There's a worrying article about recent Intel chips in this month's UK edition of PC PRO magazine (no. 361, cover date October 2024). It says that "... There have long been reports of instability with Intel's "Raptor Lake" processors, with systems crashing under heavy load, and even dead chips." Apparently, it appears to mainly affect the bigger i7 and i9 versions

I've been following that story. If I understand correctly, the chip can push too much voltage to the cores under certain circumstances causing failure if it happens for extended period. There's a microcode patch out now and also motherboard bios updates. I've read that Arrow Lake has the voltage locked at 1.2V but I'm not clear if that was earlier silicon or the final chips. I checked the voltages on my 13th gen and they looked fine before any patching, even running all cores with CPDN tasks, so it might be a more complicated use case.

I noticed the article specially, because I've just attached host 1552676 - i5-14400. The system builder has built it round a Gigabyte motherboard, and I see they have a BIOS update dated 7 August - when the machine's quiet I'll perhaps look to see if they've applied that - and maybe flash it before loading it to the max.

Definitely improve the cooling if you hit 90C without all the cores running.

It is maybe a single spike hitting 90C every five minutes or so. according to Psensor if I go above 12 cores running. 75% of the time it is below 80C

It is maybe a single spike hitting 90C every five minutes or so. according to Psensor if I go above 12 cores running. 75% of the time it is below 80C

You probably know that AMD chips will throttle usually above 90C (assuming it's a desktop). Varies slightly from chip to chip. Might still be worth investigating additional cooling. Ought to be able to run all cores without throttling.

Might still be worth investigating additional cooling.

I will first look at adding an additional exhaust fan. Probably should not have gone for a midi case!

Might still be worth investigating additional cooling.

I will first look at adding an additional exhaust fan. Probably should not have gone for a midi case!

What cooling have you got on the CPU? Might not be the case itself.

Curiously, what are these machines that take 1000+ hours to complete a task?

Curiously, what are these machines that take 1000+ hours to complete a task?

A couple of machines ago, I had a core 2 duo. I suspect that would have taken a while. An N2830 Celeron I found was taking over 600hours a task.

Longest time I found was on computer 6,515,846 A core 2 duo with 1809 hours for a task but it was completing the majority of tasks in a fifth of that time. When I ran tasks on an old netbook they would take months to complete tasks that finished on my core2 duo in a couple of weeks.

Further to heat issues, if I restrict activity by percentage of CPUs BOINC is allowed to use, the temperature goes up a lot higher than if I have the same number of tasks running but I am suspending tasks to stop them running. That is using WINE. I will when current tasks finished, see if the same happens with a VM or with native Linux tasks.

I am not coming up with any logic behind what I am seeing.
Edit: I shall over the next few days do some more investigation using top and other tools to see if I can find any clues.

Is the CPU is throttling on the higher load?

Not that I can see. But restricting by suspending tasks rather than setting number of CPUs in use leads to between 5 and 10C lower temps the bigger difference with more cores in use. It will be interesting to see if the same applies with native LInux or with BOINC in a VM. But that will be a week away as I want to get the tasks I have finished first.

Been delving a little more into the temperatures on my Ryzen9 and what the figures in Psensors mean. I found this Tctl, the highest temperature seen seems to be the temperature in the core of the chip. Edge, sensor1, sensor2 and composite temperatures have never been above 50C. The discussion I read on the link seems to suggest that Tctl going above 90C is nothing to worry about.

I believe 7950X max temp. is 95C which I think means that it's sustainable for long stretches. Modern CPUs are highly unlikely to overheat and break, they'll throttle. You can probably run your CPU between 90-95C just fine for long stretches of time. One thing to look into is undervolting. You'll likely find that you can run the CPU at a given speed a good amount cooler.

I have a 5900X which has a 90C max temp. I've had it run at or slightly higher than 90C on really hot days, as per CoreTemp. I might reduce the load a bit if it looks like it's going to be really hot for days in a row but otherwise I don't worry about temps.

With 14 cores running, my 7950X has a couple of spikes every five minutes above 90C. Mostly it is below 85C.
Till I started doing some reading about what the different temperatures indicated I had decided to cut down a bit. I have gone back up to 45% of CPUs now.

With 14 cores running, my 7950X has a couple of spikes every five minutes above 90C. Mostly it is below 85C.
Till I started doing some reading about what the different temperatures indicated I had decided to cut down a bit. I have gone back up to 45% of CPUs now.

If your running Windows you can install Ryzen Master and set to ECO Mode. It will run a lot cooler and use less electricity.

https://www.amd.com/en/products/software/ryzen-master.html

You can acheive the same results in Linux but you'll have to change things in the BIOS.

I am using Linux and WINE to run BOINC. I will have a play with the bios once the current batch of work is finished. I am assuming RyzenMaster won't work in a VM.