Panasonic GH5 - all is revealed!

[philipd]

6 hours ago, Cinegain said:

You ever noticed this on the SanDisk UHS-II's tho ('4K @ 500 Mbps')?

Hi

Quote

1. Full HD (1920x1080) and 4K UHD (3840 x 2160) video support may vary based upon host device, file attributes and other factors.

That is the footnote they put to that statement.  It's basically something SandDisk has made up for marketing, as that card doesn't show a V rating, we know it doesn't support the new real-time recording protocol and there is no guaranteed minimum write speed.

There are 2 problems with using cards that don't support the new Video (V) protocol.

1) The GH5 may not be able to reach or sustain the required bit-rates full stop unless the camera can switch to using the Video protocol mode with a supported card.  The new protocol changes completely how the camera writes to the card in order to ensure high speed writes that are sustained and never drop below the guaranteed minimum.  The GH5 firmware may even block the use of cards that aren't rated at V60 or better.

2) All-Intra is a sustained constant (more or less) write speed, the card and hardware have to sustain the write speed without any hiccups.  IPB used in cameras is a variable bit rate that allows borderline cards chance to catch up and buffers to empty as the write bit-rate ebbed and flowed so we got away without needing to worry too much about guaranteed minimum write speeds, this is not the case with All-Intra at these high recording rates, it is much less forgiving.

The V moniker shows a guaranteed minimum write speed, achieved using new hardware in the camera and card (for example a  memory card with the V rating reports to the camera about itself so the best writing scheme can be negotiated and both card and camera can switch to a new real-time streaming recording protocol).

Bench mark tests are also going to be something we should be wary of as new V type cards arrive and are tested and reviewed.  The reason being a card may benchmark much slower than it's V60 or V90 rating suggests it should and so look a poor choice for cameras like the GH5, but this is simply because the benchmark software/hardware will not engage the Video recording protocol, so isn't testing it for video recording. The software and readers will catch up in time, but it's something to we need to consider if looking at reviews.

For real-time video recording on cameras like the GH5 the only rating we should now look for is the V rating, it's the whole reason it was brought in, because all other ratings show the burst or average speeds and guaranteed absolutely nothing for real-time video.

Anyone investing in memory cards now for use with the GH5 and All-Intra based on some headline speed on the packaging where that card doesn't show the required V moniker , is going to be wasting their money.  There is a reason why V60 and V90 cards are arriving now and being sold at a premium, they are different to normal cards.  For example Sandisk can't take their current high speed cards and simply print V60 or V90 on there as it's more than just a printed a number and their current cards don't have the hardware/firmware to support the new standard.

Regards

Phil

 

[Orangenz]

18 hours ago, philipd said:

Hi

With regard to V30, V60 and V90 cards, these specify a minimum write speed, regardless of the state of card (fragmentation, erased state etc) which is the important bit of course for real-time recording, so will be lower than the headline marketing speeds.  Also it is important to note that to obtain these guaranteed minimum speeds these cards need to be written to using a different protocol, and the card itself needs to specifically support this protocol as well as the camera (of course the GH5 does).  This means buying a card that doesn't show the V rating, because it appears fast enough based on it's other speed ratings  or some website bench-marking, is not the best route to take and will likely see problems.

Regards

Phil

That sounds interesting. Is that something on the net you can link to? I imagined the UHS-II sd card is the definer of the physical protocol and U3 or V30 is a speed class that a card is meant to be able to achieve (as min speed). I would suggest that memory write speed is a physical parameter not related to a variable protocol being followed by the GH5. The advertising video directly from the SD association makes the association between UHS-I with V30 and UHS-II with V60 and V90. 

Anyway, current sandisk uhs-ii cards have been measured in the GH5 at about V70 which is more than sufficient for the new codec. Pretty easy to test too. Single long video files should be even faster than raw files. The only uhs-ii cards that fail V60 equivalent speeds and putter along at around V30 are Delkin and Lexar (the only card CRFTSHO decided to test - sigh).

Why doesn't the Sandisk card have the V60 logo then? No idea. Maybe with the current build there is enough variation between cards to dip down sometimes? It seems everyone is bypassing the spec though and wanting to go straight to V90. I suspect that is when their current cards will get the name change. 

[philipd]

29 minutes ago, Orangenz said:

That sounds interesting. Is that something on the net you can link to? I imagined the UHS-II sd card is the definer of the physical protocol and U3 or V30 is a speed class that a card is meant to be able to achieve (as min speed). I would suggest that memory write speed is a physical parameter not related to a variable protocol being followed by the GH5. The advertising video directly from the SD association makes the association between UHS-I with V30 and UHS-II with V60 and V90. 

