ram passed memtest but locks up in 3dmark

ok i got my pqi turbo 2x512 DAH overclocked at 225mhz with 2.9volts. it passes memtest with no problems or errors but when i go to try to run 3dmark05, it craps out and my system is restarded. the highest speed i can run 3dmark at is 215mhz and even there, i get lower fps and lower 3dmarks05 than when it was at 200mhz.

now, im not a benchmark ***** or anything like that but having just built this system, i dont have any games to test it with (world of warcraft sub ended, need gamecard; still downloading hl2 off steam) so im using 3dmark05 to see how it does. im just afraid this problem in 3dmark will forshadow problems i will face when these games are installed.

now my oc settins when this happened was...

HTT: 225
multiplier : 7x (9x is stock)
ht multiplier (right word?): 4x
memory was at 2.9v and timings were at first set to auto but then later set to 2.5-3-3-7.

i tried multiplier at 9x after thinking it was my cpu but no, same problem.

any ideas on whats going on? bad ram? i thought about that but it did pass memtest...
__________________
DFI Lanparty UT NF4 Ultra-D 3/10
A64 939 3000+ CBBID 0507FPAW
Zalman 7000B Al-Cu
PQI Turbo 2 * 512mb DDR400 PC3200 ram
sparkle 6600gt pci-e
maxtor 20gb 7200rpm IDE
onboard sound
Enermax EG465P-VE 24 pin; +3.3V and +5v 35A, +12v 33v

It could still be that your ram running in excess of what it is capable of. Just because it passed Memtest at a certain speed doesn't mean it will boot or pass other tests at that speed...it is more of a max. It certainly does not mean it is 3D stable.

Try running it a 200, with the tightest possible timings and the 166 memory dividers. This will allow you to run your cpu at 240 with the ram still at 200. Test it and see how it does

One other thing is that PCI card your using. I don't know if that is part of the problem. I know it is a short term thing, but it may be causing you some issues. I don't know that I've heard of anyone else running a PCI card on an NF4 rig. Anybody you can borrow a newer card from just long enough to test?

hehe actually its 6600gt, i copied the specs from my signature off another forum and i forgot i didnt update it yet.

well acutally im running the divider at 150 (dfi mother board supports other ones besides from normal 200, 166, 133) now. so im running 260 * 9 @ 1.664v. memory is downclocked to 195mhz however, this is stable for me. i just got hl2 installed and theres no problems with fps around 70 with settings maxed. its also able to go through 3dmark without a hitch so should i just keep these settings and forget about overclocking the ram?

also, u mentioned tightning up my timings? there seems to be a lot of timings in my bios and i got now idea what it is... maybe someone can help me out...

DRAM Frequency Set - 150 (DRAM/FSB:3/04)
Command Per Clock (CPC) - Enable
CAS Latency Control (Tcl) - 2.5
RAS# to CAS# delay (Trcd) - 03 Bus Clocks
Min RAS# active time (Tras) - 07 Bus Clocks
Row precharge time (Trp) - 03 Bus Clocks
Row Cycle time (Trc) - auto
Row refresh cyc time (Trfc) - auto
Row to Row delay (Trrd) - auto
Write recovery time (Twr) - auto
Write to Read delay (Twtr) - auto
Read to Write delay (Trwt) - auto
Refresh Period (Tref) - auto
Write CAS Latency (Twcl) - Auto
DRAM Bank Interleave - Enabled

so i only changed the 4 i did know which i believe is making my ram run at 2.5-3-3-7 right now. i dont know what the other ones do...

If it passes memtest then it's probably not the RAM - How is your CPU doing, does it pass Prime95?

http://www.xtremesystems.org/forums/showthread.php?t=57317
you must have your timings right

cpu has passed 12 hours of prime95. my ram is rated at 2.5-3-3-7 and i believe thats what i got it at.

could be something else

any ideas, people? if its bad ram then i need to get some others...

Yes, how's your chipset voltage doing? Is it running hot or not enough voltage? Increasing FSB put a lot of stress on northbridge...upgrade your cooling if you must or increase voltage to it a little.

well odd as it may be, you may be giving it too much voltage. i had the same problem with my PQI TCCD where when i ran it at anything over 2.63v @over 200fsb, it would be memtest stable but not 3dmark01 stable. it was a VERY odd problem i thought. right now i run my PQI @ 2-2-2-5 2.5v 210fsb. i bench @222 FSB and it is stable there also.

worked for me so i thought i'd suggest it

Well, a lot of times people complain they pass memtest, but then crash at those memory speeds when in windows. The reason can be because memtest might not be increasing the ambient temps quite as much.

When you run 3d programs, your video card heats up, as well as your processor. There is a good chance the ambient temp will get quite a bit hotter, to the point where you may encounter a failure where there was none before.

Of course, it may not be the memory at all. A superpi 32M should be a good test. It generates a lot of heat, and tests the memory well. If you pass superpi 32M, there probably is nothing wrong with your Ram... or at least nothing that will quickly grind you to a halt.

ahh i see. but the thing is, i ran 3dmark05 right after turning on my comp. it couldbt possible have raised the temperature that much that fast especially since i have the side panel off as well.

Well, no side panel doesn't always make it cooler.. It can make it worse in some spots, where the air no longer circulates.

However, it is unlikely that the graphics card will raise the ambient temperature enough in a matter of seconds...

Still, I'd do a superpi 32M test.

Really we have a few possibilities here..
1) Your graphics card - take everything down to stock and see if it happens. You should only test one component at a time to determine an error.
2) Your memory - memtest is good, but also make sure you do a superpi 32M. If your memory has errors, they come up there.
3) Power supply - If things start getting hairy when there is a lot of power draw, but nowhere else, this could be your culprit.
4) Motherboard - This is your other component to look at when things start getting hairy under higher loads/power draw. Like all other components, its usually fine when its not doing much. There are a plethora of things that can go wrong, and it is a viable test.
5) Sometimes a program is just screwed up! In this case, that would seem rare being that its more than one... But you never know; You could have spyware, a windows operating system problem, and a whole bunch of other things I dont have time to list here.

So really there are a lot of things that could be wrong. The smart thing to do is to single each part out, one by one. Keep everything at stock first, then see how you do. If it fails at stock, start testing components. If it doesnt fail at stock, then slowly toggle each and every single value back in the BIOS ONE AT A TIME. Test after each change.

Let me know how you do at stock, then we can take it f rom there.

It hjas been proven that if you have a GOOD case fan setup that things run cooler with the case side on not off...]

Duke

I've had modules pass MemTest86 but fail Gold Memory, and vice-versa, but I've never had problems with any modules that passed both tests. However, one person said he had some memory that passed both tests but fail Microsoft's memory diagnostic.