Solved Gfx slow and then BSOD.

[aggixx]

Pulled out one of the two sticks and I'm now testing to see if they're both bad. So far it looks like it's only one of them.

What surprises me is how MemTest found 12.5 mil errors, and Windows Memory Diag said it was fine.

Does the windows utility often suck that bad?

Will happily post back here after I am able to RMA one (or both) sticks, to give a status report.

 

[Maguscreed]

I personally prefer memtest86, but the built in diag is what we have a tutorial handy for.

 

[aggixx]

According to the shipment tracking, my RMA arrived at G.SKILL 4 or 5 days ago, yet I have not recieved any emails from G.SKILL acknowledging this. Is this typical?

PS. Not sure whether to start new thread for this or not, sorry

 

[aggixx]

First off, I'd like to say that I am NOT happy with G.Skill's customer service. I sent them an email asking the status on my RMA and they asked for my tracking number (the postal one, I provided the RMA number in the original email). I then replied with the tracking number and a full copy/paste of the confirmation email I recieved from USPS.

I didn't recieve a response for a few days, so I sent another email with the tracking number. Eventually, they sent me an email saying that they had shipped it, but I actually received that email the day of it's arrival :sarc:. All in all it took about 3 and a half weeks for the whole RMA procedure.

Anyways, I installed the new module yesterday, and I just now got another BSOD (DRIVER_IRQ_NOT_LESS_OR_EQUAL, or something like that). I hadn't gotten a single BSOD the whole time I was running on just the one stick. Seems to me likely the new stick is faulty as well. I will be running a memtest overnight or ASAP.

What're the chances that I've gotten two duds in a row?

 

[Maguscreed]

It's completely possible you could get two bad ones in a row, though not really likely. Then again I have absolutely zero personal experience with gskill products.

It is conceivably possible the memory slot you are actually putting the ram in is bad as well.

 

[aggixx]

Yeah, a fellow gamer just suggested that to me and that's something I'll definitely need to test when I do the memtest.

 

[aggixx]

So I did actually do a memtest that night. I ran a 12 pass test on one stick (no errors), 4 pass test on the other (no errors), and then 2 or 3 passes on both sticks (no errors). I would've posted about this sooner but I felt like I was missing a piece of the puzzle and I waited for more data.

It seems to me that the cycle has started over. I started off with relative stability, no BSODs for a couple days. Then I end up getting an occasional bluescreen (especially while booting) and eventually it turns into a full blown "just can't boot successfully" because it's throwing so many errors at me. I got a few "IRQL_NOT_LESS_OR_EQUAL" and a couple "BAD_SYSTEM_CONFIG_INFO". Edit: Also got a few "Overclocking Failed!" from the post screen (had them OCd to DDR1600, speed they were designed to run at.)

I then decided to take out the new stick of ram and run only on the old one, and now I am running fine.

Meant to do a overnight run on both sticks last night, but I forgot

Is it possible that one of the sticks just doesn't play well with the other? What would cause gradually increasing rates of BSODs considering I can just take one of the sticks out and run the computer flawlessly? Any other ideas/suggestions of what the problem could be, I'd love to hear them.

PS: While running these two current sticks, the only BSOD I've recieved was the one 3 posts ago. All of the other ones happened exclusively during bootup.

Edit: Putting this out there now since it may be of relevance, the RAM I am using is G.SKILL DDR3 1600 Ripjaws (model F3-12800CL9D-4GBRL, 9-9-9-24 timings)

 

[aggixx]

Still running on (an annoying) 2 GBs of RAM, no BSODs.

 

[mgorman87]

I would update to SP1 and then try your memory again. Maybe it was something else conflicting with the memory.

How to update to SP1

 

[Dave76]

Is there any difference if you change the RAM cards, if you have been running one 2GB card switch to the other card.

Often when running 2GB RAM cards, 4GB RAM or more you need to bump the memory controller voltage.
With the below information, we can try adjusting a few settings.

When you run memtest86+ you are looking for bad RAM cards, bad motherboard slots, so you need to follow the testing procedure in the tutorial.
Sometimes errors from these tests can be corrected by adjusting the settings.
RAM - Test with Memtest86+

Can you download CPUZ and post a snip of the CPU, Mainboard, Memory, and SPD tabs.

For posting the information , enter 'Snipping tool' in the Start button > Search box, hit enter. Select the area you want to post and save to a convenient place.
How to Post a Screenshot in Seven Forums


Have you been checking your temperatures?

Use Speed Fan or Core Temp, run one of these and check the CPU temps when at idle, under load and/or testing.
Let us know the minimum and maximum temps you get.

Go to your BIOS and tell us what the CPU, RAM and North Bridge (NB) voltages are.



Which version of NVIDIA nForce 680i motherboard do you have?

 

[aggixx]

Installing SP1 for general performance and security increase.

