SSD Problems

Here's the Wildfire again, both are on the SiImage 3124:

I started fighting with the SSD's in my big Precision T7400 workstation last summer, and subsequently learned a lot about SSDs. The story starts almost two years ago when Dell replaced a very old Precision 650 that it could not make work (replaced all parts inside twice). The replacement was a gigantic T7400 with two 80GB disks and a small amount of memory that was like the 650, but two quad-core Xeon X5942's (3.4GHz, 3400Mhz front/side bus). Wow, Christmas came early. No warranty or service was available, since it replaced a 4-year old system with a just-now expired contract. But, I decided to keep it and added memory (total 12GB) and two Crucial M225 256GB SSD's, about $675 each back then. Put them in as RAID 1 off the on-board SAS 6/iR controller. All happy. Well ok, it ran a bit warm since this beast sucks up about 300 watts. But it was quiet and fast. Fast forward to last summer when suddenly the system seemed like it was running real slow. It took forever to boot. Apps took a long time to launch. Bench32 indicated the drives were running slow, mostly for write. Sometimes verrrry slow, other times just slow.

Subsequently: learn about how SSD's work, TRIM commands, and that most RAID controllers don't pass those through so SSD performance can degrade significantly over time. Eventually decide to break the RAID and run on just one of the two disks. Left the second one installed but unused. Then came that terrible Monday morning when nothing worked. Long story short: some process had leaked memory all weekend, and filled up the single M225 with a big paging file, then the disk stopped working. I was able to boot the other unused disk, but that had a very old OS image by now. Eventually I was able to restore a current image from backup. When up and running on the second M225 I tried all kinds of things fix the trashed one; at times it seemed to work then not. Firmware 2030 seemed to help. Nevertheless, read reviews and ordered a Wildfire 128GB SSD, thinking both the M225's are untrustworthy, or at least old technology. 100 hours later . . . . still obsessed with getting an SSD to work in this box. Use vacation time to work on it!

Current drives:

Patriot Wildfire 128GB FW 3.3.2

Crucial M225 256GB, FW 2030 (both now dead, dead, dead)

Intel SS DSA2CW120G3

Controllers:

Dell SAS 6/iR Integrated Workstation Controller

SiImage 3124 Controller

Syba SY-PCX400009 (could not make work in T7400).

Blog from most recent back:

2012.10.03

Still having problems with the Wildfire dropping off the system; usually no problem, but today that corrupted my 7GB outlook.ost file. Spent an hour with the tool to repair it, but then it still wasn't working, so deleted it. Reload all from server was faster. Notice FW 5.0.4 was released in September; consider installing it.

2012.03.15

Attempt to do a DISKPART CLEAN ALL on the M225 . . . the disk disappears before it is finished, and does not return. This is how the other one expired. Will try power off/on later.

2012.03.13

The M225 improved a lot over night, but still doesn't run nearly as fast as "before":

Wow; best numbers yet for the Wildfire. (Drive did not drop off during 4K QD32 write test.) Switched the Wildfire with the M225 (changed E: to J: then F: to E:, reboot). The M225 runs better now, sequential at 189/169, but 4K random writes still very slow, 6.8 and 8.6 for the QD32.

2012.03.12

I've been running with the Intel drive as C: and the M225 as E: without problems for about a month. Today I wrote about 86GB to the Crucial, and its speed went way south, then the system blue-screened. On restart, no drives found at all on the SiImage controller (the Wildfire was on it too, empty); power off/on and they re-appear. But the M225 is still running really slow; CDM says 112 MB/s for sequential read and 9.973 MB/s for sequential write. The Wildfire is still at 262/252. Maybe TRIM commands need to catch up on the M225. Wait overnight.

2012.02.14

Patch Tuesday; install all MS fixes. On reboot system goes into Macrium Reflect recovery partition instead of Windows Active partition. Recovery code can't see Intel SSD. Try a bunch of things then end up plugging it into the Dell SAS 6/iR controller and then "fixing boot problems" in the Macrium Reflect recovery console. Don't know how the partition information got screwed up. Meanwhile, with the Intel connected to a different controller, let's see if it runs differently:

Curious.

2012.01.25

Couple more BSOD's over the last couple days, and had to reboot. Time to try something else. Order an Intel 120GB SSD.

