On all those I did, OS is on the same disk. I believe others are the same. I don't believe it affects the results much as the test SW is probably in RAM. It would affect small disks as the area used for testing would be far away from the FAT table needing long seeks. Of course CF, SSD has no seeks.

I think Transcend is running things in parallel inside and avoiding erase/writes. Erase takes a long time to do in flash.

The Transcend 32GB SSD from Newegg.com came in and it is the new version. I will run benchmark again on the X30 and see if I can come close to sugo's results in the T42.

Results of the Transcend 32GB SSD in the X30 turned out to be very comparable to the T42:

--------------------------------------------------
CrystalDiskMark 2.1 (C) 2007-2008 hiyohiyo
Crystal Dew World : http://crystalmark.info/
--------------------------------------------------

Sequential Read : 78.240 MB/s
Sequential Write : 45.288 MB/s
Random Read 512KB : 76.109 MB/s
Random Write 512KB : 30.840 MB/s
Random Read 4KB : 12.630 MB/s
Random Write 4KB : 1.934 MB/s

Test Size : 100 MB
Date : 2009/02/14 0:27:28

sjthinkpader, good to hear you got similar performance results. Thanks to this Transcend SSD, I think I can keep using my trusted T42 for another year.

I picked up one of those Transcend SSDs yesterday (they're now up to $90 at NewEgg), and I've installed it as the boot drive in my X31 (2672-cbu). Fresh install of the OS (it was a very fast install, btw), with XP SP3, AVG antivirus, and wireless connection (not sure if either of those make a performance difference). Running CrystalDiskMark 2.2 (didn't find 2.1 on the website). Results are:

Sequential Read: 76.04
Sequential Write: 45.08
Random Read (512kb): 74.00
Random Write (512kb): 31.12
Random Read (4kb): 12.81
Random Write (4kb): 1.928

-godling

Thanks for sharing your data. Looks close enough to other runs.

Here is data from a WD 160GB drive added:

Added test data from a Hitachi HTC426060G8CE00. This 60GB drive came from an iPod Classic. The test platform is my new X41t, 1866-5GU. Performance is very close to the Hitachi 1.8 inch 40GB drive tested earlier.

godling, would you mind sharing how you installed the OS? Did you create a service partition first or some other way? Also, how do you like the cards performance compared to a HDD after having it a few days?

Was thinking about buying one for an X31 here!

Nope, no service partition. I worked from an nLite version of XP I made. First, remove the internal wireless card because for some reason XP doesn't like to install fresh with a "strange" wireless card. Then install the drivers, and then re-install wireless card.

As for performance now compared to a HDD, it's still spectacular. My battery life is up around 4+ hours (from 3+), program access is close to instantaneous, and boot time with SP3 is under 30 seconds (that's with anti-virus, .net, etceteras ... it was under 20 before I installed those).

There is the occasional slow response that I haven't figured out -- and by slow, I mean it might take 2 seconds to access something -- but that might be because I'm reading up on "tweaks" to the OS to improve SSD performance, and not all of those tweaks really make the system perform better. Haven't nailed down what is and is not working for me, yet.

My suggestion: if you want a fast drive, get the SSD. That Transcend is very nice for $90. There are faster models coming out, like the Intel X-25M MLC. Intel is also supposed to be putting out an X-25E, which will be the SLC version of that SSD. Those are much pricier though, I think. I'm considering picking up a couple for my desktop, if I stumble on a spare grand somewhere.

-Jon

So you used nLite, have not tried that. Boy the performance boost of having an SSD sounds nice, but what I like more is not having to think about sudden moves damaging the HDD with a TP sitting on my lap! BTW, noticed Neweggs price went up again on the Transcend drive, it's now $93.

Hope you'll keep us updated on performance tweaks for the OS as well. Here's one you'll like:

It seems when XP installs it often does NOT properly configure for your particular L2 cache! Amazingly it sets the registry entry to ZERO!! To enable XP to take advantage of your L2 add this to your registry:

for BANIAS CPU's
;start copy below this line

Windows Registry Editor Version 5.00

[HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Session Manager\Memory Management]
"SecondLevelDataCache"=dword:00000400

;end copy above this line
This entry will set your L2 cache for XP to 1MB

---

for DOTHAN CPU's
;start copy below this line

Windows Registry Editor Version 5.00

[HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\Session Manager\Memory Management]
"SecondLevelDataCache"=dword:00000800

;end copy above this line
This entry will set your L2 cache for XP to 2MB

paste your selection into a new NOTEPAD file and save to desktop as:
L2-CACHE.reg

Double-click L2-cache.reg, accept, reboot, done.
Enjoy the faster speed!

sojourner
Wow!

