AMD 3500 overclocking problem

[z-zar]

Hi, my first post here....

I did quite a lot of reading up on overclocking and I've been reading about all these people/websites apparently effortlessly pushing their 3500 to 2.5 or even 2.6GHz but I'm still not able to push my 3500 newcastle (yes, i know, it's not a winnie) stable beyond 2365MHz. It fails prime95 within an hour or less on 2.4GHz. Changing vcore to 1.6V doesn't help. It will start windows up to 2.5GHz but not pass prime95. I can play games for maybe an hour or two but it will eventually freeze up at 2.4 or 2.5GHz.

I've just successfully completed 11 hours prime95 on 263x9=2365MHz at stock CPU voltage (1.5V) so I know at least my memory can take FSB=263. When I try 253x9.5=2.4GHz, prime will fail within the hour, sometimes within five minutes. I also tried 240x10 figuring maybe something with the half divider, but no difference. I upped the vcore to 1.55 then 1.6 but no change. I upgraded my powersupply from "came-with-chassis" to a 480W Antec trueblue 2.0. No difference.

My CPU idle temp is 37-38C, measured with speedfan. Temperatures after 30 minutes of prime95 are about 52C at my 2365MHz overclock settings. After the 11 hour run it was 56C. After 30-60 minutes at 2.4GHz the temperature is pretty much the same 52-53C.

My system specs with settings are listed below. Other than that, I set the HTT multiplier to 3x, coolNquiet disabled, dynamic overclocking disabled, and am using SATA socket 3 which is locked on my MSI board so no harddrive corruption.

Any suggestions? I thought with this top of the line (for air cooling) HSF and nice case I should be able to do better. Here's my system specs:

Athlon 3500+ Newcastle @ 263x9=2365MHz, 1.5V
Thermalright XP90 + 92mm panaflo + arctic silver 5
MSI K8N Neo2 Platinum
Corsair XMS 2x512MB TWINX3200XL, 2.5-3-3-10-1T @ 263MHz, 2.85V (this memory rocks!)
Thermaltake Tsunami case w. Antec Trueblue 2.0 480W
eVGA 6800 Ultra at stock settings/voltage

 

[(cf)Eclipse]

curious. from what i can tell... that's the limit of the cpu

though.. it's very odd that increasing vcore isn't helping. have you tried 1.65v?

 

[z-zar]

haven't tried 1.65V. Was a bit afraid of frying my cpu...

 

[(cf)Eclipse]

with an xp-90, i'd stop at 1.65v.. i would have slight reservations going over that though. just keep an eye on temps

 

[z-zar]

I've been running 1.65V vcore now for 2 hours or so playing games, so far no hick-ups. I'll run prime tomorrow when I'm away at work. Temps right now are 55C if I can believe speedfan. I noticed if I run prime95 and get a certain temperature, then restart windows, both my idle and load temperatures reported by speedfan are suddenly about 5 degrees lower than before the restart. Seems like the CPU needs to cool down completely before returning accurate temperatures upon power-up.

 

[(cf)Eclipse]

no. it's highly unlikely that the sensor will ever report correct temps. i'll be willing to bet that on a warm reboot, idle is ~38c, no?
http://www.hardforum.com/showthread.php?t=880183

 

[CitizenInsomniac]

What happens at lowers FSB speeds and higher dividers, such as 11x240, 10x250?

 

[NEODARK]

I have pretty similar hardware to yours.

