Live MB OC Report :: ASUS Rampage Extreme

[CryptiK]

Asus won't repair my board, they found LN2 traces on the PCB :p

My 2nd REX I have here is a dud;, works fine till 500FSB with high ram clocks. Once I go over 500FSB I need to keep ram clocks sub 1600

LN2 traces? What does that mean - like residue of some sort?

[Leeghoofd]

I got no idea man... I can't see anything on it... To me it's a cheap way of ditching a client...

I'll try via Asus HQ and not via the shop... (maybe the latter fooled me) I haven't seen the official RMA refusal paper ( Alternate just told me this on the phone )

[JohnZS]

8GB not sure, 4GB starts to load it up pretty hard, and I had trouble with 460 FSB+ stability with higher (DDR3-1500MHz +) ram speeds with my Q9650 and 2 x 2GB sticks. However if you go for 400 FSB perhaps you could get 8GB stable at DDR3-1600 MHz. It's a brilliant board though, and should be much better than what it replaced.

Thanks for the input CryptiK
I guess if the worst comes to the worst I may have to drop a divider (a feature which does not work on the P5E3 standard). So I guess 1600FSB will work with RAM speeds of 1334? or 1200? however ideally I would like to run 1:1

Even if it means replacing my RAM.

I am guessing I would require a bump in NB Voltage, to around the 1.33-1.45V region for 8GB @ 1600FSB?

Leeghoofd did you spill LN2 on your board?
Fingers crossed you get the problem resolved...
John

[Expat GriZ]

Asus won't repair my board, they found LN2 traces on the PCB

WHHHAATT????? Maybe eraser?? Grease??? Sounds like a load of crap man. They've always treated me pretty well.... Is this from the shop in Aarteslaar on the Boomsesteenweg??

Let me know.......

[Leeghoofd]

Jup that's the one... picking it up on thursday, I want to see the official paper. He said they found LN2 traces on the board and some erasor leftovers in the socket... The latter which would be very awkard to me... But let's just wait and see before going bezerk.

Anyway I will take the board to the Asus tech conference in Frankfurt. Either someone will stand up and tell me what designed for overclocking really means and repair the board. Or it will soar around in the room :p Ninja style :p

[Expat GriZ]

Ratz....too bad I'm not home....you could of stopped in for a bevy...I'm 5 minutes from their shop...

Good luck with it in Frankfurt....

[JohnZS]

So then guys, what settings are you able to realistically obtain with 8GB and a QX9650 on the Rampage Extreme with Air cooling (Xigmatek S1824EE)
Thanks
John

[CryptiK]

I think you are going to have to be the guinea pig John. I have searched the thread, and can't find anyone pushing a quad and 8GB. Dual and 8GB yes (DerekT) and Quads + 4GB too (few of us but nothing really well tested).

[zsamz_]

So then guys, what settings are you able to realistically obtain with 8GB and a QX9650 on the Rampage Extreme with Air cooling (Xigmatek S1824EE)
Thanks
John

its gonna be a pain but you have unlocked multi so you should do around 4.0
gonna need 1.6-1.65 on nb

[JohnZS]

Thanks for the input CryptiK
It looks like I shall be the guinea pig unless some Xtreme lurker is hiding in the shadows somewhere with a Quad, 8GB and a Rampage Extreme

Ah zsamz_ it sounds impossible to me as 1.60V on the NB is way beyond my comfort zone.

I had a little play this evening, but to no success!

I get failure in memtest #5
NB = 1.31V
DRAM = 1.80V
DRAM Latency. 8-8-8-28-10-88-8-8
Now word of warning, it is wise (and yes you guys already know this, but I thought I would type this anyway in case the novice like myself is reading this) to MANUALLY set your CPU PLL and other voltages.
FYI @ 400FSB the motherboard [AUTO] For CPU PLL IS 1.70V!!!! This will most likely degenerate a 45nm CPU in a short period of time.
I personally would not be comfortable taking the PLL over 1.58V, but that is just me
I may visit this over the weekend, perhaps slacken the RAM off somewhat and introduce more DRAM Voltage and maybe a bit more on the NB. Perhaps raise that PLL to 1.56 or 1.58V (as I tested on 1.52V)

