There have been many comments about the speed of the different brands of Compact Flash solid state memory and the IBM Microdrives in magazines and on the web. I decided to test them myself to discover the performance under carefully controlled conditions. They were tested in Nikon's D1 & D1X professional digital cameras.

I did an initial study about a year ago. There have been many improvements in the Compact Flash performance even in the short time between the two tests. The developed test procedure is easy to duplicate for someone wanting to run their own test. The test could even be run in a store while buying a Compact Flash card. The Compact Flash cards were obtained from various sources for the evaluation, and I have included a detailed list of the models and the date when I obtained them.

There are two main issues regarding speed while taking pictures: How fast will the camera save a picture and be ready for the next one, and how much time does it take to get ready for the first picture (similar to shutter delay).

Also of interest is, how fast a card can be read and copied into a computer or other storage device. I evaluated it using a FireWire to Compact Flash adapter and saved the pictures on a PC. Typically it is much less important than the picture taking speed, which is critical for live action photography. The card copying is usually done in an office setting and the speed is not critical.

Interestingly, there is a big speed difference between the D1 and the D1X while writing to the Compact Flash cards. The D1X is 2.1 times faster than the D1, which is good since the D1X file size is twice as big as the D1's.

Disclaimer: There is no guarantee that the data taken is typical of a brand of cards since normally only one card is available for testing. Different file types or file sizes can also result in different timing results. Also, if the cards are used in another type of camera that is slower, the speed difference might not be noticeable. This report is not intended as an endorsement or recommendation of any brand, it is just the result of my testing.

Time to Save a Picture:
This measurement evaluates the time it takes for the camera to save a 4 MB or 8 MB file to the Compact Flash card. By measuring the "raw" unprocessed files (NEFs) from the camera, any variable processing delays are avoided and the file size is large enough to get reliable timing. It is also the preferred format if best picture quality is to be achieved. To keep the test as accurate as possible and to remove as many variables as possible, all the measurements were made on the Nikon D1 and D1X.

The data was taken with an electronic timer, optically monitoring the green "Busy" LED next to the card. It can easily measure the light "on" time to one thousand of a second duration. At the same time, I monitored the camera's current drain. Before starting each test, the card was formatted in the camera.

I tested the CF cards both by measuring individual pictures, and then also by saving a burst of 10 or 6 NEFs. Interestingly, some brands are extremely repeatable, picture after picture, some brands have an obvious pattern. For example, every 4th picture takes 0.3 seconds longer, and one card has a much more variable timing (+/- 0.5 second) without an obvious pattern. This was especially true on the previous test, about a year ago, but today's cards are much more uniform. I have attached the detailed test procedure in the appendix.

Time to Wake Up:
Besides the speed of saving images to the card there is another important issue for some photographers: The time to "wake up" the camera and take the first picture.

If one can anticipate the action, it is easy to wake up the camera a second or so in advance and the D1 will fire "instantly" when the shutter release is pushed.

If the action happens without warning, the time to wake up might be critical. Surprisingly, it is affected by the brand and type of card used. I really expected the camera design to be independent of the card since it has an internal buffer. Apparently, the camera checks the card when it wakes up and waits for the response before taking the picture. Most brands respond to the status requests at once, while some cards do a detailed check and therefore take a significantly longer time. The extra delay caused by the card varied from zero to 1.5 seconds, which feels like eternity while you are waiting for a picture to be taken. Most camera models do not check the card on wake-up so the extra delay is usually not seen in them.

The tests were done by electronically measuring the delay from the time that the shutter release was pushed, to the time that the camera fired. To make it repeatable, the test was done in manual focus and manual exposure mode. The camera happily takes pictures without a card and I used that to my advantage. The time without a card is consistently 0.304 seconds for the D1 and 0.330 for the D1X. That time is then subtracted from each card to get the extra delay caused by the card.

Lexar cards have a significantly longer wake up delay than many other cards. This was surprising since they were the fastest saving cards. Lexar has developed a new internal format for the cards, and the cards can be updated by Lexar to avoid the time delay in a D1 camera. If you have, or will purchase a larger, faster size of a Lexar CF card, contact Lexar and they are very helpful about updating the card. After the update, the cards are stamped "D1" on the back and have zero delay. Excellent service!

Lexar claimed that they are working on a solution to allow users to download software and update the Lexar cards themselves. Lexar does not have a release date for that software. According to Lexar, the new 512 MB cards do not have the wakeup delay, but I have not tested it myself.

Summary:

There has been a big improvement in most of the "slow" brands evaluated about a year ago, compared to the latest group of tested cards. The time to save a picture on the solid state cards varied between 4.6 to 11.1 seconds, while a year ago they varied from 5.2 to 12.5 seconds. Although the speed leader, Lexar, has improved its performance by 14%, most of the competition has made even greater gains and are closing in on them.

If Sandisk is excluded from the tested units, there are only minor differences between them when used in the D1X camera.

Similarly, there have been big improvements in the wake up time. All the solid state cards had zero delay, while a year ago, there were delays up to 1.5 seconds! One important note: Off-the-shelf Lexar cards still have the 0.5 second delay, the cards must be sent to Lexar to be updated to zero delay. Call Lexar for the details, they are very helpful.