Anyway, current sandisk uhs-ii cards have been measured in the GH5 at about V70 which is more than sufficient for the new codec. Pretty easy to test too. Single long video files should be even faster than raw files. The only uhs-ii cards that fail V60 equivalent speeds and putter along at around V30 are Delkin and Lexar (the only card CRFTSHO decided to test - sigh).

Why doesn't the Sandisk card have the V60 logo then? No idea. Maybe with the current build there is enough variation between cards to dip down sometimes? It seems everyone is bypassing the spec though and wanting to go straight to V90. I suspect that is when their current cards will get the name change. 

Hi

Copying files are not a true test to see if a card has a sustained data-rate suitable for high bit rate video, for example the copy may dip to almost zero bps then burst a lot higher, giving a high overall average but would break when writing real-time video that can't be halted or would see a buffer overflow.  This is why they've introduced a new standard so we know the card will work at a sustained minimum.  File copies are different to different to real-time streaming.

Some details are in the white paper https://www.sdcard.org/downloads/pls/latest_whitepapers/Video_Speed_Class-The_new_capture_protocol_of_SD_5.0.pdf  it is part of the SD 5.0 specification, from page 7 onwards it goes into more technical details, and as explained it requires a card and hardware to understand the new protocol to enable it.

If the card doesn't show V30/V60/V90 it doesn't have support for SD 5.0 and real-time video recording, this of course hasn't mattered previously but does now where 400Mbps constant writing speeds are required. I bet there is no current SDXC that will work in the GH5 on 400Mbps intra recording, unless it is has V60 or V90 on the label, regardless of how much it costs or how high a speed it is advertised at.  I also wouldn't be surprised if on the released firmware the camera refuses to do anything unless the card reports it is V60 or better to avoid complaints of 'it isn't working'.

Regards 

Phil

[Orangenz]

14 minutes ago, philipd said:

Hi

Copying files are not a true test to see if a card has a sustained data-rate suitable for high bit rate video, for example the copy may dip to almost zero bps then burst a lot higher, giving a high overall average but would break when writing real-time video that can't be halted or would see a buffer overflow.  This is why they've introduced a new standard so we know the card will work at a sustained minimum.  File copies are different to different to real-time streaming.

Some details are in the white paper https://www.sdcard.org/downloads/pls/latest_whitepapers/Video_Speed_Class-The_new_capture_protocol_of_SD_5.0.pdf  it is part of the SD 5.0 specification, from page 7 onwards it goes into more technical details, and as explained it requires a card and hardware to understand the new protocol to enable it.

If the card doesn't show V30/V60/V90 it doesn't have support for SD 5.0 and real-time video recording, this of course hasn't mattered previously but does now where 400Mbps constant writing speeds are required. I bet there is no current SDXC that will work in the GH5 on 400Mbps intra recording, unless it is has V60 or V90 on the label, regardless of how much it costs or how high a speed it is advertised at.  I also wouldn't be surprised if on the released firmware the camera refuses to do anything unless the card reports it is V60 or better to avoid complaints of 'it isn't working'.

Regards 

Phil

Thanks for that. Interesting read. Of course the buffer test is not the best test I can imagine but it is, nevertheless, a test. A scalable test involving video will not be possible, only a duration one for one speed. As per page 5, I would not be surprised if the Sandisk U3 cards are quite capable of sustained V60. The example they give is a C10 managing V30. 

On page 7 it's a bit unclear. One of those things I'd love to speak with an engineer directly about. 

Quote

How multiple NAND devices are organized and accessed in parallel is the domain of SD memory card manufacturers,
the internal organization of the NAND memory is the domain of NAND memory manufacturers,
and the application developers get to figure out how to use the result efficiently

Most of the protocol seems to not deal with the above but with rewrite issues and parallel streams (like video with interleaved gps). At this stage I am satisfied that buffer tests showing double 30MB/s (actually 70MB/s) will be more than sufficient given that 50MB/s is the maximum sustained required. As I have a sandisk UHS-II card I will obviously test this myself on the firmware in a couple of weeks. The two takeaway points seem to be if you have the option of the Sandisk UHS-II U3 and a Sandisk V90 then go for the V90. But don't sweat it. The other point would be to follow normal sd card practise and not delete individual files or in another device. Use the camera to format the whole card when space needs to be reclaimed. 

So if that document is the only basis for saying "If the card doesn't show V30/V60/V90 it doesn't have support for ... real-time video recording" then I think that would be incorrect and misses the point of what the standard is trying to do. 

