New Project - New Problems or The day the Reaper died

Question

Ok, this is more like telling a story than a cry for help.
The story is about a new project, new (great) results and the disaster that followed.
Well story starts October 21st with me upgrading rig 1 with new parts. They came for my Opteron165 setup which oc'ed nicely but didn't make it far enough. So board, cpu and ram of rig 1 (sig) got replaced by the following parts:
The new cpu is an Athlon X2 5200+ EE (TDP 65W), 2.6ghz, 2x 1024kb L2 cache, 90nm, F3 stepping.
The board is from MSI (of course), a K9N SLi Platinum. Very nice board and luckily it is PCB version 1.2 which features a heatpipe while the older version just had a very bad cooling, passive aluminium heat sink. The older version is the one still to be seen on MSI product site.
The new one looks therefor way better and stays cooler:
In addition I threw a pair of the finest ddr2 ram in it. A 2gb dual channel kit of OCZs new DDR2 PC2-6400 Reaper CL3 Edition. DDR2-800 with 3-4-4-15 (stock), guranteed to work with 2.3volts.
So suppositions for oc were great and I started:
First of all upgraded the bios from 1.8 to the latest 1.10. Then I thought why go step by step, why not starting big. So I set HT clock to 250mhz, resulting in a cpu clock of 3250mhz (multi 13), Hyper transport link set to 800mhz (x4), ram very relaxed (divider 667, 5-5-5-15 2T). It didn't post but I thought solution is easy, forgot to set extra voltage for cpu. Next try at 3250 mhz with 1.25v and set 0.1v extra in bios. It didn't post. I set voltage to maximum but no luck. Tried again but this time I lowered the HT link to 600mhz (x3). Now it posted even with 1v extra and it also booted XP but seemed to be not very stable. So set extra voltage to 0.15v which enhanced stability greatly.
After minor driver problems my old XP installation worked pretty perfectly with the new setup. Time had come to see if it was stable and I started Orthos. It seemed to be ok, 1hour no problems, 2hours, 3hours and bam... 3hours 40 minutes and that's it. I thought "Damn, but it will make it with more voltage". I added more voltage (0.2v extra) and started Orthos again but no way, at 4hours 32 minutes it stopped again.   
I'm not easy to discourage and so I added 0.25v to the 1.25v vcore and tried another time. Still 250x13 = 3250mhz.
And it worked!
Orthos now easiely worked six hours when I stopped it myself to see if it would also be stable with more intensed ram settings. I set ram divider to 800mhz (1:1) resulting in DDR2-929 and timings to 4-4-4-15 2T.
I started Orthos and left the machine alone... Outstanding 16 hours later it was still working!
After 17hours 5 minutes and 50 seconds of running Orthos I knew I had a stable beast!
It also worked with ram set to divider 667 (resulting in 406mhz or ddr2-812) and 3-4-4-15 2T. However 1T was a real killer, thats why I left command rate at 2T and were happy so far.
Performance was really good and I wanted to know if this would increase my 3dmark06 results which were already great with old setup because of nicely oc'ed 8800GTS 320. The opteron could do 9572 3dmarks with it.
The new setup easiely broke the 10.000 points by reaching 10306 points with cpu at 3250mhz, ram at ddr2-929 (4-4-4-15 2T), vga as seen in sig.
It really seemed to be finished. The rest would be just playing around and going for suicide clocks. At least that was what I thought at that time.
Before maxing it out I wanted to have the Vista 64 Ultimate installation also running with the new setup. I hadn't booted vista since upgrading to the am2 system, so I tried for the first time....
AND NOW EVERYTHING GOT F***** UP
Vista welcomed me with a blue screen instead of a logon screen. I was pi**** but aware it might not like changing from NForce4 SLi to NForce 570 SLI. Next step was trying safe mode but vista thought of blue mode and showed a screen of that colour. I thought "ok, maybe you dislike setup change" and so I went for a clean install. I even formated the partition and gave it a go. Installation was fine 'til first start. Well, it wasn't a start to be honest as vista told me it was not able to do this. Repair wasn't successful either.
Assumption was now system is not making the oc clocks although proved to be stable under XP. So I removed any oc and loaded fail-safe mode in bios. Relaxed ram timings to 5-5-5-18. I started installing vista again, thought I could oc again when it is installed.
Oh boy, how wrong I were...  The 2nd installation turned out to be as faulty as the first and I tried again and again... After installation failed nine times I were finally discouraged.
I just wanted to boot XP for now, forget about installing vista and maybe ask here if it this was a known issue.
Well, I didn't get to see XP. I was just informed about a missing file that also prevented to boot safe mode. I took my XP cd and started automatic repair. After this was nearly finished the first errors and notifications about unknown, missing or corrupted files appeared. Although XP made it to finish install and booted successfully. But this XP wasn't a real XP as many things didn't work. I couldn't install drivers because loads of errors and crc crap made working with it impossible. Clear install time! But no way, same issues like seen in vista.
Now I were pretty desperate and I urgently needed an idea.  I came to the solution maybe I could try with single channel, using just one mem stick. I removed one and tried installing XP again. It worked without any issue. Even driver installation and other software worked without a problem. Now I were even able to install Vista.
But still I didn't know what caused all the trouble. I had three suspects: Mobo, ram, cpu. I doubted in faulty board but cpu (failing to control dual channel) or the ram stick I removed were my favorites. As it was the easiest to try I put the other stick back in but used a slot combination that would cause single channel.  Booting Vista now turned out to be a disaster. Errors, corrupt files and so on. Same with XP.
Conclusion: Dual channel is not the problem. Added mem stick was now suspect no.1. I took the first one out and started memtest. Memtest didn't show errors after 8hours. I wondered and started Vista. Result was the same as seen before with two sticks. Errors and access violations. Replaced this stick with the other one and everything was fine again.
Next day I tried with a borrowed MDT 2gb dual channel kit which also worked without a problem. 
Sad but true one of the Reapers is K.I.A. . Damn....
Now I'm sitting here with a borrowed kit of value ram which luckily allows me to let the cpu run at 3250mhz with mem divider at 667mhz (result: 406mhz) and which I can luckily keep 'til I get another kit of Reapers. I have to send my current Reapers back to vendor these days and wait for the rma process and wait and wait....
So what started greatly ended not so great. For now I'm sick and tired of seeing bios screens or installing windows and I'm also pretty pi**** because of all the data I lost. Some crucial ones among them.
Main problem remains: I have no idea WHY the Reaper died. I never set any higher vdimm than 2.3v what should be ok for it. It never felt hot and was directly in the main airflow. Also I have no clue why memtest didn't find any error with the faulty stick.

