[LINK] ASUS Eee PC, $499 in Myer for Christmas

steve jenkin sjenkin at canb.auug.org.au
Wed Dec 19 13:09:48 AEDT 2007


Scott Howard wrote on 19/12/07 12:16 PM:

> I haven't seen any performance stats for the Flash in the Eee, but
> for the most part current generation SSD drives are generally faster
> than their HDD equivalents - especially if you're talking about
> sub-notebooks, where vendors are generally using small/slower hard
> disks (eg, 1.8" 4200RPM)
>
> A quick search on Google finds no end of hits such as
> http://www.ultramobilelife.com/index.php?option=com_content&task=view&id=607
>
>
>
>
>
> For some of the Flash drives write performance generally does lag
> behind HDD, although that gap is narrowing quickly.
>
>
> Scott.

Below is some (simple) testing I did in June this year on USB flash
read/write speeds.
NAND flash memory (as in USB, CF and SD card) have asymmetric read/write
speeds.
Writes are internally 'page-based', and there there is an on-going
logical-to-physical page mapping.
['load levelling' I believe].  This means the notorious FAT file system
can't wear out 'track/block 0'.
Yes, Virginia, flash devices have 'limited' number of rewrite cycles. 1M
or more. With load-levelling, close to infinite life.

NOR flash, as used in the BIOS, are smaller, much more expensive and
write 1-byte at a time.

All EEPROM ('flash') requires an "erase" cycle before writing.

If you want 'performance', you can buy "Extreme III" CF cards used in
high-end digital cameras.
e.g. <http://www.dpreview.com/news/0602/06022614sandiskextreme.asp>

> SanDisk Extreme III CompactFlash and SD cards have minimum write and
> read speeds of 20 megabytes per second (MB/sec.)

I've never tracked down the specs for those expensive SSD's
[solid-state-disks]. Out of my league :-)


Elsewhere, in email on CLUG list, I saw someone quoting read/Write for
flash of: 40Mb/sec, 10Mb/sec. [no idea of blocksize]

Effective HDD r/w speeds vary all over the place. Flash wins on 'seek'
times :-)
IDE/SATA 2.5" are 5400rpm, ATA/SATA 3.5" are 7200 and SCSI 3.5" can be
10k or 15k [but small] - and a lot of FibreChannel [1Gbps or 2Gbps] is
used there.

Add cache effects, controller cards, cabling to the raw interconnect
speed and if you're running RAID in a NAS... - it's complex :-)
As your disks age, bad-block mapping makes some read/write cycles very
poor. Retries are problematic as well.

The current 'zoning' of disk drives  - constant lineal density like
optical drives, not constant sectors/track - means asymmetric access
speeds depending on block location.  In the center == fast, at the outer
edge == slow.
As your disk fills up, your system runs noticeably slower.


File System layout & caching have a very large effect on top of that.

And soon, we are going to have to worry about the raw BER (Bit Error
Rate on all transfers) of drives - it hasn't changed from 10^13 in 2-3
decades. It's measured at the disk: add cabling, controllers, memory,
backplane errors and "it's a problem".

HTH
sj

Testing on my PPC mini-mac using 4Gb USB.

> Seems standard devices are very sensitive to block size on *write* -
> but not on read...
>
> For some good reference material: Wikipedia and Kingston:-
>
> <http://www.kingston.com/products/pdf_files/FlashMemGuide.pdf>
> <http://en.wikipedia.org/wiki/Flash_Memory>
>
>
> Device used for these tests: 4Gb Kingston 'Data Traveller'. 512Mb
> file used for most timings (1Mb for 1b test).  No FileSystem Cache
> issues.

> USB Flash - WRITE performance

> 1b    88,827 bytes/sec

> 8k 1,035,601 bytes/sec

> 128k 3,819,827 bytes/sec

> 512k 4,155,635 bytes/sec



> USB Flash - READ performance

> 8k 9,467,708 bytes/sec

> 16k 9,448,265 bytes/sec

> 32k 9,386,199 bytes/sec

> 64k 9,475,763 bytes/sec

> 128k 9,466,469 bytes/sec

> 256k 9,453,077 bytes/sec

> 512k 9,415,421 bytes/sec

> 1024k 9,482,868 bytes/sec



-- 
Steve Jenkin, Info Tech, Systems and Design Specialist.
0412 786 915 (+61 412 786 915)
PO Box 48, Kippax ACT 2615, AUSTRALIA

sjenkin at canb.auug.org.au http://members.tip.net.au/~sjenkin




More information about the Link mailing list