Z68 - SRT cache and boot volume on a single SSD disk - it works!

[BonzaiDuck]

Correct me if I'm wrong about the assertions I make here.

I myself purchased the lesser 510-series Elm Crest -- for its speed first of all. But [and you can argue that my research was incomplete, but I don't think that's fair] -- BUT!!. . . . I also chose Intel because I was assured of TRIM support with "older OS's."

I have initially built my Z68-system around VISTA-64-SP2. Supposedly, Win-7-32/64 has TRIM support "built-in." No problem: Intel provides it with SSD-Toolbox and "SSD-Optimizer."

Just one hitch. If you check Intel's web-site on the Toolbox, Optimizer and TRIM, they note that [my paraphrase] "We are investigating the issue of making TRIM work with RAID configurations . . . currently, it doesn't . . . " and that's what the Toolbox and Optimizer report when you bring them up running.

This suggests to me that TRIM under Windows-7 doesn't work with your SSD-caching either.

Or -- just to clarify for the curious and casual reader: SSD-caching is a "RAID-0"-type configuration. So .. . "no TRIM" . . .

Anyone have any thoughts and clarifications?

 

[espex87]

Trim doesnt work in the cache but with most SSD's it is not a problem because of a background (NON OS) garbage collector (replacement of trim).

I have no idea about raids etc, so this is why I am posting this question:
When I use a Kingston 96GB SSD als boot&SRT, there is mentioned that SRT only works towards 1 other HDD

My setup: 1 SSD as boot & SRT, 2 or 3 HDD's as storage (say HDD A, B & C). Does it mean that when I run games from HDD A, they are cached. But if I run anything from HDD B & C I have no performance gain?

I never do anything in the bios with drives, I just connected the cables and turn my pc on. Hence I know nothing about raids...

 

[Chaoticlusts]

OMG! I begin by checking running services. Not there. I check the directory you mentioned. Not there. I go to install the files from the Asrock site, says I already have a newer version installed, I proceed anyway, reboot, AND IT WORKS! I have the RST application now!

Thank you so much Aszu (and Chaoticlusts for convincing me to keep working on this)!

Oh and the best part... acceleration is already enabled for my hard drive like I suspected from the BIOS but still very happy to know for sure (and have a GUI to manage it).

Awesome glad it's sorted for you weird solution but hey it works (nice catch Aszu )

Correct me if I'm wrong about the assertions I make here.

I myself purchased the lesser 510-series Elm Crest -- for its speed first of all. But [and you can argue that my research was incomplete, but I don't think that's fair] -- BUT!!. . . . I also chose Intel because I was assured of TRIM support with "older OS's."

I have initially built my Z68-system around VISTA-64-SP2. Supposedly, Win-7-32/64 has TRIM support "built-in." No problem: Intel provides it with SSD-Toolbox and "SSD-Optimizer."

Just one hitch. If you check Intel's web-site on the Toolbox, Optimizer and TRIM, they note that [my paraphrase] "We are investigating the issue of making TRIM work with RAID configurations . . . currently, it doesn't . . . " and that's what the Toolbox and Optimizer report when you bring them up running.

This suggests to me that TRIM under Windows-7 doesn't work with your SSD-caching either.

Or -- just to clarify for the curious and casual reader: SSD-caching is a "RAID-0"-type configuration. So .. . "no TRIM" . . .

Anyone have any thoughts and clarifications?

It's a little complicated...TRIM *definitely* doesn't work on the cached drive...unfortunately it's kind of hard to tell whether it works on the non-cached segment the wording is a little ambigious...things about TRIM working on RAID volumes that are not in an actual RAID array (since the OS partition isn't in an actual RAID array...) the biggest problem with telling is the way TRIM works from what I understand is it's always on (unless you turn it off) always sending signals and doesn't report if the signals don't make it too the SSD (which is what happens when they run into a RAID controller and why it doesn't work in RAID arrays...but if it's just a RAID volume not in an array it passes through the controller untouched and does work) the only way to tell if TRIM is or isn't making it through is watching SDD activity extremely closely and trying to notice a TRIM command running (I have no idea how to do this have seen it mentioned elsewhere sounds extremely complicated and beyond what I know at this point ) so basically..windows always runs TRIM unless told otherwise and won't tell you it's not working...so....ummm....hope? Or if you know how this stuff works a lot more than I do you could actually look into it really closely find out and tell us! but it *definitely* won't work on the cache partition atm as has been stated previously that's setup similarly to a RAID 1 array so it should cause the same problems that stop TRIM

^_^

 

[bo_jangles]

I've gotten this working, too. Can others post results?


i7-2600k (stock settings)
Asrock Z68 Extreme4
Kingston SSDNow V+100 96GB
WD Blue 640GB
2x Maxtor 250GBs (not really important to this discussion, but they are there)
16GB of RAM (yeah, it's silly, but it was cheap)

Regarding the installation, I basically did everything backwards but eventually got there.

