Almost daily BSODs in last two weeks

Topic

New Reply

Thanks, PK, here’s the info on the BSODs which occur at random times, including when system is doing nothing.

128 GB RAM total = 4 x 32 GB. Win10 Pro x64 22H2, Build 19045.3086

2016 custom build with GeForce GTX 950 video card, replaced in 12/2022 with used Nvidia GeForce GTX 650Ti. Device Manager shows card is “working properly.”
Nvidia driver (signed by MS Windows Hardware Compatibility) ver. 27.21.14.5671 dated 9/30/2020.
Tried to update driver via Windows update but nothing more recent, nor could I find more recent driver at Nvidia. Bluescreen View seems to point to VGA driver issue.

CHKDSK and SFC scan found nothing. Ran “DISM /Online /Cleanup-image /Restorehealth” and got “The restore operation completed successfully.”

Windows Memory Diagnostic: “Your computer has a memory problem.”

Ran “Memtest86+” for 3 hours and got 610 errors. Will run again overnight.

Thanks in advance to PK and the other experts.

I have a 3.9 MB zip file with all the dump file information.  Also the link below on Google Drive should work.

Google Drive: Sign-in

Reply | Quote

Replies

Have you read this thread?
https://www.askwoody.com/forums/topic/how-to-open-a-memory-dmp-file-for-analysis-in-windows-10-please/

cheers, Paul

2 users thanked author for this post.

cmar6, satrow

Reply | Quote

Thanks, Paul, for the link. I also have Blue_screen_view. I uploaded to this thread the zip file of the dump files, hoping someone here can point to possible source of the BSODs. I don’t know how to interpret Blue_screen_view output, assuming it’s useful in the first place.

I have removed 2 of 4 RAM sticks and am running Memtest86+ on the remaining sticks.

Reply | Quote

Please follow the “Results & Extended Testing” section of the opening Post here to determine which of your RAM sticks/slots are causing the errors.

If MemTest86+ runs for 8 passes without throwing up an error then it’s more or less a certainty that your RAM is good. If however MemTest86+ does find errors then it’s a case of determining if it’s actually your RAM that’s faulty or one or more of your DIMM slots. To determine this you will need to test every stick of RAM in every DIMM slot.

Reply | Quote

Satrow, As a first cut, I’m testing system with 2 dimms out, so now testing only “Slot 0” and “Slot 2” according to Memtest86+. I pulled out the two right-most dimms. Are you implying that sometimes it’s not the dimms themselves but the slots that are causing BSODs?

Reply | Quote

I’ve just had a quick look at the BSOD types, 3/4 are 0x50:
“The PAGE_FAULT_IN_NONPAGED_AREA bug check has a value of 0x00000050. This indicates that invalid system memory has been referenced. Typically the memory address is wrong or the memory address is pointing at freed memory.” https://learn.microsoft.com/en-us/windows-hardware/drivers/debugger/bug-check-0x50–page-fault-in-nonpaged-area

So it’s vital to get the memory checked and any faulty sticks removed/cleaned/replaced ASAP, once done we can wait to find out how the machine behaves under normal usage again.

Yes, either/both of the RAM sticks and/or slots might be damaged or possibly just dirty (pencil eraser to gently rub down the gold contacts on the sticks), cleaning slots is much trickier. Some people have washed RAM sticks – feasible but a good ‘recipe’ needs to be followed – outside of my experience.

1 user thanked author for this post.

cmar6

Reply | Quote

Satrow, very useful and educational. Thanks for checking the BSOD type. I’ll run Memtest86+ overnight on Slots 0 and 2 and report back to you tomorrow. You are implying that faulty sticks can be cleaned or replaced? I didn’t know that.

I recall G.Skill sales (or propaganda?) on their site to the effect that the kit of 4 x 32 GB = 128 GB is a matched set, implying you can’t replace any sticks. Maybe that’s just my interpretation. I can live with 2 x 32GB = 64 GB if I have to. It’s good news that there may be options to restore or replace a faulty DIMM.