Dave76:

Just realized my specs on my profile are outdated and are from my previous rig. I will update my system specs and post the requested information ASAP. I have already run multiple MemTests on both sticks of RAM using all 4 motherboard slots, with no errors.

Edit: Specs updated. CPU-Z snips added. SpeedFan temps added (note idle means chrome, ventrilo, explorer, skype etc). Note that I am currently running stock CPU fan because I temporarily can't use my Cooler Master 212+

 

[Dave76]

Do you get memtest86+ errors with both 2GB RAM cards installed?

Everything looks good, in BIOS what is your VDDNB voltage set at?
Should be default 1.10v unless you have changed it.
This is the North Bridge memory controller, bumping the voltage a little will help with RAM stability.
Don't go too high you can cause serious damage.

If your VDDNB is set at 1.10v then increase the VDDNB to 1.20v.
Install both RAM cards and try to re-create the BSoD scenario or run memtest86+ if it has not passed memtest86+ with both RAM cards installed previously.

Let us know the results.

 

[aggixx]

My RAM has passed 10+ passes with both sticks installed, and 5+ with the two sticks individually.

All voltages in the BIOS are set to Auto (and I'm starting to realized Auto is the devil).

 

[Dave76]

Are your RAM timings set to auto as well?

You will get better stability if you manually set the first four timings as per your SPD tab, leave the rest on Auto.
Which is 9-9-9-25.
You will also get better stability by setting the 'Command Rate' to 2T from the current 1T.
These changes will not noticeable affect performance, but will help with stability.

The RAM frequency is currently at 666.7 MHz, this is fine, no need to change it.

Are you going to change the VDDNB voltage to 1.20v?

Memtest86+ runs different tests with each pass, it is best to let it complete 8 passes.

Disclaimer:
Although memtest86+ is probably one of the best RAM testing software available, there is no software that is 100% accurate in finding RAM errors/faults.
Just because RAM passes multiple tests doesn't mean the RAM is not causing BSoDs.

Adjusting timings and voltages does improve stability. Auto settings are known to be the cause of hard to find errors and BSoDs.

Let us know which adjustments you are changing, it is your choice. The previously suggested changes are widely given on many sites and can improve stability.

 

[aggixx]

The RAM is designed to run at 800 MHz, but needs to be overclocked to reach that speed.

I plan on changing the timings and command rates tonight.

 

[Dave76]

I would strongly recommend not to change the RAM frequency to 800MHz (RAM speed 1600MHz), there are known issues with this speed on AMD CPUs.
If you want to try the RAM at 1600MHz, get everything else adjusted so your system is running 100% stable.
When you try the 1600MHz for RAM, then you will know the cause of BSoDs.


There is a known issue with AMD CPUs integrated memory controllers having problems running RAM at 1600 MHz and above.

VERY IMPORTANT INFORMATION REGARDING AMD AM3 CPU's and RAM SPEEDS

 

[mgorman87]

I was working on a friends computer who had an AM3 and his RAM was also 1600mhz but the BIOS would default to 1333mhz. When I bumped up the BIOS to 1600mhz I would run into a lot of memory errors. I don't think you would see that much of a difference between 1600 and 1333.

 

[Dave76]

Read the above link, it explains the issue.

Strange but, a lot of people are experiencing it.

 

[aggixx]

Thanks for that info, something I definitely didn't know. I changed the timings and command rate like you suggested, and I'm going to throw some stress onto this thing to see if it holds.

Edit: BSOD'd with those settings, after some trouble getting it to boot again, I changed the VDDNB voltage to 1.2000. I then had some trouble with "Overclocking Failed!" errors during post, so I changed the RAM to ganged as suggested in that link you posted. It turns out after looking at the link again, it was talking about ganged causing OC fails and not unganged, so . Either way, I'm booted now and I'll see if the voltage change helps. The Overclocking Failed! error seems to just be in the way and not really mean much.

AMD recommend running RAM in Unganged mode.
From my own testing, RAM above 1333 Mhz can be extremely fussy if you insist on trying to run in Ganged Mode and will often return an Overclock warning at start-up and revert to 1333 Mhz., by default. In most modern mult-threaded applications there is actually no loss in performance when running in Unganged Mode and, in some instances, Unganged mode is actually superior.

Edit #2: Another BSOD, changed it back to unganged mode. Removed 1 of the sticks so I can run stable until I can get more suggestions. This time I'm running on the new stick of RAM, and the old one is sitting in an anti-static bag.

 

[Dave76]

Did you run with the RAM timings set in manual, unganged, Command Rate 2T, VDDNB voltage 1.200v, and both RAM cards installed?
The 'ganged' mode may have caused the crash.

What is the DRAM voltage in BIOS?
The voltage numbers in CPUZ are recommended and not actual.

If you are running as mentioned above and have a crash, let me know the DRAM voltage, bump the VDDNB to 1.240v and test for stability.