RAID 0 SSD Array Slow...

[ParseMeHard]

Hey guys, just ran some benchmarks on my RAID 0 Array, and got some pretty disheartening results. I will list the computers specs first, then the benchmark results.

Setup:

Gigabyte GA-890FXA-UD5
AMD Phenom II X6 1090T @ 4GHz
4x 2GB GSKill DDR3 @ 1625MHz
2x OCZ Vertex 3 120GB SSDs in RAID0 w/ 128k
AHCI/RAID Mode in BIOS

First test is with Anvis SSD Benchmark set to 0% Compression

Second test is with the same program, at a 46% Compression

Third test is with ATTO, set to Queue Depth 10

And the final test is with CrystalDiskMark, set to 0-Fill

Why are these scores so low? This is about what one drive should score.

 

[GarfieldtheCat]

Are you on SATA II or III ports?

Which drivers are you using?

 

[VirtualLarry]

When I was running two 30GB Agility drives in RAID-0 on my ICH9R P35 board, they degraded without TRIM to be running slower than a single drive, after a week.

How long have you been running the RAID-0?

 

[ParseMeHard]

Are you on SATA II or III ports?

Which drivers are you using?

I am using the native SATAIII Ports.

I am using the newest drivers from Gigabyte.

 

[ParseMeHard]

When I was running two 30GB Agility drives in RAID-0 on my ICH9R P35 board, they degraded without TRIM to be running slower than a single drive, after a week.

How long have you been running the RAID-0?

I would say about a month now.

 

[VirtualLarry]

My understanding, from reading a certain review comparing the m4 and the SF 2nd-gen controller, is that after leaving the drive idle but powered for an hour, the speed on the m4 drives returned to normal, but the SF drives needed TRIM to recover.

This could be why your array has degraded in speed, due to lack of TRIM.

 

[el-Capitan]

My understanding, from reading a certain review comparing the m4 and the SF 2nd-gen controller, is that after leaving the drive idle but powered for an hour, the speed on the m4 drives returned to normal, but the SF drives needed TRIM to recover.

This could be why your array has degraded in speed, due to lack of TRIM.

I have the same drives in RAID 0 and have twice the performance, 7 months after install. Haven't dealt with AMD systems in a while can't help I'm afraid...

 

[Zxian]

Trim should only affect the write performance on the drives. Read performance should be the same regardless of the drive having been trimmed or not. I've used a Vertex3 drive in my previous desktop for over a year without any indication of the behaviour that VirtualLarry describes.

Unless you're continuously writing data to the drives, TRIM support matters very little these days.

I also haven't dealt too much with AMD systems to help with specific debugging information. My honest guess is that for some reason the SATA interfaces are only running at 3Gbps instead of the 6Gbps that they should be if you are sure you've got them plugged into the correct ports (should be the blue ports labeled SATA3_0 and SATA3_1 - not the white ones labeled GSATA2_6 and GSATA2_7).

Have you checked that your BIOS is up to date and that the drives are running the latest firmware from OCZ?
Are you using SATA 6Gbps cables?
Have you tried different cables?
Have you tried plugging the drives into the SATA3_2 and SATA3_3 ports?

 

[sequoia464]

AMD recently released a new raid rom that allows read ahead and write back to be enabled. You might want to make sure you are on the latest bios from Gigabyte, and the latest AMD raid and AHCI drivers - get those directly from AMD.

I don't know if the new raid rom will be incorporated in your boards bios or not but it is worth a shot. Recently Asus updated some of their boards to include the new raid rom - it probably varies by manufacturer and board model so it might not have been included yet for your board.

Install AMD's RaidExpert to enable the read ahead and write back, also make sure that NCQ is enabled through RaidExpert.

Don't know if this is the answer for you, but it can't hurt.

 

[ParseMeHard]

Thanks for the replies!

Zxian:

I'm not really writing much to the drives. It's pretty much a light gaming machine, but mostlly for basic tasks. I use it a lot for work. I am a Network Administrator, so I have a TON of VNC/RDP sessions open at once. So I don't write much to the drive at all.

The SATA interface, according to RAIDXpert, shows the drives running at 6.0GB/s, and NCQ being enabled. So I beieve the drives are operating at 6.0GB/s speeds. The are plugged into the SATA3_0/1 ports.

The BIOS is up to date.

