Bios corrupted, CPU downclocking under load??

[Filip Georgievski]

As title says, im suspecting the BIOS as the culprit for this issue.

Last night, i came across some weird FPS drops when playing PUBG. So i naturaly expected my GPU or CPU were overheating.
So i took out the panel and used a thermometer with a probe to mesure the temps, cause i dont trust software.
And to my suprise, temps on both GPU and CPU were sitting within 65C (+/- 2C).

Today, in the morning, ran some stress tests using Prime 95 and CPU ID/HW as monitors for CPU since i did search on google, and google said CPU.

So i got ahead and did the tests, and in the middle of them, i noticed my CPU downclocking to like 1.6Ghz and usage droping from 100% to like 40%.

What gives?? I got SpeedStep and Turbo Boost disabled, plus temps are within range, and voltage on the CPU is like 1.25V which is plenty for my OC.

Suggestions on this matter are welcome.

 

[Bill_Bright]

It is extremely rare for the BIOS/UEFI firmware to become corrupted after a successful flash. Even corruption during a flash/update is very rare these days.

So something else likely happened. Nothing wrong with GPU temps of 65°C. CPU temps of 65°C is pretty warm, but not really considered "hot".

The fact your CPU throttled down during Prime95 indicates to me the thermal protection features are working. Since you did not tell us what your temps normally sit at, or what they were during the Prime95 run, we have nothing to compare.

Is the case interior clean of heat trapping dust? Are all fans spinning properly? What happens with other games? Has your ambient (room) temperature changed?

 

[Filip Georgievski]

Case is maintained clean every month, room as well so very little to no dust in case.

Fans are ramping up and down as expected on idle or load, with my ASUS board being hybrid, and case fans are set to ramp up and down.

Othet games is the same story, stutter for about 10 to 15 sec, and back to normal fps, drops from 60 to 20 and back to 60fps.

Ambient room is a little bit warmer since is winter, range from 30 to 35C in my room.

Every stress test does the same odd drop in clocks, multiplier and voltage (if it is not set manualy in BIOS, from 1.2 it goes to 0.9V).

Idk what can cause this, i was torn between corrupted BIOS (SpeedStep doing some funky s..t since i noticed i was stuck on Disabled, and couldnt change it to Enabled in Bios) or VRMs getting hot (since Mobo on sensors is also following CPU temp going to 65C).

 

[Reeves81x]

As title says, im suspecting the BIOS as the culprit for this issue.

Last night, i came across some weird FPS drops when playing PUBG. So i naturaly expected my GPU or CPU were overheating.
So i took out the panel and used a thermometer with a probe to mesure the temps, cause i dont trust software.
And to my suprise, temps on both GPU and CPU were sitting within 65C (+/- 2C).

Today, in the morning, ran some stress tests using Prime 95 and CPU ID/HW as monitors for CPU since i did search on google, and google said CPU.

So i got ahead and did the tests, and in the middle of them, i noticed my CPU downclocking to like 1.6Ghz and usage droping from 100% to like 40%.

What gives?? I got SpeedStep and Turbo Boost disabled, plus temps are within range, and voltage on the CPU is like 1.25V which is plenty for my OC.

Suggestions on this matter are welcome.

doesnt matter if you trust the software temps, your computer trusts them, if it reads too high a temp it will throttle. why dont you check the software and get back to us. heat sink temp and TJ temp are very different...

 

[cdawall]

Take off the OC and verify it works at stock correctly first. Could be something as simple as you triggering OCP on the mosfet for the board. Considering that board is only a cheap 3 phase H55 board and you are overclocking in essence a quad core i7 I doubt the board can handle it at 3.7ghz. I say this after having a higher end 4 phase H55 motherboard and the same CPU and smoking the mosfets on it.

 

[Filip Georgievski]

Software reading are acurate as my probe thermometer, 65C max on CPU, and 65C max on GPU as well.

I can even push my Xeon to 4Ghz and get it to 85C and not throttle, but it does the same thing over and over, going back and forth from 3.6 to 1.6Ghz, and back again, not changing temp, 65C on 3.6Ghz, it is like it is bootlooping, but when it is under load.

Edit: Same at stock, ramps to 3.4ghz with SpeedStep, and for some reason, being cooler than my OC, 55C at load, it still goes into downclocking and hampering CPU performance.

 

[Regeneration]

Clock throttling has nothing to do with BIOS corruption. Could be chipset/VRM/CPU ovearheating.

Try to disable all power saving (SpeedStep, C-states, C1E) and thermal protection features from the BIOS, clean all fans and reapply thermal paste.

 

[Reeves81x]

check your vrm with your probe. ^beat me to it After looking up you mobo. I would say it's almost definitely a power issue. That 3 phase board is pretty weak for oc'ing, It could very well be beginning to fail.

