MSI Pro Z790-A Max increased temps after BIOS update

[music2024]

I set up an Intel based media and music composition computer earlier this year with a similar set up to @arli14f902cb with thanks to @PaulieDC for helping me set up everything. I am using an MSI Pro Z790-A Max MB with Intel 13700K processor and Cooler Master Hyper 622 Air cooler and BeQuiet 500FX case. I do not have a GPU, using the integrated Intel.

No problems all these months, but now after reading all the Intel 13/14 series issues, it made me wonder about my computer temps and possible CPU degradation, so I was following arli's thread on this issue and later on based on that Paulie and I decided to update my BIOS since it was getting outdated.

1)
I was using BIOS 7E07vM1 from November 2023 all these months. No real problems cooling anything with Cinebench R23 tests or my production testing (which so far isn't really CPU intensive since my music production is mostly RAM and SSD dependent). C R23 score 29771.

2) New BIOS Update version 7E07vM52. With the new BIOS it allowed me to select different cooling modes and using the Intel Power scheme with the 253W setting.

After the BIOS update, tons of reds in HWInfo including Package and IA Core going to 102. It's around 10-20 degrees change compared to the old bios Cinebench test data in the prior image. My Cinebench R23 score also dropped by around 800 points probably due to the heat and throttling?

What do you experts make of this? Was the old BIOS just not reporting accurately, or is the new BIOS having bugs with the power? I should note I'm only having these issues with intense Cinebench testing. On my day to day productivity and media production tasks that do not really punish the CPU, no cooling issues and fans only ramp up moderately or very low.

3)

I'd also like to ask what's causing these Core Temperature and CPU Package graph drops during the Cinebench R23 test where there are momentary large dips in temps? Is it due to throttling then it cools down momentarily or some kind of issue to be concerned over?

Thanks for any input on these issues.

 

[citay]

What do you experts make of this? Was the old BIOS just not reporting accurately, or is the new BIOS having bugs with the power?

Higher core voltage in the new BIOS version, leading to higher power draw, leading to higher thermal load and eventually thermal throttling, leading to lower frequencies and lower score. I believe the reporting is accurate. This is because MSI chose some suboptimal settings as defaults.

Sadly you don't include all the sensor data in the first screenshot, then it would be more obvious. But these are not "bugs" per se, this is most likely MSI trying to fix something and actually making things worse for you as a result.

If you read here, i list two steps to 1) protect your cooling, so you find out what amount of heat it can get rid of and set that as the power limits in the BIOS (because the Intel numbers are more or less arbitrary, they don't correspond to your CPU cooler's abilities whatsoever), and 2) to shave off some of that additional VCore again.

The drops in the graph are probably an effect of the thermal throttling, yeah (at some point it can become really aggressive).

 

[music2024]

Higher core voltage in the new BIOS version, leading to higher power draw, leading to higher thermal load and eventually thermal throttling, leading to lower frequencies and lower score. I believe the reporting is accurate. This is because MSI chose some suboptimal settings as defaults.

Sadly you don't include all the sensor data in the first screenshot, then it would be more obvious. But these are not "bugs" per se, this is most likely MSI trying to fix something and actually making things worse for you as a result.

If you read here, i list two steps to 1) protect your cooling, so you find out what amount of heat it can get rid of and set that as the power limits in the BIOS (because the Intel numbers are more or less arbitrary, they don't correspond to your CPU cooler's abilities whatsoever), and 2) to shave off some of that additional VCore again.

The drops in the graph are probably an effect of the thermal throttling, yeah (at some point it can become really aggressive).

Thank you so much for your informative and from what I have seen of your posts, extremely knowledgeable responses. What you are saying is what @PaulieDC told me was likely the case. I appreciate you confirming his excellent help and analysis. I wanted to post this here just to get some further feedback because it took me a long time to build and save for this PC build, but this Intel stuff all over the news had me worried that my CPU may be causing issues that I was not aware of.

Thank you for that link. I read it and definitely will be looking into that with help from Paulie.

In your opinion based on what I posted in the OP, do I have anything to worry about either in general or this whole Intel CPU problem going on in the news recently?

 

[citay]

Well, a 13700K is a true Raptor Lake CPU, and an i7 already, so it is "somewhat" affected (not all CPUs, but some), even though the most affected ones are the i9 models, because they are more "on the edge". It's also preferable that it's not a 14700K, because that is also much more extreme than a 13700K. Once you dial down the power limits to what your cooler can comfortably handle, which won't be anywhere near the 253W that Intel state (which, as i said, has nothing to do with the cooling, it's just a matter of preventing the worst), then you already take the edge off things. With your cooler, it's better to start with their Baseline profile for the i7, meaning 188W, this is something that's much more realistic. Even though it's also an artbitrary number by Intel when it comes to the cooling, because they have no idea what actual cooler people use. But it's a much better starting point for such an air cooler. 250W, that's firmly in AIO territory already.

 

[TXF]

or this whole Intel CPU problem going on in the news recently?

