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

[fmsam]

Many of us moved to the R3E from R2E, why ............. because it has Sata 3 6gb/s and USB 3, those are the main changes, well guess what, the marvell 9128 is almost useless if you are using SSds in raid, ask any one, as for usb 3 yes it works, its quick but for what purpose, oh my back up drive is really quick now........whoope friggin do !
You best run your vertex 2s on the intel for now.
I dont think the firmware we need will ever be released.

[zinsco]

Hi all

Any of you using the Sata 3 Marvell 9128 controller.
I'm wondering if you can help me out with what drivers give the best performance and are most stable on that Marvell. Ive heard there have been mixed results. I'll be using a OCZ Vertex 3 ssd and I'm still using BIOS version 1005 on the R3E.

Your vertex 3 will be seriously limited on the marvell because it only uses 1 lane. In fact, it may even do better on the intel sata 2 ports. If you want your vertex 3 to perform as it should you will need to get a raid card.

I would suggest going to the ocz form about this. There is tons of talk about this subject.

[WangChung]

The Marvell firmware update was released, sort of. Asus has yet to implement it into a BIOS release though. I'm running a modded BIOS from Mr. Toad for the Rampage III Gene, contact him and he may be able to mod the Extreme BIOS for you too. To be used at your own risk of course, but I've got nothing but smooth sailing going on.

[zinsco]

The Marvell firmware update was released, sort of. Asus has yet to implement it into a BIOS release though. I'm running a modded BIOS from Mr. Toad for the Rampage III Gene, contact him and he may be able to mod the Extreme BIOS for you too. To be used at your own risk of course, but I've got nothing but smooth sailing going on.

Ya, but are you using a vertex 3. I don't care what is done to the firmware. The marvel 9128 is limited to only 1x PCIe lanes and the vertex 3 needs more. Now, the marvel 9182 uses 2x PCIe lanes,
but even at that the numbers are still not as good as intel and our board doesn't have the 9182.

[donmarkoni]

Ya, but are you using a vertex 3. I don't care what is done to the firmware. The marvel 9128 is limited to only 1x PCIe lanes and the vertex 3 needs more. Now, the marvel 9182 uses 2x PCIe lanes,
but even at that the numbers are still not as good as intel and our board doesn't have the 9182.

PCI-E 2.0/2.1 has theoretical throughput of 640MB/s, which is 500MB/s practically. Now, Vertex 3 has maximum read speed up to 550MB/s.
I wouldn't call it seriously limited. Even if you use two of them, non RAID, you shouldn't experience any bottlenecks, as they are rarely used at the same time.

[xpower]

hi extreme friends...
higher bclk so far with this board???

[xpower]

hi extreme friends...
higher bclk so far with this board???

[Leeghoofd]

Bclock is hit and miss mate, sometimes it's CPU limited ( but rare), mostly it's chipset dependant, sometimes mobo design dependant ( eg MSI Xpower has issues to boot at high bclocks) ... if you have a lesser IOH chipset ( X58) then it could be you are limited to 220 or even lower... no setting (besides QPI Slow mode) will get you over that... Cold might do the trick...

I have done 233-235 (no slouch mode) on both my RE3s with a bloomfield Xeon CPU : so it's not the board design. Some have failed at 220 or lower, while their cpu did far better on other boards : Chipset limitation...

[xpower]

my R2E is stuck on 220,221. tried with slow mode but no go...........................

[donmarkoni]

How much voltage did give it (vIOH)?

[xpower]

@donmarkoni
tried 1.2,1.3 and auto as well.

[zinsco]

PCI-E 2.0/2.1 has theoretical throughput of 640MB/s, which is 500MB/s practically. Now, Vertex 3 has maximum read speed up to 550MB/s.
I wouldn't call it seriously limited. Even if you use two of them, non RAID, you shouldn't experience any bottlenecks, as they are rarely used at the same time.