The exciting thing is that in a very short time we will have 1000 testers across the world. :D 

[philipd]

43 minutes ago, Orangenz said:

Thanks for that. Interesting read. Of course the buffer test is not the best test I can imagine but it is, nevertheless, a test. A scalable test involving video will not be possible, only a duration one for one speed. As per page 5, I would not be surprised if the Sandisk U3 cards are quite capable of sustained V60. The example they give is a C10 managing V30. 

On page 7 it's a bit unclear. One of those things I'd love to speak with an engineer directly about. 

Most of the protocol seems to not deal with the above but with rewrite issues and parallel streams (like video with interleaved gps). At this stage I am satisfied that buffer tests showing double 30MB/s (actually 70MB/s) will be more than sufficient given that 50MB/s is the maximum sustained required. As I have a sandisk UHS-II card I will obviously test this myself on the firmware in a couple of weeks. The two takeaway points seem to be if you have the option of the Sandisk UHS-II U3 and a Sandisk V90 then go for the V90. But don't sweat it. The other point would be to follow normal sd card practise and not delete individual files or in another device. Use the camera to format the whole card when space needs to be reclaimed. 

So if that document is the only basis for saying "If the card doesn't show V30/V60/V90 it doesn't have support for ... real-time video recording" then I think that would be incorrect and misses the point of what the standard is trying to do. 

The exciting thing is that in a very short time we will have 1000 testers across the world. :D 

Hi

Of course it depends if the final GH5 restricts use to V60/90 cards only or attempts to write to any card on a best effort basis as to whether we have a chance with any fast card.  We know already that a Sandisk 95MB/sec UHS-I card, so that is a 760Mbits/sec data-rate, almost double that actually required for 400Mbits/sec All-Intra, failed, that was tested by this site.  So there is a world of difference between the marketing of these cards and what they achieve in reality.

It may well be a V60 card isn't even enough or 100% reliable as that is 480Mbits/sec, so not much headroom, and explains why it seems those testing the firmware 2.0 update were all given Panasonic V90 rated cards to use.  Panasonic only confirm tested with a V90 card so far.

Regards

Phil

[Cinegain]

Actually, the 95 on the UHS-I is readspeed. It's R95/W90. And as I interpolated/logically tried to deduce, you'd need around W180MB/s or higher cards for sustained 400Mbit/s, because indeed, there's no direct translation between top out writespeeds MB/s or Mbit/s and the max. sustained ones (90MB/s -> 720 Mbit/s (not 760) you'd think would cover 400Mbit/s, but the '720' is a top out speed, not a tested minimum sustainable one). But comparetively the SanDisk 280/300 (R) cards rank in those speeds as V90 classified ones, so I'm feeling rather confident that they'll reach at least V60 speeds for the 400Mbit/s (50MB/s) required by the GH5 firmware upgrade. That officially you can not conclude any such thing, I am aware of, that's why I said:  'draw your own conclusions'. My conclusion personally is that's it's probably going to work with SanDisk 280/300 cards, but like Orangenz said, we'll find out in due time. Your conclusion is that:

Quote

Anyone investing in memory cards now for use with the GH5 and All-Intra based on some headline speed on the packaging where that card doesn't show the required V moniker , is going to be wasting their money.

That remains to be seen.

Quote

There is a reason why V60 and V90 cards are arriving now and being sold at a premium, they are different to normal cards.

Exactly. And it's a good marketing ploy to ask good money for these more rare magic cards... but that doesn't mean all that much to me, to be honest. It's like Leica rebranding Lumix cameras or Hasselblad did with the Sony cams. Still the same, can't do anything more, just has a fancier name and look with a more premium price.

Quote

For example Sandisk can't take their current high speed cards and simply print V60 or V90 on there as it's more than just a printed a number and their current cards don't have the hardware/firmware to support the new standard.

Are you positive anything has changed from the previous to the current UHS-I R95/W90 with or without V30 logo on there..? Like I said, I'm not claiming a W90 card gives sustained 90MB/s, but if that gives V30, 30MB/s sustained... V60 probably needs a 180MB/s writespeed and that's all you need for the new GH5 firmware upgrade. The Panasonic V90 card and others typically do R280/W250. So to say SanDisk Extreme Pro SDXC UHS-II with R280/W250 or R300/W260 won't work with the GH5 400Mbit/s period... well, howabout we just wait and see? Remember, in the GH2 hack days, the only ones that mattered were SanDisk Extreme Pro cards to keep up with all the continuous writing; I'm not giving up on them so quickly. Might be wrong of me, but we'll find out soon enough. But just because it's not officially guaranteed, doesn't mean it won't work just fine. I'm not all that black 'n white.