I started with installing Win7 x64 on my 96GB V+100 in AHCI mode.

Crystal Disk showed this:

Kingston SSDNow V+100

-----------------------------------------------------------------------
CrystalDiskMark 3.0.1 x64 (C) 2007-2010 hiyohiyo
Crystal Dew World : http://crystalmark.info/
-----------------------------------------------------------------------
* MB/s = 1,000,000 byte/s [SATA/300 = 300,000,000 byte/s]

Sequential Read : 214.236 MB/s
Sequential Write : 159.649 MB/s
Random Read 512KB : 195.688 MB/s
Random Write 512KB : 154.355 MB/s
Random Read 4KB (QD=1) : 18.298 MB/s [ 4467.3 IOPS]
Random Write 4KB (QD=1) : 27.897 MB/s [ 6810.8 IOPS]
Random Read 4KB (QD=32) : 19.822 MB/s [ 4839.4 IOPS]
Random Write 4KB (QD=32) : 28.019 MB/s [ 6840.7 IOPS]

Test : 1000 MB [C: 67.3% (60.1/89.3 GB)] (x5)
Date : 2011/06/21 20:00:55
OS : Windows 7 Ultimate Edition SP1 [6.1 Build 7601] (x64)


Note the fast sequential reads of over 200MB/s.

Then I switched from AHCI to RAID to support two other HDDs I had. This caused some nice BSOD stop code 7b action, which I used RAIDfix to solve instead of reinstalling the OS. I used this setup for a while without running another bench on the SSD that I can remember. I had the latest Intel RST installed to have my RAID-0 working, but I didn't have an "Accelerate" choice.

To get SRT working, I used cloning software to take the contents of my SSD and put them on a different HDD. I booted with that and then saw the Accelerate tab. I set it to grab 18.some GBs of the Kingston SSD and use them as a cache. I then cloned the OS from the HDD and put it back on the 70-some GBs left on the SSD as the boot drive.

This works. Somewhere along the way (probably the cloning), the alignment was screwed up, so I used the Paragon SSD Alignment program to repair that.

Here's what it looks like in the RST (Array_0002 is just a normal RAID-0, so ignore that; I'm accelerating a single HDD):


However, my annoyance is that RAID mode, required to get this or just a basic RAID-0 to work, appears to slow the SSD a lot (particularly sequential reads) compared to AHCI mode. Here are my new bench results with the "Accelerated" setup:

Drive C: (Non-cache 70GB or so of Kingston SSDNow 96GB V+100)

-----------------------------------------------------------------------
CrystalDiskMark 3.0.1 x64 (C) 2007-2010 hiyohiyo
Crystal Dew World : http://crystalmark.info/
-----------------------------------------------------------------------
* MB/s = 1,000,000 byte/s [SATA/300 = 300,000,000 byte/s]

Sequential Read : 90.967 MB/s
Sequential Write : 184.787 MB/s
Random Read 512KB : 101.110 MB/s
Random Write 512KB : 156.064 MB/s
Random Read 4KB (QD=1) : 16.681 MB/s [ 4072.5 IOPS]
Random Write 4KB (QD=1) : 29.253 MB/s [ 7141.9 IOPS]
Random Read 4KB (QD=32) : 18.419 MB/s [ 4496.8 IOPS]
Random Write 4KB (QD=32) : 30.717 MB/s [ 7499.3 IOPS]

Test : 1000 MB [C: 61.5% (43.5/70.7 GB)] (x5)
Date : 2011/07/10 22:57:12
OS : Windows 7 Ultimate Edition SP1 [6.1 Build 7601] (x64)



Drive D: (WD 640 SRT cached with 18GB of Kingston SSDNow 96GB V+100 as "Maximized" )

-----------------------------------------------------------------------
CrystalDiskMark 3.0.1 x64 (C) 2007-2010 hiyohiyo
Crystal Dew World : http://crystalmark.info/
-----------------------------------------------------------------------
* MB/s = 1,000,000 byte/s [SATA/300 = 300,000,000 byte/s]

