X79 Storage - the Intel C600 series chipset SATA RAID Controller

[Anvil]

Hi guys!

I'm working on it, it has changed from earlier Intel series drivers and it is related to how write throughs are handeled.
(just like write through vs write back on caching raid controllers)

The thing is, is this behaviour the correct one (I think so) which means that previous drivers with or without the option checked would cause loss of data if power was cut/lost.
(the old ones performed more or less the same and they shouldn't)

The new driver really flushes data beyond any earlier version, still, the wording of the two options doesn't realy make sense.

[Anvil]

RSTe 3.1.0.1085 has been made official by Intel

4/26/2012
"This download provides support for the Intel® C600 chipset on both AHCI Capable SATA Controller and SATA/SAS Capable Controller in RSTe mode"

Link to Intel Enter RSTe download and click search downloads

[tistou77]

Hi guys!

I'm working on it, it has changed from earlier Intel series drivers and it is related to how write throughs are handeled.
(just like write through vs write back on caching raid controllers)

The thing is, is this behaviour the correct one (I think so) which means that previous drivers with or without the option checked would cause loss of data if power was cut/lost.
(the old ones performed more or less the same and they shouldn't)

The new driver really flushes data beyond any earlier version, still, the wording of the two options doesn't realy make sense.

I didn't understand, there are better versions of drivers?

Thanks

RSTe 3.1.0.1085 has been made official by Intel

4/26/2012
"This download provides support for the Intel® C600 chipset on both AHCI Capable SATA Controller and SATA/SAS Capable Controller in RSTe mode"

Link to Intel Enter RSTe download and click search downloads

I took these drivers from April 26th

[Computurd]

I agree anvil, the wording is one of the most nebulous things Ive ever read. Seems to be a chipset issue, when i try to write incompressible data with Iometer it does the same thing if the drive isnt configured correctly.

[Anvil]

Anvil, with Crystal Disk Info, you can tell me the info you have with the Samsung 830 (for Standart) with RSte
just curious

Attachment 126694

Thanks

Here is a screenshot of CDI with the Samsung connected to an LSI 9211, will check later today connected to one of the Intel ports.

edit: It was connected to an Intel port, I just picked the wrong drive to capture. (was trying to capture a 256GB drive)

samsung256_cdi.PNG

[Anvil]

I didn't understand, there are better versions of drivers?
Thanks

No, there are only the ones you have seen.
I'm talking about how it behaves vs the 10.x or 11.x series drivers on older chipsets (X58, P67,Z68,...)

[tistou77]

Here is a screenshot of CDI with the Samsung connected to an LSI 9211, will check later today connected to one of the Intel ports.

samsung256_cdi.PNG

this is like me for "Standard" (to see in "single")
Weird, some with Samsung 830, have ATA8, but nothing important I think

No, there are only the ones you have seen.
I'm talking about how it behaves vs the 10.x or 11.x series drivers on older chipsets (X58, P67,Z68,...)

Ah ok, the 3.0 is the best for the X79 (compared to 11.x)?

Thanks for your feedback

[Anvil]

The drive was connected to an Intel port, I had a 256GB Samsung connected to the LSI but apparently I picked the wrong drive in CDI.
So, it was an 128GB Samsung 128GB in raid-0 on the X79.

In my opinion the 3 series driver is pretty good, it lacks a little bit on sequential reads, not much though.

The main issue is how it performs in some benchmarks (including mine), it will take a few more days for me to check out all the options as I need to start from scratch on a fresh install.

[tistou77]

ok and thank you again for your work

[flamenko]

Just got back to reinstalling my X79 Fatal1ty and, once again, normal RST versions don't work at all. Tried testing main drive in just MSAHCI mode and my low 4k's seem a bit low... Going to try with the secondary as this was the stopper ALWAYS in the past. Well I'll be damned...half decent scores out of both the primary and secondary drive for the first time ever and on Anvil as well.... 4k's are a bit low but I won't tell you that I have easily EASILY two weeks of trying to get this problem fixed.

Question is now whether installing RSTe will improve my speeds....

EDIT: Just did a side by side with a M4 Crystal test between the new Z77 Intel and the X79 ASRock. The ASRock is showing close to 27MB/s lower write speeds including the low 4k's...but testing is still good...

[Anvil]

Last night I updated my Windows installer USB stick. (using DISM for adding updates and features)

I noticed the process was dead slow and found that I had forgotten to Check the "Write-Cache Buffer Flushing" on the drive it sits on.

So, today I ran a test using the same drive (Samsung F3 1TB) and copied (from another drive) a folder consisting of 627MBs of data, 8471 files in 622 folders.
(this to compare performance with and without buffer flushing)

Buffer Flushing Unchecked (Enabled) : 40 seconds

Buffer Flushing Checked (Disabled) : 6 seconds

The X79 without buffer flushing Disabled is dead slow.

This is on my system though, I'd like someone to repeat the test with similar test sizes and most importantly a lot of files.

You should restart the computer to make sure that the setting is in effect and that there is no caching on the files you copy!

