Intel Revises PCIe Cooling Driver To Reduce Link Speed When Running Too Hot

[binarybanana]

What's this? Bus overheat? Bridge? Controller chip in device?

IIRC it's mostly the pcie controllers pulling a lot of power. It takes a surprising amount of energy to push data around quickly, especially when you need to do computing for signal processing. PCIe is flexible but power hungry, unlike accessing RAM.

if I let them, the memtest86+ runs would probably have climbed to about 80°C and then throttled to maintain temperature equilibrium.

Sounds like throttling works as expected? 80C is lower than my laptop (tops out at 95C, a bit too much).

[coder]

Contemporary transistors are increasingly plagued by thermal constraints and performance bottlenecks. To address these challenges, it's imperative to innovate and develop new technologies that can be produced on a massive scale, enhancing or potentially supplanting the existing models.

You mean like phase-change immersion cooling?

https://www.servethehome.com/2-phase-immersion-cooling-halted-over-multi-billion-dollar-health-hazard-lawsuits/

Whoops!

[coder]

I am not sure about this feature. We might see similar errors and complaints as on windows, where users report their GPU PCIe 16x to be slower than they should be according to the official specification. Reason: The PCIe Bus clocked down to lower speeds due to power saving modes.

I guess you've never heard of PCIe ASPM. It's not new!

https://en.wikipedia.org/wiki/Active_State_Power_Management

[ssokolow]

Sounds like throttling works as expected? 80C is lower than my laptop (tops out at 95C, a bit too much).

1. I said probably (I never let it get that high. I'm just guessing based on the information in the UEFI, where Vorke exposed every tunable under the sun including programmer-only "constants" for devices that are supported by the chipset but not present.)
2. I much prefer its behaviour under an OS like Debian that can run thermald to flip the bit to put it in performance mode and have the fairly quiet fan cap it at around 52°C, even while allowing it to run perpetually at full boost clock.
3. On some occasions, I've had it get stuck at full fan in an OS without DPTF support instead. Why is it so hard to just have the firmware default to a balanced thermal profile that doesn't require OS assistance?

[binarybanana]

1. ]On some occasions, I've had it get stuck at full fan in an OS without DPTF support instead. Why is it so hard to just have the firmware default to a balanced thermal profile that doesn't require OS assistance?

I had a similar problem once. Not sure what helped but I think I went as far as updating my firmware. Very annoying.

As far as your other points I think it's fair to say that the throttling (assumingly) works at least in some capacity, just that the parameters aren't to your liking. If these things were easily configurable at the firmware level that would be nice indeed.

[ssokolow]

I had a similar problem once. Not sure what helped but I think I went as far as updating my firmware. Very annoying.

As far as your other points I think it's fair to say that the throttling (assumingly) works at least in some capacity, just that the parameters aren't to your liking. If these things were easily configurable at the firmware level that would be nice indeed.

I have configurable UEFI parameters coming out my ears with the Vorke mini-PCs. The problem is that it shouldn't be so hard to get a sane "do the sensible thing until the OS driver comes online and suppresses that" behaviour. Hell, in one case, I had to upgrade off an LTS distro's stock kernel because it was too old to provide the DPTF interfaces thermald needed, so it was stuck on "silent mode" (i.e. throttle instead of turning on the fan).

[M.Bahr]

I guess you've never heard of PCIe ASPM. It's not new!

https://en.wikipedia.org/wiki/Active_State_Power_Management

My objection is not about the ability of regulation per specification but about the quality of the driver for that.

[binarybanana]

I have configurable UEFI parameters coming out my ears with the Vorke mini-PCs. The problem is that it shouldn't be so hard to get a sane "do the sensible thing until the OS driver comes online and suppresses that" behaviour. Hell, in one case, I had to upgrade off an LTS distro's stock kernel because it was too old to provide the DPTF interfaces thermald needed, so it was stuck on "silent mode" (i.e. throttle instead of turning on the fan).

I'm just saying a laptop that throttles at 80C by default is not really insensibly configured. Mine easily hits 95C if I fully load both CPU and GPU, but it's a gaming laptop meant for performance above all else and I don't use thermald or anything like that. However, if a laptop is stuck in base clocks without a special driver that is terrible indeed.

[waxhead]

Well, no, because the thing heating up the might not be the disk or GPU itself. Managing the thermals of a tightly-packed device is difficult and it takes coordinated effort.

I disagree. But let me clarify. Firmware is also software so firmware should override to prevent hardware failure and if not , the hardware should protect itself from "general use" and not melt down just because some piece of software is not running (and some other software IS running).