***Asus Rampage III Extreme Owners Thread***

[fgw]

It is not so simple.
IIRC, if you put 3 VGA's (Tri-SLI) in slots 1, 2 and 3, you will have following configuration:
slot#1 @ 16x
slot#2 @ 8x
slot#3 @ 8x

not x8, x8, x16 that might be expected and what diagram shows. It has been argued before, asked Asus, and they said it's like that.
IIRC.

don't get me wrong donmarkoni, but i doubt this is correct.

if slots 1 and 2 share 16 lanes and also slots 3 and 4 share 16 lanes there is no way that slot 1 and slot 2 might get 24 lanes in total! might be wrong, but so might be the guy at asus who told you this.

anyway, as most of us wont use 3 way or even 4 way its more or less a hypothetical discussion. but if somebody with three cards at hand is willing to test this for us i'm curious to see the results.

ps. still waiting for my cpu and ram, so cant test this myself right now. besides that, my old cards wont support 3 way sli, so i'm lost here anyway ...

[Jeff Flowerday]

cool...vlad...I am seriously thinking about getting the 1200AX as well. I originally ordered Enermax 1250w rev 85+. Had to RMA as the fan was messed up. Readh JG's review but he failed to say anything about the noise. Then did some digging and while no can say the psu is not good. Its so freeking loud. I need the power and the silence. Corsair have an awesome PSU, just not something u would stick in a serious home workstation.

Silverstone 1500 is stealthy quiet. 19db right up to 1000W draw.

[kurumi_Japan]

Hi

what is the best current bios to use? the 878 or 0002?

thank you

I think "0878" is very nice one.
I just tested only 2D benchmarks, its improvement DRAM efficiency.
The result reported at my blog.
"0878" improve Super Pi 1M finished time 0.05sec faster than "9124".
(When I check SuperPi 1M time, I run these benchmark program over 20 - 30 times.
Same MEM setting, Same CPU clock.)

But I'm very sorry, don't know 24/7 OCing stability.

[donmarkoni]

don't get me wrong donmarkoni, but i doubt this is correct.

if slots 1 and 2 share 16 lanes and also slots 3 and 4 share 16 lanes there is no way that slot 1 and slot 2 might get 24 lanes in total! might be wrong, but so might be the guy at asus who told you this.

anyway, as most of us wont use 3 way or even 4 way its more or less a hypothetical discussion. but if somebody with three cards at hand is willing to test this for us i'm curious to see the results.

ps. still waiting for my cpu and ram, so cant test this myself right now. besides that, my old cards wont support 3 way sli, so i'm lost here anyway ...

Don't worry, I won't.
Just came home, moved my GTX 480 from slot #1 to slot #2 and what I found out is that bassplayer's statement "Slots 2 and 4 are hardwired for x8" is absolutely right.
I'm not sure about configurations of lanes when using Tri-SLI and could be wrong. I just might not remembered correctly, but anyway it would be a little stupid if MB could not re-route lanes to be 16x-8x-8x and run @ 8x-8x-16x, when it has 32 for those 4 slots.. Why the fcuk would I need third card running @ 16x
However, this will probably stay a hypothetical discussion as you said, if someone doesn't step up with 3 VGA's to test.

[kpablo]

It is not so simple.
IIRC, if you put 3 VGA's (Tri-SLI) in slots 1, 2 and 3, you will have following configuration:
slot#1 @ 16x
slot#2 @ 8x
slot#3 @ 8x

not x8, x8, x16 that might be expected and what diagram shows. It has been argued before, asked Asus, and they said it's like that.
IIRC.

You are wrong in there... It should be:

Slot #1 @16X
Slot #3 @16X

Or:

Slot #1 @16x
Slot #3 @8x
Slot #4 @8x

Or if you use Slot #2 it should be: (This i am going to test it to be 100% sure).
Slot #2 @8x
Slot #3 @8x
Slot #4 @8x

I was right, #2 and #4 will always be @8x Max.

Look at this 2 screenshots i just took:

[bingo13]

Warning : I also flashed to BIOS 0002 to find that the BIOS would not save its settings to the CMOS.
The previous BIOS before flashing was BIOS 0009. So I then flashed to BIOS 0878 and then back again to BIOS 0002,
and now it does save it's settings.

But I also noticed the BIOS dated on these.
BIOS 0009 2010-08-13 01:08:52
BIOS 0878 2010-08-13 05:34:32
BIOS 0002 2010-08-10 04:09:12

But bingo13 tells us that BIOS 0002 is " 4. Newer Code than the 0009 BIOS leak "
If this is so, then the dates and times on these are of no relevance. And thus, no way for us to know which is newer.

