Why cant Canon offer a Buffer upgrade? It cant be to difficult can it. Cant think its anything other than some memory chips. I guess that the current bodies have the memory fixed to the main board. But I cant see why they cant be removable like most laptops etc
Didn't nikon do a memory upgrade on the D3 after it came out?
That is exactly what I've asked the Canon pro division guys here in Canada in the past... but if the UDMA interface and internal processing is fast enough, it might mitigate the problem somewhat. It could have been that past Digic chips were also a limiting factor, not just the buffer, though I doubt Canon will ever confirm or deny that.
I think Nikon's offered the upgrade on at least some of the D2 and D3 cameras.
Stan has made it clear that he's not interested in the SanDisk Extreme (non Pro) or cards less than 64GB, but here are my tests.
1D Mark IV
Sandisk Extreme (non pro) 16GB
ISO 400
Tv = 1/800s
RAW = 27 shots and about 9 seconds to fully clear the buffer.
RAW = 28 shots (at 8 fps) and about 9 seconds to fully clear the buffer.
Large JPEG10 = 70 shots and about 7 seconds to fully clear the buffer.
Large JPEG8 (default) = 149 shots and about 7 seconds to fully clear the buffer.
32067dlm wrote:
Stan has made it clear that he's not interested in the SanDisk Extreme (non Pro) or cards less than 64GB, but here are my tests.
1D Mark IV
Sandisk Extreme (non pro) 16GB
ISO 400
Tv = 1/800s
RAW = 27 shots and about 9 seconds to fully clear the buffer.
RAW = 28 shots (at 8 fps) and about 9 seconds to fully clear the buffer.
Large JPEG10 = 70 shots and about 7 seconds to fully clear the buffer.
Large JPEG8 (default) = 149 shots and about 7 seconds to fully clear the buffer.
David
Thanks for the test. That's in stark contradiction to this:
... which I never believed for a second, and exchanged a number of PMs with the poster. There's no way that the cards could vary _that_ much in speed...
stanj wrote:
Usually raw + S, but in this particular job just raw.
Well, it seems like if you shoot jpegs, you can put just about any card in there. To my earlier point. I can't imagine any card adding more than one or two files more than mine. That is based on my own imagination and the fact that changing from 10 to 8 fps didn't add but one more file. To be honest, the speed felt slower with the raw files. The faster cards might speed up the fps a tad, and they might get you a file or two. The guy on the other thread must have been smokin' something. The buffer cleared surprisingly fast.
stanj wrote:
Usually raw + S, but in this particular job just raw.
32067dlm wrote:
Do you shoot to separate cards? Do you think this changes the buffer capacity? Just so you'll know, I took the SDHC card out for my tests.
David
I never use SD card unless in a complete emergency. I shoot raw+S because my workflow is tailored for that, and I am not planning on changing it; I can do raw only when buffer depth, or in my upcoming job card capacity, is an issue, but it screws up my workflow then. Sometimes that's the right price to pay.
I have seen that on the 1Ds3 the buffer depth is actually _bigger_ with raw+S than with raw only, an oddity that nobody could explain to me even though I asked. I have seen that on the 7D the actual buffer depth (frames until stutter) is much bigger with raw (23) than with raw+S (8), which is another thing that someone ought to explain to me (the factor of 3 is simply obscene).
In an ideal world I'd like to get a 3s buffer at raw+S, but what's almost more important is how fast a full buffer empties. The 7D seems to be really good on both fronts with raw only; I hope that Canon pulls another rabbit out of the hat with the 1Ds4 - again, the Ds has always been faster than the D so far, so there's hope. According to Canon, the 1D4 can take full advantage of the 90MB/s cards, and your test was with the 60MB/s flavor; I guess time will tell
stanj wrote:
Usually raw + S, but in this particular job just raw.
32067dlm wrote:
Well, it seems like if you shoot jpegs, you can put just about any card in there. To my earlier point. I can't imagine any card adding more than one or two files more than mine. That is based on my own imagination and the fact that changing from 10 to 8 fps didn't add but one more file. To be honest, the speed felt slower with the raw files. The faster cards might speed up the fps a tad, and they might get you a file or two. The guy on the other thread must have been smokin' something. The buffer cleared surprisingly fast.