Sequential Read : 91.563 MB/s
Sequential Write : 126.426 MB/s
Random Read 512KB : 91.904 MB/s
Random Write 512KB : 128.831 MB/s
Random Read 4KB (QD=1) : 18.899 MB/s [ 4614.1 IOPS]
Random Write 4KB (QD=1) : 27.816 MB/s [ 6790.9 IOPS]
Random Read 4KB (QD=32) : 20.043 MB/s [ 4893.3 IOPS]
Random Write 4KB (QD=32) : 28.007 MB/s [ 6837.7 IOPS]

Test : 1000 MB [D: 0.0% (0.1/596.2 GB)] (x5)
Date : 2011/07/10 23:09:03
OS : Windows 7 Ultimate Edition SP1 [6.1 Build 7601] (x64)



For reference, here's an old benchmark of that same drive D: disk without any acceleration:

Drive D: WD 640 (1st partition, so outer tracks, no acceleration)

-----------------------------------------------------------------------
CrystalDiskMark 3.0.1 x64 (C) 2007-2010 hiyohiyo
Crystal Dew World : http://crystalmark.info/
-----------------------------------------------------------------------
* MB/s = 1,000,000 byte/s [SATA/300 = 300,000,000 byte/s]

Sequential Read : 112.363 MB/s
Sequential Write : 110.867 MB/s
Random Read 512KB : 43.747 MB/s
Random Write 512KB : 59.222 MB/s
Random Read 4KB (QD=1) : 0.580 MB/s [ 141.7 IOPS]
Random Write 4KB (QD=1) : 1.284 MB/s [ 313.5 IOPS]
Random Read 4KB (QD=32) : 1.556 MB/s [ 380.0 IOPS]
Random Write 4KB (QD=32) : 1.395 MB/s [ 340.5 IOPS]

Test : 1000 MB [D: 78.3% (203.2/259.3 GB)] (x5)
Date : 2011/06/22 12:45:52
OS : Windows 7 Ultimate Edition SP1 [6.1 Build 7601] (x64)



The "Accelerated" single disk has slower sequential reads with the SSD in front of it than without. Yes, the other specs are better, but why would reads be slower than the no-cache case? If I'm trying to cache game loading, I'm not sure this is an improvement.

Also, the really annoying part is the non-caching part of the SSD used as my boot drive is now super slow compared to its original performance. Reads are less than 100MB/s compared to the over 200MB/s before. That's cheesy.

 

[quaddragon]

I wanted to put my 2 cents in. I managed to get this running thanks to this wonderful guide and a little help from the interwebs. It does work! the OS boots wonderfully. The ONLY problem I am see is that now and again my build will hang pausing for a minute or 2 then working normally again. I am thinking of going back to a solid state OS some apps with my 3tb seagate as data only. I just got a vertex 3 120, I think it would be a waste to use it as cache and scratch drive.

 

[BonzaiDuck]

I wanted to put my 2 cents in. I managed to get this running thanks to this wonderful guide and a little help from the interwebs. It does work! the OS boots wonderfully. The ONLY problem I am see is that now and again my build will hang pausing for a minute or 2 then working normally again. I am thinking of going back to a solid state OS some apps with my 3tb seagate as data only. I just got a vertex 3 120, I think it would be a waste to use it as cache and scratch drive.

I've watched occasionally, casually as the SSD's came out. People were starting to RAID0 with them. Certainly the performance gains were incredible over HDD's, whether in a RAID or alone.

The problem I see -- which isn't so much a problem, but a "trade-off" -- is that to get a single 500GB volume (standalone SSD or RAID0 with two or more), you'll pay ~$2.00/GB. If you use a smaller SDD volume or especially the smallest and only invest maybe $120, you can barely get the OS and some programs on there. Further, you will split your files and programs among more logical drives or volumes.

Now a person MAY want to split files on different logical volumes. For instance, it is recommended for video capture and editing to have a separate physical disk for that purpose. It would probably need to be a large disk, but that's not a major issue. It doesn't need to have the fastest performance if its performance more than keeps up with the demand for speed by the application to which it is put.

So if an SSD running full-throttle as SATA-III providing 500MB/s in sequential reads will outperform an SSD SATA-II, and an SSD SATA-II paired with an HDD SATA-II shows a 400% ("as much as . . ") improvement over the HDD's standalone performance, or 80+% of the performance of the SSD, then what sort of performance increase would you get with the SSD SATA-III (which has double+ the throughput of the SSD SATA-II)?

