Wed, Dec 4, 10:39 PM CST

Renderosity Forums / Photography



Welcome to the Photography Forum

Forum Moderators: wheatpenny Forum Coordinators: Anim8dtoon

Photography F.A.Q (Last Updated: 2024 Nov 26 6:56 am)



Subject: Flash memory speed


MGD ( ) posted Fri, 26 October 2007 at 3:29 PM · edited Wed, 04 December 2024 at 10:27 PM

Greetings,

A little while ago, TerraMatrix inquired about converting a DSLR into a video
camera
.  

In that message thread, I suggested that flash memory speed was one of
the problems that would need to be solved. 

I decided to test my assertion by doing some timing tests. 

I used as a source, 30 jpeg files on my HDD that aggregated about 34MBy
and used Windoze Explorer to copy those files into a new test folder that I
created on each of 3 of my flash memory cards.  The copies were done one
at a time; each test was run twice, and I got the timings by manually counting
off the seconds (thousand one, thousand two, ...).  If that's not sufficiently
scientific, you have my permission to repeat the expirement.  grin

I chose to use 30 files as the source because that would be 1 second worth
of video at 30 frames per second (30 fps). 

These are the (shocking) results ...

Lexar 80x 1 GBy CF Platinum II ... 27 seconds

SanDisk 256MBy CF ... 36 seconds

Transcend 1GBy SD ... 13 seconds

As you can see, the flash memory I own isn't anywhere near fast enough to
record video at 30fps. 

If anyone has (or knows of) faster flash memory, I would like to know about it. 

--
Martin


3DGuy ( ) posted Fri, 26 October 2007 at 5:48 PM

Hmm I transferred 35MB worth of jpg's to a SanDisk Ultra II 256MB CF card and it took all of 6 seconds. My slowest card is a Sandisk 512MB Card, nothing ultra or X-speed. Took 14 seconds.
If the specs on the sandisk site are true, an ultra III card should do it in 3 seconds.

What is a friend? A single soul dwelling in two bodies. - Aristotle
-= Glass Eye Photography =- -= My Rendo Gallery =-


inshaala ( ) posted Sat, 27 October 2007 at 1:46 PM

There are bottlenecks other than the "speed" of the disk.  Your processor speed, your harddrive read capabilities, and your card reader write speed/if you are using firewire, USB or USB2... so even tho i could say your manual counting was flawed in the experiment, you need to also take into account the overall speed of each test could have been slowed considerably by these other factors.

"In every colour, there's the light.
In every stone sleeps a crystal.
Remember the Shaman, when he used to say:
Man is the dream of the Dolphin"

Rich Meadows Photography


MGD ( ) posted Sat, 27 October 2007 at 2:11 PM

When I saw that 3DGuy had different results, I decided to retry my testing.  

I used the same USB 2.0 memory card reader with a different PC.  for this 
second test, a Dell notebook 1.4GHz, Windoze XP Pro; versus a 
Dell desktop 1.3GHz, Windoze ME in the earlier tests.  

I chose to use the same 30 files as the source because that would be 1 second 
worth of video at 30 frames per second (30 fps). 

These are the (new) results ...

Lexar 80x 1 GBy CF Platinum II ... 18 seconds

Lexar 12x 256MBy CF ... 31 seconds

SanDisk 256MBy CF ... 48 seconds

Transcend 2GBy SD ... 36 seconds (the original test was on this device: 2GBy not 1GBy)

As you can see, the flash memory I own still isn't anywhere near fast enough 
to record video at 30fps.  

For that matter, I am unable to explain (not to mention somewhat disturbed by) 
the inconsistent results.  

Whenever I wrote computer programs (which is just about all that I ever did in my 
entire working career), they always did exactly what I and the end user intended.  
My, my how the world has changed.  BTW, the programs I wrote ranged among 
business applications, device drivers, compilers, investment banking transactions 
in the secondary mortgage market (CMOs), ... and everything inbetween.  

