DFI 680i LT Quad Core overclocking

[virtualrain]

How is the 723 BIOS? The Nemesis said it didn't do too well above 390 FSB.

I have not tried it personally. My board is in for RMA at the moment... Code 88 Dead! Seemingly for no reason.

At least DFI offers a cross-ship program.

[famich]

Hello, I am running ¨my QX6700at 3600MHz with just 1.44VC
on water , temps over 80C are too much for my liking anyhow, my screens as well.I got the ES , it s running abit cooller than retails.

[justin_c]

yeah, i managed to lower my voltage all the way down to 1.38V without 4 core prime (with roundoff checking) error-ring or crashing. for 3.6ghz 400x9.

[justin_c]

update: if i tweak the cpu voltage special add to exactly the right amount and maintain at least 1.38V, it passed 3 hrs of 4core small fft prime at 3.6ghz. so i think i have some room for 3.8ghz prime, but will need another day of gtl tweaking. gtls for 3.7ghz (411x9) were 142/79/77, also 3 hrs stable.

[jonbass]

I have finally managed to assembled my SS on the CPU,

This weekend I assembled my SS on my CPU, I used to have some problems with my SS but now it seems to work fine. The only problem right now is the high temperature, I am not sure what is causing this but I will try to find the source to why the temp are so high.

Anyway, everything got finished this Sunday so i did some initial tests, must say that i am pretty happy with the results so far.
It seems like this CPU aren’t just great on air it also over clocks great when putting it under some cold, now I haven’t done any serious test yet just some initial tests so I could se that everything work like It should.

It seems like this CPU not only loves low voltage with air, it also like low voltage with more extreme cooling. Only with 1.41v I managed to run 3D06 in 4200Mhz, then I raised the voltage to 1.44v and managed to run 06 in 4282Mhz and Super PI in 4450Mhz.
Now I believe this isn’t the limit with this voltage, I stopped benching because the clock was so much not because the CPU was getting unstable. I believe I will be able to run 3D 06 in 4300-4350Mhz with 1.44-1.45v, and I am very happy with this.
I hope i will be absle to continue my test today, my goal is to get this CPU 3D mark stable around 4400-4500Mhz

[jonbass]

Like said before, i am pretty bummed out that my SS don’t cool my CPU as good as it used to, last winter with my QX6700 temperature was really good but something happened and it doesn’t cool my QX6850 as good. But despite of higher temperatures I am very satisfied with the results so far, I am still using pretty low voltage but this CPU just keep on going.

Here are the results from last night, I will off course continue raising this weekend and will post more screens when I have something to show.


On these results both the FSB and GPU are running low because I want to be sure it’s my CPU causing problems if the system crash, because I’m still trying to find the limit for the CPU I don’t push the rest of my system. I will off course push the rest of my system as soon as I know how high my CPU will go.



Jonbass 2007-09-01
I started warming up the system with some Super PI 32m, now this was the first 32m run so I wasn’t sure how high I could go, I started with 4245Mhz and 1.43-1.44v and this worked fine.



Jonbass 2007-09-01
Then on with 3D mark 03, 4450Mhz 1.46v,


Jonbass 2007-09-01
3D Mark 06, 4411 Mhz 1.5v, nice to pass 4400 Mhz.


Jonbass 2007-09-01
3D Mark 06 CPU test, 4450Mhz 1.5v,


Jonbass 2007-09-01
The last run was Super PI 1M, I managed to raise the CPU to 4611 Mhz, this with 1.475-1.48v.

I will post more results later this weekend.

[justin_c]

i was able to run a suicide bench at 3.95 ghz for 3dm06, getting 19637, but IVE FORGOTTEN THE GTL REF VOLTAGE numbers for that clock. now i have to find them again, hopefully i have the driver and tweaks enough to get 20000.

[virtualrain]

Wow... this board is a PITA! (but satisfying when you finally get something that works)

I got my RMA replacement board and of course started with my old settings trying to replicate my previous overclock:

OC: 400x9 (2CTST521 BIOS)
Vcore: 1.6V
VTT: 1.6V
GTL: 195/195/95

Well, of course those weren't stable. Prime95 25.4 was giving me a rounding error within a few minutes. I upped the Vcore to 102%... same problem... upped the Vcore to 104%... much better.

After this, Prime95 (Small FFT) would crash (not just error but crash) after about two hours. So I started playing with GTL starting with settings that had worked for others in the past...

170/95/95: 1.25 hrs to crash
200/115/95: 2 hrs to crash
200/170/100: 2 hrs to crash
220/200/100: 2 hrs to crash
220/200/180: 8+hrs stable

So, oddly, it seems my GTL settings for NB GTL are radically out of line from anyone else. WTF?!

Now trying Prime95 Blend to see how stable that is.

If that's stable, then I'm going to try lowering Vcore a bit and see what happens.

It's rediculous how long it takes when you have to wait 2+hrs to see if you have good settings. @#$%!