I was able to push mine to 260x10 for 2.6ghz on 1.525 -1.550 What bios are you using? the latest 1.8 seemed to make things more stable for me. (this pased prime and memtest stable 24hrs)

I also have an XP90 and have experienced the odd temp drop upon reboot, however with bios 1.8 that seemed to go away. (although my bios always reports 39C lol)

try overclocking just the cpu or memory at the time, not both so you can determine how far yours can go individually then find a balance. One thing is worth noting is that with previous bios' I tried, it would get unstable when overclocking both CPU and ram together to speeds that they were stable by themselves.

right now a good 24/7 setup for me is 237x11 = 2604Mhz with a 4xHT and my CPU is a NC core as well. (my ram does all the way to 275 though, but with the higher multi 11x I get better results than with 10x)

PS my temps hover around 35C idle - 45 load with XP90 and 92mm Tornado @ 2500RPM's (speedfan/everest)

this board doesnt like the 166 divider and another tip is make sure you un-hide the bios options and disable nvidia/ati speedup. Also kicking your agp voltage to 1.6 doesnt hurt

 

[z-zar]

I have pretty similar hardware to yours.

I was able to push mine to 260x10 for 2.6ghz on 1.525 -1.550 What bios are you using? the latest 1.8 seemed to make things more stable for me. (this pased prime and memtest stable 24hrs)

I also have an XP90 and have experienced the odd temp drop upon reboot, however with bios 1.8 that seemed to go away. (although my bios always reports 39C lol)

try overclocking just the cpu or memory at the time, not both so you can determine how far yours can go individually then find a balance. One thing is worth noting is that with previous bios' I tried, it would get unstable when overclocking both CPU and ram together to speeds that they were stable by themselves.

right now a good 24/7 setup for me is 237x11 = 2604Mhz with a 4xHT and my CPU is a NC core as well. (my ram does all the way to 275 though, but with the higher multi 11x I get better results than with 10x)

PS my temps hover around 35C idle - 45 load with XP90 and 92mm Tornado @ 2500RPM's (speedfan/everest)

this board doesnt like the 166 divider and another tip is make sure you un-hide the bios options and disable nvidia/ati speedup. Also kicking your agp voltage to 1.6 doesnt hurt

hmmm, at least you're having success with a similar setup then.

My overclock seems to be CPU limited; my RAM can take up to FSB=270 tested. I'll try using the higher CPU dividers, see if it improves things. Right now I'm running 267x9=2.4GHz at vcore=1.625 while I'm at work. Tried 1.65V yesterday but my temps approached 60C after 30 minutes prime95 so I dialed it down to 1.625V. I'll know tonight.

I'm running bios 1.4. Didn't know it was already up to 1.8. My nvidia speedup is set to auto right now, tonight I'll disable it and see if that helps. Thanks.

 

[z-zar]

OK, I got home from work and found that my prime95 lasted the whole of 11 minutes at 2.4GHz/1.625V. Bummer.

I did download the bios 1.8 as suggested and that dropped my apparent temperatures reported by speedfan about 3 degrees, i.e. 35C at idle instead of the default 38, and also load temperatures are a bit lower than before.

With bios 1.8 I tried 2.4GHz at stock settings; lasted 1 minute in prime95. Strange since I made 11 hours before I stopped it at 2.365GHz...

Right now I'm running 220x11=2.42GHz as someone suggested to use higher CPU multipliers. I'll see how that works out tomorrow.

Thanks for all the feedback. Guess I just had bad luck with this CPU.

 

[CitizenInsomniac]

In your original post you mention:

I've just successfully completed 11 hours prime95 on 263x9=2365MHz at stock CPU voltage (1.5V) so I know at least my memory can take FSB=263

This is not necessarily so. If you ran the small FFT torture test, memory wasn't actually used much because most data fit into cache. In order to truly verify your memory can function at that speed, run Memtest86 test #5 for at least 20 cycles, and then run Prime95 *large FFT* for 24 hours. It might turn out that you're giving your memory more credit than it deserves. Perhaps you need to use a memory divider at such high FSB speeds after all.

How did your 11x220 test go? You should also try these combinations:

11 x 240 with mem divider set to 6:5 (aka 166 MHz, aka DDR333)
10 x 250 with mem divider set to 6:5 (aka 166 MHz, aka DDR333)

You probably won't need more than 0.125V above stock voltage for any of these.