[philipd]

1 hour ago, Cinegain said:

Are you positive anything has changed from the previous to the current UHS-I R95/W90 with or without V30 logo on there..? 

Yes, it is in the white paper published by the SD Association, its a new protocol and hardware specification specifically for real-time video recording, so cards of certain age could never have this addition anyway as it wasn't invented :-) 

It isn't just another way of advertising the speed of the card, it denotes the card supports the new video recording protocol that is optimised for video recording. The card enters a different mode of operation for video recording on cameras that also support this standard. 

Other cards may work absolutely fine but unlikely, the problem is cards not using the new video recording method very quickly in use get to a point they cannot sustain a fast write data-rate without pausing to do other things.  This happens all the time today in the cameras, and it's fine with VBR files not going much more than 150Mbits/sec for short bursts as the cards are faster enough to catch up and buffers big enough in the cameras to ride these drops in writing performance.  When you are needing to push 400Mbits/sec constantly (not VBR) no matter how fast modern cards are at the moment, any pause in accepting data gives rise to a situation that the card can't run ahead and catch up quick enough, and very quickly the camera is going to have a buffer overflow and quit recording either within seconds are randomly as you try and use it.  Cards with the V moniker, like V90, enter a different operating mode that allows them to sustain the minimum data-rate required.  

The reason why V90 cards are not readily available (same for V60), is that it isn't just a new sticker on the same old cards, but actually something very challenging and expensive to produce, i.e. it requires the best NAND memory and fast controller hardware and good firmware on the card with a low yield rate no doubt further increasing the cost and lack of availability.

We will find out soon enough who is right, I hope I'm wrong as I don't want to pay a premium and replace good memory cards if I don't need to either..  However it was Panasonic instrumental in introducing this new protocol in SD version 5.0 or higher specifically to allow cameras like the GH5 to run at very high bit-rates to SDXC memory cards, as it couldn't be done otherwise.

Regards

Phil

 

[jonpais]

I've been shooting a bit with ETC mode when I realized it was actually giving me more like a 1.37x crop rather than 1.4x. A comparison I did recently between the Sigma 30mm DC DN in ETC and the Leica 42.5mm f/1.2 showed that there was significantly less crop than 1.4x. The 30mm should have been 84mm with ETC, but it was 82mm. And shooting ETC with the Olympus 75mm, the camera was saying it was a 206mm 35mm equivalent. So I did some quick math and it turns out the crop factor is more like 1.37x, not 1.4x. That is, assuming there is zero crop in 4K mode, which most reviewers seem to corroborate. However, right now, I've got the Leica 42.5mm on my GH5, and shooting without ETC, it's saying it's a 99mm 35mm focal length equivalent, which obviously is not correct. Has anyone else checked this out? There seems to be a bug here or something...

[tweak]

2 hours ago, jonpais said:

I've been shooting a bit with ETC mode when I realized it was actually giving me more like a 1.37x crop rather than 1.4x. A comparison I did recently between the Sigma 30mm DC DN in ETC and the Leica 42.5mm f/1.2 showed that there was significantly less crop than 1.4x. The 30mm should have been 84mm with ETC, but it was 82mm. And shooting ETC with the Olympus 75mm, the camera was saying it was a 206mm 35mm equivalent. So I did some quick math and it turns out the crop factor is more like 1.37x, not 1.4x. That is, assuming there is zero crop in 4K mode, which most reviewers seem to corroborate. However, right now, I've got the Leica 42.5mm on my GH5, and shooting without ETC, it's saying it's a 99mm 35mm focal length equivalent, which obviously is not correct. Has anyone else checked this out? There seems to be a bug here or something...

Possible it's talking about FOV in a diagonal sense (which makes sense). Are you also thinking diagonally? Or horizontally? As FOV and hFOV will obviously be different. If you are talking about regular FOV then there is a crop on GH5 shooting 4K 16:9.

[Jan G]

Hi Jonpais, i also noticed this, so I did a comparison in photo mode using a 25mm prime from panasonic. First I took a shot in 4:3 mode en the exif says 25mm is 50mm in the 35mm equivalent then I took a shot in 16:9 mode and now the exif says 25mm is 54mm in the 35mm equivalent this is the same crop as in movie mode.  So this confirms tweaks answer and is not a bug.