By using the SSD as cache, you not only get the enhanced performance, but you don't need to deal with the limitations of SSD as storage or the cost of SSD. You also provide a "simplification" for needing a fewer number of logical disk volumes -- operating as "standalone" whether SSD, HDD, SSD/HDD in conventional RAID0/1/etc. You can cache as many other disks (or volumes) insofar as you have double the number of SATA ports to provide a caching SSD for each disk or volume.

So . . . . a tradeoff. A need for a strategy. An accommodation of performance with storage size, storage need, and the need to actually separate one group of files from another.

 

[piquadrat]

I'm using RST driver no: 10.6.0.1002 at the moment on intel 320 Series 80GB ssd in hybrid (system / cache configuration). Had no problems.
Indeed there is a performance hit. Setting the ssd up into cache mode (with that pseudo-raid array) results in sequential read hit (from 260MB/s to around 200MB/s) and small random (1k) read decrease (from 20Mb/s to 18,5MB/s). Random writing fall to 36 from 40 (sequential is intact - a little weird if you ask me). So the most important aspects (random performance) is reduced by around 10%. Sequential performance is not worth a lot to me as this is a system drive. Above results are constant, I can see no further degradation in time (two weeks testing).
Cached volume (raid-0 of two quite old seagates) is noticably faster. I'm on igpu so didn't even try to start some demanding games (installed originally on mentioned raid-0) but for instance loading Half-Life2 random level is reduced by half comparing to 4disk raid-0 capable of 300Mb/s sequential read. Cached volume results around 250MB/s read in synthetics (Crystal Diskmark) and 90MB/s in writing. So it writes in ssd speed and reads in something in between ssd (originally 260) and raid (around 170MB/s). But again this volume is mainly for reading. And you can feel the speed bump. You do not have to guess if ssd caching works. This is noticeable difference.

 

[the_electrician]

Im having abit trouble getting this working some help would be appreciated.

wiped my vertex2e 60gb
fresh install win7 x64 on wd green 160gb
installed rst and partitioned ssd 18gb cache
rebooted with wd hhd disconnected
win7 on usb stick
win7 installation doesnt see ssd asks for storage drivers?

Can anyone provide a link to the storage drivers required for win7 x64so that partitioned ssd can been recognised so i can install win7 on it.

 

[piquadrat]

You need RST driver discette on pendrive and load it through F6.
http://downloadcenter.intel.com/Det...duct=Intel®+Rapid+Storage+Technology&lang=eng

 

[BonzaiDuck]

You need RST driver discette on pendrive and load it through F6.
http://downloadcenter.intel.com/Det...l%c2%ae+Rapid+Storage+Technology&lang=eng

That would basically be the RAID driver for the intel controller.

The SSD drive needs to be unpartitioned -- even if it has spare space for a partition later.

So once OS Win7 installed on HDD separately, connect the SSD after powering down, invoke the SRT software in Win, choose the SSD for the cache and size of cache and the HDD to be accelerated. If your machine is still "in the works," in process of being tweaked and OC'd, be sure to pick "Enhanced" over "Maximum" Mode.

For Electrician, the drivers will be on the Z68 motherboard installation disc, but you should instead go to the mobo manufacturer's web-site, get "support" for your Z68 mobo and pick "downloads." You want the latest driver revision, and the latest ISRT software revision. If the download is a ZIP or EXE file, it will need to be run or decompressed to a hard-drive folder, from which you would locate the appropriate x64 drivers in an obviiously labeled subfolder. Just copy them to a USB flash drive and have them ready for OS installation. No need to identify which driver (ACHI vs RAID, etc.) Just copy them all. Win7 will find the right driver given you have correctly chosen RAID mode in BIOS for the Intel controller and drives connected to it.

 

[bo_jangles]

Thanks for the data, piquadrat. Your performance hit doesn't seem as bad as mine.

I also noticed that even enabling SRT in general (not even doing this OS/cache combo setup we're discussing here) slows the sequential read performance of the SSD.

I'm leaning toward giving up on this combo OS/cache on one SSD scheme and just using Steam Mover to "manually" migrate less-used games that don't fit on the SSD. It's sad since I thought of the combo config before ordering parts and might have just saved a few bucks on a P67 if I knew SRT wouldn't let me have the best of both worlds. I suspect sequential reads are important for game load times, and when they are more than cut in half, it's no good.

