Optimized media taking 12 hours!

... or just buy the studio version and enjoy GPU decoding. You may not even need optimized media at all with H264 at 100Mbit.

He shouldn't even need the studio version.

However, 12h doesn't sound so wrong for 350 GB.

Sent from my SM-N950F using Tapatalk

Really? Even on a high end machine?
What about professionals who take terabytes of footage? Do they go for a sabbatical to Peru?

lol

350 GB of highly compressed footage. That means it has to be uncompressed, meaning lots of decoding time + normal render time. Even on a modest workstation, I would expect about that much. Drone footage is usually small in file size, which means @350GB we are looking at a long runtime (just educatedly guessing here). Here's my math 350GB ~ 350000MB/7.5MBs(4k DJI Bitrate)/60 for min/60 for hours = 12.96 hours of drone footage. That assumes the 4K bitrate is 60mb/s. Your machine is rendering 4K DNX at a little over realtime if my math is correct. Sounds right.

What's the speed of the volume you are rendering to? Could it be that you are rendering to a slow volume and it can't render any faster due to saturating average write speed?

[Peter Chamberlain]

Hey guys, lets keep the discussion civil.... focus on the issue not the individuals.

[Peter Chamberlain]

Nathan, you can render to DNxHD from any source so my recommendation was to understand more about the system. Rendering to DNxHD lets you then use that clip as a source to play in a new timeline, at HD, or at 4K and see that you can get 24fps. Check the CPU usage %.

If that all works as expected then the focus is on the source codec, which I can now see from your media info report is a heavily compressed UHD H.264 @23.976.

So, assuming your timeline is at 23.976, and you play the clip with no cache, you can see what % of the CPU is busy. My guess is the six core, overclocked or not, is going to be the bottleneck.

You can then decide if the Studio version will help you, if you need a better CPU, or other options to improved the playback.

Also note, render cache or optimized media to 1/4 res means the UHD image is made to a much lower res, and then Resolve GPU has to up-res to your timeline res, assuming you are using HD or UHD. There is false economy there. If you are working in HD timeline, make your optimized media HD. If you want to render UHD later, turn the optimized media off, reset the timeline to UHD and render. At least during editing and grading it will be faster.

... or just buy the studio version and enjoy GPU decoding. You may not even need optimized media at all with H264 at 100Mbit.

We don't even know what's causing my issue though. That's a lot of money to spend on something which may or may not leave me with exactly the same problem.

It's ridiculous that there isn't even an option to try it for a limited time

... or just buy the studio version and enjoy GPU decoding. You may not even need optimized media at all with H264 at 100Mbit.

He shouldn't even need the studio version.

Thank you. This is what I've been led to believe through all the research I've done. It's a bit rich for someone to suggest I spend close to AU$500 for something that may not solve my problem.

Nathan, you can render to DNxHD from any source so my recommendation was to understand more about the system. Rendering to DNxHD lets you then use that clip as a source to play in a new timeline, at HD, or at 4K and see that you can get 24fps. Check the CPU usage %.

If that all works as expected then the focus is on the source codec, which I can now see from your media info report is a heavily compressed UHD H.264 @23.976.

So, assuming your timeline is at 23.976, and you play the clip with no cache, you can see what % of the CPU is busy. My guess is the six core, overclocked or not, is going to be the bottleneck.

You can then decide if the Studio version will help you, if you need a better CPU, or other options to improved the playback.

Also note, render cache or optimized media to 1/4 res means the UHD image is made to a much lower res, and then Resolve GPU has to up-res to your timeline res, assuming you are using HD or UHD. There is false economy there. If you are working in HD timeline, make your optimized media HD. If you want to render UHD later, turn the optimized media off, reset the timeline to UHD and render. At least during editing and grading it will be faster.

Thank you Peter for the first genuinely helpful comment on here.
Although I'm not sure how seeing CPU usage can help me decide whether to buy the studio version if I don't know what the performance difference will be.

Isn't there any way I can try the studio version for a limited time (even just 24 hours) to justify the purchase?

[Peter Chamberlain]

No, we don't offer that option.

