Numerous RAM Errors

[MisterEd]

MBD: Gigabyte GA-990FXA-UD3
CPU: Phenom II X4 955
RAM: GSkill RipjawsX F3-12800CL6D-4GBXH (2x2GB)
Note: RAM is running at 1333 MHz

My computer worked fine for 6 months but slowly started crashing more and more. I tested the RAM with Memtest86+ with RAM in slots 1 & 2. Over 32000 errors were detected. Testing in slots 3 & 4 resulted in Memtest86+ crashing after only a few minutes.

Each stick was tested singly in slot 2 with no errors. Testing a single stick in slots 1, 3, or 4 gave inconsistent results. Sometimes the test passed while other times errors were found.

I might add that the CPU came from another computer that was struck by lightning a year ago. The other computer had 2 sticks of DDR2 RAM of which 1 stick failed after the lightning strike.

I was planning swapping out components in the following order to see what is causing the problem:
1. RAM
2. CPU
3. Motherboard

Any suggestions?

BTW, my existing RAM is not on any recommended list for this motherboard. Is there any possibility that this RAM is not stable because of that?

 

[Psykoikonov]

It would help us help you if you posted CPUZ screenshots of the Memory and SPD tab, the Processor tab wouldn't hurt either. There is a lot of Ram on the market so it not being on the recommended list means it may have not been tested. Looking at the manufacturers site the Ram has tight timings 6-8-6-24, maybe too tight for the 955. What is puzzling is that it was ok and now it's not. Either something has happened recently or the memroy errors have slowly wreaked havoc on the OS.

 

[MisterEd]

It would help us help you if you posted CPUZ screenshots of the Memory and SPD tab, the Processor tab wouldn't hurt either. There is a lot of Ram on the market so it not being on the recommended list means it may have not been tested. Looking at the manufacturers site the Ram has tight timings 6-8-6-24, maybe too tight for the 955. What is puzzling is that it was ok and now it's not. Either something has happened recently or the memroy errors have slowly wreaked havoc on the OS.

 

[Psykoikonov]

Well, I expected timings were set too tight, but most (the important ones) are right on. Anyway I would try setting the timings manually to JEDEC#2 (8-8-8-25-33) 1T or 2T shouldn't really matter. This is done in the bios of the motherboard, press the delete key during POST (initial start up of computer), press enter on the first selection (M.I.T.) and scroll to DRAM configuration. There you will find the timings of the RAM. In order I would set 2T 8 8 8 25 Auto Auto Auto Auto Auto 33 rest Auto. Also be sure on the previous page that the DRAM voltage is picked up correctly at 1.50V, if not manually set it. Reduce bus speed to 200 as opposed to 201.

If that produces any errors with memtest86+ then I would drop the speed of the ram to 1066 keeping the same timings (8-8-8-25-33) and/or swap the Ram and see if the lightning struck CPU's cache is at fault (I'm doubtful of this, shouldn't have taken 6 months to show up).

 

[trents]

Sounds to me like slots 1,3 and 4 on the motherboard are bad.

 

[Psykoikonov]

Sounds to me like slots 1,3 and 4 on the motherboard are bad.

+1 Could be but doesn't explain the 6 months of working.

 

[trents]

+1 Could be but doesn't explain the 6 months of working.

Things go bad. It must happen at a certain point in time.

 

[larrymoencurly]

I don't think the memory went bad after 6 months. I'd first look for any new software installed, even updates, because they cause the memory to be used differently and can reveal bad bits that were there all along. Because of the lightning strike, you may have one of the few CPUs with a defect in it, but I still doubt it's bad. Still, the only way to rule out the CPU or motherboard is by testing with known good memory (my usual rant: modules with chips easily identifiable by brands like Samsung, Elpida, Hynix, Micron, etc.), rather than with modules that look Nike shoes.