All three BIOS releases are built on the same base code that will be officially released shortly and combines the 0002/0878 changes.

The difference between BIOS 0002 and BIOS 0003 is temp reporting and a few minor changes on the LN2 side. BIOS 0878 does not have the same temp reporting changes in the 0002/0009 BIOS code but does have the open memory settings and will be just a bit faster in certain operations than the other two betas.

0009 was an internal build that never should have been leaked and due to the way it was leaked, it will slow down future beta releases from us until this problem is solved.

[bassplayer]

Don't worry, I won't.
Just came home, moved my GTX 480 from slot #1 to slot #2 and what I found out is that bassplayer's statement "Slots 2 and 4 are hardwired for x8" is absolutely right.
I'm not sure about configurations of lanes when using Tri-SLI and could be wrong. I just might not remembered correctly, but anyway it would be a little stupid if MB could not re-route lanes to be 16x-8x-8x and run @ 8x-8x-16x, when it has 32 for those 4 slots.. Why the fcuk would I need third card running @ 16x
However, this will probably stay a hypothetical discussion as you said, if someone doesn't step up with 3 VGA's to test.

I've always wondered why they don't give them all 16 physical lanes and then dynamically adjust them for best results.

However, if you look in slots two and four, you'll see that the pins only travel about half of the slot. Bummer

[tistou77]

I think "0878" is very nice one.
I just tested only 2D benchmarks, its improvement DRAM efficiency.
The result reported at my blog.
"0878" improve Super Pi 1M finished time 0.05sec faster than "9124".
(When I check SuperPi 1M time, I run these benchmark program over 20 - 30 times.
Same MEM setting, Same CPU clock.)

But I'm very sorry, don't know 24/7 OCing stability.

ok thanks

but I'd like to know 24/7 OCing stability

I have 802 actually

thanks

[donmarkoni]

You are wrong in there... It should be:

Slot #1 @16X
Slot #3 @16X

Or:

Slot #1 @16x
Slot #3 @8x
Slot #4 @8x

Or if you use Slot #2 it should be: (This i am going to test it to be 100% sure).
Slot #2 @8x
Slot #3 @8x
Slot #4 @8x

I was right, #2 and #4 will always be @8x Max.

Look at this 2 screenshots i just took:

http://img834.imageshack.us/img834/1903/84068001.png

http://img6.imageshack.us/img6/1873/40091534.png

I know. It is just plain stupid if it can't be re-routed to be 16x-8x-8x when using slots #1, #2 and #3. If you have 3 Tri-SLI capable VGA's and you are willing, please test them in first three slots, so we all know configuration of lanes.

I've always wondered why they don't give them all 16 physical lanes and then dynamically adjust them for best results.

However, if you look in slots two and four, you'll see that the pins only travel about half of the slot. Bummer

That would be best.
I didn't know that! Never took a look at them. However, it wouldn't pose a problem for 16x-8x-8x configuration.

Hmm... while writing, I looked at slots and they seem to have all pins. What is going on? Are there two versions of MB?

P.S. Will try to take a picture(s).

EDIT: Here are the pics:

slot #2: , slot #4:

EDIT2: There is a dead mosquito on slot #4 I didn't even noticed. This is so embarrassing.
But the bright side is that my Nokia N86 8MP takes such good photos.

[leighl]

Hi Guys I have joinned the club and got one of these Mb which I think are great.

This is my rig
i7 930 in the process of over clocking
MB R3E
6GB Gskill 1600 Mhz ram
GTX 480 SLI
Silver stone 1500W strider
Case HAFX.

Any way I have been overclocking this thing over the past month and last night I noticed that while stressing @3200MHZ Blck 153 with prime95 my core vid in CPU-Z was jumping from 1.085, 1.091 , 1.098 to 1.105 when I have set the bios to 1.091. (note with the V core at 1.091 I have not had any crashes or issues)

Would any one be able to explain why this happening and if it is normal.
I am hoping to get around 4.2Ghz out of the CPU and am pushing the Blck up 10Mhz each run untill I need to up the V core and then up again.

Im new to i7 OC so is this right so far.

Cheers for any help.

[Hondacity]

vid changes with load...idle...or temp...

[Samson]

First of all, a big Thank You to all of you who have replied to me on this issue trying to help.