You could use FFMPEG to check how fast your system could do the transcode with GPU acceleration. It's likely in a similar ballpark to what Resolve Studio should be able to do. Then again you may not need optimized media then anyway. But I'm repeating myself. Do whatever you feel is appropriate for your case.

Sent from my SM-N950F using Tapatalk

Thank you. This is what I've been led to believe through all the research I've done. It's a bit rich for someone to suggest I spend close to AU$500 for something that may not solve my problem.

I said you "shouldn't" that's not an absolute statement. Not to mention i said that yesterday as well.

At this point I don't see what you hope to gain from continuing in this thread. You have given no additional details about your hardware, even though its been asked for.

Yes, it would be nice to know which motherboard is used and especially how the Samsung 970 EVO NVME SSD and Samsung 870 EVO NVME SSD are connected! Something is wrong if you can't playback optimized media Cineform - DNxHD / HR in real time! Maybe perform a Blackmagic Disc Speed test !

The first thing that comes to mind is that your MB might be screwing you over by sharing lanes between your drives and gpu, but i would need to know what mb you are using to validate that.

The MB does not share PCIE lanes between the GPU and SATA drives. AKA the other stuff that is connected to the chipset.

he isn't using sata drives, he's using pcie drives. 970's require 4 pcie lanes each to run at max throughput.

Doesn't matter. Still ultimately going through the DMI port. Those PCIE lanes are coming from a southbridge chipset. If taken from the CPU PCIE offering then a GPU can never run with x16 PCIE lanes.

Again, I don't think max throughput is an issue as stated. The source bitrate and the output DNx birates combined are very small in the scheme of things. The OP talked about trying 1/2 res for optimized media with no change. Even a single SATA 6Gbps channel is more than enough and SSD's can handle a tons of I/O ops per second as well. So even if only 2 PCIE lanes (~10GT/s) are assigned to the M2 drive that is faster than SATA.

Those PCIE lanes are coming from a southbridge chipset.

Maybe not always !

Asus x99 delux motherboard

http://techreport.com/review/26973/asus-x99-deluxe-motherboard-reviewed

Most motherboards' M.2 implementations are restricted to dual Gen2 lanes
from the chipset. On top of that, they're bound by limited interconnect
bandwidth shared with the X99's other I/O components. The Deluxe's M.2
socket bypasses those handicaps by hooking directly into the CPU.

M.2 And SATA Express, M.2 PCIe

https://www.tomshardware.com/reviews/samsung-xp941-z97-pci-express,3826-2.html

Ultra M.2, which is a second slot tapping into a Haswell-based CPU's 16 lanes
of third-gen PCIe, too. This slot isn't affected by the chipset. And if you
drop a PCIe M.2 drive into the Ultra slot, you can still use SATA Express,
which is wired into Z97. In exchange, you can't run a graphics card using
the processor's 16 lanes, instead bumping it down to eight.

Ole Kristiansen is correct, depending on the motherboard you can get all kinds of weird pcie allocations. Hence the reason why knowing what MB someone is working with can be important.

Also as the benchmark I linked to earlier shows even an m.2 ssd can slow to a crawl if you are doing any amount of random read write.

... or just buy the studio version and enjoy GPU decoding. You may not even need optimized media at all with H264 at 100Mbit.

Does the GPU decoding still only work for 8-bit 4:2:0 H264 sources and not 10-bit 4:2:2?

Thanks!

... or just buy the studio version and enjoy GPU decoding. You may not even need optimized media at all with H264 at 100Mbit.

Does the GPU decoding still only work for 8-bit 4:2:0 H264 sources and not 10-bit 4:2:2?

Thanks!

Yes, you are limited to 8-bit 4:2:0 for h.264, This is not a limitation of resolve but of the Nvidia Api.

... or just buy the studio version and enjoy GPU decoding. You may not even need optimized media at all with H264 at 100Mbit.

Does the GPU decoding still only work for 8-bit 4:2:0 H264 sources and not 10-bit 4:2:2?

Thanks!

Yes, you are limited to 8-bit 4:2:0 for h.264, This is not a limitation of resolve but of the Nvidia Api.

Thanks for the confirmation!


Sent from my iPhone using Tapatalk