X2 4200 and Prime95 unstable in OC and stock. (!)

[supertux1]

Hi,

 

I have a few issues, mostly related to stability and a few questions related to hardware.

 

I've been having difficulty getting Prime95 stable at stock and overclock speeds on the dual core X2 4200.

 

I have tried both the 623-1 and 704-1 BIOSes, cleared CMOS and all that basic stuff. I have everything at 'optimized defaults' in the BIOS, Genie screen, except DRAM timings and divider for various tests. My RAM sticks are in the orange slots.(Does anyone know what kind of RAM I have (check sig) ? I'm not sure I'm using the right -1 -2 -3 option BIOS. It's cheap crap, I know that for sure.) SPD timings are 3-3-3-8 so that's what I use for stabiity.

 

I do my tests on a plain Windows box with all the necessary hardware drivers loaded, for a disk I use the Silicon Image RAID 0 with two 36GB 10K WD's. (I could not ever get the nVidia RAID 0 or even SATA working -- BSOD's or hangs the Windows XP install while still in text mode, but that's a separate issue.)

 

Anyway, the deal is this, I have two installs of Prime95 in separate installation directories such that I can run separate processes each with an appropriate core affinity. I run core 0 doing the mix test which tests the maximum memory, and core 1 doing the maximum heat test -- haven't tried maximum FPU tests much yet.

 

At stock I run about 1.35v @ 38C and maximum overclock is 1.55v @ 50C. At all 'stock' defaults, eg no tweaking, the best I've done with two Prime95 instances before one of them fails is 5 hours. One of the cores will eventually fail with 'rounding was 0.5, expected less than .4' Sometimes the .5 is .49 and a long string of more numbers.

 

Overclocking yeilds the same results, only faster, usually within 25 minutes.

 

Here's my best overclock to date which ran dual Prime95 for 25 minutes.

 

246x11 2706Mhz 1.55v 50C

133/200 DRAM (159.2Mhz) 2-2-3-5 @ 3.2v

 

Incidently, I played several hours of FarCry on this configuration, AquaMark3, SiSoft Sandra, 3DMark05 without issues. I'd like to say it's stable, but Prime95 says it ain't so...

 

I actually got 2.8+ ghz working but was only stable enough for a CPU-Z screen.

 

I am afraid to go above 1.55v on air.

[Keno5net]

Check out this thread:

http://www.dfi-street.com/forum/showthread...ime95+instances

 

I got my 4200X2 stable 24hr at 240 but my memory won't keep up. The above thread will explain how to set up and use Prime95 to test a dual core proc.

 

Ken

[AdamT]

I'd RMA the cpu if it can't pass Prime at stock speeds.

[THunDA]

SurePower 600W 24pin PSU

 

Do you have a link to that psu with its specs ?

[sasquatch1313]

Aside from the above valid suggestions, I would also look into the ram. "Value" ram has historically not done great in the board. Have you ran Memtest to evaluate it? I would suggest setting the cpu and ram to stock settings and making sure there are no errors in Memtest before worrying about the cpu. Both the cpu and memory are stressed by Prime95 so either could be the culprit.

 

In response to one of your other questions, I am not sure exactly what chips are on your ram, but in general a Bios that is -2 plays the nicest with the widest range of ram. It might be worth trying out 704-2BTA (check the sticky in Bios Factory) since its optimized defaults are a bit more compatible.

 

Good luck

[supertux1]

PSU Specs:

 

http://www.mdmm.com/products/techietoyz/po...rePower_600.asp

 

and

 

http://www.mdmm.com/products/techietoyz/Specs/MD-600SPCS.htm

[supertux1]

Aside from the above valid suggestions, I would also look into the ram. "Value" ram has historically not done great in the board. Have you ran Memtest to evaluate it? I would suggest setting the cpu and ram to stock settings and making sure there are no errors in Memtest before worrying about the cpu. Both the cpu and memory are stressed by Prime95 so either could be the culprit.

 

The way I test memory settings is I run test #5 (block moves) and 6 (inversions) in memtest.

At 98% in test 5 it will start dumping errors if things are not right.

I go straight to test 5 because I've found a wide range of seemingly optimal settings will

pass tests 1 through 4 but fail on 5. Or pass on 1 through 5 and fail on 6.

So I go straight to test 5 and if that works I do test 6, and if that works I boot into Windows.

 

How long do you guys usually let memtest run for before declaring the ram stable?

 

In response to one of your other questions, I am not sure exactly what chips are on your ram, but in general a Bios that is -2 plays the nicest with the widest range of ram. It might be worth trying out 704-2BTA (check the sticky in Bios Factory) since its optimized defaults are a bit more compatible.

 

Thanks, I'll try that. I've been having better luck recently with all of the DRAM settings I'm not familiar with set to 'auto', and just messing with the divider and the usual 4 timing values.

I'm currently testing @ 246x11 with 3-3-3-8 timings and the divider 100/200 to rule out

any possiblity that it might be the CPU. I've seen other posts that say 246x11 isn't

prime stable but 238-240 is, so I'll try that next if this one fails.

 

What's the difference between 'auto' on everything and optimized defaults?

Is there something that says how 'auto' actually 'calculates' a value for each particular setting?

[sasquatch1313]

For memory testing the general response around here seems to be run all tests for 8 hrs (overnight) in a loop. On my system I got around 14 passes by running for 8 hrs. It checks to make sure the system doesn't become unstable overtime due to heat buildup.

 

And before anyone else mentions it, the best advice is to make sure the system passes Memtest, Prime 95, AQ3, and 3 3dMarks (some would add S&M and OCCT) at completely stock settings before ever overclocking.

 

Keep at it, I built my system a couple of months ago and it took me a week to figure out the bios and how my system would run the best. You will get there!

[nghisus]

I'm running at 2.6Ghz with 2 gigs of Patriot XBLk's at 238 FSB. My LDT/FSB is at 4x. I'm running also at 1.42 (1.425 + 104%) volts which is giving me temps with ITEGuardian at 32-34 C. Mem timings are 2T (4 sticks) at 2.5-3-3-7. Runs stable with 3dmark and hours of B.o.S 64bit and EQII. I can run tighter timings at 2.5Ghz, 2-3-3-8 i think. Still pushing in though, but i heard takes sig more voltage to run at 2.7Ghz and above.

[supertux1]

I'm running at 2.6Ghz with 2 gigs of Patriot XBLk's at 238 FSB. My LDT/FSB is at 4x. I'm running also at 1.42 (1.425 + 104%) volts which is giving me temps with ITEGuardian at 32-34 C. Mem timings are 2T (4 sticks) at 2.5-3-3-7. Runs stable with 3dmark and hours of B.o.S 64bit and EQII. I can run tighter timings at 2.5Ghz, 2-3-3-8 i think. Still pushing in though, but i heard takes sig more voltage to run at 2.7Ghz and above.

 

The higher you go, the less linear it is with CPU speed and voltage. It's more exponential.

For example, at 2640Mhz I've been able to run it at 1.5v no problem, but to go that

last 50Mhz requires 1.55v and to get even close to 2750 or 2800 requires well

over 1.6v. Of course the heat increases with voltage. I've got a watercooled setup

now, so I'm going to see if I can get 2.8Ghz stable.

[DeepGan7]

Hi

 

Have the same "problem" here

 

Running 264x10@Vcore=1.5 fully stable

270x10 requires Vcore>=1.6

275x10 totally unstable even with Vcore=1.675

[nghisus]

The higher you go, the less linear it is with CPU speed and voltage. It's more exponential.

For example, at 2640Mhz I've been able to run it at 1.5v no problem, but to go that

last 50Mhz requires 1.55v and to get even close to 2750 or 2800 requires well

over 1.6v. Of course the heat increases with voltage. I've got a watercooled setup

now, so I'm going to see if I can get 2.8Ghz stable.

 

I think you definitely hit the nail on the head with that one. I do believe past 2.6 is where that curve starts taking that upward turn in terms of voltage. I've been curious to see how far I can push this thing on air, I really want to see your results on water, but I'm really curious how far you pushed this thing with the XP-120. ITE Guardian notes me at 2.62 at 1.42V as does CPUz. I've run 3dMark stable but haven't tried Prime yet. I figured if it runs stable enough for me when I'm playing my favorite games, I really don't care too much whether it's prime stable or not (blasphemy?!).

 

Side note, the first time i overclocked i had this bugger at 2.5 Ghz, with timings of 2-3-3-7, but now i can't seem to get it to boot at 2.5Ghz with my CAS tightened up to 2. I'm running Patriot so i heard it dosen't like anything more then 2.7 anyway. So any reason you think?