Burnt by a slower card (or carelessness) on my 20D...

[FlyingPete]

Well, despite earlier comments about not requiring fast cards for the 20D due to the size of the buffer, I was shooting in a situation on Saturday morning where a faster card (or a little more care on my part ;) ) costed be a number of shots.

I was doing some portrait work, and shooting in raw to my 1GB card (around 88 shots including a JPG in 'M' resolution). I was using rapid fire up to about five shots, I was happily shooting away, then my camera stopped, I had a look at the display CF FULL it read, so I quickly flipped the door :eek: to pop the card and replace it with my second card. At about the instant I popped the door, I realised that the camera hadn't debuffered around three shots, and remembered reading a gripe on DPreview.com about loosing shots when the door was opened, bye bye shots, nice knowing you!

Perhaps if I had a quicker card then it might have debuffered (I was using a Kingston Pro - 66x), then again I should have been patient before flipping the door open. I promise never to do it again :(

[slin100]

With a faster card, the camera might have realized that the card was full a little sooner. Either way, I think you would have lost the shots.

[JZaun]

Check here. Even the fastest Kingston tested was only half as fast as the fastest CF on the 20D shooting raw.

http://www.robgalbraith.com/bins/multi_page.asp?cid=6007-7303

JZ

[pcasciola]

A fast card might have helped, but even on the speedy 20D it takes 9 seconds to empty a completely full buffer, so if you're in a hurry, and forget and pull the CF card right after you see the CF full message, there's not a card or camera made today that will save you in that situation. I wonder if any of the 1-series have anyway of protecting against this. A CF door that locks while the buffer is unloading would be a nice custom function.

[RDKirk]

Perhaps if I had a quicker card then it might have debuffered (I was using a Kingston Pro - 66x), then again I should have been patient before flipping the door open. I promise never to do it again :(

Faster cards don't get you around the buffer limit with the 20D fast enough not to feel the limit, unless you slow your shooting down to about 2fps.

[PacAce]

Well, despite earlier comments about not requiring fast cards for the 20D due to the size of the buffer, I was shooting in a situation on Saturday morning where a faster card (or a little more care on my part ;) ) costed be a number of shots.

I was doing some portrait work, and shooting in raw to my 1GB card (around 88 shots including a JPG in 'M' resolution). I was using rapid fire up to about five shots, I was happily shooting away, then my camera stopped, I had a look at the display CF FULL it read, so I quickly flipped the door :eek: to pop the card and replace it with my second card. At about the instant I popped the door, I realised that the camera hadn't debuffered around three shots, and remembered reading a gripe on DPreview.com about loosing shots when the door was opened, bye bye shots, nice knowing you!

Perhaps if I had a quicker card then it might have debuffered (I was using a Kingston Pro - 66x), then again I should have been patient before flipping the door open. I promise never to do it again :(

A 1DmkII is rated to write at less than 40x. I doubt the 20D could do any better. Your Kingston Pro at 66x is more than fast enough for the 20D.

[pcasciola]

A 1DmkII is rated to write at less than 40x. I doubt the 20D could do any better. Your Kingston Pro at 66x is more than fast enough for the 20D.
You can't go by the manufacturer's speed rating, and up until the newest firmware release for the 1DMkII, the 20D was actually faster. Now both cameras are over 5MB/sec. The Kingston Pro Elite 66x writes to the 20D at 3.8MB/sec, which is actually 25x, while the Sandisk Extreme II writes at almost 6MB/sec, over a 50% improvement.

If you want a better source of card speeds than using the inflated x-ratings on the CF cards, check here:
http://www.robgalbraith.com/bins/multi_page.asp?cid=6007-7303
(http://www.robgalbraith.com/bins/multi_page.asp?cid=6007-6896)

[FlyingPete]

You can't go by the manufacturer's speed rating, and up until the newest firmware release for the 1DMkII, the 20D was actually faster. Now both cameras are over 5MB/sec. The Kingston Pro Elite 66x writes to the 20D at 3.8MB/sec, which is actually 25x, while the Sandisk Extreme II writes at almost 6MB/sec, over a 50% improvement.

If you want a better source of card speeds than using the inflated x-ratings on the CF cards, check here:
http://www.robgalbraith.com/bins/multi_page.asp?cid=6007-7303
(http://www.robgalbraith.com/bins/multi_page.asp?cid=6007-6896)

Yep checked Rob's site before making my purchases, I reasoned (and so far correctly except for this little incident) that with the large buffer I didn't need the extra speed of the SanDisk, esspcially for the extra cost. The company I work for during the day is a reseller of Kingston, so I get it very cheaply, around 50% cheaper than the equivillent SanDisk.

From now on, I will be waiting until the buffer is clear before swapping cards, sounds like a faster cards may have only save me an extra shot, as you said there is not a card made today that would have saved me from that. If it had been film, I would have had to wait for it to rewind!

