Web-exclusive article addendum: "Memory cards: designing with a full deck"
This'll be a "living" document; I'll periodically update it with evaluations I make using new small-form-factor removable mass-storage devices and under new hardware and operating-system platforms.
Brian Dipert -- EDN, May 25, 2000
May 25, 2000Digital-camera benchmarking
I've performance-tested the Type 1 CompactFlash cards listed in my article sidebar with two digital cameras; the Kodak DC-120 I used with my July 3, 1997 article, and Nikon's new Coolpix 990. Multiple sets of numbers achieve, in my mind, several objectives:
I've performance-tested the Type 1 CompactFlash cards listed in my article sidebar with two digital cameras; the KodakKodak DC-120 I used with my July 3, 1997 articleJuly 3, 1997 article, and Nikon's new Coolpix 990Coolpix 990. Multiple sets of numbers achieve, in my mind, several objectives:
-
You can correlate the write speeds seen with the cards new to this study against the performance results I saw with the cards used in the 1997 study.
-
Comparing the results obtained with the two cameras gives you an indication of to what degree the performance is limited by the card itself, versus system overhead
-
The Coolpix 990, being a newer camera, should complete its pre-storage image processing functions faster than the three-year-old Kodak (in fact, several manufacturers urged me to include a newer camera in this year's study for this very reason). However, the Nikon camera's higher resolution means that it captures larger images than the Kodak camera.
You can correlate the write speeds seen with the cards new to this study against the performance results I saw with the cards used in the 1997 study.
You can correlate the write speeds seen with the cards new to this study against the performance results I saw with the cards used in the 1997 study.
You can correlate the write speeds seen with the cards new to this study against the performance results I saw with the cards used in the 1997 study1997 study.
Comparing the results obtained with the two cameras gives you an indication of to what degree the performance is limited by the card itself, versus system overhead
Comparing the results obtained with the two cameras gives you an indication of to what degree the performance is limited by the card itself, versus system overhead
Comparing the results obtained with the two cameras gives you an indication of to what degree the performance is limited by the card itself, versus system overhead
The Coolpix 990, being a newer camera, should complete its pre-storage image processing functions faster than the three-year-old Kodak (in fact, several manufacturers urged me to include a newer camera in this year's study for this very reason). However, the Nikon camera's higher resolution means that it captures larger images than the Kodak camera.
The Coolpix 990, being a newer camera, should complete its pre-storage image processing functions faster than the three-year-old Kodak (in fact, several manufacturers urged me to include a newer camera in this year's study for this very reason). However, the Nikon camera's higher resolution means that it captures larger images than the Kodak camera.
The Coolpix 990, being a newer camera, should complete its pre-storage image processing functions faster than the three-year-old Kodak (in fact, several manufacturers urged me to include a newer camera in this year's study for this very reason). However, the Nikon camera's higher resolution means that it captures larger images than the Kodak camera.
I set the Kodak camera to its 'uncompressed' mode, in which it stores a minimally-processed 'dump' of the 836,400 rectangular-pixel Bayer pattern CCD sensor (adapted from those used with camcorders) output, at 8 bits per pixel. Rectangular-to-square pixel interpolation to create the final 1280x960 (1,228,800 pixel) image, as well as 24-bit color conversion and other processing functions, subsequently occur on the PC within the camera's Twain driver.
I set the Kodak camera to its 'uncompressed' mode, in which it stores a minimally-processed 'dump' of the 836,400 rectangular-pixel Bayer pattern CCD sensor (adapted from those used with camcorders) output, at 8 bits per pixel. Rectangular-to-square pixel interpolation to create the final 1280x960 (1,228,800 pixel) image, as well as 24-bit color conversion and other processing functions, subsequently occur on the PC within the camera's Twain driver.
In order to create files small enough for the 4-Mbyte CompactFlash cards to handle, I first set the Nikon camera to its 'fine' image quality mode, which creates a conservatively-JPEG compressed file (in this case, approximately 1.2 Mbytes in size). I then re-ran the suite of cards through the Nikon camera set to its 'hi' image quality mode, which does no JPEG compression (thereby minimizing camera pre-processing and maximizing file size) and generates a ~9.3-Mbyte file. In both 'fine' and 'hi' quality modes, I set the camera to its 'full' (2048x1536) image size option, to maximize the resulting file size, thereby accentuating any write performance differences between the cards.
In order to create files small enough for the 4-Mbyte CompactFlash cards to handle, I first set the Nikon camera to its 'fine' image quality mode, which creates a conservatively-JPEG compressed file (in this case, approximately 1.2 Mbytes in size). I then re-ran the suite of cards through the Nikon camera set to its 'hi' image quality mode, which does no JPEG compression (thereby minimizing camera pre-processing and maximizing file size) and generates a ~9.3-Mbyte file. In both 'fine' and 'hi' quality modes, I set the camera to its 'full' (2048x1536) image size option, to maximize the resulting file size, thereby accentuating any write performance differences between the cards.
The scene I shot was the same one I used in 
, a consistently-illuminated corner of my home office with shades drawn to eliminate any JPEG file size variability due to changing outside light conditions. I attached each camera to a tripod to maintain a consistent captured image, again to eliminate any JPEG file size variations that might make a given card inaccurately appear faster or slower than its peers. I also set both cameras to a comparable zoom lens focal length so that they'd capture equivalent images.
The scene I shot was the same one I used in , a consistently-illuminated corner of my home office with shades drawn to eliminate any JPEG file size variability due to changing outside light conditions. I attached each camera to a tripod to maintain a consistent captured image, again to eliminate any JPEG file size variations that might make a given card inaccurately appear faster or slower than its peers. I also set both cameras to a comparable zoom lens focal length so that they'd capture equivalent images.
I began the testing for each card by pre-formatting it using my desktop PC and a Sandisk Imagemate model SDDR-31 USB-interface CompactFlash adapter. To minimize system overhead, I first partially depressed the shutter button to allow the camera to auto-focus and capture exposure. Then, I'd initiate the stopwatch on my Suunto Vector watch at the same time that I fully depressed the shutter. When the LED quit blinking, I stopped the timer. If the card had sufficient density, I'd capture five images on it; otherwise I captured as many as the density would allow. Then I'd reformat the card and continue. After capturing 10 images and 10 times, I calculated the average, which you'll find below.
I began the testing for each card by pre-formatting it using my desktop PC and a Sandisk Imagemate model SDDR-31 USB-interface CompactFlash adapter. To minimize system overhead, I first partially depressed the shutter button to allow the camera to auto-focus and capture exposure. Then, I'd initiate the stopwatch on my SuuntoSuunto Vector watch at the same time that I fully depressed the shutter. When the LED quit blinking, I stopped the timer. If the card had sufficient density, I'd capture five images on it; otherwise I captured as many as the density would allow. Then I'd reformat the card and continue. After capturing 10 images and 10 times, I calculated the average, which you'll find below.
| Memory card | Kodak DC120 (sec) | Nikon Coolpix 990, fine (sec) | Nikon Coolpix 990, hi (sec) |
