Retro's PC upgrade

[Retro]

Sometimes it's just not your day and today turns out to be that day for me: my system SSD has just died without warning. It's a Samsung Pro 256GB that I've had for just over a decade and has been flawless in all that time. It was at 90% health and no bad blocks, no degraded performance and now boom! It's just failed completely.

The BIOS can sense it as the power-good beep takes about a minute to happen while it hangs the system, but it can't be seen by it. Moving to a different port makes no difference. Booting off the temporary SSD and then connecting the Samsung and doing a hardware scan causes a long hang and is then again not detected. Disconnecting it and then doing a hardware scan or connecting another drive and then scanning makes the scan complete immediately and the newly connected drive show up. This drive is dead for sure and without warning, it just turned off like a light switch. I'm feeling very frustrated right now as it's really screwed up my plans.

I'd done everything I'd needed to do on the temporary OS copy on the spare SSD and was ready to sort out the original SSD to make it boot etc. Thankfully, I've not done anything too radical on the temporary SSD so the OS copy on there is perfectly functional. Thing is, I've got W11 on it now and wanted to go back to W10 for now on the original SSD as I prefer it in some ways with the interface and wanted to do W10/W11 comparitive benchmarks, but now can't do it so easily. I'll have to make another copy and then roll W11 back and hope the PC still works after that.

So, I figured out how to activate Windows on the new hardware as it turned out that the digital license actually is associated with the account after all since when I connected the SSD to the old PC Windows activated and stated that association. I just had to use the activation wizard to tell it that the new hardware is now the current one. I don't remember seeing that option previously, or perhaps I just missed it, I'll never know.

So, W11 is currently running on a cheap £15 Patriot SSD which I'm sure won't last a decade of hard use, hence I'm gonna have to buy a Samsung replacement now instead and image it over to that. An expense I really didn't need now.

That Patriot SSD:

https://www.amazon.co.uk/gp/product/B087K9L3R5

My new Samsung SSD:

https://www.amazon.co.uk/dp/B08PC5DKZQ

Both drives are from Amazon, not Marketplace.

My fellow PC enthusiasts will click the link and note that this actually isn't the expected superfast M.2 SSD of some description, but an 870 EVO 1TB SATA drive and give me strange looks along with "R U mad bro?!" pronouncements. Especially so as the 990 Pro 1TB M.2 is only a tenner more, so getting that is a no-brainer, right? Not quite, as much as I'd like to have one.

The problem with the M.2 format that I've noticed from the start, is that it's not easy or convenient to disconnect the drive from the PC and looking at the BIOS on this mobo, I can't see a way to disable it there, either. Since I do need to boot off different operating systems and drives at times for testing and so on, not being able to disable the original boot drive is a major inconvenience to the point of showstopper as I'm not gonna take my PC to pieces to do it every time and then again every time I need to re-enable it.

Besides being screwed into the mobo with a cover sitting over it, the big, heavy, graphics card is on top of it too, so would have to be removed each time as well. Oh, and of course, M.2 drives are by definition, not hot pluggable. So, I'm forced to trade off performance for functionality. However, it's not such a bad tradeoff as even this cheap Patriot boots up lightning fast.

In the meantime, I mourn the death of my trusty Samsung SSD, and am glad that I took a copy just in the nick of time. Note that this calamity was going to happen anyway, regardless of my PC upgrade and if anything, a little sooner since the drive has had a rest while the Patriot was being used as a scratchpad.

What I have to do now is finish off the driver installations, play some games and then run those benchmarks and publish them here. Hopefully no further calamities await me.

 

[i1stadam]

Brilliant read, looking forward to updates

 

[Retro]

Smallish update as things aren't quite finished yet.

Since the 256GB system SSD died, I've bought its replacement from Amazon and installed it, a Samsung 870 EVO SATA 1TB SSD and imaged W11 onto it from the Patriot SSD. It's working fine and I just have to adjust the partition sizes to make use of the extra space.

I've played a few games and they're playing beautifully, with Rainbow Six | Siege on Steam achieving 144fps+ at 4K on my RTX 2080 at times and the low framerate issues on Call of Duty have disappeared, playing fast and smooth now. The second game was the one that really spurred me into upgrading now.

I still have to install things like chipset drivers etc which I'm trying to get from the original mfr's website wherever possible. Intel make this surprisingly difficult, forcing a registration to do so for the Z790, which I'll be doing. I do this to ensure that I get the latest driver, not just the out of date one from the mobo mfr. However, for the Realtek sound driver, I have to rely on MSI as that company doesn't even have the drivers. They've only got a small handful for what appear to be outdated hardware that's not sold anymore.

