The computer keeps crashing during the compilation process of chromium-123.0.6312.58_1, but never crashed when I was building version 123.0.6312.58!

The instructions said that in order to stop mprime, I should hit CTRL+C. That's what I did to stop it, but I don't remember what message it printed when it stopped.
 
Oleg_NYC said:
Are you absolutely sure these crashes happen because of overheating, and not because of something else?
Click to expand...
Don’t think it’s about anyone being “sure” but thermal issues are often to blame for issues like you are seeing, and your CPU does seem to be getting quite toasty. But it could be dodgy power supply, some other hardware fault, some issue in FreeBSD.
 
Yeah, it ran for a couple of minutes.... But then I decided to run it again, and this time it ran for like an hour. I hit CTRL+C to stop it, and results.txt got bigger and it was filled with "passed" messages for 10K, 18K, etc.
 
shkhln said:
(That is apparently the same product as Thermaltake UX100.)
Click to expand...

That thing's tech specs don't even mention the maximum wattage it can dissipate. And it is being used on the most power-hungry 10th gen Intel CPU they have.
thermaltakeusa.com

UX100 ARGB Lighting CPU Cooler

UX100 ARGB Lighting CPU Cooler combines 9-high air flow blades with solid hydraulic bearing. 15 Built-in ARGB LEDs delivers a vibrant lighting effect that also supports RGB motherboards. Compatible with universal Intel and AMD socket.
thermaltakeusa.com thermaltakeusa.com

I'd replace it.
 
Oleg_NYC said:
While Thermaltake was inside the tower, the square thing was below it, but not inside the tower. I guess it's called a dust filter, but I am not completely sure I am using the right word to describe it. The filter collected a lot of dust, and I removed the dust from it.
Click to expand...
Can imagine how that looks. Air intake right below the PowerSupply. Air goes in, goes thru the PS and goes out agan. Airflow is nonexistant.
Not really sure from the description, but I've seen such, and a dust filter at that place wouldn't make sense otherwise.

Oleg_NYC said:
But since it was below the power supply, this thing doesn't participate in attempting to lower the CPU's temperature, right?
Click to expand...
You got it.
Those people building these cases usually don't get it.

Folks, all of you: why isn't there a heads-up imbecile-alert on the attack-screen when people get the idea that an air-mover needs "true RGB leds"??? Whatfor? How does that improve the airflow?

Anyway, it is difficult to cool a consumer tower for sustained compute, even when trying to do it right. Mine is 105 Watt (plus 20 disks), and I cannot run it full speed for prolonged time. I would love to insert a second CPU, but that would need a different case first, i.e. a real server case, Chenbro-SR209 or something.


Oleg_NYC said:
I was noticing that some of the cores still reached the temperature of 100 degrees during the compilation proces. Are you absolutely sure that when compiling something, CPUs are not supposed to reach the temperature of 100 degrees?... Are you absolutely sure these crashes happen because of overheating, and not because of something else?
Click to expand...
Nothing is sure unless one would have a look at Your actual gear. But it doesn't seem unlikely to me.
 
The Sahara cooler that is in my computer looks like this:

61MghKetxVL._SX522_.jpg


shkhln said:
The obvious test would be to disable powerd and set lower operating frequency level (dev.cpu.0.freq) to get lower temps.
Click to expand...
While compiling chromium using the MAKE_JOBS_NUMBER=7 setting, I killed powerd++, ran the command sysctl dev.cpu.0.freq=3700 and saw values such as these:

Code: 
dev.cpu.19.temperature: 49.0C
dev.cpu.17.temperature: 52.0C
dev.cpu.15.temperature: 51.0C
dev.cpu.13.temperature: 47.0C
dev.cpu.11.temperature: 52.0C
dev.cpu.9.temperature: 50.0C
dev.cpu.7.temperature: 50.0C
dev.cpu.5.temperature: 50.0C
dev.cpu.3.temperature: 45.0C
dev.cpu.1.temperature: 50.0C
dev.cpu.18.temperature: 49.0C
dev.cpu.16.temperature: 52.0C
dev.cpu.14.temperature: 51.0C
dev.cpu.12.temperature: 48.0C
dev.cpu.10.temperature: 52.0C
dev.cpu.8.temperature: 50.0C
dev.cpu.6.temperature: 50.0C
dev.cpu.4.temperature: 49.0C
dev.cpu.2.temperature: 45.0C
dev.cpu.0.temperature: 50.0C

But setting that number to 3701 with sysctl dev.cpu.0.freq=3701 caused a huge increase in temperature:

Code: 
dev.cpu.19.temperature: 88.0C
dev.cpu.17.temperature: 80.0C
dev.cpu.15.temperature: 85.0C
dev.cpu.13.temperature: 98.0C
dev.cpu.11.temperature: 100.0C
dev.cpu.9.temperature: 100.0C
dev.cpu.7.temperature: 94.0C
dev.cpu.5.temperature: 86.0C
dev.cpu.3.temperature: 73.0C
dev.cpu.1.temperature: 94.0C
dev.cpu.18.temperature: 88.0C
dev.cpu.16.temperature: 80.0C
dev.cpu.14.temperature: 85.0C
dev.cpu.12.temperature: 98.0C
dev.cpu.10.temperature: 100.0C
dev.cpu.8.temperature: 99.0C
dev.cpu.6.temperature: 95.0C
dev.cpu.4.temperature: 86.0C
dev.cpu.2.temperature: 73.0C
dev.cpu.0.temperature: 93.0C

Why would changing the number from 3700 to 3701 cause such a huge increase in temperature?
 