Special thanks to donmarkoni who took the trouble of removing his card and place it in slot #2 to test.
And for also taking the PCI-E socket pictures.

I see that all of you are saying now that PCI-E slots #2 will only do 8x

on my board in slot #2 GPU-Z shows X8, in slot #1 GPU-Z shows X16. i know because i changed it over last night b/c of this problem. had to cut the tubes too, not flexible enough to bend and fit in slot #1.

However the situation is not so simple or so clear-cut.

With my two Sapphire Radeon HD5970 always in slots #2 and #4
I downloaded GPU-Z ver 0.4.4 (the latest version I could find), and as you can see in the image bellow
it says I am getting
PCI_E 2.0 x16 @ x16 2.0

Way at the bottom corner of the GPU-Z screen there is the drop down menu where I can select my GPU
ATI Radeon HD 5900 series

I can select all four of my 5970 GPU's and for all four I get the same result
PCI_E 2.0 x16 @ x16 2.0


So, I called up Sapphire U.S. Tech Support today (1-909-594-3128) and explained about my odd result of
Maximum Bus Setting PCI Express 2.0 x8 in the Catalyst 10.6 driver. Unfortunately the guy I spoke with
didn't know much about Video Cards it seems as he told me to try the newer Catalyst 10.7 driver
and if that didn't work to RMA both of my cards. I Thanked him very much and called up Asus U.S. Tech Support again.
1-(812)-282-2787

This time I spoke with "Bill" (a different person than yesterday). He told me the diagram on Page 2-28 is just
a general guideline. He also confirmed that all four PCI-E slots should do 16x for a SINGLE card provided
nothing else is plugged to the other PCI-E slots. (But this contradicts your results).

He also says that if GPU-Z is showing me 16x in slots #2 and #4 then it should be correct.

He also explained that the ATI driver is giving me that result because these are Dual-GPU cards.
Each of the GPU's on one card (so he says, remains to be confirmed) is like it is using 8 lanes for a
total of 8 + 8 = 16 PCI-E lanes.

It is true that in my Catalyst 10.6 driver the result
Maximum Bus Setting PCI Express 2.0 x8
is shown four times for each of the GPU's in my system (you must scroll down). I was aware of that.
I told him so myself.

This is why I am asking someone else with a Dual-GPU Radeon card in slot #1 to confirm.

From donmarkoni's pictures I have to agree that all the PCI-E sockets look fully wired.

To kpablo , the difference is that for you in your second screenshot of Everest for GPU it clearly shows
Buss Type PCI Express 2.0 x16 @ x8

Whereas for me, even in the most recent version 2232 (downloaded today) I get (Please see image bellow)
Buss Type PCI Express 2.0 x16 @ x16

but in the Chipset (North Bridge) section I still get
PCI-E 2.0 x4 port 5 In Use @ x8
PCI-E 2.0 x4 port 5 In Use @ x8

But since you people have tested slot #2 with a Single card and are getting only x8 then what is the truth here ????

Was today's Asus Tech Support guy "Bill" right about what he explained or mistaken ?!

[kill_a_wat]

ok thanks

but I'd like to know 24/7 OCing stability

I have 802 actually

thanks

I was using 0704 then have been using 0002 which I think is better for 24/7 o/c. Better stability I think.

[leighl]

vid changes with load...idle...or temp...

Thanks so its normal then to fluctuate.
What would the safest voltage be for 24/7 OC

[Hondacity]

i run mine stock

less heat less noise..100% stable

intel specs 1.35-1.4v max.....depends on temperature...if you're @ 20c and below you can do 1.35-1.4v safely.

[spoolindsm127]

Hi Guys I have joinned the club and got one of these Mb which I think are great.

This is my rig
i7 930 in the process of over clocking
MB R3E
6GB Gskill 1600 Mhz ram
GTX 480 SLI
Silver stone 1500W strider
Case HAFX.

Any way I have been overclocking this thing over the past month and last night I noticed that while stressing @3200MHZ Blck 153 with prime95 my core vid in CPU-Z was jumping from 1.085, 1.091 , 1.098 to 1.105 when I have set the bios to 1.091. (note with the V core at 1.091 I have not had any crashes or issues)

Would any one be able to explain why this happening and if it is normal.
I am hoping to get around 4.2Ghz out of the CPU and am pushing the Blck up 10Mhz each run untill I need to up the V core and then up again.

Im new to i7 OC so is this right so far.

Cheers for any help.