 

[.:Blitzkrieg:.]

I have that same mobo, cpu, ram and the only real difference is i have a xp-90c. I made my goal anything over 2.5 but have yet to reach it, the farthest I've got is 2.49. You might be able to get yours higher than you are at now but don't count on hiting 2.5 or above.

 

[Technetium]

I cant get my winnie 3500+ stable with even a 5% overclock...cant even budge anything, graphics in games will artifact and eventually lock up (5-10) minutes...

it could be due to..

a) I dont know what the fuck im doing.
b) I bought cheapest ram possible (corsair value select)

 

[2Fresh]

Search for the modded bios that allows 18% on voltage. Set your divider ro 133 and 2T on mem. bump your mem voltage up and cpu voltage untill you get to hot. also run the ht or whatever it is at 3.

 

[Technetium]

was that directed to the original poster or me :\

 

[FanATIc]

Hi, my first post here....

I did quite a lot of reading up on overclocking and I've been reading about all these people/websites apparently effortlessly pushing their 3500 to 2.5 or even 2.6GHz but I'm still not able to push my 3500 newcastle (yes, i know, it's not a winnie) stable beyond 2365MHz. It fails prime95 within an hour or less on 2.4GHz. Changing vcore to 1.6V doesn't help. It will start windows up to 2.5GHz but not pass prime95. I can play games for maybe an hour or two but it will eventually freeze up at 2.4 or 2.5GHz.

I've just successfully completed 11 hours prime95 on 263x9=2365MHz at stock CPU voltage (1.5V) so I know at least my memory can take FSB=263. When I try 253x9.5=2.4GHz, prime will fail within the hour, sometimes within five minutes. I also tried 240x10 figuring maybe something with the half divider, but no difference. I upped the vcore to 1.55 then 1.6 but no change. I upgraded my powersupply from "came-with-chassis" to a 480W Antec trueblue 2.0. No difference.

My CPU idle temp is 37-38C, measured with speedfan. Temperatures after 30 minutes of prime95 are about 52C at my 2365MHz overclock settings. After the 11 hour run it was 56C. After 30-60 minutes at 2.4GHz the temperature is pretty much the same 52-53C.

My system specs with settings are listed below. Other than that, I set the HTT multiplier to 3x, coolNquiet disabled, dynamic overclocking disabled, and am using SATA socket 3 which is locked on my MSI board so no harddrive corruption.

Any suggestions? I thought with this top of the line (for air cooling) HSF and nice case I should be able to do better. Here's my system specs:

Athlon 3500+ Newcastle @ 263x9=2365MHz, 1.5V
Thermalright XP90 + 92mm panaflo + arctic silver 5
MSI K8N Neo2 Platinum
Corsair XMS 2x512MB TWINX3200XL, 2.5-3-3-10-1T @ 263MHz, 2.85V (this memory rocks!)
Thermaltake Tsunami case w. Antec Trueblue 2.0 480W
eVGA 6800 Ultra at stock settings/voltage

Alternatly what you can do to find out if its the CPU is to set the memory divider to its lowest possible setting so the DDR is running at its absolute slowest possible DEFAULT speed (DDR200 is what this board gives me). Do not leave the timings or voltage at your preference, make sure its all defaulted and the divider is the only thing changed. Increase the multiplier to its highest (11x), change the HT Link to 3X and increase the HTT in increments. Keep doing this until your system locks in a boot up. Take ~100mhz off that lock boot up speed you had, set the memory timings back to DDR333 (3:4) and loosen timings, and try running prime.

Try DDR333 (3:4) and do 2.5-3-3-7 2T or 2-4-4-10 2T after you find the HTT max with the above method.