I started a thread in the storage forum explaining my problem without all the extra talk about our combo scheme. So far it hasn't gotten any answers with data... http://forums.anandtech.com/showthread.php?t=2178040

 

[BonzaiDuck]

Thanks for the data, piquadrat. Your performance hit doesn't seem as bad as mine.

I also noticed that even enabling SRT in general (not even doing this OS/cache combo setup we're discussing here) slows the sequential read performance of the SSD.

I'm leaning toward giving up on this combo OS/cache on one SSD scheme and just using Steam Mover to "manually" migrate less-used games that don't fit on the SSD. It's sad since I thought of the combo config before ordering parts and might have just saved a few bucks on a P67 if I knew SRT wouldn't let me have the best of both worlds. I suspect sequential reads are important for game load times, and when they are more than cut in half, it's no good.

I started a thread in the storage forum explaining my problem without all the extra talk about our combo scheme. So far it hasn't gotten any answers with data... http://forums.anandtech.com/showthread.php?t=2178040

I'm forgetting precisely what I wrote earlier in this thread, so forgive me if I repeat anything.

Here's a benchmark review from Bjorn 3D in May. They paired an SATA-II SDD with an SATA-II Seagate 7200.10/11 (whatever):

http://www.bjorn3d.com/read.php?cID=2055&pageID=10557

We're already seeing "hybrid" drives that put the two technologies into the same HDD package. But I don't think it will be so soon that we see an end to electromechanical HDDs. We'll see capacity grow, even alongside SSDs-- which will have less capacity.

So far, I'm satisfied with what I got out of the combination, even for paying a premium.

I'd been in touch with an Intel Tech-Rep during the last week. Someone here had suggested that you wouldn't gain in gaming, but this was overturned by the Intel Rep. The more significant factor, though, is sequential SSD throughput -- not for itself -- but what it means for the HDD acceleration. The Bjorn 3D tests were made using SATA-II SSD and HDD. But the sequential throughput for an SATA_III SSD can be more than twice the approximately 200 MB/s we've mentioned here. Maybe slightly more than twice. IF the benchies show "sometimes" a 400% increase over standalone HDD performance with the SATA-II's, what would that translate to in terms of an SATA-III SDD? It would hardly matter whether the HDD was SATA-II or SATA-III, but matters a lot more for which SSD you choose, and the controller port you use.

All I can say -- it's a lot, lot faster than my four-drive RAID5 on 3Ware 9650-SE controller, and a lot faster than my RAID0 on another superb machine. NOt just a "lot, lot" faster, but a tremendous "lot more." The Bjorn 3D benchies have to be lame in comparison, yet nothing to sneeze at either.

 

[BonzaiDuck]

Here's a news item I picked up in a web-search this morning:

http://techreport.com/discussions.x/18653

This was published more than a year ago. Apparently, you can have an SSD configured under BIOS "RAID" mode, and TRIM will work. But it won't work if the SSD's are part of any sort of array structure -- which would include the RAID0 defining ISRT HDD caching.

This would mean that periodically, one would have to "unhinge" the caching (and also lose any partition if there had been additional drive space used conventionally). Then, if use of a separate utility is called for (as with Intel SSD-ToolBox and SSD-Optimizer), you would run it. After that, you would "re-hinge" the SSD to HDD acceleration caching, re-create the extra partition, and restore the files that had been stored on the latter.

The only thing left to do is "try it."

 

[bo_jangles]

Yeah, I don't think it's a TRIM issue. I can get the bad SSD read speeds of 65 or 70MB/s with SRT enabled and within seconds of disabling and releasing the SRT cache, I can get read speeds of over 200MB/s without doing anything else. Also, the Kingston is supposed to have aggressive internal garbage collection to reduce the need for TRIM (good for older OSes). If it was a TRIM problem, I'd expect the performance to be good at first and then get worse over time. Instead, it turns bad the instant SRT is enabled. It just occurred to me that I have an old SSD (40GB, no TRIM) I can try to see if the performance drop can be repeated with that. I'll have to do some secure erasin' first, but I'll report back.

 

[bo_jangles]

I remembered I had an old 30GB non-TRIM SSD sitting around, so I tried the same experiment. Now I am starting to wonder if there's a flaw in all Z68s, a flaw in my particular AsRock Z68 Extreme4, if I've made some configuration error, or if that's just how it works.


