New 240 MHz preset for 5D3 / EOS M / 100D (Download available)

Started by theBilalFakhouri, September 07, 2022, 05:26:20 PM

Previous topic - Next topic

0 Members and 1 Guest are viewing this topic.

vastunghia

Quote from: theBilalFakhouri on September 08, 2022, 02:34:19 PM
Reported write speed during recording isn't accurate, I don't rely on it, it will always be higher in most cases.

Just ran a new test, with Global Draw disabled I get continuous rec. and

  • Reported write speed: ~89 MB/s CF, ~60 MB/s SD, total ~148 MB/s
  • Actual write speed (calculated from final MLV file size): 90.1 MB/s CF, 60.7 MB/s SD, total 150.8 MB/s
So reported write speed looks pretty reliable to me, at least on my 5D3.

Also, I do not understand if my results are in line with your expectation. When you say "in LiveView" you mean with Global Draw enabled, right? So cannot compare these with your 135 MB/s I guess.

Thanks for your clarifications

Sergio
5D3 for video
70D for photo


koopg

no, u nailed it, i was lost in the firat post

Sent from my SM-N975F using Tapatalk


theBilalFakhouri

Update: 160 USD have been collected from eight people.
Remaining amount: 105 USD.

Note:
PayPal seemed to take fees which I weren't aware of, I was updating the collected amount without fees. currently we have 160 USD (after fees).
To avoid fees, I think you need to use "Friends and family" option as payment type.

Thanks

theBilalFakhouri

Update: 189 USD have been collected from nine people.
Remaining amount: 76 USD.

theBilalFakhouri

Update: 194 USD have been collected from ten people.
Remaining amount: 71 USD.

Note: There is no low limit, any amount is welcomed.

theBilalFakhouri

Update: 229 USD have been collected from 11 people.
Remaining amount: 36 USD

You are so close to the new 240 MHz by now!
There is no low limit, any amount is welcomed

theBilalFakhouri

Buyers credit list will be posted after reaching 265 USD.

theBilalFakhouri

Update: 265 USD have been collected from 12 people.
Remaining amount: 0 USD

Goal reached!

The new module will be posted shortly

theBilalFakhouri

Testing time:

The a compiled version of the new module + source code have been posted in Reply #1:

TAKE ME THERE!

Happy testing!

TrincaEspinhas

EOS M with Samsung PRO Plus 128 GB

160MHz
Write≈68MB/s
Read≈75MB/s

192MHz
Write≈77MB/s
Read≈86MB/s

240MHz
Write≈ 93MB/s
Read≈ 23MB/s (got too hot probably or doesn't handle it correctly)

I'll later edit the post with some RAW video recording results at 192MHz.

theBilalFakhouri


theBilalFakhouri

Users:
Please report your results/test here!

Thanks

Walter Schulz

EOS M @240 MHz
Working with
1) Sandisk Extreme Pro R170/W90 128 GB (SDSQXCY-128G-GN6MA)
2) Sandisk Extreme Plus R170/W90 128 GB (SDSQXWG-128G-GF6CA)

not working @240
- Sandisk Extreme Pro R200/W140 256 GB
- Sandisk Extreme R190/W130 256 GB
- Samsung Pro Plus R160/W120 256 GB

reddeercity

Quote from: theBilalFakhouri on September 10, 2022, 03:20:28 AM
For Devs: The Magic Trick has been revealed, is it risky?

Yes , did you check it in quem? you don't know for sure what going on so , it just a "poke in the dark"
and you got lucky until someone burns up there cam. In my opinion you need to find out what going on , a very good way is Log files
with I/O trace build , it log all activity for 30 to 60 seconds can't remember right now .
That how i got crop rec ,Liveview , High frame rate(48fps FHD) 3x1 5k & 4K full res. to work on 5d2
and crop_rec on the 50d . Plus including the very early compressed raw on d4 cams , there's just no other way to be sure.

My 2 cents


gabriielangel

EOS M @240 MHz, with lvface hack
Tested with Danne's Aug 4, 2022 and Bilal July 31, 2022 Builds

I have two cards, both Sandisk Extreme Pro  170 MB/s, 128GB

1st card does not work @240 MHz  (Records about 2 sec regardless of resolution , benchmarks at W 20.9 MB/s)

2nd card works:
I can record 2.8k 14bit continuous (Green)
3K 1:1 19.99fps 14bit Continuous (Green)
3K 1:1 23.976fps 12bit, I stopped after 3:30 minutes (Orange), recorded three clips in a row

As usual, the first clip recorded after preset change doesn't record long, then it records Correctly afterwards.

I use a power bank and dummy battery, the temp reached 54 degrees C after recording 3 clips (Ambient temp was 28 degrees).
@192 MHz, it usually peaks at 52 degrees when 3K 1:1 14 bits, 20fps or 48 degrees when 2.5K 14bits, with 28 degrees ambient temp.

I will run a full test to check frame corruptions next week.

theBilalFakhouri

Quote from: reddeercity on September 10, 2022, 06:22:07 AM
Yes , did you check it in quem? you don't know for sure what going on so , it just a "poke in the dark"
and you got lucky until someone burns up there cam. In my opinion you need to find out what going on , a very good way is Log files

What is quem?

Logs doesn't and won't help, changing clock speed is something normal in DryOS and I am not expecting I would got anything from LOGs, e.g. Canon does change SD clock speed many times during SD initialization.

My wondering if changing clock speed many times is harmful or not for SD cards and SD controllers, also as I said, think about it as underclocking, and in general underclocking is something considered to be safe.
My 700D does work at 240 MHz all time for both read and write, by using lower clock speed for reads it would seem safer to me, because it's underclocking.

