Best File Location for Performance

Did you record 4K 60P on the GoPro9 ? I think that is h265 which may be the reason for your performance issues. Studio should make a difference.

Yes, I did shoot it in 4K/60 fps.

Just bought the license.... WOW!!!! took only 2:57 to export the file.
So the Studio DOES makes a difference.

The settings I used are attached

Use longer fades, don't use saturated "anime" colors (specially red). Actually white (or rather very bright grey) looks the best on black background. Cuts, fades etc. link to music tempo.
Use higher bitrate for export as you have your fades badly compressed (maybe it's YT, but I think it's already on the source).

Thank you for the tips about transitions, I was too excited to get this 1st video out... LOL
And yes, I need to match the video to the music in a better way

Regarding the colors, I was trying to play with the color correction because I did not use red filter on the camera and the image was with greenish ambiance unless it was close to my lights. There is long way for me to learn

I think YT created several files and compressed them again. I have to play with more exports and uploading to find the best match.

According to this comparison at:https://artgrid.io/insights/davinci-resolve-free-vs-studio/, Studio does utilize GPU accelerators compared to the free version:
"...One of the key advantages of the studio version over the free version is its use of GPU acceleration, including being able to use multiple GPUs. The studio version has GPU accelerated encoding and decoding of the widely used H.264 and H.265 formats that can greatly speed up editing and rendering performance..."

Are you saying it is not true?

The issue is that article is a 35,000 foot overview.

In the real world it's a lot more complicated than that. Take a look at this article, by a well respected organization!
https://www.pugetsystems.com/labs/artic ... olveStudio

Specifically take note even with the latest generation gpus only 8bit 4:2:0 h.264 decode is hardware accelerated. Something like 10 bit 4:2:0 or 4:2:2 from a gh5 is decoded by the cpu.


and it goes even deeper.....
https://developer.nvidia.com/video-enco ... matrix-new

What gpu you have, and what generation it is, can play a part in what you can encode and decode. Some formats can't be decoded, but they can be encoded.

Apart from de- or encoding, there's only one case where speed will be improved with the Studio version: multiple GPUs. Apart from one old MacPro model, the free version is supporting only one to compute.

Apart from de- or encoding, there's only one case where speed will be improved with the Studio version: multiple GPUs. Apart from one old MacPro model, the free version is supporting only one to compute.

But the main value IS decode / encode above UHD. For me that was the sole reason to get the Studio Max. I wanted to make sure that by spending money on new equipment I would get the ability to edit native h264 and h265 files at 10 bit 4:2:2 in realtime. Choices here are limited to a M1 or M2 Mac with hardware decode/encode or Intel QS. At the time I bought my Studio Max it was about $300 more than a 3090 that was not available other than to buy a full Dell system at almost three times the cost. Choice was simple even though I am really a Windows guy. My first Apple product as my phones are Android.

Enjoy.

I wanted to make sure that by spending money on new equipment I would get the ability to edit native h264 and h265 files at 10 bit 4:2:2 in realtime.

The build in my sig will happily do that using the cpu to decode 10 bit 4:2:2 h264, and it's nothing special.

I wanted to make sure that by spending money on new equipment I would get the ability to edit native h264 and h265 files at 10 bit 4:2:2 in realtime.

The build in my sig will happily do that using the cpu to decode 10 bit 4:2:2 h264, and it's nothing special.

I am surprised for such an old system. My old Threadripper will also just manage one track with nothing applied to it but that is not good enough for real editing. Would like to know what camera files of this type you have actually tried.

I am surprised for such an old system. My old Threadripper will also just manage one track with nothing applied to it but that is not good enough for real editing. Would like to know what camera files of this type you have actually tried.

For personal projects (what the machine in my sig gets used for), I use an A & B camera and a seperate audio recorder, with everything synced via time code.