 

[Filip Georgievski]

VRMs follow CPU temps, so at stock under load, 55C and at OC load, 65C.

 

[Solaris17]

Software reading are acurate as my probe thermometer, 65C max on CPU, and 65C max on GPU as well.

I can even push my Xeon to 4Ghz and get it to 85C and not throttle, but it does the same thing over and over, going back and forth from 3.6 to 1.6Ghz, and back again, not changing temp, 65C on 3.6Ghz, it is like it is bootlooping, but when it is under load.

Edit: Same at stock, ramps to 3.4ghz with SpeedStep, and for some reason, being cooler than my OC, 55C at load, it still goes into downclocking and hampering CPU performance.

So basically you are arguing all of our suggestions thus far are wrong without any concrete testing as of yet. What is it exactly you want us to just tell you?

 

[Filip Georgievski]

Not arguing, answering all questions asked. So i make it easyer on you guys to give me some pointers.

SpeedStep, C1 State and C1E disabled on my OC profile, still same stuff.
Enabled by defaults, still same stuff.

 

[newtekie1]

Could be chipset/VRM/CPU ovearheating.

Looking at the board and the really bad VRM design, I'd almost guarantee the VRMs are starting to overheat or even starting to go bad after the years of abuse.

 

[Regeneration]

Reset BIOS configuration and see if it also happens on stock settings.

 

[Filip Georgievski]

Looking at the board and the really bad VRM design, I'd almost guarantee the VRMs are starting to overheat or even starting to go bad after the years of abuse.

That was one of my initial suspects.

Reset BIOS configuration and see if it also happens on stock settings.

Did that as well, ran Prime 95 stock for 10 minutes and still same results.

1 thing i forgot to point out.
Usage when downclocked dropes to 40% in Task Manager from 100% while doing stress tests, and after a few secs goes back up as clocks and voltage return to normal.

 

[Reeves81x]

After looking up you mobo. I would say it's almost definitely a power issue. That 3 phase board is pretty weak for oc'ing, It could very well be beginning to fail.

 

[cdawall]

VRMs follow CPU temps, so at stock under load, 55C and at OC load, 65C.

You pulled the heatsink and checked VRM temps or pointed a wildly inaccurate temp gun at a heatsink and got 65C?

 

[Reeves81x]

you don't happen to have Phase control in your bios do you? dunno if asus has it.

 

[Regeneration]

Did that as well, ran Prime 95 stock for 10 minutes and still same results.

So its either overheating, power-related issue, or maybe a BIOS bug. Try a different BIOS version, you can also try using ThrottleStop. Take in mind H55 is a bad chipset for overclocking.

 

[Filip Georgievski]

You pulled the heatsink and checked VRM temps or pointed a wildly inaccurate temp gun at a heatsink and got 65C?

Used a thermometer with a wired probe to measure VRM temps.

you don't happen to have Phase control in your bios do you? dunno if asus has it.

I dont think so, i only has voltage regulation and a preety simple (Enable\Disable) Line Load Calibration.

 

[Regeneration]

If you have "Xtreme Phase Full Power Mode" option in the BIOS, try to play with it.

 

[Filip Georgievski]

I

So its either overheating, power-related issue, or maybe a BIOS bug. Try a different BIOS version, you can also try using ThrottleStop. Take in mind H55 is a bad chipset for overclocking.

I am well aware of thaf that is why i resorted to 3.6 / 3.7Ghz oc since it is not far of original turbo boost clock of 3.4Ghz.

If you have "Xtreme Phase Full Power Mode" option in the BIOS, try to play with it.

It does, but it is just, enable/disable thing, and does not give any extra features to play with.

 

[Regeneration]

It does, but it is just, enable/disable thing, and does not give any extra features to play with.

So try both settings...

 

[Filip Georgievski]

Ok ill try tonight do some more scenarios, and play around with some of the settings, but at this point, all points out to a bad mobo, and im not happy about that.
Ill get back here tonight with some more info and screen shots with all info about temps, clocks, voltages and stuff.

 

[Bill_Bright]

Every stress test does the same odd drop in clocks, multiplier and voltage

And in the past with this same machine and setup, were you able to play any game, run any stress test and your system didn't have an drop in clocks, multiplier and voltage?

If so, how far in the past?

Lousy VRM "design" would not cause throttling to start now when it used to work fine. Lousy VRM "design" would be apparent the first time you tasked the system in this same manner.

Poor quality components used in the design could result in premature failure or degradation of parts.

Do you have a decent quality, spare PSU of sufficient power to swap in?

 

[Filip Georgievski]

I would have to buy a new one to do that.
Or maybe i can get my friend's Evga 650 gold nova g2 which is same output as mine, but more efficient and newer, so i could do that.