Canon 5D Mark III / 5D3 / Firmware 1.2.3

[kontrakatze]

Hi,
on my 5d Mark III with 1.2.3 installed and Nightly 2014 Mar 17:

- Powersave in liveview can't be enabled or disabled (always grayed out), submenues work, Battery Level is only informational, but selectable

Maybe this can be made somehow more obvious (a hint at the bottom in green?):

- Preview mode ML Grayscale in mlv_rec works only if global draw is set to allow (of course)

Best results recording raw I seem to achieve with:
Global draw OFF
Memory hack OFF
Extra Hacks ON
Buffer fill method 3
CF-only buffers 0
Card spanning ON

This appears to be the highest working resolution for my camera: 1792x1008 16:9 @29.97 fps continuouisly without any bad frames (CF 128GB Lexar 1066x and 128GB SanDisk Extreme 45MB/s).

Others seem to get full 1080p @ 30fps continuously, but I absolutely have no clue how they get there...

Regards,
kontrakatze

[Stedda]

This appears to be the highest working resolution for my camera: 1792x1008 16:9 @29.97 fps continuouisly without any bad frames (CF 128GB Lexar 1066x and 128GB SanDisk Extreme 45MB/s).

16GB & 32GB cards seem to be faster than 128GB cards.... you may want to search around for some of the threads on different brand cards and speeds if full HD is important to you.

[kontrakatze]

True, but I can't change cards every couple of minutes. That's why I choose the 128GB 1066x. In photo mode it reaches about 115MB/s solid, therefore I think a faster card would not change anything. Read speed is about 158 MB, which is about the max of UDMA7 as far as I know.

Maybe it is he cpu which is to heavily utilized in video mode, so there are to less cycles left for writing?


Regards,
kontrakatze

[Stedda]

Then I guess you're stuck with 1792x1008 as a max resolution....

My Transcend 16/32/64 do full HD plus.

[kontrakatze]

they do this at 30fps? So you get a write speed of 106MB/s in video mode?
With or without spanning?
Could you please tell me your settings regarding buffers, hacks and so on?


Regards,
kontrakatze

[pind]

to se how good is your CF card simply make a test in ML menu...

my kpb 64mb  have avg speed about 92mbps


WHAT IS IMMPORTANT is -----> Write speed (Buffer= 16384 K) if you look my picture.

https://www.dropbox.com/s/lp0w5qvn0pfmxp2/VRAM4.BMP

in my picture you will see  the  Write speed is (Buffer= 16384 K)---> 94.6MB/s

so that is the speed of card - the tasks what your 5d3 needs to run to make movie and what data rate can get out for raw recording.

[eyeland]

magiclantern-Nightly.2014Mar17.5D3123 running on Sandisk xtreme UHS 16GB - CF: Komputerbay x1000 64GB exfat quick format
(i) First 50GB of shooting, monitoring on nexus 7 (DSLRcontroller) worked flawlessly, then started crashing alot. (on 1.1.3 it usually crashes alot from the beginning)
(ii) This build seems to have had an effect on my setup in terms of writing speeds.  I can barely get 1920x1080@25fps continuous with all hacks off and GD on.
Low level format and full ML reset does not seem to have an effect of any of the issues I am experiencing. Having some odd numbers in my benchmarks (not sure if posting these has any relevance) - will do some more comprehensive testing later.
So far, I have observed that there is a big difference in bench speeds between 1080/25 and 720/50 (in canon menu) all other things being equal. (720~65mb/s - 1080~85mb/s)  (this might be expected behavior, but I only did a bench in 720 recently)

(iii) Getting some Err 70 and Err 80 once in a while that require battery pull - have not been able to determine the conditions but they are much more frequent than they ever were on 1.1.3

Having a hard time determining a  good testing methodology. Would it make sense if we compile a standard list of testing parameters? This could improve both testing and the posting of results.
(Edited out some tired nonsense

[kontrakatze]

Hi,
today I did some more comprehensive testing with my 5d mark III.

I benchmarked both my lexar cards, a 32gb 1000x and a 128gb 1066x on the latest nightly build for the canon firmware 1.2.3 and 1.1.3

I think the idea "eyeland" had, to compile a standard test setup is quite nice and convincing.
Therefore I took some time to think about it and came up with this setup I used:

Canon
Safety Shift OFF
Tone Priority OFF
Auto Power off OFF
HDR OFF
1920x1080 @ 30fps All-I

Magic Lantern
Global Draw OFF
FPS override OFF
LV DIGIC peaking OFF
Card settings Preferred card CF, others OFF
Modules: mlv_rec, mlv_play, nothing else

almost everything else is set to OFF
Benchmark in movie mode (live view on)

Camera was not moved during the benchmarks and the lighting was constant. I don't know if this is relevant, but it might be as the liveview is on during the benchmark and therefore the cpu has things to do.