I can confirm though that this board is leaps and bounds better than the horrific P5E3 Standard!!
John

[zsamz_]

john i never ran below 1.63 on any of my x48 boards
i liked that xtra pl
its after 1.7 i be scared
but i've pumped in 1.8
btw 1.55 is really safe
nb gonna get hot with 8 gigs

my e6850 degraded @ 1.67pll some moron poopsted it was ok n .....my 4.2 chip can only do 3.99 now took 2 months
avoid auto if ya can is

[CryptiK]

Basically, errors in test #5 are ram related - try more voltage, different timings. Errors in test #4 and test #7 are often FSB/NB related - try changing vNB & vFSB, sometimes though its ram too. I found the REX will error in test #4 if you have vNB too high also, and they are quite heat sensitive (this may have been the cause ie: high temp and higher than necessary voltage). Once it gets over ~56*C you may run into issues as a result so make sure to actively cool the NB.

I found though that my board would do 625 - 630 FSB at vNB less than 1.60v (with a dual core obviously) and it just didn't need any more unless I was going for PL8 or something, so try other things before just cranking up the vNB.

[JohnZS]

Thanks for the input and advice guys
First of all I must say this is a VERY fine board. I have not had a lot of time to do any serious tweaking and OCing. But I have found with such a small bump in NB voltage (up to 1.27V) I am able to complete Memtest! @ 1600FSB and 1600Mhz on the RAM with a QX9650! (Quad Core).
There were a few other settings required for this to work.

RAM CAS Timings.
9-9-9-28-10-88-8-8
2N Command Rate.

CPU PLL = 1.56V
CPU Voltage, = 1.30V (I might be able to lower this a bit)
FSB Termination Voltage = 1.20V
CPU Ratio = 8 (for now to try and eliminate limiting factors)
Static Read = Disabled
Dynamic Writes = Disabled
Common Performance Level = 8
DRAM Voltage = 1.82V

Here is my memtest pass


Why a CPU ratio of 8 I hear you say. Well I have found that my system is 100% Stable with a ratio of 9.5 on STOCK FSB/RAM. (1333Mhz). So therefore I tried to find a clock similar to this with the FSB of 400.

It is not all good news as I cannot make it through 10 runs of linx. I get to run 8 and suffer an error, however I am able to gracefully close linx shut the PC down so at least it is not a BSOD or anything like a freeze. I am sure that this might only require a slight voltage adjustment somewhere (perhaps NB) to get me through.

However the fact I can get through memtest shows I am heading in the right direction. For those of you who do not believe how useless the P5E3 Standard is, I was unable to run 8GB at STOCK without an increase in NB voltage AND DRAM VOLTAGE... even then I would suffer random application crashes. After swapping out every component and testing in other PC's I finally found the issue was down to the board. The Rampage Extreme is a breath of fresh air. 1.25V on the NB is fine for QX9650 x 9.5 and 8GB of RAM @ CL8 and Performance Level of 7. This is completely game and linx stable and memtest stable.

However there is a slight BIOS Bug with the Rampage Extreme, which ASUS resolved for me on the P5E3 (so I may have to re-report it for this board). IF you set a manual multiplier which is higher than the default, when you COLD BOOT you have an error on POST about the CPU VID not matching the Ratio

Here is an example of the error message


John

[CryptiK]

Nice work so far John, keep at it.

[JohnZS]

Thanks CryptiK I shall try again at the weekend
If anyone else is able to reproduce the manual multiplier "VID" error message I took a picture of (which only happens on a COLD boot when a manual CPU ratio is set), please could you also report it to ASUS TSD. I have e-mailed them, but as of yet I have not had a response.
As with all things, this issue probably will only be resolved if a number of people are able to reproduce and report the problem.
Thanks
John

[Leeghoofd]

Yes I have the VID issue too with the QX9650, had it too with the P5K and P5Q Dlx boards...