• A camera - Panasonic GH5, shooting 29.97fps UHD, 4:2:2 10bit IPB h.264 at 150Mbps
• B camera - Panasonic G7, shooting 29.97fps UHD 4:2:0 8bit IPB h.264 at 100Mbps
• audio recorder - Zoom f4 recording 2 mics at 24bit 96Khz

the cameras are synced to the recorder with tentacle syncs.

• source footage is stored on a drive consisting of 3 hdds in Raid 0
• cache is on a second drive consisting of 3 other hdds in Raid 0
• deliverables/output is rendered to a 3rd drive consisting of 2 other hdds in Raid 1
• The os and all application files all live on the C drive thats is an nvme m.2

Editing wise, I use a lut to get a good starting point and then light color grading as needed. cross fades quickly render and live on cache.

Audio is more of an issue for me, because im doing a lot to it, dereverb, noise reduction, loudness normalization, Eq, compression etc.

The new machine sitting in the other room in boxes (still waiting on some parts to arrive) will blow my current machine away.

I run V Log on both the GH6 and the GH5S. Use color management and set input color space in media page. Do not need to do much other than just set levels of cameras in color page afterwards. AX1000 often needs noise reduction though as it is not as sensitive as the GH cameras. Fusion titles I think stress the Studio Max more than the PC.

Fusion titles I think stress the Studio Max more than the PC.

Most likely because it doesn't have any dedicated vram, and doesn't clock very high.

Fot tasks which require pure GPU power and fast memory it will suffer, but for every day tasks which share load between GPU+CPU it will shine as data is in the same memory. Nothing is perfect or universal.

Let me put it this way: I am using drive that is slower than yours (Samsung 970 Evo Plus, which is PCIe 3.0 drive on a PCIe 3.0 bus with theoretical max of "only" 4GB/sec) yet not once so far I wished for more drive speed when using Resolve, it is GPU and CPU that are my bottleneck. That is why I am saying that once one identifies which layouts are good (or not) further focus on drive benchmarks and figures becomes more theoretical than practical.

How does it measure with the Resolve Speed Test app? Try that and see what it says -- I find it's a good barometer of potential performance.

Do you mean RAW speed test? If yes, here are my results. On their own figures seem good enough. It's not the fastest system in the world but it seems decent enough. However, in certain situations (like when rendering cache) my CPU maxes out across all cores for a very long time while SSD is barely breaking a sweat.

BlackmagicRAWSpeedTest.png (513.03 KiB) Viewed 2462 times

There are also free Resolve Standard Candle Tests out there: straightforward projects and media that you can run on different machines to get a real-world idea of how they compare.

DaVinci Resolve 'Standard Candle UHD': Rocket Science
https://forums.macrumors.com/threads/da ... e.2273175/

Thank you VERY much for pointing me to these, I wasn't aware of them

@4EvrYng:

That's RAW decode test, I believe it runs purely from RAM, no disk involved.

I believe this is the right test, it's Blackmagic Disk Speed Test, here are my Kingston A2000 results for 1GB file:

Kingston A2000, 1GB file:

BMD_Disk_Speed_A2000_1GB_file.jpg (85.6 KiB) Viewed 2413 times

With 5GB and 10GB files, cumulative writes are 500 and 350 MB/s respectively but the progression varies drastically between runs, one time it starts slow then speeds up, another time it stars fast then slows. Oh woe.

@4EvrYng:
With 5GB and 10GB files, cumulative writes are 500 and 350 MB/s respectively but the progression varies drastically between runs, one time it starts slow then speeds up, another time it stars fast then slows. Oh woe.

If your machine has enough ram some OSs will use it as a big write cache, and the drives themselves have onboard cache. So depending on your ram and disk cache usage you can get dramatically different results.

@4EvrYng:

That's RAW decode test, I believe it runs purely from RAM, no disk involved.

I believe this is the right test, it's Blackmagic Disk Speed Test, here are my Kingston A2000 results for 1GB file:

The attachment BMD_Disk_Speed_A2000_1GB_file.jpg is no longer available

With 5GB and 10GB files, cumulative writes are 500 and 350 MB/s respectively but the progression varies drastically between runs, one time it starts slow then speeds up, another time it stars fast then slows. Oh woe.