 

[z-zar]

Thanks for all the feedback guys, all very helpful. Here's my update and response to suggestions etc:

Related to my comment that I can run 263x9 stable and thus conclude my memory can handle FSB=263:

This is not necessarily so. If you ran the small FFT torture test, memory wasn't actually used much because most data fit into cache. In order to truly verify your memory can function at that speed, run Memtest86 test #5 for at least 20 cycles, and then run Prime95 *large FFT* for 24 hours. It might turn out that you're giving your memory more credit than it deserves. Perhaps you need to use a memory divider at such high FSB speeds after all.

By default, I always use the large FFT as my stress test. I've used that for all experiments in discussed in this thread. For practical purposes, I've never had a frozen application running 263x9 on stock CPU vcore so I feel comfortable there. Nevertheless, I think at this point it makes sense to at least confirm 100% with memtest so I'll get that going.

Several people suggested using lower memory dividers like 5:6. I tried that but that actually makes my benchmarks slower so that kind of defeats the purpose...

How did your 11x220 test go? You should also try these combinations:

11 x 240 with mem divider set to 6:5 (aka 166 MHz, aka DDR333)
10 x 250 with mem divider set to 6:5 (aka 166 MHz, aka DDR333)

You probably won't need more than 0.125V above stock voltage for any of these.

My 220x11 failed with1.625V. However, the good news is that I successfully ran 6 hours of prime large FFT at 250x10 with 1:1 memory divider and vcore set at 1.65 (1.5+10%; anything lower will fail). I stopped it six hours into it. Temps were in the 55-56C range so still not burning (yet). The interesting - and slightly scary - thing was that I noticed with the 1.65V setting, both speedfan and corecell were reporting vcore at 1.70V! I also looked at MBM which agreed on CPU temps but generally reported 0.02-0.05V lower on CPU voltage so not sure which one is correct. I also noticed that the realtime vcore actually *goes up* from 1.65 to 1.7V as soon as I start prime95. Has anyone noticed these kind of voltage hikes when starting applications?

I also ran 267x9 at 1.65 vcore but that failed after two hours, suggesting that that maybe pushing my memory too hard. I'll test that with memtest tomorrow.

Finally, one last thing I noticed is that upon cold startup, prime95 fails quickly when I push my overclock. However, it seems to run better after the PC has been used for a while and has been warmed up. I can repeat the same prime test and it will last much longer the second time.

So, next up: memtest for my memory

PS. for the person having trouble with even 5% overclock: probably the cheap memory. I used to have Kingston value select which I couldn't even overclock more than 3% at default timings.

 

[Cytrix]

I also overclocked my processor ( 3500+ newcatle ) just to 2310, so I run prime on a torture test, but I have to go to work in a few hours but when my PC still runs prime95 and things go wrong does prime stop?

 

[z-zar]

I also overclocked my processor ( 3500+ newcatle ) just to 2310, so I run prime on a torture test, but I have to go to work in a few hours but when my PC still runs prime95 and things go wrong does prime stop?

yes

 

[NEODARK]

yeah it just stops ^

Now z-zar, let me give you some more insight into my experiences.

From what I've seen, the neo 2 is really, really picky about stuff you change on the bios, and I mean this.

I ran into the same thing with prime95 and coold boots vs a warm pc, not sure why this is, but I saw it happen. Dont put much faith into it though.

as for memtest86, I've ran my setup at 275x10 @ 2750MHZ 1.6vcore for over 24 hours doing full loops of mem test with 0 errors. HOWEVER, I would get errors in prime 95 and superpi almost imediately when it was 100% stable for memtest86 at least. So I just came to the conclution that for me, at least memtest86 would not be a way to test for cpu stability (only memory) guess thats why they call it "memetest"

your vcore will also go higher as the load increases on the CPU as its just pulling more power (if it ever goes lower under load, you have a weak PSU)

try this and tell me how it goes:

EDIT: I just saw you tried something similar to what I was going to suggest and it failed "220x11" you should really find out what your stuff is individually capable of first, then find out whats a good combination afterwards

 

[z-zar]

EDIT: I just saw you tried something similar to what I was going to suggest and it failed "220x11" you should really find out what your stuff is individually capable of first, then find out whats a good combination afterwards

OK, I took your advice and started doing a better job at finding individual capabilities of my RAM and CPU.