Probably not if your CPU was manufactured in 2023.

 

[music2024]

Well, a 13700K is a true Raptor Lake CPU, and an i7 already, so it is "somewhat" affected (not all CPUs, but some), even though the most affected ones are the i9 models, because they are more "on the edge". It's also preferable that it's not a 14700K, because that is also much more extreme than a 13700K. Once you dial down the power limits to what your cooler can comfortably handle, which won't be anywhere near the 253W that Intel state (which, as i said, has nothing to do with the cooling, it's just a matter of preventing the worst), then you already take the edge off things. With your cooler, it's better to start with their Baseline profile for the i7, meaning 188W, this is something that's much more realistic. Even though it's also an artbitrary number by Intel when it comes to the cooling, because they have no idea what actual cooler people use. But it's a much better starting point for such an air cooler. 250W, that's firmly in AIO territory already.

Thank you!

 

[music2024]

Probably not if your CPU was manufactured in 2023.

I don't know when it was manufactured but I bought it last year (2023) on Black Friday sale.

 

[atentatora]

I did upgrade to the same BIOS ver. 7E07vM52(Beta version) with the same MB (PRO Z790-A MAX WIFI) and observed default elevated values for AC/DC.
So, in order to achieve 40500 pts. in CB23 with 13900K & Noctua NH-D15 G2, the settings that worked for me were:

Cooling option: Intel performance settings
PL1&PL2: 253W
Iccmax: 400A (using 307A capped the power at around 180W)
Load line calibration: Auto (which I believe is level 8)
DC loadline: 98 (as per this post and it seems that VID and Vcore match almost perfect for LLC auto, but if you're on another LLC level, you should change DC loadline as well)
AC loadline: 25 (with 20 CB24 froze, but CB23 didn't freeze)
IA CEP: disable (it's now enabled by default)

With these settings, max Vcore on light load is 1.36V and on full load floats around 1.19V. Max CPU temp during CB23 & CB24 is 85 degrees.

 

[music2024]

I did upgrade to the same BIOS ver. 7E07vM52(Beta version) with the same MB (PRO Z790-A MAX WIFI) and observed default elevated values for AC/DC.
So, in order to achieve 40500 pts. in CB23 with 13900K & Noctua NH-D15 G2, the settings that worked for me were:

Cooling option: Intel performance settings
PL1&PL2: 253W
Iccmax: 400A (using 307A capped the power at around 180W)
Load line calibration: Auto (which I believe is level 8)
DC loadline: 98 (as per this post and it seems that VID and Vcore match almost perfect for LLC auto, but if you're on another LLC level, you should change DC loadline as well)
AC loadline: 25 (with 20 CB24 froze, but CB23 didn't freeze)
IA CEP: disable (it's now enabled by default)

With these settings, max Vcore on light load is 1.36V and on full load floats around 1.19V. Max CPU temp during CB23 & CB24 is 85 degrees.

Thank you for your input!

 

[music2024]

Higher core voltage in the new BIOS version, leading to higher power draw, leading to higher thermal load and eventually thermal throttling, leading to lower frequencies and lower score. I believe the reporting is accurate. This is because MSI chose some suboptimal settings as defaults.

Sadly you don't include all the sensor data in the first screenshot, then it would be more obvious. But these are not "bugs" per se, this is most likely MSI trying to fix something and actually making things worse for you as a result.

If you read here, i list two steps to 1) protect your cooling, so you find out what amount of heat it can get rid of and set that as the power limits in the BIOS (because the Intel numbers are more or less arbitrary, they don't correspond to your CPU cooler's abilities whatsoever), and 2) to shave off some of that additional VCore again.

The drops in the graph are probably an effect of the thermal throttling, yeah (at some point it can become really aggressive).

Post Bios update
View attachment 190633

Ok we read your comments and info both your above link and what you suggested to Arli's set up. I needed guidance on it from PaulieDC given his experience helping with the music technical aspects and working with Arli's on their similar set up.

Our first step was to approach it first starting at 220W for Long Duration Power Limit and Short Duration. Current Limit was put at 307 instead of Auto. My Cinebench R23 score dropped 1200 which is not too big a deal to me since I'm not looking for maximum performance, but temps and peak temps dropped a lot which seems good to me compared to the post BIOS update with all the reds. Comparison vs the above below.

Then per your above link to check for stability I ran OCCT CPU tests in short bursts at various modes then did CPU+RAM on Large/Extreme for an hour which ran a bit hotter than Cine. No errors or issues in either Cine R23/OCCT as far as crashing or anything.

What's your opinion on the new Cine Monitoring results compared to the prior one post BIOS update?

Thanks for your viewpoint.

 

[citay]

Looking pretty good. I think you could lower the Long Duration power limit by another 20W to be even safer about your cooling. Then you got the power limits part taken care of, and could attempt to lower the voltage a bit by lowering CPU Lite Load and checking for stability.

