Magic Lantern Forum

General Discussion => General Chat => Topic started by: reziarfg on April 11, 2016, 05:02:55 AM

Title: 80D Theoretical raw video hardware analysis
Post by: reziarfg on April 11, 2016, 05:02:55 AM
Hi all,
As it stands I know the 80D may be a long way from ML support, and I hope this post doesn't breach any rules, but I thought we could do some basic theory-crafting on its hardware capabilities.
Here's what my research has uncovered so far based on internet benchmarks and existing ML implementations for other cameras:


I do have an 80D and Sandisk Extreme Pro 95MB/s UHS-I SD card. If anyone knows how I can run an effective benchmark I'll give it a try. I plan on attempting to simply time how many RAW files I can write in a certain burst duration and do some math on how many MB/s it's writing. Not sure how well this will turn out.
Title: Re: 80D Theoretical raw video hardware analysis
Post by: Audionut on April 11, 2016, 05:23:44 AM
Moving this to general chat since there is no port, and the other sections are for working ports and features on those ports.

If you can find someone with a 70D (with a fast card) who can time a burst sequence and the time it takes to finish writing the files to the card, and then time the same on your 80D, it won't be overly accurate, but should hint whether the 80D is faster or not.
Title: Re: 80D Theoretical raw video hardware analysis
Post by: reziarfg on April 11, 2016, 05:54:30 AM
Thanks, I'm still pretty new to posting. Sorry for putting this in the wrong spot.

I just ran a test with a similar methodology as the one I linked in my above post (http://www.cameramemoryspeed.com/canon-7d-mark-ii/fastest-sd-cf-card-comparison/)
I ran a burst of RAW files(RAW only, no JPEG) in high speed mode with the lens cap on and manual focus engaged. I measured the time from when the red "activity" light was on until it turned off after the burst completed. Note that I stopped capturing images before the buffer filled each time so the capture rate doesn't slow down.

I ran a few tests with somewhat mixed results. If anyone can make any recommendations on improving my process, I'm open for suggestions.
RAW file size is consistent at 22.0MB per image It looks like the RAW file gets bigger and results get worse as ISO goes up. I guess this makes sense? For the sake of this test I manually set ISO to 100, F2.8, 1/200 shutter speed.

Test 1:

Test 2:

Test 3:

Test 4:

I feel like it's already been pretty well established, but If anyone's got a 70D and a sufficiently fast card and wants to try out the same test I'd be interested to see the results.
It looks like these results are slower than the ones published for the 7D MK II in my linked benchmark. But I can't say for sure if that ~10MB/s difference is a result of my testing process being flawed or a hardware difference.
Title: Re: 80D Theoretical raw video hardware analysis
Post by: Malakai on May 16, 2016, 05:55:18 PM
Just checked the video files from my 80D and it looks like the video bitrate peaks at about 92mbps, the 70D is about 42mbps. Would we have to take into account the buffer too?
Canon recommend using a UHS3 card for video to cope with the higher bitrates.
Based on this, I would hazard a guess that the 80D is twice (if not a bit more) as fast as the 70D in terms of writing to the card.
Title: Re: 80D Theoretical raw video hardware analysis
Post by: Walter Schulz on May 16, 2016, 06:53:09 PM
Canon recommend using a UHS3 card for video to cope with the higher bitrates.

Source? Link, please. It's nonsense because 80D will force UHS-III cards into UHS-I mode and some cards (Sandisk ...) will be outperformed by fast UHS-I cards. See www.cameramemoryspeed.com
Title: Re: 80D Theoretical raw video hardware analysis
Post by: nikfreak on May 16, 2016, 06:55:13 PM
Don't remember the link but 750D/760D and 80D do really improve wrote speed for sdcards. Something around 70MB/s was said to be possible...
Title: Re: 80D Theoretical raw video hardware analysis
Post by: Greg on May 16, 2016, 07:52:59 PM
http://www.cameramemoryspeed.com/canon-t6s-t6i/sd-card-comparison/

760D write speed - 75MB/s
80D RAM - 1024MB ?
(20 x 24MPx = 800MB)
Title: Re: 80D Theoretical raw video hardware analysis
Post by: reziarfg on May 17, 2016, 01:30:56 AM
Just checked the video files from my 80D and it looks like the video bitrate peaks at about 92mbps, the 70D is about 42mbps. Would we have to take into account the buffer too?
Canon recommend using a UHS3 card for video to cope with the higher bitrates.
Based on this, I would hazard a guess that the 80D is twice (if not a bit more) as fast as the 70D in terms of writing to the card.
92mbps? That's only 11.5MB/s so I don't know if it's relevant in terms of max write speeds. Since video recording is often measured in mbps and not MBps it's an important distinction.


http://www.cameramemoryspeed.com/canon-t6s-t6i/sd-card-comparison/

760D write speed - 75MB/s
80D RAM - 1024MB ?
(20 x 24MPx = 800MB)

