Menu

Show posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.

Show posts Menu

Messages - intrinsic

#1
Just tested latest build (06/27) with Transcend 1000x 16Gb
with "small hacks" on
vs 06/25:
1880x1058 FPS 23.9 > 265/330
1880x1058 FPS override 20 > 807/2269 (write speeds up avg from 60 to 64 Mb/s)
1728x972 FPS 23.9  > 686/1720


Indeed huge performance breakthrough!
Still wonder, why lowering FPS significantly increases write speed, any ideas?
Anyway, 1880x1058 @20fps looks very good for me )
#2
Is this buffering aspect has anything to do with increasing card write speed at lower FPS using FPS override? Or is it CPU problem?
I've first noticed that on 5dMkII, and 550D shows the same, it goes up to 22Mb/s at the beginning of the recording (than drops around 21), @FPS 8-12
#3
Quote from: dariSSight#1 on June 20, 2013, 12:56:29 AM
Thanks for all the work you guys do, I want to get a clear understanding on the best possible resolution and what keeps the 5D Mark II From utilizing the full Sensor functions? I know the full framing of the Canon 5D Mark II is 3:2, so can you explain why the camera gets RAW 1880 instead of 1920? ???
Here:
http://www.magiclantern.fm/forum/index.php?topic=6215.50
You may push larger resolution (2144x1078) utilizing 5x crop mode.
As it was mentioned here before, Canon's original 1920x1080 H.264 is upscaled from 1880.
Doesn't seem like a big deal to me, it's only 40 pixels. Nothing compared to RAW sharpness and detail even at the lower resolutions.
#4
Quote from: Yoshiyuki Blade on June 19, 2013, 10:23:38 PM
I just noticed that reducing the frame rate with FPS override also increases the card's write speed slightly.
Yepp, same thing here.
I usually got around 57-58Mb/s @24FPS and decreasing FPS gives me up to 60 (16FPS) - 61Mb/s (6-9FPS).
Twixtoring result very much depends on the nature of the filmed subject.
In the more or less static scenes (nature) it is possible to twixtor from 12fps.
My street tests (moving cars, people, etc) shown me that FPS should be at least 18, with further twixtoring to 75% speed.
But even in that case you should avoid fast moving large subjects close to the camera.
About the write speed.
I've found in the 550D thread A1ex posts about buffer optimisation and how different buffering strategies affect card write speed.
From here:
http://www.magiclantern.fm/forum/index.php?topic=5582.msg52455#msg52455
Question, is there some sort of connection betweeen FPS override and different buffering?
#5
Quote from: eatstoomuchjam on June 13, 2013, 07:39:36 PMEven if you can't push 1728x972, I'd imagine that you could still push 1600x900 which is still at least a bit better than 1280x720
Yes, 1600x900 is stable in my case, and it's MUCH better than 1280x720 especially in terms of details in shadows, noise and dynamic range.
I also use Tokina 11-16 EF-S, it helps to avoid issue with cropping on lower resolutions, so even at 1600x900 I still have a full wide angle.
#6
Quote from: eatstoomuchjam on June 13, 2013, 04:05:10 PM
You can shoot 1728x972 reliably with a 1000x card on the 5D Mark II.
It is probably reliable resolution with some particular 1000x cards, for instance my 16Gb Transcend 1000x at 1728x972 gives me only around 500 frames.



#7
Hi everyone!
I've got transcend 1000x 16Gb UDMA7 and ML benchmarks shows that it's capable of write speed up to 70Mb/s
In real action it shows no more than 58Mb/s
As I understand write speed depends on buffer size? And this speed/buffer size combination is unique for every card?
Is it theoretically possible to squeeze maximum from my card (this case 70Mb/s) by some sort of buffer size optimization or the only option here is to buy "legendary" Lexar 32Gb 1000x?

#8
Hi, guys!
Managed to test the latest release from mk11174.

Was able to get pretty stable 1728x576@12fps.
For me 12fps twixtored to 24fps is a good deal for the most cases I need.

Details:

- Movie mode, 1280x720@60fps or 640x480@60p
- Global draw off
- SanDisk ExtremePro 8Gb 45MB/s
- Photo Quality set to JPEG S (don't know if it affects anything)


Using previous build I've got 55Mb of free buffer cache, with the last one it was 62.
I've noticed also that at some magic point I became able to write normally at about 20MB/s, as earlier it was no more than 9-10MB/s.

About difference between RAW and standard H.264 Canon video.
For me it's mostly about noise and dynamic range which is goddamn good at low-light shooting and high contrast scenes.
So far I haven't been able to see a huge difference in detail and sharpness between 1080p H.264 and upscaled RAW I've got from 550D.
Hope with the higher stable resolutions it will become obvious )
So for now, the main feature here which makes it worth the all the efforts, is ability to get clean picture with more film-like noise especially in shadows and underexposed areas. Because trying to get the same picture in terms of shadow/highlight as I've got with RAW footage, using H.264, will result a huge digital-looking noise at shadows, especially terrific at ISO rates higher than 400.

Once again, god bless ML team and mk11174!
From Russia, with love!