X79 Storage - the Intel C600 series chipset SATA RAID Controller

[Anvil]

It's time for some initial benchmarks and info on the new Intel storage controller.

First, they have changed things a bit this time, there is a new series of both "hardware" and software and it is not compatible with the older platforms. (it can be done but it will not install on older platforms by default)

Both the option ROM and the new RSTe driver have changed from 10.n/11.n to 3.n.n

The option ROM for my MB is 3.0.0.1184 and the latest RSTe driver available is 3.0.0.2003
The driver is not available from Intel yet.

The first thing I noticed when I installed W7 was that it did not detect any storage devices, not even the DVD and so I had co copy both the chipset drivers and the RSTe to a USB stick before being able to install, not an issue but a bit annoying.
Slip-streaming or keeping the drivers on a USB drive will both work, Slip-streaming is probably the better option.

Once the OS was installed I had a first look at the new GUI for the "Intel Rapid Storage Technology enterprise" manager.



More on this later...

So, W7 was installed onto a Samsung 830 128GB and the initial benchmark was a huge disappointment.



It turns out that either there is a bug in the driver/GUI or one needs to disable "Turn off Windows write-cache buffer flushing..." to get small block (random) writes back at acceptable/expected levels.



and so the result for writes are now back on track.



I was still unhappy with the sequential read performance and was starting to think there was a bug somewhere as it was about 10% lower than expected.
The Samsungs easily do 500+MB/s on Sandy Bridge (Z68/P67) and both the 128GB and the 256GB Samsung ended up with 450-460MB/s.

I'm still not sure why the Samsungs under-performed sequentially on the new chipset but having tested a Crucial m4 I'm confident that it's something that can be fixed. (could be that it's just on my MB)

In other words, the Crucial m4 performed as expected and was able to bring the speed back up to 500MB/s on reads.

Without ticking write buffer flushing the m4 performed poorly on writes but much better vs the Samsung?




That about what I've found, so far.
More to follow during the weekend!

[Anvil]

I almost forgot, SMART does not work like it used to, not sure what has changed, maybe it's due to that drives look to be handled differently. (more SCSI like)

This is what CDM looks like.



Interestingly, the Intel SSD Toolbox is not capable of reading/displaying SMART.
(both Drive Details and SMART Details are disabled)


It is noted that the SMART feature is disabled but there is nowhere to enable it except for an option in the "bios" and that option is Enabled already.

[some_one]

FWIW I also found a reduction in performance on P67 with 3.x driver and Agility 3. The performance seemed very similar to what was had with 11.5 driver. 11.0 had better throughput and IIR correctly Anvil bench marks were something like ~7500 with 11.0 and ~5600 with 11.5 / 3.0.

[Computurd]

wow this is very interesting, I am awaiting the next stepping of chips to upgrade, so i have been wondering about any differences in performance. You would think with this being based off of a server chipset that it would be UBER....

[Anvil]

@Computurd
I think it's early days yet, I'm getting very good vibes from this combo, it feels like a 6 core 2600K, and then some.

As for TRIM, I had a sneaking suspicion that all the changes to the drivers etc might have brought changes to TRIM but it is clearly working, so another box ticked.



9.85GB copied to the drive results in this...



and then after deleting those same (2) files...



...all is back to normal.

Will do some raid testing tonight.

[Lord Micron]

FWIW I also found a reduction in performance on P67 with 3.x driver and Agility 3. The performance seemed very similar to what was had with 11.5 driver. 11.0 had better throughput and IIR correctly Anvil bench marks were something like ~7500 with 11.0 and ~5600 with 11.5 / 3.0.

Unless you hacked the Inf files, I'm not sure how you installed it on a P67, being that the RSTe 3.x drivers are only intended for the C600 chipset that only the the X79 boards currently have.

[some_one]

No need to hack anything, just goto the device manager and install the driver manually. Windows will probably tell you it's not for that HW but will let you install it anyways. For me there is a performance hit so not much point using it at this time but good to know it can be done should TRIM in RAID0 not eventuate for P67.

I did try to run some Anvil BM's with different drivers but something went astray on W8.

[Anvil]

...
I did try to run some Anvil BM's with different drivers but something went astray on W8.

I'll have a look at it, quite a few things have changed on W8 so I will most likely have to make some changes.
I've got W8 in a VM but no time to play, it is a bit early as well, most likely things will change in February.
You could try running it "as Administrator" just to see what happens. (meaning it could be related to user-rights)

From what I'm seeing I'd say stay on the 10-11 series orom's/drivers on the Z68/x67, the 3.n does not allow for "Smart Response Technology", short for SRT (using an SSD for cache) and from the looks of it it is taking a new direction.
There might not be full support for the 3.n series chipset/PCH on the "plain" Sandy Bridge platform.
(time will tell, TRIM on arrays in particular)

[Anvil]

A bit more on the new GUI

How to create a raid volume (raid 1/5/10/0)

Step 1

Selecting controller ! (wonder what other controllers they have in mind)

Selecting volume type



Except for the GUI changes there are no new features vs the older driver.