Also interested to see benchmarks from a new ThinkPad X220 tablet with an Intel SSDSA2BW160G3L (whatever that is). Not too bad:

Not much different; here's the one done as a native drive again, so you don't have to scroll up:

Criminey Nutley! Look at those random write speeds compared to the one right above. WTF. Maybe the controller is doing some write cacheing. Want to go home, but have to try making the Intel a SiI Concatenation logical drive too . . .

which is a little slower than a while back. Go back to the SiI BIOS and take the Wildfile logical drive "SiI Concatenation" out, put the Crucial back in. Now can't boot off the Wildfire, guess because the drive signature changed or something. Put that back as a logical drive and boot ok. Run CDM on the M225 again to see if there is a difference:

Un-define the Crucual M225 to the SiI controller so it shows up native. WildFire boot disk is still a "SiImage SCSI Disk Device". Here's how it runs now:

Let's try the Wildfire again to see if it can finish without a BSOD . . . sequential read comes out the same, but the WildFire blows it away with 252 for sequential write, 253/249 for 512K random read/write, 32.8/71.7 for 4K random write, and 149/156 for the 4K QD32 (AND NO BSOD this time!).

So it is a little slower in some areas, but not much.

2012.02.08

Repeating CDM runs on the WildFire after CLEAN ALL and format all as one partition. Set test runs to 9, No errors 2 times. On third pass, Write Error, device not connected; was pass 1/9 of Random Write 512KB. Drive reconnected a second later. Repeat:

ok
disconnects during 9th pass of the random read QD=32 test
no error but didn't do 4K QD32 write test at all
ok
gets write error and disconnects during 8/9 of sequential write.
ok
ok
write error, disk disconnects during 4K random write
ok
ok

2012.02.07

Still running C: boot drive off the Intel; no problems. Have run CDM about 10 times on it, and also no problems. Repeat again on the Wildfire. First run ok, then "disk not ready error" on start of second; a second later the OS sees a newly attached disk. Second run ok; third failed with read error on 3rd pass of random read test, QD-32. Drive gone from system; re-connects some 15 seconds later. Very many "The driver detected a controller error on \...\DR8." system event log entries. Next run ok. Next run, get "Write Error. 0x0000048F: The devices is not connected." during the 4th run of the first sequential write test; drive ready a second later.

Delete both partitions and run a CLEAN ALL in DISKPART (9:18 AM-9:26 AM). Activate disk, create single simple partition, do quick format. Repeat CDM; runs ok 4 times (5 passes each test). Change test runs to 9. First run ok; second has disk read error, but drive did not go offline; third ok; 4th ok; 5th ok; 6th ok; 7th ok.

2012.02.06

Have switched my boot drive to the Intel, and undefined the "SiI Concatenation" logical drive for the WildFire in order to see if the controller or something else besides the Wildfire SSD is causing the occasional reboots and BSODs. Also found that any disk connected to the SiI controller shows up fine after booting, in order to boot from it a "SiI Concatenation" logical drive has to be defined.

CDM 3.0.1 x64 gets a disk I/O error after completing all benchmarks. Try again and get "drive not connected" right away, then it is connected. Third try, and it fishes without error. The numbers come out about the same as when it was defined as a SiI Concatentation logical disk.

2012.01.27

The Intel 120GB SSD arrived. Slap it in the T7400 (love all the extra power connectors). I'm a bit confused because I didn't define it to the SiI 3124 controller, which I thought I had to do. But it shows up fine and CDM says:

2012.01.18

Got a BSOD yesterday and one today, in page kernel error. Removing paging file from Wildfire drive and put it back on the Crucial M225. Check Patriot Memory; 3.3.2 FW is still most current.

2011.10.31

Still running fine with the WildFire 3.3.2fw as long as I don't run the CrystalDiskMark benchmark. Was surprised to see Crucial now has a 5 year warranty on the M225. Called for RMA, but was told if there is no sticker on the back of it (I think there never was), then the warranty is void, even though I have a packing slip and invoice. Then look in a Crucial forum where an employee said some of them had ID stickers on the box, and the lab would know which ones. So I posted the serial number hoping mine was one of those.

2011.10.24

Try the SY-PEX40039 SATA III card again. BSOD again after inserting card; don't understand pre-installation instructions which indicate OS needs to be reinstalled to install driver. Find newer drivers for Asmedia 106x, 1.3.1.0 at http://www.station-drivers.com/page/asmedia.htm (dated 10/17/11). No help, BSOD again. Move to Gen 2 slot. BSOD. Boot to system recovery console, BSOD before it finishes loading. Maybe an address conflict: Remove SY-PCX400009 and plug SSD's back into onboard SAS/iR. BSOD. Try switching BIOS from AHCI to ATA mode. Nope, BSOD. Try starting in safe mode, no, BSOD. Uninstall driver altogether, get BSOD. These are all MACHINE_CHECK_EXCEPTION. Enough for today; send a note to Syba support.

2011.10.21

Received the SATA III card; not a Syba that I thought I ordered but box says IOCREST. Model, however, is a SY-PEX40039 and it has 2 internal SATA III ports on a tiny 1x PCIe card. Need to work up courage to Wildfire on it. Better boot from Crucial M225 first to be sure system still works. After putting the card in where a USB 3.0 adapter was, boot, and see:

Asmedia 106X SATA/PATA Controller Ver 0.89 AHCI Mode

S.M.A.R.T. Supported

Using PCIE Gen 1

SATA PM 1 Port 0 Patriot WIldfire

...

No area available for Command List/Command Table/FIS

AHCI BIOS not installed

Oh, guess that slot isn't a 2.0. But what's with "AHCI BIOS not installed"? Go into BIOS, verify SATA operation is AHCI, turn SMART Reporting on. Restart into the boot menu and pick the M225, oops, blue screen MACHINE_CHECK_EXCEPTION. Maybe the partition clone isn't good. Put Wildfire back on the SY-PCX400009, with no drives, no complaint from the SY-PEX40039. Except get another machine check as Win starts to boot. Remove the SY-PEX40039 and all ok. Download drive package and read readme.txt which says to pre-install the driver with the "F6 method" which is to book a diskette or CD. They must be joking. I installed the driver anyway and put the card back in, restart (still no drives connected to it) . . . BSOD again. Try repair again (AHCI off).

Just checked the forum and Patriot released FW 3.3.2 . . . let's try that! Gosh, I'm tired of moving these disk to a different system to update them, but I can do it with my eyes closed now.

Ok . . . one run of CDM with 5 passes of the 4k QD32 test .. . 178/132. Second ok at 179/84. Third run . . . BSOD, 0x000000F4 stop. Try CDM setting Test Data set to All 0x00. The 4K QD32 comes in a lot different at 148/185 and finishes ok. Second . . . . bsod when first pass of the QD32 write test starts.

2011.10.19

Made 3 runs of CDM on the Crucial M225; no BSOD or restarts. The random write QD32 test, which causes problems with the Wildfire, comes in at only 10 to12 MB/s, one tenth the speed of the Wildfire.

2011.10.18

No crashes today (no running benchmarks either). Order a Syba 2 Port SATA II PCI-e x1 controller (no RAID) to see if that would run the WildFire faster or better. Worth a try at only $20. Patriot support said they've only tested LSI controllers and they all seem to be in the $300 range, and all do RAID. I can't seem to find a PCI-x 2.0 x4 SATA III controller anywhere. One that ASUS announced that had USB 3.0 too would be great; no one seems to sell that. An x1 controller will max at 500MB/s (see Wikipedia), less than the SATA III max of 6Gbit/s so I don't expect to see a great performance boost, but since I see a big difference for the Wildfire between the on-board SAS/iR controller and the Syba SY-PCX400009 another data point will be informative.

2011.10.17

Patriot sent me a link to the test 3.3.0 firmware. Was going to wait until they tested it some more, since I was not having trouble other than running CDM (where the system always BSOD'ed or rebooted at the last write test). Put disk in an Optiplex 790 to update firmware (update can't see disk through the RAID controller). Update went fine. Back in T7400, booting Windows 7 is again very fast. The CDM finishes now:

I don't know if the 107.6 for the 4K QD32 write benchmark is significant compared to 160.8 reported with the 3.1.0 firmware (see below under 2011.10.12). Rest of the numbers are similar. Guess I don't care as long as it doesn't crash the computer.