Reply | Quote

Depends on what’s damaged/dirty, damage may not be fixable, dirt may be.

What was the cleanliness of the PC case innards like generally, much dust/grime? How about the installation location, free from vibration, unlikely to be knocked?

Regarding replacement, best to check with G Skill whether they could provide a matched pair that should work with whichever pair you’re left with, or maybe they could just find a single. Perhaps they’ll man-up and supply a replacement set…

I know RAM compatibility overall has improved greatly in recent years, I’m not sure if the matched pair/matched four isn’t just a hangover from the past – as you say, (or propaganda?)

Reply | Quote

Satrow:

Thanks for the useful post. Did you look at my zip file of the BSOD dumps? I ran Memtest86+ overnight for 10 hours with these results: 171 errors; Status: Failed. Pass 2. Pass 56%; Test 29%. 12 failing addresses.  L1 Cache 32 KB; L2 cache 512 KB; L3 caches 64 MB; Memory 127 GB.

Thanks, cma

Reply | Quote

I’ve never touched RAM in this custom built system of 11/2021 so there are no manuals, but I’ll look online. Next is to figure out where the RAM is and how to remove it systematically. In another system I destroy 4 RAM sticks by vacuuming out the system.

RAM: G.SKILL Trident Z Royal Series 128GB (4 x 32GB) 288-Pin DDR4 SDRAM DDR4 2666 (PC4 21300) Intel XMP 2.0 Desktop Memory Model F4-2666C19Q-128GTRS

Mobo: Gigabyte X570 Aorus Elite Wifi

Thanks alejr and satrow.

 

Reply | Quote

Alejr:

This is hugely useful. I will put it on one screen of another system so that I can hopefully pull out the two RAM sticks tomorrow and get started once the thunderstorms have passed.

Reply | Quote

cmar6 wrote:

You are implying that faulty sticks can be cleaned or replaced? I didn’t know that.

The contact points on both the RAM stick and the slot are “gold plated” so it’s extremely rare for there to be any corrosion build up on them. And, most of the time, simply unplugging a RAM stick and plugging it back in is sufficient to remove any crud that may have built up on the contacts.

A more likely cause of RAM stick failure after an “extended period” of use, is overheating of the RAM stick itself; which is much more common on newer motherboards where the manufacturers have “increased” the memory voltages to make the RAM more stable at faster speeds.

This problem is also common on the “smaller” form factor motherboards (i.e. Mini-ITX or smaller) where other “heat producing” components on the motherboard are closer to the RAM slots but, since your motherboard is an ATX, that shouldn’t be an issue in your case.

cmar6 wrote:

I recall G.Skill sales (or propaganda?) on their site to the effect that the kit of 4 x 32 GB = 128 GB is a matched set, implying you can’t replace any sticks.

Actually, the term matched set doesn’t mean you can’t replace one or more sticks, it means all the sticks in that particular set were tested together “as a group” up to their max rated speed [i.e. DDR4 3200 (PC4 25600)] and there was no measurable different in the read/write response times between each stick.

95% of the time, unless you’ve “overclocked” your RAM, a defective stick can be replaced without encountering any problems as long as the replacement stick is the exact same make/model as the defective one.

BTW, all G.skill RAM comes with a Limited Lifetime Warranty from defects in material and workmanship (as long as they’re being used as intended, i.e. not being overclocked) so, if one or more of yours test bad, you should either contact whomever you bought them from for replacement or use G.skill’s RMA Request Form.

1 user thanked author for this post.

satrow

Reply | Quote

Hi Alejr: I am currently running Memtest86+ with 2 modules removed (from Slots B1 & B2): 4 passes so far without errors. I do have a “newer” mobo. I have not O/C RAM: SMT 24T (Par) and CAS 19-19-19-43.  Good news from you about being able to replace 1 or 2 defective sticks of RAM.  So far G.Skill tech support has been responsive, sending me the link to their RMA Request Form, so I’ll keep my fingers crossed.

