Could the AI7 be the cause?Suggestions?

Over the last few months I started to upgrade my system with an AI7, 2.4c mo, two 256 Kingston hyperx 3200 bh-5, a fortron 550 psu and anATI 9800pro card.
After playing with it for a while I found that the sweet spot (stable) with the CPU on stock cooling was at 3.2 (267fsb) 1.55V with 5:4 div. 2 5 2 2 timing at 2.8V. Because summer is coming and I live in a desert I decided to upgrade my cooler to a sp94 with a panaflow fan. With that setup I could clock to 3.3(276fsb) stable at 45C load with 1.625V. After one day suddenly the computer rebooted got stuck in windows and after rebooting it became immediately unstable with the blue screen saying windows had to shut down to prevent damage to the system…….Normally this comes up with a new hardware/software install which has conflicts. I had to take the fsb back to 250 were it is stable at 1.6 V. It will not be stable with prime even with the 3:2 divider any higher. If I go to about 265 or so, it will reboot and this blue screen is coming up. First I thought it might be the CPU, which degraded but this blue screen is not typical for the CPU, it normally freezes, getting to hot. The temp is not over 42 C load. So I used one stick of memory and voila it runs fine at @3.3 prime for hours. So does the other stick . In the moment I put them together, I start to get memory errors after a few minutes above fsb 255 and prime fails shortly after, at higher speeds I get the blue screen mentioned above.
I upgraded the bios no effect. I am stumped. Could it be the mother board? I just put some AS5 under the north bridge cooler (does that void my warranty?). Any suggestions ?

Thanks
puji

I have the exact same setup as you. I'm running an AI7 with a 2.4c m0 with 2x256 of HyperX PC3500. My overclock also topped out at 12x267 using 1.55v. I also just bought a SP94, but I opted for the Vantec tornado fan. I also can't run the 3:2 divider, I get all sorts of errors. I have the memory at 2-2-2-6 using 3.0vdimm.

I take it from your post that you cannot even get it stable back at 267 like you have it previously? That's quite strange. I would suggest relaxing your memory timings or try jucing the voltage a bit.

I was going to suggest putting some AS under your northbridge heatsink, but it seems you already did that. I did as well, I'm not sure about the warranty though.

If you want to really start over, clear your bios, and reconfigure everything to make sure you haven't missed some setting.

As far as the hot weather goes, I usually find the area where I'm just on the border of stability and then I back it down 100mhz or so so I can be sure it will be stable when the temperature goes up. For example, if your case you could run 3.3, but I would have backed it back down to 3.2 just to ensure stability.

I dislike running on the fringe, strange things can happen from time to time. I prefer stabililty over a 3% gain in clock speed.

Arch said:

I have the exact same setup as you. I'm running an AI7 with a 2.4c m0 with 2x256 of HyperX PC3500. My overclock also topped out at 12x267 using 1.55v. I also just bought a SP94, but I opted for the Vantec tornado fan. I also can't run the 3:2 divider, I get all sorts of errors. I have the memory at 2-2-2-6 using 3.0vdimm.

I take it from your post that you cannot even get it stable back at 267 like you have it previously? That's quite strange. I would suggest relaxing your memory timings or try jucing the voltage a bit.

I was going to suggest putting some AS under your northbridge heatsink, but it seems you already did that. I did as well, I'm not sure about the warranty though.

If you want to really start over, clear your bios, and reconfigure everything to make sure you haven't missed some setting.

As far as the hot weather goes, I usually find the area where I'm just on the border of stability and then I back it down 100mhz or so so I can be sure it will be stable when the temperature goes up. For example, if your case you could run 3.3, but I would have backed it back down to 3.2 just to ensure stability.

I dislike running on the fringe, strange things can happen from time to time. I prefer stabililty over a 3% gain in clock speed.

Click to expand...

