X79 Storage - the Intel C600 series chipset SATA RAID Controller

[tistou77]

the testing I've done with a samsung 830

RSTe (3.1.0.1085) and "Turn off Windows write-cache buffer flushing ..." unchecked
as the screen, in the first pages

with RSTe and "Turn off Windows write-cache buffer flushing ..."checked



with RST (11.1.0.1006 modded)




boot and shut down is longer with RSTe

[Anvil]

The RSTe 3 series / Samsung benchmark looks very close to mine. (I installed onto a 830 128GB last night)

Not much difference except for sequential read, most drives end up at 450-480MB/s on the 3 series drivers, SF drives are an exception though, they easily do > 500MB/s*.

I might end up reinstalling again to check out Samsung 830 in raid. (in order to compare with my previous results)

[tistou77]

Anvil, you've done tests with RSTe and Samsung 830?

thanks

[Anvil]

Yes I did, both as a single drive and as raid-0.
(as XS was down I didn't get to post it and a few days later...)

From what I recall, "speed-wise" (boot-up) it was not much different from the Plextor M2P, in general there were no issues.

I'll post the benchmarks later tonight.

[tistou77]

Thank you very much for your feedback

From what I recall, "speed-wise" (boot-up) it was not much different from the Plextor M2P, in general there were no issues.

With the Samsung 830, the boot is "similar" with the Plextor and RSTe?

[Anvil]

RSTe 3.1.0.1068, and I've kept it. (for now)

ATA SAMSUNG SSD 830 SCSI Disk Device_128GB_1GB-20120503-2347.png

Intel Raid 0 Volume SCSI Disk Device_243GB_1GB-20120505-1305_2R0_Samsung830_128GB.png

The boot time is about the same on the Samsung and the Plextor.

[tistou77]

thanks

and you have compared with the RST 11.1.0.1006?

[Anvil]

I'll check RST 11.x when I get the time.
(there's just a lot going on, so don't hold your breath )

The 3 series driver is rather OK, it's just different.

[tistou77]

Anvil, with the Samsung should always check "Turn off Windows write-cache buffer flushing ..."?
I have not retested with RSTe

And I sent an email to Samsung to see if they knew why the SSD have problems with the X79 chipset

[Anvil]

All drives (SSDs, HDDs, iow: the lot) require the "Turn off Windows write cache buffer flushing" to be checked, so it's got nothing to do with the Samsung.
(or were you thinking of some other issue?)

Here's an example using the Intel 520 180GB without the buffer flushing disabled. (unchecked)

ATA INTEL SSDSC2CW18 SCSI Disk Device_180GB_2GB-20120210-2125.png

Now with buffer flushing disabled. (checked)

ATA INTEL SSDSC2CW18 SCSI Disk Device_180GB_2GB-20120210-2146.png

[johnw]

Anvil:

Are you confusing "Enable write caching on the device" for "Turn off Windows write-cache buffer flushing on the device"?

I always run my SSDs with "Enable write caching" checked and "Turn off...buffer flushing" unchecked, and I never saw anything like the terrible write performance you show in your first ASU screenshot.

[Anvil]

I'm afraid not!

Both needs to be checked on the X79, the terrible write result is due to the buffer flushing being active. (unchecked)
(a bit confusing naming unfortunately)

[johnw]

I'm afraid not!

Both needs to be checked on the X79, the terrible write result is due to the buffer flushing being active. (unchecked)
(a bit confusing naming unfortunately)

Ah, I knew this thread was about the X79, but I did not realize that the X79 was different than other chipsets for write cache buffer flushing.

That is unfortunate, since disabling write-cache buffer flushing is not the safest thing you can do. But it looks like you almost have to do it with the X79 unless you want terrible performance.

EDIT:

I see you even mentioned that in the first post. Sorry for wasting space rehashing the issue!

[Anvil]

It looks to be buried deep in the C600 chipset.

It could have been a bug but this is the 3rd driver I'm using so unless there is something up with the Option ROM this is how it's meant to be, using the 3 series driver that is.
From what I have seen, the 11.x driver behaves like we're used to so many have opted for that driver on the X79.

[tistou77]

All drives (SSDs, HDDs, iow: the lot) require the "Turn off Windows write cache buffer flushing" to be checked, so it's got nothing to do with the Samsung.
(or were you thinking of some other issue?)

