I been getting a ton of people asking me in PM's and in my posts about advice and such for overclocking the 680i chipsets, and how I can get the results I have from these boards.
so, I'll try to write down anything I can think of about these boards thats worth knowing.... atleast, that I personally have tested and know 100% for sure is working for me.
however, what works for my board, may not work for yours, so dont think this is a conclusive post or anything.
First, Alittle about me and the 680i.
I use the Asus P5N32-E Sli. stock default bios.
no mods (I would if I could find any, and havent got curious enough to start looking for them myself)
I have No SLI cards.
why this is important to mention?
cause lately, I see a whole lot of 680i "reports" that have little to nothing to do with the board, and get overshadowed by the 8800GTX's or such being used.
which tells me nothing.
zilch. nada... zip, as useless as if I didn't even see the post.
a SLI board that can run SLI... wow... amazing... neato...
but what about the board?
well, thats where I come in... kinda.
see, I've spent the last week, doing nothing.... notttttttttthing related to videocards, no 3Dmark, nothing related to anything other then whats in the BIOS and what will get me more speed.
I spend atleast 3-4 hours a night booting, and rebooting that board over and over and over again, changing 1 setting at a time to find what works and what doesnt and where the glitchs are, and where they arent.
sometimes, the results are very curious...
heres a few.
1. CPU HT Voltage.
easily the most important voltage on my board when it comes to breaking 500 FSB's with it.
no voltage, not Ram, not vcore, not chipset.... no voltage does as much for this board as this one voltage, CPU HT Voltage.
stock voltage is 1.2 volts.... tops is 1.55.
what its related too?
not a clue... but what it does, is very clear.
it gets me high FSB's and makes them boot everytime.
take this voltage down even alittle, and you find a quick loss in overclockability with this board.
I just leave it at 1.55, it doesnt seem to affect the CPU temps, but it does have a direct effect on the results, and not a small one either.
in my case 1.2 volts will work fine to about 420.
1.55 volts will work fine to 525.
thats not a small range as you can see...
so thats where I suggest you start first, if you want to try going higher.
I dont think the reference 680i boards from EVGA, BFG, ECS or anyone are the same caliber as this Asus board, since they are all stock Nvidia boards, so I dont know how much this will help anyone with a Stock board, but if you have a Asus 680i, definatly give this voltage a shot.
2. CPU Multiplier.
this is where things get very tricky with this board.
it took me a few days to notice this, but the CPU multiplier has a direct effect on the bandwidth of your ram.
in my case, going from 7 x 500, to 6 x 500, loses me about 20% in bandwidth, I've tested this with Crystal CPUID, Sisoft sandra, and Everest Ultimate Edition.
its the same everytime...
the lower your multiplier goes, the less bandwith you get for your ram.
so 1000 mhz on the ram, at 6 x 500, is not the same as 1000 mhz at 7 x 500.
Lowering your multiplier on these boards has a drawback that you might not wanna live with, especially if your overclocking a high multiplier CPU on Air, like a E6700, where you been lowering it on the 965/975 chipsets with no penalty to your speeds in things like Super Pi, but if you lower the Multiplier from its stock default multiplier on the 680i, you will get a slower score then you might be expecting, and it's probably directly related to the bandwidth loss you suffer by lowering your multiplier.
3. Ram Speed Glitch.
Theres a Glitch that one can Exploit when Overclocking with the 680i boards, and it's signifigant enough that its worth mentioning.
your ram speed will be reported wrong in a few programs like Ntune, if you lower your multiplier.
set 1000 mhz in bios, it boots at 1 ghz, but you ran a 6 multiplier, instead of 7.... when you look in Ntune, it will report 1200 mhz.
this only happens when you lower the multiplier.
I havent taken the time to see if this glitch effects CPUZ yet.
(its thanksgiving, and I just got home)
so this is something to be aware of, to avoid fakers and false results.
4. 1T Command Rate.
This ones a iffy one, cause some people will want to not believe it.
1300 mhz, cas 5-4-4-10 at 2T.... is only alittle faster then 900 mhz, cas 3-3-3-8-1T.
regaurdless of the speed used, linked or unlinked... 1T is very fast, when you get it up to 900+ Mhz.
I tested this at 3.5 ghz (7 x 500) many many times.
unlinked mode always, and setting my ram speeds manually.
cas 3 at 900 mhz, and 1T.... while not exactly easy to get stable, is almost as fast as 1300 mhz at cas 5-4-4 at 2T.
while I admit, I have not tested this in gaming yet... in everything else, this has been confirmed for me over and over.
heres where it gets alittle fuzzy.
1300 mhz... while you may not believe it, is alot easier on these boards then it sounds. I am starting to think, that pretty much any High Quality ram has a real good shot at 1200 mhz easy on these at reasonable voltages, and 1300 mhz if you push it alittle.
but at the same time, cas 3 at 900 mhz at 1T, isn't so easy.
it takes extra voltage usually, and is harder to keep stable.
but should prove to be great for extreme benchmarks should you choose the 1T/low latency route.
so it's not so easy to say one is faster then the other, cause high MHz is definatly easier to achieve, then moderatly high speeds with low cas at 1T is.
it wont suprise me if this is a aguement that is very hard to have a conclusion with.
