Memory generating bit-errors (not always) thus BSOD or app crash

[fslateef]

I recently built a machine and from first day I am facing random crashes and issues. I tested everything before and found that my 4x2GB RAM (G.Skill RipjawsX CL7 1600MHz 7-8-7-24, 1.6v) generates random bit-errors in memtest and they are working fine in pairs even later memtest for 9 passes was ok with full 8GB RAM. Then I found that there are some issues with my motherboard (Asus M4A89GTD Pro) and replaced it with new motherboard from Gigabyte (GA-990FX-UD3). For first 4-days with new board the system was stable then when I started gaming (Crysis with extreme settings at 1080p) I faced system freeze.

Then I ran Furmark which is fine and still having game crash/freeze and random application crash or system BSOD. Yesterday on new board I ran memtest86 and found bit-error in pass-6 aroung 6GB RAM address.

At the moment I am running RAM at specs timings and voltage. Which is 1600MHz 7-8-7-24 2T @ 1.6v and CPU is at 3.7GHz with CPU-NB 2.6GHz @ 1.41v.

Now my confusion is: should I try increasing RAM voltage a bit ? Or this is can be due to my power-supply ?

Here's my system specs:

-- Motherboard - GA-990FXA-UD3

CPU - Phenom II X6 1090T CPU 3.7GHz, CPU-NB 2.6GHz, HT 2GHz and CPU VCORE 1.41v; Cooler is: Coolermaster V6GT CPU cooler

-- RAM - 4 x 2GB G.Skill 1600MHz DDR3 7-8-7-24 @1.6v

GPU - Asus Radeon HD 6970 2GB GDDR5
Storage - 500GB WD 7200rpm (16MB cache) and 1TB WD 5400rpm (64MB cache)
Case - Coolermaster CM-690 II Advanced
PSU - Thermaltake XT-775W
Monitor - AOC 2436Vwa (1080p, 60Hz)

Waiting for your response.

Thanks in advance

 

[fslateef]

My exact RAM model is: F3-12800CL7D-4GBRM (2-kits)

Last time the memtest86 error was reported in Test # 7 and Pass # 6.

Can someone help me in this regard ? Is this can be RAM voltage related due to all 4-slots are full ?

Also can anyone tell me how much I can increase RAM voltage from spec/rated which is 1.6v ?

Thanks

Here's my memtest86 screen-shot with error @ 1.6v (rated voltage)

 

[henydiah]

Test again with another slot, test it one by one, remember if you test it must under default, and see it having error ? RMA or get new ram.

 

[fslateef]

Thanks henydiah for replying.

I tested them in pairs and in different slots, they work fine in pairs.

On my previous motherboard, when I first tested all 4-modules together I got some errors then after doing pair testing I tested all 4 modules together for 18-hours and they worked perfectly fine in memtest86 but occasional (once in a day) system crash/BSOD was there.

This time after changing motherboard I faced freezing during games and then I started memtest86 yesterday and got this error. Thats why I am actually confused that might be its a voltage setting issue (as modules are fine in pairs). Because this error and application/system crash is very very occasional; so whenever the memory error happens then it crashes the application if its using that memory OR system if its a system/kernel process.

 

[jaquith]

Test #7 is the CPU IMC errors -> http://www.memtest86.com/tech.html

IMO - first try a tighter CAS 1600MHz DDR3 7-8-7-24 @1.6v ==> 1600MHz DDR3 8-8-8-24 @1.6v. If that fails then start looking at your OC; there's no reason to use HT, Turbo, etc.

edit: the test that you really need to focus on is Prime95 for 24 hours, and then Memtest.

 

[fslateef]

Is this can be due to CPU-NB voltage ? I did overclocking of CPU-NB also thus need adjustment in its voltage? I did increased ddr3 voltage from 1.6 to 1.61v and played Crysis without crash for 1.5 hour and no crash previously I can't run it for more than 30mins.

I will post at gskill forum also to ask about maximum voltage I can go for those modules.

For now I will do prime95 test first and then memtest86 if that is fine, altough previous 5hour test was fine.

Also I think test # 7 is block move which IMO creates stress on RAM thus need some power related issues ?

 

[jaquith]

IMO - You have a CPU OC issue and not a RAM issue. Also again, HT can be left alone for 3.8GHz and your vCore is high 1.36~1.39v should be sufficient.

Nice guide -> http://www.overclockers.com/step-guide-overclock-amd-phenom/

Memory:
Either write-down your OR save your OC Profile in the BIOS. Then set everything to Stock e.g. Load Optimized/Defaults. Make certain you're running BIOS F2; update if needed. Do NOT use Easy Tune6, Cloud OC, AMD Overdrive, Easy Energy Saver, etc - nothing BIOS invasive; set the OC 100% manually there's not much to set.