The buffer depth is one thing - the time it takes to empty is another. With the 7D, which I used for some tests yesterday, the 60MB/s SanDisk card empties the buffer in 4.5s, while with an adata 32GB 533x card (rated 50MB/s write) it took 9s. So while the buffer was the same (23 raw), in one instance I could be shooting at full speed in half the time. Projecting to the 1D(s)4, if I can get a similar improvement with the Pro cards, I'm game.
stanj wrote:
I have seen that on the 1Ds3 the buffer depth is actually _bigger_ with raw+S than with raw only, an oddity that nobody could explain to me even though I asked.
Wow thats crazy! Never noticed because I never shoot RAW+JPEG but I'll have to try when I get home tonight.
stanj wrote:
So, in short, I'd also be very interested how many frames at 10fps raw (only) the 1D4 gets onto a Pro card before it stutters.
ISO 400, 1/1000 @ 4.0, RAW only, SanDisk Extreme Pro 16gb (UDMA 6), no in-camera noise reduction on, OneShot AF to eliminate focus-related shutter delay...
The buffer count shows 23, but I can actually shoot 29 frames before it first stutters.
The time from first stutter to the buffer is finished writing is about 7.5 seconds. This is based on when the red light goes out. Actual time measured was 7.79s but includes an extra frame or 2 written as I'm unable to release the shutter fast enough after the stutter stops to avoid capturing extra frames.
How do you come up with 4s? I want to make sure my information isn't misleading. When I first noticed a delay in the shutter (buffer full), I started a stopwatch and released the shutter. I stopped the clock when the red light went out 7.79s later.
olearydj wrote:
How do you come up with 4s? I want to make sure my information isn't misleading. When I first noticed a delay in the shutter (buffer full), I started a stopwatch and released the shutter. I stopped the clock when the red light went out 7.79s later.
That's because it was early morning and I misread your statement. I read time from the first _shutter_ and not _stutter_. Since it would take ~3s with the 1D4 to get to 30 frames, subtract that from the 7.79 (taking into account the slight delay for the last two frames or so), and I came up with about 4s.
Since it's 7.8s, it's actually slower than the 7D with the slower non-pro cards, which is quite disappointing, as it has more pixels yet writes noticeably faster. Damn.
olearydj wrote:
How do you come up with 4s? I want to make sure my information isn't misleading. When I first noticed a delay in the shutter (buffer full), I started a stopwatch and released the shutter. I stopped the clock when the red light went out 7.79s later.
olearydj wrote:
ISO 400, 1/1000 @ 4.0, RAW only, SanDisk Extreme Pro 16gb (UDMA 6), no in-camera noise reduction on, OneShot AF to eliminate focus-related shutter delay...
The buffer count shows 23, but I can actually shoot 29 frames before it first stutters.
The time from first stutter to the buffer is finished writing is about 7.5 seconds. This is based on when the red light goes out. Actual time measured was 7.79s but includes an extra frame or 2 written as I'm unable to release the shutter fast enough after the stutter stops to avoid capturing extra frames.
Hope this helps.
Dan, thanks for taking the time to do the test. I have one question: are you sure the buffer count was 23? My camera indicates 26 at ISO 400.
Perhaps unrelated but I noticed a possibly odd thing: when I set high ISO noise reduction to the strong setting, the buffer increased from 26 to 28 for any ISO between 100 and 640 (27 from 800 to 3200). The only thing I can think of is the NR makes for a slightly smaller file size after RAW compression, therefore allowing more in the buffer? Or if there is no affect on RAW, the NR will have an affect on the embedded full rez jpeg. I did a comparison test at ISO 1600 - without NR the RAW measures 21.5MB, with strong NR it is 21.3MB... I wouldn't think that's enough of a difference to allow another two frames...
stanj wrote:
Since it's 7.8s, it's actually slower than the 7D with the slower non-pro cards, which is quite disappointing, as it has more pixels yet writes noticeably faster. Damn.
I don't have any experience with the 7D, but I would guess it has a smaller buffer and similar write speeds, therefore a longer time to dump after a full buffer. Maybe not since you're comparing to the UDMA 5 or slower cards.
I'm curious what kind of specialized project you have that places such high demands on the buffer speeds. Maybe you should be using a movie camera?
rscheffler wrote:
Dan, thanks for taking the time to do the test. I have one question: are you sure the buffer count was 23? My camera indicates 26 at ISO 400.
Yes - 23. I just double-checked it. Are you also using the SanDisk Extreme Pro 16gb?
No, I primarily use a bunch of Sandisk Extreme IV 8GBs, but it's the same if I use an old Extreme III 4GB, Transcend 8GB 266x, Sandisk Extreme III 30MB/s 16GB.... None of my cards are UDMA 6. Not sure if that makes a difference.
