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

[adamsleath]

looks like dfi really needs new bios....to state the bleeding obviarse.
the evga with p28 looks like a no brainer (for quad oc) according to nugzo's experience.

[Jobeo]

hey VR thats some nice info you're accumulating over on your blog.

at the moment i've become very gunshy about overclocking this board. i was tweaking bits and pieces here and there and not getting close to the voltages you posted a bit ago in your screenshot.

something went wrong and the BIOS on boot was telling me the system was in safe mode and to check the CMOS. i loaded settings i had saved which i knew worked and it kept saying it was booting in safe mode. so i went and loaded optimized defaults, everything at stock speeds and it still reported it was in safe mode on boot.

so i figured something was wrong with the BIOS and i should flash it again to make sure. the flash went fine, it completed and told me to reboot. after rebooting i got a blank screen. nothing. i powered down and when i turned it back on a horrendous beep was coming from the rig. the post code showed "02" which is a reserved bios code. i tried clearing the CMOS by holding down reset/power for a few seconds, that didn't help. tried the cmos clear jumper and that didn't help either.

to make a long story short (if possible) i had to remove the CMOS battery for a few minutes to get the thing up and running again. i'm not sure why all this happened but i'm glad its working now even if its just at stock. i'll need to build up some courage before trying again =/

[virtualrain]

I guess I take a different approach to overclocking than most... I basically crank everything up to safe maximums and see what the processor will do and then scale things back one at a time... that's why my initial OC settings seem like really high Voltages... they are!

It turns out the max my X3220 will do under coarse adjustments right now is about 3.5 GHz (390x9) - I was hoping for 3.6 and perhaps that will come with some fine tuning.

I'm now in the process of scaling back my voltage settings to see what the minimums are to keep it stable at 3.5GHz and then I'll see if I can tweak higher.

I've found the BIOS to be pretty solid... if a setting is off and it won't post, the easy reset works for me. Most of my problems have been in getting Prime stable which is time consuming. I've found that using small FFT's is the most taxing on the CPU and if it's going to fail Prime, it will do so on this test in about 5-10min. The blend test takes longer because it doesn't hit the small FFT's until about 10 min. in.

Sometimes when the board fails, it fails hard!!! As Nugzo reported, it will simply shut itself off. I found this happened most often if I was pushing voltages to the extreme (like CPU VTT at 1.575 or so).

I hope to have some quad settings to share tomorrow, but I'm still confused about the GTL settings. They don't make any sense why there is a "CPU GTL 1" and a "CPU GTL 0/2/3"... the articles I've read seem to imply that 0/1 and 2/3 are ususally tied together and should have close to the same value. With the board, the recommended value for GTL 1 is twice the value of the 0/2/3 value.

[Tony]

The GTL options are labelled wrong in bios.

Option 1 is core 0 and 1
Option 2 is core 2 and 3
Option 3 is the NB

There is as far as I know no way to have the GTL voltages working how thwy are labelled in bios.

If i get time I will alter the wording with modbin so it coorect reflects what is going on

[virtualrain]

Is it true that GTL 0/1 should be twice the value of GTL 2/3? This doesn't make sense to me either unless the voltage bias is skewed to begin with.

We need some frame of reference like this in order to intelligently tweak... http://csd.dficlub.org/tikiwiki192/t...page=GtlRef&bl

[Raja@ASUS]

Is it true that GTL 0/1 should be twice the value of GTL 2/3? This doesn't make sense to me either unless the voltage bias is skewed to begin with.

We need some frame of reference like this in order to intelligently tweak... http://csd.dficlub.org/tikiwiki192/t...page=GtlRef&bl

your response is at the club, for further insight ask @ TTR, kris is the man for gtl..


regards
Raja

[palese]

One thing I will give Evga with their revision A board, it sure is easy to overclock I ran my Xeon 3220 27/7 @ 3.2Ghz with out a hiccup, I wish I could say the same for my new DFI Hell I just want a simple OC but nothing is simple with this board.

[virtualrain]

Yeah, I'm not a big fan of GTL... which is what makes this board so complicated.

I don't know how EVGA and Intel boards get great quad core overclocks without it, but there has to be a better way than relying on users to tweak GTL. It will be interesting to see if people with EVGA boards can match or exceed their quad clock on the DFI or if they just give up in vein and go back to their EVGA. Giving people the tools to achieve great overclocks (such as tweakable GTL) is only good if the tools are well understood and/or some guidance exists... otherwise, they are more of a hinderance. I must admit, I'm finally making some headway, but it's been a frustrating weekend with this board to be sure and none of the current guidance was much help.

I think if you can overclock a DFI board, you should get some kind of Scout's honor or badge or something because this ain't no basic rope work here boys and girls!

[Apollonas]

Giving people the tools to achieve great overclocks (such as tweakable GTL) is only good if the tools are well understood and/or some guidance exists...

OR when it can get you higher than the others by tweaking more. Otherwise when you know that you going to have the same result or less you going with the easiest solution.

[palese]

Yeah, I'm not a big fan of GTL... which is what makes this board so complicated.

