Stability testing my q6600 build

[KingGheedora]

q6600
Abit ip35 pro
2 4GB kits G.SKill DDr2-800 (8GB total)
Corsair HX620
Vista Enterprise 64 bit

I can't get the system stable at stock speeds. I tried running only 4GB of ram at a time and whenever I run dual orthos blend and leave the system, I come back to a windows login screen. I am using the stock retail Intel cooler, and not overclocking anything. I've tried upping ram voltage and MCH voltage a little, but doesn't seem to help.

System is a little more stable with the RAM running much lower than 800 but I haven't had time to do really thorough tests.

What would cause these reboots?

 

[SerpentRoyal]

Use CPUz to confirm that RAM is running at 400MHz. Run Memtest86 test #5 for two hours.

 

[KingGheedora]

I did use CPUz to confirm, and memtest doesn't find any errors. After hours of running test 5 & 8. Orthos causes the crashes though.

 

[Acanthus]

Originally posted by: KingGheedora
I did use CPUz to confirm, and memtest doesn't find any errors. After hours of running test 5 & 8. Orthos causes the crashes though.

Are there any dumps? Vista will typically generate one when there is a hardware failure.

 

[CurseTheSky]

My guess is power supply, overheating at load temperatures, or bad memory.

First, try installing just ONE stick of memory, in the slot closest to the processor. Don't worry about using dual channel kits for now. If one fails, try a different stick. Try all four sticks until all fail, or one (or multiple) work. If one seems to be dead, ask for more advice about them here, then RMA.

If that doesn't work, check your temperatures as you run Orthos. At stock speeds, depending on your climate, I'd hope to see idle temperatures of 25-40C, and load of 35-60C. Any more than that, and you may need to reinstall the heatsink.

Finally, see if you can find another power supply to test with. The crashes at load may be due to fluctuations in power; your PSU may not be putting out enough juice for everything, or may be dying.

 

[SerpentRoyal]

I'd swap the MB. Bad CPU is very rare. Also make sure that there is adequate air flow in the case. You should put the MB on the shipped cardboard box and retest after flashing to BIOS 11.

 

[Mr Fox]

Originally posted by: KingGheedora
q6600
Abit ip35 pro
2 4GB kits G.SKill DDr2-800 (8GB total)
Corsair HX620
Vista Enterprise 64 bit

I can't get the system stable at stock speeds. I tried running only 4GB of ram at a time and whenever I run dual orthos blend and leave the system, I come back to a windows login screen. I am using the stock retail Intel cooler, and not overclocking anything. I've tried upping ram voltage and MCH voltage a little, but doesn't seem to help.

System is a little more stable with the RAM running much lower than 800 but I haven't had time to do really thorough tests.

What would cause these reboots?

Set the O/S like this :


Control Panel/System/Advanced System Settings/Start-up and Recovery/Settings

Un-check the "Automatically restart" under system failure. still have it write an event to the log.

Ok it and apply it on your way out.

That will force it to BSOD and Dump.


I'm going to guess that 8 Gigs is way too much.... and you might not get it all to play with each other.

 

[KingGheedora]

I will try some of the above. But the problem is not the 8GB, as I cannot get stable even with only 4GB installed.

 

[cmdrdredd]

Try giving your Northbridge more voltage. That is the MCH chipset in the BIOS. Bump it up to around 1.37 which is perfectly fine. Also make sure your memory uses the same IC or is very similar. Keep your voltage at the recommended value and set your timings correctly.

 

[Mr Fox]

Originally posted by: KingGheedora
I will try some of the above. But the problem is not the 8GB, as I cannot get stable even with only 4GB installed.

Most MOBOS do not like all 4 slots occupied... some Chipsets are more friendly to it than others....

We also need to know more about what Timings are you running voltages etc....

Also are all your updates done ? The last two updates have made Vista x64 much more stable.

Stability comes from 4-5 things, and they all have to be right.

 

[KingGheedora]

All updates are done for Vista, that was the first thing I did. Latest drivers for everything.

I understand about the mobos and the 4 slots, but like I said, I am testing with only 4gb installed (2gb per slot), and it's unstable.

Default timings are being used: 5-5-5-15, 2T. I verified this with cpu-z.

Voltages are all default (for this ram it's 1.8v) , but I've tried upping the DIMM voltage to 1.9v, 2.0, and 2.1v.

I also got the same reboots with running stock but lowering the mem ratio so the ram runs at the next setting below DDR2-800.