Here are results of DiskSpeed test for my system... From left to right (all Samsung Evo 970 Plus 2 TB):
C: (Windows system), 4KB cluster, test with 5 GB file
E: (cache volume), 64KB cluster, test with 1 GB file
E: (cache volume), 64KB cluster, test with 5 GB file

DiskSpeedTest.jpg (717.57 KiB) Viewed 2324 times

Yes, my disks are faster and sustain longer -BUT- you can say "woe is me" -ONLY- if your disk is not -FAST ENOUGH-, if it is the weakest component in the link. Is it, though? If you feel it is you can solve your problem for less than $200, good Samsungs these days are around $100 per TB.

In my case disk is obviously not a problem, it is not even breaking a sweat, yet I have been frustrated with playback performance problems last days.

[*]A camera - Panasonic GH5, shooting 29.97fps UHD, 4:2:2 10bit IPB h.264 at 150Mbps

264 is less compressed than 265. Try testing it with a sample 4.2.2 265 file.

Here are results of DiskSpeed test for my system... From left to right (all Samsung Evo 970 Plus 2 TB):
C: (Windows system), 4KB cluster, test with 5 GB file
E: (cache volume), 64KB cluster, test with 1 GB file
E: (cache volume), 64KB cluster, test with 5 GB file

DiskSpeedTest.jpg

Yes, my disks are faster and sustain longer -BUT- you can say "woe is me" -ONLY- if your disk is not -FAST ENOUGH-, if it is the weakest component in the link. Is it, though? If you feel it is you can solve your problem for less than $200, good Samsungs these days are around $100 per TB.

In my case disk is obviously not a problem, it is not even breaking a sweat, yet I have been frustrated with playback performance problems last days.

Actually, it's not influencing my Resolve workflow that much, it seems, because cache files are usually short and mostly don't go beyond 1GB. Surely, 2000 MB/s reads and 1500 MB/s writes are plenty, however badly they compare to a Samsung EVO.

I was just frustrated that my Puget 4K Media benchmark is half of what it should be because of stalled writes.

Sent from my Mi 9T using Tapatalk

[*]A camera - Panasonic GH5, shooting 29.97fps UHD, 4:2:2 10bit IPB h.264 at 150Mbps

264 is less compressed than 265. Try testing it with a sample 4.2.2 265 file.

You are quoting Dan's shooting. My A camera is the GH6( C4K or 5.7K ) and B camera GH5S ( C4K ) and C camera Sony AX100( shooting 60P HD ). I only shoot 60P so either 4:2:2 10bit or 8bit depending on camera and only choice I have for h265 is the GH6 and that is 4:2:0 10bit. My PC has no problem with 29.97 P h264 either but it is not a frame rate I use. The Panasonic GH cameras shoot V Log too. So even simple playback involves decode and color space conversion. So to be clear. My PC can work fine with the GH5S h264 in either 4:2:0 150Mbps or 4:2:2 29.97 that I do not use but checked. It can even just about play the GH6 h265 10bit 5.7K files occasionally dropping a frame. However put all three cameras in a multicam 1920x1080 project and the PC will just stop. This is where the hardware decodes of the Studio Max come in as the same project ( and I mean the same project all from the USB C SSD plugged into to either the PC or Studio Max ) multicam is smooth full speed. Are there other differences? Yes , Fusion titles may actually be a little quicker on the PC and I am also not sure if NR or sharpening is also a little faster on the PC ( these are usually applied to the AX100HD file ). I have no test proof or any interest in finding out really but I expect the Threadripper/1080Ti may have the advantage with some things.

The whole point of my comments were that the drives are of no concern if the source files are compressed h264 or h265. These data rates are so low that a thumbdrive is fast enough. Remember most cameras record to an SD card. Different if the source is ProRes etc when drive speed is very important but then load on CPU and GPU is low. When I export to either ProRes or GrassValley HQX that goes to an NVME drive on the PC or internal on the Mac.

