|
Problem description: Performance of my SATA SSDs and HDDs is way below normal and it seems to degrade to outright terrible the longer the computer is powered on. I know there's nothing wrong with the drives themselves because I've tested them in my USB3 dock and they run without issues there. It happens to all drives on SATA whether old or new. Another worrying thing is that the POST time has become unpredictable - sometimes seconds, sometimes entire minutes with a black screen before it gets to the standard BIOS screen showing RAM and connected devices. But the actual boot time after post is still only a few seconds. Check out these graphs. Left is after a cold boot, right is after 16 hours of idle time. See those dips? I can hear the absence of HDD activity during those dips, especially when HD Tune is running the access time test and the HDD is being very loud. All activity stops dead for about half a second. This doesn't happen with drives connected to the dock over USB3. This is what the benchmark looks like when a drive is connected through my USB3 dock - completely normal. When the very same drive was on a SATA port the result looked exactly like the others I showed earlier with the weird dips. The disk activity light seems to dim in sync with those dips on the graphs. Even when not in use it never ever stops blinking and it loops through a visually obvious pattern about once per second. I've never seen that before a few weeks ago. It started happening after I installed a Visual Studio update and restarted but I doubt it's related to that, more than likely a coincidence. There's nothing suspicous on my computer, no antivirus warnings, no malware scans showed anything even in an offline scan. All temperatures and voltages are well within the norm according to HWinfo64. Any suggestions? I suspect a dying motherboard, or at least the SATA controller. Failing that maybe the PSU has got a problem... It's all very strange. Attempted fixes: - Changing SATA ports around in lots of different combinations. No effect. - Removing all drives except the boot drive. No effect. - Reinstalling chipset drivers. No effect. - Fresh Windows install on spare SSD - seemed a little better, but power cycling has the same effect on my regular install as you can see above. I think it would degrade the same way if I left it idle, but I need to use my computer so... Todo list: - Get my hands on a spare PSU and see if things change. Recent changes: - There was a power cut a few days ago and it may have been a brownout. - The never-ending blinking pattern of the drive activity light started after I installed a Visual Studio 2015 update a few weeks ago. Surely a coincidence but maybe symptomatic of another problem. - Last hardware change was two more 8GB sticks of RAM. But that was about three months ago. I would've noticed if it directly caused an issue. Operating system: Windows 7 Pro x64 System specs: Mobo: Asus P8Z77-V LX CPU: i5 3570K @ 4.2GHz GPU: MSI GTX 1070 RAM: 32GB Corsair SillyName, 4x8GB DIMMs SSD: Samsung 850 Evo 1TB HDD1: WD Black 2TB HDD2: WD Black 640GB PSU: Seasonic S12II-520 Location: Europe I have Googled and read the FAQ: Yes
|
# ? Sep 22, 2016 20:57 |
|
|
# ? Apr 26, 2024 09:30 |
|
Run Crystal Disk Info and make sure that none of your drives show Caution. If that looks good, verify the system can pass at least one full pass of Memtest86+ without errors. If that is good, reset to stock CPU and RAM clockspeeds.
|
# ? Sep 23, 2016 03:50 |
|
Thanks, I gave your suggestions a try. All drives show "good" in CDM and it passed the Memtest86+ test. I tried resetting the OC to stock and nothing changed. Just to be sure I linpacked the crap out of it with the OC enabled, but it didn't show any instability then either. I also tried raising the core voltage slightly in the BIOS, no effect. So I don't think it's the OC. I also tried updating to the latest bios. No change. (E: actually this fixed the long POST times somehow) I tried a fresh Windows install again on a different drive (one that shows the performance problems in my current install) and it performed perfectly. I think it really is some obtuse software problem on my Windows install. I even set up all the drivers that are on my real install so it would be a fair shake. What the hell, could VS have really turded up my machine? Spatial fucked around with this message at 00:57 on Sep 26, 2016 |
# ? Sep 25, 2016 21:30 |
|
I decided to go on the prowl with Process Explorer in search of offenders. Turns out that WmiPrvSE.exe (WMI Provider Host) is doing periodic IO at about the same frequency as I'm seeing the dips. I had the bright idea of suspending the process. Guess what happens? The dips disappear and performance goes back to normal! I can literally toggle it on and off and see the dips disappear and reappear in real time on the HD Tune graph. I can't just kill a core Windows component without loving something up. Any pointers? Any idea what's going on here? e: Here are the graphs with the obvious dips where I've unsuspended the process: Spatial fucked around with this message at 23:23 on Sep 25, 2016 |
# ? Sep 25, 2016 22:57 |
|
Okay, I seem to have found a solution. Hopefully someone else will find this useful. The hidden symptom here was the Windows Management Interface service. Every second it would get simultaneous exclusive access to all drives and read a few kilobytes of data for no apparent reason. I saw three handles like " \Device\Ide\IAAStorageDevice-1" showing up in Process Explorer at regular intervals. The accesses became longer over time until they became the main source of IO contention and my system took a giant stuttering poo poo on itself. After way too long googling, I learned that WMI has a 'repository' which defines all the stuff it does. Something was configured incorrectly there presumably by Visual Studio. I reset it by halting the service, running "winmgmt /resetrepository" and then restarting it again. This resets WMI back to the state it has after a fresh install. Apparently doing this can have negative side effects on some installed programs, but it didn't on my machine. It just resumed working normally. Spatial fucked around with this message at 11:29 on Sep 26, 2016 |
# ? Sep 26, 2016 11:26 |