I have not tried the SATA3_2 and SATA3_3 ports. Do you think I could move the drives over without having to break the Array?

sequoia464:

I am on the newest BIOS from Gigabyte, and the latest AMD Driver from Gigabyte. I didn't think to check directly from AMD. I am currently getting the newest version form AMD for both the driver and the RAIDXpert. The version that I'm getting from AMD isn't making much seance for matching version of the SB. But I'll give it a shot, and update.

I have not tried more cables. These are the cables that came with the board. I was just assuming the are SATAIII 6.0GB/s rated. I may need to pick up a few, just in case.

 

[ParseMeHard]

After updating the drivers, the speeds didn't really change any. They actually went a bit down, but I'll blame that on not letting the controller run Garbage Collection after all these benchmarks. A few questions.

First:

After installing the new RAIDXpert driver, these are some new options. What should these be?:

Second:

I installed all three of the following drivers.

I can see that the RAIDXpert driver updated, but I cant seem to find this 12.8 revision date anywhere in my system. When my RAID Controller Scans for drives on POST, it says "RAID Option ROM Version 3.2.1020.39". Where can I verify that is the newest version?

Third:

When I go into Device Manger, and check the Proprieties of the RAID, this is the version I'm showing:

Is this right?

 

[Old Hippie]

The problem is you're running an AMD board with OCZ drives.

Sorry.

 

[Zxian]

The problem is you're running an AMD board with OCZ drives.

Sorry.

Take your trolling and product fanboisms elsewhere please. AMD boards are fully capable, and OCZ Vertex3 drives are perfectly stable with modern firmware releases. There have been compatibility issues with certain SSDs and nVidia chipsets, but this is the first of any potential incompatibilities with an AMD system.

@ParseMeHard - Turn on Read Ahead cache and Write back cache.

As for the driver revision, that would be found under the IDE ATA/ATAPI Storage Controllers section of device manager. The disk drive is simply what that controller presents to windows as the "drive", and will AFAIK always be a Microsoft driver.

 

[Old Hippie]

Take your trolling and product fanboisms elsewhere please. AMD boards are fully capable, and OCZ Vertex3 drives are perfectly stable with modern firmware releases.

LOL!

AMD's RAID has always been @ 10% slower than Intel's and the OCZ has always been "liberal" with their drive's speeds....check it out here.

When was the last time you saw SSD testing done on an AMD board? Maybe never? And why would that be? Maybe because the reviewers want the best speeds?

While you may have other problems (no AHCI?) there's no difference between SATA2 & 3 cables.

I'm no fanboi or troll but I know what I'm talking about.

 

[Makaveli]

LOL!

AMD's RAID has always been @ 10% slower than Intel's and the OCZ has always been "liberal" with their drive's speeds....check it out here.

When was the last time you saw SSD testing done on an AMD board? Maybe never? And why would that be? Maybe because the reviewers want the best speeds?

While you may have other problems (no AHCI?) there's no difference between SATA2 & 3 cables.

I'm no fanboi or troll but I know what I'm talking about.

Agreed, your first post may not have been that helpful

But you are 100% correct with OCZ drives you never know what you are getting and AMD chipsets are still not at the level of intels.

As for the OP do you have a drive image done?

Maybe you should then break apart the array and test the drives one by one.

 

[Zxian]

OCZ has always stated their write speeds using fully compressible data. The test from Anandtech shows incompressible data only. There is obviously going to be a discrepancy. The tests done by the OP were using ATTO and CrystalDiskMark with 0-fill - both of which are fully compressible data. There is no reason why he's getting half the performance he should.

You yourself mentioned a slightly slower speed using AMD raid. Again, this doesn't account for the large discrepancy the OP is seeing.

All RAID controllers automatically enable AHCI when configured in RAID mode.

When was the last time you actually noticed the difference between any modern system's IO outside of benchmarks? I upgraded my desktop from a single OCZ Vertex3 to two Intel 520's and I can barely tell the difference.

Your post was completely unhelpful, and simply pointed out that he didn't have "the best" possible configuration. It didn't actually provide any additional insight into why he's getting lower performance than he should be given his hardware setup. That's why I asked you to kindly take your opinions elsewhere. You might know about reviews and benchmark statistics, but in this case, you provided zero constructive input.

 

[Andle Riddum]

The culprit is OCZ junk and in RAID no less.

 

[el-Capitan]

The culprit is OCZ junk and in RAID no less.

I beg to differ:

 

[jhansman]

I beg to differ:

Ah, there you go, confusing the guy with facts.

OP, have you posted your question on the OCZ support forum? Sumpin' definitely ain't right.