Thank you for your reply. A few things, I started with the kingston hyperx 3500 and that is ch-5 and found out (lots of posts about that here) that ther is a problem with the divider and ch-5 . It basically will not work even at default speeds. So I got the 3200 with is bh-5 and normally works really well . I could go 244 fsb and 1:1 and with the 5:4 divider over 285 . But the cpu was not stable (freezing up) so I backed down to comfortable levels and had no problems for two month . So you problem can only be solve by getting bh-5 chips or high 1:1 like the adata chips.
I suspect my motherboard has now trouble with the divider and bh-5 because I can go 244 1:1 2522 and start to have trouble at 255 5:4 with two sticks,but can go 1 stick 278 5:4 stable.

I don't think my HyperX PC3500 has the CH-5 chips on it. I bought it at this time last year so its over a year old. Its also my understanding that CH-5 doesn't work at any speed using 2-2-2-6, instead it requires 2-3-2-6.

What slots are you using for the memory? I've heard people having greater success using 2-4 instead of 1-3.

Arch said:

I don't think my HyperX PC3500 has the CH-5 chips on it. I bought it at this time last year so its over a year old. Its also my understanding that CH-5 doesn't work at any speed using 2-2-2-6, instead it requires 2-3-2-6.

What slots are you using for the memory? I've heard people having greater success using 2-4 instead of 1-3.

Click to expand...

You are probably right with your hyperx . I tried all slots, it does not change. The strange thing is it still works perfect in single channel mode but not in dual.

The system is back to normal. My suspision is the northbridge. Several hours before my system failed I had exchanged the heatpad on the zink for some arctic silver 5. It worked after that and I do not know what actually triggert the failure. So today after everything else failed I removed that zink and there was a liberal amount on the zink. I put it back and moved it around to make good contact. Nothing changed first. So I set the system to the best stable settings: 243 fsb and 1:1 ratio. Then I "cooked"the system. Actually I just tried different things for airflow and it heated everything up to the high 50th for a while. Then I had to try the old setting again(I did not expect any changes at this time ) but to my surprise and delight 278fsb at 5:4 stable.
So my guess is, the zink did not have good contact for a while., which seems to make the dual channel 5:4 divider situation unstable, but not the 1:1.

Interesting thread!

I'm stuck at 267 as wel... it can even do hours of memtest86 without errors at 285/233 2-5-3-3 with 2.95v, but Prime95 fails instantly above 267. I guess it's the CPU since I was also stuck here with my IC7 and 2×256 XMS3500.

Going for an M0 and an SP-94 next week

Another interesting thing: I leave it on 24/7, but when I turn my system off for some reason, it will be unstable when overclocked for about 10 minutes. Blue screens. After a few minutes of idling (I enter BIOS setup first to avoid damage to windows) it runs perfectly.... has anyone seen anything like this?

Csybe said:

Interesting thread!

I'm stuck at 267 as wel... it can even do hours of memtest86 without errors at 285/233 2-5-3-3 with 2.95v, but Prime95 fails instantly above 267. I guess it's the CPU since I was also stuck here with my IC7 and 2×256 XMS3500.

Going for an M0 and an SP-94 next week

Another interesting thing: I leave it on 24/7, but when I turn my system off for some reason, it will be unstable when overclocked for about 10 minutes. Blue screens. After a few minutes of idling (I enter BIOS setup first to avoid damage to windows) it runs perfectly.... has anyone seen anything like this?

Click to expand...

I can share some of my experiences. When I used the stock aircooling (only arctic silver 5 instead of pad) I used to get unstable above 3.2 no matter what Vcore as if the higher heat through higher voltage negated the positive effect. Also when it got only 3 degrees warmer in the room prime became unstable . Sometimes when it was quite warm in the room and I shut down the sytem, the next morning I could not boot, because my bios setting had changed misteriously , as if the latent heat triggert something after I shut the computer down. I had to lower the fsb a few notches and it did not happen again. This all dissapeared after I installed the sp 94.
So the chances are that it is the same heat relatet problem in your case and you will be able to go to 275/288 without a problem once you can improve your cooling. You might even go first only with the sp94 and all will be ok (save the money for a prescott second edition). My sytem with cover off is now sys 31 cpu 34 pwm 35 no load and sys 31 cpu 44 pwm 45 full load(100%). Before it was sys 31 cpu 38 pwm 39 no load and sys 31 cpu 58 pwm 60. I also think it is not the temp you can read which is important ,it is more the heat transfer which is so much greater with the sp 94.

