PC Processors

[harmony]

Hello everyone,

Recently I upgraded from R5 3500 to R5 5600G after updating to the latest available BIOS of my Gigabyte B450 Aorus M motherboard (v F64 ). I am also using Deepcool AG400 LED cooler for enhanced cooling for the processor. Everything runs fine except the idle temperature shoots up & down continuously in the background without the system doing anything. Due to non stop temperature spike, the fan speed also ramping up & down continuously making the system noise a bit annoying in a quite working environment.

When I monitored CPU temp through both HWiNFO64 and Core Temp, nothing abnormal spike can be seen. But when I monitored the CPU through AIDA64 this continuous temperature spike can be seen in the graph quite clearly and also the corresponding fan speed spike.  What could be the reason for this behavior ? I have also updated the AMD Chipset driver to V 5.08.02.027 without help. 

Attached the Screenshot of the behavior for your reference. Please note that the system is in idle state without any background task running.  

5600G Idle Temp Spike

Regards

[zhembree]

Normally, temperature spikes like that would be correlated with at least some CPU activity, but it looks flat toward the end. It would be helpful to see how the frequency, power, and temp vary during the stress test. Generally, an uneven cooler mount or uneven thermal paste distribution could cause thermal issues, but I don't know if either of those would cause periodic random temperature spikes.

Cleaning, repasting, and remounting the cooler might work. Just make sure you have enough paste, and that the screws are evenly tight.

[harmony]

Last week while upgrading my DRAM, after cleaning I re-applied MX4 thermal paste as part of routine maintenance. The cooler is seated rather well. As requested, I have done stress test using Prime95 and noticed something very unusual. There are spikes happening as before but this time some thermal spike is as high as 127c according to AIDA64. This reading cannot be found under Core Temp or HWiNFO64. I have attached the screenshot for reference. What could be at fault here ? 

-My new DRAM is Corsair Vengeance Lpx DDR4 3600MHz 32GB (16GB x 2) Model CMK16GX4M1Z3600C18.

-The power supply Antec NE650G 650 Watt (NE650G M)

-Currently waiting for new GPU to arrive and the graphics is driven by the 5600G itself.

Regards

[harmony]

When I reset BIOS and didn't load Corsair XMP 3600mhz profile and running at default DRAM frequency of 2667mhz and this Thermal Spike phenomenon completely gone both at idle and load. Attached screenshots for reference.

Idle, Corsair XMP Not loaded

Load, Corsair XMP not loaded

Regards.

[zhembree]

It's a relief to hear the thermal spikes you showed in your other post have stopped. At the risk of stating the obvious, temperatures of 127C are extremely dangerous, and could easily kill your new CPU. If those readings are accurate, and your board is behaving correctly, over-temperature protection (OTP) should've been triggered. If that had happened, then your system should've immediately cut power. No Windows shutdown, no warnings, just a black screen and silence.

Assuming OTP is working properly on your motherboard, I'd guess that either those temperature readings are erroneous or the temperatures weren't sustained long enough to trigger OTP. If those temperatures are accurate, the fact they were resolved by resetting your BIOS and leaving XMP/DOCP off suggests your board might've been supplying far too much voltage, probably CPU VDD or SoC VDD.

I'd recommend using the "load optimized defaults" option in your BIOS if you haven't already, leave DOCP off, and save. Run another stress test, but this time use a utility like CPUID HWMonitor alongside AIDA64 to monitor package power, voltage, and temperature. If you see temps like that again, in either application, stop immediately. If you don't see the problem recurr, reenable your RAM's DOCP profile and retest.

[harmony]

I did "load optimized defaults" and in both idle and stress test is fine without any thermal spikes. Screenshot below for reference.

Default BIOS with no XMP loaded.

But when I loaded Corsair XMP, thermal spike at idle again started showing in the AIDA64. This time I do not have the courage to try Prime95 again. 

To try something different, I again reset the BIOS and this time I manually set DRAM clock speed (3600mhz), timing and voltage of 1.35V and when monitored through AIDA64 the thermal spike was very much happening at idle. I also tried down-clocking to 3200mhz at the same voltage and timing and the problem continue to persists.

Should I try downgrading my BIOS to the previous version ?

[zhembree]

At least your CPU's thermals, voltage, power, and frequency appear to be in spec. I checked the memory QVL for your board here, but couldn't find a list for Cezanne chips, and the Vermeer list, for 5000-series CPUs without graphics, doesn't list your memory kit as being supported.

If I had to guess, I'd say either your BIOS UI is somehow misrepresenting the voltages being supplied, or it's doing something painfully stupid in the background (like juicing the SoC voltage) because it's trying to deal with an unsupported memory kit. If it's a new bug, which I doubt, downgrading might work. The good news is that if you leave the memory at the default JEDEC speeds, your system appears to run normally. I wish I could offer more insight, but I'm out of ideas.

[zhembree]

On second thought, I'm not entirely sure that 127c reading was accurate. As it turns out, +127 is the largest value that can be represented by an 8-bit signed integer, and the graph you posted seems to show a near-instantaneous 60c temperature delta. That's suspicious by itself, and the fact you said those temps weren't reported in other software seems to reinforce that suspicion. It also explains why the motherboard didn't trigger OTP.

I'd recommend re-running those stress tests with HWMonitor and Ryzen Master. If they don't show those temp spikes, then I'd say the values AIDA64 reported are incorrect. If you're using third-party fan curve software for your cooler, then it may also be misreading the CPU temperatures and causing the periodic fan speed ramping you described.

[harmony]

Thank you for your insight. I just downgraded my BIOS to the next lower level. But the same behavior continues. I tried stress testing in AIDA64, while monitoring through HWMonitor and AIDA64. As usual only AIDA64 reports the thermal spikes.  Screenshot attached for reference.

Older BIOS, Factory Defaults, XMP loaded 

-I would have ignored AIDA64, but what continue to concern me is, this idle temp spike can be seen even inside BIOS menu and the corresponding fan speed spike. Things will be normal if I don't load XMP and only use standard DRAM speed of 2667Mhz. I don't know if I should continue using XMP and ignore this spike or just use standard DRAM speed of just 2667mhz and forget the extra speed available on DRAM.

Regards

[zhembree]

I suppose whether or not you want to reenable XMP ultimately depends on whether you'll actually miss the performance, and if that difference is worth the effort. Third-party temperature monitoring software can be unreliable, but it's worth noting that AIDA64 and HWMonitor appear to be showing drastically different readings in your last screenshot. The 100+ readings persist in AIDA64, but HWMonitor is showing a peak of 75C. In fact, everything in HWMontior looks perfectly normal, voltages, power, everything.

[harmony]

I explained this to Gigabyte support. Lets see what they have to say, if anything.

I may stick with the standard jedec 2667 speed in the meantime. 

Thank you.