Makes you wonder if Canon listens to 'the people', Phil Askey at dpreview.com has been on about the card door since the D30, even some of the compacts have features to save you from such screw ups.

[theflyingkiwi]

I think this is the case of being patient. At my last shooting experence, I was waiting for the card to be come full and the camera was taking it's time. like 1-1.5 mins to figure out if there was enought space.

Even tho it was anonying, I think it is a case that we expect equipment to react as fast as we do, what a shame it doesn't.

[pcasciola]

The Sandisk Extreme III is a little "extreme" in price for the modest performance gains it offers. I use the Ultra II and original Extremes, which can be had for around $75 last time I checked and are pretty close to the top of Galbraith's lists for the 20D/1DMkII/1DsMkII.

Unfortunately (well, maybe), Sandisk has switched over to a cheaper chip for their latest Ultra II line, so I'm not sure how they'll pan out in terms of reliability.

[PacAce]

You can't go by the manufacturer's speed rating, and up until the newest firmware release for the 1DMkII, the 20D was actually faster. Now both cameras are over 5MB/sec. The Kingston Pro Elite 66x writes to the 20D at 3.8MB/sec, which is actually 25x, while the Sandisk Extreme II writes at almost 6MB/sec, over a 50% improvement.

If you want a better source of card speeds than using the inflated x-ratings on the CF cards, check here:
http://www.robgalbraith.com/bins/multi_page.asp?cid=6007-7303
(http://www.robgalbraith.com/bins/multi_page.asp?cid=6007-6896)

Well, it seems like the 1GB Kingston Pro Elite that was test by Rob was the 40X version. Now, I'm not sure if there is or isn't a 66x version of the Kingston Pro Elite, but if there is, I would imagine that it would do a little better then the 40x version, don't you think? (BTW, how do you know that the version in Rob's test is the 66x version?)

At any rate, looking at Rob's test results, the 1DmkII is not able to take advantage of the faster speeds of any of the cards that are rated at higher than 50x. But that doesn't necessarily mean that the manufacturers ratings are over inflated, does it? A poorly performing card on one camera may perform much better in another camera or a totally different type of devide, such as a PDA.

[FlyingPete]

Well, it seems like the 1GB Kingston Pro Elite that was test by Rob was the 40X version. Now, I'm not sure if there is or isn't a 66x version of the Kingston Pro Elite, but if there is, I would imagine that it would do a little better then the 40x version, don't you think? (BTW, how do you know that the version in Rob's test is the 66x version?)

At any rate, looking at Rob's test results, the 1DmkII is not able to take advantage of the faster speeds of any of the cards that are rated at higher than 50x. But that doesn't necessarily mean that the manufacturers ratings are over inflated, does it? A poorly performing card on one camera may perform much better in another camera or a totally different type of devide, such as a PDA.

Kingston don't t rate their cards with the 'x' format, the 66x is based on two facts, the first that x=150KB, and second they have a claimed throughput of 10MB/s read, and 8MB/s write. 10MB/s / 150K/s =66.666, but they are really only around 50x write based on the 8MB figure.

The normal (non Elite Pro) are rated at 3.7MB/s read/1.5MB/s write, so they are 24x or 10x depending on how you look at it. I have a few of these cards, and they seem to work fine in the 20D due to the large buffer size.

[PacAce]

Kingston don't t rate their cards with the 'x' format, the 66x is based on two facts, the first that x=150KB, and second they have a claimed throughput of 10MB/s read, and 8MB/s write. 10MB/s / 150K/s =66.666, but they are really only around 50x write based on the 8MB figure.

The normal (non Elite Pro) are rated at 3.7MB/s read/1.5MB/s write, so they are 24x or 10x depending on how you look at it. I have a few of these cards, and they seem to work fine in the 20D due to the large buffer size.

That's fair enough. That's why I asked because I couldn't find anything on a Kingston Elite Pro rated at 66x. However, I did find many references to the kingston Elite Pro (made by Toshiba) that's rated at 6+ MB read and 5+ MB write, which makes it about a 40X.

[pcasciola]

At any rate, looking at Rob's test results, the 1DmkII is not able to take advantage of the faster speeds of any of the cards that are rated at higher than 50x. But that doesn't necessarily mean that the manufacturers ratings are over inflated, does it? A poorly performing card on one camera may perform much better in another camera or a totally different type of devide, such as a PDA.

That's my point, though. The speed ratings are useless when trying to pick a card that will be used in a digital camera, and luckily we have Galbraith's benchmarks to go by. A good example of the 40x/80x farse is Lexar. I feel sorry for the people than ran out and bought Lexar's first edition 80x card based on the 80x rating, only to find out it was performing more like a 45x card. Apparantly the second edition is much better, but how many people got stuck paying top dollar for a brand new "80x" card that is about 35-40% slower than people who waiting a couple of months and got the newer, faster model of the same card for about half the price. Bottom line is, I completely ignore the speed ratings, and go by Galbraith's benchmarks for my camera to calculate the speed differences.