That's true about the throughput but I have spent a lot of time reading in the ocz forums about the vertex 3's and no one has been able to get the full performance
from them with the marvell. The reason has been given that it only uses 1 lane. There must be something more to it than the fact that 1 lane has throughput of
640. I don't pretend to be an expert on any of this, or to know more than (especially you) any of you here, but I've read every page on the thread below and it is the consensus that the marvell
won't give full performance.

Here's on the intel controller

Here's on a LSI raid card

Here's on the Marvell controller 9128

[donmarkoni]

@donmarkoni
tried 1.2,1.3 and auto as well.

Sorry that I didn't mention it right away, but what is your PCI-E speed when you're trying to go over 221?
I had best results if keep PCI-E speed half of the BCLK.

That's true about the throughput but I have spent a lot of time reading in the ocz forums about the vertex 3's and no one has been able to get the full performance
from them with the marvell. The reason has been given that it only uses 1 lane. There must be something more to it than the fact that 1 lane has throughput of
640. I don't pretend to be an expert on any of this, or to know more than (especially you) any of you here, but I've read every page on the thread below and it is the consensus that the marvell
won't give full performance.
Here's on the intel controller
Here's on a LSI raid card
Here's on the Marvell controller 9128

Hmm... Strange... As you said, there must be something more. I wouldn't be satisfied with the answer: one PCI-E lane is to blame.
If there was any other way to test...
I think I found one: Buy ASRock SATA3 Card for 25-30$ (sorry for saying so directly: go and spend money ). It is PCI-E x1 SATA3 controller with Marvell 88SE9123 / 88SE9120 chipset. Similar chipset, still one lane used, but different aproach.
Maybe you get more conclusive results.

[zoson]

Sorry that I didn't mention it right away, but what is your PCI-E speed when you're trying to go over 221?
I had best results if keep PCI-E speed half of the BCLK.

I think this is board dependant. I can get to 221BCLK without changing PCI-E speed. I can get 232BCLK with 104 PCI-E.

Hmm... Strange... As you said, there must be something more. I wouldn't be satisfied with the answer: one PCI-E lane is to blame.
If there was any other way to test...
I think I found one: Buy ASRock SATA3 Card for 25-30$ (sorry for saying so directly: go and spend money ). It is PCI-E x1 SATA3 controller with Marvell 88SE9123 / 88SE9120 chipset. Similar chipset, still one lane used, but different aproach.
Maybe you get more conclusive results.

I agree with this also. The PCI-E lane itself is NOT the limiting factor. The controller itself just sucks. The way around it is use two drives in raid0. Neither drive performs at 'full' speed, but the overall reads and writes are still just as fast, or faster than any sata6 ssd.

My two 80gb x25-m g2's outperform a vertex3 in every possible category except sequential write speed. The marvell has no problem saturating my ssd's write speed at 140mb/sec, and gets ~450mb/sec-470mb/sec read speed.

Overall, I think the complaining about the marvell's performance is just nitpicking. Sure it doesn't give you the best possible performance, but it's still very good performance.

[xpower]

is booting with 223 bclk with diff volts and diff conbination of PCI-I of 107 110 115 but is freazes.
tried ioh 1.35v and vtt 1.39v.
still testing .....................



sorry for second post above^^^^^^^^^^^^

[xpower]

220 booting fine but when trying to stress with linx it freezes.
"uncorrectible hardware error"
trying to low the mem as when it auto its reaching 1800.
trying :12x220 easy
pci freq:115

any other setting to stablize 220bclk?

[Leeghoofd]

why on earth would you need 115PCI-E for just 220 Bclock... are you sure your rams can handle 1800mhz ? What are the temps of the IOH ? 1.35 IOH seems a bit out of whack too for just 220 and single GPU...

[rhpsystems]

Yesterday I've using my computer and after a suddenly reboot my R3E say this:

Please enter setup to recover BIOS settings
CMOS DATE/TIME Not set
Press F1 to run SETUP
Press F2 to load default values and continue

Load defaults, reboot, and get into Windows. After a while, reboots again, but I've no more response, even keeping pressed the power button onboard or Reset, NOTHING. The board stay turned on but no response of any kind. I need to remove the battery, shutdown power supply and then Windows loads againg, but after some minutes this issue shows again. This occurs with any BIOS, doesn't matter if I change between chips.