| AMD 8 Mbyte (UltraNAND) | 24 | N/A (1) | N/A (1,2) |
| Hitachi 8 Mbyte (AND) | 9.2 | 7.2 | N/A (2) |
| Hitachi 96 Mbyte (MLC AND) | 23 | 8.5 | 47.5 |
| Kingston Technology 128 Mbyte (MLC AND) | 22.4 | 6.4 | 33 |
| Lexar Media 64 Mbyte (NAND) | 10.2 | 6.8 | 29 |
| Lexar Media 160 Mbyte (NAND) | 15.5 | 4.5 | 26.5 |
| Micron Technology 10 Mbyte (NOR) | 7.1 | 4.5 | 45 |
| M-Systems 32 Mbyte (NAND) | 27.2 | 16 | 101 |
| Sandisk 15 Mbyte (NOR) | 8.5 | N/A (3) | N/A (3) |
| Sandisk 32 Mbyte (MLC NOR) | 12.2 | 7.6 | 52.5 |
| Silicon Storage Technology 8 Mbyte (NAND) | 6.5 | 4.2 | N/A, see June 15, 2000 addendum entry below (2) |
| SimpleTech 128 Mbyte (MLC AND) | 25.4 | 8.5 | 43 |
| Smart Modular Technology 4 Mbyte (NAND) | 28.5 | N/A (4) | N/A (2,4) |
Memory card
Kodak DC120 (sec)
Nikon Coolpix 990, fine (sec)
Nikon Coolpix 990, hi (sec)
Memory card
Memory card
Memory cardMemory card
Kodak DC120 (sec)
Kodak DC120 (sec)
Kodak DC120 (sec)Kodak DC120 (sec)
Nikon Coolpix 990, fine (sec)
Nikon Coolpix 990, fine (sec)
Nikon Coolpix 990, fine (sec)Nikon Coolpix 990, fine (sec)
Nikon Coolpix 990, hi (sec)
Nikon Coolpix 990, hi (sec)
Nikon Coolpix 990, hi (sec)Nikon Coolpix 990, hi (sec)
AMD 8 Mbyte (UltraNAND)
24
N/A (1)
N/A (1,2)
AMD 8 Mbyte (UltraNAND)
AMD 8 Mbyte (UltraNAND)
AMDAMDAMDAMD 8 Mbyte (UltraNAND) 8 Mbyte (UltraNAND)
24
24
2424
N/A (1)
N/A (1)
N/A (1)N/A (1)
N/A (1,2)
N/A (1,2)
N/A (1,2)N/A (1,2)
Hitachi 8 Mbyte (AND)
9.2
7.2
N/A (2)
Hitachi 8 Mbyte (AND)
Hitachi 8 Mbyte (AND)
HitachiHitachiHitachiHitachi 8 Mbyte (AND) 8 Mbyte (AND)
9.2
9.2
9.29.2
7.2
7.2
7.27.2
N/A (2)
N/A (2)
N/A (2)N/A (2)
Hitachi 96 Mbyte (MLC AND)
23
8.5
47.5
Hitachi 96 Mbyte (MLC AND)
Hitachi 96 Mbyte (MLC AND)
Hitachi 96 Mbyte (MLC AND)Hitachi 96 Mbyte (MLC AND)
23
23
2323
8.5
8.5
8.58.5
47.5
47.5
47.547.5
Kingston Technology 128 Mbyte (MLC AND)
22.4
6.4
33
Kingston Technology 128 Mbyte (MLC AND)
Kingston Technology 128 Mbyte (MLC AND)
Kingston TechnologyKingston TechnologyKingston TechnologyKingston Technology 128 Mbyte (MLC AND) 128 Mbyte (MLC AND)
22.4
22.4
22.422.4
6.4
6.4
6.46.4
33
33
3333
Lexar Media 64 Mbyte (NAND)
10.2
6.8
29
Lexar Media 64 Mbyte (NAND)
Lexar Media 64 Mbyte (NAND)
Lexar MediaLexar MediaLexar MediaLexar Media 64 Mbyte (NAND) 64 Mbyte (NAND)
10.2
10.2
10.210.2
6.8
6.8
6.86.8
29
29
2929
Lexar Media 160 Mbyte (NAND)
15.5
4.5
26.5
Lexar Media 160 Mbyte (NAND)
Lexar Media 160 Mbyte (NAND)
Lexar Media 160 Mbyte (NAND)Lexar Media 160 Mbyte (NAND)
15.5
15.5
15.515.5
4.5
4.5
4.54.5
26.5
26.5
26.526.5
Micron Technology 10 Mbyte (NOR)
7.1
4.5
45
Micron Technology 10 Mbyte (NOR)
Micron Technology 10 Mbyte (NOR)
Micron TechnologyMicron TechnologyMicron TechnologyMicron Technology 10 Mbyte (NOR) 10 Mbyte (NOR)
7.1
7.1
7.17.1
4.5
4.5
4.54.5
45
45
4545
M-Systems 32 Mbyte (NAND)
27.2
16
101
M-Systems 32 Mbyte (NAND)
M-Systems 32 Mbyte (NAND)
M-SystemsM-SystemsM-SystemsM-Systems 32 Mbyte (NAND) 32 Mbyte (NAND)
27.2
27.2
27.227.2
16
16
1616
101
101
101101
Sandisk 15 Mbyte (NOR)
8.5
N/A (3)
N/A (3)
Sandisk 15 Mbyte (NOR)
Sandisk 15 Mbyte (NOR)
SandiskSandiskSandiskSandisk 15 Mbyte (NOR) 15 Mbyte (NOR)
8.5
8.5
8.58.5
N/A (3)
N/A (3)
N/A (3)N/A (3)
N/A (3)
N/A (3)
N/A (3)N/A (3)
Sandisk 32 Mbyte (MLC NOR)