The Andvanced Tab is no different from 10/11


You need to tick "Proceed with deleting data" in order to create the volume.



Once the volume is created we can have a closer look at what we can do/change.

As I did not utilize the full size of the array there is still space left for another small volume or one can Increase the size, just like on the 10/11 series.
Disk data cache can be Enabled or Disabled. (have not checked this one out yet)

I have tried to find out if this has any effect on benchmarks and so far there is no change, on top of this it is not possible to Disable this option, you can make the change but once you reboot it is back to Enabled.

"quote from the help file"
Enabling Disk Data Cache
Enabling the disk data cache for all disks on the array allows you to enable cache memory physically present on the disks and use it to speed up data access. This action is only available from the Array Properties pane because the data cache must be in the same state across all disks that are part of a single array.In the Array Properties pane, the disk data cache is reported as enabled or disabled for all disks in the array. In the Disk Properties pane, the disk data cache is reported as enabled or disabled for a specific disk that is part of that array. The option to change this setting is only available from the Array Properties pane.




If we click on the volume (inside the array) we get to the Volume properties. (as shown below)
The properties looks to be exactly the same as on 10/11.



All in all there is not much new on the configuration side except for the GUI and I personally prefer the new one.

[Anvil]

To test the 6Gb/s raid ports I re-installed onto a pair of Crucial m4 64GB

The results did not disappoint although I had hoped for a bandwidth of > 1000MB/s, it was pretty close.







More to come later as I have already tested Vertex 3 240GB 2R0...

[Computurd]

damn impressive for 2 devices. 3 years ago who would have thought it would be that impressive?

[DooRules]

Two of my 128 gb M4's, ud7 x79, enterprise drivers as well, writes are just all over the place. I tried Anvil's suggestion above but it doesn't seem to give any consistency whatsoever, in terms of writes. Reads are where they should be. Tries the 11.5 alphas IRST, installed no issues with the same new control panel, but still no consistency in writes. Back on the enterprise drivers again.

[Anvil]

@DooRules

So you ticked the "Turn Off Windows Write Cache Buffer Flushing" and Enabled Write Back in RSTe?
(all things point to you have done this correctly, otherwise random writes would have been on par with HDD's)

I'm not sure if this is a bug and so results using other MB's would have been interesting!



Your random results looks pretty decent but the 16QD (Threads) should normally have been higher.

[DooRules]

Yes, I did both. Same thing with as ssd, writes are just nerfed quite a bit. Thing is I can run either your bench or as ssd and get a nasty low score, run it right away again and writes go up quite a bit, but not where they should be. I think better drivers will alleviate most if not all of this. Not to worried, still lost in all those bios settings for the chip.

[Anvil]

My MB's been pretty consistent during benchmarks but I expect we'll see several updated drivers and OROMs. (as usual)

--

Here's 2R0 Vertex 3.

[DooRules]

A little better after a new install. Not to worried about it at all, will only get better with newer drivers I should think.

[DooRules]

New RSTe version up on station drivers...
http://www.station-drivers.com/forum...hp?f=32&t=3540

[Anvil]

Thanks for the link.

I downloaded it last night, will re-check some of the benchmarks to see if there are changes.

[DooRules]

Up on Intel now Anvil...

http://downloadcenter.intel.com/Deta...?DwnldID=20507

[DooRules]

Writes are starting to come back some.

[Anvil]

There is a new SSD Toolbox as well but it won't download Link
I was hoping it might fix the issue with reporting SMART.

[Live or Die]

Heres my 2 Corsair Force 3 drives in Raid0, In some benchmark i see really slow speed then in others like ATTO i see the speeds below i have no clue.

[Anvil]

The Force 3 reacts to the level of compressibility used, ATTO is best case and AS SSD or my benchmark using the 100% (Incompressible) will give you worst case.
(all SF based drives perform according to the compressibility of the data used)

[Live or Die]

So these are normal speeds? or is the C600 controller lacking in production.

[DooRules]

Started playing around a little today. I downloaded the new 11 series irst from station drivers.
I went to device manager to update driver, the updater would not look at the irst installer but would accept the x64 file.
This was the achi and iastor new files. Had to reboot twice. Control panel seems to be not working for now.
Still other than control panel all seems well. I do not have the windows cache buffer box ticked.
Till now on my x79 the writes have been some what low, here are my new results...






These are by far the best scores yet on my x79.