Updated: 1300 Mhz, 32M Stable. 2.45 volts. Asus 680i. Mushkin PC6400. 56k Warning.

[Kunaak]

I been testing the Asus P5N32-E SLI board pretty much non stop every hour for the last 2 days to figure out how to get the best OC from it, and what works and what doesnt and where to get things stable... and basically just trying to figure out the quirks of the board, and how to work around them.

so far, there are a few drawbacks, but nothing that would make me reconsider my buy in this board, it's a excellent board, and should only get better with additional bios updates. easily better then my P5WDH, and almost better then my P5B Deluxe... I am still very drawn between both boards. the P5B OC's better on FSB, but does worse on ram OC's.
the 680i does really good on the FSB, but not as good as the P5B, but is definatly in its own league when it comes to ram tweaking.
so I really cant say what boards better... it really comes down to what kinda overclock are you looking for really...

the ram, is my Mushkin PC6400.
same ram I used to get the WR for cas 3-3-3 overclocking last month, at 1060 mhz at 3.47 volts... if you wanna take that WR, go ahead, cause I am not touching that kinda voltage again, once is enough.
I think it took a week before my heart went back to normal after that day...
thats alot of money to put to that kinda risk, and I certainly cant replace something I kill, if its real expensive.
so once is enough...

Hmm, random hardware used...

E6300 that tops out at 3.87 ghz on P5B Deluxe on air.
Corsair 620.
ATI Rage Pro PCI card.
2 x 120 mm fans on the PWM and Ram.
1 x 40mm fan on the NB.
lots of trial and error.

first off, I want to say this...
I have no way to verify anything with this board.

CPUZ, Everest, Sisoft Sandra, Crystal CPUID, Even Nvidias own Ntune... nothing works.

nothing will show me the Cas, or the Dividers used.
all I have to show the MHz, is Bios, and Ntune, cause while Ntune 90% doesnt work on this board, it will show the CPU MHz, and Ram MHz.

so other then this, this is the best I can do to prove to you guys, this is what I say it is.
the Bios Matchs the information I get in windows, theres no errors... and the results do produce tangible improvements which I will also try to show.

first, the Bios shots to show what settings I used.

These are the voltages set right from bios.



these are the results of the voltages set from bios, after reboot.
all voltages overvolt, except the Vcore, which isnt even close to what you set in bios, and is usually off by about .1 volts. everything else, overvolts.



this is the cas used. forgive the shot, its taken from a CRT, and occassionally I get the big black bars blocking various info, and I just dont care enough about this particular shot to redo it, its the best of about 10 I took, and I am sick of trying to get the perfect shot.



this is the setting I used. Unlinked. so the divider is determined by the speed I set in. even dividers are extremely stable, odd ones are hit or miss.
1200 mhz = extremely stable. 1222 mhz = kinda stable = 1250 mhz = extremely stable, 1272 mhz, very unstable, 1300 mhz = mostly stable, 1333 mhz = barely stable for a few minutes, 1400 mhz = no boot.

[Kunaak]

and the Actual Boot Screen Mhz.

first 1200 mhz. cas 4-4-4

this was just testing....
about the only way I can show that bios and windows show the same thing...



and 1300 mhz, this is where the majority of my testing has been done lately. cas 5-4-4

[Kunaak]

and finally the results.

these are to show, that theres definatly a difference between 1200 and 1300 mhz, and not just some tiny questionable difference.

Super Pi First.

1200 mhz cas 5-4-4 = 14 minutes 43 seconds.



1300 mhz, same cas, 5-4-4 = 14 minutes 30 seconds.




and now Individual Ram test.

Crystalmark.

I personally like this benchmark, cause its really really easy to spot when a latency has a effect on the results.

1200 mhz, cas 5-4-4 = 18,088.



1300 mhz, cas 5-4-4 = 19434.

[Kunaak]

and finally to anyone curious... these are not suicide shots.
they seem to be extremely stable, and are very repeatable and boot every time.

its only when I pas 1333 mhz that it becomes extremely unstable...
weird coincedence, that that just happens to be Intels next FSB??

I'll leave that to mulder and skully...

ok, let the onslaught begin, you guys may skin me alive now....
bring on the pitchforks and flamethrowers, cause I expect this is gonna be alittle hard to believe for now, atleast, till some others start getting and using these 680i's

I've tested this for the last 24 hours over and over... the results are always the same... theres no variation, and I waited all day to test them and retry them and do my best to verify them...
I held off to post them, cause I honestly dont think anyones gonna believe this....