Further, before OC'ing anything make certain that 'Stock'/'Rated' works FLAWLESSLY. Then start OC the CPU once the RAM works at Rated Frequency.

Using stock CPU - Set & Test your RAM Manually 1600MHz DDR3 7-8-7-24 @1.6v. Run Memtest or whatever testing Apps you have: Prime95, 'Temp App', Memtest, etc.

NB Voltage Control - Auto.
CPU NB VID Control - Auto ; this increases the CPU IMC voltage and 'maybe' a +0.10v increase but is not typically needed for DDR3 1600. Max safe is 1.35v and >1.35v degrades the CPU. IMO Auto or 1.25v.

IF the RAM checks-out then OC your CPU and check your RAM Frequency {Memory Clock}.

 

[fslateef]

Thanks for suggestions. I will follow what you said. Just for info as you said I do overclocking manually.

My BIOS options are weird, means I can't set RAM voltage manually leaving all other voltages AUTO; for changing a single voltage I have to set everything to manual where NORMAL option is present (not AUTO).

Also my CPU voltage was default around 1.475V and I set it to 1.375V in BIOS but still in HWMonitor I see its going to 1.41V. Why this happening ?

For DDR3 1600MHz I read in G.Skill forum that you have to increase frequency from 200MHz to 240MHz and reduce CPU multiplier if you don't need overclocking. Is this correct ? If wrong then isn't default CPU memory-controller is supporting maximum 1333MHz speed ?

At the moment I did some voltage increase of 0.01v in NB voltage and CPU NB VID. RAM is also already on 1.61v which is 0.01v increase. Running Prime95 at the moment and will check for correct voltages in BIOS sometime later. If it crashes then I will go for what you said in the meanwhile it will be good if I can get answer to my above confusion.

Also can you tell me the maximum voltage limits for 1090T ? and CPU-NB and so on ? I just want to know this for safe side

I am really thankful to everyone who are taking time in responding to my issue. Thanks

 

[jaquith]

The CPU Voltage varies on the CPU's 'Load' unless set Manually -- and even then it varies; the 'problems' are vDroop and under/over vCore voltage. A valid vCore reading is when Prime95 is running in the background and the CPU 'is' under 'Full Load.' IF you can obtain 3.7GHz without touching vCore then do it that way. Review both the Black Editions and Thuban Overclocking Help from the link above.

There's no reason for you to OC the FSB unless you had DDR3 1800~2000 RAM; keeping in mind your 6-core 1090T CPU.

The 1090T BE has an unlocked CPU Multiplier and a 3.7GHz is a 'small OC':
3200/200 = 16
3800/200 = 19
1600/200 = 8

BIOS:
Load Optimized
CPU Clock Ratio -> 19 ; 3.8GHz OC
Set Memory Clock -> 8

DRAM Configuration /Enter

DDR3 Timing Items -> Manual
1T/2T Command Timing -> 2T
CAS# latency -> 7
RAS to CAS R/W Delay -> 8
Row Precharge Time -> 7
Minimum RAS Active Time -> 24

DRAM Voltage Control -> 1.60v ; DDR3 has a tolerance of ±0.1~±0.2v

CPU NB VID Control -> Auto OR 1.20~1.25v ; if you get more 'Test 7' errors.

Save & Exit

edit: vCore Voltage - the're no 'Magic' value, EVERY MOBO is different. Ideally the LOWEST STABLE vCore wins! Less = More!

 

[fslateef]

Thanks jaquith for detailed reply

Some updates:

I was running Prime95 with settings in screen-shot below; it was running fine for around 7-hours and then I went to sleep and during night Windows did some update installation and rebooted my machine hence when I woke-up in morning I saw Windows login screen and after logging-in there was no BSOD/dump and got a Windows update notification. So I ran that test Prime95 again and now in office. Will update about its results as soon as reach home.

Your given link is really good _and_ I haven't got enough time yet to read it but will read/follow it asap.


Now have some questions if you don't mind:

1) -- If I run Prime95 (blend) fine for 15-hours or so then can I say my system is stable ? Or 24-hours is minimum ? As it will be difficult for me to keep my system on for 24-hours or more.

2) -- Are the voltage settings in screen-shot below are safe and good for now ? I think these are quite low; right ?

Thanks again for helping.

 

[fslateef]

A update on Prime95 with above settings; its still running after 15-hours or so. Test is still running and will try to run it for at-least 24-hours.

