Mon May 28, 2018 9:08 am
Having noticed some playback slow-down on the Color page with 15 Beta 3 again (details in my last posts of these two threads: viewtopic.php?f=21&t=71032&p=413506#p413506 and viewtopic.php?f=32&t=74406) - I'd started to doubt my own understanding of how exactly various caching modes work. This is a very frustrating feeling - having worked with Resolve for almost 2 years now, and suddenly getting full of doubts whether what I have come to understand as a logical and relatively simple mechanism in one of the rudimentary Resolve functionalities! So please, both fellow users AND BMD moderators in this forum - please help me re-gain confidence again!
Let me start with the example I used in the former of the two links provided above, i.e. a scenario like this:
- a clip (XAVC-I UHD@50p) without anything done to it yet plays back with full, solid green dot, 50 fps speed (from the source drive, which in my case is an internal RAID 0 array of 4 WD Blacks, 3TB each with a solid 700 MBps sustained read transfer)
- if I force Resolve to cache this same clip (into DNxHR 444 HDR format - and since Resolve "feels" it doesn't need caching, achieved by ticking the "Cache clip color output" and caching mode to User), it can only play back with just some 40 fps !!! (of course, now it is playing back from my cache drive rather than the source matrix)
- I'm stressing it again that it can not be my cache drive's fault, as this is the fastest possible, NVMe SSD disk (plus I'm still perfectly able of playing my cached media back from it at the perfect and solid 50 fps with Resolve 12.5.3 and 15 Beta 1)
All the above would indicate the culprit to be lack of code optimization in some Resolve versions. BUT, I continue playing with this simple clip for testing purposes, and here is what I'm observing (this is where I'm totally losing confidence in my true understanding of the mechanism behind the caching functionality in Resolve
):
- I create just a single node, and implement some quite heavy, 4 frames TNR plus Spatial Noise Reduction in the "Better" mode; of course, I can now use Smart Caching (with the "Cache clip color output" option cleared)
- low and behold: when I play the now cached clip back from my NVMe disk, it plays back with the full and stable 50 fps again!
I'm really embarrassed not being able to understand the difference; could anyone help me please? TIA
Piotr
PS. All those differences in playback speed of media cached with different modes has nothing to do with the Release 14.3 bug which prevented full speed even of even just 25p (not to mention 50p) UHD media in Color page, whenever the Scopes were active - this bug has been confirmed by BMD, and successfully fixed in the first Beta 15. But having just noticed the achievable fps speed, depending on the caching mode used - I must (for now) call off my suspicious warnings of the bug re-appearing with the 15 Beta 3 release!
Let me start with the example I used in the former of the two links provided above, i.e. a scenario like this:
- a clip (XAVC-I UHD@50p) without anything done to it yet plays back with full, solid green dot, 50 fps speed (from the source drive, which in my case is an internal RAID 0 array of 4 WD Blacks, 3TB each with a solid 700 MBps sustained read transfer)
- if I force Resolve to cache this same clip (into DNxHR 444 HDR format - and since Resolve "feels" it doesn't need caching, achieved by ticking the "Cache clip color output" and caching mode to User), it can only play back with just some 40 fps !!! (of course, now it is playing back from my cache drive rather than the source matrix)
- I'm stressing it again that it can not be my cache drive's fault, as this is the fastest possible, NVMe SSD disk (plus I'm still perfectly able of playing my cached media back from it at the perfect and solid 50 fps with Resolve 12.5.3 and 15 Beta 1)
All the above would indicate the culprit to be lack of code optimization in some Resolve versions. BUT, I continue playing with this simple clip for testing purposes, and here is what I'm observing (this is where I'm totally losing confidence in my true understanding of the mechanism behind the caching functionality in Resolve
):- I create just a single node, and implement some quite heavy, 4 frames TNR plus Spatial Noise Reduction in the "Better" mode; of course, I can now use Smart Caching (with the "Cache clip color output" option cleared)
- low and behold: when I play the now cached clip back from my NVMe disk, it plays back with the full and stable 50 fps again!
I'm really embarrassed not being able to understand the difference; could anyone help me please? TIA
Piotr
PS. All those differences in playback speed of media cached with different modes has nothing to do with the Release 14.3 bug which prevented full speed even of even just 25p (not to mention 50p) UHD media in Color page, whenever the Scopes were active - this bug has been confirmed by BMD, and successfully fixed in the first Beta 15. But having just noticed the achievable fps speed, depending on the caching mode used - I must (for now) call off my suspicious warnings of the bug re-appearing with the 15 Beta 3 release!
AMD TR 2990WX CPU | MSI X399 CARBON AC | 64GB RAM@XMP3200 | 2x RTX 2080Ti GPU | 4x 3TB WD Black RAID0 media drive | 3x 1TB NVMe RAID0 cache drive | SSD SATA system drive | AX1600i PSU | Decklink 12G Extreme | Samsung UHD reference monitor (calibrated)