When system case was open, I did not notice any obvious dust, to Satrow’s point.  After current Memtest run finishes, I was considering using either soft brush or compressed air on open Slots B1 & B2. Is compressed air a bad idea with two RAM sticks in place?

Reply | Quote

cmar6 wrote:

Is compressed air a bad idea with two RAM sticks in place?

As long as it’s “canned air” and the can is held “vertically” while using it (to prevent any of the liquid propellant from spraying things), it won’t be a problem.

You should never use a brush on anything inside a PC as it can cause a build up of static electricity which can damage things!

Just FYI, about 2 years ago, I started using this OPOLAR Cordless Air Duster/Vacuum instead of canned air.

It’s only “slightly” larger than a typical can of compressed air, has a 6000mAh “rechargeable” battery (supports 10W fast charging), has 2 speeds (30,000 & 60,000 rpm) and, with the blue vacuum attachment removed (it twists off/on), blows air as powerfully as canned air without the possibility of spraying liquid propellant on anything!

1 user thanked author for this post.

Lars220

Reply | Quote

Bad news on this test. After 26+ hours and on Pass 6, I had an error with Slots A1 and A2 occupied.  So now I will remove a RAM stick from Slot A2 and run Memtest on only Slot A1 = Slot 0. Can I assume one error is enough to define bad RAM (or defective slot?)

Reply | Quote

Memtest errors point to bad RAM and even one error should not be accepted.  But, I have seen a time or two when the memtest errors were caused by poor CPU cooling or accidental overclocking or undervolting of RAM or CPU.

Reply | Quote

cmar6 wrote:

Can I assume one error is enough to define bad RAM (or defective slot?)

As pointed out above, even 1 error is too many. Good RAM sticks/slots should be able to run multiple passes of memtest with no errors at all!

At this point, you should test each individual stick in slot DDR4_A1 to determine whether the problem is just one or more RAM sticks (i.e. only some of them fail) or if the slot itself may be bad (i.e. all 4 of them fail.) Be sure to keep track of which ones fail/pass (maybe a “small” marker tick or piece of tape somewhere noticeable on the stick itself.)

cmar6 wrote:

So far G.Skill tech support has been responsive, sending me the link to their RMA Request Form, so I’ll keep my fingers crossed.

That’s good to hear! After you verify which sticks are bad, you should ask whether they’re willing to replace the whole matched set of 4 x 32GB or only the defect sticks?

As I pointed out previously, 95% of the time replacing a defect RAM stick from a matched set with the exact same make/model will work just fine but…

If they’re willing to replace the “whole set“, and you can deal with the downtime waiting for them to send the replacements, I’d go for it as the new set will have been tested to be 100% compatible with each other.

2 users thanked author for this post.

cmar6, satrow

Reply | Quote

RAM stick #1 in Slot A1 had no errors in 9 passes, so RAM #1 and Slot#1 are both good. Now I’m testing RAM stick #3 in Slot A1.

1 user thanked author for this post.

satrow

Reply | Quote

RAM stick #3 had no errors in good Slot A1. However, with RAM stick #2 in Slot A1, system won’t boot up and won’t even go to mobo screen. Can I assume that #2 is bad?

“BIOS has been reset.”

Now testing module #4 in Slot A.

Reply | Quote

cmar6 wrote:

Can I assume that #2 is bad?

Yup, stick #2 is definitively bad and needs to be replaced!

Reply | Quote

@cmar6 –

AFTER you’ve cleared up the bad memory issue, there is indeed a driver update for your video card, which was released back on June 26th of this year!

You didn’t get any results from Windows Update, because back in October 2021, Nvidia cut your card series off from any more feature updates, but will continue offering driver security updates for it until September 2024 (details below).

Here’s a link to Nvidia’s page with a download link for the update as well as other important information: https://www.nvidia.com/download/driverResults.aspx/206792/en-us

The link above should open in a separate tab. Once you’re on that page, you should see a green button labeled, simply, “Download”. Clicking that will begin the driver’s download to your computer.

