coolermaster HAF 922 case.

No spare parts to test available.
 
ashetos said:
I'm using Arctic Cooling Freezer 7 PRO Rev. 2 in a Coolermaster HAF 922 case.

No spare parts to test available.
Click to expand...

Hmm. Have you tried any non-AVX versions of software to test your CPU? Like an older version of Prime95?

Or have you poked around in the BIOS to see if there are some temp limit settings you can tweak?
 
DrMrLordX said:
Hmm. Have you tried any non-AVX versions of software to test your CPU? Like an older version of Prime95?

Or have you poked around in the BIOS to see if there are some temp limit settings you can tweak?
Click to expand...

The BIOS settings are very limited, the temperature limits are way too comfortable, something like 110 degrees for the voltage regulator etc. There is really nothing relevant I can tweak, remember also that H67 is a locked chipset and 2500 is a locked chip. In theory, I'm not hitting any power limit of the CPU, the package power was well under 95 Watts on average on many of my experiments. Y-cruncher was also about 20 Watts lighter than linpack and it still throttled.

I could look for a non-AVX benchmark to try. From memory, I can say that non-AVX workloads don't throttle.
 
I still have a 2500k with less then 600 power on hours when ever that 2500k hits 72c something with the VRM and PWM circuits on that cheap asrock pro3 causes it to drop from 5000 Mhz to 4400 Mhz.
This 2500k is used less then 15 min a week by my wife.
 
john3850 said:
I still have a 2500k with less then 600 power on hours when ever that 2500k hits 72c something with the VRM and PWM circuits on that cheap asrock pro3 causes it to drop from 5000 Mhz to 4400 Mhz.
This 2500k is used less then 15 min a week by my wife.
Click to expand...

That sounds similar... Maybe my CPU cooler blowing with the direction of the airflow instead of towards the board is not helping...
 
ashetos said:
The BIOS settings are very limited, the temperature limits are way too comfortable, something like 110 degrees for the voltage regulator etc. There is really nothing relevant I can tweak, remember also that H67 is a locked chipset and 2500 is a locked chip. In theory, I'm not hitting any power limit of the CPU, the package power was well under 95 Watts on average on many of my experiments. Y-cruncher was also about 20 Watts lighter than linpack and it still throttled.

I could look for a non-AVX benchmark to try. From memory, I can say that non-AVX workloads don't throttle.
Click to expand...

Okay, so BIOS tweaking won't help. Try Prime95, after reading this:

https://www.mersenneforum.org/showthread.php?t=21462
 
I tried Prime95 after adding CpuSupportsAVX=0 in local.txt but it seems there is no difference whether this line exists or not for everything.

Voltage, power and throttling are identical. I think the option does not work, but there is no way to confirm, nothing gets logged by the programme...
 
Ok, so i debugged prime95 parser by adding bogus input in local.txt. It seems that somehow it remembers old state of local.txt even after you correct the error and save the file. The reliable way is to delete local.txt and edit local.txt again.

AVX and no AVX both throttle, and there is not much difference in power, maybe 5 Watts, but there is about 0.02 Volts difference. It starts throttling when the CPU hits about 70C so there is definitely some kind of a pattern here.
 
If I were in your shoes, I would either:

a). use this as an excuse to upgrade, since it's either your CPU or board doing it, or
b). get a cheap-arse sandy chip, maybe an i3-2120, pop it in there, and see if you can replicate the behavior
 
LTC8K6 said:
Try using XTU to possibly spot why its throttling?
Click to expand...
I did, it was useless. The other utilities proved to be more useful. There is no CPU throttling being reported anywhere. For some reason, the BD PROCHOT signal is responsible for the throttling.
 
Well, BD PROCHOT signal is pretty much always the board telling the CPU to throttle. If you can't find out why the board is sending the signal, then you need a new board.
 
Is there a "temp offset" in BIOS/UEFI somewhere? Maybe the BIOS is reading the 70C, and "adjusting" it to 90C, before it makes its throttling and fan-speed calculations.
 
VirtualLarry said:
Is there a "temp offset" in BIOS/UEFI somewhere? Maybe the BIOS is reading the 70C, and "adjusting" it to 90C, before it makes its throttling and fan-speed calculations.
Click to expand...

That's nice thinking but I remember the stock cooler hitting 98C 7 years back and throttling due to that. So I had to change it.
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top