Storage card speed optimization – long-time & real-world tests. A MUST!

[Guest Menneisyys]

After having downloaded the latest Shadow of Legend beta, I’ve realized it has 9422 files and 424 directories. If you’ve followed my articles for more than a year, you may already know that this means really slow transfer rates to even the fastest cards. (For example, to a (cheap) 1 GByte Kingston SD card, formatted with the most storage-saving FAT32/512 format (and not using a backup FAT table to speed up operation), it took exactly 2:28 (h:mm) to transfer all these files. (Only to find out 14M was left free, which resulted in the immediate crashing of the updater program because there simply was no free storage to download updates to. The SoL web page should be updated and the 1+ GB storage requirement increased to at least 2GB so that others don't even try to copy it to an 1GB card.) If you’ve read my three-year-old, highly recommended article on optimizing storage card speeds, you may already know you can heavily decrease the time needed to transfer several thousand files to a card by using another file system.

Note that I will NOT explain all the stuff I’ve already elaborated on in the above-linked article once more. Read it so that you understand what I’m writing about, what tools you should use to format your cards with etc. This test is pretty much like the previous one, except that, of course, I’ve tested the latest high-capacity microSD cards to find out how they compare to each other. For real-world tests, I’ve benchmarked transferring 48 files residing in 18 directories to these cars. In addition, I’ve, instead of some meaningless desktop file transfer speed testings, I’ve run some real-world tests using my Canon IXUS 960IS (SD 950IS) 12 Mpixel camera in superfine mode, using continuous shooting. I use continuous shooting a lot when taking social photos so that I have several shots to select the best from; then, card writing speeds have a tremendous effect on the continuity and speed of shooting. In this respect, this test will be of real importance to people looking for the fastest possible card to be used in their (high-end – don’t forget low(er)-end cameras like that of HP simply can’t make use of the high speed of cards) digital cameras.

Symbian and BlackBerry users: note that this article is applicable to your operating systems too. It’s just that you won’t have access to the Windows Mobile apps to format your cards. However, you can do the same on your desktop. And, of course, you may also face the problem of having to copy to thousands of files to your card – or, for that matter, optimizing it for speed when used in a digital camera (or a desktop card reader).

(Note that I, generally, only take shots in fine mode only; then, the cards’ transfer speed doesn’t have that big an impact on the speed of taking photos in continuous mode as in Super Fine mode. The latter uses roughly twice the storage for shots than Fine mode.)

The cards I’ve tested (click the image for a higher-resolution one)

Now, let’s take a look at the results. In the first part, I only elaborate on the file creation speed. You’ll want to check this section out if you want to optimize your cards for file creation - but not necessarily massive, multi-Gigabyte file transfers; see the second section, the real-world camera tests, for that. Note that optimizing for file creation speed doesn’t necessarily result in being optimized for massive file transfers, as will also be shown in the camera tests.

Section 1 – optimizing strictly for file creation speed

Lower-capacity SD cards

2GB Connect3D:

FAT32 / 512 / -: 32s

FAT32 / 512 / +: 45s

FAT32 / 2k / +: 40s

FAT32 / 4k / +: 40s

FAT32 / 16k / +: 43s

FAT16 / 32k / +: 25s

FAT16 / 64k / +: 26s

FAT16 / 32k / -: 26s

FAT32 / 4k / -: 34s

Recommended: FAT16 / 32k for file creation speed (with or without backup); otherwise, FAT32 / 512 / - (12.5% slower).

1GB Kingston:

FAT32 / 512 / -: 23s

FAT32 / 512 / +: 32s

FAT16 / 16k / +: 9s

FAT16 / 32k / +: 16s

FAT16 / 32k / -: 9s

FAT16 / 16k / -: 8s

FAT32 / 4k / -: 20s

Recommended: FAT16 / 16k / - for speed; otherwise, FAT32 / 512/ - (280% slower)

1GB Sandisk:

FAT32 / 512 / -: 9s

FAT32 / 512 / +: 22s

FAT32 / 4k / +: 22s

FAT16 / 16k / +: 19s

FAT16 / 16k / -: 7s

Recommended: FAT16 / 16k / - for speed; otherwise, FAT32 / 512/ - (28% slower)

High(er)-capacity, higher-speed microSD cards:

SanDisk 2GB, no class given:

FAT32 / 512 / -: 4s

FAT32 / 512 / +: 11s

FAT32 / 4k / -: 5s

FAT16 / 32k / -: 5s

FAT16 / 32k / +: 10s

Recommended: FAT32 / 512 / - for both speed and capacity; backup FAT introduces at least a 100% speed hit

SanDisk 8G SDHC class 4

FAT32/32k/+: 4s

FAT32/32k/-: 4s

FAT32/4k/-: 4s

FAT32/512/-: 4s

Kingston 4G SDHC class 4

FAT32/4k/+: 24s

FAT32/4k/-: 11s

FAT32/512/-: 22s

FAT32/512/+: 29s

Optimal: FAT32/4k/-; backup FAT introduces 100% speed hit; 512byte is slow

Sandisk 256M miniSD:

FAT16 / 4k / -: 12s

FAT16 / 4k / +: 13s

FAT16 / 16k / -: 12s

FAT32 / 512 / -: 11s

FAT32 / 512 / +: 12s

Optimal: FAT32/512/-; other settings aren’t considerably worse, though

Section 2 - Camera tests

Shooting continuous images (the screen of an LCD monitor; this guarantees the output will consist of huge JPG files) for 2 minutes. I’ve tested the cards with the parameters optimized for file creation speed and for FAT32/32k/+ (the default mode the IXUS960 formats all cards to). The former is before and the latter is after the slash. I’ve also marked how many shots were taken in two minutes and how much storage they took. As an example, let’s take a look at the first row,

1GB Sandisk 82 shots (FAT16 / 16k / -) for 651,106k / 82 shots (default) for 669,843k

just below. The section before the slash,

1GB Sandisk 82 shots (FAT16 / 16k / -) for 651,106k

means the Canon took 82 shots, totaling 651106k, when using the FAT16 / 16k / - file system with the 1GB Sandisk SD card. Continuing with the part after the slash (/),

82 shots (default) for 669,843k

states (also) 82 shots have been taken with the default (Canon) FAT32/32k/+ file format, totaling, this time, 669843kbytes. (Note that we’re speaking of photos; this is why they aren’t of exactly the same size.)

Now, the results (SD and microSD; I haven't benchmarked the miniSD card in this test):

SD:

1GB Sandisk: 82 shots (FAT16 / 16k / -) for 651,106k / 82 shots (default) for 669,843k

1GB Kingston: 82 (FAT16 / 16k / -) for 630,179k / 79 (def) for 657,524k

2GB Connect3D: 51 (FAT16 / 32k / -) for 421,809k / 47 (def) for 415,823k

microSD:

Sandisk 2GB: 89 (FAT32 / 512 / - ; about 1-3s warm-up time) for 518.584k / 88 for 624,291k

Kingston 4GByte Class 4 SDHC: 128 (FAT32/4k/-) for 781,932k / 98 (FAT32/32k/+) for 703,071k. That is, in this case, the camera default is somewhat suboptimal.

Sandisk 8GB Class 4 SDHC: 95 (FAT32/512/-; takes 16 secs to check the card on every restart!!) for 797.178k / 142 (FAT32/32k/+) for 1,051,712k

As a rule of thumb, in most cases, the default file system is optimal when used with the Canon. (With for example the Kingston 4Gbyte Class 4 microSDHC, I’ve got somewhat better results in the other way around. It’s still about 34% worse than the Sandisk 8GB Class 4 results in the default mode, which is, incidentally, way better than the performance delivered by all the other, tested cards.) Of course, for pure file creation speed, you’ll want to look at the results in the first section and format your card(s) accordingly. Don’t forget the Canon, as has already been pointed out, uses a backup FAT, which results in, with some cards, even two or even three times worse performance when copying a large number of small files than without the backup FAT. Keep this in mind when transferring for example Shadow of Legend to your card.

Finally, note that if you use the, otherwise, most storage saving FAT32/ 512byte combo with really high-capacity cards, you may encounter slowdowns upon powering up the devices using the cards. For example, the Canon camera needs approximately 16 seconds (!!!) to gain access to the Sandisk 8GB microSDHC when formatted to this mode. With the 2GB Sandisk microSD, this warm-up time was about 1-3s secs. You’ll encounter the same issue with desktop card readers (you’ll need to wait that more(!) to gain access to the contents of the card) and, probably, even handhelds / handsets (if they do power down the card when suspending). This means you’ll need to carefully test whether heavily optimizing for storage results in huge slowdowns at startup / resuming your mobile device.