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

[Jobeo]

ya there are times i wish i had just gone with the EVGA ... although the early revisions on that one were crap. i think i'd have a much easier time on the A1 than this DFI board...

when the x38 boards are released i'm gonna seriously consider EVGA...

[virtualrain]

Originally Posted by MikeMK

Yup, well i think it will take me longer to get to grips with the DFI and the EVGA, still it will be interesting to see how it all pans out and if the pain is worth the gain!

@MikeMK... Since you have the luxury of an extreme dual core (intead of a cursed fixed-multi quad) you may find that both boards serve you equally well on the OC front and you end up choosing one with the nicest layout/cooling for your needs. 1T RAM performance on the DFI may be the clincher depending on your RAM.

@Jobeo... I think DFI may have a compelling X38 offering (if it's available before it's obsolete!). Have you heard EVGA might do an X38 product? AFAIK, I thought they were only doing nVidia reference boards.

[GNU]

Originally Posted by virtualrain

60-deg under load is fine.

If your system is not stable, you probably need more voltage... you still have a bit of room to move there... try increasing your CPU and NB voltage a notch or two and test it... stress test it at night when it's cooler.

My system is stable but I can't run any CPU Stress Tools (hot cpu warning).
OCCT burned my CPU until 70°C between 8min then break.

And more Voltages gives CPU and NB more heat, that's bad.

I will retest occt.

[virtualrain]

Originally Posted by GNU

My system is stable but I can't run any CPU Stress Tools (hot cpu warning).
OCCT burned my CPU until 70°C between 8min then break.

And more Voltages gives CPU and NB more heat, that's bad.

I will retest occt.

Well then you need to backoff or improve cooling by the sounds of it.

[NeoGeo333]

Just got the board a couple days ago along with a E6420 and 4x1 gig Ballistix PC-8000. Man, this board is a lot harder than the Ultra-D NF4 from same company. Cant even past 3.2Ghz even with voltage at 1.5v. I guest my cpu is holding me back but july 22 is so close and a quad for 266.00 seems too unresistable.

[Jobeo]

Originally Posted by NeoGeo333

Just got the board a couple days ago along with a E6420 and 4x1 gig Ballistix PC-8000. Man, this board is a lot harder than the Ultra-D NF4 from same company. Cant even past 3.2Ghz even with voltage at 1.5v. I guest my cpu is holding me back but july 22 is so close and a quad for 266.00 seems too unresistable.

lol ... you thought it was hard with a dual core wait till you fiddle with a quad on this board..... *sigh*

[NeoGeo333]

hopefully by then DFI will have acceptable bios to fix all the bugs and easier o/c.

[virtualrain]

Originally Posted by Jobeo

lol ... you thought it was hard with a dual core wait till you fiddle with a quad on this board..... *sigh*

QFT!

BTW, if you guys are interested, I did some power measurements with an AC watt meter on my system and posted them in the PSU forum...

http://www.xtremesystems.org/forums/...d.php?t=147546

[GNU]

Originally Posted by virtualrain

Well then you need to backoff or improve cooling by the sounds of it.

Improve cooling not possible (only Aircooler not more or another room with lower temps 21°C or the ultimate solution -> Vapo

But I can't do more for a watercooler.

[MikeMK]

Originally Posted by virtualrain

@MikeMK... Since you have the luxury of an extreme dual core (intead of a cursed fixed-multi quad) you may find that both boards serve you equally well on the OC front and you end up choosing one with the nicest layout/cooling for your needs. 1T RAM performance on the DFI may be the clincher depending on your RAM.

Ive actually grabbed myself a Q6600 to play with - fancied having a play with a quad and seeing what they can do - hence the decision between the EVGA and the DFI as pretty much all other SLI boards dont seem to have a great amount of luck as far as Quads are concerned.

I took note of the 1T results earlier in the thread... very impressive. I have 2gb of Cellshock PC28000 D9GKX, and another 2gb of Crucial Ballistix PC8500 D9GMH sat around here, so im hopefull at least one of those kits will get me decnt 1T results.

[Jobeo]

Originally Posted by GNU

I will retest occt.

OCCT is irrelevant ... in my opinion it proves nothing if you're system is OCCT stable. i posted about my experiences with it a page or so ago and i'll never use it again.

tx VR for the link i've used prime95 before but that version you linked is better for quads. i've been running a lot of tests recently and getting much better results. BTW is there any way to have it monitor core temps and cutoff at a certain termp like OCCT does? that's the one feature i miss...

as far as some mythical DFI BIOS goes that has an "easy quad overclock" button i don't think it will ever happen. the bios is the way it is and i don't see any major changes coming. they might tweak a few things here or there but we'll always be stuck balancing VTT's and GTL's... i hope i'm wrong =/

[Borgod]

I actually like OCCT alot.

My QX6700 is OCCT stable 6 hours @ 3600Mhz. I was running it at this frequency all weekend playing games, video encoding, listening to mp3's and downloading torrents all simultaneously with no problems whatsoever.
As this is likely to be the heaviest work load I ever put on my CPU and knowing I'll never reach 70+ degree temps and 100% load I'm happy to rely on OCCT as an indication of system stability. 6 hours I determined to be the longest stint of non-stop gaming I'll ever play, and even then it would never be with all 4 cores.

I'll still use Prime95, but not testing for as long as 8 hours.

This is my experience and personal opinion for my own personal use

[Jobeo]

ya i'm not sure there is any perfect test... other than running the programs you use and seeing how well it goes.

on my most recent 400 FSB clock i crunched small FFT's a while then did 90 minutes of a blend test... everything seemed stable and working well.

until i was just playing supcom for about an hour and had a ton of units on the screen and it reboot crashed. =/ =/ i tweaked a couple volts and when i go to bed tonight i might just leave it blend testing...

