truly stumped by prime05 errors :confused:

[deja100]

Hi all. Im truly stumped by the prime95 errors im getting error pic (154kb so i linked), and was wondering if anyone more familiar with prime95 may have some insight.

 

First a small background. If the parts seem outdated its because i built this system a year and a half ago. Ive been fighting with it on and off ever since. Never could get it running at 500mhz FSB, till i discovered DFI anyway :nod: After i got that working it was solid, until i started playing games or anything that used heavy graphics. turns out the problem was that my ANTEC Truepower480 couldnt supply enough power to the system once a meager radeon 9600 started going all out. Never thought id see the day when 480 watts of power wasnt enough, lol. since then ive replaced that with an OCZ powerstream 600watt beast and a Geforce 6800 GT and that problem is solved. anyway, the full current lineup is in the sig. now to the problem.

 

Im running an Athlon XP 3000+ (Barton core, rated to run 2100mhz). Through Prime95 testing(test 1 to test processor only and lowering of mem/fsb to 200) ive established that the ceiling on this core is somewhere between 2255 mhz and 2300 mhz (at least without raising the voltage beyond spec). The memory is 1GB(2X512MB running in dual channel mode, slots 2&3) OCZ PC4000 Gold. Through sisoft sandra burnin (using successive memory benchmarks only and lowering proc to 1500mhz) ive established the ceiling on the memory at 504mhz (again without raising voltage).

 

My current setup is with FSB/MEM at 1:1 set to 250x9=2250 mhz. With this setup im 15 hours prime stable (test 1). could have gone longer but i shut it off because i needed to go to bed and computer makes the room too hot, lol. a proof of this run is here proc 2250 mhz FSB/MEM 500mhz stable 15 hours pic is 365kb I also ran sandra burnin for 9 hours straight. again i could have gone longer but i needed to turn it off for the same reason. proof here: 2250/500FSB memory stable pic is 240kb ish.

 

So heres the actual problem. Im pretty convinced that the proc and mem are stable at 2250mhz and 500mhz respectively both by themselves and when run together. the problem came when i ran prime95 with test2, which uses some memory. problem is it fails within minutes of running it. Im totally confused as to why this happens, as my sandra runs seem to prove that the memory is stable. over the years this has always proved to be a reliable test of memory stability in my experiences.

 

heres my complete list of BIOS settings:

*FSB/MEM 1:1

*FSB: 250

*multiplier: 9

*AGP clock: 66mhz

 

*vcore: 1.65 (tried upping this, didnt have any effect except to heat up my CPU more)

*vAGP: 1.5

*vmem: 2.8 (these older modules are rated for 2.8v max, not 3.0 like newer OCZ modules)

*vchipset: 1.8 (1.9 didnt improve anything so i lowered it back)

 

*CPU Interface: Aggressive (Optimal didnt improve anything)

*Command per clock: off (unbootable when enabled)

*super bypass: enabled (unstable when disabled)

*sync mode memory bypass: enabled (just slower when disabled and doesnt improve stability)

*data scavenged rate: FAST (unstable when set to normal, go figure)

 

these are the rated timings of the modules, raising them doesnt improve things and lowering them makes it worse

*TRAS: 7

*TRCD(read): 4

*TRCD(write): 4

*TRP: 4

*CAS: 2.5

 

I read somewhere that TRC should be set to TRAS + TRP -> 7+4=11 and that TRFC = TRC + 2 so ive set them respectively. someone correct me please if that info was incorrect. I tried playing with them nonetheless and again nothing improved.

*TRC: 11

*TRFC: 13

 

Oh the mystery timings. Ive tried these on both auto and manually set. Ive currently got them manually set. got the original numbers from another persons system, then had to raise a few to get sandra stable. nonetheless ive tried raising anything and everything and no improvement.

*TDOE: 2

*TRRD: 3

*TW2P: 3

*TW2R: 3

*TREXT: 3

*TR2P: 3

*TR2W: 4

 

*Driving Stregth: originally set to auto for used DIMM slots. tried playing with the whole range of options and stability was either the same or worse. put it back to auto on all DIMM slots.