The Panasonic GH cameras shoot V Log too. So even simple playback involves decode and color space conversion.

This to makes almost no difference as all you are doing is applying a lut!

Actually, it's not influencing my Resolve workflow that much, it seems, because cache files are usually short and mostly don't go beyond 1GB. Surely, 2000 MB/s reads and 1500 MB/s writes are plenty, however badly they compare to a Samsung EVO.

I was just frustrated that my Puget 4K Media benchmark is half of what it should be because of stalled writes.

I understand why benchmark results can result in frustration. In my previous life I was a person responsible for hundreds of servers whose performance was critical and I frequently had to evaluate performance issues, identify culprits, and recommend correction plan. Due to it I have come to say to people "there are lies, darn lies, and benchmarks". Benchmark can help identify bottlenecks IF used cautiously. OR they can lead you down the rabbit's hole. It is irrelevant how "slow" benchmark says one system is when compared to another, what matters is does system meet expectations of the performance, does it deliver results fast enough, by the user. My 970s aren't fastest NVMEs out there, their Puget scores will be lower. BUT when I look at performance counters while playing back / rendering their bandwidth is nowhere near saturated, disk queue is practical zero, and response times are practically sub-millisecond. So I ignore their "lower" benchmark score.

The whole point of my comments were that the drives are of no concern if the source files are compressed h264 or h265. These data rates are so low that a thumbdrive is fast enough. Remember most cameras record to an SD card. Different if the source is ProRes etc when drive speed is very important but then load on CPU and GPU is low. When I export to either ProRes or GrassValley HQX that goes to an NVME drive on the PC or internal on the Mac.

ProRes is still not the end of the world, but uncompressed YUV/RGB or DPX/TIFF/EXR will hit you very badly.

The whole point of my comments were that the drives are of no concern if the source files are compressed h264 or h265. These data rates are so low that a thumbdrive is fast enough. Remember most cameras record to an SD card. Different if the source is ProRes etc when drive speed is very important but then load on CPU and GPU is low. When I export to either ProRes or GrassValley HQX that goes to an NVME drive on the PC or internal on the Mac.

ProRes is still not the end of the world, but uncompressed YUV/RGB or DPX/TIFF/EXR will hit you very badly.

Yes agreed and on my GH6 C4k 60P ProRes is 1.9Gbps as compared to h264 at 200Mbps quite the difference load on drives.

The Panasonic GH cameras shoot V Log too. So even simple playback involves decode and color space conversion.

This to makes almost no difference as all you are doing is applying a lut!

Well colour space conversion ( colour management or CST ) is a better way of converting colour spaces than a LUT and then apply a LUT if you want for "look " but decode is real and does make a difference. If you are running a codec that can be decoded with CPU then fine. However I have ample proof on my systems for the files I use that my Threadripper/1080Ti is not up to the performance level I want to edit multicam but the Studio Max does work. The main difference is the hardware decoders/encoders. You can believe what you want but I have both systems to compare with exactly the same files and project from the same SSD drive. Both are using the same level of USB interface on their systems. The limitation is not interface or drive it is CPU/GPU/hardware decode/encode. As I mentioned I think that looking at encoding fps the PC is a little faster than the Mac for some things like Fusion titles.

However I have ample proof on my systems for the files I use that my Threadripper/1080Ti is not up to the performance level I want to edit multicam but the Studio Max does work. The main difference is the hardware decoders/encoders. ... The limitation is not interface or drive it is CPU/GPU/hardware decode/encode.

After several frustrating days with my Nvidia 2080 Super based Windows PC I am starting to feel like you. My decoding/encoding are GPU accelerated but I don't need acceleration just at the beginning and end of the workflow, I need it at every step possible, and as far as I know none of the codecs available for render cache are hardware accelerated. So I am speculating that my pain would be addressed with Mac's ProRes engines (if I can ever afford the move).

