AMD SB850/890GX with SSD RAID-0 first look

[GullLars]

Wich is just retarted. WHY on earth would AMD set NCQ as disabled for RAID by default? And then require you to download a tool to activate it. It should at least be accessible in BIOS/RAID-BIOS.

[felix_w]

Nah, i think it's disc specific.

The first two Vertex i bought had this setting each one differently. The first i connected was enabled, the second was disabled.

I don't know the logic behind this, but i guess it's disc specific set.

[F@32]

It is in the RaidXpert, if you highlight each drive and then pick settings, it has acces to write-cache and NCQ settings to enable/disable.

The drives come default NCQ off.

OK, with NCQ on, C1E off, SATA3 off, Cool'n'Quiet off - getting sequential read/write performance closer to ICH10R in ATTO. When I will have time, I will reset disks and re-bench properly.

[GullLars]

Do a run of AS SSD, newest version. 4KB random IOPS and 4KB-64thread (QD 64) will show you how well the NCQ is working.

[CrimInalA]

I was thinking of not putting my ssd's in raid on my new 890fx board , since it kills trim .
Do any of you think the drivers will be trim enabled in the near future ?
I guess it's no priority for AMD to support this .. ?

[F@32]

I think AHCI drivers support TRIM, but I can't test with the current SSD I have around. Wouldn't make sense that first native SATA3 controller doe not support TRIM. I wouldn't hold my breath for RAID TRIM. But at the same time, are you that concerned with write degradation over time? To me random and sequential reads are the most important things for gaming machine. And once in a while I perform Tony TRIM.

[F@32]

Any one tried to run HDD Erase on SB 850?

[CrimInalA]

I tried both hdderase from a bootdisk and from within the asus raid config tool .

From within asus raid config tool at post takes the longest . (20 min)
From bootdisk took 1 minute per disk .

I had to put both my ssd drives on sata 5 and 6 and ide mode or else they didnt get recognized by the tool .

[F@32]

I tried both hdderase from a bootdisk and from within the asus raid config tool .

From within asus raid config tool at post takes the longest . (20 min)
From bootdisk took 1 minute per disk .

I had to put both my ssd drives on sata 5 and 6 and ide mode or else they didnt get recognized by the tool .

Thanks, that confirms my preliminary conclusion that v3 of HDD Erase is not compatible with SB850 (I tried IDE, AHCI, RAID modes). Folks should attach their drives to JMicron controller in non-RAID mode to secure erase SSDs.

[CrimInalA]

i used v 4.0 , but what you say is true

[sequoia464]

Gigabyte 870A-UD3 ... had to be on plugs 4 and 5. I'm using HDDERASE 3.3 with the enhanced erase option, ports 5 and 6 of the 850 southbridge work for me.

Whats up with SATA III not recognizing my Intel SSD's?? Is AMD going to fix this??

[stinkas]

Are there gunna be any updates to the SB benchmarks soon?

[F@32]

Are there gunna be any updates to the SB benchmarks soon?

Will try doing it this w/e if I get free time.

EDIT: OK, updated first post. No matter NCQ, it still way behind Intel. IOMeter pretty much didn't change, so I kept original results. On other hand HDTune and ATTO looking better, but we know not to trust in those, right?

Someone on oczforum posted 3Rx0 with Vertex on SB850. Not impressed...

[Praz]

100GB Vertex LE and 120GB Vertex 2E in Raid 0 on the Crosshair IV Formula. NCQ enabled via RAIDXpert. IOMeter test file is 4k random write.

[F@32]

Argh, getting a lot of Task 30 time out errors in RAIDXpert since F7B BIOS...

[CrimInalA]

try altering SB voltage ? i have/had the same on my crosshair IV ..

[F@32]

try altering SB voltage ? i have/had the same on my crosshair IV ..

Figured it out. After flashing BIOS I flipped SATA3 mode back on when setting everything up. And it screws RAID up. Now with 1024K alignment and SATA3 off I think I unlocked the performance of the SB850 I expected all along. Posting fresh benchies shortly...

