Ed Jeffersen
Active member
Vectorscope
GH4 REASONABLE requests for future firmware
- Thread starter tacotim
- Start date
I asked a Panasonic rep in Hollywood about enabling 200Mbps for 4K- their response was "too processor intensive". 100Mbps is too low for 4K- macroblocks and artifacts are clearly visible for motion shots. Those with experience with codecs understand very well that lower bitrates are much harder and more CPU intensive to hit than higher bitrates. In terms of compression, 200Mbps is less CPU intensive than 100Mbps. Additionally, the CPU expense as bitrate goes down tends to be non-linear: lower bitrates are a lot more expensive: e.g. 4x more CPU for 2x bitrate reduction. If the same CPU(s) are used to copy data from buffers to the memory card, then 200Mbps is 2x the CPU vs 100Mbps, so that could be an issue. However, given the parallel nature of these systems I would expect the buffer copy to be handled by specific DMA hardware running effectively on a separate thread and CPU, and thus not related to compression CPU load. My take is it's limited for business reasons: if we want better 4K we'll need to spend more (external recorder or RED, F55, etc. to record internally). The Panasonic Varicam 35 appears to be priced around $60k: that's the Alexa market... Camera manufacturers can't give us everything at once- how will they make sales for future camera upgrades?
Darktide Media
Active member
I think RAW output to the Shogun is something needed for certain projects. Also I would think a LumixLog would be nice, Cine D is good but something more will be nice when shooting 10 Bit out.
...Assuming the SAME 100 Mbps image data is being stored at the 200 Mbps data rate, as I can't image storing 2x as much image data with a 200 Mbps video data stream is going to be less CPU intensive. So 200 Mbps with the SAME artifacts and macro-blocking as the 100 Mbps video stream would likely lower the CPU usage, but doubling the 100 Mbps image data being stored in the 200 Mbps video stream would logically require MORE CPU power. ( no "free lunch" when encoding more data )
Let's make no assumptions:
The GH4 can handle 200Mbps video pipeline throughput, from sensor to memory card.
It doesn't matter if we started with 4K or 2K- it's 200Mbps of bandwidth.
The GH4 uses IPB VBR H.264 compression.
I-frames take less CPU time than P or B frames. P and B frames are algorithmically expensive optimization search problems (specifically, the variable DCT block size motion estimation).
If we double the bitrate we can double the bits for I-frames or double the number of I-frames, increasing quality with either the same CPU load or decreasing it.
Increasing the bitrate puts less demands on the VBR optimization search as less data must be discarded, allowing more high-frequency information to flow through with less CPU search time (optimizing which bits to discard while preserving as much quality as possible).
Bitrate control is also related to adjusting DCT quantization: throwing away more or less bits takes the same CPU time if only adjusting quantization coefficients.
In summary, as bitrate increases, the search for the optimal bits to discard decreases, thus lowering the CPU demand. Bandwidth requirements aren't as expensive as memory copies are highly optimized in hardware.
In theory this sounds do-able, but is the hardware actually able to support processing and storing this extra data continuously without overheating ?
I think only Panasonic engineers are able to answer this question definitively.
FWIW, I doubt that Panasonic intentionally crippled the 4K ability of the GH4 in any way.
True North
Well-known member
Well the camera currently manages the same data rate flow now so why wouldn't the processor be able to handle it?
Agreed
This is the confusing part. Panny seemed to try and go above and beyond to give everyone what they wanted i.e. higher internal bit rates with 10 bit 4:2:2 external output, 4K, Slo-Mo, etc.... It seemed odd to many people that they would offer a 200Mbps Internal HD Codec and limit the 4K Internal to 100Mbps which is arguable where that 200Mbps would be most needed. Logically speaking, it just doesen't make sense that the camera's processor can't manage the load.
We obvioulsy can't know all of the technical limitations of the camera but outside looking in it certainly appears as if 4K 200Mbps would be doable. Oh Well, there is always the GH5...lol!
Again, higher bitrates can mean lower CPU load, as the optimization search time is decreased, so the hardware could run cooler. One business reason not to include 200Mbps 4K at launch could be the scarcity of U3 SDXC cards. The only write speed issues I've read about were with 200Mbps 1080p ALL-I (something folks aren't using much as 4K => 1080p offers the best quality (except for fast motion)). It sounds like Panasonic engineering isn't allowed to release much detail to sales & marketing, so we might not know until or if there is a firmware update.
We have H.265 to look forward to in future cameras. It will provide greater than 2x less data for the same perceived quality. While it's a lot more complex than H.264, it has been designed for a parallel implementation, so embedded solutions will be feasible in the near future (a firmware upgrade to H.265 is thus unlikely). The GH4's 10-bit 422 external output will become handy as rec.2020 10-bit consumer TVs become more popular: this will also provide an option to increase 4K recording bitrate beyond 100Mbps (Atomos Shogun, etc.); an external recorder makes the system no longer portable and stealthy; we can't have everything... yet
David W. Jones
Veteran
Like the GH3, when you utilize the HDMI port on the GH4, your audio monitoring is switched to recorded sound only, so you loose the ability to monitor realtime without the delay.
Having the ability to switch between realtime and recorded sound while using an external monitor would be nice.
Having the ability to switch between realtime and recorded sound while using an external monitor would be nice.
I can.
You have to remember that the data is not being 'thrown out' first, and then the remaining image encoded. The data is being 'thrown out' during the encode process, according to an algorithm. I'll be the first to admit lack of expertise on this subject, but it does seem theoretically sound that 'throwing out' less data would be less computationally expensive, and thus require less processing power, just as throwing out none of it would require the least processing power.
Yes, if you are using the exact same compression mode and parameters, a higher final bitrate will take less CPU power to compress to than a lower bitrate. If you think of the simple analogy of the relative amount of effort that it would take to pack 3 suitcases worth of clothes into 1 suitcase, vs packing the same amount of clothes into 2 suitcases instead, you get the simple idea of why this is.
However, if you are using a different packing method (All-I vs IPB), that may not apply. Unless you know that All-I uses a lot less CPU power (and RAM) than IPB, you don't know if the CPU will be able to handle the quadrupling (plus) in number of pixels when going to 4K All-I 200mbps. Remember that they need to do so safely, considering heat dissipation and all.
However, if you are using a different packing method (All-I vs IPB), that may not apply. Unless you know that All-I uses a lot less CPU power (and RAM) than IPB, you don't know if the CPU will be able to handle the quadrupling (plus) in number of pixels when going to 4K All-I 200mbps. Remember that they need to do so safely, considering heat dissipation and all.
Last edited:
Mike Harvey
Veteran
Finally am able to start playing around with my GH4... my request is (assuming it's a software and not a hardware thing) to do away with the 4GB limit on exFAT SDXC cards. I get it for AVCHD, but there is also metadata to string those clips back together in my NLE. The lack of metadata for the 100mbs and 200mbs modes could make this a pain in the rear-end to string along clips during long interviews or events.
Camerapilot
Active member
If it's not a hardware issue, I would love to see the video out delay much minimized. Right now it's the worst I have seen in any camera.
I strongly support the OP request no. 1, and I wish to add one of my own, although it concerns only the users of the european recording-time-crippled cameras.
Say that I have a big card where I already recorded some video. The problem is that I have no way to know how much space is still available.
All I can see is that I can still record 29:59 of video (the maximum EU limit), but in reality I will be able to record much more (albeit in more sequences).
Then this is my request: please make it possible to display the total amount of consumed/available memory in the card (in GB).