[AndreYang]

I will try my ROG mobo. Maybe it's CPU problems. I didn't use E6300.

[EnJoY]

Yea, those results are nuts. Why such little improvement in pi times when going from 1200mhz all the way to 1300mhz with same timings?

How did the sticks fair on 965/975?

If this isn't bugged (and I'm not saying it is at all, or doubting your talents), then it will really be astounding.

[erwinz]

nice kunaak.. tnxs for sharing..

[Kunaak]

Yea, those results are nuts. Why such little improvement in pi times when going from 1200mhz all the way to 1300mhz with same timings?

How did the sticks fair on 965/975?

If this isn't bugged (and I'm not saying it is at all, or doubting your talents), then it will really be astounding.

the sticks topped out at 1180 mhz on the P5B deluxe, and is in the 32m database. which you can see here.

http://img179.imageshack.us/my.php?i...as54410mz9.jpg

as for Super Pi... I dont know.
I am not a tweaker for that particular benchmark.
I dont aim for the best times, but just to run the benchmark.

I know this however.
with my gigabyte DS3, and ram at 1 ghz, and same OC, the board was doing 15 minutes, 55 seconds.
I remember, cause it drove me nuts trying to replicate a OC someone showed me, where they did 15 minutes exactly, same hardware, same board... however, that person disappeard cause thier OC's began looking fishy, so I got worked up trying to figure why my board was slow... when it wasn't....

this being at 14 minutes 30 seconds, is extremely good isn't it?
I mean, its a allendale, at only 3.5 ghz.

[G H Z]

There's only one 32M score up on hwbot that even comes close at this Mhz so I'd say the performance is good. It looks like Frank has a little project ahead for these async boards. Nice job kunaak, when you get the urge let the Maximizer loose on this thing and see what happens

[EnJoY]

the sticks topped out at 1180 mhz on the P5B deluxe, and is in the 32m database. which you can see here.

http://img179.imageshack.us/my.php?i...as54410mz9.jpg

as for Super Pi... I dont know.
I am not a tweaker for that particular benchmark.
I dont aim for the best times, but just to run the benchmark.

I know this however.
with my gigabyte DS3, and ram at 1 ghz, and same OC, the board was doing 15 minutes, 55 seconds.
I remember, cause it drove me nuts trying to replicate a OC someone showed me, where they did 15 minutes exactly, same hardware, same board... however, that person disappeard cause thier OC's began looking fishy, so I got worked up trying to figure why my board was slow... when it wasn't....

this being at 14 minutes 30 seconds, is extremely good isn't it?
I mean, its a allendale, at only 3.5 ghz.

Well it isn't the overall time that bothers me, I'm not actually sure how good it is comparatively myself either. However, the seemingly miniscule difference between your scores at 1200 and then at 1300 are what make me wonder if it's really doing what it says it is.

I forgot your previous result was in my db, my bad.

[Kunaak]

for anyone curious...

I talked to frank about this board this morning, and basically, he's working on it, but Nvidia isnt giving up any information, so it's gonna be fighting tooth and nail to get Nvidia and CPUZ to work together, so till Frank or someone figures out how to get the information read...
we're all screwed on overclocking and having a easy way to show results with these boards... which is really sad, cause they are damn good boards...

by the way... this board doesnt need the maximizer.
it does all the way to 3.5 volts right from bios, as you can see here



by the way, my review post on this board, is here....

http://www.xtremesystems.org/forums/...d.php?t=123227

[Eldonko]

Wow these 680i boards are huge for mem clocking... crazy results :o

[metro.cl]

looks really good, 1300mhz is a new place to bench now

[Kunaak]

I just got done reinstalling my OS, cause I was trying to get 1333 mhz stable, and I screwed up the OS, cause it was doing a "begining physical dump of memory" and I got impatient and just rebooted it... which screwed it up.

anyways... now I am testing 32M from 800 mhz, to 1300 mhz, at the same cas and speed...

aaaaaaaaaaand I am pretty sure the thing just got faster...
Super PI 1M is faster across the board.

if anything, it looks like my 14 minute 30 second run... was slower then it could have been. I guess I'll find out in about an hour after running 32M a few more times (800 then 1000, and 1200, then 1300 all cas 5-4-4-12 2T TRC 6.)

[charlie]

should be interesting results on the 32M runs. I'm really likin' this board.

[EnJoY]

Kunaak, Tony has said that 1T performance is really where it's at with these boards from his testing. He's achieved 900mhz 1T with cas 3 (iirc) with reasonable volts on his ram/board and says it performs better than the high mhz. I'd be interested to see your results on that as well if you can get your ram clocking well at 1T.