Answer 1

Quote from: flobelix on 30-October-07, 06:12:31
Be aware that you have to keep an eye on shader clocks when oc'ing the gpu. With common tools you're oc'ing gpu and shaders at the same time. Since I'm suspecting a misproportion of gpu/shader clocks to be one of the reasons for problems with factory oc'ed Geforce 8 you should try with different combinations. With the latest Riva Tuner you can change 'em seperately without having to mod the bios.
Memory is at it's end I suppose, that's the problem with most high end cards, no potential. Seems as you have to live with that.
That's true for HD2900 series I think but as far as I know 8800GTX needs it even for stable working.
Could be your psu is too weak, maybe really time for the new Corsair. On the other hand should be strong enough I guess.
You're ungrateful! That good old 3500+ oc'ed so well and was once a real performer and you call him crap.... bad boy.
"You're ungrateful! That good old 3500+ oc'ed so well and was once a real performer and you call him crap.... bad boy. "
yes you are right the CPU perform great, but its just old..(is not CPU fault in direct sense )
"Be aware that you have to keep an eye on shader clocks when oc'ing the gpu. With common tools you're oc'ing gpu and shaders at the same time. Since I'm suspecting a misproportion of gpu/shader clocks to be one of the reasons for problems with factory oc'ed Geforce 8 you should try with different combinations. With the latest Riva Tuner you can change 'em seperately without having to mod the bios."
Yes im aware about it, but not this is the case.. i got instability even with 2Mhz on the GD more...SD remain @ stock 1350 and memory stock as well.. and go instability...
Reversed story if i put everythink is stock, but +2Mhz more on the memory i got the same instability....
"With the latest Riva Tuner you can change 'em separately without having to mod the bios."
hmm nope here, its locked increasing GD clock inscrease SD clock as well. they are locked and increased in parallel.(etc, both in same time)
"Memory is at it's end I suppose, that's the problem with most high end cards, no potential. Seems as you have to live with that."
mem. chips used are for 1000Mhz the VGA stock, MSI is not OCed them.. so they must have potencial.. also VGA is adverted with DOT... at least its designed to handle till 5-6% OC, i have done few review(don't trust much here,but anyway its an info) and user opinion no problem to OC till 2100Mhz(1050). So i doubt that is memory problem.
"Could be your psu is too weak, maybe really time for the new Corsair. On the other hand should be strong enough I guess."
Yes that is the only conclusion which i can stuck at the moment... Good news is that i have talked with distributor today and they ensured me that will delivered tomorow for me. So tomorow i will know more about that mistery when i install the more powerful(and hope so great) PSU.