Aging OC XP-M Barton requires more voltage?

[SniperMerc]

Sience you allready have heat spreaders on your ram, I wouldn't hesitate to pump some more voltage to them, 2.7-2.9 should be easily doable.

Definately bump that cpu voltage up, XP's can handle a good ammount, I find that with my setup 1.8 Vcore keeps me stable, if a bit warm @ 12.5x200 for 2500 MHz but it is doable considering how much airflow I have in case. For CPU sink I'm using the Zalman AlCU.

During warmer days (75 F and up) if I start doing any gaming I start cranking up the fans on my fan bus, cpu starts to get finicky around 55C so I like to keep it under that.

BTW I don't think I ever saw what cpu sink your currently using, post it up.

As for Prime, I don't even bother using it any longer for long term testing, it's not as helpfull as people think. I like to run several hours of 3d mark 2k1 looping, followed by a cpu stress program that a friend gave me, I'll only run memtest if I feel like there is an actual memory problem. Also good one to put stress on the GPU is the real time high dynamic range lighting demo, that thing will just bake a gpu and make your CPU swealter with the heat coming off the video card. Running rthdrld along with the cpu stress program just butchers both at the same time.

Even with all that testing tho, the only real way to make sure your system is running properly is to use it. I've built systems that I've overclocked and ran every concievable test on before shipping to a customer, had it show up at their front door and start crashing a few days later.

http://www.daionet.gr.jp/~masa/rthdribl/ the lighting demo, it will run on dx 8 based cards and stress the hell out of em, but it wont be as pretty as it would be on dx 9 cards.

Edit: pic of the lighting demo running on my sys with mbm 5 temps http://snipershide.us/temp/lighting_demo.jpg

 

[trexpesto]

I have no trouble failing the stress tests, so I may as well stick with Prime, yes? Memory-intensive Prime test failed at 2 hours+ with 2.7 VDimm, 1.65 VCore, MMX440 GPU without drivers installed, and just the PSU, CPU, and one 80mm case fan pointed at my NB passive sink.
The HSF is a copper Thermalright SLK947U with a "54 CFM" 92mm fan. 18 month old AS5. Temp hit 51 on a hot day with 29C board temp. Zalman passive NB sink gets warm but not hot to the touch with the case fan blowing on it.
The original OC was at 216x11 @1.65Vcore, which I am trying to get back to. Maluckey pointed out that I may have a different chipset driver now, BIOS rev possibly too. So maybe 1.7 VCore is inevitable, but the failures seem to point at mem/NB so I am starting there, slacked RAM timings and adding VDimm so far. No prob with 9+ hours of memtest at 2.6V Maluckey also said he runs OC at 1.5V chipset, which is weird because my defaults to 1.6 when I load failsafe BIOS settings.

One question: if Prime runs twice as long, does that mean it's really more stable? With the increase in VDimm I got twice as long before failure. 2 hours with 2.7V versus one hour with 2.6V. Trying to be systematic here.

Also with PC3500 ram at 216 FSB, it isn't overclocked and slacking the timings alone only got me an hour of Prime, so growing more suspicious of the NB.

 

[SniperMerc]

From everything I've seen you post so far, it's all starting to center on your ram, could be a bad or degrading module that is the problem.

Try 3-3-3-8 on mem timings and see if it stabilizes.

Also something of note, on NF2 Ultra chipsets, there is a weird but noteable quirk when the TRAS is set to 11 or 12 that will actually increase the bandwidth and stabilize the ram better. So you may want to try 3-3-3-11 or 3-3-3-12 might even try going back down to cas 2.

 

[trexpesto]

Failed at 1.7VCore, 2.7VDimm, timings 2-4-4-9. Memtest ran fine for 9+ hours.

I'm going to go back to basics, take everything apart, clean and reinstall sinks with new AS, start at fail safe, memtest, etc. Thanks for the support.

 

[maluckey]

trexpesto,

Are you in dual channel?? If so, Tras 11 is the money maker (as mentioned by Sipermerc). Seting it too low doesn't allow sufficient time and errors out. With my setup, less than tRAS 10 in dual channel @ 200 mHz FSB is unstable.

I also feel that much of your issues center of RAM. RAM can go bad at any time. Try removing one of the modules, or swapping out one you have for a known good one and see if at 2.8v in BIOS if you can stabilize @ 200 mHz FSB. Overclockers RAM almost always needs more juice to do the job...

 

[trexpesto]

I will ty it, but it's not overclocked. PC3500/DDR 433. It's actually at 432, reading in cpu-z.

I am afraid its the northridge, hope not. Because I would be somewhat happy to get two 1GB sticks; my virtual memory is always kicking in on BF2. I have already reseated them BTW.

 

[trexpesto]

heh at 3.3.3.11 it ran for 0 minutes, tried twice. Running OK so far at 2.3.3.11.
Self-test 1 passed.

I still have both modules in there.
Stock voltage and stock speed at 2.6V/432MHz because it is PC3500 == DDR 433
VCore is back at 1.7 in BIOS, cpu-z reports 1.664 - 1.68V, saw one drop to 1.648

 

[maluckey]

I'ts PC3500 == DDR 433

Yes, and mine is PC2700...It's all in the chips, not the ratings. Just because your are using PC3500, doesn't mean that you don't need more juice to stabilize the RAM. My particular RAM was also rated as PC3200, PC2700, PC3000 etc. Be that as it may, and despite the rating of 2.7 volts, 2.8 volts is often used by most for stability with these particular modules. It has always seemed that the higher the CPU mHz, then the more work that the RAM is asked to do as a result of faster operations and processing power.