12.2
7.6
52.5
Sandisk 32 Mbyte (MLC NOR)
Sandisk 32 Mbyte (MLC NOR)
Sandisk 32 Mbyte (MLC NOR)Sandisk 32 Mbyte (MLC NOR)
12.2
12.2
12.212.2
7.6
7.6
7.67.6
52.5
52.5
52.552.5
Silicon Storage Technology 8 Mbyte (NAND)
6.5
4.2
N/A, see June 15, 2000 addendum entry below (2)
Silicon Storage Technology 8 Mbyte (NAND)
Silicon Storage Technology 8 Mbyte (NAND)
Silicon Storage TechnologySilicon Storage TechnologySilicon Storage TechnologySilicon Storage Technology 8 Mbyte (NAND) 8 Mbyte (NAND)
6.5
6.5
6.56.5
4.2
4.2
4.24.2
N/A, see June 15, 2000 addendum entry below (2)
N/A, see June 15, 2000 addendum entry below (2)
N/A, see June 15, 2000 addendum entry below (2)N/A, see June 15, 2000 addendum entry below (2)
SimpleTech 128 Mbyte (MLC AND)
25.4
8.5
43
SimpleTech 128 Mbyte (MLC AND)
SimpleTech 128 Mbyte (MLC AND)
SimpleTechSimpleTechSimpleTechSimpleTech 128 Mbyte (MLC AND) 128 Mbyte (MLC AND)
25.4
25.4
25.425.4
8.5
8.5
8.58.5
43
43
4343
Smart Modular Technology 4 Mbyte (NAND)
28.5
N/A (4)
N/A (2,4)
Smart Modular Technology 4 Mbyte (NAND)
Smart Modular Technology 4 Mbyte (NAND)
Smart Modular TechnologySmart Modular TechnologySmart Modular TechnologySmart Modular Technology 4 Mbyte (NAND) 4 Mbyte (NAND)
28.5
28.5
28.528.5
N/A (4)
N/A (4)
N/A (4)N/A (4)
N/A (2,4)
N/A (2,4)
N/A (2,4)N/A (2,4)
Notes:
Notes:
-
Received 'this card cannot be used' message when I attempted to use the card with the Nikon camera.
-
Card density too small to store file created by Nikon camera
-
Received 'image cannot be saved' message when I attempted to use the card with the Nikon camera.
-
Received 'system error' message, along with an inability to power on system, when I attempted to use the card with the Nikon camera. I've also noticed incompatibilities between this card and a few other digital cameras and other CF-enabled systems, as well as some CompactFlash readers.
Received 'this card cannot be used' message when I attempted to use the card with the Nikon camera.
Received 'this card cannot be used' message when I attempted to use the card with the Nikon camera.
Received 'this card cannot be used' message when I attempted to use the card with the Nikon camera.
Card density too small to store file created by Nikon camera
Card density too small to store file created by Nikon camera
Card density too small to store file created by Nikon camera
Received 'image cannot be saved' message when I attempted to use the card with the Nikon camera.
Received 'image cannot be saved' message when I attempted to use the card with the Nikon camera.
Received 'image cannot be saved' message when I attempted to use the card with the Nikon camera.
Received 'system error' message, along with an inability to power on system, when I attempted to use the card with the Nikon camera. I've also noticed incompatibilities between this card and a few other digital cameras and other CF-enabled systems, as well as some CompactFlash readers.
Received 'system error' message, along with an inability to power on system, when I attempted to use the card with the Nikon camera. I've also noticed incompatibilities between this card and a few other digital cameras and other CF-enabled systems, as well as some CompactFlash readers.
