Memory Card Speed: You Thought Your Laptop Slots Were Fast?

[Jason Dunn]

If you've got a newer laptop, it probably has some sort of memory card slot - SD, CF, perhaps Memory Stick. I'm preparing to do a speed shoot-out comparison of 512 MB SD cards from various manufacturers, and I wanted to lock down a testing methodology that was fair and accurate. My Fujitu P5010D has built-in CompactFlash and Secure Digital slots, and I thought that would be the best way to test the card speeds. After all, the SD slot is integrated into the laptop, directly onto the system bus (right to the metal so to speak), and surely that must be the fastest method of moving data to and from an SD card, right? Wrong. 8O

After a few tests, I've discovered something quite ugly: using a USB 2.0 external memory card reader is 387% faster than the SD slot on my laptop. Using a Sandisk USB 2.0 (http://www.sandisk.com/retail_eightinone.html) memory card reader, I achieved the following speeds using HD_Speed from Steelbytes (http://www.steelbytes.com/) while testing a Sandisk Ultra II 512 MB SD card with 512 KB block size reads:

Built-in SD Slot: 1586 KB/s
Sandisk Reader: 6144 KB/s

Quite the difference, no? Things were even worse with the CF slot where a low-end 128 MB no-name CF card gave me 1484 KB/s in the USB 2.0 reader and a meagre 972 KB/s in the internal slot.

I'd be very interested to hear from owners of other brands of laptops to know whether this is something common to all laptops, or if Fujitsu cheaped out and other laptops can do a better job at this. Although I'll certainly put up with slower speeds in exchange for not having to carry a memory card reader with me, I'd be much happier if the memory card slots were faster!

[Lee Yuan Sheng]

It's common Jason. My IBM X20 reads CF pretty slowly as well.

The best test of speed should be in the devices that will use the cards, so I think it's time to get the cameras out and popping the cards into them.

[James Fee]

The best test of speed should be in the devices that will use the cards, so I think it's time to get the cameras out and popping the cards into them.
That would be very interesting. How would you go about such a test?

[Lee Yuan Sheng]

Well, since one can't have all the cameras all the time to test all the cards, I suppose it could be a database with contributions from everyone.

Either that, or a few of the review team passes the cards around themselves and posts the results. I'm sure there'll be enough different digital SLRs around the review team. =P

[Jason Dunn]

Well, since one can't have all the cameras all the time to test all the cards, I suppose it could be a database with contributions from everyone.

I think he means how would you test the card speeds with the camera? Take the exact same picture over and over, then time the write speed of the camera? What about read speeds? I agree this would be the best way of testing it, because it's the most realistic, but I can't come up with a good method of actually performing the tests.

[Lee Yuan Sheng]

Read speeds? Well, that can be tested via a card reader, since I suppose most people use a card reader to retrive their images? Or do most of you use the camera? (time for a poll; Gary, where are you..)

As for writing speeds, switch to continous, fire a full burst, time it, then calculate the write speed via the size of the files divided by the time taken to write them to the card.

[Jason Dunn]

As for writing speeds, switch to continous, fire a full burst, time it, then calculate the write speed via the size of the files divided by the time taken to write them to the card.

As someone who's done a lot of benchmarking, what you're suggesting would be insanely agonizing. You'd have to shoot the same scene over and over while changing the camera's image quality setting (to simulate different file size speeds, and you'd have to test each card with each camera in this way. The sheer number of permutations is mind-boggling. 8O

You're welcome to give it a go, but I'm going to stay far, far away from that. :lol:

[Lee Yuan Sheng]

No, you need to run the test just once. ok, a few times to average out reaction times, but that's it. Why do you need to use different compressions etc?

[James Fee]

What I was getting at was what does it prove if the D100 is much faster than the 4300 at writing a JPG? Why is it faster? I can guess, but the why would need to be answered before you could come up with a test.

[Lee Yuan Sheng]

I'm didn't say I want to prove if a camera is faster than another. I'm saying that some cards perform differently in some cameras, and using one camera or a card reader is not a reliable indication of a card's speed in your camera.

[James Fee]

I'm didn't say I want to prove if a camera is faster than another. I'm saying that some cards perform differently in some cameras, and using one camera or a card reader is not a reliable indication of a card's speed in your camera.
Oh I know, but one might blame problems on the card, not the camera.

Come to think of it, why would that matter? If a camera has trouble with a card for whatever reason, it should be avoided.

[Lee Yuan Sheng]

Oh no, I didn't mean troublesome cards. What I meant to say is that not all cards perform at the same speeds in all cameras. A Lexar might be faster in one camera, but a Sandisk Ultra II faster in another.

[Jason Dunn]

No, you need to run the test just once. ok, a few times to average out reaction times, but that's it. Why do you need to use different compressions etc?

Because you're assuming that an SD card will write a 1 MB JPEG image to the card as fast as a 16 MB RAW image - and based on my tests so far, that's not the case. I'm testing 256 KB, 1 MB and 16 MB synthetic file sizes, and the results vary widely.

Test measurement would be tricky - you'd have to shoot the same scene each time, and any variation in file size would invalidate the test results.

If there was a way to load software onto the camera (I remember my old Kodak DC265 could load scripts) and generate synthetic files for reading/writing and then measure it, this method would work.

[Jason Dunn]

I didn't say I want to prove if a camera is faster than another. I'm saying that some cards perform differently in some cameras...

If a memory card performs differently in different cameras, that should be because of the camera, not the memory card. In benchmarking you need to have one or more constants and one or more variables. In this case, my variable is the SD card - the constant is the laptop, the USB 2.0 card reader, and the file sizes being read/written.

