SSDs - State of the Product?

[Mercutio]

Huh? Isn't having a fast page file and temp files one of the main reasons for having an SSD?

I'd say it's much more important to have fast access to operating system files that are in constant use than whatever userland data. As far as the pagefile, Windows wants there to be one, and constantly shifts 4kB blocks in and out of it even on systems with tremendous amounts of RAM (8GB, in this case). I was hoping that not having that constant disk activity might contribute to longer drive life at the expense of some small amount of performance (and probably not much, given how that machine was used).

 

[Santilli]

This one: http://www.ocia.net/reviews/kingspecssd

I am not noticing any stuttering. I am seeing about 78MB/s reads with it. Not bad given the interface.

That's what I've been looking for. Going to give it a try in the Panasonic CF-51.

I just checked, and I can get my install probably down around 20 MB, with programs.

I suspect that will be enough of a jump to let me wait until they really get the bugs sorted out on the i7 laptops, or, even longer...

http://www.memoryc.com/products/des...PATA_IDE_SSD_Solid_State_Disk-MLC_/index.html

 

[Chewy509]

Anandtech has a review on the new Corsair Force SSD (uses a SandForce SF-1200 controller).

http://www.anandtech.com/show/3656/corsairs-force-ssd-reviewed-sf1200-is-very-good/1

In short - very fast, not so cheap.

 

[Pradeep]

Good article showing the difference between the SF-1200 and SF-1500, at least on paper. Interesting thing is in the event of one of these drives getting degraded, the reads will be impacted, and not writes.

 

[Pradeep]

I'm curious why Anand didn't add the 128GB version of the C300 at $419 into that table.

 

[Pradeep]

http://www.tcmagazine.com/comments.php?shownews=33772&catid=2

"While all focus is on SandForce-based drives, OCZ has decided to also introduce an updated version of its Colossus 3.5-inch SSD. Known as Colossus LT, the new drive have 34nm MLC NAND Flash memory chips, two (Indilinx) controls, internal RAID, 128MB of cache, a SATA 3.0 Gbps interface, a MTBF of 1 million hours, and offer maximum read/write speeds of 260 MB/s (15,000 IOPS on 4K random reading/writing)."

When I first saw this I thought of Coug with a pillowcase full of Micropolis heavy iron. Or was that Merc?

 

[ddrueding]

http://www.tcmagazine.com/comments.php?shownews=33772&catid=2

"While all focus is on SandForce-based drives, OCZ has decided to also introduce an updated version of its Colossus 3.5-inch SSD. Known as Colossus LT, the new drive have 34nm MLC NAND Flash memory chips, two (Indilinx) controls, internal RAID, 128MB of cache, a SATA 3.0 Gbps interface, a MTBF of 1 million hours, and offer maximum read/write speeds of 260 MB/s (15,000 IOPS on 4K random reading/writing)."

When I first saw this I thought of Coug with a pillowcase full of Micropolis heavy iron. Or was that Merc?

And at the moment, the "LT" is cheaper.

 

[Bozo]

I haven't read this yet but I thought it migh be of interest to some of you.

SSD comparison

 

[Handruin]

I haven't been following SSD drives that closely, but I was really surprised to see Plextor has an SSD now!?

 

[Pradeep]

Based on the review Bozo linked to it appears to be one to avoid.

 

[Handruin]

I noticed that also. My comment was just in surprise, I never really expected Plextor to enter the SSD market. Competition is a good thing, so maybe their 2nd gen will be better.

 

[CougTek]

When I first saw this I thought of Coug with a pillowcase full of Micropolis heavy iron. Or was that Merc?

The Micropolis pillow was Merc's idea, not mine. The subtle sledgehammer is my trademark.

 

[LunarMist]

The last Plextor I had was a SCSI CD-ROM that sounded like a Hoover. I hope their SSDs are quiet.

 

[LunarMist]

When I first saw this I thought of Coug with a pillowcase full of Micropolis heavy iron. Or was that Merc?

Micropolis? Wow, that has been a few years now. In the 1980s I recall them being around the bend on the north side of Nordhoff. I suppose they don't exist at all now.

