Windows 8.1 = disaster??
 

As some of you know, I've been battling stability issues on my Haswell machines. In the faint hope that perhaps there was an OS issue, I let one box upgrade from Windows 8 to Windows 8.1. Big mistake.

According to task manager, about 2/3 of one core is now dedicated to servicing System Interrupts. Iteration times for prime95 worker #1 went from 15ms to 43ms. Thanks, Microsoft.

I think my next step is to install Windows 7 over Windows 8.1. Hopefully I can do that without wiping the disk clean.

[QUOTE=Prime95;359871]
I think my next step is to install Windows 7 over Windows 8.1. Hopefully I can do that without wiping the disk clean.[/QUOTE]

Maybe. But don't count on it... :no:

My 8.1 box does not exhibit that behavior.
I'd let it run for a while/overnight and see if it settles down; maybe it is the OS re-indexing for the search function.

I haven't seen that in 8.1 either.

8.1 is miles better than 8, but I'm still up in the air over 7 vs. 8.1.

Rather strange. Are you familiar with the event viewer? Does such a thing exist in Windows 8?

If you have the time you may be able to figure out if anything weird has been happening. My 11-month old box hasn't gotten to this point yet, but older computers I have dealt with have needed a pass through the event viewer to see if anything was causing errors (a lot of programs that shouldn't exist were failing to run because they weren't meant for this computer, for example).

Have you run a disk cleanup? If you've been having a bunch of crashes, there's probably lots and lots of error logs or something the computer is trying to sort through or something. Run some general maintenance, leave it alone for a bit (maybe go back to stock settings) and see if things keep going badly.

Screw that "Disk Cleanup" use CCleaner and reboot :razz:
:smile:

It looks like I prematurely jumped to a conclusion. Sorry, MS. Windows 7 is suffering the same fate. I guess a device has started to spew continuous interrupts. Time to do some research to see how to pinpoint the culprit (or if some system service is causing it). I'll let it run overnight and start investigations tomorrow.

At least it wasn't a complete waste of time as I did want to try Windows 7 or Linux to see if Windows 8 is related to my instability.

The interrupt storm did not disappear overnight. However, I accidentally discovered (I'd forgotten to turn off Windows 7 power save) that after the machine goes to sleep and I wake it the interrupt storm has disappeared. Strange.

Does that give anyone any ideas as to either cause or cure?

Are there any details available on the type of interrupts?
I have been having an intermittent and very rare issue with IRQ 29, Hex 1D. But, ever since I turned off Prime95 throttling nine months ago I have not had this problem.

Long shot, but is throttling below 100% on your Prime95?

Does the event viewer give you any idea as to which program or service is causing the headaches? You could try running through your task manager and start killing off tasks one by one until something changes, depending on what the frequency is.

Sleep has never been flawless but for a problem to disappear AFTER putting the computer to sleep...

I don't know what sleep physically does on a hardware or software level. It does kill the hard drive and CPU (and GPU) while keeping the memory fed to maintain it.

Do the issues come back after a reboot? Do they ever come back after a sleep (hours later)?

[QUOTE=Prime95;359932]The interrupt storm did not disappear overnight. However, I accidentally discovered (I'd forgotten to turn off Windows 7 power save) that after the machine goes to sleep and I wake it the interrupt storm has disappeared. Strange.

Does that give anyone any ideas as to either cause or cure?[/QUOTE]

SysInternals Process Explorer [I]might[/I] give some indications of what is going on. It does display CPU time taken up by "Hardware Interrupts and DCP's." Even if your runaway process is software, Process Explorer gives a more comprehensive view of who is doing what, and consuming memory and CPU time.