Which codecs you use for proxies / render cache on your Mac and do they end up (noticeably) accelerated, please?

One of the points of the easy to edit codecs is you don't need HW acceleration.

What sort of CPU are you using?

One of the points of the easy to edit codecs is you don't need HW acceleration.

What sort of CPU are you using?

I have i9-10920X 12-core 24-threads with all threads running at 4.3 GHz in X299 chipset motherboard with 128GB of memory. I need to process 4K60 10-bit footage. Which codec(s) you believe I should be using for proxies and render cache, please?

I'd suggest DNxHR.
SQ or LB won't need much space for proxies. But with your pretty good CPU and enough space, you could even try to transcode to HQX 10 bit and use it for finishing.

I'd suggest DNxHR.
SQ or LB won't need much space for proxies. But with your pretty good CPU and enough space, you could even try to transcode to HQX 10 bit and use it for finishing.

DNxHR HQX at half resolution is what I initially figured and tried. It encoded very simple cache of 4 minute clip in approximately under 2 minutes with all CPUs at approximately 45%. Playback was a disaster though. Clip I could play at 60fps without render cache was getting around 50fps playing with DNxHR HQX render cache. Two of CPUs were at 100%. Then I tried DNxHR LB and I was still getting dropped frames (less of them, but still), one CPU was still at 100% and other at approximately 40%. Hence my wish to get hardware acceleration of render caching.

Hmm, it seems that hardware acceleration for mezzanine codecs is not such a bad idea.
Only Apple does it until now.

Hmm, it seems that hardware acceleration for mezzanine codecs is not such a bad idea. Only Apple does it until now.

I would argue that acceleration of anything is always a good idea I have zero knowledge of what is hardware accelerated in Resolve vs. Premiere but I would like to think that if Adobe can do something that Black Magic already isn't then Black Magic should be able to do it and even better. So, Black Magic, what are we waiting for?

Hardware acceleration must be offered by your computer these days. The times of a Red Rocket or anything like that are over. Nobody would make money with such a niche product, as far as I can see.

OK, I admit that there's still the Fairlight Audio Accelerator and Avid has something similar for ProTools.
But for video? Even the Avid Mojo DX Media Composer Accelerator can be found only second-hand.

Everybody relies on general GPUs these days and capabilities built into these or the CPU.

Hardware acceleration must be offered by your computer these days. The times of a Red Rocket or anything like that are over. Nobody would make money with such a niche product, as far as I can see.

OK, I admit that there's still the Fairlight Audio Accelerator and Avid has something similar for ProTools.
But for video? Even the Avid Mojo DX Media Composer Accelerator can be found only second-hand.

Everybody relies on general GPUs these days and capabilities built into these or the CPU.

I didn't mean BMD should create hardware product. What I meant is that in general anything that can be accelerated should be accelerated, if there is a functionality/feature in Premiere that takes advantage of hw acceleration but Resolve isn't then Resolve should.

I'd suggest DNxHR.
SQ or LB won't need much space for proxies. But with your pretty good CPU and enough space, you could even try to transcode to HQX 10 bit and use it for finishing.

DNxHR HQX at half resolution is what I initially figured and tried. It encoded very simple cache of 4 minute clip in approximately under 2 minutes with all CPUs at approximately 45%. Playback was a disaster though. Clip I could play at 60fps without render cache was getting around 50fps playing with DNxHR HQX render cache. Two of CPUs were at 100%. Then I tried DNxHR LB and I was still getting dropped frames (less of them, but still), one CPU was still at 100% and other at approximately 40%. Hence my wish to get hardware acceleration of render caching.

I'd wager you've got something else going on.

Drives? What does task manager show while this is happening? Are the drives keeping up?

The whole point of my comments were that the drives are of no concern if the source files are compressed h264 or h265. These data rates are so low that a thumbdrive is fast enough. Remember most cameras record to an SD card. Different if the source is ProRes etc when drive speed is very important but then load on CPU and GPU is low. When I export to either ProRes or GrassValley HQX that goes to an NVME drive on the PC or internal on the Mac.