I ran 12 hours of memtest86 (33 passes; all tests) with 0 errors at 267x8 (i.e. CPU running below stock at 2.1GHz) with memory settings 2.5 3 3 10 1T @ 2.85V, and 1:1 memory divider. So that's good news, I think that takes care of confirming my memory capabilities.

Now onto the CPU. I set my memory divider to 5:6 and running 240x10 which makes my memory run at stock 200MHz and CPU at 2.4GHz. I started evaluating CPU performance while slowly raising my vcore:

vcore=1.500V: doesn't boot into windows
vcore=1.525V: boots into windows; fails prime95 large FFT in 4 minutes
vcore=1.550V: prime fails in 1 minute

My clockgen doesn't allow me to go above 1.55V so I have to reboot my PC now to raise vcore. I'll continue updating this list in the next post...

 

[z-zar]

here's the complete list with 240x10, mem divider 5:6

vcore=1.500: fails booting windows
vcore=1.525: boots windows OK; prime95 large FFT fails in 4 minutes
vcore=1.550: prime fails in 1 minutes
vcore=1.575: prime fails in 3 minutes
vcore=1.600: prime fails in 3 minutes
vcore=1.625: prime fails in 1 minute
vcore=1.650: prime fails in 3 minutes
vcore=1.675: prime fails in 14 minutes (actual voltage reading 1.70V)

At the last test, my temps went up to 58C so that's pretty much it for upping the voltage. The last result is a bit strange though since I successfully ran 6 hours of prime at 250x10 two days ago.

Another interesting test was in the tests above, at vcore=1.65, prime failed quickly when choosing large FFT. I decided to also run small FFT and that didn't show any problems until I stopped it at 21 minutes. Another round of large FFT failed again in 7 minutes. Anyone know what that means?

I'm pretty much at the end here with all these exercises. It just looks like I can't reliably get to 2.4GHz unless someone has any bright ideas. Bad luck with the CPU I guess.

 

[NEODARK]

Large FFT is memory related (I think) However I think your theory of just having a bad CPU is probably more accurate. Thats really a lot of volts you are putting trough that chip.

in comparation, I'm able to boot just fine into windows @ 2.7Ghz with 1.525 vcore. It wont be 100% stable, but if I kick it to 1.55 it is.

If you are giving your CPU that much juice and you still arent seeing any improvements, you either have a really bad clocker, or you are overlooking something else. Maybe you could post pictures of your bios settings?

Also you know all the other tweaks of the mobo right? (like using sata 3 and 4, not 1 and 2... also keep in mind your mobo could be the culprit as well, my neo2 had like 0 thermal compound under the northbridge, but AS5 fixed that. Also when testing your components, to me it seems you still have not ruled out the possibility of your mem craping out. When testing memory try this:

1- set CPU multi to 6X memory divider to 200 and 2.5-3-3-10 1T
2- set HT to 3X
3- kick up the HTT in 10Mhz increments
4- launch memtest86 and run test #5 in a loop for at least 10 loops, if that passes, boot your PC and run SupePI32 32M test, if it pases, run Prime95 LARGE FFT test, if it pases you are pretty much as stable as it gets. You can also use 3dmark2k1 "lobby" test in a loop, if that passes a couple of hours you are set.

for CPU:

1- put the cpu multi to 11X and HTT to 3X and vcore to 1.55 to start
2- put mem divider in 100mhz or 133mhz and 2.5-3-3-10 1T
3- boot into windows and run superpi 32M test, then try prime large fft. Use 3dmark2k1 "lobby" test in a loop

give that a try and post results

 