 

[udaman]

bootable PCI-e 2TB = $8.8k (Amazon has it for $10.2K!)

http://www.xpcgear.com/ocz-oczssdpx-zd2p882t.html

1.4GB/s R/W...slower on IOPS, however.

 

[Mercutio]

The sack full of Micropolis drives were to be vigorously applied to Prof. Wizard.

 

[LunarMist]

The sack full of Micropolis drives were to be vigorously applied to Prof. Wizard.

Was that at SR or here?

 

[LunarMist]

Intel X-25 prices seem to be slowly falling, or I am not remembering clearly?

 

[timwhit]

Was that at SR or here?

Both.

 

[Santilli]

Kingspec got here. Ordered SATA by accident.

Not bad for 100 bucks, with shipping.

Looks like I have a new scratch/pagefile disk...

 

[LunarMist]

.

Looks like I have a new scratch/pagefile disk...

Maybe so, maybe not. STR is not the best indicator of scratch disk performance.

 

[Santilli]

Maybe so, maybe not. STR is not the best indicator of scratch disk performance.

It's an SSD. With that kind of transfer rate, what do you think the access time is going to be?
.o1????

What I should really do is create a Ramdisk, and put the pagefile on that, but, IIRC, the Ramdisk is created after the system is booted, and, therefore the OS uses an existing source, prior to boot.

 

[timwhit]

It's an SSD. With that kind of transfer rate, what do you think the access time is going to be?
.o1????

What I should really do is create a Ramdisk, and put the pagefile on that, but, IIRC, the Ramdisk is created after the system is booted, and, therefore the OS uses an existing source, prior to boot.

None of that matters. You need to be concerned with random read and random write performance, which can't be accurately estimated by looking at transfer rate and access time.

 

[Handruin]

Greg, access time is not an exact correlation to the measured STR. You can have 400MB/s STR with horrible access times of 20ms+. The ATTO benchmark will show 400MB/sec in sustained transfer, but the second you switch to a random access, the STR will drop. This is why Intel's I/O meter can be a very good tool to benchmark specific read/write ratios with different levels of random access. If you can determine the pattern that suites your needs, you can then begin to look for a drive or storage subsystem that can perform the best for you. A specific example would be something like Oracle and/or SQL Server. Based on your application needs, STR might not be at all useful, but fast 64 KB reads and writes might make a huge difference.

If your file access and usage patterns via application or direct file access are 50-90% random vs 1-2% random, you'll have better performance when there is less random access required. Any time the drive has to randomly access data, there will be a wait involved.

 

[Santilli]

Greg, access time is not an exact correlation to the measured STR. You can have 400MB/s STR with horrible access times of 20ms+. The ATTO benchmark will show 400MB/sec in sustained transfer, but the second you switch to a random access, the STR will drop. This is why Intel's I/O meter can be a very good tool to benchmark specific read/write ratios with different levels of random access. If you can determine the pattern that suites your needs, you can then begin to look for a drive or storage subsystem that can perform the best for you. A specific example would be something like Oracle and/or SQL Server. Based on your application needs, STR might not be at all useful, but fast 64 KB reads and writes might make a huge difference.

If your file access and usage patterns via application or direct file access are 50-90% random vs 1-2% random, you'll have better performance when there is less random access required. Any time the drive has to randomly access data, there will be a wait involved.

OK:
Using HD Tach these are the relevant facts:

2 X-25M 2's, in raid 0
435 MB/sec burst
280 AVG MB/sec reads
Access time is 0.1 MS

Compare with 3 Vertex Turbos:
131 MB/sec read avg
224 MB/sec burst
0.2 ms access time (100% slower;-0)
and the

Kingspec:

157 Mb/sec AVG
0.2 ms access time.

Using the Kingspec for a part of the pagefile eases a bit of wear on the X-25's, and, I have not noticed the 100% increase in pagefile access time...

 

[LunarMist]

I'm sure that will help. :roll:

 

[Santilli]

iIt's better then having it sit around and do nothing...

