User Friendly Computers
of Royal Oak


Small Office - Home Office - Computer Consulting

The Case of the Slow Flash Cards:

By: Bill Woelk
Date: 01/24/2009

flashcards


While recently working with a photographer client of mine, I noticed her camera images were uploading very slowly from her Lexar Pro flash cards to her new custom-built 64-Bit PC. We were using the built-in 52-in-1 Rosewill card reader on her computer. We also tried the card reader built into her Dell 2408 24" LCD monitor. Both readers were very slow. The photographer had standardized on both 2GB and 4GB Lexar Professional 133x flash cards for her Nikon D200. These are premium grade flash cards that carry a higher than average purchase price. As such you would expect them to perform better than standard CF (Compact Flash) cards. Lastly, when any of her Lexar cards was inserted, Windows XP would no longer prompt for what she wanted to do, which was to open Adobe Photoshop CS3.

We next tried reading from both of her built-in USB 2.0 card readers using a 512MB Sandisk Utlra-II CF card I had with me. Both times Windows XP opened the normal dialog box and asked what we wanted to do with the card's contents. I next opened a Windows File Explorer window and tried copying both to and from the SandDisk Ultra-II card. The card was able to write and then read back all of the image files in a matter of seconds. We made sure to copy enough files to avoid running in burst mode.

What was going on here? A brand new 4GB UDMA, high end Pro card was being out-done by a three year old mid-line 512MB card. I decided then and there, that I had to figure out what was was causing these issues. Lexar is a highly trusted brand, certainly they were not over-rating their card's abilities??? So this incident prompted me to bench test eight different Compact Flash (CF) cards, that I had available to me. To round out my testing I also purchased a high end SanDisk Extreme III card and a Trend Micro - Transcend 4GB card rated at 266X speed. The Transcend card was highly rated by the end-user reviewers on www.NewEgg.com. It sported an excellent five-egg rating.

Here are some of the questions I wanted answers to:

  1. Is there a difference in performance between various USB 2.0 card readers?
  2. Is there much difference in read/write speeds between the various brands of CF cards?
  3. Are the 266x and faster cards worth the premium prices being charged?
  4. Is USB 2.0 limiting modern flash card data transfer speeds?
  5. Are UDMA cards compatible with USB 2.0?
  6. Are 2GB or smaller CF cards faster, than 2GB or larger CF cards?
Designing the Test: Please keep in mind the reason for these tests was to determine the continuous data transfer rates of the various cards under test, not their burst rates. I wanted to know why the Lexar cards were so slow at reading back data to my client's image editing PC. There are many other web sites that have measured the burst rates of various CF cards, if you are interested in burst rates, please use a search engine like Google to learn more. Please note: Burst rates are important when snapping digital pictures, to allow the camera to save the image file quickly and enable it ready to take the next picture ASAP.

In addition to the eight flash cards, I also wanted to compare the data transfer results for three different USB 2.0 card readers I had available to me. One was a portable I carry with my laptop, the other two were both Internal readers that I use in my system builds. The internal 52-in-1 AFT model XM-4U was the most expensive at around $30.00, the other internal Rosewill RCR-102 52-in-1 reader was mid-priced at about $18.00, and the portable Crucial Technology reader was purchased for less than $10.00.

The Test System:
I started by setting up a test computer I normally use at UFC for formatting and partitioning new hard drives. The basic system consists of an AMD Athlon 64-Bit 6000+ dual core processor with 4.0GB of DDR2-800, an GF-7300 PCI-E graphics card and a Seagate 80GB 7200 RPM hard drive. Pretty average, nothing special to write home about. It did have several things going for it though. For one, it had an easy to remove side door that allowed for easy access to the internal USB headers on the Gigabyte motherboard. This would make it easier for me to connect the two internal card readers and to observe their results. The system had recently been reformatted and the operating system was just freshly installed. So it had very little application software installed on it, that might skew the test results. For example it lacked both anti-virus and anti-spyware protection. I later installed AVG 8.x and MalwareBytes anti-spyware. This made no detectable difference in the test results. Note the system was not connected to the Internet while being used for testing. Unlike newer operating systems, the Windows 2000 Professional installed on this computer has very few background tasks running, that could interfere with my flash card benchmark testing. There were only 29 tasks listed in the Task Manager window. The last thing I needed was a warning window, or Windows Update, or who knows what else popping up in the middle of a test run and blowing my results.

The Test Procedures: First I selected 150 random, fine mode JPEG files taken with my Canon Rebel Digital SLR. These consumed 441MB of hard disk space according to the Windows File Explorer. Before testing, I disabled write-back caching on the computer's hard drive. I next shut down as many unnecessary background applications as I could. Windows disk indexing was also disabled. C-Cleaner was used to remove unnecessary files from the hard drive, then the hard drive was thoroughly defragged.