First test, no SRT enabled, just benching the SSD

-----------------------------------------------------------------------

CrystalDiskMark 3.0.1 x64 (C) 2007-2010 hiyohiyo

Crystal Dew World : http://crystalmark.info/

-----------------------------------------------------------------------

* MB/s = 1,000,000 byte/s [SATA/300 = 300,000,000 byte/s]

Sequential Read : 177.875 MB/s

Test : 1000 MB [ 77.3% (21.6/28.0 GB)] (x5)

Date : 2011/07/14 14:12:02

OS : Windows 7 Ultimate Edition SP1 [6.1 Build 7601] (x64)





Second test, 19GB SRT cache enabled, benching the remaining data partition:



-----------------------------------------------------------------------

CrystalDiskMark 3.0.1 x64 (C) 2007-2010 hiyohiyo

Crystal Dew World : http://crystalmark.info/

-----------------------------------------------------------------------

* MB/s = 1,000,000 byte/s [SATA/300 = 300,000,000 byte/s]

Sequential Read : 63.650 MB/s

Test : 1000 MB [ 0.8% (0.1/9.4 GB)] (x5)

Date : 2011/07/14 14:15:04

OS : Windows 7 Ultimate Edition SP1 [6.1 Build 7601] (x64)





Third test, disable SRT and release cache, no other adjustments:



-----------------------------------------------------------------------

CrystalDiskMark 3.0.1 x64 (C) 2007-2010 hiyohiyo

Crystal Dew World : http://crystalmark.info/

-----------------------------------------------------------------------

* MB/s = 1,000,000 byte/s [SATA/300 = 300,000,000 byte/s]

Sequential Read : 181.509 MB/s

Test : 1000 MB [ 0.8% (0.1/9.4 GB)] (x5)

Date : 2011/07/14 14:31:01

OS : Windows 7 Ultimate Edition SP1 [6.1 Build 7601] (x64)


The following review shows that SRT improves the sequential read speed of a HDD (but isn't as fast as the same Kingston V+180 64GB alone):
http://www.techspot.com/review/395-asrock-z68-extreme4/page12.html

My setup, however, ends up being so slow that the HDD alone actually has faster sequential reads than the SRT mode.


Can more people please run a similar test on their machines? Just bench the SSD data partition with SRT enabled and compare to a bench of the same data partition with SRT disabled (and cache released). I want to have a larger dataset to see if I have a defective board or what.


Thanks in advance to anyone who can help study this and provide data!

 

[BonzaiDuck]

I remembered I had an old 30GB non-TRIM SSD sitting around, so I tried the same experiment. Now I am starting to wonder if there's a flaw in all Z68s, a flaw in my particular AsRock Z68 Extreme4, if I've made some configuration error, or if that's just how it works.


First test, no SRT enabled, just benching the SSD

-----------------------------------------------------------------------

CrystalDiskMark 3.0.1 x64 (C) 2007-2010 hiyohiyo

Crystal Dew World : http://crystalmark.info/

-----------------------------------------------------------------------

* MB/s = 1,000,000 byte/s [SATA/300 = 300,000,000 byte/s]

Sequential Read : 177.875 MB/s

Test : 1000 MB [ 77.3% (21.6/28.0 GB)] (x5)

Date : 2011/07/14 14:12:02

OS : Windows 7 Ultimate Edition SP1 [6.1 Build 7601] (x64)





Second test, 19GB SRT cache enabled, benching the remaining data partition:



-----------------------------------------------------------------------

CrystalDiskMark 3.0.1 x64 (C) 2007-2010 hiyohiyo

Crystal Dew World : http://crystalmark.info/

-----------------------------------------------------------------------

* MB/s = 1,000,000 byte/s [SATA/300 = 300,000,000 byte/s]

Sequential Read : 63.650 MB/s

Test : 1000 MB [ 0.8% (0.1/9.4 GB)] (x5)

Date : 2011/07/14 14:15:04

OS : Windows 7 Ultimate Edition SP1 [6.1 Build 7601] (x64)





Third test, disable SRT and release cache, no other adjustments:



-----------------------------------------------------------------------

CrystalDiskMark 3.0.1 x64 (C) 2007-2010 hiyohiyo

Crystal Dew World : http://crystalmark.info/

-----------------------------------------------------------------------

* MB/s = 1,000,000 byte/s [SATA/300 = 300,000,000 byte/s]

Sequential Read : 181.509 MB/s

Test : 1000 MB [ 0.8% (0.1/9.4 GB)] (x5)

Date : 2011/07/14 14:31:01

OS : Windows 7 Ultimate Edition SP1 [6.1 Build 7601] (x64)


The following review shows that SRT improves the sequential read speed of a HDD (but isn't as fast as the same Kingston V+180 64GB alone):
http://www.techspot.com/review/395-asrock-z68-extreme4/page12.html

My setup, however, ends up being so slow that the HDD alone actually has faster sequential reads than the SRT mode.