I actually believe you're talking about VDroop. It's a feature of i7 cpus that fluctuates the voltages rather than keeping them at a constant value. For me it's something that doesn't make sense, but still makes sense. Did that make any sense? Well anyways what you want to do if you don't want the voltage fluctuating so much is to turn on LLC (Load-Line Calibration). This is what Asus calls it, other mobos have different terms ie VDroop Control. You will find this setting in the Extreme Tweaker section. And for QPI voltage, the LLC has to be enabled via a jumper on the board, not thru the bios. Although once the jumper is set, the bios will show QPI LLC: Enabled (before will show : Disabled). The jumper is near the 24 pin ATX connector. People have found it easier to get stable clocks with LLC enabled, but some people also like to leave it off. See if it helps in your case

[Samson]

Hi Guys I have joinned the club...

... my core vid in CPU-Z was jumping from 1.085, 1.091 , 1.098 to 1.105 when I have set the bios to 1.091. (note with the V core at 1.091 I have not had any crashes or issues)...

Would any one be able to explain why this happening and if it is normal...

Welcome!

Your voltage fluctuation are of no significance and should not concern you. There is nothing
wrong with your system and it (at least this small variation) has nothing to do with any feature
of the Core i7.

People. We are talking about fluctuations here of 0.006 to 0.014 volts. When dealing with voltages
this low do you know how difficult it is to maintain some fixed value ? We are dealing with laws
of Physics here. Electricity.

A Pentium 4 will have Vcore fluctuations of 0.04 ~ 0.05 Volts and that is normal !

Can it be stabilized? Of course! To a degree though, and this would involve complex and more
expensive circuitry. When dealing with Life-dependant (or some types of Industrial) equipment
the expense is justified. But when dealing with a "Gaming" PC. Well then you can play with
LLC and raise it a bit (read previous posts for more info.)

I was right, #2 and #4 will always be @8x Max.

Look at this 2 screenshots i just took:

By the way, from your image here it looks to me that your results for
PCI-E 2.0 x4 port #5 In Use @ x8 <---- this is slot #2
PCI-E 2.0 x4 port #7 In Use @ x16 <---- this is slot #3

are for slots #2 and #3. Not slots #2 and #4.

My results in the image bellow are for slots #2 and #4 ( notice the port #'s )

Slots 2 and 4 are hardwired for x8. No way around it, bud

Just came home, moved my GTX 480 from slot #1 to slot #2 and what I found out is that bassplayer's statement "Slots 2 and 4 are hardwired for x8" is absolutely right.
... it would be a little stupid if MB could not re-route lanes to be 16x-8x-8x and run @ 8x-8x-16x, when it has 32 for those 4 slots...

It is in fact true that some boards are hard wired that way. The reason why the MoBo can't re-route lanes
"would" be because the lanes physically are not there. The board's design simply trace out only 8 PCI-E
lanes to that particular PCI-E slot. This is done to significantly reduce the cost of production.
In such a case, the 8 lanes that go to that slot are taken from the 16 available lanes of that pair, and
become unavailable to the slot that has all 16 lanes traced out to it. Thus, droping them both to x8/x8
when used at the same time.

I hoped that a board of this caliber and quality, however, would not be so cheaply made.

This is why I asked both Asus Tech Support guys if sticking one SINGLE card on any of the
four PCI-E slots would give me 16x. And both have answered Yes. But I know that sometimes (even often)
they can be mistaken and you guys here know more than they. Your results of testing would contradict them.

However, my own results make everything ambiguous to say the least because if in fact slots #2 and #4
have only 8 lanes traced out to them then I should never be getting results like this in GPU-Z

Bus Interface: PCI_E 2.0 x16 @ x16 2.0 (image shown in previous post)

Nevertheless, your findings make me wonder and worry since I am still getting this (image) bellow that
confuses the matter.

[Samson]

I am sorry for the double post but I have been asked a question and think it best to post for everyone.

Anyone that wants to correct me on anything they feel is mistaken, please do so.

I am now trying my Corsair 2000MHz C8 set. Again I don't want you to explain everything for me, I should do my own research. But why is 1600MHz or CAS6 preferred over say CAS8 @2000MHz? Is it more about the frequency or latency?

If the case is that the preffered latency is CAS6, then I will start at 1600MHz and see how far I can go on the frequency, correct?

RAM brings data to the processors to be processed, (or stores data for "fast" later use). The idea of faster
RAM is to be able to keep pace with the processing speed of your CPU. So that the CPU is always fed data and
never sees an idle time (when under load) simply because your RAM can not deliver data to it fast enough, and
thus, the RAM becomes a bottleneck.