My benchmark testing consisted of opening a Windows 2000 file explorer window and copying the 150 image files from a hard disk folder to the flash card's root directory under test. I next copied the images from the flash card back to a C:\TEMP folder on my hard drive. The two copy sessions were both timed using a free stop watch program called Timer2. After selecting the files to be copied, I then dragged them to the target drive. As soon as I released the left mouse button, I started the timer. When Windows closed its copy window, I immediately stopped the timer and recorded the elasped time results. The timer was reset before each new test run. After a card was tested for both write and read back times, I closed out the explorer window. I then removed the flash card and inserted the next card to be tested. I then opened a new explorer window and repeated the process again until all eight cards had been tested in the three readers. The tests produced 48 separate timed results. If I got results that varied greatly from the mean average, then I repeated the tests and threw out any non-repeatable values. If the two tests were similar, then I averaged the results of the two readings. If they were still off, I re-ran the test a third time.

The Windows File Explorer in my experience is a fairly buggy program. It was buggy in Windows 95 and its still buggy even today in the latest versions of Windows Vista. I guess Microsoft is too busy adding new features to Windows 7, to fix the older OS features. If I did not close and then re-open a new Explorer window for each new flash card test, I would occasionally obtain weird, unrepeatable results. Even hitting F5 to refresh the drive tree did not help. Explorer would sometimes display ghost images of files that had recently been deleted, or the copy times would suddenly slow down to a crawl. I rebooted the system whenever I switched to a different flash card reader.

The Benchmark Results: I compiled all of my test data into an Excel spread sheet and then created the color chart below to better illustrate the results. Please keep in mind that I designed this test to compare large data transfer rates, not burst rates. Please see the chart below:

Flash Card Results Chart


Note: The label for the AFT model XM-4U (top most reader), did not render properly in the chart above.


Results: The AFT XM-4U was the oldest card reader used in these tests and its slower overall results bare that out. The card test results above are arranged the fastest read-times to the slowest. The fastest results are at the top of each the three card reader test series. I was not too concerned about write times, since that takes place inside of the camera for the most part and is subject to individual camera variables. The low-end Kingston Flower CF card produced the best read speed results on the built-in AFT XM-4U reader. The 133X Lexar Pro 4GB comes in last place on this reader. The Ultra-Fast Transcend 266x card falls into the upper middle range at one minute and thirty two seconds.

With the Rosewill built-in reader, the Transcend card moves to the top of the list on read times. The inexpensive MicroCenter bargain card is at the bottom at 2:24.

The real shocker in this test series was the third inexpensive Crucial Technology (CF only) card reader. I bought this reader to augment an Asus Eee PC laptop I own, that only has a built-in HD-SD card reader. I paid less than $10.00 for it from NewEgg.com. Its very compact and shaped like a fish head, with the eye of the fish being the activity LED. This external reader out-performed both of the Internal readers. On this reader the Sandisk Ultra-II card turned in the best read results, at 60 seconds flat. The Lexar Pro 2.0GB brought up the bottom on this test. Note the skewed write time results between the various brands. I don't know how to account for this adnormality. The low end Kingston Flower card came in last place in the write tests.

I strongly suspect that the particular USB chipset used inside of the card reader makes a big difference as to which readers performed the best during these tests. None of my readers to my knowledge are rated as UDMA readers. On a slow card reader, all of the cards performed poorly. On a faster reader most of the cards posted better times. With these results in mind, my photographer client purchased a new Lexar Professional UDMA dual-slot USB 2.0 card reader (model RW035), for me to test. The Lexar UDMA model RW035 produced some very interesting results. The most amazing was the Transcend CF card by Trend Micro. This card posted an astounding :31 second write to flash time for the 150 image copy test. That is 441MB of image data copied in only 31 seconds! I was so surprised by this result, that I repeated the test a second time. The second test was identical with all files written in only 30 seconds. The next fastest writing card was the Sandisk Ultra-II at 00:54 seconds. The Sandisk Ultra-III came in at close at 00:59 seconds. Twice as slow as the less-expensive Transcend card. Both Lexar Pros were in the poky 01:16 second range. The bargain priced Kingston flower brought up the rear at a very sluggish 04:02.

On the all-important read times, the 266X Transcend card tested the fastest at 01:13 seconds. The Sandisk Ultra-III was in second place at 01:15 seconds. The two Lexar's averaged 01:18 seconds each. The Kingston Elite did fairly well at 01:16 seconds. The Sandisk Ultra-II dropped to third from last place at 01:28 and the MicroCenter card came in last at 01:29. I am surprised that all of the cards read back data slower than they could write it. This makes no sense to me at all. Other than their must be some internal optimizing going on to support burst modes on digital cameras better. Most flash cards have an interface system that varies which memory cells are written to, in order to spread the data storage over all of the flash cells evenly. Flash memory has a limited number of lifetime write cycles per cell. Too many writes to one cell and that cell can be burned out. Reading on the other hand is not supposed to pose a longevity problem with flash memory.