Received 'system error' message, along with an inability to power on system, when I attempted to use the card with the Nikon camera. I've also noticed incompatibilities between this card and a few other digital cameras and other CF-enabled systems, as well as some CompactFlash readers.
When analyzing the results, keep in mind that the flash memory inside the card is only one of several factors that influence the write performance I saw. As mentioned in my May 25 article, the memory controller, if well-designed, can assist an otherwise-slow flash memory by programming multiple bytes and erasing multiple blocks in parallel, buffering the file (or subsets of it) in temporary RAM, etc. Since I measured only performance, not power consumption, I wasn't able to expose the use of these performance boosting, but potentially battery-draining, techniques. Conversely, a poorly-designed controller can performance-strap an otherwise fast flash technology.
When analyzing the results, keep in mind that the flash memory inside the card is only one of several factors that influence the write performance I saw. As mentioned in my May 25 article, the memory controller, if well-designed, can assist an otherwise-slow flash memory by programming multiple bytes and erasing multiple blocks in parallel, buffering the file (or subsets of it) in temporary RAM, etc. Since I measured only performance, not power consumption, I wasn't able to expose the use of these performance boosting, but potentially battery-draining, techniques. Conversely, a poorly-designed controller can performance-strap an otherwise fast flash technology.
System-level optimizations and bottlenecks also impact the results. If the camera's firmware uses only a generic ATA or IDE command set to communicate with the card, the effective transfer speed may be slower than if the camera identifies the presence of a specific vendor's card, and switches to a proprietary higher-performance command set. The camera may also not be capable of feeding the card data as fast as the storage media could best-case accept it. Image processing overhead, such as color correction, interpolation and compression, also minimizes otherwise-evident card-to-card differences.
System-level optimizations and bottlenecks also impact the results. If the camera's firmware uses only a generic ATA or IDE command set to communicate with the card, the effective transfer speed may be slower than if the camera identifies the presence of a specific vendor's card, and switches to a proprietary higher-performance command set. The camera may also not be capable of feeding the card data as fast as the storage media could best-case accept it. Image processing overhead, such as color correction, interpolation and compression, also minimizes otherwise-evident card-to-card differences.
With these qualifiers in mind, I encourage you to both compare one card's results against another using the same camera, and evaluate the same card across multiple cameras and camera settings, to get a full sense of each storage technology's capabilities and limitations. In the paragraphs that follow, I'm not going to subject you to a line-by-line analysis, but I'll point out trends that I found particularly interesting. For example, the M-Systems and Smart Modular Technology cards were among the consistently slowest performers, while the Silicon Storage Technology card was the fastest, yet both employ NAND flash technology. This contrast highlights the importance of good memory controller design, as well as system-level support for that controller.
With these qualifiers in mind, I encourage you to both compare one card's results against another using the same camera, and evaluate the same card across multiple cameras and camera settings, to get a full sense of each storage technology's capabilities and limitations. In the paragraphs that follow, I'm not going to subject you to a line-by-line analysis, but I'll point out trends that I found particularly interesting. For example, the M-Systems and Smart Modular Technology cards were among the consistently slowest performers, while the Silicon Storage Technology card was the fastest, yet both employ NAND flash technology. This contrast highlights the importance of good memory controller design, as well as system-level support for that controller.
If you only looked at the Kodak numbers, you'd think that Hitachi's multi-level-cell (MLC) AND technology (used in the vendor's 96-Mbyte card) was significantly slower than the conventional AND technology used in Hitachi's 8-Mbyte card. Look at the Nikon numbers, however, and you'll probably reach a more MLC-optimistic conclusion. More generally, each card's Nikon 'fine'-derived numbers were faster than the Kodak equivalents, even though in the former case the camera was JPEG-compressing the image prior to storing it, and in spite of the fact that even with JPEG compression, the resultant Nikon-generated files were still roughly 50% larger than the Kodak ones.