I don't know how EVGA and Intel boards get great quad core overclocks without it, but there has to be a better way than relying on users to tweak GTL. It will be interesting to see if people with EVGA boards can match or exceed their quad clock on the DFI or if they just give up in vein and go back to their EVGA. Giving people the tools to achieve great overclocks (such as tweakable GTL) is only good if the tools are well understood and/or some guidance exists... otherwise, they are more of a hinderance. I must admit, I'm finally making some headway, but it's been a frustrating weekend with this board to be sure and none of the current guidance was much help.

I think if you can overclock a DFI board, you should get some kind of Scout's honor or badge or something because this ain't no basic rope work here boys and girls!

virtualrain do I need to change the default GTL settings just to get a mild overclock And if so does each core on the quad have it's own setting?

Right now when I bump up the FSB and clock speed to say just 3.0 the system goes into safe mode and says hit F1 if I want to continue, but when I do and it goes into windows nothing is changed

[OldGuy]

F1 will continue on in "safe mode" hence no change. It is there to allow you to recover and go into the bios and make changes.

virtualrain do I need to change the default GTL settings just to get a mild overclock And if so does each core on the quad have it's own setting?

Right now when I bump up the FSB and clock speed to say just 3.0 the system goes into safe mode and says hit F1 if I want to continue, but when I do and it goes into windows nothing is changed

[virtualrain]

I'm still trying to find something that is Prime95 stable for at least a few hours but here's what was working for me...

Voltages:
-CPU VTT: 1.53x (something high around here - but not too high)
-CPU special add: Auto
-CPU Drive: Strong
-RAM Drive: Weak
-RAM Voltage: (whatever your RAM likes)
-NB Core: 1.5x (something high)
-SB Core: 1.62
-SB Dual: 1.52
-LDT Voltage: 1.22
-CPU VTT: 1.52
- GTL1: 180
- GTL0/2/3: 180
- NB GTL: 90

I know these GTL settings are totally wacked from what has been recommended by DFI and Tony, but they are the only settings that seem to provide any semblance of stability under PRIME. I choose equal values for GTL1 and GTL0/2/3 based purely out of frustration tweaking values closer to default. They seem to work better than anything else I've tried. Either I have a wacked CPU, got extremely lucky, or the implementation and guidance on GTL for this board is off... I have no idea what.

Note that at least we've had it clarified that GTL1 is actually the GTL setting for Core 0/1 and GTL 0/2/3 is the setting for Core 2/3. They are mislabled in BIOS.

Go with high voltages for your CPU and NB and choose a high FSB that you think your CPU can do. You can always scale back the voltages after you find stable GTL settings. You don't want to be fighting both low voltages and bad GTL settings at the same time in my opinion. You will find there is a large gap between what you set in BIOS and what you get (as shown in BIOS). Furthermore, there is significant droop under load.

I started at 400FSB but couldn't get it to be stable at any GTL settings (and I tried a lot!). I then lowered the FSB to 390 and got a lot more stable, particularly with the GTL settings mentioned above.

If you can't even boot windows with these GTL settings, then you should tweak them significantly to try and find some bootable values. I found that I could boot windows with a wide range of GTL settings... I just couldn't find any that were PRIME stable. For example, I could boot at Tony's recommended sweet spot and also the radical settings I have above.

Best I've done is 2hrs PRIME Stable (hadn't had a chance to test longer yet) at 390 FSB with above settings.

Use Prime95.2 on SMALL FFT to save time... I found it would error out by test #5 (about 5min.) if I had unstable settings. Every system is likely different in this regard, but you will get a feel for what will make your computer fail quickly soon enough! For long term stability testing, use the Blended variation as it will stress the NB more.

Don't overclock your RAM until you get a stable FSB overclock on your CPU. Give your RAM enough volts to keep it happy but not stress it and if you are using Sync/Linked on your FSB the RAM clock should be under your RAM's ability to perform. There's no reason to have RAM playing a role in stability while you are trying to find a stable FSB.

I really hope someone starts finding the key to GTL on quad core here soon. Post your results please!

[palese]

F1 will continue on in "safe mode" hence no change. It is there to allow you to recover and go into the bios and make changes.

So that about says it all then, if you do not fiddle with the GTL settings then your going no where

[scarface]

enjoying your beta testing fellas? lol

[The Nemesis]

INstOCK @Newegg $299.99 + shipping.

[palese]

I'm still trying to find something that is Prime95 stable for at least a few hours but here's what was working for me...

Voltages:
-CPU VTT: 1.53x (something high around here - but not too high)
-CPU special add: Auto
-CPU Drive: Strong
-RAM Drive: Weak
-RAM Voltage: (whatever your RAM likes)
-NB Core: 1.5x (something high)
-SB Core: 1.62
-SB Dual: 1.52
-LDT Voltage: 1.22
-CPU VTT: 1.52
- GTL1: 180
- GTL0/2/3: 180
- NB GTL: 90

I know these GTL settings are totally wacked from what has been recommended by DFI and Tony, but they are the only settings that seem to provide any semblance of stability under PRIME. I choose equal values for GTL1 and GTL0/2/3 based purely out of frustration tweaking values closer to default. They seem to work better than anything else I've tried. Either I have a wacked CPU, got extremely lucky, or the implementation and guidance on GTL for this board is off... I have no idea what.

