Distinguishing between CPU or Memory Issue

[dadi_oh]

Just got my RMA ASUS P5N32E-SLI board back (finally). Installed an E2180 that I had already used in my P5K-E board at >3GHz and also some OCZ SLI PC6400 memory that I had also used in the P5K-E (can't remember the exact overclock but it worked well).

Well I can't seem to get this to work stable at default 200X10. At first it just crapped out during driver install. Knowing how hot the 680i runs I touched the heatsink and almost burned my finger. So I put a 92mm fan resting on top of the video card pointed at the northbridge and that brought the temp down to ~45C (as measured on the heatsink). Installed the rest of the drivers fine.

Next tried running Prime95 and it bombs out after about 3 passes. Temps on the CPU are fine with the cores getting to about 50C with Prime95. I have the memory as per the EPP @ 4-4-4-15-1T @ 400MHz. Memory is at 2.1V which should be fine (EPP states 2.0V for the 4-4-4-12-1T timings).

I checked the BIOS and it was really old. They shipped it to me with 0605. I flashed to 130x (can't remember if it was 1302 or 1304... the machine is off now). Was afraid to use the latest 1403 since it created real problems with the E2180 multiplier not being able to be adjusted (on my original board).

So now I am wondering if this is a memory instability or CPU instability (on this board).

If I were to run small FFT (8kB) I assume that would keep it in L1 cache so I could run that and see if it is stable. That might rule out the CPU. Then I could run large FFT to get it out into memory and see if that is unstable. Or is there another way to see if this is memory timing related?

Looking for ideas.

 

[EnglishLion]

Does it just bomb out of prime or crash windows too? I find that cpu crashes freeze windows or instant reboot whereas memory crashes to blue screen (BSOD).

Only sure way of knowing is to swap in some alternative memory, got any spare?

 

[alexp999]

How many sticks of ram you got?? Try running memory at 2t instead of 1t

 

[dadi_oh]

Does it just bomb out of prime or crash windows too? I find that cpu crashes freeze windows or instant reboot whereas memory crashes to blue screen (BSOD).

Only sure way of knowing is to swap in some alternative memory, got any spare?

Don't really want to take the memory out of my other machine but I don't know if it is necessary now. This memory was unning fine in my P5K-E motherboard.

Sometimes it just fails the Prime95 test (usually within the first 3 iterations) but sometimes it completely crashes (no BSOD), just resets the machine and reboots. Happened to my son last night playing COD4.

I am thinking it is motherboard related and more specifically the Northbridge. When I first got the board I only got part way through loading drivers before it was giving me intermittent crashes which lead me to believe heat related issues. Placed a fan on the back side of my video card pointed at the NB heatsink which seemed to make it "more" stable. Using an IR probe I am now measuring the NB and SB at about 48C on the heatsink which seems reasonable to me. I have some of the optional ASUS fans on order that clip onto the rads on the heatpipes so I will try one of those to see if I can get the NB cooled without a fan propped up on my video card.

Another odd thing. The E2180 that I am using has a FSB wall around 380MHz (as measured on my P5K-E motherboard) but I can't get the FSB past 290MHz on this 680i board. I set the multi to 6X and start increasing the FSB. At 300MHz it will post but not boot. At 310MHz it will not post. Tried all the way up to 400MHz just to be sure there was not a FSB "hole" that I could skip over. Also tried increasing the NB voltage past 1.2V default but that actually makes it worse.

Based on all the evidence so far I am thinking it is the northbridge and more specifically it is thermal limit on the NB. So either this motherboard has a poorly attached heatsink or I have a bum 680i. I bought this off an online store on EBay and already had to return the first one since it would not POST at all. Took several weeks to get it back so I think I will just see what I can do to get this running properly. Maybe an aftermarket heatsink/fan on the NB? But then I need something on the power section since this is an all-in-one heatsink that covers NB and power. I have a Jing-Ting NB cooler that I had used on my A8N-E board but not sure if it will handle the 680i power.

http://www.bigbruin.com/2006/jts0006_1

Also, although I have modded this case to get the best airflow I can, it is still lacking in that department so I might look for a better case with 120mm fans in and out.

If I hadn't wanted the SLI feature I would have stuck with another P5K-E (or other P35 based) board. They OC really well and the P35 NB runs waaaaay cooler than this 680i. Oh well.