Quick preview:
4K random read QD16 before 1024K alignment and SATA3 mode turned ON


4K random read QD16 with 1024K alignment and SATA3 mode turned OFF

[CrimInalA]

This is mine on SB850 (CHIV)

[F@32]

This is mine on SB850 (CHIV)

Niiice. OK, first post has been updated with benchies using correct set-up. I think I'm finally happy with SB850

[Oliverda]

Nice results!

[TassadarDavid]

Hi all,

First of all, I have to say I'm new in this forum, this is my first post, although I've read you many times.

I come from some spanish forums like hardlimit, Noticias3D, Esoeslodemenos or hard2mano, where my nick is Tassadar. I suppose anyone of you will know who I am.

I'm here becouse I've spent a lot of time trying to configure correctly this mobo:

-Mainboard: Asus M4A88TD-V EVO/USB3.
-Processor: Phenom x6 1090T
-2x SSD Corsair Force F60 in raid 0
-8 Gb 1600 mhz Mushkin CL6

And I'm not capable (is this the appropiate word?) of reaching the results of the master: F@32

-I have activate NCQ in both disks in RaidXpert
-I've set off C1E and C&Q in bios
-I've configure SATA2 in bios instead of auto or SATA3
-I've disable auto-defrag in win7
-I've enable write back caching in RaidXpert
-I've align partition (I really have use this calculator, and it says my partition is ok)

Here is a shot RaidXpert, in which can be appreciated that NCQ y and write back caching are on (using driver RAID 3.1.1540.151):

RaidXpert view:



View of RAID driver used (the one that comes with AMD Catalyst 10.7):



Nevertheless, using revisions 3.2.1540.24 and 3.2.1540.35, ever when are newer, AMD RAID console doesn't appear on device administrator...



...as a consequence, it becomes impossible activating NCQ, since disks are not recognized under RaidXpert:



My best results are those:





As you can see, can't even the results of F@32:





In this point, I don't know what to do.....

The only stranger thing I see is in the shot of "AS SSD Benchmark", it is in the detail under "ahcix64s"; where I have "103424 K -OK" and in F@32 image the number is "1024 K - OK". Al the rest seems the same. Can this be my problem?

Any help????

thank you very much to all, and I hope you can sorry my english, whose is not as good as I would like

[TassadarDavid]

Hi again,

I've download a newer version of RaidXpert, the 3.2.1540.10, and it support the hard drives even when using driver RAID 3.2.1540.35.

As I told in my first post, my results are not good at all, and I would like you to help me

Today I've made tests with driver 3.1.1540.151 and 3.2.1540.35 and with stripe size 64 kb and 128.

Those are my results, in every picture there are 4 different scenarios, using different stripe sizes and raid drivers:

CrystalDiskMark:




ATTO BenchMark



AS SSD



I sincerely would appreciate and thank any help

[Oliverda]

v3.2.1540.75 WHQL certified RAID/AHCI driver is available.

donwload

[felix_w]

Hi again,

I've download a newer version of RaidXpert, the 3.2.1540.10, and it support the hard drives even when using driver RAID 3.2.1540.35.

As I told in my first post, my results are not good at all, and I would like you to help me

Today I've made tests with driver 3.1.1540.151 and 3.2.1540.35 and with stripe size 64 kb and 128.

Those are my results, in every picture there are 4 different scenarios, using different stripe sizes and raid drivers:


I sincerely would appreciate and thank any help

Did you wipe drives before each scenario ?

Also, can you update AS-SSD results with image showing all 4 scenarios ?

Thanx

[[XC] Oj101]

Not really completely on topic, but there seems to be a reason that NCQ is disabled by default. ALL benches go down on my array (2x 1TB Samsung F3 7200RPMs in RAID0 on SB750/790FX by a considerable amount, with HDTune's bench often hovering around 1.9MB/s. Disabling NCQ takes my minimum to around 60MB/s for a very brief time. It's not just the one bench, everything crawls. For example if I double click on a folder, there can be a pause of a few seconds where the hard drive activity light is on solid but I can hear the drives are idle. The drives themselves are fine, they were each tested for a solid seven days before being put in the array and all problems disappear with NCQ disabled.