(Sorry for the wall of text, but people seem equally critical of missing details)
After about 2 years without trouble, I've been having issues recently with system instability, BSODs, random application crashes, graphical glitches etc. After excluding a few other possibilities I ran MemTest86+ and got a whole screen of red lines, hundreds of thousands of errors on multiple tests. So I set out trying to work out if I could get by on only one stick for a month or so until I have the money for a new kit. This is where it gets bad. I swap the sticks and the errors all stay at exactly the same addresses - a cluster of addresses somewhere around 6 GB (in a 2x4 GB kit). I swap them round a couple more times, but the bad addresses stay the same. After a little reading around, I decide it's probably a processor/motherboard issue. I figure I can't hit an error at a ~6 GB address if I only install one 4 GB DIMM, so that was the temporary fix. And it worked - for about two months. Now I have a cluster of bad addresses around 1.8 GB (whichever DIMM is installed), and system instability as bad as before.
This is probably the point to mention specs:
- Processor: Intel Core i5 2500K, currently clocked at stock 3.3 GHz.
- Motherboard: Asus Maximus IV Gene-Z / Gen 3, with latest stable BIOS
- RAM: Corsair Vengeance 2x4 GB DDR3-1866MHz kit, model on MOBO QVL list, timings as per manual, installed in a recommended slot configuration (A2 & B2, size of CPU cooler prevents installation in A1 & B1)
- Graphics card: NVIDIA GeForce GTX 680 (Asus GTX680-2GD5)
- PSU: XFX PRO850W Core Edition
- OS: Windows 7 Professional, all updates, all latest stable drivers
Also, before running MemTest, I tried a fresh install of Windows 7 and all relevant drivers on multiple hard drives to exclude either OS or hard drive. It's worth noting that the processor/motherboard are on the LGA 1155 socket, meaning that if I try to replace only one of them, I'm a little limited in my choices. Worse, if it doesn't fix the problem and I have to replace the other as well it looks like I'll actually end up spending more to just tread water than it would cost me to upgrade both to something newer on LGA 1150. It's probably just best to replace both or neither.
Normally I would use the desktop for all my home computer use, but with this recent trouble I've moved just about everything important over to my laptop. The laptop can't play graphically demanding games as well as the desktop, but aside from this my desktop isn't being trusted with anything vital anymore. I don't need water-tight system stability, and occasional crashes are acceptable so long as I have time to enjoy a couple of hours of gaming before they occur. I'm essentially looking for a stop-gap that will keep some gaming alive until I can justify spending out on a new motherboard/processor pair, probably around Christmas (I know, a long time for a temporary fix).
Ideally I'm looking for:
- any way to troubleshoot and positively identify the location of the error: RAM, motherboard, processor
- a Windows equivalent of BadRAM - if I can tell it to stop using ~1.8 GB and ~6 GB addresses this will solve my problem well enough for the time being (don't suggest /maxmem or anything similar as I'd need to set it below a crippling 1.8 GB to be effective)
- something that can harmlessly reserve the bad address ranges, assuming the computer makes it through start up (not as safe as the previous option, but I'll risk it on a gaming box)
- any strategy for replacing either or both of the motherboard and processor that minimises the upfront cost without going backwards in spec. (least favoured / most realistic option)
- any simple potential fixes, like reseating the processor? I haven't tried it yet as I don't have any thermal paste on hand, but if anyone knows of this fixing the issue I'll order some and give it a try. I once successfully revived a laptop by baking the graphics card in the oven - so wild suggestions, supported by experience, are welcome.
I realise I'm probably asking for miracles. Based on only my own experience the right move is to replace the processor (and it's on-board memory controller that is most likely at fault). I'm hoping that someone else knows something relevant that I don't.