To rule out the disk, this is where I would typically boot from a Linux live DVD (I usually use Fedora for this, but most any pre-built live distro should be fine), and use the Gnome Disks utility to launch a SMART extended test on the drive. If it's a new drive, I'll do a (destructive) badblocks test first, and then do the SMART extended test. Regardless, this is best to do outside of Windows so you that can isolate disk access, since Windows loves writing to the drive constantly. If you used the bootable distro of SeaTools, then you probably covered this anyway. But yeah, after all of that, the main SMART attributes to watch out for are pending reallocations, reallocated sectors, and read/write/CRC errors. Pending or reallocated sectors mean your drive is dying, while the others indicate either a controller issue or a loose cable. Also bear in mind that if you have a Seagate drive, they fuck with the bitmask for the raw numbers, so a non-zero number in the reallocated sectors field (as shown outside of SeaTools, e.g., in Gnome Disks tool) doesn't necessarily mean that there were reallocations -- only if the number is extraordinarily high, because the count starts somewhere in the upper bit range, while the least significant bits are some kind of usage counter. Anyway, ANY reallocations in SeaTools = dying drive that needs to be backed up and replaced ASAP, even if SeaTools doesn't say that the drive is failing (and assuming SeaTools isn't showing the raw value, e.g., something ridiculous in the 10k range).
MemTest86+ is my go-to for RAM testing, and seems to be somewhat more actively maintained than MemTest86, although they are based on the same original code. Either way, how long did you let it run? I like to do a single pass with all tests, and then beat the snot out of it for several passes with test #7 if I suspect a bad DIMM. Are you using XMP profiles in your BIOS?
What GPU do you have? 3DMark may be worth using as a stress test to see if it can duplicate the issue -- that will point to either drivers or cooling. Furmark may also be an option. If it's happening while using 3D modeling software, it could be an OpenGL issue, in which case you're probably boned, since you're stuck with the manufacturer-specific OpenGL drivers.
Finally, you may need to stress test the CPU/RAM interface -- a Prime95 stress test in Large FFTs mode will either find an issue, or cook your PC (especially if you're on a stock cooler or a laptop), but it's something to consider.
I let it do its thing overnight- the version I used was the free version of the non-open source one though so it didn't look like I could tweak parameters much. I'll try again with MemTest86+
Nah, if it survived an overnight MemTest86 run, that's basically the same tool -- your RAM is OK in isolation. Consider the Prime95 Large FFTs test, though -- could be something in the CPU/RAM interface. At the moment, I'm leaning towards "shitty OEM OpenGL driver," but we should at least rule out hardware and cooling first.
To rule out the disk, this is where I would typically boot from a Linux live DVD (I usually use Fedora for this, but most any pre-built live distro should be fine), and use the Gnome Disks utility to launch a SMART extended test on the drive. If it's a new drive, I'll do a (destructive) badblocks test first, and then do the SMART extended test. Regardless, this is best to do outside of Windows so you that can isolate disk access, since Windows loves writing to the drive constantly. If you used the bootable distro of SeaTools, then you probably covered this anyway. But yeah, after all of that, the main SMART attributes to watch out for are pending reallocations, reallocated sectors, and read/write/CRC errors. Pending or reallocated sectors mean your drive is dying, while the others indicate either a controller issue or a loose cable. Also bear in mind that if you have a Seagate drive, they fuck with the bitmask for the raw numbers, so a non-zero number in the reallocated sectors field (as shown outside of SeaTools, e.g., in Gnome Disks tool) doesn't necessarily mean that there were reallocations -- only if the number is extraordinarily high, because the count starts somewhere in the upper bit range, while the least significant bits are some kind of usage counter. Anyway, ANY reallocations in SeaTools = dying drive that needs to be backed up and replaced ASAP, even if SeaTools doesn't say that the drive is failing (and assuming SeaTools isn't showing the raw value, e.g., something ridiculous in the 10k range).
MemTest86+ is my go-to for RAM testing, and seems to be somewhat more actively maintained than MemTest86, although they are based on the same original code. Either way, how long did you let it run? I like to do a single pass with all tests, and then beat the snot out of it for several passes with test #7 if I suspect a bad DIMM. Are you using XMP profiles in your BIOS?
What GPU do you have? 3DMark may be worth using as a stress test to see if it can duplicate the issue -- that will point to either drivers or cooling. Furmark may also be an option. If it's happening while using 3D modeling software, it could be an OpenGL issue, in which case you're probably boned, since you're stuck with the manufacturer-specific OpenGL drivers.
Finally, you may need to stress test the CPU/RAM interface -- a Prime95 stress test in Large FFTs mode will either find an issue, or cook your PC (especially if you're on a stock cooler or a laptop), but it's something to consider.
I let it do its thing overnight- the version I used was the free version of the non-open source one though so it didn't look like I could tweak parameters much. I'll try again with MemTest86+
Nah, if it survived an overnight MemTest86 run, that's basically the same tool -- your RAM is OK in isolation. Consider the Prime95 Large FFTs test, though -- could be something in the CPU/RAM interface. At the moment, I'm leaning towards "shitty OEM OpenGL driver," but we should at least rule out hardware and cooling first.