Edit: A note about PSU and shutdown isssues... I learned something further about the enormous power this sucker consumes with a quad under heavy OC/load... Unlike my original testing with this board, I'm running dual 500W PSU's... one is only connected to the 24pin ATX motherboard connector and the 8-pin ATX CPU connector... thus powering the motherboard and CPU. The other PSU is powering everything else... pumps, fans, storage, GPU, etc. As soon as I started Priming Small FFT's the board would shutdown. It turns out that my PSU's are true dual-rail designs with 22Amps rated current per rail. The two rails appear to be split with one rail serving the CPU 8-pin connector and the other rail serving everything else. Thinking that maybe the Small FFT's were tripping the rail limiter on the PSU, I rewired the 8-pin connector so 4-wires were now coming from the other rail (stole them from the PCI-e GPU connector) and the other 4 were left as is. Now the 8-pin CPU power connector had 4 wires from each rail and thus a total for 40+amps available to it. Voila... no problem. The power consumption is just mind-boggling though.

Stay tuned.

[Kampfzerstorer]

I can't imagine how you guys are getting clocks like these outta this board. I can't even get this crappy e6850 to do 3.7 100% stable.

Great work.

[virtualrain]

I can't imagine how you guys are getting clocks like these outta this board. I can't even get this crappy e6850 to do 3.7 100% stable.

Great work.

It ain't easy... that's for sure!... Days of trial and error! (Also not every processor is going to do 3.7GHz... you know that right? )

Here's an update to my trials and tribulations...

From my last post, small FFT's were 8+hrs stable at...

OC: 400x9 (2CTST521 BIOS)
Vcore: 1.6V + 104%
VTT: 1.6V
GTL: 220/200/180

On my pre-RMA board, my settings were

OC: 400x9 (2CTST521 BIOS)
Vcore: 1.6V
VTT: 1.6V
GTL: 195/195/95

When I tried blend on the new settings, it ran for 4hrs and crashed... I was a bit disappointed and got to thinking about the GTL logically. If I understand things correctly, the three GTL settings control the reference termination voltage for the first CPU die, second CPU die and the NB. Hence, it made sense to me that if I needed 195/195 on my old board for die 1 and 2 respectively then since I'm using the same CPU on my new board, I should use the 195/195 settings on this board.

Since the NB silicon is different between the two boards it would make sense that the third (NB) GTL setting would have to change and this was obvious from my small FFT stability work earlier. So on this hunch, I tried the following combination of old and new GTL settings:

GTL: 195/195/195

Blend is now running stable after 6+hrs. I'll kill it and be satisfied in a couple more hours.

Then I'll set to work on trying to lower Vcore.

[Borgod]

Your identical GTL1 & GTL2 was always an abnormality with your first board according to recommendations by Tony, Victor etc. But 2 boards with equal successfull results is certainly interesting!

I'll be revisiting the OC on mine soon!

[virtualrain]

Your identical GTL1 & GTL2 was always an abnormality with your first board according to recommendations by Tony, Victor etc. But 2 boards with equal successfull results is certainly interesting!

I'll be revisiting the OC on mine soon!

I know! Now my NB GTL needs to be the same... which is wildly different from anyone else AFAIK.

The good news is that it's now blend and small FFT stable for 8+hrs.

I tried lowering Vcore without much success. It would crash within an hour to a few minutes depending on how much I lowered it. So my idle Vcore is 1.6V and under load it droops to 1.55V... Good thing I'm running a dedicated CPU cooling loop with a PA120.3! Temps are still manageable at low 60's under full load.

So, to summarize, my replacement board required more Vcore and radically different NB GTL settings to reach the same overclock as my old board.

[virtualrain]

Here are my results/settings for the replacement board...

Prime Stable 8+hrs:



BIOS Settings:

[jonbass]

I haven’t been benching for about two weeks, the air here in the north of Sweden have been so moist so i had problems with condensation so i decided to take a short break. During these weeks i decided to volt mod my graphic cards and also tune my SS to get better temperatures, and everything seems to be working great now, my CPU temp are almost 4-5 degrees lower than before, also my better 8800Ultra seems to answer really great on the voltage modification.


Lets start with my 3D mark 03 result, this is with my CPU running 4375Mhz with 1.47v, the graphic card is running 750-1217Mhz with 1.37v.
I am also running my CPU lower so I can be sure it wont cause any problems when I am trying out my GPU.




And off course my 3D mark 06 result, this one I am very happy with, now I am pretty sure it will go higher but like I said this was only an initial test. CPU is running 4375Mhz and the GPU 760-1217Mhz with 1.37v.

[denden]

hi what u mean SS to mod? does that mean volt mod?

I haven’t been benching for about two weeks, the air here in the north of Sweden have been so moist so i had problems with condensation so i decided to take a short break. During these weeks i decided to volt mod my graphic cards and also tune my SS to get better temperatures, and everything seems to be working great now, my CPU temp are almost 4-5 degrees lower than before, also my better 8800Ultra seems to answer really great on the voltage modification.


Lets start with my 3D mark 03 result, this is with my CPU running 4375Mhz with 1.47v, the graphic card is running 750-1217Mhz with 1.37v.
I am also running my CPU lower so I can be sure it wont cause any problems when I am trying out my GPU.




And off course my 3D mark 06 result, this one I am very happy with, now I am pretty sure it will go higher but like I said this was only an initial test. CPU is running 4375Mhz and the GPU 760-1217Mhz with 1.37v.