Let me show You something, that is called "FlashPoint SSD accelerator":
http://forum.thinkpads.com/viewtopic.php?f=3&t=41568&p=502632#p502632

Please spreed this information to many low-end SSD users,
Thanks.

Some more benchmark data ...

Hitachi 5K500.B-500 (ThinkPad X200s):
Sequential Read : 51.009 MB/s
Sequential Write : 47.764 MB/s
Random Read 512KB : 25.178 MB/s
Random Write 512KB : 25.030 MB/s
Random Read 4KB : 0.350 MB/s
Random Write 4KB : 0.946 MB/s

OCZ 2GB 150x SD card (ThinkPad X200s, built-in SD card reader): ... just for fun!
Sequential Read : 16.358 MB/s
Sequential Write : 12.701 MB/s
Random Read 512KB : 16.705 MB/s
Random Write 512KB : 5.811 MB/s
Random Read 4KB : 3.047 MB/s
Random Write 4KB : 0.081 MB/s

I will run a benchmark on my T60's HTS722020 drive and start a new chart for SATA drives in may be the T6x forum.

How about the general forum? I think that would be the best home for it, really?

I tested the RiDATA 8GB 233x CF card running in my daughter's X40 in the past 6 months or so. It now has 6.6GB of data on it with 1.4GB of free space. It definitely feels slower but the computer is still very usable.

I split the growing table to two:

1.8 inch/CF size


2.5 inch size


In the 2.5 inch table, you can see the importance of free space and defragmentation on performance. If the disk is full, the head has to travel further from the FAT area to the track being written. So a full disk will always be slower. This should affect Flash/SSD less as they have no heads to move.

The new Samsung 40GB HS040HB ZIF HDD crapped out. And I past the 7 days warranty the vendor gave so I will add the HS122JC data when it gets here.

Apologize for the small tables as I tried to stay within the 50KB image size.

I received and tested the 5400 RPM Samsung ZIF drive HS122JC. It is indeed very fast.



I didn't have time to take apart the computer so it was tested in an Ultrabay tray in the Ultrabase X3.

It affects flash memory even more, because linear writes become more random, and especially slower. Incorporation of write buffer to SSD may significantly reduce random write penalty on fragmented volumes.

Please note - most newer SSDs are MLC ones, with hardware write buffer, but MLC itself has very long addressing time for random writes.

Addressing time for random reads is very short (it is major benefit of any flash when compared to spinning plater media), so fragmentation is not a problem in this case.

Spinning platers media have very similar time to access random reads and random writes, it is why standard programs for testing HDD's performance are not suitable for flash memory. And this is why CrystalDiskMark became so popular with SSD (flash in general) benchmarks.

Thank you for the suggestion. What you said is valid for sure.

Nevertheless write buffer is not a new technology. OS and drivers don't use it mainly due to reliability. If there are any interruption during write and content of buffers are lost, the disk may be corrupted.

This issue was widely discussed for many years in the industry.

The $76 RiDATA 32GB CF card from Newegg.com arrived today. Windows 7 RC was loaded on a X40 and CrystalDiskMark data collected. Here is the updated chart:



The sequential write performance improved a lot. The random write large and small file size are about the same as other CF cards. Generally web browsing seemed pretty fast. Playing DVD has some jumpiness to it. Here is the Windows Experience Index:

Processor Intel(R) Pentium(R) M processor 1200MHz 2.2
Memory (RAM) 1.25 GB 4.1
Graphics 1.0
Gaming graphics Not detected 1.0
Primary hard disk 22GB Free (30GB Total) 5.6

I've ordered a Samsung 4200RPM 80GB HDD (HS082HB) and a Sandisk 8GB SSD. I will report those data once they arrive.

I bought a Sandisk 8GB MLC SSD out of a HP Mini 1000. Performance is only average.



It is not a standard shape.



Here it is in a X41 getting tested:


That IDE-ZIF adapter has the 3.3V regulator bypass since X40/X41 are 3.3V already. The Sandisk SSD board was tested with a USB-ZIF adapter first to make sure it is working. It took a while to get it cloned from an 8GB RiDATA CF/IDE setup. Finally it worked in a A31p with old drive (8GB CF/IDE) in the right side bay and new drive (8GB SSD/USB) off a USB2.0 port. It only took 25 minutes to copy 6.5GB of data.

Next in line to be tested will be a Samsung HS082HB, 4200RPM drive. It would be nice to test a SLC SSD.