I don't know if it's a heat issue for me, but it was stable @ 275 for about two months, since then Prime gives me rounding error above 267...

I will get the SP94 and I might be able to replace my CPU to an M0 stepping, if I can do that I won't miss the chance. It will cost only a few bucks since it is the same type...

270 -> error after 20 mins
275 -> error after 5 mins
above -> error immediately (but still boots into windows)

Csybe said:

I don't know if it's a heat issue for me, but it was stable @ 275 for about two months, since then Prime gives me rounding error above 267...

I will get the SP94 and I might be able to replace my CPU to an M0 stepping, if I can do that I won't miss the chance. It will cost only a few bucks since it is the same type...

270 -> error after 20 mins
275 -> error after 5 mins
above -> error immediately (but still boots into windows)

Click to expand...

There is a chance, that it is heat related. The zalman has a weak spot in the center, because ther are no fins, so the heat has to travel longer before it dissipates. That could be the deciding factor weather it will be more stable even at the the same temp. readings.
So if that works for you, why not waiting for a 2.8 prescott with the next generation steppings ( intel is working hard on it) . they have a much larger cache and might even overclock better if they can resolve the heat issue. Priices are the same as Northwood.

You're right about the Zalman

Need a 2.4C, I want high FSB, and with my current board & RAM I can go for a reasonable 300...

I also have an ai7, 2.4c m0 combo with hyperx pc3200 ram (bh-5).
I can run prime for hours at 284fsb 2/3 ratio (2 2 2 5), memtest 86, super pi, and 3dmark2003 without artifacts. I can even run at 292fsb with loose timings and have run these tests stable. But Neverwinter nights will give me a lot of artifacts, which I can't fully cure until I run at a lower fsb and or loosen the ram timings.

I have a sp-94 cpu cooler and at load I am at 46c max on the cpu, 29c on the system, and 56c on the power supply (purepower 480).

It seems like it must be a nortbridge problem, but raising the agp/northbridge volts doesn't seem to help. Mabey a bigger cooler on the northbridge would help, but the current temps don't seem high.

With one stick of ram I can even boot into windows at 300fsb, but it wasn't stable.

Any Ideas?

Interesting thread.

I just yesterday stuck in a second stick of PC3500 in DC. I started seeing intermittent errors that I hadn't seen in single-channel mode. Strange thing is, memtest86 shows no errors on multiple passes.

I'm running a much lower FSB (230 FSB), but the phenomenon was the same: DC in 1:1 gave some spurious behavior, whereas either single stick produced no problems.

I raised the DDR voltage by .05V, and that seems to have helped a bit. Might it be possible that running more sticks is sagging your DDR voltage under load?

I might just try the tip of raising the AGP / NB voltage. --- Paul

Many people report poor memory performance. Hearing all these memory issues make me believe the AI7 has some serious problems.
Also, I usually get rounding errors in Prime when I push the memory hard.

A question to all of you:
Have you disabled the last two settings under Advanced Chipset settings, CPC and tRDA (if there still is that kind of setting). Having it enabled on the IC7 causes unstable performance and brick wall at ~260.
With it disabled, I reach 300FSB.

I don't have first hand experience with AI7, but IC7 require more vdimm when running for example 300FSB 5:4, then 240FSB 1:1, even though the memory is running at the same speed.

Jags said:

I also have an ai7, 2.4c m0 combo with hyperx pc3200 ram (bh-5).
I can run prime for hours at 284fsb 2/3 ratio (2 2 2 5), memtest 86, super pi, and 3dmark2003 without artifacts. I can even run at 292fsb with loose timings and have run these tests stable. But Neverwinter nights will give me a lot of artifacts, which I can't fully cure until I run at a lower fsb and or loosen the ram timings.