ProRes is still not the end of the world, but uncompressed YUV/RGB or DPX/TIFF/EXR will hit you very badly.

Yes agreed and on my GH6 C4k 60P ProRes is 1.9Gbps as compared to h264 at 200Mbps quite the difference load on drives.

Yes, but 2gbit/s is still “just” 250MB/s so not a problem for simple single SSD.

However I have ample proof on my systems for the files I use that my Threadripper/1080Ti is not up to the performance level I want to edit multicam but the Studio Max does work. The main difference is the hardware decoders/encoders. ... The limitation is not interface or drive it is CPU/GPU/hardware decode/encode.

After several frustrating days with my Nvidia 2080 Super based Windows PC I am starting to feel like you. My decoding/encoding are GPU accelerated but I don't need acceleration just at the beginning and end of the workflow, I need it at every step possible, and as far as I know none of the codecs available for render cache are hardware accelerated. So I am speculating that my pain would be addressed with Mac's ProRes engines (if I can ever afford the move).

Which codecs you use for proxies / render cache on your Mac and do they end up (noticeably) accelerated, please?

I do not need proxies or cache for normal workflow multicam with GH6 h265 5.7K 10 bit 60P, GH5S C4K 60P 10 bit 4:2:2 files and the AX100 XAVC-S HD file which isn't a problem anyway, but is in the multicam to be dealt with. They just play fine in a 1920x1080 60P timeline. My whole point in moving to the Mac. Hardware decode while editing. Of course noise reduction or even Fusion will slow it down ( I even think it slows more than the Threadripper/ 1080Ti ) The final render is faster on the Mac ( because of the hardware decode I am sure ) but not something that bothered me on the PC anyway. I often bring project to the PC to render as I can then put output on one of the NVME drives for further encoding for DVD and Bluray. If I want to see how Fusion titles etc run I let it cache to ProRes LT full resolution but then switch off as I do not want it caching everything !! The Studio Max of course has hardware decode/encode for ProRes too.

I think the point I want to make and Andrew keeps saying too, if source files are compressed then the drives could be almost anything as they do not figure in the performance. CPU,GPU and hardware decode will be the control of performance on the timeline. Very different for uncompressed files.

I'd suggest DNxHR.
SQ or LB won't need much space for proxies. But with your pretty good CPU and enough space, you could even try to transcode to HQX 10 bit and use it for finishing.

DNxHR HQX at half resolution is what I initially figured and tried. It encoded very simple cache of 4 minute clip in approximately under 2 minutes with all CPUs at approximately 45%. Playback was a disaster though. Clip I could play at 60fps without render cache was getting around 50fps playing with DNxHR HQX render cache. Two of CPUs were at 100%. Then I tried DNxHR LB and I was still getting dropped frames (less of them, but still), one CPU was still at 100% and other at approximately 40%. Hence my wish to get hardware acceleration of render caching.

I'd wager you've got something else going on.

Drives? What does task manager show while this is happening? Are the drives keeping up?

Thing is I can't find anything going on besides CPU. Task Manager, Resource Monitor and performance counters are first things I looked at and when running with render cache on system disk had practical zero utilization, disk with source media had absolute zero utilization, and disk with render cache was active less than 2% of the time reading 50-60 MB/sec with average 0.2ms response time, average disk queue less than 0.01, and zero split I/O.

At the same time two CPUs were at 90% and four around 20-40%. So only thing I can suspect is CPU.