Can more people please run a similar test on their machines? Just bench the SSD data partition with SRT enabled and compare to a bench of the same data partition with SRT disabled (and cache released). I want to have a larger dataset to see if I have a defective board or what.


Thanks in advance to anyone who can help study this and provide data!

I'm not set up on this rig yet to do that sort of benchmark. I'll see what I can find, though, tomorrow or when I have more time.

I'm just wondering if -- since you're running the bench-test software against a RAID configuration that is . . . anomalous . . . that it may not be giving the correct results. That is, maybe they're not real. It is also possible that the SSD is a model that doesn't work with it, but then, you DO have it set up for caching, don't you?

All I can tell so far: A game that I install on "new" machines I've also installed on this one. Usually, even with my 4-drive RAID5 w dedicated controller (256MB cache), there are delays between menus and screens as you drill down through the game. The sort of things you'd expect from a hard disk, even as the RAID5 is quite fast.

Second time I play the game and after rebooting (although that wouldn't make any difference by itself -- but figure all my volatile RAM is "cleared") -- it's as though the game is "in memory." No delays. No hesitation. As if . . . . it's in RAM.

 

[the_electrician]

Cheers guys got my rig configured.

i5 2500@4500Mhz
Asrock Z68 Extreme4
8Gb G.Skill Ripjaws X DDR3-1600
MSI 5850
Seasonic X 750
OCZ Vertex 2 60Gb
WD1600avvs-63l2b0
Dell 2209WA

Here is come screenshots and benchmarks

[URL=http://imageshack.us/photo/my-images/684/irst.png/] Uploaded with [/URL]ImageShack.us



Uploaded with ImageShack.us

 

[BonzaiDuck]

Not up to speed about using ImageShack.US, yet. This also prompted me to reinstall WS_FTP and fiddle with my web-pages, which I haven't done in a couple years. But haven't uploaded the screenies yet.

What screenies? I followed Electrician and bo_Jangles to download CrystalDiskMark and ATTO. I'll just summarize the results.

First, my RAID5 with a 3Ware 9650-SE controller "seems" fast enough, but the benchies show that it has reads maxing out at around 150 MB/s and writes consistently at about 225 MB/s. I can see how this new ISRT thing got me all enthused and animated . . .

Now -- the new Z68 system with ISRT. SATA-III 510-series SSD and a WD Caviar Black SATA-III HDD:

The CrystalDiskMark, for the "accelerated" HDD in "enhanced" mode (with slower writes), shows:

Sequential read: 244.7 MB/s
Sequential write: 95.47 Mb/s
512K Random reads: 210.00 MB/s
512K Random writes: 42.53 MB/s

In "Maximized" mode -- probably something we wouldn't do until we're 100% that our over-clock and everything else is rock-stable:

Sequential read: 211.0 MB/s
Sequential write: 190.4 MB/s
512K Random reads: 189.3
. . . .Random writes: 181.8


ATTO shows perfectly consistent for "Enhanced" mode with the CrystalDiskMark score. And I took a comparison ("enhanced" only) in ATTO of the SSD's formatted partition versus the "accelerated" HDD:

For all but the smaller transfer-sizes of .5K to 4K, write speeds are consistently between 212 and 220 MB/s, while read speeds excluding the two smallest transfer sizes exceed 300 MB/s with the 4 and 8KB reads exceeding 465 and 480 MB/s respectively.

So -- YES -- it might seem that for the expenditure, one could ALMOST lean toward using the SSD as a boot drive only.

But the proof of the pudding is how the ISRT "RAID0" disk acceleration trumps an HDD standalone performance. Easy enough to do: unhinge the SSD array, reboot, run the benchies on the HDD, and then rehinge the SSD to HDD acceleration. Haven't done that yet.