 

[dadi_oh]

How many sticks of ram you got?? Try running memory at 2t instead of 1t

Yeah I actually relaxed the memory timing to 6-6-6-15-2T at one point to completely rule out memory timings. But I ran this memory for a few weeks in my P5K-E board at 400MHz 4-4-4-12-1T so I am almost positive this is not the memory.

 

[alexp999]

A lot of mobos have trouble running at 1T. Especially if you have 4 sticks.

 

[dadi_oh]

A lot of mobos have trouble running at 1T. Especially if you have 4 sticks.

OK. I only have 2 sticks but I had relaxed to 2T command rate at one point but didn't seem to make a difference. I could just relax it to 2T until everything is sorted out to avoid potential issues.

Thanks

 

[EnglishLion]

Sounds like nb and heat problem then, if upping the nb voltage makes it worse (more heat).

On the system cooling front:

If you want to add an additional intake fan you can use the scythe kama bay (see my sig for a review) where ever you have three free 5.25" bays to add intakes. Exhausts can be added by pci slot coolers.

 

[dadi_oh]

Sounds like nb and heat problem then, if upping the nb voltage makes it worse (more heat).

On the system cooling front:

If you want to add an additional intake fan you can use the scythe kama bay (see my sig for a review) where ever you have three free 5.25" bays to add intakes. Exhausts can be added by pci slot coolers.

Thanks. I thought about adding more cooling capability to the case but it might be cheaper to just buy a better case with 120mm fan capability. I am probably at the point of diminishing returns on this case and once I add the second video card for SLI it will be asking a lot for this case to breathe properly. Also, we are just heading into summer here in Ottawa, Canada and temps in the 30C range are just around the corner. We just jump in our pool to cool off but that might not be an option for the computer

I like the Antec cases but I will see what is available around here for a reasonable price.

 

[AddSub]

CPU vs. RAM, problem-wise? 9 out of 10 times it turns out to be a RAM/motherboard problem, in my experience at least. CPU's themselves are rarely the cause of any problems. People don't realize how durable and reliable these little things are.

Quality control when it comes to manufacturing of RAM IC's is much less comprehensive vs. manufacturing of CPUs, and RAM IC manufacturers themselves have less rigorous testing procedures and methods vs. that of AMD/Intel. Also, when it comes to RAM vs. motherboard issues, it's usually the motherboard that is at fault and not the RAM itself. I've encountered many motherboards that had screwy BIOS'es. For instance a random bug in a BIOS on one of my old motherboards caused problems every time I set CAS timing to 3 because the CAS would in reality be set to 2 by the BIOS, but it would show 3. CAS 2 is something my old RAM could not handle and it caused many problems until I tracked down what it was.

Again, CPU's themselves are rarely the cause of hardware problems. Poorly designed motherboards, a buggy BIOS or even screwy RAM is more probable. Much MORE probable.

 

[dadi_oh]

Thanks. I thought about adding more cooling capability to the case but it might be cheaper to just buy a better case with 120mm fan capability. I am probably at the point of diminishing returns on this case and once I add the second video card for SLI it will be asking a lot for this case to breathe properly. Also, we are just heading into summer here in Ottawa, Canada and temps in the 30C range are just around the corner. We just jump in our pool to cool off but that might not be an option for the computer

I like the Antec cases but I will see what is available around here for a reasonable price.

Well I picked up an Antec Sonata II case used so I will move the system over into that. I bought an additional 120mm fan to put on the front so I think that I should have a well ventilated system. That will help but I am still concerned that maybe the NB heatsink is not mounted properly. Since I will have the motherboard out of the case to move it I will probably remove the heatsink and remount with AS Ceramique. Still waiting for the optional ASUS fan for the radiator on teh end of the heatpipe but I might try screwing a small fan onto the NB heatsink itself as well.

Anyone with this same board know how much they had to bump NB voltage to get FSB up in the 350 ish range?

 

[ntdouglas]

A lot of mobos have trouble running at 1T. Especially if you have 4 sticks.

Actually, 1t does not even support 4 dimms being populated. Using all 4 dimms you have to use 2t.

 

[dadi_oh]

Well I picked up an Antec Sonata II case used so I will move the system over into that. I bought an additional 120mm fan to put on the front so I think that I should have a well ventilated system. That will help but I am still concerned that maybe the NB heatsink is not mounted properly. Since I will have the motherboard out of the case to move it I will probably remove the heatsink and remount with AS Ceramique. Still waiting for the optional ASUS fan for the radiator on teh end of the heatpipe but I might try screwing a small fan onto the NB heatsink itself as well.

