Failing desktop - repair or replace?

J

jamespetts-2

n00b
Joined
Apr 17, 2016
Messages
34
I have a self-built PC from late 2010 which has been upgraded over the years: it is an overclocked first generation i7 with Windows 7, 12GB of RAM, 2x 980Ti in SLI and various peripherals.

Over the last year or so, there have been increasing problems with the computer, mainly in the form of random crashes/BSOD errors. I replaced the original 850w PSU (which had visibly melted capacitors when I removed it) with a new 1200w PSU, but this seems to have made no difference. The blue screen errors continue to occur mostly at random, but they tend to happen more often when the computer is performing a computationally intensive task and in warm weather.

Having ruled out the PSU as a cause, I tried running the latest version of Memtest86 to see whether the RAM was the problem but, running it overnight for 6 passes on the default settings, no errors were detected. Similarly, running Windows' own memory checking software found no errors. Either, therefore, there is a problem with the RAM that this software does not detect, or the issue is with the motherboard/CPU.

Thoughts turn to starting again with a new computer - I can salvage much of the hardware (including the new PSU), but I am concerned about modern operating system choices. Windows 7 does not seem to be a sensible long-term option for a new PC, and running Windows 10 as my main operating seems to be insane given the sinister level of control over other people's computers that Microsoft have arrogated to themselves.

Linux has much going for it - I use it (Ubuntu 18.04) on my Skylake NUC that I use in work quite successfully, but there are a number of things that I will not be able to use with it, including using Lightroom (or perhaps Capture One Pro - I know that Darktable is a thing, but, having tried that, it pales in comparison to Lightroom in terms of develop tools), Visual Studio, certain games, and some fairly obscure software such as SONAR Artist (MIDI software), Traincontroller Gold (model railway control software) and possibly Premiere Elements (although I am not sure how good that Linux's video editing tools are these days). Also, I am not sure how well that Linux will work with my new multifunction printer/scanner with its two sided scanner, which I find very useful for scanning documents sent to me by post very quickly.

Some of the specialist software can possibly be run successfully in a Windows 10 virtual machine under Linux (I have tested Traincontroller with this, which seems successful so far, although this is not very computationally demanding), but photograph editing is potentially more of an issue, as Lightroom/Capture One Pro are very computationally intensive and it is painful to use them if they are other than very fast.

Ideally, it would be good to be able to continue to use my current computer for another year or two while Windows 7 remains supported by security updates and I have more chance to test using Windows 10 under a virtual machine on my Skylake NUC, by which time the upgrade options would have become a little better (I am most interested in the large number of cores being offered by the latest Ryzen CPUs, as some of the amateur open source game development work that I do on a game called Simutrans-Extended is both very computationally intensive and capable of a high degree of parallelisation).

I should be interested in anyone's thoughts on the various issues mentioned here, as it is difficult to know where to start with the virtualisation and fault diagnostic issues in particular.
 
jamespetts-2 said:
they tend to happen more often when the computer is performing a computationally intensive task and in warm weather.
Click to expand...
based on that, sounds like overheating. if you want a new system, transfer parts over and use your 7 key during win10 install for a free upgrade. just do it.
 
jamespetts-2 said:
running Windows 10 as my main operating seems to be insane given the sinister level of control over other people's computers that Microsoft have arrogated to themselves
Click to expand...

While in theory I agree with the thrust of your comment here and also share the opinion that Microsoft has gone too far with the telemetry and control over Windows updates and such that they exerted with Windows 10, it is still a perfectly valid operating system. You are running dual 980 Ti in SLI, so I can only presume you care about gaming performance, and like it or not Windows 10 is the premier gaming OS. It has better drivers and performance overall for gaming than any other OS, and is the only OS to support DX12.

I would vote for Windows 10 and just perform some of the steps necessary to take back a bit of control and disable some telemetry if it matters to you.

With that said, you might try simply backing off on the overclock to see if stability returns. Obviously you OC'd because you want the performance, but as your system has aged perhaps it's lost a step.
 
Thank you for your thoughts.

As to overheating, I tried experimentally running Two Point Hospital with Open Hardware Monitor running: it recorded a peak CPU temperature in core no. 1 of 79C (averaging circa 69C), and motherboard peak temperatures of between 47C and 65C. Presumably, however, the tendency to overheat would not increase with time? I do blow the dust out of the heatsinks whenever I open the case.

As to clocking back - I did try reverting the overclock a while ago, but the blue screens persisted, so I do not think that this is the issue alone.

Has anyone much experience of Windows 10 running in Virtualbox?
 
Check for failing / exploding / leaking capacitors on the motherboard. I had a PC that worked fine for many years and then a few years into it, the capacitors were leaking. I RMA'ed the mobo and issues went away.
 
How about the hard disk drive? They tend to fail after some years. Mayhaps run chkdsk to see
 
jamespetts-2 said:
Thank you for your thoughts.

As to overheating, I tried experimentally running Two Point Hospital with Open Hardware Monitor running: it recorded a peak CPU temperature in core no. 1 of 79C (averaging circa 69C), and motherboard peak temperatures of between 47C and 65C. Presumably, however, the tendency to overheat would not increase with time? I do blow the dust out of the heatsinks whenever I open the case.

As to clocking back - I did try reverting the overclock a while ago, but the blue screens persisted, so I do not think that this is the issue alone.

Has anyone much experience of Windows 10 running in Virtualbox?
Click to expand...
that temp should be fine and if the bsod happens without oc that's not it. id do the two suggestions above and id also test/replace the cmos battery. a dying one can cause all sorts of odd issues. maybe post the bsod.
 
79C cpu temp is a bit warm for just a game. Are you running the stock CPU cooler? Is it dusty? Have you ever replaced the TIM? What are your GPU temps in game? If you just want an excuse to upgrade, go for it. New hardware is always nice.

I am assuming you are on x58 with a i7 920? If you want to cheap out, you might be able to replace the cpu with a x5650 or similar for 6 core goodness and a clock speed bump.
 
Thank you for your replies - I will have to perform the visual inspection of motherboard capacitors as suggested when I have a moment - this is a good idea.

As to the storage devices, the main system drive is a newish (2017) SSD, so failure of this is unlikely; also, the symptoms are not consistent with storage failure.

As to the cooling, I am not using stock cooling: I do not think that this overclock would have been possible with stock cooling. I am using a huge Noctua CPU cooler with two very large fans, which is what I have used since the computer was first built. The CPU is an i7 950 overclocked to ~4.1ghz.

Incidentally, the reason for the 2x 980Ti in SLI was not because I am a hardcore gamer, but rather because I upgraded my monitor to a 4k unit in 2016, and at the time, the 980Ti was the only thing capable of producing useful framerates on such a high resolution. In retrospect, I wish that I had waited a month or two for the 1080 to be released, but I did not know that its release was so close when I bought the 980s. I do like 3d games from time to time (and I like to get decent framerates when I play them), but they are not the main focus of my computing activity.
 
What does it say about the BSOD in computer management event viewer? Perhaps that will give a clue to the cause of the crashes. You could also remove one of the 980ti and see if it still crashes.
 
Have you tried removing the HSF, cleaning the hell out of the old thermal compound, and using brand new stuff? I try to do it every 2-3 years just to be safe, as thermal compound loses its effectiveness over time (or so I've heard).
 
Yup, I would remove 1 of the GPU's at a time ... then redo the TIM on CPU...........
 
You must log in or register to reply here.
Back
Top