This is really cool. The t6s wasn't there when I did my test and since the 80D falls between T6s and 7D MK II in terms of placement in Canon's lineup it stands to reason it's also capable of the same speeds. Unless for some reason they put a faster SD card controller in the T6s. But that seems unlikely. Looks like good news all around!
Title: Re: 80D Theoretical raw video hardware analysis
Post by: Malakai on May 17, 2016, 11:28:04 AM
Source? Link, please. It's nonsense because 80D will force UHS-III cards into UHS-I mode and some cards (Sandisk ...) will be outperformed by fast UHS-I cards. See www.cameramemoryspeed.com

Page 5 of the canon 80D user manual (http://gdlp01.c-wss.com/gds/0/0300022720/01/EOS_80D_Instruction_Manual_EN.pdf)
Title: Re: 80D Theoretical raw video hardware analysis
Post by: Walter Schulz on May 17, 2016, 11:41:02 AM
Sorry, you mixed it up.

UHS Speed Class 3 = At least 30 MByte/s sustained write rate
UHS-III = New interface type, but useable in UHS-I devices.

http://kb.sandisk.com/app/answers/detail/a_id/1996/~/difference-between-speed-class,-uhs-speed-class,-and-speed-ratings
https://en.wikipedia.org/wiki/SD_Memory_Card#Speed_class_rating

Title: Re: 80D Theoretical raw video hardware analysis
Post by: Malakai on May 17, 2016, 12:33:15 PM
Doh, well spotted, teach me to skim through the manual lol
Title: Re: 80D Theoretical raw video hardware analysis
Post by: BeanBread on July 31, 2016, 06:44:06 AM
Maybe this is worth something.

Canon 80D video testing, ALL-I, 24fps
Lexar SD Class 10, UHS Speed Class 3, 1000x UHS-II SDXC

I've shot about 100 small videos today, up to 2min long.
First record button press of the day camera's buffer or whatever filled up and the recording stopped.

The cache memory level indicator showed up maybe 12 times in total out of which 3-4 times it concluded with a stop record.
Could be a bed card, maybe Lexar is cutting corners, becoming greedy, could be something else.
I haven't done the math but I was expecting a "Write Speed 80 MB/s Maximum" to be enough room for the 80D.
Reading the manual also, the said SD card should have been enough.

Do you guys know of some free software to test my SD card performance ? I have a Lexar USB3.0 card reader.

Thanks !
Title: Re: 80D Theoretical raw video hardware analysis
Post by: Walter Schulz on August 02, 2016, 02:35:25 PM
Card description far from being detailed.

Use cardreader and benchmark program to determine card speed.
PC: ATTO, CrystalDiskMark
OS X: Black Magic Disk Speed Test
Title: Re: 80D Theoretical raw video hardware analysis
Post by: KelvinK on August 02, 2016, 02:52:52 PM
I guess, they had to increase SD writing speed to make 1080/60p safe recording. My guess it's 60-70MB/s, not more, but already good for RAW recording :)
Title: Re: 80D Theoretical raw video hardware analysis
Post by: BeanBread on September 15, 2016, 01:20:20 AM
Camera: Canon 80D
SD card : 128Gb Lexar SD Class 10, UHS Speed Class 3, 1000x UHS-II SDXC
Video: 24fps ALL-I FullHD

Crystal Disk Mark 5.1.2 x64 - Banchmarking
Using Lexar USB 3.0 Dual Slot Reader LRW400U Rev.A

Seq Q32T1 50MB
One(single) test run
Read 149MB/s
Write 136MB/s

One(single) test run
Seq Q32T1 4GB
Read 11.5MB/s
Write 11.0MB/s

One(single) test run
Seq Q32T4 4GB
Read 11.4MB/s
Write 6.8MB/s

So .. the Canon 80D manages to choke the above Lexar card. 24fps ALL-I FHD
Not always, 1 time out of 5 maybe, and it's usually right at the start first 15 secs.
Enough to destroy my confidence and confine myself to IPB.

The Canon 70D seem to be fine with the above card, and video setting above.

Will update my findings once I have something new relevant.
Title: Re: 80D Theoretical raw video hardware analysis
Post by: reziarfg on September 29, 2016, 12:36:32 AM
Just in: Cameramemoryspeed.com has their numbers on the 80D

It's just slightly faster than the 7d mark II which is great news.
http://www.cameramemoryspeed.com/canon-80d/sd-card-comparison/

Since they report "avg write speeds" we'll just call that sustained. Making some liberal assumptions, this puts theoretical RAW video performance at around the original 7D (which is quite good!)
Here is the comparison chart from elsewhere on this forum for drawing conclusions:
https://docs.google.com/spreadsheets/d/16cgnRivbUv7nA9PUlCLmLdir3gXdIN3pqzCNAAybepc/edit#gid=5
Relevant thread:
http://www.magiclantern.fm/forum/index.php?topic=6215.0
Title: Re: 80D Theoretical raw video hardware analysis
Post by: LesterL on December 05, 2016, 04:17:14 PM
Those speed reports are great news! With the new development of 12-bit and 10-bit raw on other Magic Lantern supported cameras, the 80D could be quite an impressive camera for the price. It's dynamic range is already more impressive than almost all of the Canon predecessors.