Here's screenshot after 15-hours:

 

[jaquith]

Now have some questions if you don't mind:

1) -- If I run Prime95 (blend) fine for 15-hours or so then can I say my system is stable ? Or 24-hours is minimum ? As it will be difficult for me to keep my system on for 24-hours or more.

2) -- Are the voltage settings in screen-shot below are safe and good for now ? I think these are quite low; right ?

ANS1 - I always allow 24 hours before I certify everything good. Obviously, I saw both of your posts and all looks fine.

ANS2 - If your Temps and system is Stable then clearly you don't need to touch anymore voltage settings. You always have the option of trying them lower. IF your BIOS allows 'Profiles' then save your settings and optionally try other settings; Profile1 = Stable 3.8GHz, Profile2 = Low Voltage 3.8GHz, Profile3 = Crazy 4.4GHz, etc.

I did see where you posted this in other forums, but keep in mind people 'copy' and apply non-BE FSB OC'ing ignoring the advantages aka simplicity of an unlocked CPU multiplier. Frankly, the other settings e.g. Multipliers are not needed - the Auto takes into account the CPU Frequency; some BIOS 'might' not know the 'tables' but I'm pretty certain your GA MOBO BIOS would calculate them correctly.

However, IF you later decide to OC >4GHz then indeed additional settings would be required.

As I suggested all along, there is no 'exact' vCore - every MOBO/CPU/#Phases all has some slight variance, and so in the future just simply remember -> Stable and {Less = More}. Note the variances in the second spreadsheet "P67 Mobo OC Results" - http://www.overclock.net/intel-motherboards/916189-update-official-intel-p6x-motherboard-comparison.html

 

[fslateef]

I did see where you posted this in other forums, but keep in mind people 'copy' and apply non-BE FSB OC'ing ignoring the advantages aka simplicity of an unlocked CPU multiplier. Frankly, the other settings e.g. Multipliers are not needed - the Auto takes into account the CPU Frequency; some BIOS 'might' not know the 'tables' but I'm pretty certain your GA MOBO BIOS would calculate them correctly.

However, IF you later decide to OC >4GHz then indeed additional settings would be required.

As I suggested all along, there is no 'exact' vCore - every MOBO/CPU/#Phases all has some slight variance, and so in the future just simply remember -> Stable and {Less = More}. Note the variances in the second spreadsheet "P67 Mobo OC Results" - http://www.overclock.net/intel-motherboards/916189-update-official-intel-p6x-motherboard-comparison.html

I understood everything and will keep that in mind. I posted this (not same post) in one other forum and then I marked that as solved as now I am satisfied with tomshardware forum. Can you tell me where you saw this posted by me ? Just asking due to curiosity. Because I haven't posted details like this any where else.

 

[jaquith]

It's no big deal -> http://www.overclockers.com/forums//showthread.php?p=6911603 I agree with @Mjolnir to a point but I know 3.7GHz~3.8GHz can be run just fine 'simply' on a 1090T BE.

I often search before I post my first response to see if others reported a similar issue -- in your case attributable to your GA-990FXA-UD3, but I am also aware of Test #7. Once in a while I see the test #7 error problems and most of the time it's either a CPU OC or >DDR3 1600 frequency e.g. 1866 or faster.

 

[fslateef]

It's no big deal -> http://www.overclockers.com/forums//showthread.php?p=6911603

Now I realised what you were referring too, but isn't my that post and this one is completely different ? That one was related to my CPU cooler where my CPU was heating even @ stock speed and the final problem was cooler's pull/push fans direction _not_ memory issue. I know I discussed about my CPU voltages there but at that time I was unable to do any overclocking.

Anyways Prime95 was fine after 20hours and then again the windows updates restarted machine which is annoying so failed to run prime95 for 24 hours.

Now will try to use system normally and gaming and if it crashes then will run memtest86+ v4.20. Will post back about it status soon.

Thanks for bearing with me and help.

Edit: I started this discussion basically due to my RAM issue and it ended up in overclocking and voltage settings thing (which happens might be due to the fact that my RAM was/is behaving incorrectly due to wrong overclock settings). Thats why its similar to that thread

 

[jaquith]

Yes and No, as I stated the Memtest errors IMO are coming from an OC {CPU} issue; test #7 is the CPU's IMC. You hit the 'nail on the head' with your 'Edit:' comments.

I felt for a 3.7~3.8GHz that the BIOS is/was a little over zealous.

Your Prime95 is good - focus on the Memstest but with any CPU OC don't be too surprised about #7 errors.

In addition, look at your Event Manager for the prior Critical errors -- the BSOD could have been caused by a bad driver.

 

[fslateef]

I understand that test#7 failure can happen due to CPU IMC.

