Solved Constant Random BSODs on a 5 days old system

[Dave76]

Open and run Core Temp during all testing, note the temps and let us know.

Disclaimer: If you don't feel comfortable running stress tests on your computer, then don't.
There is a potential to damage components. Proceed at your own risk.

I Read your post,
I'll run the memtest series when I get back(how many passes is advised?) Both modules in both slots one at a time.(slot 1 and 2 only or slot 3 and 4 also? Given your reference?)

If stick 2 fails in two slots, I'd say its bad.
You should see something by the end of two passes, if no errors let it run for 7 passes.
This will let us rule out the bad RAM possibility.

If you get to the stage of testing two 'good' RAM sticks in slots 1 & 3, run memtest86+ on both of them, 7 passes, overnight.

What's AHCI mode?

AHCI, Advanced Host Controller Interface

In your BIOS, Integrated Peripherals, (motherboard manual pg 48) OnChip SATA Type.
Check what the setting is and let us know.

The crashes are kinda random, somtimes within 30 minutes, sometimes after like 2 hours. It's really strange.

So the crashes are random.

http://www.ixbt.com/mainboard/gigabyte/770ta-ud3/board-sm.jpg
You see the coolblock that says gigabyte, I noticed that that actually becomes pretty damn hot.
But then again IMO the GPU becomes pretty hot too...

Yes, the block that says Gigabyte.

GPUs normally run pretty hot, that's normal. The north bridge shouldn't be very hot, should be able to hold your finger on it.

This is going to take a loooooong while to test xD But atleast I already have the clearence from the shop to stop by and trade them in, or I think so

Thanks again for helping and answering questions. When I get back I'll first run the memtests and try voltages, when the system is steady I'll do a clean install.w

After you run the memtest86+, let us know the results.

According to the info from AMD, let's increase the DRAM voltage first.
Set it to 1.55v,
Run prime95, enable 6 cores, torture test - Blend.

In the Prime95 advanced tab, check 'Round off Checking'
Options tab, Torture Test, select Blend, Number of Torture Test threads to run: 6​

*Have Core Temp open and running, watch the temps closely when the test starts, then until it levels off. Should be in the 50 - 60c range. Max CPU temp is 62c.
If the temps are 35ish, it's not running correctly and will not stress the CPU or RAM.

It should show and error within a couple of minutes of starting if the system isn't stable.
30 minutes should be good enough for this test.
If Prime95 stops or the computer crashes it's a fail.

Do not exceed 1.65v.

If Prime95 fails or the system crashes go to 1.6v, same test as above.
Your BIOS should have a max of 1.64v, it should increase in .02v intervals, the 1.66v setting should be in red color.
Go to 1.64v, same test as above.

If stable at any voltage, Prime95 keeps running, no crash. Stop test after 30 min, test with most stressful game for several hours.

If it's not stable at 1.64v then bring it back to 1.5v and let us know the steps you have taken and the temperatures at each.

 

[iTitus]

Ok here I go:
I did a lot of memtests:

Stick 1(the one that seemed good while testing SC2)
Passed all slots with 2,5 passes each time. Due to shortage of time I only did 2,5 passes

Stick 2(the one that seemed to mallfunction while testing SC2)
Passed slot 1 2,5 passes
failed at slot 2, 2 different times, one short run and one long run(see pics for error codes)
passed slot 3, 2,5 passes
failed at slot 4(tested 1 time, 3 passes i thought)

The pics are attached.

Onchip Sata type: "Native IDE".
So AHCI is disabled i think

Also I touched my NB cooling block during ram test, and IMO I couldn't hold my finger there(although my friend could)

I didn't raise voltages yet... First I'd like some info on the errors

 

[Dave76]

Good job on the memtest86+ tests. A little short on run times, but the information is useful.
Normally the errors are enough to indicate the problem and a long test is not necessary.
When there are no errors, running the tests longer can show errors later in the test, this is very useful information.

Test #6 will detect data sensitive errors.

RAM #2 had errors in slots 2 and 4 which is in two different channels and both memory controllers, when running unganged.

This indicates the DRAM voltage may help.

Bump the DRAM voltage as described.

According to the info from AMD, let's increase the DRAM voltage first.

Let's work on RAM #2 first.
When it passes all tests, then we can if check RAM #1 will run at those settings.

Set it to 1.55v,

First run memtest86+ for 7 passes or stop it if your getting errors.
* Always have Core Temp and HWMonitor open and running, watch the temps closely when the test starts, then until it levels off. Should be in the low 50's. Don't let it go past 60c. Max temp for your CPU is 62c.
Also run CPUID HWMonitor, this is important, there are reports that some AMD processors don't give accurate CPU temp information. You need to confirm the temps.