-- 
Martin

p.s. I think I'll take a liedown and contemplate my ... ... LOL


MGD ( ) posted Sat, 27 October 2007 at 2:33 PM

While I was composing my 2nd message in this thread, inshaala mentioned that,

There are bottlenecks other than the "speed" of the disk

Boy oh boy!!! Is that ever the truth. 

--
Martin


3DGuy ( ) posted Sat, 27 October 2007 at 4:39 PM

It also depends on what codec you're going to use and what resolution you're going to shoot at. You're saying every frame is encoded in it's entirety, but alot of codecs just don't work that way. This translates to quite less data per second.

Think about it. You're saying 1s will take up 34MB, that translates to about 2GB per minute. That would mean you fit just over 2 minutes worth of video on a DVD. Let's take a movie of around 90 minutes in DVD quality, that'll take up round 4GB (including sound yes). Divide that by 90 minutes (average movie length) and that will give you about 45MB per minute. or 760KB/S.. including sound.

Granted, HD movies will take up far more but still. Blu-ray (single layer) = 25GB. Now even if the entire 25GB is devoted solely to a 1.5 hour movie, that still translates to 4.5MB/s. Which is still way less than the 34MB/s you're looking at.

What is a friend? A single soul dwelling in two bodies. - Aristotle
-= Glass Eye Photography =- -= My Rendo Gallery =-


MGD ( ) posted Sat, 27 October 2007 at 5:24 PM

I see that 3DGuy offered some insightful information by saying,

It also depends on what codec you're going to use and what resolution
you're going to shoot at

... along with some good comparisons and explanations about DVD encoding. 

I'm sorry that I failed to explicitly state some of my assumptions. 

I was trying to estimate the recording data rate that would be needed for
images that are about 4 mega pixels each.  In my experience, that would
be a jpeg file that is about 1 MBy to 1.5 MBy. 

What is the pixel count for standard (non HD) video images?  What is it
for HD images? 

Thanks for any additional comments.

--
Martin


3DGuy ( ) posted Sat, 27 October 2007 at 6:10 PM

HDTV = 1920x1080 (full HD), so just over 2MP. Regular DVD = TV = 720×576 (PAL) or just over 0.4MP. If you're writing each frame as a jpg, then you're right... maybe, what if you up the compression?. Most codecs work differently though. Basically they record a keyframe (1 complete frame) and then just record the differences between the keyframe and the next frame. Every so many frames a new complete frame is inserted.  There's variations to the theme, but that's basically it.

My point is, with video you can't really say "a frame is xxxx bytes" so 30 frames is 30 times xxxx bytes".

What is a friend? A single soul dwelling in two bodies. - Aristotle
-= Glass Eye Photography =- -= My Rendo Gallery =-


TomDart ( ) posted Sat, 27 October 2007 at 8:53 PM

Quite interesting thread, even if I only pick up on half of it.  This still makes sense.  I believe when the opening of the Matrix movie was done, the scene was shot with a bank of Canon DSLR's, all correlated to exact timing. Many stills from the Canons made the opening scene work well.  Not video in shooting but a movie in final cut...to think about.


TomDart ( ) posted Sun, 28 October 2007 at 5:00 PM

Ok, since some of you are more up on the tech of flash memory speed, I have two questions.

My latest CF cards are Delkin Film Pro, 305x, supposedly with "lightning fast performance".  I really can't tell a difference but perhaps I am too slow to pick up on it..

And, as for Video, the normal consumer video shoots what resolution of image? Most stills I have seen from "non pro" video are certainly lacking in resolution.  It is closer to me shooting at the smallest camera image size, something like 800px wide or so.

If this is so, how would card speed compare with such smaller images?   Just curious, not a techy but can understand it when explained.         Tom.


Privacy Notice

This site uses cookies to deliver the best experience. Our own cookies make user accounts and other features possible. Third-party cookies are used to display relevant ads and to analyze how Renderosity is used. By using our site, you acknowledge that you have read and understood our Terms of Service, including our Cookie Policy and our Privacy Policy.