So, the only way to know is by asking SD experts.

names_are_hard

Quem is typo of qemu.

The IO trace log reddeercity is talking about is not a normal log, and can give you lots more info than you normally see, and can help.  Fundamentally, he's saying that since you don't understand why it works, you can't know if it works safely.  Which is a reasonable argument.

I do wonder if this is really related to dropping the speed for reads, it seems weird that would help (possible!  But strange - the card is rated to do higher speed reads than writes).  Here's a few tests you could run:
- do a 240 write, then change clock to 240 before doing a read: perhaps it's not 160 that fixes the read, but doing a mode change.
- do a 240 write, then change clock to 160 (or whatever is supposed to be good), but don't do any reads, then change to 240, then do the read.  Perhaps 160 is relevant, but not required to be the final selected mode.
- do a 240 write, don't change clock for the read but do trigger a small pause.  Perhaps the change to 160 helps simply because it takes some time, and this lets some other part of the system become happy.

These kinds of tests could be combined with IO trace to help us work out what the critical factor is.

reddeercity

Quote from: names_are_hard on September 10, 2022, 09:31:51 AM
Quem is typo of qemu.
......
Fundamentally, he's saying that since you don't understand why it works, you can't know if it works safely.  Which is a reasonable argument.
sorry I have dyslexia  :-[
Exactly!! :D
I'm using a1ex's locgic here , that the same argument he would make when many people come up ways of doing things that
couldn't be explained "just work when you set xxx parameters in the reg's"  will that didn't fly with him until it was understand fully.
Case in point the compressed raw , it was known about almost right from the beginning but it was a blackbox "still is to a certain extent"
e.g. 4.2.2. MJpeg (AVI) it there ,only cam to show promise so far (40D) , i have being poking at it for years so i hope 40D MJpeg development continues
once it work ok then i'll  try to get it on 5d2 & 50d . (I don't have time to do reverse engineering anymore but i do have some time that i can code it to
5d2 &50d.
I could go on & on about examples of short cut that have some success in certon fearture but at best there posed a dangerest risk
and I don't what to be that guy that coded a short cut that caused damage or just simple Softbreak the cam. 

Just trying to promote safe reverse engineering & coding  :)

SebastianC

My 5D mark iii 3k 3.3k 3.5k UHD 14bit or 12bit can't more than 77mb(SD) at 240MHz.
Global draw off

1920 1080 mode can achieve 90-100mb(SD),CF is 61mb.
240MHz.
Global draw off

5760 3240 anamorphic fullres mode can achieve 117mb(SD)
CF is 61mb.
240MHz
global draw off

Benchmark
write is 95mb
read is 72.5mb
Global draw On


All mode 12bit or 10bit can record more times.

I used Sandisk micro sd 128G V30 A2,
CF card is Sandisk 128G Ext pro 160mb

Thank you!

Grognard

I've tried.

It works flawless!
On 5D mark iii it's very stable, but it's not a real improvement because of the 133mb/s limit.

But I've noticed something intresting. My SanDisk 256g 170mbs use to work now with the brand new sd-uhs hack while it wasn't the case before, even in 160mbs. But use to work before on Eos M with old sd_uhs but not the New one!

on EosM it's very usefull! Now we can reach continuous 3k 10bits or continuous 5k anamorphic 14 bits lossless.

thank you Bilal.

Grognard

Quote from: SebastianC on September 11, 2022, 02:12:23 PM

All mode 12bit or 10bit can record more times.


Really ? I didn't notice any improvement about records times.  By the way it was already continuous in 3.5k 12bits and uhd 10 bits.

anto

EOS M
ML Nighty build 2022Apr30
Glabal Draw off

Card Sandisk extreme pro 170 MB/s 512GB

192 Hz
write speed 59 MB/s

240 Hz
write speed <20 MB/s

Danne

Could you test this version?
https://www.magiclantern.fm/forum/index.php?topic=25781.msg240247;topicseen#msg240247

Same code more or less. Test it with 2.8k preset and turn bitdepth to OFF to enable 14bit.

theBilalFakhouri

Recap users reports for the new sd_uhs:

5D3:

192 MHz and 240 MHz works without issues.
-SebastianC's report.
-Grognard's report.
-vastunghia's report.

EOS M / 100D:

1. 240 MHz doesn't work with all Sandisk Extreme PRO 170 MB/s SD cards:
-SebastianC's report on 100D using Sandisk Micro SD 128G V30 A2?, 240 MHz seems to work with new sd_uhs
-Walter's report using Sandisk Extreme PRO/Plus 170 MB/s on EOS M, 240 MHz does work with new sd_uhs
-gabriielangel's report using two Sandisk Extreme PRO 170 MB/s cards on EOS M, one of them worked at 240 MHz, the other one didn't work.
-anto's report on EOS M using Sandisk Extreme PRO 170 MB/s 512GB, it didn't work with 240 MHz.
-Danne's report using four Sandisk Extreme PRO cards on EOS M, all of the cards worked!

2. Some SD cards now works at higher OC presets with the new sd_uhs:
-TrincaEspinhas's report using Samsung PRO Plus 128 GB on EOS M (only 160 MHz was working with old sd_uhs)
-GianlucaM83's report using Sandisk Extreme Pro 170 MB/s (only 160 MHz was working with old sd_uhs)

Thank you all for testing and feedback!

I expected the new sd_uhs and 240 MHz preset to work on EOS M / 100D using all 170 MB/s SD cards, unfortunately that didn't happen.
This a new issue to me, no idea why certain Sandisk Extreme PRO 170 MB/s doesn't work while other do work.

This would require more digging. I hope someday we will reach perfect 240 MHz on EOS M / 100D.