A Samsung HS082HB 80GB, 4200RPM, 8MB cache HDD from a Dell netbook arrived. This is a 5mm thick single platter drive. Areal density must be pretty high yielding pretty good performance results, close to the 5400rpm sibling HS122JC.

Got a Samsung 120GB HS120JB 2MB cache, 4200rpm drive. The smaller cache had reduced performance quite a bit, almost to the same level as the 60GB Hitachi. Slower than both the 5400rpm and 4200rpm 8MB cache Samsung drives. Although small file random read and write are on par with everything else.

A Sandisk 32GB SSD arrived (SSD UATA 5000 32GB). This is my first real SSD and cost just over $100 delivered. It was an almost new disk removed in an upgrade and left on the shelf a few years. Windows 7 was loaded and it is pretty fast, especially for the 4K random read/writes where it beat pretty much everything written on this X2x/X3x,X4x forum. The absence of seek time really shines thru. This is an older design and the sequential and random large file read/write time are a little slower than newer drives.


The Samsung HS120JB data was ran in a Ultrabay tray and didn't look right. So I reran the benchmark in the HDD slot and got better results. It still is a little slower than the single platter HS082HB. Since HS120HB has 60GB on each platter, the areal density is actually lower than the 80GB HS082HB, both being 4200rpm drives. The 5400rpm 60GB/platter HS122JC still has a slight edge. This side by side comparison points to the importance of areal density as related to performance.

BestBuy.com has the 320GB WD WD3200BEVE on sale for $69.99 so I placed order for one and elected to pick it up at the local store. Curiously the store has it on the shelf for $99.99.

This drive is almost identical to the 160GB version in every way except areal density. Because of that it gave the fastest HDD performance that I've seen in 2.5 inch size, close to the Transcend SSD for sequential. Understandably it loses out in random since the SSD has no heads to move.



As compared to the higher mechnical performance Hitachi 7K100, it moves the media 1/3 slower across the head but the 4x areal density more than made up for the slower rotational speed. The Hitachi has a faster track to track step rate and made up some different in the small file random read.

I attribute the higher write speeds to the improvements in the write channel, cache management and control electronics in the past few years.

Two more devices on the 1.8 inch size table:

1. A Toshiba MK8009GAH HDD. This is a 4200rpm, two platter drive with 40GB per platter. It comes with either 2MB or 8MB cache. Performance is right in between the Hitachi and the Samsung drives.

2. An 8GB intel SSD model Z-P230 that came from an ASUS netbook. The read performance is close to the Sandisk 8GB SSD but random write is slower.



Sorry about the eye chart. I have to find a way to make a bigger chart with small file size.

Stumbled on this thread via Google, registered just to thank you for the charts. Now I think I would be buying a Samsung HS122JC instead of a SSD drive for my X40.

On reducing the filesize of the chart, try pasting it into IrfanView, click Decrease Color Depth in the Image menu, decrease it to 2 Colors (black/white). Then save it as a GIF.

Welcome to the forum.

My newest HS122JC was produced in April, 2009 so it is still in production. I just fried a perfectly good unit by putting it in a 5V Ultrabay tray with a no-regulator adapter. It worked for a little while, got really hot then stopped working. So be careful working with these 3.3V drives.

I was in Hong Kong two weeks ago. Followed a HK member's tip posted here, went to the Sham Shui Po computer market and got a few things. Not quite as big as Akihabara in Tokyo or Guan Hwa market in Taipei but pretty specialized. Seem like a place for professional electricians.

I will try your tip for a smaller file size.

An 8GB Centon 200x CF card went on sale this past weekend and I got one. As expected, it is faster than the Kingston 133x but slower than the RiDATA 233x. Installing WindowsXP took significantly longer.



I did find a better way to post these charts by a bmp/jpg printer driver called Xan Image Printer. It creates a direct .bmp and I used XnView to automatically crop the blank edges.

I got an ADATA Class 10 SD card. The adapter is this type:
http://cgi.ebay.com/ws/eBayISAPI.dll?Vi ... 0338295434

The trial and error went on for several weeks without success. Every time the card came up as Removable and refused to boot. This adapter also only work on 5V so the trials in X40's 3.3V slot were complete failures. Finally I came across this web page after reading thru several forum discussions:
http://www.vampiric.us/articles/Hardwar ... hTools.asp

The Bootprep was the utility that allowed me to make a MBR on the SD card that enabled it to appeared as Fixed device. Then I proceeded to load WinXPP and run XtalDiskMark.



The ADATA spec calls for 10MB write and 20MB read speed. The result didn't come close to that. I am not sure whether this is due to the card being slow or the adapter is the bottle neck. Feel in real use is actually not bad.