 

[KingGheedora]

Originally posted by: cmdrdredd
Try giving your Northbridge more voltage. That is the MCH chipset in the BIOS. Bump it up to around 1.37 which is perfectly fine. Also make sure your memory uses the same IC or is very similar. Keep your voltage at the recommended value and set your timings correctly.

What do you mean by this?

 

[KingGheedora]

Originally posted by: Mr Fox
Set the O/S like this :


Control Panel/System/Advanced System Settings/Start-up and Recovery/Settings

Un-check the "Automatically restart" under system failure. still have it write an event to the log.

Ok it and apply it on your way out.

That will force it to BSOD and Dump.


I'm going to guess that 8 Gigs is way too much.... and you might not get it all to play with each other.

Fox, I did this, and started up Orthos blend with pretty much stock settings except RAM running slower than stock (only 4GB installed). After just 12 minutes I got a BSOD. The error had to do with uguru.sys. I had been getting error messages every time I opened uguru, and had trouble with one of the installs of it and had to install twice.

I found some forum posts via Google, of a guy who was also having random reboots of his ip35 pro. Another poster said to delete the \windows\system32\wbem\repository\ folder b/c the error had to do with WMI service, and he claimed the WMI repository was corrupt.

After doing this the repository rebuilt itself and I've been running Orthos blend for over two hours with no issues. I can also use uguru without errors, so things are looking better now. I'll update after more testing.

EDIT: Still getting the reboots. After three hours of Orthos I changed fsb to 400MHz, RAM ratio to 1:1 and multiplier to 6, and am still getting reboots after a few minutes to dozens of minutes of orthos. WTF. Damn my impatience, it takes so long to test for stability it's hard to be methodical.

 

[KingGheedora]

Still testing with only 2 x 2GB installed...

I was able to run Orthos blend for 3 hours at 333FSB / 400MHz RAM / 2.4GHz CPU and stock voltages. It didn't crash at this setting but I stopped the tests manually.

I get crashes with Orthos Blend when I change to 400MHz FSB / 400MHz RAM / 2.4GHz CPU, but at these same settings Orthos Small FFT ran for 15 hours -- what would this indicate is wrong? I did bump up the MCH voltage by two steps and that didn't seem to get the system stable. I didn't see any need to increase the DIMM voltage in this case because the RAM was stable at this voltage at 400MHz when the FSB was at 333. Does this mean it's not the ram, that it's the board?

Even after following Fox's advice, and setting the system to not reboot, the machine reboots. Could it be one of the chipset chips overheating? I read that the heatpipe coolers on these boards are not seated to make good contact with the chips. Is 400MHz FSB a lot to expect? I thought I'd seen people running at 450 and 500FSB.

 

[Mr Fox]

Originally posted by: KingGheedora
Still testing with only 2 x 2GB installed...

I was able to run Orthos blend for 3 hours at 333FSB / 400MHz RAM / 2.4GHz CPU and stock voltages. It didn't crash at this setting but I stopped the tests manually.

I get crashes with Orthos Blend when I change to 400MHz FSB / 400MHz RAM / 2.4GHz CPU, but at these same settings Orthos Small FFT ran for 15 hours -- what would this indicate is wrong? I did bump up the MCH voltage by two steps and that didn't seem to get the system stable. I didn't see any need to increase the DIMM voltage in this case because the RAM was stable at this voltage at 400MHz when the FSB was at 333. Does this mean it's not the ram, that it's the board?

Even after following Fox's advice, and setting the system to not reboot, the machine reboots. Could it be one of the chipset chips overheating? I read that the heatpipe coolers on these boards are not seated to make good contact with the chips. Is 400MHz FSB a lot to expect? I thought I'd seen people running at 450 and 500FSB.

That was where I was going to direct you next, and when you set your memory do it in the BIOS... get rid of the Software Crapola ... PERIOD

pull the board... and pull the Chipset cooling stuff and replace with whatever thermal goop you have.. be sparing when you put them back on the mobo make sure that you get them on there straight, and square to the chipsets.

Set your Memory Voltage to 2.0 volts see if that helps.. Timings set manually 5-5-5-15 for now... No Auto or SPD...

If this does not fix it you might have some bad Memory.... but it is still too soon to tell

You are doing the right thing as you have narrowed it to two items.. heat or hardware..

 

[Acanthus]

Uguru.sys is the Abit software overclocking app.

I wouldnt use it, it sucks anyway.

Modifying settings in the BIOS is always better, regardless of the brand/software.

 

[KingGheedora]

I wasn't using uguru anyways, I thought the software was crap the first time i fired it up and I never used it since. I ditched the old 80GB Maxtor for a 320GB Maxtor (re-badged Seagate) and went with Windows 2003 server (32 bit). I was getting too many issues with Vista x64 and couldn't put up with them while trying to get this system stable.