If it passes memtest86+:
Run prime95, enable 6 cores, torture test - Blend.
In the Prime95 advanced tab, check 'Round off Checking'
Options tab, Torture Test, select Blend, Number of Torture Test threads to run: 6
*Have Core Temp open and running, watch the temps closely when the test starts, then until it levels off. Should be in the low 50's. Don't let it go past 60c. Max temp for your CPU is 62c.
Also run CPUID HWMonitor, this is important, there are reports that some AMD processors don't give accurate CPU temp information. You need to confirm the temps.

If the temps are 35ish, it's not running correctly and will not stress the CPU or RAM.

It should show and error within a couple of minutes of starting if the system isn't stable.
30 minutes should be good enough for this test.
If Prime95 stops or the computer crashes it's a fail.

Do not exceed 1.65v.

If Prime95 fails or the system crashes go to 1.6v, same tests as above.
Your BIOS should have a max of 1.64v, it should increase in .02v intervals, the 1.66v setting should be in red color.
Go to 1.64v, same tests as above.

If stable at any voltage, Prime95 keeps running, no crash. Stop test after 30 min, test with most stressful game for several hours.

If it's not stable at 1.64v then bring it back to 1.5v and let us know the steps you have taken and the temperatures at each.

 

[iTitus]

So if I understand correctly:
Put stick 2 in slot 2
Raise DRAM voltage(only that one) to 1,55V
Run memtest 7 passes or when errors reoccur
Then: How can I run that software during Memtest? Memtest is booted from an usb device and the software is windows related??

If it passes 30 minutes prime95:
When I ran the prime95 test the other time for 12 hours, Max temp was about 36C. Same for each time I run Prime. Note that I DO have a 3rd party cooler. I don't know if it lowers the temps by 30C but ...

If it failes raise to 1,6 and repeat till 1,64
I attached a screen of the normal temps according to CPUID hardware:

 

[iTitus]

I Raised the DRAM voltage to 1,56V With stick 2 in slot 2 and did a memtest.
Is this enough to call it a "fail" for 1,56V?

 

[Dave76]

Yes thats a fail. Raise the voltage to 1.6v and test again.

Ya, you won't be able to run a temperature monitor when doing the memtest86+.

Be sure to monitor temps when running Prime95.

Something is not right if Prime95 only gives you 36c.
That is a nice cooler, same as mine, but you should have temps in the 50s when running Prime95.
Are you using CUPID Hardware Monitor when testing?
It should be accurate, run Core Temp at the same time to check/verify the temps.

What does Windows Task Manager > Performance tab > Resource monitor, say when your running Prime95?

All cores should be 100%, RAM also.

 

[iTitus]

I'm going to raise the voltage to 1,6 now. Then 1,64.
I ran a program called "LinX", which was a suggestion from GEIL. Gave me a error too, stick 2 in slot 2. Here's a screen.

When I visit task manager during a prime95 test, it says both CPU and RAM 100% usage.
So does Core Temp.

brb with results

 

[Dave76]

LinX is good also, it will should give you higher temps that Prime95.

Be sure to monitor your temps during testing with either LinX or Prime95.

Those are some low temps, I have read some posts about AMD not giving the correct temps in some cases.

Let us know the results and the temps.

 

[soaringink]

I am having exact same behavior on ASUS M4A785TD-M EVO mobo. Just now finalized both mobo bios and video driver update (direct from ATI). (The card is Sapphire 512mb, windows had loaded its own WDM drivers). My CPUID returned 9,9,9,24,33 so I will try those settings and the volt bumps. Here though, I have four brand new builds, exactly the same. In the 7 days since install, three crash repeatedly while one has not crashed ever. Sure smells like unstable RAM production.

 

[iTitus]

Dear Dave,

I Raised the voltage to 1,6V and ran 8 passes memtest with NO errors.
Please advise on the next step, the prime95 test. I fear that i'll still get the 35ish temperatures... But I'll turn it on for half an hour now

See you in Half an Hour

EDIT:
2 pics before and after running prime. No errors or something, but still 35ish temperatures

 

[Dave76]

The results look good.

If Prime95 is running all cores at 100% then it's working.
The low temps aren't bad, just unusual, maybe thats normal for your CPU.

Run Prime95 or LinX for 30 minutes, as you mentioned.
If it passes, install the other RAM stick and run Prime95 or LinX again, if it passes you can use your computer normally and see if it remains stable, play your most computer stressing/demanding games.

Memtest86+, Prime95 and LinX are good testing apps but it's not impossible for the problem to return.
There can be several reasons for this, hopefully it will run stable.

If the tests don't pass, give detailed info here and we'll proceed to the next step.

Let us know the progress.

Edit:
That TMPIN2 82c, is a bit high.
Will depend on your motherboard what that is for, from your pics it's not CPU, GPU or HDD.
Might be NB temp, if your friend can hold his finger on it, it can't be 82c.
Might be case temp, which would tell me that the sensor is bad.
Has it always been like that?

