Question Sanity check before I scrap this old PC for parts please!

[mikeymikec]

Spec:

Phenom II X2 550
ASUS M4A78T-E (latest BIOS)
RAM: 2x2GB DDR3-1333 (tried a spare set of known good modules too)
SSD: Tried two (original: Samsung 870 EVO, Samsung 850 PRO)
PSU: Tried two (original: Corsair VX 450, Corsair CX430)
GFX: Tried two (original: GeForce 8400GS, plus onboard graphics)
OS: Win10, multiple clean installs
switched SATA cable going to SSD

This PC belonged to a customer and I built it in 2009. It ran XP flawlessly on a hard drive to my knowledge, then eventually I convinced the customer to upgrade to Win10 along with a memory upgrade and SSD.

I say 'to my knowledge' because the customer isn't a terribly reliable witness IMO, for example they said nothing while the computer was apparently crashing on a daily basis on Win10. Anyway, I think it can be fairly said that ever since the Win10 upgrade this PC has never been right: Multiple BSOD codes such as NTFS_FILE_SYSTEM, MEMORY_MANAGEMENT, IRQL_NOT_LESS_OR_EQUAL, system freezes, etc. Having said that, I had maintained their computer on XP over the years and I would have noticed at some point if it liked to crash a lot.

I've posted about it before:

I can read an NTFS file system on an SSD with Linux but not Windows

I'm kinda stumped for how to proceed with this one. Windows 10 clean-installed on an SSD (Samsung 860 EVO) ran into problems and automatic repair doesn't fix the problem. Where it really gets interesting is that: 1 - booting from Windows setup media results in one CPU core being saturated...

forums.anandtech.com

TLDR on that thread: event log entries included IO retries to disk which strongly suggested a faulty SSD so I talked to Samsung. They said nah, disable NCQ and you'll be good. That at least got rid of the disk errors from the time(s) I had it in for testing, and what seemed to improve stability further was switching from AHCI to IDE (this happened accidentally, the BIOS resetted itself while the customer had it), and I was at a point with it whereby the BSODs had seemingly stopped, no more freezes, but the Windows install was mangled from all the prior crashes and I couldn't get it working properly (e.g. Windows 'apps' like Calculator wouldn't start, and it couldn't do a Windows feature upgrade without throwing an install error). I took it in again with the plan to do a clean install of Windows, but while I had it in, it froze a couple of times (and one was outside of the Windows install, I think it froze during the first stage of setup before partitioning). I convinced the customer to replace it.

A bit of extra side info: I built several PCs based on this board, I know that it should be able to handle Win10 + SSD + AHCI. They weren't picky about drivers or anything like that either.

So I've started playing with it again in the hope that I'd get to the bottom of what's wrong with it, but as you can see from the spec I've tried switching out a lot of hardware in the hope of narrowing down the issue. It's had multiple runs of Prime95 without issue (max heat option), multiple runs of memtest86 4.3 and 5.01, and the current symptoms are much like what they were, mainly freezes and weird BSODs that point at drivers (SYSTEM_SERVICE_EXCEPTION, DPC_WATCHDOG_VIOLATION) despite often none being installed; the most recent BSOD (the DPC one) happened today after a clean Win10 install of 1607 without an Internet connection (so no VGA drivers yet, as vanilla an install driver-wise as can be) during an upgrade to 1803 via USB. Before that crash, I gave the PC a good vacuum-clean. The CPU HSF wasn't in a terrible state, but Prime95 would get the CPU going near 70C, so why not give it a good clean.

Personally I think it's time to scrap certainly the board, probably the CPU as well as I can't ever imagine selling a slightly better dual-core to a customer with an Athlon II X2, what's the point in that. I had hoped to get it stable so I could test out my spare 960T processor from another similar-era PC that's no longer stable.

 

[Insert_Nickname]

This PC belonged to a customer and I built it in 2009. It ran XP flawlessly on a hard drive to my knowledge, then eventually I convinced the customer to upgrade to Win10 along with a memory upgrade and SSD.

Are you using the original DIMMs along side? More ranks automatically lower how much frequency you can run them at. Might try at something really slow like DDR3-800 or thereabouts. Old memory can occasionally get flaky. But since you've used a known good set, that seems unlikely to be the culprit.

Is it stable if you run XP on it currently?