If you only looked at the Kodak numbers, you'd think that Hitachi's multi-level-cell (MLC) AND technology (used in the vendor's 96-Mbyte card) was significantly slower than the conventional AND technology used in Hitachi's 8-Mbyte card. Look at the Nikon numbers, however, and you'll probably reach a more MLC-optimistic conclusion. More generally, each card's Nikon 'fine'-derived numbers were faster than the Kodak equivalents, even though in the former case the camera was JPEG-compressing the image prior to storing it, and in spite of the fact that even with JPEG compression, the resultant Nikon-generated files were still roughly 50% larger than the Kodak ones.
Another odd AND observation bears mentioning. Hitachi's 96-Mbyte, Kingston's 128-Mbyte, and SimpleTech's 128-Mbyte cards all employ Hitachi's 256-Mbit MLC AND chip and matching memory controller. However, Kingston's card was consistently faster than the other two. This discrepency may reflect faster-writing chips inside the card. It might also reflect a controller running at a higher clock rate, or one otherwise performance-tuned.
Another odd AND observation bears mentioning. Hitachi's 96-Mbyte, Kingston's 128-Mbyte, and SimpleTech's 128-Mbyte cards all employ Hitachi's 256-Mbit MLC AND chip and matching memory controller. However, Kingston's card was consistently faster than the other two. This discrepency may reflect faster-writing chips inside the card. It might also reflect a controller running at a higher clock rate, or one otherwise performance-tuned.
Lexar Media's '10x' 160-Mbyte card was actually slower than the vendor's '8x' 64-Mbyte card with the Kodak camera, although Lexar Media's higher-is-faster marketing claims matched reality on the Nikon camera. Many of you will closely scrutinize the Sandisk-vs-Lexar Media results, since these two companies are among the most aggressive competitors of all the vendors who participated in my study. When you do, keep in mind that the Sandisk 32-Mbyte card uses MLC flash technology, which (all other factors being equal) will take longer to write a given file than a conventional technology such as that used in the 15-Mbyte Sandisk card or either of the Lexar cards.
Lexar Media's '10x' 160-Mbyte card was actually slower than the vendor's '8x' 64-Mbyte card with the Kodak camera, although Lexar Media's higher-is-faster marketing claims matched reality on the Nikon camera. Many of you will closely scrutinize the Sandisk-vs-Lexar Media results, since these two companies are among the most aggressive competitors of all the vendors who participated in my study. When you do, keep in mind that the Sandisk 32-Mbyte card uses MLC flash technology, which (all other factors being equal) will take longer to write a given file than a conventional technology such as that used in the 15-Mbyte Sandisk card or either of the Lexar cards.
Just as my study didn't quantify power consumption, it didn't compare per-megabyte-card pricing. Lower media cost at a given density is the most important claim offered by vendors employing MLC flash, and may in fact be more critical in your application than highest-possible write speeds. While we're on the subject of Sandisk, I was surprised to find that the Nikon camera refused to talk to the 32-Mbyte card, because Sandisk was the initial developer of the CompactFlash memory card form factor and interface. I wish I could have gotten the Sandisk conventional flash-based card to work, to see how it stacked up against the Lexar Media cards with the Kodak camera.
Just as my study didn't quantify power consumption, it didn't compare per-megabyte-card pricing. Lower media cost at a given density is the most important claim offered by vendors employing MLC flash, and may in fact be more critical in your application than highest-possible write speeds. While we're on the subject of Sandisk, I was surprised to find that the Nikon camera refused to talk to the 32-Mbyte card, because Sandisk was the initial developer of the CompactFlash memory card form factor and interface. I wish I could have gotten the Sandisk conventional flash-based card to work, to see how it stacked up against the Lexar Media cards with the Kodak camera.
In general, the magnitude of the discrepency between fastest and slowest cards, as well as the inconsistency of fastest-to-slowest trends from one camera (and setting) to another, were both more significant than I'd predicted in advance. I'd expected that the smallest cards would be at a disadvantage compared to their larger siblings, because there was less 'spare' media available to enable multi-component parallel operations, as well as to hide the occasionally-required cleanup of 'dirty' bytes and blocks caused by deleted files. I'd strived to obtain similar-density cards from all the study participants, although this goal ended up being unachievable. However, the results didn't seem to bear out my preconception that smaller cards would be unfairly handicapped.