Note that at least we've had it clarified that GTL1 is actually the GTL setting for Core 0/1 and GTL 0/2/3 is the setting for Core 2/3. They are mislabled in BIOS.

Go with high voltages for your CPU and NB and choose a high FSB that you think your CPU can do. You can always scale back the voltages after you find stable GTL settings. You don't want to be fighting both low voltages and bad GTL settings at the same time in my opinion. You will find there is a large gap between what you set in BIOS and what you get (as shown in BIOS). Furthermore, there is significant droop under load.

I started at 400FSB but couldn't get it to be stable at any GTL settings (and I tried a lot!). I then lowered the FSB to 390 and got a lot more stable, particularly with the GTL settings mentioned above.

If you can't even boot windows with these GTL settings, then you should tweak them significantly to try and find some bootable values. I found that I could boot windows with a wide range of GTL settings... I just couldn't find any that were PRIME stable. For example, I could boot at Tony's recommended sweet spot and also the radical settings I have above.

Best I've done is 2hrs PRIME Stable (hadn't had a chance to test longer yet) at 390 FSB with above settings.

Use Prime95.2 on SMALL FFT to save time... I found it would error out by test #5 (about 5min.) if I had unstable settings. Every system is likely different in this regard, but you will get a feel for what will make your computer fail quickly soon enough! For long term stability testing, use the Blended variation as it will stress the NB more.

Don't overclock your RAM until you get a stable FSB overclock on your CPU. Give your RAM enough volts to keep it happy but not stress it and if you are using Sync/Linked on your FSB the RAM clock should be under your RAM's ability to perform. There's no reason to have RAM playing a role in stability while you are trying to find a stable FSB.

I really hope someone starts finding the key to GTL on quad core here soon. Post your results please!

CPU VTT: 1.53x seems high is this to cover droop?

[virtualrain]

CPU VTT: 1.53x seems high is this to cover droop?

Yeah, on small FFT PRIME the Vcore will droop to around 1.44V with this kind of BIOS setting... at least on my board/CPU. Use Speedfan to monitor your Vcore. Let me know if you see the same thing.

Even after setting Vcore to 1.53x in BIOS when you reboot and come back to BIOS and look at the actual reported voltage (bottom of the Voltage Control screen) I see the actual Vcore is only 1.5!

I haven't got out the DMM yet... perhaps later today.

[The Nemesis]

Most Nvidia 680i boards have large vrdoop. It kinda makes oc more of a challenge because you have to set the vcore so high to get enough voltage during load. Really suxs with "B" chips which tend to get hot with more vcore. I have to set a little over 1.5v on my Abit because it drops down to 1.42v @ load.

[Roger_D25]

Interesting Nemesis, I was very suprised with the small amount of V-Droop my board has (even compared to my 3x prior boards)! Although I havn't seen it posted anywhere I was able to find a CPU voltage read point (after along time of testing different spots) and with BIOS set to 1.45v (actual voltage only 1.39v) it will only droop to 1.374v underload which is better than I have seen on most of my other boards (0.016v difference when underload).

I imagine guys that if most people find their boards drooping underload there will be a mod very soon for it (especially since this is an enthusiest board and will probably be very popular).

Virtualrain,
Thanks very much for taking the time to share your results with us, that info could save someone hours of tweaking (at least to get some baseline settings)! If you have the time please keep us posted with your progress, you'll be running at 450Mhz in no time!

[Jobeo]

So that about says it all then, if you do not fiddle with the GTL settings then your going no where

well at a minimum you have to fiddle with the VTT... for stability you might have to tweak the GTL i'm not sure. but that is all for lower-end overclocks. for a decent overclock it seems both the VTT and GTL better be spot on.

i haven't read this thread in a bit... the posts recently have been very informative though. if i could pull off a 390 clock i'd be happy at this point.

i'm gonna start overclocking it again sometime this week. since my last post about the CMOS crash i'm still hesitant... having a stock system that runs seems better than having a dead board =/

[djt]

INstOCK @Newegg $299.99 + shipping.

I just ordered one from Newegg. I hope I’m not sorry I did.

I plan on using it with a X6800.

[djt]

I forgot to ask, what’s latest on the X-Fi issues, if there are any?

[Spazilton]

Quick question. I have one coming in on Thursday.

Could someone measure the spacing on the NB holddown loops?

From the pictures they look identical to the spacing for the 975x chipset. I am wondering if I will be able to use my Chipset WB from my BX2.

[95blackz26]

INstOCK @Newegg $299.99 + shipping.

didnt even see your post. just checked newegg and they had it so i ordered it.
ill be using it with a E6700 with mushkin redline.
also i have a x700 pro in my old rig would it be beneficial to use it as a physics card.

[Wiggy McShades]

is anyone else having trouble getting 400fsb with an E6700 ive tried up to 1.4v on the northbridge but it gets more unstable after adding more than that. whats everyone getting with single dual core cpu's for fsb on these boards i think im doing something horribly wrong