Today's Memory Cards

 

There're a lot of misconceptions around today's CF and SD technology. A lot of people swear that there're drastic differences in between the speed and the power consumption of the cards. This is why we started working on a series of test to find out whether there's significant difference in speed and power consumption of the cards.

 

One common misconception states that the larger the capacity, the more power is needed to power the writing of a card. Another common misconception states that SD cards are orders of magnitude slower than CF cards. In our test, we've proved both misconceptions false.

 

We've tested the following brands/cards:

Kingston 128 SD (age: about 6-7 months)

SanDisk 16 SD (age: about 9-10 months)

Pretec 48 CF (this one is pretty aged; was manufactured more than two years ago. This certainly shows in both the very bad performance and the comparatively high power consumption. Pretec's newer cards must be much faster.)

Ridata 256 CF (bought it some 18 months ago)

Unbranded 512 CF (bought it some 3-4 months ago). I coulnd't determine the original supplier - I have neither a Mac nor an SCSI card reader (see the Links section).

 

 

Figure 1: the cards I've tested

 

1. Access Speed

 

ActiveSync speed with the h2210

 

I've computed the ActiveSync write speed of both the Ridata 256 CF card and the Kingston 128 SD. The Kingston's write speed was around 170 kbytes/s; Ridata's around 220 kbytes/s.

 

In-PDA filecopy speed

 

File Explorer turned out to be very buggy under WM2003. It is very slow at writing to SD/CF cards under WM2003, independent of the source (mem/another card). It must be a bug in either the h2210 or WM2003 itself.

 

This is why I only made few tests with File Explorer and used Total Commander (http://www.ghisler.com/ce.htm) for all the remaining tests. (Note that the lack of progress in the scrollbar can be misleading!)

 

I've reset the Pocket PC between all individual tests. (Given that the h2210 is pretty slow at booting up, it was pretty tiring job). I've created a large WAV file and copied it to and  from the cards in most possible combinations.

 

The speeds are all in Mbyte/sec. (2*) means I've run the test two times to make sure I've got the right result.

 

 

        from:

to:

Kingston 128 SD

Pretec 48 CF

Ridata 256 CF

Main memory

Kingston 128 SD

 

 

0.052

0.051 (2*)

Pretec 48 CF

 

 

 

0.049

Ridata 256 CF

0.198

 

 

 

Main memory

0.47

 

 

 

Table 1: Card speeds with h2210 and File Explorer

 

 

from:

to:

Kingston 128 SD

SanDisk 16 SD

Pretec 48 CF

Ridata 256 CF

Unknown 512 CF

Main memory

Kingston 128 SD

 

 

0.25

0.30

0.302

0.35

SanDisk 16 SD

 

 

 

0.37

 

 

Pretec 48 CF

0.22

 

 

 

 

 

Ridata 256 CF

0.56

0.5

 

 

 

0.99

Unknown 512 CF

0.503

 

 

 

 

0.755

Main memory

0.64 (2*)

 

 

0.54

0.37

 

Table 2: Card speeds with h2210 and Total Commander

 

I've also run some CF <--> main memory tests with the iPAQ 3660 to find out whether it's only the WM2003 version of File Explorer that is slow. Note that I wasn't able to test the unbranded 512 CF in my 3660 because it didn't recognize the card. (I use the PCMCIA sleeve, btw). Incidentally, it was my only one PDA that refused to recognize the CF; even my oldest handhelds (e.g., the OmniGo 100, the HP 320LX etc) recognized it.

 

The results are not much worse than those of h2210 (note that I've run the tests with the inherently-slower File Explorer and not with Total Commander):

Ridata CF -> mem: 0.30 MByte/sec

mem -> Ridata CF: 0.194 MByte/sec

 

I've also tested whether the difference in speed of two CF cards do affect casual (non-filecopy!) programs. I've chosen Mobipocket Reader 4.5 for this test (this older version is a bit faster at searching than later versions because it uses internally 8-bit and not 16-bit characters) and I used a very large, 150 Mbyte-long document compressed into a 70 MByte archive. The two cards fared equally fast; this means it's very rarely that the card write/read throughput speed will be the bottleneck in a given application. The CPU is just too slow for most tasks so it's still the CPU that is the bottleneck, and not that of the CF card.