What's going on?

[Raja@ASUS]

Yesterday I've using my computer and after a suddenly reboot my R3E say this:

Please enter setup to recover BIOS settings
CMOS DATE/TIME Not set
Press F1 to run SETUP
Press F2 to load default values and continue

Load defaults, reboot, and get into Windows. After a while, reboots again, but I've no more response, even keeping pressed the power button onboard or Reset, NOTHING. The board stay turned on but no response of any kind. I need to remove the battery, shutdown power supply and then Windows loads againg, but after some minutes this issue shows again. This occurs with any BIOS, doesn't matter if I change between chips.

What's going on?

Try another CMOS battery and see what happens.

-Raja

[rhpsystems]

Try another CMOS battery and see what happens.

-Raja

It's done, but issue persist.

[xpower]

why on earth would you need 115PCI-E for just 220 Bclock... are you sure your rams can handle 1800mhz ? What are the temps of the IOH ? 1.35 IOH seems a bit out of whack too for just 220 and single GPU...

110 115 dosnt help. use ram to 1323mhz that i dont think is the problem.
any other settings to run that bclock.

[Brama]

It's done, but issue persist.

You can try to use second BIOS chip and check that cpu is well seated.

[Lee_Kay]

6Gb/s controller...

I feel sorry for Marvell they are being dragged through the mud on this for nothing. Highpoint do a 640 Rocket raid controller card based on the 9128 and it performs great. Well above that of the motherboard.

Sapphire Black board works as promised.


Asus even to this day post

"True SATA 6Gb/s Support
Experience the future of storage!
Supporting next-generation Serial ATA (SATA) storage interface, this motherboard delivers up to 6.0Gb/s data transfer rates. Additionally, get enhanced scalability, faster data retrieval, double the bandwidth of current bus systems." - Taken this morning from Asus webpage on the R3E.

The above statement can be taken as the following.

Get enhanced Scalability... - When you raid 0 you get less performance than with one drive.

Double the Bandwidth of current bus systems.... - I’m the ICH10R scales over 500MB/s when used in raid 0. The limit on the 6Gb/s controller if max was ever possible is 500MB/s. Not 1000MB/s.

Faster data retrieval... - Only if your using mechanical drives.

this motherboard delivers up to 6.0Gb/s data transfer rates.. - THIS IS IMPOSSIBLE. Its capped at 5Gb/s due to that being the max that one PCI 2.0 lane will ever achieve.

The sale of all motherboards by Asus, Gigabyte, Intel are sold under false information.



I bought my Asus Rampage Extreme 3 as an upgrade from an extreme 2. Because it had 6Gb/s controller and I had every intention of using it when the drives became available. I bought two C300 drives and ran into the issue where they where no longer detected on the Marvell chip after a windows install. So the drives where returned to Newegg for a refund after restocking fee. So fast forward to 3 weeks ago when I bought 2 Intel 510 120GB drives I bought them with the purpose of making use of Asus ROG super speed drive app (raids 2 drives to raid 0 on the same Marvell controller). Its now on its way back to Asus because I can not get the board to run above 330MB/s in raid 0 on the Marvell controller (something i believe is due to limitations on the board not the drives). On a standalone LSI 9240 4I I get 770MB/s in raid 0. The only drawback to this is I have to impede on the performance of my SLI setup to be able to use the card.

in summary
I bought the board purely for the 6Gb/s controller because it was advertised. It failed to do just that. I should have stayed with my rampage 2 Extreme and bought the card I ended up buying anyway and saved a lot of money.

Asus support has sucked through out all of this.
My motherboards had to go back to asus. They had no stock to process a Advanced RMA so I had to buy another motherboard in the mean time.

Some Mod on Asus website insists its ok to get 250MB/s sustained transfer on the Marvell controller because the rest is lost to system overhead.

Tech support informed me that they would not upgrade or downgrade to a board that meet the advertised statements because its not Asus policy. He then said to sell my board and go and buy a new one for my needs.

