stability and voltages

[trodas]

I have some stability problems, that are hard to track, since they are - well, longtime only Like todays, crash after 12 hours. Damn.

 

I starting to think that I running too low chipset/agp voltage, perhaps? So, guys, what settings are you using?

 

I running BH-6 2x512 memory at 11-2-2-2-2-13-15-0-0-3-3-2-0-3-E-E-F-E settings, 3.0Vdimm.

 

Mobile Barton 2500+ at 1.850Vcore and 2200 - 2600Mhz (it make 2600, but the mobo not...)

 

PSU - Enermax 431W modded 5V on both fans, modded rail, possible setup the 3.3V rail to almost anything. I tried going from 3.31V to 3.46V but still crashing. Is recommended going back to 3.3V for lover heat generation, or not?

Heat IS a issue there, since my case if and stays fanless.

 

...and I laved chipset and AGP voltages at AUTO. Good? Bad? Ideas???

 

 

 

Except voltages, another my problem could be overheating of mosfets or SB. My SB getting amazingly hot, 52 degrees measured with my Delta Trak toy http://deltatrak.com/infrared_thermo_8.shtml

Should I glue there Zalman NB 32J for better stability???

 

Another two spots are mosfets bellow CPU. I already make a heatsink for them: http://ax2.wz.cz/show.php?p=badthingshappe...47&c=8&d=1&v=v2 (step back 5 pics to see the making progress ) but did not glued it yet, because I wonder if I should do the CO (Current Overload) mod, that basicaly replace the 2.2k resistors with 4.7k ones and some folks reported less Vcore fluctuations??? Should I care, when I use only 1.850Vcore and never go higher???

 

Last spot is the overheating mosfets bellow the memory banks. I can make a nice, but smaller (every Ultra graphic card, and even mine, are long, so it get into this space, so I can't use too high heatsink...) heatsink on these too. Could be this a issue? :confused:

And how I could try to find the culprit? Run memory at 3.3V and see if crashes come soon?

 

At 200x12.5 my mobo with 1.850Vcore/AGP AUTO/CHIPSET AUTO/3.0Vdimm hold only 20 minutes and then crashed at blue screen of death (BSOF)

 

Advice need.

[FDM]

Have you ran memtest at all? I would do that and rule out if it is the mem causing the crash. Also, increasing your chipset voltage might help.

 

You should stick a sink on any chip that gets hot on the board, and definitly the mosfets if you haven't already. Can't hurt anything

[trodas]

Yes, memtest run and passed well over 225Mhz, so even at 225Mhz Im well in safe border.

 

But what if the voltage fluctuate too much (3.0V for 2x512 DIMMs are lell a lot of power to draw) and there could be a error then??? Heatsinking the mosfets bellow rambanks need

 

But maybe not. And I think, that I push them as far, as 3.3V to see, if crashes will be more common. What do you think???

 

 

Yes, I have only AGP and chipset voltages at AUTO, witch give 1.5V, I believe... Should I push up? What settingy you, guys, use? What about... leave the AGP alone and push to 1.8Vchipset??? :confused:

[FDM]

with 250fsb I run 1.9 to the chipset, and for AGP I run 1.6 normally. 1.8 should be more than plenty for 200fsb.

[trodas]

Okay, I leaved for now (just one change at once rule) the AGP at auto and pushed the poor chipset too 1.8V Stable now, but I run at pretty low settings now

200 x 11 and only 1.775Vcore and 2.7V for memory... Let me see it it last a day...

 

Anyone tried reach better stability using bigger SB cooler??? :confused:

[trodas]

And what about the OC modd??? :confused:

 

I mean, could it help me a little to get more "clean" voltage w/o fluctiations?

 

 

...some ppls ever reported that these "fluctuations" are bios rounding error, since when measured with multimeter, the voltage seems to be stable??? :confused:

 

I running my other testing mobo at 1.9Vcore and it kicking well, but ONLY 2100Mhz - more the oldie chip did not make stale, of course 100% load all day. And it seems holding well, so... :confused:

 

Should I just glue the mosfet heatsinks??? :confused:

[LoMbArD]

what kind of temps are you getting? IMHO you can never have too much cooling, even if it is passive. running higher voltages means higher proc temps, which in turn means higher possibility of instability,

[SAE]

Yeah. Trodas, first try to run the memory with CPC disabled as it's double sided ram and maybe that's what is causing your lockups.

 

If that does not help, you should definately try to sink all the mosfets (6 below the cpu, 4 below the dimm banks, the one next to the agp is vdd and gets very hot). Additionally put a fan there blowing directly on the sinked areas...

 

About the OCP mod:

 

It does NOT reduce voltage fluctuation!!! It's to get the CPU more current (what's not always healthy) as Abit uses 3.3k ones for that purpose...

 

The Vdroop mod reduces fluctuation, but you are right, when measuring the real vcore at the rightmost toroid below the cpu socket (it's insulated, you have to check the soldered connections) you 'd very likely notice there next to NO fluctuation.