In general, the magnitude of the discrepency between fastest and slowest cards, as well as the inconsistency of fastest-to-slowest trends from one camera (and setting) to another, were both more significant than I'd predicted in advance. I'd expected that the smallest cards would be at a disadvantage compared to their larger siblings, because there was less 'spare' media available to enable multi-component parallel operations, as well as to hide the occasionally-required cleanup of 'dirty' bytes and blocks caused by deleted files. I'd strived to obtain similar-density cards from all the study participants, although this goal ended up being unachievable. However, the results didn't seem to bear out my preconception that smaller cards would be unfairly handicapped.
I mentioned in the main article that some of the cards came from my initial study of three years back. Rest assured that in the interim, the cards had collected dust on my office shelf; I hadn't used them (which might, by heavily cycling the flash memory inside the card, degrade the cards' write performance). Notice any other interesting interpretations of the results? Drop me a line and I'll include your comments in the next revision of this addendum. I'll also be adding benchmarking results using desktop and notebook PCs, which will enable me to include the various SmartMedia cards, IBM MicroDrive, Iomega Clik! drive and Sony Memory Stick.
I mentioned in the main article that some of the cards came from my initial study of three years back. Rest assured that in the interim, the cards had collected dust on my office shelf; I hadn't used them (which might, by heavily cycling the flash memory inside the card, degrade the cards' write performance). Notice any other interesting interpretations of the results? Drop me a lineline and I'll include your comments in the next revision of this addendum. I'll also be adding benchmarking results using desktop and notebook PCs, which will enable me to include the various SmartMedia cards, IBM MicroDrive, Iomega Clik! drive and Sony Memory Stick.
More information on Nikon's Coolpix 990 digital camera
More information on Nikon's Coolpix 990 digital camera
The pace of technology innovation really is amazing. Just three years ago, $1000 bought you Kodak's DC120, with a native resolution of just under 1 megapixel (just over a megapixel interpolated). Megapixel-resolution cameras are well below $500 today, and $1000 (or less) gets you the smaller, lighter and faster 3.34 megapixel Nikon Coolpix 990, with a whole host of improved and added features, including:
The pace of technology innovation really is amazing. Just three years ago, $1000 bought you Kodak's DC120, with a native resolution of just under 1 megapixel (just over a megapixel interpolated). Megapixel-resolution cameras are well below $500 today, and $1000 (or less) gets you the smaller, lighter and faster 3.34 megapixelThe pace of technology innovation really is amazing. Just three years ago, $1000 bought you Kodak's DC120, with a native resolution of just under 1 megapixel (just over a megapixel interpolated). Megapixel-resolution cameras are well below $500 today, and $1000 (or less) gets you the smaller, lighter and faster 3.34 megapixel Nikon Coolpix 990, with a whole host of improved and added features, including:
-
3x optical zoom with variable aperatures, supplemented by a 4x digital zoom
-
256 element matrix metering auto-exposure, along with optional shutter-priority, aperature-priority, and fully manual exposure modes
-
Multi-area autofocus, as well as optional manual focus (including macro to 0.8 inches!)
-
A 40 second video capture mode at 15 fps (QuickTime file format) or continuous image capture at 2 fps (full resolution) to 30 fps (QVGA resolution).
-
Built-in USB connectivity
-
A diverse family of accessory lenses and external flash units
-
NTSC and PAL video output capability
-
An extensive built-in set of options for user interface, exposure, image editing and display
3x optical zoom with variable aperatures, supplemented by a 4x digital zoom
3x optical zoom with variable aperatures, supplemented by a 4x digital zoom
3x optical zoom with variable aperatures, supplemented by a 4x digital zoom
256 element matrix metering auto-exposure, along with optional shutter-priority, aperature-priority, and fully manual exposure modes
256 element matrix metering auto-exposure, along with optional shutter-priority, aperature-priority, and fully manual exposure modes
256 element matrix metering auto-exposure, along with optional shutter-priority, aperature-priority, and fully manual exposure modes
Multi-area autofocus, as well as optional manual focus (including macro to 0.8 inches!)
Multi-area autofocus, as well as optional manual focus (including macro to 0.8 inches!)