[zoson]

The above statement can be taken as the following.

Get enhanced Scalability... - When you raid 0 you get less performance than with one drive.

Double the Bandwidth of current bus systems.... - I’m the ICH10R scales over 500MB/s when used in raid 0. The limit on the 6Gb/s controller if max was ever possible is 500MB/s. Not 1000MB/s.

Faster data retrieval... - Only if your using mechanical drives.

this motherboard delivers up to 6.0Gb/s data transfer rates.. - THIS IS IMPOSSIBLE. Its capped at 5Gb/s due to that being the max that one PCI 2.0 lane will ever achieve.

This is wrong.
The theoretical maximum of a single PCI-E lane is 640MB/sec, which is above 5.0Gb/sec.
PCIe 2.0 delivers 5 GT/s, but employs an 8b/10b encoding scheme which results in a 20 percent ((10-8)/10) overhead on the raw bit rate, which brings transfer down to about 500MB/s.

The sale of all motherboards by Asus, Gigabyte, Intel are sold under false information.

I bought my Asus Rampage Extreme 3 as an upgrade from an extreme 2. Because it had 6Gb/s controller and I had every intention of using it when the drives became available. I bought two C300 drives and ran into the issue where they where no longer detected on the Marvell chip after a windows install. So the drives where returned to Newegg for a refund after restocking fee. So fast forward to 3 weeks ago when I bought 2 Intel 510 120GB drives I bought them with the purpose of making use of Asus ROG super speed drive app (raids 2 drives to raid 0 on the same Marvell controller). Its now on its way back to Asus because I can not get the board to run above 330MB/s in raid 0 on the Marvell controller (something i believe is due to limitations on the board not the drives). On a standalone LSI 9240 4I I get 770MB/s in raid 0. The only drawback to this is I have to impede on the performance of my SLI setup to be able to use the card.

My two x25-m G2's get up to 480MB/s reads on my marvell chip with superspeed. You're doing it wrong. Maybe drive alignment or stripe size.

in summary
I bought the board purely for the 6Gb/s controller because it was advertised. It failed to do just that. I should have stayed with my rampage 2 Extreme and bought the card I ended up buying anyway and saved a lot of money.

Asus support has sucked through out all of this.
My motherboards had to go back to asus. They had no stock to process a Advanced RMA so I had to buy another motherboard in the mean time.

Some Mod on Asus website insists its ok to get 250MB/s sustained transfer on the Marvell controller because the rest is lost to system overhead.

Tech support informed me that they would not upgrade or downgrade to a board that meet the advertised statements because its not Asus policy. He then said to sell my board and go and buy a new one for my needs.

Sounds to me like you didn't do your research before you purchased and got burned. You won't find any pity here.

[Lee_Kay]

This is wrong.
The theoretical maximum of a single PCI-E lane is 640MB/sec, which is above 6.0Gb/sec.
PCIe 2.0 delivers 6 GT/s, but employs an 8b/10b encoding scheme which results in a 20 percent ((10-8)/10) overhead on the raw bit rate, which brings transfer down to about 500MB/s.


My two x25-m G2's get up to 480MB/s reads on my marvell chip with superspeed. You're doing it wrong. Maybe drive alignment or stripe size.


Sounds to me like you didn't do your research before you purchased and got burned. You won't find any pity here.

I bought the board when it first came out back in 2010. And if you look at the post bios screen on the Marvell screen it clearly states Maximum Bandwidth on one PCI-E is 5Gb/s Also PCI-E 2.0 spec is 500Mb/s

"PCI-SIG announced the availability of the PCI Express Base 2.0 specification on 15 January 2007. The PCIe 2.0 standard doubles the per-lane throughput from the PCIe 1.0 standard's 250 MB/s to 500 MB/s. This means a 32-lane PCI connector (x32) can support throughput up to 16 Gb/s aggregate. The PCIe 2.0 standard uses a base clock speed of 5 GHz, while the first version operates at 2.5 GHz."

Im using windows 7 64 bit Ultra. so no alignment is needed.