 

I did all mods except the Vsense (does not do anything)... and I am quite satisfied. But remember, the CPU can suffer badly when giving more than 2V for a longer period of time - I know it well and my CPUs are subzero-cooled. Do you really wanna give the cpu more than that? Are the additional 50MHz really worth it? I dunno. You have to decide

[trodas]

My machine is now pretty stable, but running so so slow...

 

200x11 only, 1.775Vcore, 1.7Vchipset, 1.6Vagp and 2.8Vdimm

 

LoMbArD - I have watercooling, external, kinda neat, and my CPU is proven to be stable over 75C (ADM specs says that it can go to 85C anyway, P4 thermal diode shut down at 135C, so )

 

SAE - well, my purpose is get my machine stable at about 225 x 11 - 2475Mhz. My CPU can make 2600Mhz with 1.850Vcore. So all I want is stable voltage in the 1.850Vcore limit. I never go for more. As you say, not good for either CPU, but mainly for the mobo. In fanless case, where the only one air intake is the 92mm PSU fan at 5V - well...

So, you did not recommend my ANY mod at all?

 

I think I beginning to agree with you. My second Lanparty B mobo running now second day at 100% CPU load (folding) on my testing bench wih 1.9Vcore. (there is oldie unloacked Barton, that need 1.9Vcore to get 2100Mhz. At 2.0Vcore it can do 2200, but this is IMHO overkill for 100MHz - forget it. I fried my Epox 8RDA+ already, trying to run this way. CPU survive, but heatsink on the mosfets change it's color, the green PCB turned to be a bit brown bellow the mosfets and toroids (!) and finally, the leaking capacitors exploded. And from this time the mobo never worked again.

 

Yesterday, after day of run, I measured the temperatures of the chips on the Lanparty B on my bench desk. There are fairly low, for the 1.9Vcore, however it should come to play that the Zalman CNPS7000Cu blowing the air directly to these... at least to these bellow the CPU.

If anyone are interesed, I paint the values on picture

[ViperJohn]

Originally posted by trodas

I have some stability problems, that are hard to track, since they are - well, longtime only Like todays, crash after 12 hours. Damn.

 

I starting to think that I running too low chipset/agp voltage, perhaps? So, guys, what settings are you using?

 

I running BH-6 2x512 memory at 11-2-2-2-2-13-15-0-0-3-3-2-0-3-E-E-F-E settings, 3.0Vdimm.

 

Mobile Barton 2500+ at 1.850Vcore and 2200 - 2600Mhz (it make 2600, but the mobo not...)

 

PSU - Enermax 431W modded 5V on both fans, modded rail, possible setup the 3.3V rail to almost anything. I tried going from 3.31V to 3.46V but still crashing. Is recommended going back to 3.3V for lover heat generation, or not?

Heat IS a issue there, since my case if and stays fanless.

 

...and I laved chipset and AGP voltages at AUTO. Good? Bad? Ideas???

 

 

 

You have BH6...soften the memory timings up slightly to 11,3,3,2,2 for starters.

 

Viper

[SAE]

Originally posted by ViperJohn

You have BH6...soften the memory timings up slightly to 11,3,3,2,2 for starters.

 

Viper

 

Right. IMO CPC is the biggest problem though... (he 's on 11-27 -> no CPC available)

 

@trodas

 

I don't think a Vmod would stabilize the 1.85V. The Vmod by DFI is not even active then. There's no real fluctuation. I think CPC is the worst problem, what you can only come by flashing to 6-19 (hopefully becomes official soon) as it's really stable, and IMO there's no risk of flashing. Winflash should do it fine. I always use it. And there are many peeps that flashed this way

 

Or did you already flash 6-19?

[trodas]

ViperJohn - mime mems are fine at 11-2-2-2, dude They need more that I expected, 3.1Vdimm, to pass fine at 225Mhz, but this is okay for BH-6 chips, from what I hear

 

SAE - okay, I only glued the heatsinks then It helps. From start I can work at 225x11 for about hour, then it hang Probably overheating. The mosfet cooler are burning hot, man :eek:

And the Vdimm/SB coolers aren't cooler much, eh

 

Okay, now some interesting stuff. Picture from mosfet coolers on my second, AirCooled by Zalman CNPS7000Cu, after a day of folding with Vcore 1.9V:

http://ax2.wz.cz/show.php?p=wc&id=84&c=8&d=1&v=v2

 

Same machine, Vdimm mosfets, one 512MB Mushkin BH-6 11-2-2-2 and 3.0V IIRC:

http://ax2.wz.cz/show.php?p=wc&id=91&c=8&d=1&v=v2

 

Around these pictures you can find others like this:

 

Enjoy

 

 

And now, I have a question. I run now stable at 1.750Vcore, 1.6Vagp, 1.8Vchipset (can be 1.7V too, I switch soon, I jsut tried overvold the NB to get rid of the nasty audio clicks I have) and 3.1Vdimm - stable means prime priming, etc. Not a problem, except... Except the DAMN audio clicks!!! :mad:

I think my SB is well cooled, with the Zalman 32J NB cooler, is not it? :confused: