04-04-2013 01:03 PM - edited 04-04-2013 01:59 PM
using z420, Firmware is up to date, Windows 8 - with latest Drivers from HP Homepage.
Storage Mode is set to RAID + AHCI.
Samsung 840 PRO SSD attached at SATA Port 0 (AHCI 6Gb/s)
2x Samsung Barracuda attached at SATA Port 3 + 4 (AHCI 3GB/s) running RAID 1
The Problem is that my SSD is NOT detected in AHCI Mode (detected as IDE) - and so the Performance struggles.
When I detach the 2 Barracuda and change the Storage Mode in BIOS to AHCI only (without RAID) and restart Windows (AHCI driver activated in Registry) the SSD is correctly detected as AHCI and delivers wanted speed.
Switching back to RAID + AHCI the SSD is again detected in IDE mode.
Is there a Problem with RAID + AHCI Mode that the SSD is detected as NON AHCI?!?
Hope someone can help me on this.
Solved! Go to Solution.
04-05-2013 12:55 AM
Some more information:
- I now moved the 2 Barracuda to the SCU Ports - did not help.
- Also tried if without the Barracuda installed - no luck
Even when only the SSD is present - the Controller will present it in IDE Mode as long as RAID + AHCI is activated
04-05-2013 03:57 AM
To exclude hardware incompatibility, I just removed the Samsung 840 PRO SSD and installed the originally HP HDD which was delivered with my z420.
BIOS is in RAID + AHCI Mode - and the HP HDD is presented in IDE Mode too.
04-05-2013 11:09 PM - edited 04-05-2013 11:10 PM
I have a Z210 where i have recently installed an Intel 520 SSD and two 600Gb SATA 10K RPM drives with the later two drives in non raid mode. BIOS was set as RAID+AHCI and all works as it should with all drives running in 'AHCI mode' though it must be noted that i am using Intel's "iaStor.sys" driver rather than MS's generic "ahci.sys" driver.
I plan to convert the 600Gg HDD to RAID0 in the near future and will post results.
See this post for more detail on how to verify the drivers being used.
So how do you know you are running with IDE drivers and this is the reason for the poor performance?
[edited: clarified i recently changed my config from that shown in the linked post]
04-06-2013 09:25 AM - edited 04-06-2013 09:38 AM
thank you for your reply!
My assumption that the SSD is running in IDE Mode was because:
- write speed of SSD is 230MB/s instead of 520MB/s
- when changing from RAID+AHCI to AHCI (and activating the appropriate driver within registry) I get the full 520MB/s
- when RAID + AHCI is activated the SSD is discovered as "aSsmnu gSS D48 0RP OeSirse" - when I change BIOS to AHCI only it's detected correct (Samsung 840 PRO)
But it seems my assumption is wrong ... When I take a look at device manager - "list device by connection" I can see that the SSD is connected to the Intel C600/C220 chipset SATA RAID Controller. (see attached Screenshot)
But why do I get that bad write Speed in RAID + AHCI Mode when it's correctly presented as AHCI drive?
Is the drive using the RAID driver? Does this support AHCI?
Just don't get it why it's working correctly in AHCI Mode but not in RAID + AHCI?
- device_manager_by_connection.jpg --> Devive Manager View by Connection
- controller_driver.jpg --> Drivers used by Controller
04-07-2013 09:50 PM
Seems my previous post disapeared :?
Anyway, today i thought i would convert my 2 separate 600G 10K SATA HDD into a Raid1 Array using the Intel Rapid Storage Technology App from within Windows7 itself. Both disks had been previously formated and used. As such i migrated all data onto the first disk and thus made the second disk empty. Luckily the IRST app allowed me to migrate my data from my first disk to the new array comprising the first and second disks. As such nothing was lost during the conversion :) End result was SSD + RAID1 (comprising 2x 600GB velociraptors)
Prior to migration to Raid1, as a first step, i benchmarked my SSD using HDTune2.55 and it returned the following:
Also, the device magager listed the following, note the SSD, 2xHDD and BD burner and also note the drivers for the SSD:
I then performed the Raid1 migration and after the process was complete which took over about an hour, the IRST listed the following:
Also the device manager listed the following, note the SSD, Raid1 array and the BD burner and also not the drivers for the SSD:
I again ran the HDTune benchamark and recieved the following:
So as can be seen, there was no degradation to SSD performance in having the 2x HDD connected in non raid using the IRST intel drivers.
Having done the SSD benchmarking pre and post HDD raid construction, and not knowing the finer points of drive benchmarking, the results look odd. Intel ARK for the 120GB SSD shows:
The SSD read and write performance specified by Intel is much better than what HDTune benchmarking tool indicates, specifically Intel shows read of 550MB/s and write of 500MB/s compared to the HDTune of 300-375MB/s (read).
So the questions are:
1) Why does HDTune show ~375MB/s then a drop to ~300MB/s until 45% and thereafter it's back to ~375MB/s?
2) Why is the real? performance not even close to those indicated by the Intel ARK specification (550MB/s)?
3) Is it simply that the Intel drivers are not as good/optimised as the MS drivers?
Has anyone with more SSD benchmarking understanding than me actually benchmarked the Intel iaStorF.sys drivers against the MS ahci.sys driver. What were the results.
04-07-2013 10:32 PM - edited 04-08-2013 02:43 AM
Well it seems there may be an issue with Intel RST drivers version 11.5 that will not correct itself when installing the later RST driver v220.127.116.113! And I upgraded some weeks ago and had not notices the subtle impacts :)
According to the net, this issue seems to impact disk performance in that the SCSI filter driver "iaStorF.sys" is loaded (which can not provide trim) instead of the correct "iaStor.sys" driver. The issue also results in poorly named "device strings" within device manager (in my case it adds 'SCSI Disk Device' to the disk name).
Whether this is the real cause of my seemingly degraded performance is yet to be determined but investigations continue... Will post when I have more info...
[edited to add the following]
It seems that during the driver update process, Intel changed their drivers from a simpler structure of "iaStor.cat", "iaStor.inf" & "iaStor.sys", etc as used in version 10.6.0.1002 (available as sp57563.exe from HP) to a different structure of "iaStorAC.cat", "iaStorAC.inf", "iaStorA.sys", "iaStorF.sys", etc as used in version 18.104.22.1683 (which is not available from HP but directly from Intel). It also seems that this driver structural change occurred post RST version 11.2!
Q: why has HP not released a more recent update considering Intel has made a number of releases since HP's last release?
The Intel RST change log is rather port, unlike the Intel RSTe Customer Release Notes which have all the history listed within the one PDF, so it's difficult tracking what has changed and what dependencies exist in RST unless one does a bunch of searches and downloads.
Also the net discussions tend to be related to using RST and SSD's configured in Raid 0. Specifically whether the RST driver can pass through any OS initiated TRIM commands AND whether the OS can support TRIM commands for the SCSI mini port connected hardware (as the RST drivers present SSD's and HDD's to the OS).
So the end result is that though I thought I understood what I was doing when installing my SSD, it seems I am more confused than ever. And I still have the confusing drop in performance as reflected in the HDTune benchmarking!
Does anyone have a clear technical explanation of how to ensure correct drivers are loaded (latest Intel ReadMe indicates iaStor.sys should be loaded but this does not exist - an indication of the quality of Intel docs?).
Previous questions are still valid and I hope someone can answer.
[edited to add the following]
And it's worth noting that Z420 has a C6xx series chipset and should use Intel RSTe drivers (@ version 3.x) rather than the RST drivers (@ version 11.x) which is the drivers relevant to my Z210 machine and the basis of my comments.
Finally, whether the iaStorF.sys is really a remnant from the RST upgrade, I don't know, and have yet to find any concrete information (just reposts of the same stuff) so for now I wont play within the registry until I discover some real information (from Intel, HP or MS).
[edited to fix my spelling and added the following :)]
Just to add further confusion, running 'winsat disk' in an elevated command prompt yielded 521MB/s sequential 64 read, 266MB/s random 16 read & 502MB/s sequential writes.., somewhat different to HDTune...
04-10-2013 02:37 AM
Well it seems there may be an issue with Intel RST drivers version 11.5 that will not correct itself when installing the later RST driver v22.214.171.1243! And I upgraded some weeks ago and had not notices the subtle impacts :) According to the net, this issue seems to impact disk performance in that the SCSI filter driver "iaStorF.sys" is loaded (which can not provide trim) instead of the correct "iaStor.sys" driver. The issue also results in poorly named "device strings" within device manager (in my case it adds 'SCSI Disk Device' to the disk name).
As you say there may be in issue with RST - I'm using the RSTe Drivers - happened to know if those Problems can arrise with that too? As the crippled device name in device manager could point to this Problem?!?
And I guess you didn't find a solution for this yet?
04-10-2013 03:49 AM - edited 04-10-2013 03:52 AM
RSTe drivers are the correct for your Z420.
RST drivers are correct for my Z210.
Just wish HP would update their RST drivers to bring them into line with Intel revision... .
Finding accurate technical information is rather difficult and the posts I have read give vague and possibly inaccurate information not well supported by references to manufacturers documentation or published statements.
What seems to be the situation is that Windows 7 send TRIM commands to disk devices in some situations.
It seems to be dependent on whether the OS is aware of SSD's being used (hence 'winsat disk' will let OS know a very fast disk =SSD is available). It also depends on which drivers are used and whether the drivers can pass TRIM (I have read that MS SCSI miniport drivers can not handle passing TRIM). Then there is the SSD firmware itself which may also not support TRIM (as in early SSD devices). And if you have Raid hardware (PCIe cards) between the OS and SSD, it also depends on that devices firmware capabilities as to whether the card reports rotational speed as 0 RPM (in which case it is assumed an SSD by OS) and whether the cards firmware can handle TRIM commands.
However, I am using the Intel RST drivers and dumb(ish) Intel mobo raid , I don't know if all the links in the chain are OK. I can't even apply that MS SCSI miniport drive logic to Intel SCSI miniport drivers, so RST may indeed pass TRIM from OS to raid/c206 chipset to SSD firmware. But again nothing is concrete because I can't find documentation (and no one, HP, Intel MS) like to tell.
And if you look at the benchmarking, it may be that due to my C:\ drive being 109GB in size with 46GB used and 63GB free, this somehow relates to the speed of the benchmark tool - slower in the used portion of the SSD and faster in the empty portion of the SSD. But the occupied pages and blocks within the SSD will perform slower than the unoccupied part of the device w.r.t. writes so why it impacts reads, who knows. Maybe the OS is doing writes in the background while benchmarking?
So, nothing concrete...
If your performance was better with msahci.sys drivers and you don't use a raid configuration, by all means return to this driver. In my case I am using raid 1 for my data so am stuck with what I have. For me it's still much better than spinning platters :)
04-17-2013 02:08 PM
alright - I was able to solve the Performance Problem now!!
I replaced the RSTe iaStorA.sys with the RST iaStorA.sys version - and now the SSD deliveres the full speed ...
I attached 2 Screenshots - "before.jpg" shows my configuration with performance Problems; "after.jpg" my current configuration with full performance!
Only problem left is the crippled device Name of the SSD - but I hope this is no problem ... ?!?
06-28-2013 11:32 AM
Thank you for the information, had the same problem with HP Z420 workstation and slow disk performance.
I am now using the Intel drivers and getting better performance on all drives (ssd and standard) on the SATA bus.
Some 'strange' issues remain... 2 identical WD BLACK HDD's, one on the 6Gb/s connector and one on the 3Gb/s connector. The drive on the 3Gb/s connector performs better.
When I try to switch the controller setting in the BIOS to AHCI from AHCI/RAID. My system could not boot anymore and it was impossible to get it to work, even by manually loading the AHCI driver. I read in another article that I should ignore the fact that the Samsung software claims the SSD drive is not running in AHCI mode so I'm not going to bother with that for a while.
I hope new HP drivers will appear soon to correct these issues by default.
06-28-2013 01:43 PM - edited 06-28-2013 01:50 PM
Yes, the Samsung SSD Magician is apparently incapable of dealing with the RAID+AHCI, but theSD is definitely not in IDE mode...
BTW, the 840 Pro seems to be quite good performance-wise compared to the 830: