by Aleksey Meyev
07/30/2007 | 06:52 PM
Many of us are busy increasing the amounts of graphical information about the world around us by using a digital camera or a camcorder. Besides choosing the image capturing device proper, we have to face the problem of choice of the memory card. Owners of cheap cameras just have to choose the type and capacity they need. The speed characteristics of the card aren’t important for them since most available memory cards will suit such cameras. But people who have more serious devices, perhaps professional-level mirror-based cameras, have to approach the problem of choice more carefully. The speed of reading from the card may also be important, for example for journalists. When you’ve got some hot material on your card, you have to send the photos to your agency as quickly as possible. High speed is also needed by owners of camcorders with flash memory and by users of mobile devices like PDAs, players, media-centers and notebooks. Why do you need a memory card in your notebook? Because it is a small and universal data storage device that consumes very little power.
Last year SanDisk announced a new series of high-speed Compact Flash memory cards for this broad user audience and accompanied it with a release of two new card-readers with FireWire and USB interfaces. With capacities up to 8GB, this card series delivers read and write speeds up to 40MB/s, according to the manufacturer. For comparison, the previous high-speed Compact Flash series from the same brand, Extreme III, delivered only half that speed, 20MB/s. Compared with the CD speeds, 40MB/s equals 266x.
Is this speed really high for practical applications? Let’s evaluate it by using images in RAW format which stores raw, unprocessed data from the camera’s array and provides broad opportunities for further processing. This format results in very large files. One frame from a 10-megapixel array (such arrays are currently employed in top-end amateur cameras and in popular mirror-based cameras like Canon 400D, Nikon D80, Pentax K10D and Sony Alpha) is as large as 10-15MB in RAW format, depending on how complex the image is. Easy to calculate, the flash card is fast enough to capture at a rate of 3-4 such frames per second. For a majority of cameras, except for professional-level mirror cameras, this speed is the maximum the camera controller and the shutter can do, and the memory card won’t be a bottleneck. The speed of the card-reader integrated into the camera may also be a bottleneck, of course, but that is the camera maker’s rather than the card maker’s problem.
If you have been reading our reviews of portable storage media, you already know that the data-transfer rate of 40MB/s is unachievable with the USB interface. This is why two card-readers were introduced for the new series of SanDisk flash cards. The one with a USB interface is for ordinary speeds while the FireWire-interfaced version is for those who need the maximum possible data-transfer rate.
If you want to know a bit of history, read the following section about the story of Compact Flash format and its technicalities. And then we will tell you more about the devices we are going to test today.
The Compact Flash format was developed by SanDisk in 1994. October 11, 1995, the Compact Flash Association was established by twelve companies (Apple Computer, Canon, Eastman Kodak, Hewlett-Packard, LG Semicon, Matsushita, Motorola, NEC, Polaroid, SanDisk, Seagate and Seiko Epson) to standardize and promote it. The first revision of the specification was ratified then. The goal of the new standard was to preserve all the advantages of ATA Flash cards while eliminating their main drawback, which was their too big size. Compact Flash got a PCMCIA-compliant 50-pin parallel interface and dimensions of 36 x 43 x 3.3 millimeters. You can install a Compact Flash card into a PCMCIA slot via a simple adapter. All CF cards support two voltages, 3.3V and 5V.
The CD card is based on flash memory of the EEPROM type (Electronically Erased Programmable Read-Only Memory) whose characteristic features are:
Another feature of this memory type is that memory cells are accessed in blocks. A block of several cells is read or written to at once even if only some of the block data are actually required. If data has to be written into a partially free block, the existing information is read and merged with the new information, and then the resulting block is written in full instead of the old one. This method has a rather poor random access time, but also high sequential read speeds.
Memory cells are getting destroyed from being rewritten and their service life is about 100 thousand rewrite cycles. The controllers of modern cards feature special tracking algorithms for distributing data evenly among all the card cells, which makes the service life of the whole card longer. The controllers also keep track of the status of particular cells. When a cell is destroyed, the entire block with that cell is marked as destroyed and is substituted for a reserve block. Each card has reserve blocks for that purpose. And when there are no more reserve blocks, the card capacity begins to shrink as more blocks get destroyed from use. It is a very unlikely event for your flash card to get destroyed fully. More probably, its capacity won’t suit you anymore and you will replace it with a larger card before the described situation.
Interesting to note, early Compact Flash cards used to come in capacities of 2, 4, 10 and 15MB and the standard described a maximum data-transfer rate of 8MB/s.
The standard was evolving steadily towards larger capacities, higher speeds, and broader functionality. It quickly became the most widespread among flash card formats.
In March 1998 the Type II card specification was added (and ordinary cards began to be called Type I). The new cards were thicker (5 millimeters as opposed to 3.3 millimeters) to accommodate more storage space. Type I cards were compatible with Type II connectors and such connectors came to be used everywhere although the cards themselves were getting less and less popular.
In the fall of the same year the CF+ specs (CF 1.4) were written to describe input/output functions for devices designed in Compact Flash format. Various fax-modems, Ethernet adapters, barcode readers and, eventually, TV-tuners, Bluetooth adapters, GPS receivers and Wi-Fi became available in Compact Flash format, making it even more popular. In the spring of 2001 the standard is expanded by adding security features (Secure Compact Flash).
The version 2.0 specification was released on the 16th of June, 2003. The maximum allowable interface speed was increased from 8MB/s to 16MB/s (the then-available chips could only yield 5-7MB/s, so the authors of the specification tried to make some reserve of speed for the future). The support for a DMA interface with UltraDMA-33 mode was introduced for new devices. Considering the growth of card capacities, the makers of new devices were advised to provide support for FAT32 besides FAT12 and FAT16. This should have eliminated the possible problem with 2GB and lager cards.
These future-proof measures proved to be exhausted in just a year and a half. On the 6th of January, 2005, the version 3.0 specification appeared. It increased the speed to 66MB/s while maintaining compatibility with earlier released cards and added support for UltraDMA-66. The maximum storage capacity was increased to 8GB.
And finally, in March 2007 the version 4.1 specification was released to increase the speed to 133MB/s and add UltraDMA-133 mode. You can download the specification from the CFA website after passing a simple registration procedure.
Today, there are thousands of various devices using Compact Flash cards. The maximum storage capacity is 16 gigabytes and Lexar, one of the leaders in this field, has announced a card with a speed of 300x (or 45MB/s).
Now let’s get back to the devices we’ll test today.
As we mentioned above, the Extreme IV series includes the fastest of all cards currently made by SanDisk. Their peak read and write speed is as high as 40MB/s according to SanDisk’s internal tests and they come in 2GB, 4GB and 8GB capacities. As opposed to the previous Extreme III series, there is no 16GB model in the new series. The manufacturer declares an operating temperature range of -25 to 85°C.
The SanDisk Extreme USB Card-Reader, equipped with a USB interface, was one of the two readers introduced along with the new flash card series. This reader is declared to ensure a write speed of 18MB/s with Compact Flash cards. This is a lower speed than the new card supports, yet is quite high for comfortable work with large-capacity cards. Besides Compact Flash, this reader supports Secure Digital and Memory Stick Pro Duo. It comes with a USB cord, a user manual and a CD with the RescuePro software for restoring deleted data on external disks.
We also took a model with a FireWire interface (its connector is FireWire800 compatible, but we guess it only supports data-transfer rates up to 400Mb/s). It provides a maximum write speed of 40MB/s that allows getting the utmost from the new flash cards. Unlike the Extreme USB, this reader supports Compact Flash format exclusively.
The package contains two interface cords with different connectors, a user manual, and a CD with data restoration software. This reader comes with a 3-year warranty (as opposed to the USB model’s 2-year warranty).
We used the following software:
Using two versions of FC-Test allows to compare the results with those we got in our earlier tests while transitioning to the newer version of the testing program.
The following PC configuration was used for the test:
We took cards from the older series, SanDisk Extreme III and SanDisk Ultra II, for the sake of comparison. Once leaders in terms of performance, these cards are still being produced and are quite popular among the end users. The new card-readers will be compared with the previous-generation SanDisk ImageMate USB 2.0 and a SanDisk Ultra ImageMate FireWire, which we have already tested and discussed in detail before (for details see our article called Eleven Card Readers Roundup: Which One Is the Best?).
We’ll start out with the synthetic IOMeter benchmark. This pattern measures the sequential read and write speeds on data blocks of different size (from 0.5 to 1024KB) for each card in the different card-readers.
The USB card-readers deliver similar results with the Ultra II card. The Extreme FireWire is considerably faster than the USB models while the older FireWire reader is inferior to the others. The abrupt ending of the latter model’s graph is due to its inability to process data blocks larger than 128KB – you’ll see this in other tests as well.
The Extreme III behaves differently. It delivers its maximum performance when working in the Extreme USB reader. It is here that it surpasses the Ultra II card. Strangely enough, it does worse in the Extreme FireWire, somewhat better on small data blocks than in the ImageMate USB reader. The old model of the FireWire-interfaced reader is the slowest of all and still cannot process large data blocks.
The new Extreme IV card shows its best in the Extreme FireWire. This pair achieved the declared speed of 40MB/s on large data blocks and leaves no chance to any other card/reader pair. In the other readers this card behaves normally, yet its downright poor performance in the ImageMate FireWire is somewhat surprising.
This sequential writing test is perhaps the most important of all synthetic benchmarks for memory cards.
The Ultra II card delivers almost the same speed irrespective of the reader, except for the sluggish ImageMate FireWire. This must be the maximum write speed of this card and you cannot expect any improvements from changing the reader.
The Extreme III behaves as oddly as in the read test: it has its best result, about 50% better than the result of the Ultra II, in the Extreme USB reader. It performs slower in the Extreme FireWire and the ImageMate. The ImageMate FireWire is the slowest of the readers again.
The Extreme IV card feels at ease in the Extreme FireWire at both reading and writing, reaching a speed of 33.5MB/s on large data blocks. This is an excellent result, even though lower than specified. In the other readers the write speed of this card equals its read speed.
We’ll evaluate the speed of random read and write operations by measuring the time the drives take to perform them. The data block size is varied from 0.5 to 32768KB in this test.
It can be seen that the random read speed grows up on data blocks of 512KB and higher. As for the smaller blocks, the read speed of the old Ultra II card is far lower than that of the Extreme series products. The difference between the cards becomes negligible on large data blocks and the difference between the readers becomes conspicuous. It’s like in the previous test: the Extreme III has its highest speed in the Extreme USB reader while the Extreme IV card prefers the Extreme FireWire.
The random write test produces almost the same results as the random read test above. It’s on large data blocks that you see a big difference between the competitors. The best card/reader combinations are the Extreme III in the Extreme USB and the Extreme IV in the Extreme FireWire. The Ultra II is still slower on small data blocks although not as much slower as in the random read test.
The average read/write response time is measured in a 10-minute test to read/write 512-byte data blocks with a request queue of 1.
The outcome of this test depends largely on the card rather than on the reader. And the new Extreme IV in its favorite Extreme FireWire reader is unrivalled here. The response time of the cards is higher by a third of a millisecond in the Extreme USB and the ImageMate USB and by half a millisecond in the ImageMate FireWire. This difference matters for the Extreme series cards, but the Ultra II doesn’t care – its response time is far too bad.
The card-readers don’t differ much when performing write operations. The Extreme IV wins this test with a response time of 30 milliseconds. The Extreme III is slower by 5 milliseconds and the Ultra II has a twice higher response.
We continue checking out external storage media for their compliance with Windows Vista’s ReadyBoost technology using IOMeter. To remind you, the point of that technology is in using a flash drive or a flash card as an additional cache with a data access time lower than that of the hard disk. To qualify for this application, the external disk must meet certain performance requirements and have a capacity of 256MB and higher. The OS itself benchmarks the speed of the attached device and proposes that it be used for ReadyBoost. What are the requirements? The flash disk must ensure a data-transfer rate of 2.5MB/s and higher when reading random 4KB data blocks and a data-transfer rate of 1.75MB/s and higher when writing 512KB data blocks. So, we measure these two speeds in a 10-minute test. For better readability, the red vertical line marks the ReadyBoost-compliant speed in the diagrams:
The read diagram suggests that the Ultra II is not fit for ReadyBoost. And neither card is ReadyBoost compliant in the ImageMate FireWire reader – its read speed is too low. Note the superb performance of the Extreme IV card in the Extreme FireWire reader.
It’s simpler with writing: each card easily overcomes the required barrier. The ImageMate FireWire reader just cannot work with data blocks of the required size. The Extreme IV card is brilliant in the new card-readers.
FC-Test will help us examine the flash cards under real-life conditions. The test writes and reads a few file-sets and measures the time it takes to perform each operation. This allows to calculate the resulting speed and see how it depends on the number and size of the processed files. We use three file-sets that differ in the size (1, 10 and 100MB) and number (1, 10, and 100) of files included. Practice suggests that a 100MB file is large enough to reveal the maximum performance of a flash-based storage media and using a larger file doesn’t affect the results much
Let’s check out the write speeds, which are the most important results of this test.
The standings of the cards haven’t changed but the overall picture is different. The Extreme IV card is unrivalled if compared in the same reader. The write speeds are lower than the read ones and equal 26MB for the Extreme IV in the Extreme FireWire and 12.5MB/s for the Extreme III in the Extreme USB (this card still performs worse in the FireWire reader). The Ultra II is slow, its write speed being no higher than 5MB/s. The old ImageMate card-readers can’t stand a comparison with the new Extreme series.
The speeds get higher when we switch to larger files, but the standings do not change. The Extreme IV is almost as fast as 30MB/s while the Ultra II nearly notches a speed of 7MB/s.
It’s on largest files that the maximum of speed can be achieved. The Extreme IV does not deliver the promised 40MB/s, yet it overcomes the 30MB/s mark which is a good practical result. The Extreme III is close to 15MB/s and the Ultra II to 9MB/s. The card-readers perform just as they’ve done in the earlier tests.
The read tests come next.
Processing 1MB files, the cards all deliver the same speed in the ImageMate FireWire. It is the card-reader that’s the bottleneck here rather than the flash card installed. The old ImageMate USB is only sufficient for the Ultra II card – the others can do better as they prove in the newer card-readers. The Extreme USB yields 17MB/s, which is enough for the Extreme III card but not enough for the Extreme IV. The latter card can only shows its best in the Extreme FireWire and is very close to the specified speed of 40MB/s, leaving the other pairs far behind. Surprisingly, the Extreme III card has better results in the USB-interfaced Extreme series card-reader rather than in the theoretically faster FireWire model.
The transition to 10MB files doesn’t have a big effect on the results.
The results don’t change again when we switch to a single 100MB file.
We’ll now test the cards in the newer version of FC-Test using the same methodology.
The cards have similar, and rather low, speeds when writing 1MB files. Well, you don’t actually need a very high speed with such small files unless you are going to emulate video capture by quickly shooting into highly compressed JPEG files. The card-readers behave just like they did in the previous tests, so there’s no need for more comments.
This test provides real-life photography conditions for the tested devices. The speeds are high, even though far from the specified ones. The Extreme IV speeds up twofold, leaving the older cards behind.
The Extreme IV can’t reach even 30MB/s, let alone the specified 40MB/s, on the 100MB file in this test. The other cards do worse than in the first version of FC-Test by 10-15%, too.
Next go the read test results.
The results are so close in each diagram that they don’t have to be commented upon separately. The Extreme IV finally notches the promised 40MB/s mark. The Extreme III card still prefers the Extreme USB reader, having a read speed of 17.4MB/s in it. The Ultra II has a speed of 10-11MB/s. The ImageMate FireWire refused to pass this test, producing inadequate results.
Delivering superb performance, the Extreme IV flash card is far superior to the previous series cards in virtually every parameter. But you should be aware that the card can only give you its best when you insert it into an appropriate reader. To connect it to you PC, you should best use the Extreme series reader with a FireWire interface. The USB version of the reader cannot provide the bandwidth necessary for the card. The practical write speed of the card doesn’t reach the declared 40MB/s, being only 30MB/s, yet it is anyway two times the speed of the previous generation card, Extreme III.
By the way, if you have an Extreme III or want to buy it and save a little money, you should consider the USB version of the Extreme reader because this card performs better in it than in the old ImageMate USB reader or in the far more expensive FireWire version of the Extreme reader. It looks like different flash cards have their own preferences when it comes to the specific reader.
The Ultra II series card looks poor against the mentioned two and cannot give you a speed higher than 10MB/s.