5. Heat and Overclocking.
if you take nothing I say here serious, atleast take this one point and listen for a second.
these boards run really.... reaaaaaaaally hot.
my Asus board here has heatpipes all over it, but after about 5 minutes of it on, its clear the heatpipes on this board are doing little to nothing for this board. this board should not be run with any form of passive cooling if you intend to overclock.
active cooling is a must, and not just for the NB.
the SB and PWM area all too get very very hot when running overclocked for long periods of time. to give you an example, how I found this out was.... I was poking around the board with my flashlight looking for some ICS, so I could see if there was a clockgen available for this board.
the board was on and running...
I tapped the NB heatsink, with my flashlight hand... and I burned my hand.
not like alittle warm... I mean burned.
at the time, the board was heavily overclocked, and running some high voltages... but no different then any other overclocker would do.
to say it was hot, isnt the right word.
I've seen boards that run hot in the past... especially after days like the first prescotts.
this board makes the prescott look like a ice cube.
these boards need far better cooling then I am seeing on them at stock.
it might take a hour or so of heavy overclocking or gaming to see what I mean, but touch the heatsink and you will see what I mean.
put fans on there, anywhere you can if you overclock.
these things need it.
6. Latency and "Strap".
this board definatly does run with some form of "Strap".
however, where they are, isnt exactly clear yet.
what I mean is... passing 1066 FSB, doesnt mean your gonna automatically get the 1333 FSB strap kicking in.
when your at a low FSB like 333 or so, your gonna be around 50-55 NS for latency.
when you get to above 500 FSB your gonna be about 60-65NS for latency.
boot at 9 x 400, and you will have a faster set of ram, then if you booted at 8 x 450, cause the latency and the multiplier bandwidth loss would make it less of a good overclock then you might think, even though your ram appears to be running faster, you might find the difference is shockingly small, cause your taking a hit in 2 ways.
the latency of the chipset changes... and your lower multiplier then the stock one, also makes your ram take a high on its bandwidth.
so test your OC's alot before you settle for one, cause just setting one speed without testing it... may result in a slower speed then you thought you had, cause of things you cant actually control, like the chipset latency, and the bandwidth you lose by lowering your multiplier.
7. 1333 Mhz Wall.
this is gonna sound strange, cause its only a hunch...
but getting 1300 mhz with these boards is almost too easy.
however the closer you get to 1333... the sharper your stability falls.
I have 2 sets of ram that easily hit 1300.
but its the same in both cases. theres a drastic fall in stability once you get near, or pass 1333 mhz on your ram speed.
so drastic, that it actually seems like a wall, more then a hardware limit.
getting up to about 1310 mhz is easy with alittle stability.
1320 will work for awhile.
but 1333 is a quick and drastic change in stability.
my guess... and yes, I have no proof of this.
is that theres a hardware wall within the bios, to make 1333 mhz not possible.
why?
cause the 1333 FSB is intels next CPU FSB.
and Ive seen this before with the 925x chipset, where they didnt want people taking 200 FSB CPU's changing them to 266 and getting a cheap extreme edition.
so there was a near solid wall at 260-ish.
but that same wall, didnt exist for CPU's with the 266 FSB.
would they do it again?
sure, why not?
this gives them a reason to push the 1333 FSB even more, if they appear to OC better... when they really are just getting a free jump over a bios limitation.
so thats my little conspiracy theory.
8. Linked or Unlinked.
this is something I really dont know where to go with...
I've heard people claim that Unlinked is slower then Linked.
however, after about 20 trys with Super Pi, and various bandwith tests.... I can't see to find any differences, between linked or unlinked.
I mean none... I get no benifit from running linked, then I do unlinked....
9. Dividers.
when you run stock, you have a few usual dividers to play with, like 3:2, 5:4 and 1:1.
when you run unlinked, you have a whole range of odd dividers that kick in to determine your ram speed.
however, some dividers seem to be less stable then others.
example.
for me running 7 x 500, I run unlinked.
unlinked at 1300 mhz, and Auto for the divider I get 1300 mhz even, and uses the 10/13 divider.
but run linked, and set the divider manually and I use 3:2.... which says it gets me 1300 mhz in bios, but reboot and I get 1333 Mhz.... not the 1300 it said.
to get it to boot at 1300 at 3:2, I gotta clock down 1 FSB to make the ram seen at 1299 mhz... but that turns out to be less stable, then if I let the bios just use the 13/10 divider, when I set auto and left it unlinked.
so try different combinations for your ram, and realize, some dividers are less stable then others.
10. OS Corruption and high Ram MHz.
heres the one biggest drawback to this board so far.
with how easy it is to et past 1200 mhz with these boards, your gonna wanna push your ram farther and farther.
if your anything like me, you will just keep trying and trying till you cant go any farther.
but once you pass 1330 mhz...
you will quick start to corrupt your OS, and it wont boot anymore, or constantly gives you BSODs just after the load screen.
usually, a windows repair install wont work either, cause by then, you probably corrupted most of your boot drivers, DLLs and such... so theres nothing for windows to work with, it may not even see that theres a OS there to repair...
so be very verrrrrrrry careful if you really get into pushing your ram, otherwise you might loose everything off your hard drive.
make back ups of your information that you dont want to lose, and get yourself a copy of Norton Ghost, if you really wanna start pushing your ram on these boards.
cause Data Corruption happens very easily past 1330 mhz.
thats it for now, I am tired of typing and I am full of turkey and damn I gotta sleep... this foods making me feel 100 pounds heavier and sleepy.
Reply With Quote
Bookmarks