GA 790FXTA-UD5 -> System Unstable -> Testing Components -> CPU/RAM Errors

[thx1138]

I recently built the system outlined in my signature but have suffered occasional random restarts while playing Left 4 Dead 2. I've played Fallout 3 and experienced crashes but they could well be coding errors (Windows continued to run underneath).

So far, errors have been found when testing the CPU and also the RAM, but I'm not sure that you tell that one hasn't caused an error in the other. Indeed, perhaps they are both faulty.

CPU tests:
CPU: LINPACK - ERROR after 13 minutes (max temp. 63 C)
CPU: OCCT - No error after 1 hr at default settings (max temp. 64 C)
CPU: OCCT - ERROR after five minutes at Test Mode: Large data set (max temp 64 C)

GPU tests:
GPU: OCCT - No error after 1 hr (I see no temperature readout for my card)
Furmark - No error reported after 4 hrs (I'm not sure how Furmark reports errors or whether it detects artifacts; max temp. 81 C)

RAM tests:
Memtest86+
WallTime: 12:33:59 - Cached: 4087M - RsvdMem: 12K - MemMap: e820 - Cache: on - ECC: off - Test: Std - Pass: 14 - Errors: 107 - ECC Errs: 0

Test: 4 - Pass: 0 - Failing address: 000bd7f3f3c - 3031.9MB - Good: fbee1d66 - Bad: fbee1d67 - Err-Bits: 00000001 - Count: 1 - Chan: [BLANK]

with settings: RAM: 800 MHz (DDR1600) / CAS: 11-11-11-24 / DDR3 (64 bits)
The memory is being run at its rated clock speed with lower-than-rated timings (it is supposed to be 7-7-7-24). The voltage supplied is 1.65.

Notice that the Errors readout is 107 but only one error is elaborated upon below it. The Errors readout is increasing with time.

Edit: tested again with settings: RAM: 533 MHz (DDR1066) / CAS: 7-7-7-16 / DDR3 (64-bits) @ 1.5 volts with result:
WallTime: 1:22:59 - Cached: 4087M - RsvdMem: 12K - MemMap: e820 - Cache: on - ECC: off - Test: Std - Pass: 1 - Errors: 4 - ECC Errs: 0

Test: 4 - Pass: 0 - Failing address: 000bd7f3f3c - 3031.9MB - Good: 2a9dc386 - Bad: 2a9dc387 - Err-Bits: 00000001 - Count: 1 - Chan: [BLANK]

Do I have problematic settings, CPU or RAM? Should I obtain an RMA on either component? Please advise.

[nicolatesla]

I recently built the system outlined in my signature but have suffered occasional random restarts while playing Left 4 Dead 2. I've played Fallout 3 and experienced crashes but they could well be coding errors (Windows continued to run underneath).

So far, errors have been found when testing the CPU and also the RAM, but I'm not sure that you tell that one hasn't caused an error in the other. Indeed, perhaps they are both faulty.

CPU tests:
CPU: LINPACK - ERROR after 13 minutes (max temp. 63 C)
CPU: OCCT - No error after 1 hr at default settings (max temp. 64 C)
CPU: OCCT - ERROR after five minutes at Test Mode: Large data set (max temp 64 C)

GPU tests:
GPU: OCCT - No error after 1 hr (I see no temperature readout for my card)
Furmark - No error reported after 4 hrs (I'm not sure how Furmark reports errors or whether it detects artifacts; max temp. 81 C)

RAM tests:
Memtest86+
WallTime: 12:33:59 - Cached: 4087M - RsvdMem: 12K - MemMap: e820 - Cache: on - ECC: off - Test: Std - Pass: 14 - Errors: 107 - ECC Errs: 0

Test: 4 - Pass: 0 - Failing address: 000bd7f3f3c - 3031.9MB - Good: fbee1d66 - Bad: fbee1d67 - Err-Bits: 00000001 - Count: 1 - Chan: [BLANK]

with settings: RAM: 800 MHz (DDR1600) / CAS: 11-11-11-24 / DDR3 (64 bits)
The memory is being run at its rated clock speed with lower-than-rated timings (it is supposed to be 7-7-7-24). The voltage supplied is 1.65.

Notice that the Errors readout is 107 but only one error is elaborated upon below it. The Errors readout is increasing with time.

Edit: tested again with settings: RAM: 533 MHz (DDR1066) / CAS: 7-7-7-16 / DDR3 (64-bits) @ 1.5 volts with result:
WallTime: 1:22:59 - Cached: 4087M - RsvdMem: 12K - MemMap: e820 - Cache: on - ECC: off - Test: Std - Pass: 1 - Errors: 4 - ECC Errs: 0

Test: 4 - Pass: 0 - Failing address: 000bd7f3f3c - 3031.9MB - Good: 2a9dc386 - Bad: 2a9dc387 - Err-Bits: 00000001 - Count: 1 - Chan: [BLANK]

Do I have problematic settings, CPU or RAM? Should I obtain an RMA on either component? Please advise.

You are overheating your components
If this is caused by the "test" you mentioned, I recommend you stop testing it...Is this one of those computer stress tests?
In any case, 64 C on the processor and 81C on the Furmark is medium well indeed!
There is a severe case ventilation problem?
The fans are not running?
Try taking the side off the computer case and run it
Whats the temp now?
(I don't run the stress test software, I don't recommend you do that.)
Over heating will severely shorten the life of the components and destroy them eventually.
You need front intake and rear exhaust fans, power supply fan, large CPU cooler...to run what you are trying to run, or maybe just a much better case with much better fans?