 

2. Power Consumption

 

For this test, I've used NoteM (http://www.geocities.com/zavorine/wince/mp3.htm) because it's common for Pocket PC users to use their PDA as a large-capacity sound recorder to record entire meetings spanning several hours. In such a situation, it's essential to use a memory card that consumes the least power. I didn't run any other application in the background and started all test with a fully loaded h2210. I didn't switch off the screen, just switched off the backlight.

 

All of the cards fared quite well in this test. The aging Pretec CF produced the worst results again. Using it made the battery some 5% emptier than using the Ridata card in 128 kbps mode. The unbranded 512 CF consumed certainly more power than the Ridata CF, but the difference wasn't really big. The unbranded CF consumed approximately as much (or a little bit more) power while recording the 128 kbps MP3 stream as the Pretec 48M CF while recording the 16 kbps CF stream.

 

There was virtually no difference between using the Kingston 128 SD for 4 hours and writing into the main memory instead. This means the Kingston 128 SD's power consumption at writing is orders of magnitude lower than that of the h2210.

 

The Ridata 256 CF certainly made its presence felt. In using the 16 kbps recording mode, after 4 hours, the battery was some 6-7% emptier than with the Kingston SD (or without any card). When I switched to using the 128 kbps mode (which means 8 times more writing), this difference raised to some 9-10%. This means there's not much difference between using high- and low-quality WAV/MP3 modes, you won't rerally feel the difference.

 

You should only resort to using SD cards instead of CF cards when absolutely minimal power consumption is a must and even a difference of 3-4% is intolerable. However, beware of SD cards. NoteM wasn't able to write constant bitstream into them at both 56 kbps and 128 kbps. In the former case, the sound stuttered now and then; in the latter case, it was completely unintelligible because of the constant stuttering. NoteM records everything else (that is, WAV's) to SD cards OK. That must be another WM2003-related bug in NoteM.

 

Note that, as I checked and jotted down these values by hand (didn't want a battery monitor application to put extra burden on the battery), some of the interim values are missing. However, as the battery consumption is constant and, therefore, the battery level decreases in a completely linear fashion, you can easily extrapolate the missing values.

 

 

sd16k

cf16k

cf112k

48cf16

512CF112k

0:24

96

 

 

 

 

0:32

95

 

 

 

 

0:45

 

 

 

91

 

1:00

91

 

 

88

88

1:14

88

 

 

 

 

1:22

 

 

 

84

 

1:37

84

82

 

81

 

2:00

 

 

 

 

74.5

2:27

76

 

 

 

 

2:34

 

 

69

 

 

2:50

71

 

65

 

 

3:00

 

 

 

 

61

3:05

 

65

62

 

 

3:20

 

 

 

 

57

3:28

64

60

 

56

 

3:40

 

 

 

 

52

3:50

61

 

 

52

 

3:54

60

 

 

 

 

4:00

 

 

 

 

48

4:40

 

46

 

 

 

 

Conclusion

 

-         there is no much difference in power consumption between various CF brands.

-         unbranded CF cards are considerably slower and consume a bit more power than branded ones (Ridata was definitely one of the fastest CF carsds in 2001; now, it's in the middle).

-         the amount of writing to a CF card clearly has an impact (however low) on battery consumption

-         SD cards consume much less power than CF cards. Still, the difference is not very big.

-         WM2003's File Explorer is buggy: copying main memory to any kind of storage cards OR copying anything to a SD card is very-very slow. The situation was much better in PPC2002.

 

Links

 

There're some cool links here for memory card testing methologies and memory card-related discussion forums.

 

http://www.dpreview.com/articles/mediacompare/

http://www.dpreview.com/forums/forum.asp?forum=1023

CompactFlash Performance Database: http://robgalbraith.com/bins/multi_page.asp?cid=6007

 

A new and pretty cool German article: http://www.de.tomshardware.com/storage/20021122/

Another cool German discussion forum/(scanned) article: http://www.ppc-welt-community.info/Community/showthread.php?threadid=12527

And another cool one:

http://www.ppc-welt-community.info/Community/showthread.php?threadid=29378