Defective Heat Sinks Causing Garbage Gaming | Hacker News Reader

Defective Heat Sinks Causing Garbage Gaming

randomascii.wordpress.com

84 pointsblackhole23 comments

>"When I told the customer of my suspicions they said that they had recently replaced their processor heat sink. When they put the old heat sink back on their performance problems went away."

A person who buys and installs their own heatsink, but can't troubleshoot a temperature-driven performance problem impresses me as an odd combination.

I expect providing support to members of the "enthusiast" PC market is an interesting challenge.

There are certainly more online guides to the former than the latter. Also, heatsinks come with instructions in the box. But yeah, I'll bet the "enthusiast" label applies to a greater diversity of people than one would immediately assume....

Everyone has to learn sometimes, and learning by trying is a great way (also, sometimes a costly way...), it's not that odd a combination I wouldn't suspect.

Far too many people buy their own heat sinks because that is what guides or other people tell them they need to have for best performance. There are those who buy it so they are part of the clique.

Its not like manufacturer applied heat sinks are fully reliable. I have had more than one go bad in my day, even better my Samsung DLP was the worst of them because its heat sink glue failure was on visible on the screen as white dots.

I'd actually be capable of buying and installing my own heat sink (I have never bothered doing so; I don't really do hardware mods), but I do not know how to troubleshoot a temperature-driven performance problem.

I mean, it's just thermal paste on a fan, right? That's not exactly hard, after a fairly low threshold of know-how.

I think that you are overthinking it, this sort of troubleshooting doesn't require all that much domain specific knowledge (certainly the author of the article came at it with more than was strictly necessary).

If you install new fancy fuel injection nozzles in your car, and suddenly your car is shit, then swap them back out and see if the problem goes away. After you determine that it does, you can start investigating if the installation was botched or if they are broken, or whatever. You don't need to know anything about cars to do that, it's just generic troubleshooting.

>"In all cases the customer had said that the problem must be with our game because only our game was hitting these mysterious slowdowns."

This probably threw them off.

Cleaning the CPU heat sink and fan from dust (and in my case, cat hair) can sometimes give a surprising performance boost. Especially on cheap laptops that are often sold with inadequate cooling.

CPU thermal throttling can have a huge effect on performance.

Pulling the gaming boxen over to the garage & blasting them out with the air compressor hose is part of spring cleaning here.

Do you run into any trouble with dust/static doing it that way vs canned duster?

Not dust or static, but air compressors can build up water in the tanks and lines, if you don't drain the tank like you're supposed to.

If doing this (or using canned air) it's a good idea to use a small stick or bit of plastic to jam any fans to stop them spinning too hard and ruining their bearings or throwing a blade, both of which I've had happen. Laptop ones seem particularly susceptible, probably down to size and typically higher usage temps.

Or acting as a generator and throwing an over voltage into a circuit that isn't protected from it. (Also, don't push your robot with the motors wired unless you know you protected the motor driver circuitry.)

I had never considered this possibility -- either in my laptop's fans nor in my Arduino robot. But that's a great pro-tip I'll keep in mind. (It may be why my Arduino stops working normally after a couple hours and takes a couple days to straighten out)

Frustratingly, cheap laptops are very often constructed in such a way that properly cleaning the CPU fan requires completely disassembling the device. I hate taking apart laptops because the ribbon connectors are often very stiff and very fragile.

I was once on the verge of buying a cheap dell laptop only to discover that replacing the HD required first removing the mainboard (!)

That's a luxury if your a MacBook Air owner.

Its common in hardware enthusiast forums and discussion now when there is a performance problem for there to be 2 initial questions. The first is regards to clock speed and the second is temperature. Its really common for clock throttling to be the underlying cause of performance problems on today's systems and temperature is also the number one cause.

But the second major cause, representing about 30% of cases is software/firmware bugs in the drivers for the graphics card or the Windows scheduler doing odd things for the CPU under a particular workload. As more of the performance of components becomes dependent on boosting and clock increasing in order to save power the more we are seeing inconsistent behaviour and problems with the implementations.

> Thermal throttling is extremely difficult to detect in xperf traces. It is done automatically by the CPU or the motherboard, and the operating system (OS) doesn’t realize that it is happening. I use the xperf toolset for these investigations but its CPU frequency graphs show the CPU running at full throttle and its power management events say that all is well, and yet…

At least with a Sandy Bridge CPU on Linux you get power management events which show up in dmesg and the reduced frequency is correctly reported. Definitely a case of the drivers/monitoring software not looking for the right things rather than the hardware not exposing this information.

> I hope that some day the ETW code in Windows that provides the CPU frequency will be fixed to detect thermal throttling – and a temperature provider would also be nice. Chapter 14 in the Intel Software Developer’s Manual, Volume 3A:, System Programming Guide, Part 1 would be a good starting point…

He mentioned that in the article ...

This should be made more visible by OS vendors.

Windows popping up one of those alert baloons would be very helpful.

I recently discovered 'turbostat' for reading actual CPU frequency on Linux: http://lxr.free-electrons.com/source/tools/power/x86/turbost...

Bruce's direct testing approach has a certain elegance, but instead of timing a loop 'turbostat' queries the CPU's performance registers. So in addition to being handy if you are running Linux, it could serve as a guide for developers of other systems trying to figure out which MSR's are needed and how to interpret them.

    cat /proc/cpuinfo

This would be simpler, but doesn't work if you are running a non-standard clock. /proc/cpuinfo gives you the nominal frequency, but turbostat gives you the actual. In the example below, the 3.79 GHz reported by turbostat is the actual speed and the 2.527 reported by /proc/cpuinfo is not. I think thermal throttling is another case where turbostat is correct but the kernel info is not.

  nate@centos:~$ cat /proc/cpuinfo | grep -i mhz
  cpu MHz		: 1596.000
  cpu MHz		: 1596.000
  cpu MHz		: 1596.000
  cpu MHz		: 1596.000
  cpu MHz		: 2527.000
  cpu MHz		: 1596.000
  cpu MHz		: 1596.000
  cpu MHz		: 1596.000

  cor CPU    %c0  GHz  TSC    %c1    %c3    %c6   %pc3   %pc6
            2.82 3.75 3.79   2.92   0.57  93.69   0.00   0.00
    0   0   0.29 2.76 3.79   0.26   1.24  98.21   0.00   0.00
    0   7   0.11 3.02 3.79   0.45
    1   3   0.14 2.45 3.79   0.14   0.56  99.16
    1   6   0.03 2.79 3.79   0.25
    2   1   0.06 3.07 3.79   0.12   0.26  99.56
    2   5   0.05 2.85 3.79   0.13
    3   2   0.05 2.81 3.79  21.88   0.23  77.84
    3   4  21.83 3.78 3.79   0.11

Reply on news.ycombinator.com