Here's an example using the Intel 520 180GB without the buffer flushing disabled. (unchecked)

ATA INTEL SSDSC2CW18 SCSI Disk Device_180GB_2GB-20120210-2125.png

Now with buffer flushing disabled. (checked)

ATA INTEL SSDSC2CW18 SCSI Disk Device_180GB_2GB-20120210-2146.png

Ok, you must activate the C-STATE in the bios (which?) and check the box?

Is to Intel solve this problem then?

Samsung told me

We ask if you could please test this unit within the Magician tool as numbers provided by 3rd party software we can not comment on as they use different metrics to get their numbers

Anvil, why with your software we see this problem (buffer flushing disabled unchecked) but with Crystal Disk Mark, AS SSD there is no difference?

you could test?

[Anvil]

Who are we?

I'm testing using AS SSD and I'm seeing the exact same behaviour on a random drive I picked, the C300.
(it takes forever, I'll see if I can avoid Aborting )

Preliminary screen-shot.

AS_SSD_C300_unchecked.PNG

[johnw]

Who are we?

Why are we here?

[Anvil]

johnw,

I didn't quote tistou

Anvil, why with your software we see this problem (buffer flushing disabled unchecked) but with Crystal Disk Mark, AS SSD there is no difference?
you could test?

I'm finding this just as annoying as anyone else, on my setup it can be reproduced.
Will check CrystalDiskMark next.

[tistou77]

indeed, it's for "me" and not "we"

with the X25-E, I had no difference, I would have tested with the Samsung

and for

Ok, you must activate the C-STATE in the bios (which?) and check the box?

Is to Intel solve this problem then?

Samsung told me

We ask if you could please test this unit within the Magician tool as numbers provided by 3rd party software we can not comment on as they use different metrics to get their numbers

what do you think?

[Anvil]

I'm getting some weird results using the latest driver (3.1.0.1085(1068))

I haven't really tested that issue on the latest driver and used AS SSD to show that it was reproducible and it was, at least on QD1. (I had to abort as it took forever)
(the Intel 520 180GB tests I posted was tested on an earlier driver, will have to check which one)

My benchmark show the same results using the Vertex 4 whether the option is checked or not, which means I'll have to retest other drives, will have to continue tomorrow.

Looks like a complete reinstall is due again just to verify what's going on!
Hopefully it's fixed with the latest driver.

[tistou77]

thank you very much for your feedback and waiting for your conclusions

And for C-STATE, it's better to enabled in the bios?
Which one?

thanks

[tistou77]

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

Capture.PNG

Thanks


And I retested the RSTe 3.0, is slightly less performance than with RST 11.1

RST 11.1


Samsung 830 RST 11.1.png


RSTe 3.0


Samsung 830 RSTe 3.0.png


I activated the C-STATE, I do not know if that changes anything
And I don't know if I stay with the RST or RSTe

[Anvil]

I'll check the Samsung with CDI

If you want maximum performance/responsiveness, deactivate power savings. (disable c states)

I tend to keep C States Enabled and just run some background task or enable it if needed.
Having C states enabled helps a lot on the temps.

[Computurd]

this is very very similar to an issue that i had with my board and incompressible data. i thought that a modded bios fixed the issue, but that might not be the case, i changed so many variables.
les over at TSSR also had the exact same issue. here is the thread:
http://thessdreview.com/Forums/ssd-d....htm#post21704

and it doesnt matter if its ASU or anything else, incompressible goes in the toilet.

[flamenko]

Anvil,

You have stumbled upon something that I have been cutting my head off with for at least 3 weeks now. In fact, I am just reinstalling the ASRock Fatal1ty X79 now. The problem I had was seen first in your software when testing incompressible data and then confirmed with AS SSD. Even when it worked with AS SSD, one could tell that the first sample was no good and then the performance would click in on the next.

As well, at times, I could get the performance out of the main drive BUT NEVER the secondary....not good when testing SSDs. Glad you are involved as you have the upper hand in diagnosing this issue

Oh also.....when upgrading to RSTe drivers, one cannot go back even with a reinstall and attempted installation of lower drivr packages will stop as they are incompatible with the system.

Thanks for the link PAUL!