[2Fresh]

here's the complete list with 240x10, mem divider 5:6

vcore=1.500: fails booting windows
vcore=1.525: boots windows OK; prime95 large FFT fails in 4 minutes
vcore=1.550: prime fails in 1 minutes
vcore=1.575: prime fails in 3 minutes
vcore=1.600: prime fails in 3 minutes
vcore=1.625: prime fails in 1 minute
vcore=1.650: prime fails in 3 minutes
vcore=1.675: prime fails in 14 minutes (actual voltage reading 1.70V)

At the last test, my temps went up to 58C so that's pretty much it for upping the voltage. The last result is a bit strange though since I successfully ran 6 hours of prime at 250x10 two days ago.

Another interesting test was in the tests above, at vcore=1.65, prime failed quickly when choosing large FFT. I decided to also run small FFT and that didn't show any problems until I stopped it at 21 minutes. Another round of large FFT failed again in 7 minutes. Anyone know what that means?

I'm pretty much at the end here with all these exercises. It just looks like I can't reliably get to 2.4GHz unless someone has any bright ideas. Bad luck with the CPU I guess.

So prime fails..how do games run? Are you stable besides prime? If so, dont worry about it. I never even run prime. As long as my games, tunes, and web are fine... I am all good.

 

[z-zar]

So prime fails..how do games run? Are you stable besides prime? If so, dont worry about it. I never even run prime. As long as my games, tunes, and web are fine... I am all good.

Typically, for settings at which prime fails I can usually play ut2k4 for an hour or two before it eventually freezes and needs reboot.

 

[FanATIc]

here's the complete list with 240x10, mem divider 5:6

vcore=1.500: fails booting windows
vcore=1.525: boots windows OK; prime95 large FFT fails in 4 minutes
vcore=1.550: prime fails in 1 minutes
vcore=1.575: prime fails in 3 minutes
vcore=1.600: prime fails in 3 minutes
vcore=1.625: prime fails in 1 minute
vcore=1.650: prime fails in 3 minutes
vcore=1.675: prime fails in 14 minutes (actual voltage reading 1.70V)

At the last test, my temps went up to 58C so that's pretty much it for upping the voltage. The last result is a bit strange though since I successfully ran 6 hours of prime at 250x10 two days ago.

Another interesting test was in the tests above, at vcore=1.65, prime failed quickly when choosing large FFT. I decided to also run small FFT and that didn't show any problems until I stopped it at 21 minutes. Another round of large FFT failed again in 7 minutes. Anyone know what that means?

I'm pretty much at the end here with all these exercises. It just looks like I can't reliably get to 2.4GHz unless someone has any bright ideas. Bad luck with the CPU I guess.

Did you try 2T at all? May i also suggest you yank one of the 512 ram sticks.

 

[z-zar]

Large FFT is memory related (I think) However I think your theory of just having a bad CPU is probably more accurate. Thats really a lot of volts you are putting trough that chip.

For large FFT it reads "max power consumption and heat; some ram tested"

If you are giving your CPU that much juice and you still arent seeing any improvements, you either have a really bad clocker, or you are overlooking something else. Maybe you could post pictures of your bios settings?

Bummer, I don't have a digital camera for screenshots of my bios. Is it possible to make somehow make a screendump while in bios? Otherwise I can just write them down.

Also you know all the other tweaks of the mobo right? (like using sata 3 and 4, not 1 and 2... also keep in mind your mobo could be the culprit as well, my neo2 had like 0 thermal compound under the northbridge, but AS5 fixed that.

Yep, I use sata 3. As far as the NB heating, yes I read about people observing the marginal thermal compound situation. I haven't checked mine. It took me 2.5 hours just to install my new power supply with all the cable management a few days ago cause the whole mobo had to come out of the case to get the supply in so I'd hate to have to do that again to check my NB. But then again, if I have to I guess I will.

What I *can* say is that typically my NB temperature jumps about 5C within the first few minutes of prime 95. Typically it's running at 35C or below, depending on room temp.