Best suitable on my board is 9 X 450FSB and 4GB 1800mhz C7on the Rams. I use strap 333 and clock twister light to reduce the stress on the NB. If I go over 450FSB I need to lower the ram clocks to remain stable and use strap 400. But this REX's NB isn't the best... too bad my good one went down...

[JohnZS]

Yes I have the VID issue too with the QX9650, had it too with the P5K and P5Q Dlx boards...

Best suitable on my board is 9 X 450FSB and 4GB 1800mhz C7on the Rams. I use strap 333 and clock twister light to reduce the stress on the NB. If I go over 450FSB I need to lower the ram clocks to remain stable and use strap 400. But this REX's NB isn't the best... too bad my good one went down...

Ahh you have reproduced the VID issue too?
The annoying thing is when I had the issue on the P5E3, I reported it to ASUS and they fixed it in a BIOS update (1502). However they have not responded to me since I reported it on the Rampage Extreme

To me it seems very sloppy that they fixed it on one board, but did not carry over the fix to all effected boards.

So on air would you guys say that keeping your NB volts below 1.4V is a good idea?
John

[Leeghoofd]

If your rig is fully stable with only 1.4 then that's great. Mine needs 1.55NB voltage, NB temps go up to 50°C (with added active fan on the NB)

[Zeus]

I wouldn't mind pushing 1.5V through the NB for a daily setting. On water i pushed up to 1.78V for benching with no ill side effects.

[JohnZS]

Thanks guys
The problem for me is... temperature!! As others have correctly stated 8GB puts more stress and strain on the NB.
At Stock Clocks my NB reaches a staggering 51C during Linx (20 runs of "6896MB") and up to 54C according to the LCD Poster during heavy gaming. (Yes a Single PCB GTX 295 can make the NB rather hot when it is working hard).
When I raise to 1600FSB the NB reaches 59C during heavy gaming of 1hour and 52C during Linx.

I think 8GB + Quad Core really does hamper what you can achieve with clocks...therefore I am not going for anything Xtreme, just something which increases my performance mildly.

Leeghoofd you are a dab hand with a QX9650?
What Voltage would you suggest for the Vcore (on Air with a Xigmatek Heatpipe cooler)? I am aiming to keep bellow 1.3265V Vcore and 1.58V on the CPU PLL. Ideally I would like to acheive 3.4Ghz with 1600FSB, however I may have to sacrifice on the RAM clocks here.
Interestingly I have noticed that raising the multiplier requires a BUMP in NB volts (not CPU volts). Is this because >3Ghz Quad Core is more taxing on the NB?
Perhaps I am being too ambitious, but I am certain Intel did make a Xeon based 45nm "Core" series CPU which ran @ 3.4Ghz 1600FSB so in theory I could aim for this :p

Zeus
How hot did your NB run on air? and how much did moving to water cool things down?

Thanks again
John

[Don_Dan]

Did you change the thermal compound under the stock heatsink yet? If not, do it, it's worth 10-15°C less for me. The stuff Asus uses is very very crappy and poorly distributed.

Concerning NB volts, I'm not using the REX for 24/7, but I have been running many hours tuning for a LCC at 1.55V, I haven't seen a sign of degradation yet.

[zsamz_]

john i had a moded katana3 on my nb keepin her below 40c @ 1,63v

[JohnZS]

Thanks for the advice Dan,
I must admit I did replace the thermal stuff already on the polished side of the NB with AS5, however I did not take the whole heat pipe assembly off as I was worried about breaking something..
Anyway I might be on the way of getting some system stability.

Here is what I have managed to achieve so far.


Uploaded with ImageShack.us


Uploaded with ImageShack.us



Uploaded with ImageShack.us

As you can see I am on the way to getting stable. Once I am certain the system is stable I will post my settings in case they help people with 45nm Quads and 8GB.
I think setting "Clock Twister" To Moderate is a very good start to gain stability with very little in the way of Voltage Increase on the NB.
John

[JohnZS]

I think I might be getting there guys... any ideas?






John

[Leeghoofd]

John is you test with LinX make sure you set ram to max or a higher value then just 1024Mb, this should take over 30 minutes easily... I always test too with Hyperpi 32M (4 instances)