Delkins and Unity declined to have their cards tested. Since their cards might have different electronic controllers and internal design, I cannot guess or suggest at all how they perform compared to the tested ones. If I get the opportunity to test them, or any other untested brands and models, I will gradually update the report with the new data. I will also add data from the Nikon D1H when it becomes available for testing.

If your camera is slow, there might not be any noticed performance difference between some cards, simply because the cards are faster than the camera. It is still a good idea to try to buy faster cards, since your next camera is likely to be faster.

Please note that the Viking cards that were tested, were obtained directly from Amazon and appears to be a new high speed model. They actually are very similar to the Lexar cards, they even had a slight wakeup delay like the Lexars. They were the fastest cards tested to date! Once I find out the details about the new cards, I will include it here. On Viking's website, only "regular" speed cards are shown.

A new brand of cards called "RiData" were tested. Note, this is not the same brand as "RitData" although the cards are made by the same manufacturer (Ritek). I obtained a loaner card from the Australian distributor, Vasa Carapic, and the card performed extremely well, almost identically to the Viking cards mentioned above. Both brands outperformed the 12X Lexars. The RiData are shipped "second day" worldwide for a very low fee and they are also sold on Ebay.

Another new card on the market called "Transcend" was tested. It was obtained from "MyDigitalDiscount" together with another 512 MB card. The second card from the same supplier was claimed to be manufactured by RiData although no markings on it stated so. See #19 and #20. I wish the manufacturer would mark their cards. Click on the ID code in the second table for a picture of the cards.

Both of them performed extremely well and they are very similar to the previously tested RiData and Viking cards. I will update the report shortly with the FireWire transfer data. See the links below for information about vendor.

Remember, the IBM Microdrive is not approved by Nikon to be used in the D1 camera although it is approved for the new D1X and D1H. Many D1 owners are using the Microdrive in the D1 without problems but others have had serious problems and lost many pictures. There is nothing wrong with either the Microdrive or the D1, they simply are not fully compatible. I have only heard good comments about the Microdrive when used in the D1X and the speed is very good.

If the IBM Microdrive is used, there is an additional current drain above the CF cards. The write current drain is typically twice as large, as for the solid state cards. To put that in perspective, it is only about a 10% increase in the total current drain of the camera.

For anyone wanting to duplicate these tests, here is the procedure:
If a D1X is tested, change to 6 NEFs instead of 10 NEFs.

Set the camera to manual exposure mode, 1/125 sec, manual focus and Hi-Raw= 12 Bit NEF. Also set the camera to continuous shooting mode (CSM: 25-Ch and CSM:26-21) to get a 10 shot sequence. Format the card in the camera before each test. If it is an IBM Microdrive, see the section below.

Measure the time from when the green light turns on, until it turns off. Typically, it will be about 60 seconds. Do not include the time to take the pictures, just the duration that the green light is on. This time is then divided by 10 and used in the data table. If the card is less than 41 MB and the 10 NEFs will not fit, divide the time by the number of pictures that were taken. This procedure was repeated until the card was full, and then all the data was averaged. (I did not have the patience to fill up the 1GB Microdrive.) The data rate was calculated by dividing the file size (4.055474 MB or 8.088552 MB) by the time per picture.

To be able to accurately measure the time, you need to practice and write down the time for each test. After a few tries, you should get repeatable numbers that you can trust.

IBM Microdrive:
There have been many ideas about how to use the Microdrive in the D1. One concept is to never format the Microdrive in the camera because the camera does not understand about defective sectors. Instead, it is suggested to format it in a PC and afterwards running Scandisk. Even better, use Norton DiskDoctor to find and mark bad sectors. In day to day use, the pictures are then just deleted from the Microdrive without reformatting it. This concept sounds logical, although I have not investigated it. (I do not use Microdrives in my D1.) For the testing, I tried it both ways and did not detect any speed difference.
In the D1X, I formatted the IBM Microdrive in the camera as I did with the other cards.

Wake-up Test Procedure:
The tests were done by electronically measuring the delay from the time that the shutter release was pushed, until the time that the camera fired. To make it repeatable, the test was done in manual focus and manual exposure mode.

I noticed a small change in speed, just a couple percent, depending on the camera position; down-horizontal-up. To make a long story short; it turns out that the camera is light sensitive even in manual exposure mode. After realizing that, I added a lens cap and I even had to close the viewfinder.

The data was averaged from 10 tests and the inherent camera delay of 0.304 seconds (D1X=0.330) was subtracted and the time was rounded off to the nearest 0.1 seconds.

FireWire Reading Test Procedure:
A Lexar FireWire to Compact Flash adapter was used with a Pentium-3 PC at 733 MHz. Twelve 8 MB NEFs were saved in each card using the D1X and then they were copied onto the PC's harddrive. The tests were done by measuring the time it took from dropping the 12 NEFs into an empty folder. The test was repeated several times to assure accuracy.

Current Drain Test Procedure: The additional current drain, while the camera was writing to the Compact Flash card, was measured using professional electronic test equipment. The test was repeated several times to assure accuracy. The inactive, standby current drain was absolutely insignificant of all the tested units except the IBM Microdrive. It "wakes up" and draws 0.08A for a short time until it shuts down again.