 

[MisterEd]

Well, I expected timings were set too tight, but most (the important ones) are right on. Anyway I would try setting the timings manually to JEDEC#2 (8-8-8-25-33) 1T or 2T shouldn't really matter. This is done in the bios of the motherboard, press the delete key during POST (initial start up of computer), press enter on the first selection (M.I.T.) and scroll to DRAM configuration. There you will find the timings of the RAM. In order I would set 2T 8 8 8 25 Auto Auto Auto Auto Auto 33 rest Auto. Also be sure on the previous page that the DRAM voltage is picked up correctly at 1.50V, if not manually set it. Reduce bus speed to 200 as opposed to 201.

If that produces any errors with memtest86+ then I would drop the speed of the ram to 1066 keeping the same timings (8-8-8-25-33) and/or swap the Ram and see if the lightning struck CPU's cache is at fault (I'm doubtful of this, shouldn't have taken 6 months to show up).

Changing the timings did not solve the problem. The test ran for 3-1/2 minutes without error but locked up during "Test 5 [Block move, 64 moves]". It was at 20% completion and 90-95% of test 5 when everything froze.

I reran Memtest86+ with the memory clock at 1066MHz. 3 passes were completed without error before I aborted the test.

I should have mentioned that I had been folding with this computer for the last 4 months. Maybe the stress on the computer caused something that was marginal before to finally start to fail. During the folding the CPU temp was about 45C.

The CPU has been questionable for a long time. Hardware thermal control has to be disabled because turning it on causes the CPU speed to be stuck on 800MHz all the time. I also had to tame the CPU temp with water cooling because with stock cooling the CPU idle temp was 45C and load temp was 70-80C. Now the idle temp is 30C and load temp is 45C.

 

[Psykoikonov]

I reran Memtest86+ with the memory clock at 1066MHz. 3 passes were completed without error before I aborted the test.

I think that the Ram was never capable of running at rated speed with that processor and motherboard (should have been RMAed at purchase), and it took 6 months of RAM errors to corrupt enough of the OS and software files to show the damage. With the new Ram settings of 1066 8-8-8-25-33 or some new RAM (tested) and a fresh install of the OS will cure the problems without recurrence.

 

[redduc900]

Does bumping either the vNB or NB VID voltage, or even the Vdimm from the default of 1.5V to 1.55-1.60V help at all?

 

[MisterEd]

I think that the Ram was never capable of running at rated speed with that processor and motherboard (should have been RMAed at purchase), and it took 6 months of RAM errors to corrupt enough of the OS and software files to show the damage. With the new Ram settings of 1066 8-8-8-25-33 or some new RAM (tested) and a fresh install of the OS will cure the problems without recurrence.

The computer showed no stability problems until recently. What corrupted OS and software damage? I am not saying there is none. I am just saying there is no evidence of it yet. Slowing the RAM to 1066 has resulted in no stability problems any more. Before I could not go more that a few hours without the computer crashing. However I will not make any final judgements until I have a chance to see if any problems crop up in the next few weeks.

I bought the motherboard when it first came out. There was not much information available then on what memory worked with it. I bought what was available from the local dealer that I bought the motherboard from. Now the dealer has other RAM available known to work with it. He has following:
Corsair CMX8GX3M2A1333C9 8GB Kit (2x4GB)
Corsair TW3X4G1333C9A 4GB Kit (2x2GB)
Crucial CT2KIT102464BA1339 16GB kit (8GBx2)
Crucial CT2KIT51264BA1339 8GB Kit (4GBx2)
Crucial CT2KIT25664BA1339 4GB kit (2GBx2)
When I can I will buy some more RAM to see if that solves the problem.

BTW, I bought the motherboard planning to buy an AMD FX CPU for it. Since no FX appears to be much better than what I already have I am waiting to see if AMD's next generation of CPU is much better.

Does bumping either the vNB or NB VID voltage, or even the Vdimm from the default of 1.5V to 1.55-1.60V help at all?

Thanks I'll try that.