A little background on your current Nvidia chip, the GTX 650Ti: It’s in the Kepler family of GPUs and, as such, will stop receiving any more updates come September of next year, 2024. Nvidia stopped producing regular driver updates for it back in October 2021. This info comes straight from the driver download page linked to above. In addition to the green button labeled “Download”, there are three tabs on that page, one labeled “Additional Information”. Clicking that tab will produce the following statement:

Effective October 2021, Game Ready Driver upgrades, including performance enhancements, new features, and bug fixes, will be available for systems utilizing Maxwell, Pascal, Turing, and Ampere-series GPUs. Critical security updates will be available on systems utilizing desktop Kepler-series GPUs through September 2024.

In hindsight, if you’d managed to replace the GTX 950 with another one, you’d still be able to receive the ordinary Game Ready Driver upgrades mentioned in the above quote. From what I can tell, the last Game Ready driver update for your current card was released back in September 2021, and was version 472.12; all driver updates for your card since then are security update drivers.

By the way, to tell what the current version of your Nvidia driver is, simply use only the last 5 digits of the version listed by Windows. So, your current driver version is actually 456.71, since the last 5 digits of the driver number reported by Windows are “…4.5671“.

I sincerely hope all this helps if you’re still experiencing problems after clearing up the memory problem that’s surfaced!

Reply | Quote

Bob, this is invaluable information, especially as the BSOD logs indicate that for almost every BSOD, there was in issue involving the VGA driver! Once I work my way through RAM issues, and hopefully get new RAM from G.SKILL, the next thing will be to update the video driver and then see how I stand.   You, Satrow and Alejr have provided really high level info, which is one reason the Askwoody community is so successful.

Reply | Quote

All Memtests complete. One bad RAM#2, all DIMM slots look good. I sent RMA to G.Skill and am awaiting reply. Now with 3 good RAM modules in place, is this time to update the GTX 650Ti Nvidia driver or should I wait till I have 4 good RAM modules in place?

Reply | Quote

Always fix one problem at a time. In this case sort the RAM – I assume you can wait a week for the driver update.

cheers, Paul

2 users thanked author for this post.

cmar6, satrow

Reply | Quote

BTW, don’t “assume” the new stick is good simply because it came directly from G.Skill.

Install it in your system (I’d suggest the very first DIMM slot, DDR4_A1) and then run Memtest on all 4 sticks as a group (a minimum of 8 passes) to personally verify there are no errors. And remember, even one error is too many!

IF the RAM all tests good, then (and only then) proceed with further T/S, updating drivers, etc.

1 user thanked author for this post.

satrow

Reply | Quote

Alejr and Paul: That raises an immediate question on my RAM sticks. 3 of them are good–tested individually for 9 passes. I received late yesterday from GSKILL the RMA OK. I immediately replied, asking if the RMA was for 1 RAM module or 4 modules but won’t get a reply till Monday at earliest.

Based on what you say about not assuming that I would get a set of 4 new modules–even if GSKILL agrees to that–it may be best to keep my 3 known good modules and only return the one bad one today?  What do you think? Also, running Memtest on 4 sticks at once will take forever.

Reply | Quote

You have tested the 3 good ones so a quick test of a single replacement will be fine – alejr is being cautious, which is a good thing.

cheers, Paul

Reply | Quote

@Paul T‘s right, I’m being overly cautious, but there’s a reason for it.

Just like CPU manufacturers, RAM manufacturers don’t stress test their products.

For normal single sticks of RAM, they perform a quick single pass write/read test and, if it passes, it gets packaged and shipped out. And even “matched sets” such as you got, only get a “slightly more complicated” single pass write/read test (as a group of # sticks) to ensure there’s also no measurable different in their read/write response times before they also get packaged and shipped out.

While there’s a pretty good chance a new “single stick” would work just fine with your other known good ones, it’s not 100% which is why I recommended running a test on all of them as a group to ensure they all work together without any errors.

