Using a compact flash SSD-like disk in a laptop

Installation

AData CF Cf To Ide44 adapter CF 

Small video of the PC booting

Performances

A few HD-Tach runs: average 43mb/s, peak 48.5 mb/s, lower 37mb/s, access time 0.6ms.
CF  CF  CF  CF 

A priori comparison with a Raptor, a standard 7200rpm disk and a 2.5" disk

Using the following parameters (I took the max speed for the hard disks, but their speed would actually drop a bit for files not at the beginning of the disks):
DiskAccess time (ms)Speed (mb/s)
Raptor8110
HD1575
A very fast 2,5" HD1545
SSD (my CF)0,645
Theoric speed
I put 3 times the Raptor to show the effects of fragmented files on classic harddrives (SSD/CF do not suffer from that).
We see that SSD/CF will be faster than a raptor for files smaller than 0.6mb, and it will be faster than any 7200rpm disk for files under 1.7mb. The average file size in my Windows system32 directory is 350kb.
For fragmented files, add the number of fragment times the access time to each of the hard drive, but not to the SSD, making the SSD or CF even faster.

Lifetime of such CF disks

French version here
In normal use on a PC, one does not notice the swap, therefore one must be at maximum around ten access to the swap a second (with an access time of 12ms, one is limited to 83 accesses a second in max use, but at this stage that would row already severely), and still, only in phase of intensive use and multitasking, not at rest, where there is almost no activity on the swap (is enough to supervise the diode of disc activity to realize that at rest, it is not very often active...).

Then, the files swap, it's usually arount 1 Go, and the pages make only 4 KB, that is to say a little more than 260.000 pages. By making 10 accesses a second, one thus reaches each page approximately once every 26.000 seconds, that is to say once every 7 hours.

To reach 100.000 accesses in reading one swap page, one thus needs 79 years... By counting only the accesses in writing, one must at least be able to multiply that by two or three, because the very vast majority of the data in swap relate to storage areas containing of the static data (binary, resources...), since the data evolving/moving much are firstly stored in RAM (the mechanism of pagination most rarely discharges towards the swap the reached zones and gives priority to those which were not modified since their last loading, for precisely limiting time wasted to write in the swap).

Still add to that the mechanisms of wear leveling integrated to the SSD, and you must be able to multiply that once again by two or three… Morals of the history, you will have died well before your SSD…

And also:

Don Barnetson, Samsung: (French version)
Don Barnetson puts an end to the various allegations covering the lifespan of a hard disk flash by announcing that this type of hard disk had one average lifespan (MTBF) from one to two million hour, that is to say better than of the professional hard disks intended for the servers.

The first storage cells held hardly the 1.000 cycles of obliteration/writing whereas today the cells memory SLC (Individual Cell Level) hold approximately 100.000 cycles. By a simple calculation, one shows while taking as bases a use of 120 written Mo per hour on the hard disk, it would take 267 hours to entirely fill a disc SSD with 32 Go. The disc would become unusable at the end of 1.500 years (Via calculation, one arrives at 3.000 years but Samsung may take in account the time of obliteration or a use of 12h/24). If one supposes a card coult be writtent at up to 1 Go per hour, the disc has a lifespan of 366 years, always with cells SLC. By pushing the reasoning with its paroxism, ie 32 Go/hour, it will thus take 100.000 hours to reach the limit what corresponds to more than 11 years...

Links