 

[Psykoikonov]

MBD: Gigabyte GA-990FXA-UD3
My computer worked fine for 6 months but slowly started crashing more and more.

The computer showed no stability problems until recently. What corrupted OS and software damage? I am not saying there is none. I am just saying there is no evidence of it yet.

The first statement made me think that the OS was now destroyed, seen it before from bad Ram or running Ram too fast/too tight.

Regarding Ram there is so much of it on the market that motherboard manufacturers lack time to test it all. If the Ram runs without errors at 1066 and what you use the computer for doesn't require slightly faster Ram or more of it then I personally I wouldn't buy any.

However if you were looking for some decent Ram at a good price I would buy G.SKILL Ares Series 8GB (2 x 4GB) 240-Pin DDR3 SDRAM DDR3 1866 (PC3 14900), it isn't on the motherboard list but it should work perfectly with your current setup and with a CPU upgrade down the Road. If you do pick up some Ram be sure to test it.

Within windows a better test for Ram is Memtest. The aforementioned program has found errors where memtest86+ found none.

 

[MisterEd]

Does bumping either the vNB or NB VID voltage, or even the Vdimm from the default of 1.5V to 1.55-1.60V help at all?

I decided to go back to the original timings (8-8-8-24-31-1T) and try adjusting the DRAM Voltage Control. Upping the voltage reduced the errors with MemTest86+. Most if not all the errors came in Test 5 (Block Move).

1.5V - ~90,000 errors
1.65v - ~60 errors
1.695V - 0 errors

Can I leave the voltage at 1.695 or should I back it off to the lowest stable value above 1.65?

 

[Psykoikonov]

I'll just quote from the wiki for JEDEC standard on DDR3.

According to JEDEC[2], 1.575 volts should be considered the absolute maximum when memory stability is the foremost consideration, such as in servers or other mission-critical devices. In addition, JEDEC states that memory modules must withstand up to 1.975 volts before incurring permanent damage, although they are not required to function correctly at that level.

 

[redduc900]

What were the results of bumping the vNB and / or the CPU VID voltage?

 

[MisterEd]

I spoke to soon. The errors have returned.

I just got some new RAM. Corsair TW3X4G1333C9A 4GB Kit (2x2GB). It is on the tested list for this motherboard.

I reran memtest86+ with the memory settings on auto. Even the new RAM had over 60 errors halfway through the first pass. This is a lot less than the other RAM which had over 90,000 errors with the same settings.

I guess I will try a new CPU next.

 

[redduc900]

I'm trying to help you, but if you don't answer / ignore my questions I'm afraid I can't be of much help.

 

[MisterEd]

I'm trying to help you, but if you don't answer / ignore my questions I'm afraid I can't be of much help.

Thanks for all your comments. I tried adjusting the voltages and timings but nothing solved the problem.

I am taking online college courses right now and do not want to tear the computer apart right now. For now I have dropped the memory speed to 1066. This at least appears to be stable. When the term is over in a few weeks I will have time to do something more drastic.

I need to determine what is causing the problem for sure and act on that. When I have the time I am going to pull the CPU out of another computer and try it out on this one. That will prove whether the CPU is the problem. If that is not the problem then I will have to determine whether the motherboard is at fault.

 

[larrymoencurly]

I spoke to soon. The errors have returned.

I just got some new RAM. Corsair TW3X4G1333C9A 4GB Kit (2x2GB). It is on the tested list for this motherboard.

I reran memtest86+ with the memory settings on auto. Even the new RAM had over 60 errors halfway through the first pass. This is a lot less than the other RAM which had over 90,000 errors with the same settings.

I guess I will try a new CPU next.

I don't think it's the CPU. The fact the error count went down as the voltage went up proves that the memory was marginal, and you really need to try some known good non-overclocked memory chips and quit messing around with junk from G.Skill, Corsair, or any company that sells heatsinked modules with house branded chips on them.

Just wondering, but why did you go from G.Skill to Corsair instead of to known good memory?