[guess2098]

ROG is @ newegg
http://www.newegg.com/Product/Produc...82E16813131074

[Kunaak]

Kunaak, Tony has said that 1T performance is really where it's at with these boards from his testing. He's achieved 900mhz 1T with cas 3 (iirc) with reasonable volts on his ram/board and says it performs better than the high mhz. I'd be interested to see your results on that as well if you can get your ram clocking well at 1T.

sure I will give it a try, I certainly got ram capable of cas 3 (WR cas 3 ram .... for now....) and 1T is certainly easy at 900mhz, so I can try that.

by the way.
1311 mhz so far stable.

heres how the ram scales from 800, 1000, 1200 then 1300.









the results you see here... are correct.
I have done each one atleast 3-4 times in the last couple hours.

800 mhz, is just slow....
I have a theory on why.

with a 2000 QDR FSB, and 800 mhz, your ram has less bandwidth, then the FSB is capable of, so the ram is holding the CPU back.

at 1000 mhz ram, its near perfect for QDR 2000 FSB.

at 1200 mhz, the FSB is then slower then the ram so the CPU then starts holding the ram back, and you get a smaller boost in score then.

1300 mhz, is pretty much the same thing as 1200 mhz in this theory then....

thats my theory.

also, I have found 1 bug.
if you set the ram to 1000 mhz, and the multiplier to 6... then when you get into windows, it shows 1200 mhz.
set 1200 mhz, and it shows 1400 mhz.
set 1300 mhz, and it shows 1512 mhz.

so Ntune is bugged if you use anything besides the default multiplier.

Also, the multiplier used, effects the bandwidth.
that was a surprised me.
6X consistantly has less bandwidth then 7X.

[Kunaak]

Kunaak, Tony has said that 1T performance is really where it's at with these boards from his testing. He's achieved 900mhz 1T with cas 3 (iirc) with reasonable volts on his ram/board and says it performs better than the high mhz. I'd be interested to see your results on that as well if you can get your ram clocking well at 1T.

Tony is indeed correct.

while I am not running 900 mhz, 875 should do just fine...
I dont want to pump 2.7 volts into this ram to get 900 mhz stable, when 875 is fine at 2.45 volts (for every .1 volt I get about 10-15 mhz)

this is cas 3-3-3-Tras 3, 1T, TRC 1.

this is the most tweaked I could get, and my 1300 mhz run was also my most tweaked, so its safe to say, 900 mhz at 1T is definatly faster then 1200 mhz at 2T.

cause I have just under 900 mhz, and its almost faster then my best 1300 mhz run.

I wish I had a spare videocard though, I really want to test this in real games and such...
I got alittle to spare, but I just can't bring myself to buy a cheap videocard.

[IluvIntel]

Kunaak, are you able to get a hold of Mushkin PC8000 redline sticks and try them in this board? I was seriously thinking of this mobo & ram combo. But not sure how they would run...

[Revv23]

wow very nice and thorough testing kunaak, the clocks are great. Too bad it looks like 975x can still beat it in performance even with lower clocks.

[althes]

Thanks for the testing

[EnJoY]

That 1T performance is really cool. It's affect on latency is clearly huge for conroe.

[politenessman]

nv monitor does not show cpu core speed or fsb or ram speed correctly on my evga. it is the latest version. i am only running 1210mhz ram speed. anyone else having this prob? anybody know of a reliable utility that works?

[Kunaak]

nv monitor does not show cpu core speed or fsb or ram speed correctly on my evga. it is the latest version. i am only running 1210mhz ram speed. anyone else having this prob? anybody know of a reliable utility that works?

look at a few posts of mine ago.
if your not using the default multiplier, then it will show the wrong speed.

from my post...

also, I have found 1 bug.
if you set the ram to 1000 mhz, and the multiplier to 6... then when you get into windows, it shows 1200 mhz.
set 1200 mhz, and it shows 1400 mhz.
set 1300 mhz, and it shows 1512 mhz.

so Ntune is bugged if you use anything besides the default multiplier.

Also, the multiplier used, effects the bandwidth.
that was a surprised me.
6X consistantly has less bandwidth then 7X.

since you have a E6600, and your FSB is 1980 mhz, this means your running 8 x 495 for 3.96 ghz (impressive by the way) but this also means your ram speed is gonna be shown wrong.

set your multiplier back to 9, and then boot into windows, and you will see that your ram speed will then be reported correctly.
just like I said in my post a few times ago.