 

[music2024]

Looking pretty good. I think you could lower the Long Duration power limit by another 20W to be even safer about your cooling. Then you got the power limits part taken care of, and could attempt to lower the voltage a bit by lowering CPU Lite Load and checking for stability.

Thanks so much for looking it over and the input. Very much appreciated! For now since I do not game and am not really doing heavy CPU stuff I will keep it at this setting for now until the new upcoming Intel code. If I do start to push the system more or get into gaming I will follow up with your stronger measures to keep it under control.

One last question. I have been using the CPU (not daily and not pushing it hard) on and off since early this year, is there anyway to determine whether my CPU has degraded in this time span due to the Intel issues?

 

[polosistealth]

hi, what bios do i have to update to fix the new intel micro code bug? M.52? we have to wait for a new bios?

 

[citay]

One last question. I have been using the CPU (not daily and not pushing it hard) on and off since early this year, is there anyway to determine whether my CPU has degraded in this time span due to the Intel issues?

As for checking wether an individual (true) 13th/14th gen CPU is affected, this particular video recommends to install the NVIDIA drivers 5-10 times or so, as it contains a bunch of decompression workload, which is supposedly good to reveal any instability in this case. I think this is not the last word on a good method. For example WinRAR has an integrated benchmark (ALT-B) that also checks for errors, presumably doing compression/decompression, maybe that is something that could also work to detect problems, i don't know.

Basically, the early warning sign of CPU degradation is what people initially reported that got this whole thing rolling, namely, crashes or errors during strenuous decompression tasks. For example when launching a game, at the point where it decompresses the textures to feed the GPU. People started getting crashes there, with a message about a GPU out of memory error (but it really was because of the CPU, because the decompressed data was invalid).

If you have any signs of instability, then you want to do further tests. If you never have any instability anywhere, there's no need to worry over something that may never manifest itself.

 

[music2024]

hi, what bios do i have to update to fix the new intel micro code bug? M.52? we have to wait for a new bios?

It isn't out yet. Intel said something around August.

 

[music2024]

As for checking wether an individual (true) 13th/14th gen CPU is affected, this particular video recommends to install the NVIDIA drivers 5-10 times or so, as it contains a bunch of decompression workload, which is supposedly good to reveal any instability in this case. I think this is not the last word on a good method. For example WinRAR has an integrated benchmark (ALT-B) that also checks for errors, presumably doing compression/decompression, maybe that is something that could also work to detect problems, i don't know.

Basically, the early warning sign of CPU degradation is what people initially reported that got this whole thing rolling, namely, crashes or errors during strenuous decompression tasks. For example when launching a game, at the point where it decompresses the textures to feed the GPU. People started getting crashes there, with a message about a GPU out of memory error (but it really was because of the CPU, because the decompressed data was invalid).

If you have any signs of instability, then you want to do further tests. If you never have any instability anywhere, there's no need to worry over something that may never manifest itself.

Thank you so much.

 

[polosistealth]

7zip has also decompression benchmark

 

[music2024]

Looking pretty good. I think you could lower the Long Duration power limit by another 20W to be even safer about your cooling. Then you got the power limits part taken care of, and could attempt to lower the voltage a bit by lowering CPU Lite Load and checking for stability.

Old firmware

View attachment 190752

@citay MSI released the microcode beta which improved things last month but then my score dropped. However I just noticed MSI released a non beta firmware with the microcode patch earlier this month and just installed it. My Cine score went up and temps seem to be improved compared to the older screen. What do you think of the below compared to the above? Appreciate your knowledgeable input. Thank you.

First Run, lower Cine score but no Core/Package throttle.

Second Run right after the above, and Core/Package turned red but score improved.

 

[citay]

There's just a 3% difference, i wouldn't put too much importance on these run-to-run differences. For example, during the first run you might have had some background processes taking a bit of CPU time away, the scores drop a bit and the CPU stays a bit cooler, because whatever else is taking up CPU time will be lower load than Cinebench. Then in the second run you don't have this interruption and the score can improve a bit. Have you lowered CPU Lite Load to your CPU's actual needs? That should give the biggest score improvement.

 

[music2024]

There's just a 3% difference, i wouldn't put too much importance on these run-to-run differences. For example, during the first run you might have had some background processes taking a bit of CPU time away, the scores drop a bit and the CPU stays a bit cooler, because whatever else is taking up CPU time will be lower load than Cinebench. Then in the second run you don't have this interruption and the score can improve a bit. Have you lowered CPU Lite Load to your CPU's actual needs? That should give the biggest score improvement.

Thank you so much for the explanation and good to know everything is good at these settings. Thanks so much for weighing in as I trust your opinion much. Thanks.

I have not gotten into the Lite Load stuff yet as I am not really too familiar with those indepth adjustments. Right now the PC is performing great for productivity tasks and in my day to day use the CPU is very low load and temps are generally only in the 40- mid 50s C max when I'm working in my music program. So I decided to just keep things simple and continue the stable settings, for now anyway.