What you're suggesting is different cards (variable) in different cameras (variable) shooting different scenes (variable). In the overall scheme, that test would be irrelevant. At most, it would tell you how ONE camera performed with different memory cards - which is perfectly valid, but that wouldn't help people who own other cameras.

...and using one camera or a card reader is not a reliable indication of a card's speed in your camera.

That's correct. This test is strictly confined to determining the speeds of various brands of 512 MB SD cards in isolation of other factors (cameras). Think about it like this: you already own a camera, and you're trying to decide which brand of SD card to buy. This test will help you know which card is the fastest. That's all I'm trying to accomplish. :-) You can change the brand you buy, but you can't change your camera without buying a new one (I doubt most would in any case).

Would it be useful to know which camera has the fastest write engine on the market by using the same SD card (constant) with different cameras (variable) in a controlled environment taking the same picture (constant)? Sure, absolutely! I'm sure some cameras are faster than others when it comes to writing images. But I don't have the resources to take on such a complicated task. This SD card test is tedious enough as-is. ;-)

Hope that explains my approach and goals better. :D

[Lee Yuan Sheng]

I'm not getting my point across, so I'll just leave things as they are. Happy testing!

[Jason Dunn]

I'm not getting my point across, so I'll just leave things as they are. Happy testing!

I must be missing something then - I'll listen if you want to try again. :-)

[Janak Parekh]

I'm not getting my point across, so I'll just leave things as they are. Happy testing!
I think I understand your argument, but do you have links to some numbers demonstrating this? I'd like to know how the variances play. I understand cameras might have different write patterns, but I'd be surprised the difference would be that substantial as opposed to an "across-the-board" calculation.

--janak

[Neil Enns]

There's a CF card adapter you can buy that runs at faster speeds than regular adapters. It wasn't available in the U.S. for a long time, but I think there are distributors that carry it now. Here's the company's site:

http://www.delkin.com/news/press/cardbus_pressrelease.htm

I know there was a post on Gizmodo or something about it a while back, but I can't find it :(

Neil

[bbarker]

After a few tests, I've discovered something quite ugly: the SD slot on my laptop is 387% slower than using a USB 2.0 external memory card reader.
I'm curious about that math. If it were half as fast, it would be 50% slower. Wouldn't 100% slower be zero?

[Jason Dunn]

After a few tests, I've discovered something quite ugly: the SD slot on my laptop is 387% slower than using a USB 2.0 external memory card reader.
I'm curious about that math. If it were half as fast, it would be 50% slower. Wouldn't 100% slower be zero?

Well, math has never been my strong point, so perhaps I screwed that up. ;-)

6144 divided by 1586 = 3.87, or
1586 muliplied by 3.87 = 6144

Thus, the laptop slot is either slower by a factor of 3.87 or the USB 2.0 memory card reader is faster by a factor of 3.87. If something is 100% faster than something else, it's twice as fast (ie 1000 KB/s at 100% faster is 200 KB/s).

Does that make sense now, or am I the one that's confused? :lol:

[bbarker]

The question wasn't in jest. I've always wondered when people say something is more than 100% slower.
If something is 100% faster than something else, it's twice as fast (ie 1000 KB/s at 100% faster is 200 KB/s).
If it's 100% slower, what is it? Ie, 50% as fast as 1000 KB/s is 500 KB/s. Is that the same as 50% slower? If so, wouldn't 100% slower be 0 KB/s? Then what would 387% slower be? How can you get slower than zero?

I hope someone explains this so even I can understand.

[Jason Dunn]

The question wasn't in jest. I've always wondered when people say something is more than 100% slower.

Well, that was probably poor wording on my part. ;-) I should have said it was 3.87 times slower than the USB 2.0 memory card reader. Hope that clears it up!

[Janak Parekh]

Right -- it's (1/3.87 * 100)% as fast, or ((1 - 1/3.87) * 100)% slower, or 74.1% slower. Or something. ;)

--janak

[bbarker]

Well, that was probably poor wording on my part. ;-) I should have said it was 3.87 times slower than the USB 2.0 memory card reader. Hope that clears it up!
That doesn't seem right either. Wouldn't 0.5 times slower be 50 percent slower, or half as fast? If that is the case, then 1 times slower would be 100% slower, or zero.
Right -- it's (1/3.87 * 100)% as fast, or ((1 - 1/3.87) * 100)% slower, or 74.1% slower. Or something. :wink:
That sounds more accurate. I don't think anything can get any slower than 100% slower or "1 times" slower.

Off-topic, I know. But this has always confused/bothered me.

[Janak Parekh]

Well, that was probably poor wording on my part. ;-) I should have said it was 3.87 times slower than the USB 2.0 memory card reader. Hope that clears it up!
That doesn't seem right either. Wouldn't 0.5 times slower be 50 percent slower, or half as fast? If that is the case, then 1 times slower would be 100% slower, or zero.
Mathematically, you might be accurate, but I believe that Jason's statement above is a common characterization.

--janak

[Jason Dunn]

That sounds more accurate. I don't think anything can get any slower than 100% slower or "1 times" slower. Off-topic, I know. But this has always confused/bothered me.

You're right, nothing can be 100% slower - it was my error.

[bbarker]

Mathematically, you might be accurate, but I believe that Jason's statement above is a common characterization.
Yes. And I'm on a Quixotic mission to right that wrong in this world! :ninja:

[Jonathon Watkins]

Yes. And I'm on a Quixotic mission to right that wrong in this world! :ninja:

.... one slow SD disk at a time! :wink:

This is the reason I like my external Crucial USB 2 reader of course! :D

[ctmagnus]

What about adapters in PC Card slots? How fast are they?