I still favor using the SSD this way, though, with two caveats. I'm only guessing that my SSD standalone does indeed live up to the "near 520 MB/s" read-speed in its spec, and we might assume without further info that the RAID configuration actually depresses its performance. But the improvement on the HDD makes it pretty much a winner (and we all KNOW that these SATA II or III HDD's can't much manage more than 100 or so -- with the WD Veloci-Raptor SATA-III spec'd at 145 MB/s "sustained" read spead).

But . . . . if the SSD partition in the RAID ISRT configuration shows speeds that are "close enough to the drive's top-end spec, why is the HDD acceleration limited to what our colleagues like electrician show with their SATA-II drives?

Conclusion: If you're gonna do this, you might save some ducats by getting small SATA-II SSD's and SATA-II HDD's to use in the configuration.

 

[bo_jangles]

Thanks for the feedback, gentlemen.

the_electrician appears to have nice results without any significant decrease in sequential read speeds.

BonzaiDuck seems to have a large hit in read speeds, but since he's starting with such a fast SSD, it's still an improvement over the HDD.

My hit on read speeds is so bad, and since I'm starting from a slower pure SSD speed to begin with, the read performance of a cached HDD is worse than without. Yes, random access of the HDD is superior, but why can't I have no penalty like the_electrician or a small penalty like piquadrat?

Is this a hardware fault? A software fault? I can't think of any configuration errors I've made.

Here's a set of ATTO benches showing my problem. Upper left is the SSD with SRT disabled. Reads are over 200MB/s. Lower left is SSD data partition with SRT enabled. Reads are terrible. Upper right is HDD with SRT disabled. Reads are above 100MB/s. Lower right is HDD with SRT enabled, and it matches the terrible reads well below 100MB/s.

 

[BonzaiDuck]

Well, Bo, those are the kinds of comparisons that are useful. In your case though, it poses a more "diagnostic dimension."

Earlier, somebody observed that the RAID configuration layer depressed performance across the board. According to another review-link I posted (using exclusively SATA-II hardware), the acceleration performance "came close" to the SSD standalone performance. It appeared that as much as 4/5 or even more of SSD standalone performance was preserved. It also pointed to "in some cases" a 400% improvement over the HDD standalone performance.

I'll have some more interesting sh . . stuff to report Monday or Tuesday. Tuesday. I found a bargain on a Veloci-Raptor SATA-III. Looking at the two SATA-III drives paired with the SATA-III SSD, it may uncover some relationships between the components that the review article I linked could have missed.

Again, though -- some people I know had greater expectations for the leap from SATA-II to III. Those hopes were dashed some time ago with benchmarks showing that any gains were less than stellar -- or not too significant. Someone else here noted that SATA-III posed benefits that will only be realized in the future, and so I'd be interested to hear any observations about what was meant by that, or what that means . . .

 

[MoMeanMugs]

I got my system set up on Saturday. I took my time and got everything configured using the SRT cache with a Kingston 128 GB SSD and Seagate 2TB drive. I didn't even get a whole day of use out of it, when I get this yesterday:

I needed to reboot the system because the audio was stuttering for some reason. When trying to boot back into W7 x64, the system bluescreened (rebooted before I could see the stop message). Now it's infinitely stuck at this. I left it running for over 17 hours, but it still sat at that screen. I can see the hard drive light going intermittently, so it seems like it's doing something, but who knows what. I tried doing a Google search, but couldn't find anything. If I have to reinstall Windows, so be it. Anyone have any ideas on what causes this or how to prevent it in the future?

System specs:

Asus Z-68-V Pro
2500k
16 GB 1600 Ripjaws
Kingston 128 GB SSD
Seagate 2TB
5870 Crossfire

 

[piquadrat]

I have the same mobo, processor and ripjaws but 8GB (9-9-9-24 @ 1600) and never experienced such sensations. I could risk the thesis that this is not SRT related.

 

[bo_jangles]

Yeah, your rebooting problem doesn't sound too related to SRT. I saw that screenshot once when I did a hard reset. It's normal.

For your BSOD, do the usual F8 and disable automatic reboot... look at stop code... google...

 

[MoMeanMugs]

I don't think over 15 hours sitting on that screen is normal! I can't get past that screen, so I think something got hosed. I am going to try to reinstall Windows sometime this week.

 

[bo_jangles]

Oh sorry, I misread your post. Yeah, not normal. Can you get to the Ctrl-I screen after a hard reset and then disable SRT? Then do F8, etc., if you still BSOD.