*Slew rate: auto (dont know what this is so i left it ALONE, lol)

 

dont know how this could affect things but included it for completeness

*AGP Aperture: 64MB

*AGP 8X support: enabled

*AGP fast write: enabled

*CPU thermal throttling: disabled

*system bios cacheable: disabled

*video ram cacheable: disabled

 

whew, man i wish BIOS's had a setting dump feature, lol. anyway, ive played with just about anything that seems to have relevance and the stability is the same. It always crashes within minutes of starting the prime95 test 2. test1 runs like a charm and so does sandra memory bench burnin. the proc never gets above 50C and system temp never above 37C. Further the system never crashes, lockups, restarts, etc. i can play UT2004, code and compile programs, etc all day long without problems. nevertheless theres that little hint of instability there that i would like to get rid of. I found that this problem goes away when the mem is set to 484mhz or lower, but the modules are rated for 500 and i want it, lol. anyone have any ideas. sorry for the long post but i wanted to give you all as much info as possible.

 

thanks

[psy^]

Damn, sounds like you've been wrestling with that system for a while man. These computers sure can be biatches when they want to hey :nod: :eek2: :shake: :sad:

 

it's a different platform i know, but give these timings a try:

 

I got my RAM about a year and a half ago. Maybe through some freak occourance you and i have the same chips

[Psykoikonov]

At 200Mhz fsb my mobility barton will do [email protected] prime stable 12 hours. At 250Mhz anything over [email protected] will fail prime quickly (within 30min) and trying 2.6Ghz will fail prime on the first test everytime. I'm pretty sure you found out what alot of us have and that's that your CPU is stable at 200Mhz fsb much higher than it is @ 250Mhz fsb. Now the trick is to find the happy middle.

 

Psyko

[Mr Stom]

I am amazed that you have got that cpu to run that quick without increasing the vcore.

 

Personally I would try and back off the clocks a little and work your way up to 250Mhz, getting it stable in prime at each step.

Have you had prime stable at stock speeds on your rig? I have found that some bioses (esp 1/31) would not provide prime stable clocks at any speed.....just a thought

 

Try dropping your multi to 9, FSB to 166/200, whatever your default is, and work your way to your target giving a little extra juice as needed as prime fails along the way....

 

Mr Stomly

[deja100]

I dropped the memory clock to 100FSB(200 DDR) and tested the proc for stability. 200X11=2200 was stable on test 1, 200X11.5=2300 was not (without raising vcore anyway) after that i lowered the CPU clock to 1500mhz ish and tested the ram. its stable on both test 1 and test 2(the one thats giving me problems) up to 242 mhz (484 DDR). thats when i started playing with all the settings i mentioned, and had no luck. thanks for the thought. im not going to push the vcore at the moment. i have a fairly nice heatsink on it, which is currently keeping it at 50C max at full load. i dont like have my procs past 55C and im sure it would get that high at current settings if this was the summertime. im looking at getting a pure copper sink for it at some point. may push it some more after that. thanks for the thoughts though.

[klrgreenis]

Sounds wierd but Mr. Stom has a point I was able to go up to 245fsb with 2 x512 sticks, but I would have to start a 235 and keep uping it one fsb at a time, if I would try to just go to 245 it would fail prime...

[deja100]

ive tested by bumping it up in 1mhz increments and found the highest stable clock to be 484FSB/MEM. thats not too far off from the 500 mark so i guess thats not bad. In all fairness when OCZ sold these chips a year ago they did say that these sticks were intended for the intel chipsets (i875 i believe it is?) and i knew i was taking a chance. so i suppose its possible that these memory chips just dont work well with the nf2 chipset. i guess at this point i could test them some more, maybe see if i can get better results from lower timings and clock speeds. something to play with i guess. anyway, thanks for your help

[Mott13]

I have allways found that rounding errors usually mean you need more vcore so git yurself

a better HSF and try again:)

Oh and I wouldnt worry bout giving that ram some more volts either I had some ocz 3500el

that ran 3.2@230 fsb for months...and mine were rated for 2.8 also...

[persivore]

Have you tried running memetest? I've found that I can run prime on the CPU intensive test, even if memtest errors in the first pass.

Tbh, it sounds like the chipset could be maxing. Try using a lower Vdd because i've found that running lower Vdd can sometimes benefit more than running high Vdd.

Also, do you have a fan blowing over the RAM/northbridge area?

[deja100]

ill try lowering chipset voltage and/or raising vcore and see what happens. thanks guys. as for cooling, i have this case: N82E16811133106 - item number on newegg.com (id post the link but ive found that newegg links sometimes dont work later on after saving them. ) as you can see theres two side fans on the case. the lower one is blowing air over the bottom of the board, hitting the SB nicely and somewhat on my video card. the top fan is hitting the NB directly and a decent amout of air on the memory too. so im pretty sure cooling is under control. i also swapped out the relatively slow fans on the 2 side and the 2 rear fans that came with with this case N82E16835999110. loud little suckers till you get them on a fan speed controller. anyway, ill play with it some more and let you know what happens. thanks