[First Look] DFI 680i LT NF-680i LT SLI (56K Warning~~~)

[Jobeo]

well i got somewhere... installed 3dmark03 and loaded my sketchy settings... the reboot crash occured during the benchmark, which actually made me happy cause now i had a better way of testing it.

i tweaked the NB volts higher that didn't help. even tried running the test non-SLI and it seemed to reboot even faster. all of the reboots were occuring at different times during the test there seemed to be no pattern.

so i started fiddling with the GTL's ... i had them set at the "sweet spot" of 1.44v and 170 95 85... tried them a bit higher and the test seemed to run a bit longer before rebooting. finally ended up with GTL's of 180 100 90 and 3dmark03 passed for the first time ! ! i ran the test 3 more times and it passed each time. then ran the 3dmark05 and 3dmark06 tests and both passed. loaded an old supcom save with a lot of units and that seemed to go OK too...

SO.... to make a long story short... i mean, i knew the GTL's were important cause i've mentioned them in almost every post these past couple days.. but MAN they are _really_ important. unless i'm mistaken we're talking about tweaking millivolt settings yet it seems to have a huge impact on stability.

so i think i've found a new sweet spot at 1.44 and 180 100 90. which doesn't really add up cause the Nvidia GTL reference table says the VTT should be 1.55 at those GTL's but whatever works i'll take it... i'll probably need to test further with another prime blending session but for now things are looking good.

[C-N]

SO.... to make a long story short... i mean, i knew the GTL's were important cause i've mentioned them in almost every post these past couple days.. but MAN they are _really_ important. unless i'm mistaken we're talking about tweaking millivolt settings yet it seems to have a huge impact on stability.

Even with a dual core the GTL's make all the difference not for the CPU's sake but for the NB.

I am still working on getting 3.7GHz running 8 hours + Orthos stable & have increased my reboot/crash time from 3 to 6 hours tweaking GTL's.

I think GTL's are going to take forever to workout & understand as even the smallest variant in CPU/NB puts you back to the start again with the GTL's.

[Praz]

so i think i've found a new sweet spot at 1.44 and 180 100 90. which doesn't really add up cause the Nvidia GTL reference table says the VTT should be 1.55 at those GTL's but whatever works i'll take it... i'll probably need to test further with another prime blending session but for now things are looking good.

It really does make sense as the reference table is just that, a reference. Raju more or less stated this several pages back in this thread.

Rhyme or reason is in the Gtl article by Kris as you thought. With regards to formula's Gtl is in the 2/3 VTT ballpark as Kris has told, the rest is down to onboard resistor/component variance and tolerance levels, including thermal drift, system specific rfi/emi etc, not forgetting individual cpu variance...

[virtualrain]

well i got somewhere... installed 3dmark03 and loaded my sketchy settings... the reboot crash occured during the benchmark, which actually made me happy cause now i had a better way of testing it.

i tweaked the NB volts higher that didn't help. even tried running the test non-SLI and it seemed to reboot even faster. all of the reboots were occuring at different times during the test there seemed to be no pattern.

so i started fiddling with the GTL's ... i had them set at the "sweet spot" of 1.44v and 170 95 85... tried them a bit higher and the test seemed to run a bit longer before rebooting. finally ended up with GTL's of 180 100 90 and 3dmark03 passed for the first time ! ! i ran the test 3 more times and it passed each time. then ran the 3dmark05 and 3dmark06 tests and both passed. loaded an old supcom save with a lot of units and that seemed to go OK too...

SO.... to make a long story short... i mean, i knew the GTL's were important cause i've mentioned them in almost every post these past couple days.. but MAN they are _really_ important. unless i'm mistaken we're talking about tweaking millivolt settings yet it seems to have a huge impact on stability.

so i think i've found a new sweet spot at 1.44 and 180 100 90. which doesn't really add up cause the Nvidia GTL reference table says the VTT should be 1.55 at those GTL's but whatever works i'll take it... i'll probably need to test further with another prime blending session but for now things are looking good.

That's great. Frankly I'm totally amazed you got it prime stable without tweaking GTL.