Out of interest I decided to set up the Studio Max render and PC render as close as possible . So exported to h265 on both. Exporting to internal discs on both ( so from the external SSD attached to both systems to internal drive on Mac and NVME on PC ) Same project file. The 1 hour 59 min project encoded on the Studio Max in 1 hour 33 mins and on the PC in 1 hour 26 mins. So even with the hardware decodes on the Mac the PC was faster and I think the difference is encoding of the noise reduction, sharpening and the Fusion titles especially the scrolling titles. I could watch the Studio Max slow down to less than 3 fps for those where the PC was more like 14fps. So I expect even the 1080Ti may be more powerful than the GPU in the M1 Max ? I even think the NVENC encode looks nicer too. Confirmed my approach though . Edit on Studio Max for nice timeline response while editing, integration with iPad that I can use to watch on my TV with Airplay but do my encoding on the PC. I am sure the fastest solution would be a new PC but why do that when for just the cost of the GPU one can get a Mac that works great for the editing process. Just a RTX 4090 is close to the cost of the Studio Max in Canada . Actually $300 more than the spec on my Studio Max !!

Actually, it's not influencing my Resolve workflow that much, it seems, because cache files are usually short and mostly don't go beyond 1GB. Surely, 2000 MB/s reads and 1500 MB/s writes are plenty, however badly they compare to a Samsung EVO.

I was just frustrated that my Puget 4K Media benchmark is half of what it should be because of stalled writes.

Sent from my Mi 9T using Tapatalk

Just tested my Samsung EVO Pro 980 2TB with 1G and 5G. It looks like very similar results.
The raw test on the RTX3060 actually surprised me given the card is considered low-end.

Just tested my Samsung EVO Pro 980 2TB with 1G and 5G. It looks like very similar results. The raw test on the RTX3060 actually surprised me given the card is considered low-end.

I don't have hands on experience with 980 Pros so I can't comment should your results be higher than mine. On one side they are PCIe 4.0 compared to my 3.0, on the other side this might be reflecting differences in our CPU setup.

RTX3060 might be considered "low end" when compared to all those 3090s and 4090s out there but in the grand scheme of things it isn't a bad card.

Out of interest I decided to set up the Studio Max render and PC render as close as possible . So exported to h265 on both. Exporting to internal discs on both (so from the external SSD attached to both systems to internal drive on Mac and NVME on PC ) Same project file. The 1 hour 59 min project encoded on the Studio Max in 1 hour 33 mins and on the PC in 1 hour 26 mins.

Thank you for doing this! Personally I spend much more time in the timeline than exporting for that less than 10% difference in export to matter to me. Can you comment on render cache differences, its creation speed and timeline playback, please?

Out of interest I decided to set up the Studio Max render and PC render as close as possible . So exported to h265 on both. Exporting to internal discs on both (so from the external SSD attached to both systems to internal drive on Mac and NVME on PC ) Same project file. The 1 hour 59 min project encoded on the Studio Max in 1 hour 33 mins and on the PC in 1 hour 26 mins.

Thank you for doing this! Personally I spend much more time in the timeline than exporting for that less than 10% difference in export to matter to me. Can you comment on render cache differences, its creation speed and timeline playback, please?

I will try some tomorrow as I normally do not use any full time just use smart cache to check credit rolls etc then turn off. Will have to set up a test. The whole point for me is to run native files in the timeline. Most of my project are around 2 hours so with 3 cameras the last thing I want to do is spend several hours making proxies and then use drive space unnecessary .

Out of interest I decided to set up the Studio Max render and PC render as close as possible . So exported to h265 on both. Exporting to internal discs on both (so from the external SSD attached to both systems to internal drive on Mac and NVME on PC ) Same project file. The 1 hour 59 min project encoded on the Studio Max in 1 hour 33 mins and on the PC in 1 hour 26 mins.

Thank you for doing this! Personally I spend much more time in the timeline than exporting for that less than 10% difference in export to matter to me. Can you comment on render cache differences, its creation speed and timeline playback, please?

I will try some tomorrow as I normally do not use any full time just use smart cache to check credit rolls etc then turn off. Will have to set up a test.

Thank you in advance! If you have time to do it I would very much appreciate test with 4K 60fps 10-bit 100-120 Mb/sec bit rate footage so I can compare playback speed result against what I get on my system.