I'm running W11 and the option to roll back to W10 is unavailable for some reason, even though it was only installed 3 days ago. Therefore, to run the W10 benchmarks, I'll have to reinstall W10 fresh on the now spare Patriot SSD and update it with the latest patches. I want to run all the benchmarks together and then publish them all at once rather than piecemeal.

That's it for now, any suggestions for what to do next or comments welcome.

 

[Retro]

Ok, after another sleepless night, I thought I'd do a little more with the PC setup's fine tuning.

The other day, I installed the chipset and sound drivers. I must admit that I'm underwhelmed by that Realtek audio control panel functionality although the sound quality is good, like I expected. I'll see if I can use my old and trusty Creative X-Fi Fatal1ty sound card with W11 so then won't have to bother with the Realtek.

Unfortunately, the card is now years out of support and the outdated driver had intermittent and serious* issues in W10, so I hope it's a bit better on W11 somehow. I'm not optimistic though and it might be even worse, in which case the card becomes a paperweight and I'd be gutted since the sound quality from it is a cut above the rest, really quite noticeable.

I also sorted out the fan profile for that second fan on the CPU cooler. Previously, it would spin up to 100% every time. To fix it, I went into the BIOS and simply copied the fan profile of the first CPU fan to this and enabled smart mode. It now works properly and my Noctua cooler is more effective at keeping the CPU cool under load.

Other things I still have to do:

• Memtest the memory, but it's been faultess so far, so chances are that it will pass - all 128GB of it.
• Install W10 on the spare Patriot SSD and do the W10 benchmarks to compare with W11. I can then finish them all and finally present them here.
• See if it's possible to fix those cod6 chat audio issues once and for all, although I'm not optimistic. Did a quick test the other day and it was still the same broken experience despite the new hardware. I'm really pissed that I spent £70 on this game, yet effectively can't play it because my friends can't hear me. Rainbow Six | Siege has no such issues and plays at a really high framerate, too. While I intend to play the cod6 campaign, I specifically bought it to play online with my friends so you can understand my annoyance.

*Severe crackling and distortion that would come and go and sometimes needed a reboot, or even just a logoff / logon to fix, for a while. I even had to turn off Windows Fast Boot to reduce the problem. I have no such issues with the onboard Realtek sound card.

 

[Retro]

After being rock solid until now, I just got my first blue screen of death crash a few minutes ago. I was moving the volume slider of a YouTube video when it happened. Even one crash like this isn't a good sign and it didn't have anything informative in the message either, no reference numbers or anything. Perhaps it's the sound driver given that adjusting the volume appeared to trigger it, but it's the latest one and doesn't seem that likely to me.

I hope like hell it's not the CPU causing this already. Checking the temps again in AIDA64 look fine.

Note that I've not got round to running that memtest yet, so sooner rather than later might be a good idea now.

I also ran cod6 again the other day and the sound fault is still there, so clearly this game is just broken, or it's intentional, which I'm leaning towards. Playing Rainbow Six | Siege with Discord is faultless and runs a high framerate, even in 4K on this old graphics card, so why is cod6 like this? What does Activision have to gain by blocking Discord? It doesn't even work properly when trying to use its own sound support between PC and console, ie there's no speech audio at all.

 

[Retro]

It BSODed for a second time today. This is exactly the problem that I wanted to get away from, especially as it could be caused by anything and my original 2700K had a tendency for this that never completely went away. The temp PC showed that total stability is possible, so I'm really pissed that my new one is already showing a problem. It does make me wonder if it's the problem that's been plaguing 13th and 14th gen CPUs, but it would seem to be a bit early to show and it does have the fixed 12B microcode update since before even installing the CPU.

Anyway, I did the following:

• Took a photo of the error and Googled it.
• Disabled the Wi-Fi network as I'll never use it other than in exceptional circumstances as it's a potential source of driver errors. The antennas aren't even connected, but it still receives a pretty good signal since it's only a few metres away from the router.
• Checked the BIOS voltages and settings which all look ok and everything is still stock, as intended. I don't intend on overclocking this PC due to the high temperature that the CPU runs at when pushed, especially not now with this BSOD problem.
• There's no further BIOS updates available.
• Ran a brief memtest which was fine. I needed to use the PC, so will run it again for a longer test when I go out next. Must remember not to wipe that USB stick lol.

I might make a clean install of Windows at some point, but I really shouldn't have to and doubt it will fix this problem which I suspect is caused by hardware like it was with the 2700K. The most annoying thing about such an intermittent problem like this is that you can't pin down what's causing it. Say I could definitely determine that it was the CPU or mobo, then I'd just return them and buy new as they're still within their return periods, but can't do that without pinning it down.

Finally, as far as those comparative benchmarks go, dunno if that's gonna happen now. The bottom fell out of it when the original system SSD died and the backup couldn't be rolled back to W10 as I just can't be bothered with the faff of installing W10, drivers and everything else needed to make a working and comparable installation. If I saw a demand for it, then I might feel more motivated to do it, but given the lack of responses to this thread, I doubt anybody cares anyway.

@Geffers you might be interested in these last two posts.

 

[Geffers]

Faults like these are hard to narrow down as to the cause. Millions of instructions happening in the blink of an eye, it is a wonder it actually works at all. Haven't done any serious upgrading for years when I used to own a desktop, I fiddle a lot with raspberrypi SBCs but of course nothing is exchangeable on them, my laptop I have changed memory and SSD but that is all. Occasionally I get a complete lock up but that is seldom.

Frustrating with a new rig but of course the hardware manufacturers will blame the software and visa versa.

Geffers

 

[Retro]

Here's the reason for the BSODs:



Check out the memtest failures in red. Gonna have to start removing memory to figure out which stick(s) are bad. If it generates these regardless of sticks, then the CPU is likely bad, if I never see these again, then it's indeterminate. What fun.

Test was run for a couple of hours while I went out and looked like this on my return.

 

[Retro]

Well, this is interesting, see the screenshot below: the memory modules have very slightly different model numbers, despite me having bought the same product twice. One model ends in M161D and the other C16D. This isn't what I wanted and could be the root cause of the memory errors. I only realised this when I ran memtest which revealed the full model number of the modules. Turns out that AIDA64 Extreme that I was using first doesn't show the model number without some digging, while HWiNFO in the screenshot more readily shows it. Also, memtest shows it when running, but then replaces it with the error messages. I guess that's because it works in DOS mode so has limited real estate to work with. Can't press a key to reveal it afterwards, either.

Now, when I was getting those errors, they would start to show up in memtest after just a minute or so, while running at their rated speed of 5600MHz (XMP profile 1 in the BIOS), but the modules were arranged differently, M161D in rows 0,1 and C16D in rows 2,3 which is the correct arrangement as channel A is 0,1 and channel B is 2,3.

I then moved them into the alternating arrangement below, but now I can't set them to XMP1, their rated speed of 5600MHz since the mobo just sits there for several minutes, eventually reporting an overclocking error and falling back to 4000MHz. Tried several times now with the same result.

Later, I'll be doing more troubleshooting by removing two modules and having two of the same model in channels A&B and seeing if they work properly and so on, but not now as it's only a few hours until new year and I've got a working PC, albeit running a bit slower. You wouldn't know it from the desktop responsiveness though.

 

[Geffers]

Often fault tracing is a process of elimination.

 

[Retro]

Last night I went round a friend's house for new year's celebrations and left memtest running at this 4000MHz setting. Just came back and it's passed, having gone round the full testing loop twice. This is great news since it means that the CPU isn't damaged (it really shouldn't be) and that it's possible to run the RAM without errors at a lower speed, if I have to.

Tomorrow, I'll swap the sticks round to their correct positions, but with only two sticks installed, check I can set it to XMP1 5600MHz again and then memtest once more. If I get errors, then I'll try the other two sticks and see if they can run at 5600MHz without errors. If so, then I'll test the original two sticks one at a time and see if one of them is bad and then RMA it. However, I hope that I'll be able to run all 4 at full speed and this problem will finally be behind me. However, if the second two show errors, then I'll have to run them one at a time and see which one is bad. It wouldn't surprise me if all sticks run just fine on their own, in which case testing will be inconclusive. This is a place I really don't wanna be.

 

[Retro]

Right, now is tomorrow so I decided to go for gold: didn't bother with running just two modules. I swapped the two middle ones to get them in the right positions, set XMP, ran the test again and this time it passed. Hopefully that's the end of these errors now.

What a relief! And a nice start to the new year.

 

[Geffers]

Don't errors occur quite regular with computery things but the software compensates? I know TCP has error correction although apparently UDP does not.