USB 2.0 Bus Issues: I have read that the aging USB 2.0 bus is becoming a bottleneck as far as UDMA cards and reader performance is concerned. As my testing with the Lexar UDMA reader clearly shows. it definitely made a speed difference, even on a USB 2.0 bus. Both Lexar and Sandisk sell some rather pricey Firewire 800 readers for those users who want the ultimate performance. These readers can cost in excess of $100 to purchase. The next hurdle is how many users even have a Firewire 800 port on their computer? This means adding an expensive PCI-E Firewire 800 card to your desktop. Don't bother with IDE Firewire 800 cards, the ancient parallel ATA bus is also a bottleneck any more. If you own a laptop you are probably out of luck, unless you can locate a Firewire 800 PC express card to fit your laptop. Firewire 400 which is available on many existing PC's and laptops, is said to have a slight speed advantage over USB 2.0, due to its reduced command and control overhead. I would not recommend spending a premium on a Firewire 400 reader over a USB 2.0 reader, there just is not that much speed improvement in my experience. Firewire 800 yes, but Apple Computer owns the Firewire 1394a/b standard and charges other manufacturers a hefty royalty fee to implement the standard. This has all but limited it use to video cameras and a few external hard drives.

eSATA may be a near future solution to this USB 2.0 bus bandwidth issue on the PC. SATA-II interfaces are rated at up to 3.0 Gbits per second transfer rates in burst mode. If you have a spare SATA header on your motherboard, its an easy task to add an eSATA back plate. The back plate fits into any open expansion slot and has a short SATA cable that attaches to any open SATA connector on your motherboard. The problem is nobody manufactures an eSATA reader yet. I did find a couple of internal readers that were meant to convert a CF card into a pseudo solid state hard drive. These units had both PATA and SATA interfaces on them, but none that I found, supported UDMA transfer modes. There were many complaints from end-user reviewers about how slow these units were. Many failed to even compete with conventional 4200 RPM 2.5" laptop drives. These adapters seemed to be aimed mostly at the netbook, or embedded industrial market where speed is not a big issue.

To answer the question regarding does a CF card's size or capacity affect its speed. I could not see much difference between the two Lexar 133X Pro cards. One was rated 2.0GB capacity, the other 4.0GB. This question is still open to debate due to my small number of cards I tested. If I get access to more identical cards, I will look into this matter again. Most pro photographers purposely stick with 2-4GB cards, over the genuine fear that if they get a defective card, they will lose fewer image files until the data can be backed up to other media. Imagine losing all of the data on a 16GB flash card? It happened to me once, but fortunately I had all the files backed up on a server hard drive. Flash card technology is still not 100% reliable in my experience. Always unload any new photos as quickly as possible to a laptop hard drive, or an external hard drive when working in the field. For this reason I am extremely leery yet of solid state hard drives. I want to see some MTBF ratings first.

Recommendations: The Lexar UDMA model RW035 card reader out-performed the other three card readers tested by a wide margin. It can be purchased online from vendors like B&H Photo for under $30 and will most likely max out your USB 2.0 bus. Coupled with the Lexar UDMA reader I would strongly recommend the Transcend 266X UDMA cards as a best buy. These are available from NewEgg.com for under $35.00. The Lexar Pro 133X cards are over-priced in relation to the performance they provided in my opinion. I might note that the Lexar 2.0GB card used in my testing had to be repeatedly reformatted in order to delete the files on it. Simply using delete in File Explorer did not do it. I threw this defective card away after my testing was completed. The Sandisk Ultra-II has been discontinued to my knowledge. If you can find one at a discounted price, it worked better on the older non-UDMA readers. The Sandisk Ultra-III was kind of s disappointment considering its high cost. I also had to reformat it a second time. In its favor, it did include a mini-CD with some nice data recovery software on it. Again its performance may have been hindered by the card reader technology I used. The Sandisk Ultra-III, came in second place with the Lexar UDMA reader, which makes sense, since its a UDMA rated card. As far as the bargain priced cards go, you get what you pay for. None of the Kingston cards tested were very impressive. Avoid the cheap low cost, slower than 133X cards, unless you have a lot of patience, or an older camera. Note some older cameras that don't support the FAT-32 file format, can not format 2.0GB or larger flash cards.

Future Considerations: USB 3.0 is in the process of being approved by the IEEE standards body and promises to deliver 3.0Gbits per second performance similar to eSATA. eSATA on the other hand is here today, but no card readers exist to my knowledge that make use of it. This is a big disappointment, since solid state hard drives are available today with SATA interfaces on them. So SATA to flash memory chipsets must exist. Intel recently announced a break through in flash technology which promises a ten-fold increase in transfer speeds. So don't load up on present-day technology flash cards. Much faster flash chips and cards are on the way. These will probably end up in solid state hard drives at first, but should also migrate into flash cards at some later point. The problem will be to find a reader fast enough to do them justice. Again USB 3.0 may offer the solution.

Trivia: The X-speeds listed on many compact flash cards, actually refer back to the original Sony CD-ROM data transfer standard of 150KB per second read speed, or 1X CD data speed. So a 266X CF card is 266 times faster than an original 1x CD-ROM drive. Hard to believe that technology has come so far and has shrunk so small. I own a miniSD card that is smaller than a finger nail and can hold 32GB of data. Unbeliveable!



Web page last updated:

Web site Copyright © User Friendly Computers of Royal Oak - 2009