So I've started playing with it again in the hope that I'd get to the bottom of what's wrong with it, but as you can see from the spec I've tried switching out a lot of hardware in the hope of narrowing down the issue. It's had multiple runs of Prime95 without issue (max heat option), multiple runs of memtest86 4.3 and 5.01, and the current symptoms are much like what they were, mainly freezes and weird BSODs that point at drivers (SYSTEM_SERVICE_EXCEPTION, DPC_WATCHDOG_VIOLATION) despite often none being installed; the most recent BSOD (the DPC one) happened today after a clean Win10 install of 1607 without an Internet connection (so no VGA drivers yet, as vanilla an install driver-wise as can be) during an upgrade to 1803 via USB. Before that crash, I gave the PC a good vacuum-clean. The CPU HSF wasn't in a terrible state, but Prime95 would get the CPU going near 70C, so why not give it a good clean.

I'd say the board is going south. I had something similar happening with an AM3 board. It first got unstable. Then just flat-out refused to work at all suddenly. It happens.

 

[mikeymikec]

@Insert_Nickname No, I switched the modules, one pair for another.

XP - the original HDD is still inside the machine but completely disconnected; I considered firing it up and looking for any evidence of historical instability, but really, what's the point. I can't sell a machine that can't handle an SSD or an up-to-date version of Windows, and it's of little use to me. I am curious though

 

[VirtualLarry]

If the cause was flaky SATA ports, a dual-port ASMedia-chipset SATA6G card should work.

What specific mobo and specific chipset board was this?

 

[VirtualLarry]

GFX: Tried two (original: GeForce 8400GS, plus onboard graphics)

I don't know how problematic an 8400GS would be for Win10/Win11, but it my mind, that doesn't sound so great. I know that with an ASRock 780G mobo that I had used for a friend's build that lasted him 10 years or so, the onboard HD3200 or whatever it was was definitely problematic in Win7, nevermind Win10 or newer.

I would try a GT 710 or GT 1030, and an ASMedia SATA card, if you really wanted to save the rig.

 

[mikeymikec]

@VirtualLarry

Specs are in the OP, chipsets 790GX, SB750. I know the 8400GS is ok with Win10, again quite a few PCs I built had that card in. The onboard graphics is HD3300.

Admittedly I didn't try a different SATA port, but I would have thought that if the stability issue was purely SATA related then so would the nature of the errors. I have a SATA card I bought recently (ebay) but I don't know if it 100% works as I've never used it before.

 

[Ajay]

Personally I think it's time to scrap certainly the board, probably the CPU as well as I can't ever imagine selling a slightly better dual-core to a customer with an Athlon II X2, what's the point in that. I had hoped to get it stable so I could test out my spare 960T processor from another similar-era PC that's no longer stable.

Yes, this. I've had old systems going down the drain like this. I do all the basic tests/swaps like you have. If they are still messed up, they go into the bin. Good luck!

 

[bigboxes]

Was this a clean install of Win10?

 

[VirtualLarry]

Were you running HWMonitor? I've not been running it for the last week, and interestingly, my main rig hasn't crashed, even when mining on all threads of the CPU.

 

[mikeymikec]

Was this a clean install of Win10?

several clean installs of WIn10.

 

[mikeymikec]

Were you running HWMonitor? I've not been running it for the last week, and interestingly, my main rig hasn't crashed, even when mining on all threads of the CPU.

Yeah, part of the reason why I vacuum-cleaned it was that the CPU eventually topped out at 70C which I understand is the max for Ph2, but it carried on without errors.

- edit - I've just re-read your question, and no I wasn't running HWM the whole time, just a quick check here and there. There have definitely been Windows sessions that crashed without HWM being run during that session.

 

[mikeymikec]

I'm probably speaking too soon, but this PC's stability seems to have improved since swapping the 8400GS for a spare GT710 and changing the SATA port to a hitherto unused port.

The test routine I'm using atm is clean installing Win10 1607, up'ing to 1803 via USB, then 1903 via Windows Update, then manually installing the latest feature update via the update assistant. I'm purposefully doing it this way for maximum IO work. On one occasion it got through the whole set of updates without borking and then borked soon after, but every other time it has borked during the test. Since the latest hardware changes (see start of this post), it finished one test cycle successfully and I'm at the final feature update atm for one full test. I guess I'll run this test another two times before declaring some kind of success, and then comes the fun bit of walking back my last two changes one at a time to figure out which is the culprit.