 

[Pradeep]

http://www.semiaccurate.com/2010/05/03/sandforce-ssds-break-tpc-c-records/

IBM using a bunch of SF-1500 based drives to deliver world class performance.

 

[LunarMist]

http://www.semiaccurate.com/2010/05/03/sandforce-ssds-break-tpc-c-records/

IBM using a bunch of SF-1500 based drives to deliver world class performance.

Maybe if she wins the lottery...

 

[Santilli]

Got the KingSpec SSD Professional, PATA, thanks to Stinker's suggestion.

Installed XP, on Panasonic CF-51 no problems.
ATTO scores are really low, around 31 mb/sec, vs. 45 seconds for the 7200 rpm seagate it replaced.

Boots to login in 21 seconds, vs about 35 seconds for clean install on 7200.

Generally seems about twice as fast at normal browsing, with clean installs of XP SP3.

Seems a lot of these programs hit the hard drive, as does XP, and, when the paging is
to a SSD, there is a huge jump in performance.
Shut down is well under 10 seconds.

MSFT Word 2000 is about 2 seconds to working.

From login to working, that is connected to the net, takes about 35 seconds. This used to take about a minute and a half on the old install on the 7200 Seagate.

No stuttering. Makes this old laptop useable. Currently have used 10 gigs of 29 gigs on the drive.

So far, this appears right up with the large amount of ram for money well spent on speeding up a laptop.

Stinker, thank you for the suggestion.

It did take nearly a month to get here from China, and, I sure wouldn't try an RMA...

Have to wonder why there aren't a LOT more folks trying to get part of this huge market.

 

[Stinker]

I'm still very happy with the Kingspec. I could not believe how fast I got mine....less than 1 week through an eBay seller from China via US Mail. I've had domestic packages take longer than that! I'm sure there's a lot of variability with it. Maybe yours took a long time because of some lingering effects of the volcano ash cloud, Greg...? Nonetheless, glad it's working out for you! :salut:

 

[Santilli]

It is and it isn't. I'm not jazzed about the false speed statements, and kind of feel like they shipped me an inferior product.

A real surprise after their SATA version being exactly the speed they claim.

G

 

[Stinker]

Here's my benchmark results if it helps you trouble shoot anything. Never did an attachment before so let's see how this works...

 

[Santilli]

Mines half that, or less. Runs about 30 MB/sec on both ATTO, and Hard Tach.

 

[ddrueding]

Greg, I would test it on another system as well, might be a chipset issue.

 

[Santilli]

HMMM. It is SUCH a pain to pull the drive out of the caddie, and plug it into another machine. I guess the Athlon 3000 would have an ide port on the motherboard...

For that matter, I think the server does as well...

 

[Santilli]

Well, I pulled it, but the pins are closer together then the IDE stuff on my desktops, and won't fit.

Checked the model Number:
KSD PA25.1-032MJ, and it's the same as the one in Stinkers link.

 

[ddrueding]

Yup, 44-pin laptop IDE connection (includes power as well). Doesn't really matter either way, that is the machine it is destined for, and that is as fast as it goes there.

 

[LunarMist]

Greg, you just discovered the 44-pin notebook drive? I hope you don't have to work with 1.8" drives with a half a dozen different connectors. :bravo:

 

[Santilli]

I try and avoid working on laptops. The Panasonic is WAY faster with this drive in place. Mavis Beach 15, my typing program, is like a totally different experience then the 7200 drive.(just went 80 words a minute, no errors)

I really wonder what doubling the drive speed would accomplish on this laptop.

For my uses, this CF-51 is now useable for another couple years, at least, providing nothing else takes a dive.

Somethings are a bit bothersome. The drive caddies are about 100 bucks, or more, which is about what a used CF-51 is going for on ebay.

Speakers blew out, and crack, and need replacing. Again, I should buy another computer for parts, but, spending 100 dollars on an old laptop, with some of the deals I see out there makes no sense.

I'm waiting for the i7's to come down in price, and SATA 2 or 3 to be included.

If this drive makes such a difference at 30 MB/sec, I'd like to see a 8-10 times faster processor, with a SSD really doing 255 MB/sec or more.