Also when testing your components, to me it seems you still have not ruled out the possibility of your mem craping out.

So far I've successfully run 11 hours prime large FFT at 263x9 with 1:1 mem divider, as well as 12 hours (33 passes, all tests) memtest86 at 267x8 with 1:1 divider. All my cpu voltage tests in the previous post were done with memory running at 200MHz *with* relaxed timings so I do believe my memory is OK. Even if there is still some doubt, based on these results it's hard to imagine why my memory would cause prime95 to fail within just a few minutes if running OK for 12 hours memtest.

Do you have any experience with how the HT multiplier affects stability? I assumed that as long as I set it such that FSB x HT-mult. < 1GHz stability should not be affected. But that is an assumption at this point.

 

[z-zar]

Did you try 2T at all? May i also suggest you yank one of the 512 ram sticks.

Nope, haven't tried 2T yet. I'll give that a try.

 

[z-zar]

OK, in line with the suggestions from some of you guys I followed the procedures from cf-eclipse as given in this post

http://hardforum.com/showpost.php?p=1027479689&postcount=2

First, I tried to determine my max mobo speed. I put my cpu multiplier to 5x, my ht to 3x and my memory to 100MHz (1:2), then upped the FSB until running into trouble with prime95 large FFT. I made it up to 320MHz without problems and I called it a day there since that's plenty for my system. I took some screendumps of clockgen etc but can't figure out how to include those in my post. Bummer. If someone can tell me how, I'll post them later.

Next, I brought my FSB back to 200MHz, put my CPU multiplier at stock 11x, left my memory as in the previous test and then upped the FSB until prime small FFT failed. That happened at 222x11=2442MHz, at which time I upped my vcore to 1.55V. When prime didn't show problems within a minute or so I upped it to 227x11 at which time my PC froze. So the 222 (probably lowe) is pretty much the CPU limit.

Summarizing the tests I reported earlier:
* successfully ran 12 hours memtest (I stopped it there) at 267x8, 2.5 3 3 10 1T 2.85V
* successfully ran 11 hours prime95 (I stopped it there) 263x9 with vcore=1.5V
* failed prime95 typically within half an hour at anything above 263x9 with vcore=1.6V
* managed 6 hours successfully running 250x10 at 1.7V (!). Anything below 1.7 would fail, even at 240x10. CPU temp at 250x10/1.7V was 58C so still not smoking, thanks to my XP90.

Today I've been playing UT2k4 and Farcry for many hours at 267x9 without problems, although prime95 will fail within 30 minutes.

Someone asked me to post my BIOS settings. I don't have a digital camera but this is what I copied from the screen for my settings today. As I mentioned, not entirely prime stable at these settings but with FSB lowered to 263 it is:

DRAM settings : 2.5 3 3 10 1T 200MHz
High performance mode: manual
Aggressive timing: disabled
NV/ATI speedup: auto
DRAM drive strenght: normal
Dynamic overclocking: disabled
Spread spectrum: disabled
HT Frequency: 3x
Cool N Quiet: disabled
PCI clock auto detect: enabled
CPU ratio: 9x
CPU FSB frequency: 267
AGP frequency: 66
CPU VID: 1.550V
CPU voltage: by CPU VID
Memory voltage: 2.85V
AGP voltage: 1.50V

Please let me know if anything is out of whack with these settings (I think not), otherwise I'm calling it quits on my overclocking experiment since I'm out of ideas

PS. I *did* take the time to remove my mobo from my case (again ) and inspect the chipset thermal compound. It actually looked reasonable. I estimate 60% coverage of the Nforce chip. Either way, I replaced it with AS5 (before I ran the tests above). Not much difference.

PS2. anyone know where the "temp3", "local" and "remote" temperature points that speedfan reports are measured on the MSI Neo2 mobo? temp2=cpu, and I suspect temp3=nforce chip. How about the others?