While waiting for windows to install, I got installed fans into each of the 6 fan spaces in the case, and put the side and top panel covers on. I did this just after 2003 was set up and after starting up four instances of Prime 95. Just before putting the last side panel on, I noticed CoreTemp reported the first core temp as being 79C. It may have gone higher if I hadn't put the panel on, but after doing so it has been hovering around 72-73C. So it seems temps may have been the issue. (I'm going to try ghosting the Vista install onto another partition on the 320GB Maxtor, and re-run the Orthos to really see if the fans and closed case are helping with the problem, but in any case the system appears stable). Is it possible the CPU temp was going up to 85C and the machine was rebooting due to reaching the TJunction temp? Does the system shut down, or reboot when that happens?

In any case, it's been running for 4.5 hours (4 instances of Prime 95) so far, and no crashes. This test is a little different from what I've been running though -- Dual Orthos. I configured the Prime instances to each use 1GB of ram. Orthos was using around 1.7GB per process, so I'm not sure if there is a significant difference between what the two were doing, if the Orthos is more intensive than the Prime95 the way I had them configured. But I think it seems that temperatures have been the issue.

73C load seems high, I will probably get a Thermalright Ultra Extreme, lap it, and the CPU, and see how that goes. And also look into replacing the chipset cooling. I'm not crazy about bending the heatpipes just right so they can each sit properly on the chips.

 

[KingGheedora]

Fox, even though CPU-z has been reporting that the Auto setting is configuring the RAM speeds properly, I've had them manually set in the BIOS to rule that out as a problem.

 

[cmdrdredd]

Originally posted by: KingGheedora

Originally posted by: cmdrdredd
Try giving your Northbridge more voltage. That is the MCH chipset in the BIOS. Bump it up to around 1.37 which is perfectly fine. Also make sure your memory uses the same IC or is very similar. Keep your voltage at the recommended value and set your timings correctly.

What do you mean by this?

If one IC on a set is using Micron D9GMH and needs 2.1v and the other is using Elpida that needs 1.9v and will not function with 2.1v then you have a problem.

 

[cmdrdredd]

Originally posted by: Acanthus
Uguru.sys is the Abit software overclocking app.

I wouldnt use it, it sucks anyway.

Modifying settings in the BIOS is always better, regardless of the brand/software.

uGuru changes the BIOS directly from the software. Maybe try using it before opening your mouth.

It's a perfect way to overclock your CPU and test voltages without restarting every time. It's simple. Crash in Prime then bump voltage and try again. all without having to restart.

 

[KingGheedora]

Originally posted by: cmdrdredd

Originally posted by: KingGheedora

Originally posted by: cmdrdredd
Try giving your Northbridge more voltage. That is the MCH chipset in the BIOS. Bump it up to around 1.37 which is perfectly fine. Also make sure your memory uses the same IC or is very similar. Keep your voltage at the recommended value and set your timings correctly.

What do you mean by this?

If one IC on a set is using Micron D9GMH and needs 2.1v and the other is using Elpida that needs 1.9v and will not function with 2.1v then you have a problem.

Ok, I take it IC is integrated circuit, and that you mean using identical memory sticks? All four sticks are identical. I got 2 4GB kits of G.SKILL DDR2-800.

 

[cmdrdredd]

Originally posted by: KingGheedora

Originally posted by: cmdrdredd

Originally posted by: KingGheedora

Originally posted by: cmdrdredd
Try giving your Northbridge more voltage. That is the MCH chipset in the BIOS. Bump it up to around 1.37 which is perfectly fine. Also make sure your memory uses the same IC or is very similar. Keep your voltage at the recommended value and set your timings correctly.

What do you mean by this?

If one IC on a set is using Micron D9GMH and needs 2.1v and the other is using Elpida that needs 1.9v and will not function with 2.1v then you have a problem.

Ok, I take it IC is integrated circuit, and that you mean using identical memory sticks? All four sticks are identical. I got 2 4GB kits of G.SKILL DDR2-800.

Ok then, that's all I was saying. They need to handle the same voltages and timings.

 

[KingGheedora]

12 hours Prime95 with no crashes! Core temps after 12 hrs Prime95 are: 74,71,70,73.

Will try Vista/Orthos again, but not holding my breath. Vista has sucked.

This is all stock voltages, except for MCH voltage which is at 1.33v. I might be able to run that lower (stock is 1.25v), it was just set to 1.33v from previous testing before I found the problem to be heat.

 

[KingGheedora]

12 hours Orthos stable in Windows 2003.