Oleg_NYC said:
Why would changing the number from 3700 to 3701 cause such a huge increase in temperature?
Click to expand...
My educated guess is that there's some kind of thermal control cutoff (at upper limit of 3700) programmed/implemented into that i9 itself. OC does have the tendency to heat up the CPU, which is rated at up to 3.7 GHz under normal operation. If you let it go above that (specify 3701), thermal throttling gets shut off because it becomes pointless to try and rein in the rising temps. That i9 can be allowed to run at up to 5.3 GHz (per pcpartpicker.com data), but at the expense of rising temps that are difficult to control. This is why OC enthusiasts tend to buy water-based coolers.
 
You might want to check whether the motherboard has any power limit, auto overclock or whatever settings. Although the cooler is not great, there is likely a second contributing factor.
 
richardtoohey2 said:
But the only thing that you know of that has changed is the Chromium version that you are trying to compile?
Click to expand...

This is what I find so confusing! Ccache was disabled; 20 make jobs were enabled in the settings, and chromium-123-0-6312-58 finished compiling less than 3 hours after I started the process. There was no crash. But one day later, when compiling version _1, there were crashes with only 4 make jobs enabled. It's possible that during all that time, I was unknowingly utilizing the Turbo Boost setting. I think I need to disable Turbo Boost in the UEFI.

[USER=54069 said:
shkhln said:
Any crashes?
Click to expand...
Click to expand...
No...
So, if the CPU's temperatures are pretty low with the 3.7 Ghz setting, that means I don't have an overheating problem. Every single time I noticed the temperature of 100 degrees, it was so hot because of the Turbo Boost setting... I am not 100 % sure it's true though. More testing is needed to verify this... I don't know if disabling Turbo Boost will prevent crashes from occurring in the future... But you also suggested that it's possible the motherboard has the auto overclock feature enabled. If it's indeed enabled, I need to disable it as soon as possible.
 
Sounds like OP might want to try having 20 make jobs enabled for the _1 version as well... just to keep things consistent from one version to next.
 
Oleg_NYC said:
Why would changing the number from 3700 to 3701 cause such a huge increase in temperature?
Click to expand...
It's a bit extreme here, but otherwise normal. The hotter the core gets, the more energy it draws (for the same compute). And the more energy it draws, the hotter it gets. In other words: the higher the performance/speed, the more unefficient (and hot) it all becomes.
That is one of the many crazynesses of this whole overclocker/gamer situation: they are willing to spend twice the energy to get maybe 10% more speed.

Then this chip of yours is a 'K' chip, meaning speed is not capped. It can go up a lot more than only 10%, So the energy that it pulls at uncapped max speed (that is the 3701) could be quite more than twice as much as at nominal speed (3700). And since the energy has no other place to go than into heat, we can measure this as temperature.
The idea with the 'K' chips is, not limiting the speed to confirmed working levels, but have it run as fast as ever it can go, no matter if it does occasionally crash. I don't know why one would want that, but the gamers want it and pay extra for it.

Let's conclude: your chip runs at a nominal speed of 3700MHz. And that is what you get when setting 3700. All the chips can go faster than their nominal speed. How much faster, that depends on the quality of the individual silicon wafer. So when you set the speed to one above nominal, that means: let the chip run as fast as it thinks it can.
For normal chips that is some 10-15% more (but costs a lot of extra energy). 'K' chips however, are selected from the best wafers, and they can go up a lot more. According to the specs, this one might go up to 5300 MHz.

I never managed to understand the benefit of this. Okay, instead of compiling chromium in six hours, it might compile in 4:45. Fine, but then using twice the amount of electric power in the process - what is that good for?
But gamers are willing to pay lots of money for these chips, these boards, these PS - so Intel and ASUS do offer them (and as a compensation people can then go to demonstrations to "save the planet").
 
Oleg_NYC said:
But one day later, when compiling version _1
Click to expand...
You might have been very close to the tipping point, and this new version was just enough to trigger the new behaviour. Or some code in the new version pushes some component just enough. Or the new version just happens to appear the same day that your PSU starts acting up, or RAM, or that thermal paste just decays enough to make a difference.

It does sound like it wouldn't hurt to get things a bit cooler and see if you can get the new version to build at all, even if it takes longer.
 
BTW, what comes to my mind: have you ever checked if the mainboard BIOS has some overclocking settings enabled which aren't even supposed to be stable?
Because, a normal cpu will just decline to run such settings, but this 'K' chip probably will.
 
PMc said:
BTW, what comes to my mind: have you ever checked if the mainboard BIOS has some overclocking settings enabled which aren't even supposed to be stable?
Because, a normal cpu will just decline to run such settings, but this 'K' chip probably will.
Click to expand...

I just finished compiling chromium-123.0.6312.86. It took me 13 hours to compile it. But in my settings, I only had 7 make jobs enabled and powerd++ was probably throttling the clock frequency of the CPU. During the compilation process, the only time I noticed a spike in temperature was when I did what shkhln had suggested I should do.
And now I am gonna reboot and take a look at the settings available in the UEFI. I will definitely disable Turbo Boost.
 
You must log in or register to reply here.
Back
Top