 

[mikeymikec]

I tried a different SATA port for a couple of passes and it didn't seem to be that, so I went back to the first SATA port and switched graphics cards back to the 8400GS. The first BSOD I could have let slide (Win10 evidently uses an older driver for the 8400GS, it installed the driver then wanted to restart and I hadn't let it yet, it crashed in the 'waiting to restart' session), but then it BSOD'd again (UNEXPECTED_KERNEL_MODE_TRAP).

Good call, @VirtualLarry , I had ruled out graphics after the onboard graphics did the same thing. Now I get to test out my Ph2 960T!

 

[BonzaiDuck]

Is Phenom II a decade-old AMD processor?

Maybe somebody can get something out of the machine if it runs Win 10.

I've come to a crossroads with my brother who "lives" upstairs. I've kept him in modestly-build computers and hand-me-downs for years. I put Win 10 on the Ivy Bridge system I built for him -- 12GB RAM, but I didn't migrate his programs from Win 7. I thought HE would take care of that. So he's been bitching for at least a year that he's not using Win 7, and why wouldn't I have set up his computer so he could "learn Win10"? Well, I did! He would only have needed to migrate his Outlook *.PST file and his internet Favorite list from the appropriate Windows subdirectory.

So I was planning ot get him a new laptop, but what I find as acceptable are $900 HP, ASUS TUF and similar models. And we're short on money for January, when such a purchase must be paid off. I'm weighing the new laptop against giving him one with a Core-2-Duo Centrino Gateway Executive laptop that was released in 2007, purchase by me in 2014 and upgraded to Win10, 1TB 2.5" spinner (lower energy use than SSD), and 8GB (maximum) RAM. It's a slug when it's installing windows updates, and it's not the wireless connection -- it's the installations and updates that cause the processor to reach 100% and stay there for a half-hour. I had put the fastest wireless NIC in the laptop that I could get for it, offering 150 mbs.

At some point, one needs to break from old technology. But I'll give Bro the Gateway for an interim period, and see if I can't find him a newer one.

I'm just worn out with this sh-sh-stuff, taking care of Moms and Bro, both elderly and disabled in one or another respect. I feel disabled myself!

 

[igor_kavinski]

Now I get to test out my Ph2 960T!

How did it go?

 

[igor_kavinski]

So I was planning ot get him a new laptop, but what I find as acceptable are $900 HP, ASUS TUF and similar models.

My vote is Asus TUF.

 

[mikeymikec]

How did it go?

So far so good!

The board the 960T came from also liked to throw out various BSODs even when sitting idle in Windows, I basically took all the 'known-good' hardware from the build in the OP and transplanted to the case with the 960T and its suspect board and it kept crashing, so I took the 960T and the 'known-good' hardware back in to the OP setup and it's been going fine, so after some testing I swapped the PSU for a beefier one, swapped the GT710 for a 760, even unlocked all the 960T cores (which it has run with most of its life), all ok except for this small detail:

Question - Handbrake, nvidia encoder not working, help?

Spec: AMD Phenom II X4 960T with 6 cores activated, stock speed ASUS M4A78T-E, latest BIOS 4GB RAM 128GB SSD GeForce 760 Windows 10, latest version I think Handbrake, latest version I've used this graphics card in another system for most of my DVD ripping and I didn't have to do anything...

forums.anandtech.com

 

[mikeymikec]

@VirtualLarry

About the HD 3300 - I've got another computer I built with the same board in at the moment (SSD upgrade job), and the customer has been using on-board graphics on Win10 for ages without any display issues reported in Windows. One interesting difference though is that I haven't seen this customer in such a long time that the graphics setting in the BIOS was still set to SidePort only, and I wonder if that's the only way that this graphics chip will still work properly on Win10.

Background:
When I first started building with this board (2009), I figured that 128MB video RAM (which is the total SidePort memory soldered on to the board) should be enough for the foreseeable future so I set PCs to SidePort only, especially because I wanted to reduce the effect of integrated graphics on the main system RAM's bandwidth (this was the first time I started trusting that integrated graphics didn't completely suck). In later years I noticed that graphics usage was steadily creeping up so I re-enabled the default setting of UMA+SidePort.

One other thing I'm almost completely certain of now is that this AMD chipset does not get on with Samsung EVO drives, whereas the PRO drives (no longer available unfortunately - have worked fine) - disabling NCQ with EVO drives stops the CRC errors from building up.

 

[sdifox]

It's ancient. Let it go.