I have a sp-94 cpu cooler and at load I am at 46c max on the cpu, 29c on the system, and 56c on the power supply (purepower 480).

It seems like it must be a nortbridge problem, but raising the agp/northbridge volts doesn't seem to help. Mabey a bigger cooler on the northbridge would help, but the current temps don't seem high.

With one stick of ram I can even boot into windows at 300fsb, but it wasn't stable.

Any Ideas?

Click to expand...

The northbridge is very sensitive to heat. I just put a funnel to the panaflow fan .Either I moved the zink or the airflow got changed enough, the northbrighe is getting hotter and with that I suddenly got instability with the sytem. not as bad as before but prime was not stable . So I lapped the zink because it is really uneven. Then I put arctic silver 5 on . In the beginning I got erroers like crazy again and failure of prime even at 250 fsb. I really had to run the sytem under load for quite a bit to get the silver speading and now it gets slowly better. So my gues in your case is the heat of the videocard under game load effects the northbridge . A better cooler or lapping might solve that problem, otherwise get a cooler for the card , which funnels th heat out of the sytem.

macklin01 said:

Interesting thread.

I just yesterday stuck in a second stick of PC3500 in DC. I started seeing intermittent errors that I hadn't seen in single-channel mode. Strange thing is, memtest86 shows no errors on multiple passes.

I'm running a much lower FSB (230 FSB), but the phenomenon was the same: DC in 1:1 gave some spurious behavior, whereas either single stick produced no problems.

I raised the DDR voltage by .05V, and that seems to have helped a bit. Might it be possible that running more sticks is sagging your DDR voltage under load?

I might just try the tip of raising the AGP / NB voltage. --- Paul

Click to expand...

Memtest 86 might not show errors, because your memory is ok and the northbrigdge gets not stressed in that test. It is the northbridge especially with the divider under heat constrains which causes the problem. You might see errors using memtest which runs in windows (I do). Try to get the northbrige cooled better, It is a Known problem with the AI7 mentioned as far back as the first review from Anand Tech.

That's a very good point, and I'm inclined to agree with that possibility. At the end of the day, I bumped the DDR voltage just a hair and increased the NB fan speed a small amount as well. Thank you for the tip.

At some point, I'll probably look into larger, passive (a la Zalman) NB sink and make sure there's good enough case airflow over it. I'm not really into the buzzing NB fan concept when it can be avoided. Glad I decided to try out 1 GB of memory before moving on to 2 GB.

At any rate, I don't appear to have any problems at the moment, whether in gaming or in long computations.

One point for the others: if heat is the issue here, I'd recommend at least trying to reduce the NB voltage, rather than raise it. You might be pleasantly surprised. -- Paul

macklin01 said:

One point for the others: if heat is the issue here, I'd recommend at least trying to reduce the NB voltage, rather than raise it. You might be pleasantly surprised. -- Paul

Click to expand...

macklin01,
And that is a very interesting comment you made there. Could be on to something. Interesting thread guys. My board runs prime,memtest and 3d STABLE @ 288 but no amount of voltage will make it work right at 290 . Maybe I get a little higher beacuse my cpu is watercooled which would cut down on some heat to the n/b as well. Wonder if this would help. http://www.frozencpu.com/cgi-bin/frozencpu/vid-26.html

I would like to think I could solve this problem is caused by heat because I could remedy it pretty easy. But the artifacts in Neverwinter occur immediately, even if the game is started right after booting up. That shouldn't give the northbridge time to get hot. But I will take replace the heat pad with arctic silver 5 anyway, just to see if it helps.

I have also noticed that when I set the fsb to a high number such as 284, it will sometimes drop one to 283 the next time I start up the computer. I wonder if the clock generator is completely stable on the ai7.

My personal findings are that problems start at fsb over 270.

I chose this board over the ic7 for the solid high ddr volts, which it does supply. But with my setup, upping the ddr volts over 2.8v hasn't seemed to make any difference. So I am probably going to pick up an ic7 and see how that goes. I really think that this processor is capable of 300fsb.

Jags your PSU could be it. It has a very weak 12V rail as far as I know.