Regards,

Jan

 

 

[Orangenz]

Another GH5 video :D 

 

[tweak]

On 26/09/2017 at 7:27 PM, Jan G said:

Hi Jonpais, i also noticed this, so I did a comparison in photo mode using a 25mm prime from panasonic. First I took a shot in 4:3 mode en the exif says 25mm is 50mm in the 35mm equivalent then I took a shot in 16:9 mode and now the exif says 25mm is 54mm in the 35mm equivalent this is the same crop as in movie mode.  So this confirms tweaks answer and is not a bug.

Regards,

Jan

 

 

Cool man, thanks for checking!

[Orangenz]

On 20/09/2017 at 2:39 AM, philipd said:

We will find out soon enough who is right, I hope I'm wrong as I don't want to pay a premium and replace good memory cards if I don't need to either..  However it was Panasonic instrumental in introducing this new protocol in SD version 5.0 or higher specifically to allow cameras like the GH5 to run at very high bit-rates to SDXC memory cards, as it couldn't be done otherwise.

Regards

Phil

 

As per the picture CInegain put up, the UHS-II (non V rated) cards work fine. I can't test the dual card record function though and it might only work with identical V60 cards. So considering both those points, everyone is right.  

[philipd]

5 hours ago, Orangenz said:

As per the picture CInegain put up, the UHS-II (non V rated) cards work fine. I can't test the dual card record function though and it might only work with identical V60 cards. So considering both those points, everyone is right.  

Hi

Good news, thanks for reporting back, and I've seen successes and failures reported on non V90 cards even if the speed rating suggests it would be okay, so it might be a luck of the draw a little bit. (My Sandisk V30 cards work fine with FHD at 200Mbits/sec but failed after 20 or 30 seconds at 400Mbits/sec for 4K but that was expected). I'd be inclined not to trust anything important to a non V90 card, but otherwise if it works it works.  In terms of backup to two cards, Panasonic have stated in their manual insert update for firmware 2.0 that at least one card must be a V60 or better for the back up function to two cards.

Good on Panasonic for not blocking non rated V60 or 90 cards which I thought they might have done, so the camera at least can have a best effort attempt at any card.

Regards

Phil

[gethin]

Oh spoot. I forgot that the gh5 does not display shutter speed in aperture priority mode. This is so silly.  Does the new firmware?  Really without it you need to be able to set minimum shutter angle

 

[rotgg]

Shot on gh5

 

[Dimitris Stasinos]

Hi guys. I have an issue with 4K 60fps to 30fps conforming into FCPX. I didn't want to open a new thread because it might be something really simple i am missing here.. It worked fine when i shot some stuff at 4K 50 and conformed to 25. But yesterday i tried to do the same with 4k 60fps footage on a 30 fps timeline (with automatic speed of course) and after conforming i noticed that the playback was choppy. I zoomed in and saw duplicate frames, exactly as this was shot at 30 fps and then stretched at half speed....Weird. Do you have any idea what's going on?

P.S. I also shot some stuff in 1080p 60 on the same card (Sandisk V30) and it worked as expected in post, the problem lies only on 4K 60 files, they actually behave as 4k 30...(Firmware 2.00).

[Stathman]

Can you upload a file?

[Emanuel]

On 05/10/2017 at 7:21 AM, Dimitris Stasinos said:

Hi guys. I have an issue with 4K 60fps to 30fps conforming into FCPX. I didn't want to open a new thread because it might be something really simple i am missing here.. It worked fine when i shot some stuff at 4K 50 and conformed to 25. But yesterday i tried to do the same with 4k 60fps footage on a 30 fps timeline (with automatic speed of course) and after conforming i noticed that the playback was choppy. I zoomed in and saw duplicate frames, exactly as this was shot at 30 fps and then stretched at half speed....Weird. Do you have any idea what's going on?

P.S. I also shot some stuff in 1080p 60 on the same card (Sandisk V30) and it worked as expected in post, the problem lies only on 4K 60 files, they actually behave as 4k 30...(Firmware 2.00).

Open the thread :-)

[Dimitris Stasinos]

17 hours ago, Emanuel said:

Open the thread :-)

After some tests i figured this out. When you set GH5 to record 4K 60 fps, for some reason the shutter speed goes up to 1/30 sec and does not stop at 1/60 as it should. So if you accidentally set the shutter speed at 1/50 for example (as i did), then the camera records at 30 fps (it totally makes sense). In FHD 60p mode though the shutter speed can't be set at these values, it stops at 1/60, again as it should be to prevent mistakes. So be careful with the shutter speed setting when you are recording 4K 60 fps, it's different than shooting FHD 60 fps.