[C-N]

Originally Posted by Jobeo

ya i'm not sure there is any perfect test... other than running the programs you use and seeing how well it goes.

on my most recent 400 FSB clock i crunched small FFT's a while then did 90 minutes of a blend test... everything seemed stable and working well.

until i was just playing supcom for about an hour and had a ton of units on the screen and it reboot crashed. =/ =/ i tweaked a couple volts and when i go to bed tonight i might just leave it blend testing...

I think the only way to test this boards longterm stability is with an overnight blend test. I can run small fft's or large fft's all night with my E6400 @ 3.7GHz & my E6600 @ 3.7GHz but when you put the 2 together & run a blend it will reboot.

The longest I have been able to hold 3.7GHz down for with this board is 3 hours on a CPU that has done 3.7GHz on a DFI 975X/G for several months & its even done 3.8GHz once on the 975 & once on my ICFX3200 for an OCDB entry so I know the CPU can do it.

Regards

Craig.

[Borgod]

Yeah the limitation isn't with the CPU's... my QX6700 has done 5000Mhz and it's starting to get difficult for me at 3600Mhz on the DFI on water.

[C-N]

It sure is strange its just a good job you don't need the raw CPU MHz to get good performance out of the board.

I can boot 490fsb with the E6400's x 8 multiplier but I cant even clockgen 440fsb with my E6600's x 9 multiplier even though it runs 4.0GHz just fine for benching on the 975 & ICFX.

I think the limitation is something to do with the native multiplier setting something up at boot but as of yet I haven't found anyone with an unlocked CPU that can prove/disprove my theory.

I have seen 450fsb x 9 = 4GHz + on an unlocked x6800.

If anyone with an unlocked CPU can comment on what they can boot & or clockgen on the 8 & 9 multiplier I would appreciate it.

Regards

Craig.

[GNU]

Originally Posted by Jobeo

OCCT is irrelevant ... in my opinion it proves nothing if you're system is OCCT stable. i posted about my experiences with it a page or so ago and i'll never use it again.

tx VR for the link i've used prime95 before but that version you linked is better for quads. i've been running a lot of tests recently and getting much better results. BTW is there any way to have it monitor core temps and cutoff at a certain termp like OCCT does? that's the one feature i miss...

as far as some mythical DFI BIOS goes that has an "easy quad overclock" button i don't think it will ever happen. the bios is the way it is and i don't see any major changes coming. they might tweak a few things here or there but we'll always be stuck balancing VTT's and GTL's... i hope i'm wrong =/

It's not important if irrelevant or relevant, important is, your system runs stable with high temps and I need a solution to get lower temps. OCCT vs Prime95: My opinion is, OCCT is better for short tests and Prime better for long tests.

[Jobeo]

Originally Posted by C-N

I can run small fft's or large fft's all night with my E6400 @ 3.7GHz & my E6600 @ 3.7GHz but when you put the 2 together & run a blend it will reboot.

ya i experienced that the other day when i was prime'ing. i could run small FFT's for over an hour but the blend failed within 1 minute. i agree that blend seems to be the most rigorous test to prove stability.

i blended last night for 3 hours and everything went fine. i guess i can assume these are stable settings but only time will tell.

[speed bump]

Finally a board without a million heatpipes all over the place.

[Grinch]

I just ordered the swiftech mcw30 to watercool my SB...anybody using this to cool their SB?

[Borgod]

Are you watercooling your NB?

I'm seriously considering ordering the MIPS kit, although Im undecided if I'll fit the SB block since it only gets to 40degrees with the standard HSF and will only further restrict flow.

Btw, have you checked your clearances with your video card with a MCW30 on the SB?

[Jobeo]

i've found a new standard in stability testing ... its name is SUPREME COMMANDER and let me tell you this game will expose any faults in your clocking. all at once it tests cores, memory, NB, graphics, and audio all wrapped into one tight little package.

i had a clocking recently that proved OCCT stable 1hr test... it crashed within 1 minute of supcom play.

not only that it makes testing fun... although crashes can be annoying so remember to save often. prime blend is still the way to go for initial testing but the final proof is from supcom LOL

it won't show max temps for your OC though... it stresses one core very vigorously and 2 more cores provide audio and AI logic for your opponents... great way to test !!

[Grinch]

Originally Posted by Borgod

Are you watercooling your NB?

I'm seriously considering ordering the MIPS kit, although Im undecided if I'll fit the SB block since it only gets to 40degrees with the standard HSF and will only further restrict flow.

Btw, have you checked your clearances with your video card with a MCW30 on the SB?

here is my current setup:



I have more than enough clearance and my SB runs at 48-52C all the time

[GNU]

Originally Posted by Jobeo

i've found a new standard in stability testing ... its name is SUPREME COMMANDER and let me tell you this game will expose any faults in your clocking. all at once it tests cores, memory, NB, graphics, and audio all wrapped into one tight little package.

i had a clocking recently that proved OCCT stable 1hr test... it crashed within 1 minute of supcom play.

not only that it makes testing fun... although crashes can be annoying so remember to save often. prime blend is still the way to go for initial testing but the final proof is from supcom LOL

it won't show max temps for your OC though... it stresses one core very vigorously and 2 more cores provide audio and AI logic for your opponents... great way to test !!

It burns the CPU while the test correct? That's bad for me, 'cause I can't test it.

[Praz]

Originally Posted by Jobeo

i had a clocking recently that proved OCCT stable 1hr test... it crashed within 1 minute of supcom play.

Most people with stable rigs using OCCT for stability testing run it for 4 hours. One hour is just an indication of possible stability. Prime95/Orthos is the ultimate test of stability though because with Core2 processors those two programs are buggy as hell.