Just tested my Samsung EVO Pro 980 2TB with 1G and 5G. It looks like very similar results.
The raw test on the RTX3060 actually surprised me given the card is considered low-end.

I get 68 fps with my RX 6700, otherwise they are comparable in other benchmarks. Beside CUDA, the difference may also be partly due to mine running @PCI 3.0 x16 and not 4.0 because of the motherboard (Asus withdrew Ryzen 3xxx 4.0 support). The thinking behind this is: I observed 30% lower score with PCI x8, thus I expect some difference between 4.0 and 3.0 x16.
They're not low end GPUs, the perception of low is because Nvidia only has one RTX, and really bare, GPU beneath it.

RTX3060 might be considered "low end" when compared to all those 3090s and 4090s out there but in the grand scheme of things it isn't a bad card.

I agree with you. I could obviously pore more $$ on higher model of GPU and CPU but wanted to see if i could build a decent PC under $2K. I think the results are pretty good and I'm pleased.

Just tested my Samsung EVO Pro 980 2TB with 1G and 5G. It looks like very similar results.
The raw test on the RTX3060 actually surprised me given the card is considered low-end.

I get 68 fps with my RX 6700, otherwise they are comparable in other benchmarks. Beside CUDA, the difference may also be partly due to mine running @PCI 3.0 x16 and not 4.0 because of the motherboard (Asus withdrew Ryzen 3xxx 4.0 support). The thinking behind this is: I observed 30% lower score with PCI x8, thus I expect some difference between 4.0 and 3.0 x16.

I was actually contemplating between the RX-6700 and the RTX-3060 when designed the PC. Ended up with the 3060 because of the CUDA advantage.
I have not seen big difference between PCIe 4.0 and 3.0, maybe 3%-5%.
The Samsung 980 Pro is rated for very high r/w rates (7k/5k) so I would assume the difference is coming mostly from the new NAND chips.
By the way, in comparison, the raid5 array, in which I used the EVO 960, yielded only 650Mbps R/W, although it is connected to 6Gbps SATA bus on the motherboard. It's still fast enough for saving large files but not as fast as the 980 Pro NVME

They're not low end GPUs, the perception of low is because Nvidia only has one RTX, and really bare, GPU beneath it.

I meant low end of the high-end pack
I'm pretty pleased with the performance of the GPU

I will try some tomorrow as I normally do not use any full time just use smart cache to check credit rolls etc then turn off. Will have to set up a test. The whole point for me is to run native files in the timeline. Most of my project are around 2 hours so with 3 cameras the last thing I want to do is spend several hours making proxies and then use drive space unnecessary .

Same here.

I never understood all the people wanting to take compressed h.264/265 and transcode or generate proxies. Unless you drastically cut resolution and quality you still need the same drive space and throughput, and you lose time to processing no matter what.

imo, it's faster, easier and cheaper to buy a cpu and gpu that can handle the encodings than it is to try and jump through storage related hoops.

take a look at the DNxHR specs, specifically bit rate in the first table, and bit depth in the second.
https://avid.secure.force.com/pkb/artic ... ifications.

I never understood all the people wanting to take compressed h.264/265 and transcode or generate proxies. Unless you drastically cut resolution and quality you still need the same drive space and throughput, and you lose time to processing no matter what.

imo, it's faster, easier and cheaper to buy a cpu and gpu that can handle the encodings than it is to try and jump through storage related hoops.

take a look at the DNxHR specs, specifically bit rate in the first table, and bit depth in the second.
https://avid.secure.force.com/pkb/artic ... ifications.

I thought "quarter" or "half" size proxy means quarter/half of the resolution of the source?

The thinking behind this is: I observed 30% lower score with PCI x8, thus I expect some difference between 4.0 and 3.0 x16.

I've not compared results between PCIe 4.0 card in 4.0 vs 3.0 slots myself but I've seen number of tests by others (on YouTube and in enthusiast forums) that indicate difference isn't worth sweating over (IIRC less than 10%).