[Kostyanych]

Btw, I am thinking about Lexar 80x vs Sandisk Ultra II....
Galbraith's benchmarks shows nearly the same performance for Canon 20D. But I have read somewhere that Sandisk CF cards have high latency. I saw some benchmarks showing Random Seek benchmark. And Sandisk was the last one in the row. But all the rest results were nearly equal.
I think that benchmark was pretty old. Anyway I didn't find anything today.
Does anybody know any recent benchmark which includes Random Seek tests? Does it any matter for us?
What card do you prefer: Sandisk Ultra II or Lexar 80x? (I hope we can buy only the second edition of Lexar now.)

I am really confused...

[Ikinaa]

question : suppose you take 10 shots in burst-mode, but the card has only enough place to write 5 pics to the card. So... the cam shows 'CF FULL' and it keeps the 5 non written shots in the buffer.
What do you do?
If you open the door, the shots are lost.
If you don't open the door, the shots are in the buffer but you can't change the card.
Is there a solution to the problem?

[Longwatcher]

question : suppose you take 10 shots in burst-mode, but the card has only enough place to write 5 pics to the card. So... the cam shows 'CF FULL' and it keeps the 5 non written shots in the buffer.
What do you do?
If you open the door, the shots are lost.
If you don't open the door, the shots are in the buffer but you can't change the card.
Is there a solution to the problem?

To date my camera won't let me take pictures if there is not enough room on the CF card. It seems to presume that the file will be maximum size when it takes it and knowing how much room is left on the card will allocate appropriately. As it actually writes to the card it sees that more space is available (since files are never max size) and I sometimes get one or two more shots at the end. So the above will never happen (or at least not with a D60, 10D, or 1DsMkII)

for Kostyanych,
As to Lexar versus Sandisk. I just would like to mention that I don't trust my Lexar cards with WA, but do trust my Sandisk cards. The reason is I know I have lost some images when I used the Lexar card, although it only seems to happen if I completely fill the card. Have not had similiar problems with any Sandisk card or Lexar without WA. I am apparently not the only one who has experienced this. This has occurred on both 10D and 1DsMkII.

[PacAce]

Btw, I am thinking about Lexar 80x vs Sandisk Ultra II....
Galbraith's benchmarks shows nearly the same performance for Canon 20D. But I have read somewhere that Sandisk CF cards have high latency. I saw some benchmarks showing Random Seek benchmark. And Sandisk was the last one in the row. But all the rest results were nearly equal.
I think that benchmark was pretty old. Anyway I didn't find anything today.
Does anybody know any recent benchmark which includes Random Seek tests? Does it any matter for us?
What card do you prefer: Sandisk Ultra II or Lexar 80x? (I hope we can buy only the second edition of Lexar now.)

I am really confused...

As PCASCIOLA stated in his post, it's better to consult Rob's benchmark test for selecting compact flash card for a particular camera application. However, your mileage may vary if you get the same card and use it in a different camera, like when you upgrade the camera in the future.

Regarding the manufacturer ratings, it's not they are over-rated by the manufacturers. It's that they use different standards for determining them. For example, the Lexar 80x cards are based on the use of WA technology to determine it's top speed. WA technology is good for writing (and reading) huge bursts of data as in video applications. And, no doubt, that's how they tested the card to achieve their ratings. However, these tests are meaningless in a digital camera application where the largest burst of data in a single shot would be about, say, 8MB for a 20D or 1DmkII. And this is using the WA technology. The Canon cameras do not take advantage of WA technology so they will never realize the 80x rating. Some Nikon cameras do, but I doubt they'll be able to utilize the full 80x speed of the card even if the camera could write that fast, and they can't, because the burst data size they generate still aren't comparable in size to those generated by video applications.

Anyway, getting back to your question re random seek performance, my personal opinion is that results from randam seek tests will not really be meaningful for camera applications. Why? Because the time when you care about the performance of a card is when you are taking pictures. You want to get the image to the card as quickly as possible. Image file writes to cards are sequential write operations, not random seeks (which, btw, are read operations). Granted, there will be some random seek operations performed even when writing images to cards, as when updating the FAT directory, but these are inconsequential compared to the actual write operation.

[Kostyanych]

Thanks, PacAce. You are right that random access speed is not so much important for digital cameras.... I was just looking for some advantages from Lexar...
BUT the reliability is very important. (Thanks to Longwatcher for the caution).
So it seems to me I will get another question soon: Sandisk Ultra II vs Sandisk Extreme III. :) Does is cost the money? :D