 

[Makaveli]

I beg to differ:

Once he gets his system figured out I will be curious to see his numbers your using an intel board not AMD.

 

[Andle Riddum]

I beg to differ:

Now where are the screenshots of yer blue screens? ()
I'm really curious why numb nuts choose a storage device that's 1 step forward in performance and 5 steps backwards in reliability...and it's not like there are lack of choices either!

Here are some hard data:
Overall
- Crucial 0.82% (as against 0.8%)
- Intel 1.73% (as against 0.1%)
- Corsair 2.93% (as against 2.9%)
- OCZ 7.03% (as against 4.2%)

- 15.58% OCZ Vertex 2 Series SSD 240 GB
- 13.28% OCZ Vertex 2 Series SSD 160 GB
- 11.76% OCZ Vertex 2 Series SSD 80 GB
- 9.52% OCZ Vertex 2 Series SSD 120 GB
- 8.57% OCZ Vertex 3 Series 120 GB
- 7.49% OCZ Vertex 2 Series SSD 60 GB
- 6.61% OCZ Vertex 2 Series 3.5" SSD 120 GB
- 6.37% OCZ Vertex 3 Series 240 GB
- 6.37% OCZ Agility 3 60 GB
- 5.89% OCZ Vertex 2 Series SSD 100 GB

The Vertex 2s have the worst scores but the Vertex 3s have nothing to be proud of either. Note that over the coming period, the Vertex 3s are doing much better thanks to developments in the firmware, with a rate of just 1.01% for the Vertex 3 120 GB as things stand.

http://www.behardware.com/articles/862-7/components-returns-rates-6.html

 

[Zxian]

Why are people still comparing AMD vs Intel controllers here? The OP has this hardware. It is performing slower than it should. Nobody is questioning that a RAID0 array on an Intel controller would be slightly faster in benchmarks. Stop posting meaningless information.

@ParseMeHard - Here's another forum post that you might want to check out. I'm not sure if the motherboard you're using would suffer from the same issue, but it might be worth checking out. You should be able to swap the SATA cables to other ports without risking the array.

@Andle Riddum - If you're going to quote something, make sure you quote it. Plagarism is bad in any form, even a forum post.

The link you provided even showed that the Vertex3 has a comparable return rate (not a mention of reliability, just bulk return rates) to the other big name manufacturers recently. When was the last time you heard of a Vertex3 user that was suffering from blue screens? I've never had any, and the people I know of who have SF-2281 based drives haven't had any blue screens since firmware 2.15 was released.

Blue screen issues have been resolved a long time ago (nearly a year now), and out of all the Sandforce based drives OCZ was often the first manufacturer to offer firmware updates to try to resolve the issues that plagued all SF-2281 drives. OCZ was the only big name that got hammered on this issue due to their market share at the time. I know of several people who owned Corsair SF-2000 series SSDs who had the exact same symptoms as Vertex3 owners.

I'm also really curious as to why you feel the need to bash the OP for his choice of hardware purchases. Internet anonymity strikes again!

 

[Andle Riddum]

Let me put it this way. Would you in your right mind mass deploy OCZ (fixed/patched)crapfest in your work environment? Well I can tell you an IT admin/consultant would never use a product with such a shitty track record.

This ain't a game, where you release today and patch later. It's a storage device and people(especially OCZ "fans") seems to sellout on reliability for extra performance points on benchmarks to show it to e-buddies...:\

You accuse me of plagiarism? I even provided the source(very hard find btw...where's my thanks?

If you got an axe to grind, try harder

 

[el-Capitan]

This is a stupid conversation and does nothing to help the OP. 1+ Year on my Vertex 3 and no BSOD. If I put my drives into RAID 0 I of course do NOT store important information on there. This is my gaming rig where 120GB grew to small after a few months. Bought a second one, made it one volume - viola. If the array craps out tomorrow I wont care. All i have to do is re-install my games.

Reliability of Vertex 1+2 vs 3 are completely different animals, btw.

Can we now please try and get OP 'up to speed'?

 

[Cerb]

Let me put it this way. Would you in your right mind mass deploy OCZ (fixed/patched)crapfest in your work environment? Well I can tell you an IT admin/consultant would never use a product with such a shitty track record.

Let me put it this way: the OP has 2 Vertex 3s on an AMD motherboard, so none of that BS matters one little bit.

Now, has the OP benched the drives again with read-ahead and write-back caching?