[Jobeo]

That's great. Frankly I'm totally amazed you got it prime stable without tweaking GTL.

tx... from my experience, and this is all conjecture, the GTL's seem tied to the FSB only. when you fiddled with the FSB the gtl's have to be tweaked to support it. i think thats why my tests with 11-12-13 went OK cause i wasn't changing the FSB at all.

on the 400 FSB i've been talking about i was using a low 7x multi... i just bumped it to 8x and it passed all the 3dmark03 tests without any changes to the VTT/GTL's... they are still at 180 100 90

i think thats probably why those of us with quads have so much trouble above 400... the GTL's probably need to be even tighter, say within 1 pt of where they need to be, or the OC won't prove stable.

[simonmaltby]

I got this board last week and so far its reasonably good in most ways and very good in others.

I have found it more stable than the BFG 680i i had before and can get rock solid stability at 450FSB.

Got my chipset water blocks today... Link mofset, Nb and SB so will let you know when all setup and running.

[Grinch]

I got this board last week and so far its reasonably good in most ways and very good in others.

I have found it more stable than the BFG 680i i had before and can get rock solid stability at 450FSB.

Got my chipset water blocks today... Link mofset, Nb and SB so will let you know when all setup and running.

looking forward to it

[simonmaltby]

For info.

I suffered some really wierd issues over the weekend with the board.

First i flashed to 521 from the official site and all was well with my X6800 (indeed it did 4.5 without issue on freon). The i changed to my E6600 and could not find multi unlocking...

So i flashed to Beta 504 and sure enough the multi unlocking came up, but other parts of the bios were poor. so i flashed back to 521. On doing this the multi unlocked remained... good news i thought... No this is when the really wierd stuff happened.

At this point i could not get stable even at stock. In windows as i moved the mouse, or accessed HD the num lock keyboard loght would flash and ther would be a stuttering. This only happened when using Raid. After a few moments the system would just freeze.

I trued battery out etc - nothing worked. In the end i flashed back to 504. then flashed back to 521 - removing the battery between each flash. I also flashed in Windows XP using Winflash and did the old DFI trick of flash once and then without re booing flash again.

Anyway that got me back to stable, but no multi unlocked.... so for me 521 does not enable downward multi's which is a pain.

[Jobeo]

For info.

I suffered some really wierd issues over the weekend with the board.

hrmm... ive never heard of anything like those issues... when i flash the BIOS i just winflash it once and have never removed the battery afterwards.

as far as switching processors... ya i could see some trouble with that. are you sure you're removing the battery for long enough? the last time i did the BIOS reported a new CPU detected and installed the proper siprom table for it...

[virtualrain]

For info.

I suffered some really wierd issues over the weekend with the board.

First i flashed to 521 from the official site and all was well with my X6800 (indeed it did 4.5 without issue on freon). The i changed to my E6600 and could not find multi unlocking...

So i flashed to Beta 504 and sure enough the multi unlocking came up, but other parts of the bios were poor. so i flashed back to 521. On doing this the multi unlocked remained... good news i thought... No this is when the really wierd stuff happened.

At this point i could not get stable even at stock. In windows as i moved the mouse, or accessed HD the num lock keyboard loght would flash and ther would be a stuttering. This only happened when using Raid. After a few moments the system would just freeze.

I trued battery out etc - nothing worked. In the end i flashed back to 504. then flashed back to 521 - removing the battery between each flash. I also flashed in Windows XP using Winflash and did the old DFI trick of flash once and then without re booing flash again.

Anyway that got me back to stable, but no multi unlocked.... so for me 521 does not enable downward multi's which is a pain.

Are you using the 4CTST or the 2CTST version of 521. The 2CTST should have unlocked downward multis... at least it does for me.

[BAR*B*Q]

I got this board last week and so far its reasonably good in most ways and very good in others.

I have found it more stable than the BFG 680i i had before and can get rock solid stability at 450FSB.

Got my chipset water blocks today... Link mofset, Nb and SB so will let you know when all setup and running.

would you share your voltage settings please ? im trying to get 400 x 9 primestable. theres allways a fft length that fails.

[Jobeo]

would you share your voltage settings please ? im trying to get 400 x 9 primestable. theres allways a fft length that fails.

or you could share yours and we could see if anything is glaringly wrong .. . .

[BAR*B*Q]

or you could share yours and we could see if anything is glaringly wrong .. . .

tried so many mate. none of them really prime stable.
all i can say is that im testing with an e6600 + 2x1gb d9gmh memory on 423 bios: 400 x 9 in sync mode 1T.