Just FYI, the chances of actually getting a bad RAM right out of the box are pretty low, but they’re not zero so my policy has always been to run a stressful Memtest on any new RAM I purchase (i.e. 8 passes so they heat up and really get put thru the wringer.) That way, I’ll know if it’s flaky and can immediately ask for a replacement instead of encountering problems with it down the road like you did.

Of course, it’s always your choice on how to proceed.

1 user thanked author for this post.

satrow

Reply | Quote

Regardless of anything else, your Memtest86+ results indicate you definitely have a RAM problem and it needs to be resolved before checking whether something else may also be bad!

Remove 2 stick of RAM and run Memtest86+ again.

Note: consult your motherboard manual to determine which 2 slots are the “primary” ones with only 2 sticks of RAM (it’s not always the 1st & 2nd slots!)

Just FYI, it’s “most” likely the detect is in the 1st or 2nd RAM stick (i.e. the first 64GB) as it’s extremely rare for programs to need more that 64GB to run.

1 user thanked author for this post.

satrow

Reply | Quote

Extracted from the Gigabyte X570 Aorus Elite WiFi manual

If only using 2 RAM sticks, they need to be in either DDR4_A2 & DDR4_B2 or DDR4_A1 & DDR4_B1

Location & labeling of motherboard RAM slots.

Notes:

The “right-hand” side of the motherboard is always toward the “front” of the case.

If it’s in a Tower style case, the “top” of the motherboard is toward the “top” of the case.

If it’s in a Desktop style case, the “top” of the motherboard is toward the “right” side on the case.

1 user thanked author for this post.

satrow

Reply | Quote

I just extended my AskWoody subscription for another year. I wish one had the choice of 2-3 years. This site provides invaluable expertise in all things Window, e.g., the great job done by experts in this thread Paul T., Satrow, JustAnother, and Bob99 in resolving my BSOD issue. They helped me pinpoint the bad dimm module and showed me how to test it.

I sent the bad module to GSKILL, a great vendor. When they say “lifetime warranty” they mean it. They told me to send all 4 modules even though the other 3 were fine. They then sent me a matched set of 4 brand new 32 GB modules, installed on 8/17/23 and no BSODs since then.

I also followed the good advice to update the driver for display adapter. I doubt that was causing BSODs, but it was an important security issue. Great work from the AskWoody “team.”

4 users thanked author for this post.

windbg, TechTango, Bob99, satrow

Reply | Quote

@cmar6 –

Thanks for the update, which put a nice bow on the package that comprises this thread. BTW, from what I recall, there’s a new driver package that fixes security holes put out by Nvidia about every 3 months or so. Here’s a link to the Nvidia security page that lists their security bulletins:

https://www.nvidia.com/en-us/security/

That page has a list of security bulletins for most all of their software and hardware products. You’re looking for any bulletin that has anything along the lines of “…GPU Display Driver…” in the title. Those bulletins will tell you the driver version you need to install to fix any security holes they’ve found lately.

Keeping an eye on that page will help you avoid having Windows Update try to download and install the incorrect driver that MS has put out. BUT, there is another way around that by specifying within Group Policy (or, I think, via a registry edit) that you don’t want Windows Update to scan for and install driver updates it finds. If you have a Pro edition of Windows 10, then you can use Group Policy, but the Home edition of Windows 10 doesn’t have that “luxury”.

As you may have guessed by now, come the end of next September, you’ve got a choice to make about what to do about your video solution going forward. If your computer and, most importantly, your budget support it, you could try getting a newer video card put into your computer. Or, you could take a look into acquiring a new (or new-to-you) computer that should have a newer video card.

One final request: Can you please mark this thread as “Resolved”, so others won’t try giving an unneeded answer to a no-longer-existent problem?

1 user thanked author for this post.

cmar6

Reply | Quote

Bob, thanks for the useful Nvidia info.

Reply | Quote