Optimism premature. Second run does a BSOD. Third try ok. Enough for today.

2011.10.14

Surprised to see many file system errors in the system event log for the previous boot partition on the M225, yesterday at 5:50; must be during the SEP scan. Wonder if the Macrium disk cloning is not quite perfect. Chkdsk /F finds no errors.

The Wildfire firmware updater cannot find the disk -- no surprise since the Syba controller makes it appear as a "SiImage SCSI Disk Device". I may regret this, but start another CDM run . . . gets to the last test and the numbers look good. Uh-oh, Chome isn't responding . . . whoa, another BDOD. Yippee!

Ok. Put the Wildfire in an Optiplex 790 (no RAID controller) and updated firmware to 3.2.0 (was 2.1.0). Boot back on T7400 (data was not cleared). The docs for the firmware had:

-Improves Drive Management after unsafe power cycle.

-Improves Drive Recovery from Sleep/Hibernate failures.

-Adds support with specific Raid Cards.

-Enhances Drive power management.

Hopefully the Syba SY-PCX400009 was one of the "specific Raid Cards." Dare I try the CDM benchmark again? Better than another BSOD and losing work . . . oops, plugged it into the SAS/IR controller by mistake instead of the SiI 3124. Ran CDM without realizing that and thought perf was now 50% of pre-3.2.1. Back with the Wildfire on the SiL controller, run CDM again . . . numbers look like those below . . . nope, about halfway through the last test (the 4K QD32 write), and system reboots (no BSOD). PHOOEY!! Windows really slow to start . . . power off/on, seems ok now. Better schedule chkdsk (no errors found). Open a ticket with Patriot to see what they say.

2011.10.13

Was just marveling at how smart I was to get my SSD problems straightened out. Decided to run CDM once more to see if Wildfire numbers still look as good as yesterday. On first write, BSOD at 3:47pm, Kernel_Data_Inpage_Error, 0x0000007A. Hmm, maybe some incompatibility between the Wildfire SSD and thy Syba controller.

Schedule chkdsk /f for C: drive. No errors found. Also look for current firmware version, which is 3.2.0 (but no date so I'm not sure I need it or not, plus I don't see a firmware version in the Syba card's bios setup). Start a full virus scan just in case.

2011.10.12

Moved Wildfire to SiImage 3124 controller, cloned C: partition from working M225, adjusted driver letters and boot order, and booted it. Delete E: partition and use one on the M225.

CrystalDiskMark (3.0.1 x64) says it runs a lot faster now:

Yesterday when it was on the SAS/IR controller (and not the boot disk, and had 3 partitions), it the benchmark was:

Wow.

2011.10.11

Boot and run all day from the one Crucial M225 that seems to still be working; try to get I/O failure on the WildFire; could not. Maybe SAS 6 driver update fixed that problem. Leave it there or move it to the SiL controller?

Try many times to do DISKPART / CLEAN ALL on the M225 that doesn't seem to work; it keeps dropping off into another dimension. Conclude it is broke.

2011.10.10

- Crashed yesterday. Dang; thought running off the Wildfire drive alone would solve that.
- Install PCI-X SATA card (Syba SY-PCX400009) that came Friday; hook up both Crucial M225. Hmm, it want's to do RAID; no option for 1 drive RAID 0. Eventually set drives in concatenation mode.
- Around noon, crapload of "The driver detected a controller error on \...\DR2." events, then system powers off. Sweet.
- Look around some more. Hmm, there is a new driver for the SAS 6/iR controller . . . 1.28.3.52 7/1/2009 . . . thought I had already looked for a driver update for that.
- Syba supported replied to my question asking if the SY-PCX400009 will pass OS TRIM commands through to a single disk in concatenation mode: yes.

(Wish I had started this blog sooner.)

2011.09.28

Spent all spare time trying to clean or do low/level format of these M225's. Nothing works (tried several different computers, boot CD's, various utils, etc.). Eventually read about using DISKPART and CLEAN ALL. By golly, that runs (takes about 4 hours on an Optiplex 790). CDM benchmark is now a whole lot better:

At least on one. On the other, I don't think it was ever able to finish.

2011.08.26

Have both M225 running, but the "bad" one runs slow:

Compared to the "good" one:

Which isn't all that good.

Page updated

Google Sites

Report abuse