That's just my take on it. Also, timings greatly afect the RAM and it's voltage ratings. At 2-3-4-12 I am able to run lower voltages than at 2-2-2-11. It's the nature of the beast. The tighter the timings, the more voltage. Some RAM just won't work at ANY timings other than SPD. Yours may be that way...

 

[trexpesto]

They used to run stock just fine. Dratted entropy!
2.3.3.11@2.6V == 14 hours of Prime so far, but got the same reslut off just upping to 1.7 VCore.

I wish we had better OC tools. I feel like I am using a small sledgehammer when I need an ice pick.

 

[maluckey]

The best OC tool in the world is the one on your shoulders. Use that and disregard anything that fails the "stink test" and your on your way to stability in your OC....(again).

I wouldn't do any more Prime95 at this point. I would run the hardest games that I can find, and if you don't crash, then you are likely good. A resource hog game can crash you faster than Prime95 anyways. so why wait 14+ hours to see results at this point?

I normally run S & M for three hours and Prime for a couple three. If those pass, then large file transfers while encoding video and playing intense games is next....

 

[trexpesto]

14 hours doesnt bother me too much, I do it while out of the house. But point taken; I'll put my vid card back in and try hitting BF2 and HL2

I snugged my timings up to 2-4-4-7 at 2.6V, ran all day. May end up trying 2-3-3-6 or 7 at 2.8 if everything goes well. Thanks for your support.

 

[maluckey]

Single channel?

 

[mindless1]

Something not yet mentioned is that RAM and CPU do not usually need higher voltage (than previously) to run at the same speeds so long as temps stay as low. HOWEVER, you can't just keep using the same motherboard and have any meaningful comparison!

Two words: Electrolytic Capacitors. They degrade. It's proven, even spec'd by the cap manufacturers that their parameters degrade during normal use, during sitting on a shelf, but especially in an overclocked system that puts more ripple current through them. Raise your voltages? You wear them out even faster.

The one hope to prolong a board as much as possible is to keep the caps as cool as possible. You can run an XP chip at 1.85V no problem with the aforementioned 'sink and fan, but that's not what the board was designed to sustain long-term. PSU caps wear out too, though less often is it a noticable ceiling clock-speed reduction rather than a pop one day, or that more caps on the motherboard see even more ripple.

BTW, there are a lot of XP-M CPUs that can't do 2.5GHz on air, not stabily though extensive stress-tests. Certainly quite a few can but it's not uncommon to find one that won't go much past 2.4 unless the voltage is ramped up really steep like past 1.9V which is too high for sane air-cooler use at typical ambient temps.

 

[trexpesto]

Running DC with the same RAM. Tests OK so far with 2-4-4-7, 1.7 VCore.
I did rustle up a 512 stick of Kingston Value PC3200 to play with.

When first built, the system tested stable at 11.5x216 but required 1.75VCore, too high for me.
I am dirt poor now so I will be lowering my OC rather than put up the voltages too much. Might keep 1.7 VCore, OC the vid card.

 

[maluckey]

trexpesto

Are you SURE that you are in Dual Channel? No offense, but I can't even run an application at tRAS 7 in dual channel. I error out at tRAS 9.

Dual channel is more stable and faster overall with tRAS set to 11. Test after test after test has shown this to be almost universal with DDR I and dual channel. Sure, you can run it different, but with errors and limitations that make it not worthwhile overall.

 

[trexpesto]

yabbut dont you have overclocked PC 2700? So since these are pretty much stock, came as a tested DC kit, what is surprising about it?
They are in slot 0 and 2, what program will show.

I reinstalled the NB sink, it's a NVidia nForce2 rated at 400 MHz, so it IS getting overclocked about 32 MHz, might give it a tiny increase in voltage or get some more directed cooling on it. My CPU sink drains hot air directly on it.

 

[trexpesto]

Wow one big difference between running with the 9800Pro w/VGA Silencer and the Geforce MMX440 is my -12V rail goes from around -12 to +3Volts. What is up with that?

...
couple reboots seem to have fixed it. It was at +3.34V on the -12V line there. Trippy.

...
chr*st, on the second page. sorry folks:
What is too hot for the NB? Where is the board temp sensor again?

 

[maluckey]

yabbut dont you have overclocked PC 2700? So since these are pretty much stock, came as a tested DC kit, what is surprising about it?

Overclocked or not, in dual channel, tRAS 11 is most always more stable and faster with tight timings (2-2-2).

My PC2700 is overclocked for sure (but barely above default voltage), but it also identical to the Kingston HyperX PC3200 except for the box that it came in and the lettering on the heatspreader. During the Heydays of WinBond BH-5, you could find the stuff in everything from PC2700 to PC 3500. The difference being how much they could soak you for by rating it higher.

BH-5, BH-6 and even CH-5 was rated as PC 2700, PC 3000, PC3200 and PC3500. It all depended on how much of each was laying around at the time. The newest UTT DDR is nothing more than new chips from the old dies for BH-5 (on new boards). It still runs best at tRas 11 at 2-2-2

 

[trexpesto]

I can try it. AbitEQ reported low VCore, upped it to 1.7 in the BIOS. Still dips to 1.65Vin AbitEQ on occasion..

So, also raised the chipset voltage, as it was actually the most overclocked part, 400MHz to 432. A little bit only

 

[maluckey]

vCore droop is normal on the NF7-S v2.0. The #2 pencil vCore (NOT the VDD) mod helps it out a bit and keeps it from varying much.