[RyderOCZ]

Last night I updated my Windows installer USB stick. (using DISM for adding updates and features)

I noticed the process was dead slow and found that I had forgotten to Check the "Write-Cache Buffer Flushing" on the drive it sits on.

So, today I ran a test using the same drive (Samsung F3 1TB) and copied (from another drive) a folder consisting of 627MBs of data, 8471 files in 622 folders.
(this to compare performance with and without buffer flushing)

Buffer Flushing Unchecked (Enabled) : 40 seconds

Buffer Flushing Checked (Disabled) : 6 seconds

The X79 without buffer flushing Disabled is dead slow.

This is on my system though, I'd like someone to repeat the test with similar test sizes and most importantly a lot of files.

You should restart the computer to make sure that the setting is in effect and that there is no caching on the files you copy!

Repeatable here and for the other guys in our beta testing. X79, you want buffer flushing checked.

[Anvil]

Thanks for checking!

I was starting to believe that some people weren't seeing the issue/slowdown on X79 and needed someone to verify that it wasn't just me having the issue.

Anyways, I've added an option in ASU to relax on forcing writes. (it wasn't an issue on earlier chipsets)

It will now by Default behave like the other benchmarks like i.e. IOmeter, Crystal DiskMark, AS SSD,...
The option can be changed in settings in case someone wants to go the extra mile on writing unbuffered data.

edit:
I expect this was using several different motherboards and not just Asus in particular.
(I'm sure it was!, would be great knowing though)

[bonidinimon]

Hi all

I have the same issue (very slow 4k writings) with the Intel X79 Chipset and the Intel RSTe Driver. Is there a technical explanation for the problems?

Also, I believe that TRIM does not work, because the RSTe treats the drives as SCSI. What do you say?

Sorry for my english

[Anvil]

If you are using the correct ports and drivers TRIM does work just fine.

The screenshot shows before and after deleting the test-file on the drive.

C300_17-mai-2012_20-28_pre_deleting_file.png C300_17-mai-2012_20-30_after_deleting_file.png

Could you post a benchmark showing your issue with 4K and did you check the "Write-Cache buffer flushing" option.

[bonidinimon]

Hi Anvil

I am glad that you've tested TRIM and that its working

So then, here the screenshots from my X79 platform concerning the low 4k write performance:

Write-Cache buffer flushing ON


Write-Cache buffer flushing OFF


The problem affects not only the SSD, but also my WD Velociraptor and Green!

[tistou77]

The problem with RSTe, comes from the rom bios? disks?
Or directly from drivers Intel RSte?

And it is Intel to solve the problem

[Nizzen]

The problem with RSTe, comes from the rom bios? disks?
Or directly from drivers Intel RSte?

And it is Intel to solve the problem

New today:
Rapid Storage Technology Enterprise (RSTe)
rom
3.1.0.2101
www.Station-drivers.com

Haven't tested it yet.

[Anvil]

There is nothing wrong with RSTe.
Intel has changed how Buffer Flushing and Write Through are handled on the Series 3 driver.

All you have to do is to Check the "Write-Cache Buffer Flushing" option on each drive in the Device Manager.
It affects all drives connected to the chipset whether they are SSDs or HDDs.

I'm sure Intel has some reason for doing this and in my opinion they have done the right thing.
This way you now have an option to have buffer flushing working, it didn't work on earlier versions of the RST that's why we are seeing this difference in performance.

[Anvil]

Nizzen

I haven't found anyone that have updated the option ROM on X79s. (haven't really been looking)

I'll start looking for something for ASUS Rampage IV Extreme, I'll let you know...

edit:
There have been made requests for Option ROM updates on the ROG site, including the new one (3.1.0.2101). Link to ROG

[Nizzen]

Can it be in this thread?
http://www.overclock.net/t/1193055/a...g-updated-5-17

[Anvil]

Looks like there are a few mods for several MBs

Already found an update for an older BIOS (1201) and I'm on 1305, I'll continue reading...

[tistou77]

There is nothing wrong with RSTe.
Intel has changed how Buffer Flushing and Write Through are handled on the Series 3 driver.

All you have to do is to Check the "Write-Cache Buffer Flushing" option on each drive in the Device Manager.
It affects all drives connected to the chipset whether they are SSDs or HDDs.

I'm sure Intel has some reason for doing this and in my opinion they have done the right thing.
This way you now have an option to have buffer flushing working, it didn't work on earlier versions of the RST that's why we are seeing this difference in performance.

Agree, but there is more protection on writing?
And why the boot is longer with RSTe compared to RST?

[Anvil]

Yes, it should lead to better data-integrity in case of i.e power loss.

I don't know about the boot time, my setup takes more than a minute to load so the RSTe isn't really an issue, I power on and go get myself some coffee.
(booting includes a few raid controllers)
When I get the time I'll try without the raid controllers.

What is your total boot time (from power on) and how many seconds from Windows starts loading.

I'd say, just leave it powered on

[tistou77]

between RST and RSTe, I have a few more seconds with RSTe (when the windows logo to finished being animated)

Nothing embarrassing, but when you're used to a faster boot, you do not want to be longer