Anyone with this same board know how much they had to bump NB voltage to get FSB up in the 350 ish range?

Almost gave myself a heart attack last night I moved the contents of the old case over to the new Sonata case. While I had the motherboard out of the case I decided to check out the thermal paste application that ASUS would have made at the factory. Pulled off the NB heatsink and it looked like gray muffler cement had been applied on the chip. Had to use acetone to get it off the heatsink and the 680i chip. The bottom of the copper heatsink was quite rough. I had a mind to lap it but I couldn't find my sandpaper. So I just cleaned it all up, applied some Arctic Ceramique in a thin layer on the chip and remounted the heatsink. Put everything back together, pushed the power button and..... fans all came on full speed but no BIOS beep. Oh Oh... Tried clearing the BIOS and same story.

So I pulled the entire contents back out of the case, removed the heatsink to examine the thermal paste. One whole corner of the chip did not make contact with the heatsink based on the thermal paste transfer to the heatsink. Oh crap. I guess the heatsink does not sit quite flat on the top of the chip. Maybe that why ASUS uses chewing gum to apply the heatsink in the first place to take up the slop. Thinking about it now, having that big arm of a heatpipe sticking out the side adds some complexity to keeping the heatsink flat on the chip. The slightest mis-bend in the heatpipe and you are going to alter how it sits. I guess that is the hazard of this design.

So I added a thicker layer (than I normally do) of the ceramique, put it all together again and hit the power button.... Pause.... Beep! Cool. Back to normal. Thought I had fried the chip but I guess it has thermal protection built in. Thank you Nvidia.

But after all that I am pretty much back to tthe same sort of FSB wall that I think is on my motherboard side. This based on the fact that this particular E2180 had a FSB wall around 370-380 on my P35 based motherboard (P5K-E). I tried bumping the NB voltage from 1.2V to 1.4V and the CPU core to 1.45V but it just gets hotter and doesn't seem to help the FSB. So now I am running at 275X10. System seems stable according to Prime95.

I snapped on the optional ASUS fan on the Northbridge radiator (had to remove not one but two warning labels stating that these fans are for watercooled systems only). It doesn't seem to block my OCZ vendetta by much so I think it is just a conservative CYA recommendation.

At stock 200X10 the NB heatsink runs at about 48C with the optional fan. At 275X10 it is sitting around 63C (at default 1.2V). A bit hot for my taste but I guess as long as it works. At some point I might consider pulling off the stock heatsink, put on a fan powered heatsink and some sort of passive heatsink on the power section. Based on the thick thermal pad they put on the power devices they can't require that much cooling. Maybe even some of the video RAM heatsinks I have laying around.

So maybe I will just leave it at that. 2.75GHz isn't that bad and with the 2 SLI 7900GT (clocked at GTX speed 650/1700) I break 10,000 on 3DMark06.

 

[jtleon]

Heatpipe manufacturing challenges

Thinking about it now, having that big arm of a heatpipe sticking out the side adds some complexity to keeping the heatsink flat on the chip. The slightest mis-bend in the heatpipe and you are going to alter how it sits. I guess that is the hazard of this design.

I have wondered about consistent engagement using plastic spring pins with multi chip heat pipe systems. If I were you, I would start bending those heat pipes until I achieved perfect mating with all sinks/chips (using cheap HS compound for confirmation). If you cant achieve near perfection, it may be aftermarket time. If you do achieve good mate, then move on to lapping - not before!

As you now know, no heatsink, no matter how good, can cool without a proper interface to the chip. On the other hand, with a great interface, even a crappy heatsink can do the job!

Regards,
jtleon

 

[dadi_oh]

I have wondered about consistent engagement using plastic spring pins with multi chip heat pipe systems. If I were you, I would start bending those heat pipes until I achieved perfect mating with all sinks/chips (using cheap HS compound for confirmation). If you cant achieve near perfection, it may be aftermarket time. If you do achieve good mate, then move on to lapping - not before!

As you now know, no heatsink, no matter how good, can cool without a proper interface to the chip. On the other hand, with a great interface, even a crappy heatsink can do the job!

Regards,
jtleon

Agreement Meter turned to 11....