Multi-area autofocus, as well as optional manual focus (including macro to 0.8 inches!)
A 40 second video capture mode at 15 fps (QuickTime file format) or continuous image capture at 2 fps (full resolution) to 30 fps (QVGA resolution).
A 40 second video capture mode at 15 fps (QuickTime file format) or continuous image capture at 2 fps (full resolution) to 30 fps (QVGA resolution).
A 40 second video capture mode at 15 fps (QuickTime file format) or continuous image capture at 2 fps (full resolution) to 30 fps (QVGA resolution).
Built-in USB connectivity
Built-in USB connectivity
Built-in USB connectivity
A diverse family of accessory lenses and external flash units
A diverse family of accessory lenses and external flash units
A diverse family of accessory lenses and external flash units
NTSC and PAL video output capability
NTSC and PAL video output capability
NTSC and PAL video output capability
An extensive built-in set of options for user interface, exposure, image editing and display
An extensive built-in set of options for user interface, exposure, image editing and display
An extensive built-in set of options for user interface, exposure, image editing and display
The biggest downside I noted when using the Coolpix 990 was the long (~1 sec) delay between full depression of the shutter and when the camera actually took the picture. This lag time wasn't a problem when the camera was on a tripod and I was shooting still life scenes. However, the delay was more frustrating when handholding the camera or shooting action subjects (two-legged or, in my case, four legged dog and cat 'children', for example). I'd expected to experience a lesser, not greater, delay compared to my much older DC120. The lag time was less in the manual recording (M-REC) mode versus the fully automatic (A-REC) mode, however.
The biggest downside I noted when using the Coolpix 990 was the long (~1 sec) delay between full depression of the shutter and when the camera actually took the picture. This lag time wasn't a problem when the camera was on a tripod and I was shooting still life scenes. However, the delay was more frustrating when handholding the camera or shooting action subjects (two-legged or, in my case, four legged dog and cat 'children', for example). I'd expected to experience a lesser, not greater, delay compared to my much older DC120. The lag time was less in the manual recording (M-REC) mode versus the fully automatic (A-REC) mode, however.June 15, 2000Digital-camera benchmarking
Silicon Storage Technology provided me with a larger 16-Mbyte CompactFlash card so that I could test it with the Nikon Coolpix 990 in 'hi' resolution mode. Its average write time was 30.2 sec, one of the fastest of the group (with the exception of the two higher-performance Lexar Media cards). I didn't bother retesting the 16-Mbyte SST card with the Kodak DC120, or with the Nikon in 'fine' resolution mode.
Silicon Storage Technology provided me with a larger 16-Mbyte CompactFlash card so that I could test it with the Nikon Coolpix 990 in 'hi' resolution mode. Its average write time was 30.2 sec, one of the fastest of the group (with the exception of the two higher-performance Lexar Media cards). I didn't bother retesting the 16-Mbyte SST card with the Kodak DC120, or with the Nikon in 'fine' resolution mode.
Speaking of Lexar Media, the company gave me some interesting feedback, related to my qualifier in the article and addendum that system overhead may make some cards appear faster or slower than they really are. Apparently, one reason for the slowdown when using larger-capacity cards with the Kodak DC120 is that the camera does sector-by-sector reads of the FAT (File Allocation Tables) before beginning its write of the picture. Because single-sector reads are slow, and because larger cards have larger FAT tables, the DC120 will experience a significant decrease in performance with larger cards. By comparing cards of equal capacity with this camera, you can get a more accurate test of the cards themselves.
Speaking of Lexar Media, the company gave me some interesting feedback, related to my qualifier in the article and addendum that system overhead may make some cards appear faster or slower than they really are. Apparently, one reason for the slowdown when using larger-capacity cards with the Kodak DC120 is that the camera does sector-by-sector reads of the FAT (File Allocation Tables) before beginning its write of the picture. Because single-sector reads are slow, and because larger cards have larger FAT tables, the DC120 will experience a significant decrease in performance with larger cards. By comparing cards of equal capacity with this camera, you can get a more accurate test of the cards themselves.
Check this addendum often, and drop me an e-mail with any feedback.Check this addendum often, and drop me an e-maile-mail with any feedback.