[thx1138]

I am running the tests to determine the source of the problem. I will assess whether the errors occur with the case open and a desk-fan directed at the inside but I do not believe that the CPU or GPU are overheating. For a 125W processor with stock HSF, 64 C is not unreasonable. I would be concerned only as it approached 80 C. The GPU on the other hand, should be able to handle temperatures in excess of 100 C. Nothing here is being operated outside of spec. I do not have a temperature reading for the RAM, but the tests show errors even when it is underclocked.

[absic]

Not sure if this will help but I have encountered similar problems with a recent build using the GA-790XTA-UD4 Mobo. Memory = Corsiar Dominator CMD4GX3M2B1600C8. CPU = Phenom ii 965 (145W) C2 Stepping.

When trying to run this rig with memory set to specified ratings the system was very unstable. However letting bios handle the settings (default/auto) and with the memory downgraded from 1600 to 1333 Mhz the sytem passes all tests and is now running well.

It is very frustrating downgrading memory speed but, so far, I have found no other solution to this problem despite hours of searching and tweaking.

As a starting point I would recommend using the default settings in bios and seeing how the rig runs before trying to blame any of the hardware. These boards are new and I am sure that these memory speed issues will be resolved with Bios updates in the future.

[nicolatesla]

I am running the tests to determine the source of the problem. I will assess whether the errors occur with the case open and a desk-fan directed at the inside but I do not believe that the CPU or GPU are overheating. For a 125W processor with stock HSF, 64 C is not unreasonable. I would be concerned only as it approached 80 C. The GPU on the other hand, should be able to handle temperatures in excess of 100 C. Nothing here is being operated outside of spec. I do not have a temperature reading for the RAM, but the tests show errors even when it is underclocked.

It is unfortunate that people have been led to believe that overheating their computers with "stress tests" will locate the source of a problem.
whoever told you that a GPU would handle excess of 100C (212F) is living in a dream world.
It may be true that carefully heating or cooling individual components (one at a time) under controlled conditions and monitoring with test equipment, a technician may find a thermally intermittent (thermal failure) component. But just heating the unit past design maximum tells you what? "When we roast it, it fails?"
My 125W processor says 62C maximum. I run it 42C maximum, even when it's way loaded up. I would not expect it to run at the temperatures you are measuring, and I would not expect the components to remain in a reliable condition.

[thx1138]

Not sure if this will help but I have encountered similar problems with a recent build using the GA-790XTA-UD4 Mobo. Memory = Corsiar Dominator CMD4GX3M2B1600C8. CPU = Phenom ii 965 (145W) C2 Stepping.

When trying to run this rig with memory set to specified ratings the system was very unstable. However letting bios handle the settings (default/auto) and with the memory downgraded from 1600 to 1333 Mhz the sytem passes all tests and is now running well.

It is very frustrating downgrading memory speed but, so far, I have found no other solution to this problem despite hours of searching and tweaking.

As a starting point I would recommend using the default settings in bios and seeing how the rig runs before trying to blame any of the hardware. These boards are new and I am sure that these memory speed issues will be resolved with Bios updates in the future.

Hey, thanks for the info. Unfortunately, I'm running Memtest86+ now with the RAM underclocked to the motherboard's default settings and I'm still getting errors at the same address. This time I've got a desk fan directed into the case and all components are not too hot to touch. I'm going to return/exchange the RAM and see if my experience gets any better. If I'm right, I will only get errors for one of those sticks of RAM, so I'm going to test them individually, to be sure.

WallTime: 8:41:13 - Cached: 4087M - RsvdMem: 12K - MemMap: e820 - Cache: on - ECC: off - Test: Std - Pass: 10 - Errors: 26 - ECC Errs: 0

Test: 6 - Pass: 0 - Failing address: 000bd7f3f3c - 3031.9MB - Good: 00000010 - Bad: fbee1d67 - Err-Bits: 00000011 - Count: 1 - Chan: [BLANK]

I am running the tests to determine the source of the problem. I will assess whether the errors occur with the case open and a desk-fan directed at the inside but I do not believe that the CPU or GPU are overheating. For a 125W processor with stock HSF, 64 C is not unreasonable. I would be concerned only as it approached 80 C. The GPU on the other hand, should be able to handle temperatures in excess of 100 C. Nothing here is being operated outside of spec. I do not have a temperature reading for the RAM, but the tests show errors even when it is underclocked.

It is unfortunate that people have been led to believe that overheating their computers with "stress tests" will locate the source of a problem.
whoever told you that a GPU would handle excess of 100C (212F) is living in a dream world.
It may be true that carefully heating or cooling individual components (one at a time) under controlled conditions and monitoring with test equipment, a technician may find a thermally intermittent (thermal failure) component. But just heating the unit past design maximum tells you what? "When we roast it, it fails?"
My 125W processor says 62C maximum. I run it 42C maximum, even when it's way loaded up. I would not expect it to run at the temperatures you are measuring, and I would not expect the components to remain in a reliable condition.

There's no way I can air-cool my processor to 40-something Celsius at full load with an ambient temperature of ~20 C, using the stock HSF unless i underclock it. If you aren't underlclocking the CPU, you're either not using the stock HSF (I think the warranty of the processor depends on using the stock HSF) or you provide the computer with exceedingly cold air. The purpose of a stress test is not to overheat a component. Heating is just a byproduct of doing what it's supposed to do.

[thx1138]

With my new RAM, I also got errors when running it at full speed. I test each module individually, and each suffered no errors. I then put both back in again, and they run smoothly together, with no errors (Memtest86+ run for about 12 passes). Thus, after all that, my memory is compatible with this motherboard.