I did all benchmarks three times, but got always the same results:
Firmware 1.1.3
Lexar 32gb 1000x      w   91.7      90.6      91.9      @16384k
                                r    122.3    122.3    122.1
Lexar 128gb 1066x    w   91.9      93.4      93.3      @16384k
                                r    122.3    122.4    122.2
Firmware 1.2.3
Lexar 32gb 1000x      w   86.3      89.3      89.6      @16384k
                                r    118.2    118.2    118.5

Lexar 128gb 1066x    w   85.4      89.1      89.0      @16384k
                                r    118.3    118.2    118.4


Next I did a brief test on the power consumption and the heat up of the camera, but these test were not really scientific as I took different batteries, relied on the display (temp) used a plain simple stopwatch and had to press a button every now and than to hinder the camera to shut the display down. Therefore take the result with care.

Anyway, the powerconsumption on 1.2.3 was much higher, the heat raised significantly faster.

By now, the datarate is significantly lower, the power consumption higher and the heat (37°C compared to 45°C after 3 minutes) also.
But hey, I stay with 1.2.3  I'm sure this can be sorted out.

Tell me what you think about the setup and let us compile a standard for testing.
And no, I can't do any programming, but if I can do anything else, like serving virtual coffee, I'll do it 

Regards,
kontrakatze

[kontrakatze]

Hi,
i forgot something totally. My cf cards were both formatted with Fat32. I guess this would make a difference to exFat, but now it's too late for testing today...

Should we set exFat to the list of standards for testing?

Something to add:
It is weird, the lighting situation seems to influence the write speed during the benchmark. With a lenscap on top I get repeatedly different speeds compared to lenscap off...
Could someone test this to be wrong or true?

Regards,
kontrakatze

[KenshirouX]

Hi,
i forgot something totally. My cf cards were both formatted with Fat32. I guess this would make a difference to exFat, but now it's too late for testing today...

Should we set exFat to the list of standards for testing?

Something to add:
It is weird, the lighting situation seems to influence the write speed during the benchmark. With a lenscap on top I get repeatedly different speeds compared to lenscap off...
Could someone test this to be wrong or true?

Regards,
kontrakatze

So what's the final verdict? Is it definitely worth updating to this version or should we wait for the mystery update that might come out next month?

[jafa]

I hope the mystery update is a free C.500 to all ML users  ....No!....oh well.

In my personal opinion , no , unless you want the ability to remove the bootflag , then yes.
Ive found the MLV modules on both 113/123 not reliable enough to rely on , but this is no problem as I prefer the audio out of my h4n anyway and syncing is not such a big hassle , but I have had more stopping using  123RAW than I did with 113RAW also my camera seems to heat up faster and run higher temps if only by 3 or so degrees but the most noticeable downside if you can call it that is that 123 seems to eat up battery life allot quicker than 113 (I should add that is just what SEEMS to be happening it could all just be my poor take on using both firmwares).
So what am I running on both my Mk3's ?, 123 mainly cause I cant be arsed to revert back to 113 and cause 123 seems to be the way forward even if it is a little bit less stable at the moment , there are of course no more nightly s for it at the moment so im holding my breath in anticipation for what is next-if anything.

Cheers
jafa

[fotogemeinschaft_de]

Before ML on my Mark 3 I could easily let the camera on over several days; thats not possible with ML 123; the power-consumption is really high

[a1ex]

Before ML on my Mark 3 I could easily let the camera on over several days

also my camera seems to heat up faster and run higher temps if only by 3 or so degrees but the most noticeable downside if you can call it that is that 123 seems to eat up battery life allot quicker than 113

Proof please.

(I already did my part)

[miatch]

friggin yeah it worked, pretty awesome, just gotta get out there and shoot!
thanx.

ps- cf issue interesting, had to pull mine at present.

[tomosaiko]

Hello,

I just installed it but i cannot find how to shoot video in RAW mode?
Thanks for help!

[Walter Schulz]

Module tab -> RAW_REC.MO or MLV_REC.MO -> Restart -> Enable RAW option via movie tab

[tomosaiko]

THX!

[jakdaniel1975]

with the version of March 26, I can not record at 50 fps at 1728x606 with 64 gb Komputerbay 1000x while with the version of March 17, I recorded without interruption ... the same thing happens to someone?

[a1ex]

The two versions are 100% identical (look at the change log).

[Canon eos m]

The two versions are 100% identical (look at the change log).

I noticed that the change log was empty. Why would someone want to update the build without any change to it?

[a1ex]

Nanomad was experimenting with the build server.

[budafilms]

Canon Menu HDMI Mirror works perfect shooting .RAW
You can connect any monitor by HDMI.

It's a revolution for me.

[dubzeebass]

Canon Menu HDMI Mirror works perfect shooting .RAW
You can connect any monitor by HDMI.

It's a revolution for me.

AWESOME !

[Jakobmen]

Canon Menu HDMI Mirror works perfect shooting .RAW
You can connect any monitor by HDMI.

It's a revolution for me.

with ML overlays? 

what settings you using.?

[jbggump]

Brand new 5D MK III (literally just opened it), formatted Transcend 64gb 1000x exFat on computer.  Downloaded 1.2.3 ML compatible, and installed.  Easy as pie.  Not using SD card at all.  I am fairly new to ML so I can't really speak to how things are working, but no need for 10 minute youtube install videos.

Shot some RAW footage at the upper end of the settings and it seemed to work fine - only a few minutes though.  Will test it out more tomorrow.