Now if I get BSOD then what to do ? Let say memtest86+ shows some errors then I again have to set CPU IMC related parameters or can say that my CPU or RAM is bad ?

I also don't have any external driver installed in my fresh Windows installation besides ATI catalyst 11.6 latest. Means all drivers are from Microsoft and only external application is Crysis. Then the crash if happens then it will be hardware fault.

 

[jaquith]

BSOD, as I mentioned -- look at the Event Manager.

Drivers, use those listed at GA's site, especially the Chipset -- http://www.gigabyte.com/products/product-page.aspx?pid=3894&dl=1#driver

I seriously doubt a (1) obscure IMC error is the core problem with your BSOD(s).

Q - What listed Critical errors are contained in the Event Manager as show above?

 

[fslateef]

I have yet to see any crash. Due to prime95 I unable to run any game which I will do today and let you know if something happens.

Also at the moment I am at job thus can't access my machine. I will install drivers from GA website _but_ AFAIR the chipset driver on GA website is from AMD and they are still 8-series.

I am really taking you too much time

 

[jaquith]

Interesting, yep I looked at all of the Gigabyte chipset drivers. ASUS posted (WHQL) 3.0.816.0/3.0.812.0 2011.05.26, but Gigabytes are all 2010/04/12 even for the UD7 line.

So in that case, I take it back use the 'latest' drivers installed now.

Problem, if GA's BIOS is geared towards 'older' drivers then once the 'new' drivers are available then IMO look for a corresponding 'new' BIOS that will accommodate the 'correct & new' {edit}>chipset drivers. This could be your problem.

 

[fslateef]

Update:

One interesting thing is: my system is seems stable at all the settings I mentioned above about overclocking like 3.7GHz, CPU-NB 2.6GHz with those specific voltages. I haven't saw any crash in Prime95, running lot of other applications along-with prime95 at the same time, and playing Crysis for around an hour with everything max @ 1080p.

But when I tried to play with some more settings like increasing 100MHz in CPU frequency or reducing CPU-NB voltage a bit and increasing CPU vcore a bit; everytime I saw a BSOD OR application crashes or fails to run (while running Prime95).

So somehow my current settings are most stable and balanced and this is really strange. Might possible the BIOS update will fix something as its not tolerating a small bit of change even 100MHz increase in CPU speed

 

[jaquith]

Define How you changed -> "increasing 100MHz in CPU frequency"?

Yeah, you need a BIOS update and Driver update as soon a GA adds proper AMD Chipset drivers to its' download page. I cannot guarantee it will 'fix' an OC/BSOD issue - it depends upon the root cause and IF the BIOS address the problem(s).

 

[fslateef]

Define How you changed -> "increasing 100MHz in CPU frequency"?

Yeah, you need a BIOS update and Driver update as soon a GA adds proper AMD Chipset drivers to its' download page. I cannot guarantee it will 'fix' an OC/BSOD issue - it depends upon the root cause and IF the BIOS address the problem(s).

I just changed the multiplier from 18.5 to 19 ........ Also tried setting CPU NB VID to NORMAL and increasing CPU voltage. Nothing worked I am happy with what I have for now but just seems weird thats why mentioning that.

 

[fslateef]

I don't want to increase CPU-NB from 1.1.75v as from Thermal and Power Document the MAXIMUM CPU-NB-VID voltage is 1.175v for 1090T (part-number). Is this correct ?

 

[jaquith]

The CPU NB VID Control can safely go to 1.30v without a problem; it's for the CPU IMC control voltage. As far as being 'correct' @1.175v perhaps with DIMM standard of 1.50v then yes, but that's not the max safe voltage. The increase was to eliminate/reduce the Test #7 IMC errors in conjunction with your 'DIMM 1.6v' RAM; the disparity between voltages {delta} may have been the problem for the errors.

Q - What happens if all you do is enter the CPU Clock Ratio and set the RAM? Without all the other Multipliers/Voltages/etc. The 1090T 'Turbo frequency' is 3.6GHz so a +0.2GHz still should be okay with 'stock' settings; all dependent on your MOBO of course.

BIOS:
Load Optimized
Core Performance Boost -> Disabled
CPU Clock Ratio -> 19 ; 3.8GHz OC
Set Memory Clock -> 8

DRAM Configuration /Enter

DDR3 Timing Items -> Manual
1T/2T Command Timing -> 2T
CAS# latency -> 7
RAS to CAS R/W Delay -> 8
Row Precharge Time -> 7
Minimum RAS Active Time -> 24

DRAM Voltage Control -> 1.60v ; DDR3 has a tolerance of ±0.1~±0.2v

CPU NB VID Control -> Auto OR 1.20~1.25v