But there is the opposite situation also. That being when the RAM is capable of delivering data faster
than the CPU can process it. In that case the speed of your RAM becomes redundant (not providing any additional
benefit).
If this RAM speed was achieved by overclocking, then you are overheating your RAM for nothing and risk
data corruption (errors) and possible damage to your RAM.

So now, the million dollar question is ! What is the optimal speed of the RAM ?

Core i7 by Intel specs is intended to support Triple-Channel DDR3 RAM of 1066MHz.
The designation for DDR-3 1066Mhz is PC3-8500.

In triple-channel the (theoretical bandwidth) Data-Transfer Rate for PC3-8500 is
3 x 8500 MB/sec = 25500 MB/sec

For triple-channel DDR3-1600MHz (PC3-10600) it would be
3 x 10600 MB/sec = 31800 MB/sec

That is about 31 GB per second. In other words, A LOT !!

So now, the million dollar question again. What is the optimal speed of the RAM ?
And how much is too much ?
Of course it depends on what CPU you have ? How many cores ? What speed is it running at ?
So, I think no one but an Intel engineer that designed the Core i7 can really answer that.

But "I would think" that at 31 GB/sec it is already more than any i7 can handle.
If someone knows better, please say so.

The benefit of having an Un-Locked multiplier on a CPU is that when the RAM speed is already
fast enough, you can speed up the CPU to use up the data delivered by your fast RAM.

So now, what is better? Higher MHz on your RAM or better timings ?
When you have decided on a RAM speed that you feel is sufficient (and no one can tell you what
that exactly is) then the thing to do is to lower your timings as much as possible.
The timings are what determine how fast your RAM can search for, find, and retrieve data to
deliver to your CPU. And often, lower timings result in much better data transfer rates than
simply overclocking your RAM at higher MHz. Lower timings are often more important than faster MHz.

Determining what is more advantageous is a game of Benchmarking trial and error.

But to better answer your question specifically, I would say that 1600MHz is already fast enough,
and I would prefer 1600MHz with timings in the 6's than 2000MHz with timings in the 9's.
(Though actual values can be calculated mathematically to determine which is better I don't know
the math off hand. Read my links for that).


Keep in mind that Benchmark results do not always reflect real world application improvements, since
Benchmarks are often optimized for the task while real world apps are not. ( i.e. A Benchmark that
tests your RAM will not suddenly request data from your HDD thus causing thrashing, but a real
world app may).

To better understand RAM timings please refer here

Everything You Need To Know About DDR, DDR2 and DDR3 Memories
================================================== ==============
http://www.hardwaresecrets.com/artic...3-Memories/167

Understanding RAM Timings
============================
http://www.hardwaresecrets.com/article/26

Memory Overclocking
===================
http://www.hardwaresecrets.com/artic...erclocking/152

And most importantly !
Have FUN !!! But try not to fry anything !

[tistou77]

I was using 0704 then have been using 0002 which I think is better for 24/7 o/c. Better stability I think.

ok thank you

The 0002 will be better than 0878 (although the 0878 is newer)

[kill_a_wat]

Actually I recently did a typical linx run to see how it went and it's at least as stable as 0704 which I am happy with. It feels more stable but obviously I don't expect it to perform any miracles


By kill_a_wat at 2010-08-17

[tistou77]

Actually I recently did a typical linx run to see how it went and it's at least as stable as 0704 which I am happy with. It feels more stable but obviously I don't expect it to perform any miracles


By kill_a_wat at 2010-08-17

ok thank you for the information

did you test the 0878? or just the 0002

[kill_a_wat]

I haven't tried the 0878 BIOS yet unfortunately so can't speak from experience only comparing 0704 to 0002 so far for 24/7 oc on bloomfield. But I am now using 0002 as my daily BIOS

[bassplayer]

Samson, there are two revisions of the board. I'm not sure if the PCI-E lanes differ between them. I thought mine only had 8 lanes in the second and fourth slots, but I could be thinking of my Gigabyte board. I'll give it a look when I get home from work.

[spoolindsm127]

Thanks once again Samson great explanation!

[spoolindsm127]

@ Bootsy and Splave

Can either of you please take a look at your board itself to see what revision is stamped on it? Just started looking into this new theory of the 2.xx revision boards reaching higher BCLK... I have a 1.04G lucky me Also how about everyone else?