You can try this:

I have this same "issue" with my Phenom II x4 940 (in both Speedfan and CPUID Hardware Monitor).

Speedfan is displaying inaccurate values. There are two kinds of temperature monitoring devices in place on motherboards, thermistors and diodes. If the diode is set to read as a thermistor it will give you unusually high numbers as you're seeing, and I'd imagine the opposite is true of a thermistor being read as diode (although I haven't experienced this personally).

To remedy the issue in speedfan, select the "configure" button on the "readings" tab.

Select the "advanced" tab.

On the "chip" drop-down menu there should be two choices, "AMD K10 at..." and another that is specific to your motherboard. Choose the one that isn't "AMD K10 at..."

There should be a list of values in the box below the drop-down menu.
On my machine the first three under the "Property" column being:

Temperature sensor diode 1
Temperature sensor diode 2
Temperature sensor diode 3

The "Value" column adjacent to these indicates whether it is being read as a thermistor or diode.

You can select one of these values and then in the "Set to" drop-down menu below set the "Value" to thermistor or diode.

---------

In my case the "Property" said diode right in the description and the "Value" was set to thermistor so it was fairly straight forward to just switch the "Value" to diode and the temperature came to a much more realistic ~35C.

 

[Dave76]

I am having exact same behavior on ASUS M4A785TD-M EVO mobo. Just now finalized both mobo bios and video driver update (direct from ATI). (The card is Sapphire 512mb, windows had loaded its own WDM drivers). My CPUID returned 9,9,9,24,33 so I will try those settings and the volt bumps. Here though, I have four brand new builds, exactly the same. In the 7 days since install, three crash repeatedly while one has not crashed ever. Sure smells like unstable RAM production.

Welcome to Seven Forums.

Good call on the BIOS and video drivers.
Check for chipset drivers and update if you find a newer one.

You can set your RAM timings manually to 9-9-9-24, leave the rest on Auto.

Read the posts at VERY IMPORTANT INFORMATION REGARDING AMD AM3 CPU's and RAM SPEEDS
Including his first post on page 2, he clarified ganged and unganged info.

If your not set to unganged, change it.

D/L CPUZ and post the Memory and SPD tabs.

What is you current RAM voltage settings?
Don't recommend going past 1.65v.

Let us know your progress.

 

[iTitus]

@Dave

I'll install the other ram stick and do the prime test.
If it doesnt work with 2 sticks, does it mean that stick one doesn't support the higher voltage, or does it mean that because stick 1 is inserted too, the voltage drops a little resulting in stick 2 failing?

Notice, when I tried the work arround from that guy the temp 80 went to -78, so I guess that didn't work xD
I don't know why the temps are so low?
I changed another value and the fan started spinning like hell. (reminds me why I didn't set "fan speed" to FULL in the bios but let it be "Automatic"

Titus

 

[Dave76]

Did RAM card #2 pass the Prime95 test for 30 minutes?

Those are the two most likely possibilities and of coarse, that RAM card (#2) is bad and won't play nicely with others.

If it won't pass with both cards inserted, we can do some minor voltage adjustments.
Now we know card #2 will work with a little more voltage.
Maybe a little NB Voltage bump, you have to very careful with this, lets not do this until last.

Let me know how it's going.

 

[iTitus]

I think both test(ram 2 in slot 2 and both(1in1 and 2in2)) passed prime.
Also I think I found my "real" coretemp, being arround 40 something during prime.

I'll play some games and see if that works

Another question is: Is this higher voltage bad for either the RAM or the MOBO, because if that's the case I'll just RMA the ram/mobo(both bought at the same company). Thanks
And I might someday be interested in overclocking, due to this "start" I'm getting used to stability tests

 

[Dave76]

Good news both RAM cards passed.

Games are a good stress test, let us know if it stays stable.

Too high voltages are not good, you are still within the safe and reasonable limits.

You could RMA the RAM but, there's no guarantee that the new RAM wouldn't need the same adjustments.
Then again, they might not...

Be very careful of the overclocking bug, once bitten...
You'll be running stability tests all the time

It's an interesting learning experience, read up on the subject and then have fun.

Let us know how the system is running, this info will be useful to many others with similar issues.

 

[zipsi]

so much trouble with a very new system, wouldn't it be easier to return it and get a functioning one? I'd understand if it would be a year old system, then you would try anything to get it back working, but atm if anything fails - no refunds.

 

[iTitus]

Hi Guys,

Long time no see. I just played SC for 2 hours and upon alt+f4 ing out of the game I got a crash(not the best way to exit a game, will never do so etc etc)
But here are the files:
Everything is touchable, except the NB, which is damned hot!@

 

[iTitus]

I just figured out I was still running on DirectX 9,0c.
I Updated it now. Will let you guys know

 

[Dave76]

Thanks for the update.

Let us know how it's going.