I think the aftermarket route is more attractive. The trial and error of trying to achieve flatness would chew up more time than I have (and probably a couple of tubes of thermal paste

I have this JingTing heatsink laying around that I had on my A8N-E (NForce 4) motherboard. Wonder if it would do the job?

http://www.bigbruin.com/2006/jts0006_1

 

[jtleon]

That Heatpipe may not be needed.

Agreement Meter turned to 11....

I think the aftermarket route is more attractive. The trial and error of trying to achieve flatness would chew up more time than I have (and probably a couple of tubes of thermal paste

I have this JingTing heatsink laying around that I had on my A8N-E (NForce 4) motherboard. Wonder if it would do the job?

http://www.bigbruin.com/2006/jts0006_1

Certainly that aftermarket item would serve the NB well, but I thought that heatpipe also cools the power MOSFETS. If so, you can simply cut the heatpipe (bandsaw comes to mind) at the mosfet sink and throw away the rest. And/Or cut also at the NB and try without the heatpipe function (keep that aftermarket for another project). I would not be surprised that the heatpipe is more eye candy than a requirement - if you can achieve quality interface without it. Who knows, you may discover that there is nothing inside the heatpipe!

Regards,
jtleon

 

[dadi_oh]

Certainly that aftermarket item would serve the NB well, but I thought that heatpipe also cools the power MOSFETS. If so, you can simply cut the heatpipe (bandsaw comes to mind) at the mosfet sink and throw away the rest. And/Or cut also at the NB and try without the heatpipe function (keep that aftermarket for another project). I would not be surprised that the heatpipe is more eye candy than a requirement - if you can achieve quality interface without it. Who knows, you may discover that there is nothing inside the heatpipe!

Regards,
jtleon

Heatpipes really do work as a highly efficient means to transport heat from point A to point B. That is why they use them in laptops so extensively. I believe this heatpipe is really working based on the fact that I measure very similar temps between the NB heatsink and the radiator over the power section.

I could hacksaw it off but I thought that if I ever want to resell the board it would be nice to have stock. I could pull of the stock, replace the NB with the JingTing and then use some stick-on heatsinks on the power devices. According to the imprints on the bottom of the thermal pad there are 4 of these (MOSFETS?) being cooled. I was thinking of using some VGA memory heatsinks since I have some of those available. I think they would work at least as well as what they have on there now but I might need to deflect some air from the CPU fan down towards them.

 

[jtleon]

Heatpipes really do work as a highly efficient means to transport heat from point A to point B. That is why they use them in laptops so extensively. I believe this heatpipe is really working based on the fact that I measure very similar temps between the NB heatsink and the radiator over the power section.

I could hacksaw it off but I thought that if I ever want to resell the board it would be nice to have stock. I could pull of the stock, replace the NB with the JingTing and then use some stick-on heatsinks on the power devices. According to the imprints on the bottom of the thermal pad there are 4 of these (MOSFETS?) being cooled. I was thinking of using some VGA memory heatsinks since I have some of those available. I think they would work at least as well as what they have on there now but I might need to deflect some air from the CPU fan down towards them.

Suit yourself, however if I were in the market for a used board, I would purchase one that is a proven overclocker over one that cannot achieve stable stock speeds, any day!

Stick On heat sinks are crap and you know it! Use a quality thermal adhesive - AS makes a good epoxy.

Regards,
jtleon

 

[dadi_oh]

Suit yourself, however if I were in the market for a used board, I would purchase one that is a proven overclocker over one that cannot achieve stable stock speeds, any day!

Stick On heat sinks are crap and you know it! Use a quality thermal adhesive - AS makes a good epoxy.

Regards,
jtleon

I do have some Arctic Thermal Epoxy that I could use. Problem is, once you stick it on it is there forever as far as I know. I might do that but I am thinking that the power section is not really high demand as far as cooling goes (compared to the NB). On my P5K-E, ASUS doesn't even bother to put a heatsink on the power section above the CPU socket. I aimed my thermal probe at them and they are only in the low 40's. Most power devices are rated up in the 150C range AFAIK so probably a good decision by ASUS on that board.

I agree the stick on is not optimal. The Zalman VGA heatsinks I used on my 7900GT card had a preapplied thermal tape and 2 of them fell off at some point. Opened the case and they were laying on the bottom And I had cleaned the memory with IPA before applying.

So many options... so little time...