SanDisk Extreme III MS Pro-HG Duo (for Readyboost)

I've always been interested in Readyboost... It seems really clever to use flash memory for what it's good at, random data access. But usually flash drives and SD cards are very slow, like 3MB/sec.

Well I discovered that Sandisk makes a Memory Stick rated for 30MB/sec. I always thought Sony was foolish for creating another proprietary standard (a la Betamax), but in this case it seems to be the best for raw speed, and good for me since I can dedicate my VAIO CW's MS slot to Readyboost. I figured for $29 I might as well buy one and do some experimentation


So far I've done an HD Tune benchmark, but I can't test write speed because Windows doesn't allow you to delete a partition on a MS (HD Tune requires NO partition to exist). If anyone knows how to do this, let me know. Right now it's formatted as FAT32, but I don't know how that influences HD Tune.

HD Tune read benchmark using 64KB blocks (I think the difference at the beginning is because of the 600mb of data on the stick):


512KB blocks (why the fluctuation?):


32KB blocks:


Rudimentary write speed test, copying file from laptop hard drive:


Random access benchmark (note the high access times for bigger files):


Benchmark comparing random reads to sequential reads:



Here's what I want to know:
Why the differences in access times and read rates for different sized blocks?
How do the access times affect Readyboost?
Does this stick have wear levelling like an SSD, which would allow it to be used for a swap file?

 

BTW, I posted this here because Sony laptops have MS slots

 

You should try this, http://www.online-tech-tips.com/computer-tips/format-usb-ntfs/, to format the MS to NTFS.

+rep if it works :]

 

I am using a Sony Memory Stick Pro Duo Mark2 8GB Flash Card and I am using Win 7 64B. I am only getting 5 MB/s in my Vaio Z4. Do I have to install something or is the card the limiting factor?

Update:

Got it - the cache was disabled. Now I am getting 50-60 MB/s

 

I formatted to NTFS and ran one of the tests again, no difference. I think HD Tune accesses the drive directly rather than using the file system.

To test write speed, it requires no partition to exist. Removing the drive letter doesn't work.

That article says that NTFS requires more writing to the drive, so I guess it's not good for flash because it will cause premature wear.

 

5MB/s is probably the right speed for that stick. This is a special stick meant for cameras etc.

 

BTW, in Windows 7 a memory stick does work for Readyboost. So does an SD card.

 

Definitely works on Vista. I use old, otherwise unused cards from mobile phone, what is definitely noticeable is the speed any application opens, if it was open at least once takes 1/4 of the org time. That is already making ReadyBoost worth using for me. Before had to keep editors, audio or Photoshop open , now can close them and they open back in seconds.

 

That's pretty impressive. Have you compared that to what happens without RB, and just the RAM caching?

 

Yes, rather obvious when few times forgot to put the card back after using the slot for some data transfers.

 

Because I have XP I am using an aftermarket product called eboostr. The program lists the random read speed for my 2gig older Sandisk extreme III sd card at 10727 KB/s I have the same model 2gig sd card in a sd to usb adapter and the random read speed is 16323 KB/s I have both cards formatted to fat 32. I have tried toggling between fat 32 and NTFS and there is no difference in the random read speeds. The software site states if using over 4gig card to format to NTFS. This program also uses any unused ram to speed up other programs. I am limited to using 4 devises and the program will only use 4 gig of each devise. There is a very large difference in the speed that programs load.

 

Is there some way to track how long programs take to open, preferably collecting data over a period of time?

 

winkosmosis: The ready Boost shows more of a difference with the slower running HDDs that are running on laptops The USB ports allow higher speeds than the sd. I

 

I do not know about the memory stick speeds. as for actual time differences I do not know an accurate way to measure it. I tested the program I am using for over a month and was impressed enough with the speed improvement that I purchased it. If you are good with an electronic timer or stop watch program you could disable the readyboost and restart it and check the time. You are going to notice the difference more on larger programs. It also takes time for readyboost to gather the info as you first open an close different programs.

 

Looks like Readyboost is much better in Win 7 than it was in Vista http://www.connectedinternet.co.uk/2009/09/03/windows-7-readyboost-really-boosts/

 

ReadyBoost helps, no matter how much RAM you have. It works as a compressed backing store for your swap, and because of the much faster access times of NAND storage, paging back in from swap becomes much faster.
This benefits those who are low on memory more, simply because they will page more. However, no matter how much RAM you have, there is some benefit, because of how Windows uses the paging file actively to free up additional memory -- even with tonnes of free memory, more free memory can be used for disk caching, which increases speed.

That said, there is one situation where you should disable ReadyBoost, and that's if you defragment with a Windows certified defragmenter. The problem is that Windows' routines for defragmenting go through swap, so every block moved gets copied to the ReadyBoost device too, which slows down the defragmentation quite severely. I'm not sure if this has been fixed with Windows 7, but I doubt it -- the benefits of using swap as an intermediate when defragmenting is that (a) the defrag can safely be aborted at any time without ending up with a broken drive, and (b) you don't require large amounts of free space on the drive being defragmented.

Someone with W7 and ReadyBoost can possibly confirm that it's still a problem? Start a defrag and watch the blinkenlights.