KuraiShidosha
New Member
Can confirm this finding. It increases by 300MB at a time it seems.Start and stop many times will consume a lot of VRAM.
This only happens with the new nvenc.
https://youtu.be/Qu9xkL6MGIQ
Can confirm this finding. It increases by 300MB at a time it seems.Start and stop many times will consume a lot of VRAM.
This only happens with the new nvenc.
https://youtu.be/Qu9xkL6MGIQ
Try as an experiment to lower the frame rate to 30 and set two-pass encoding....When I use the nvenc new my stream gets blurrier than when I was using x264 on my i7 7700k on veryfast same bit rate 8,000.
...
Just use shadowplah for local recordings and obs for streamingI came here to check up on whether it was worth downloading this, but for reference I haven't actually tried the beta yet. However I saw this and wanted to chime in.
I currently stream at a different resolution to what I record, and I use NVENC for both recording and streaming. It was one of the main reasons I went to NVENC so I could have the higher quality video to use for editing for YouTube, while keeping things looking nicer at lower bitrates on Twitch streams. Not having that ability might be a bit of a pain.
Just wanted to give some more info on this specifically. I'm looking forward to seeing some performance improvements with the new version of this, but it's no longer a slam dunk for me to definitely try it straight away after seeing this. Cheers.
For those of you who want to be able to rescale the output for streaming while recording with a higher resolution: Are you all using NVENC for both recording and streaming, or are you using different encoders for each?
Fine! But I would like to see the possibility of calibration and the availability of the following options:Hey guys, I made a build if anyone wants to test it out. It's rough but seems to work...
Can you update your x264 library to version 0.157.2935 (545de2f)?I added a few of the advanced options ... I can add other options as well but these stood out the most in the nvidia documentation
{ NV_ENC_PRESET_HQ_GUID, NVENC_TWO_PASSES }, /* slow */Slow, medium and fast are not defined presets in the NVENC API, they aren't listed in the nvEncodeApi.h header file.
@jellysandwich be sure to make a pull request when the jim-nvenc branch is merged into master
Therefore, they need to be added there.Those are not defined presets in the NVENC API, those lines do not appear in the header file.
Either way, it's up to Jim if he wants to have all kinds of advanced settings for NVENC in OBS. My guess would be definitely not visible by default, but behind some kind of checkbox that would make more settings appear when checked.
// =========================================================================================
// * Preset GUIDS supported by the NvEncodeAPI interface.
// =========================================================================================
// {B2DFB705-4EBD-4C49-9B5F-24A777D3E587}
static const GUID NV_ENC_PRESET_DEFAULT_GUID =
{ 0xb2dfb705, 0x4ebd, 0x4c49, { 0x9b, 0x5f, 0x24, 0xa7, 0x77, 0xd3, 0xe5, 0x87 } };
// {60E4C59F-E846-4484-A56D-CD45BE9FDDF6}
static const GUID NV_ENC_PRESET_HP_GUID =
{ 0x60e4c59f, 0xe846, 0x4484, { 0xa5, 0x6d, 0xcd, 0x45, 0xbe, 0x9f, 0xdd, 0xf6 } };
// {34DBA71D-A77B-4B8F-9C3E-B6D5DA24C012}
static const GUID NV_ENC_PRESET_HQ_GUID =
{ 0x34dba71d, 0xa77b, 0x4b8f, { 0x9c, 0x3e, 0xb6, 0xd5, 0xda, 0x24, 0xc0, 0x12 } };
// {82E3E450-BDBB-4e40-989C-82A90DF9EF32}
static const GUID NV_ENC_PRESET_BD_GUID =
{ 0x82e3e450, 0xbdbb, 0x4e40, { 0x98, 0x9c, 0x82, 0xa9, 0xd, 0xf9, 0xef, 0x32 } };
// {49DF21C5-6DFA-4feb-9787-6ACC9EFFB726}
static const GUID NV_ENC_PRESET_LOW_LATENCY_DEFAULT_GUID =
{ 0x49df21c5, 0x6dfa, 0x4feb, { 0x97, 0x87, 0x6a, 0xcc, 0x9e, 0xff, 0xb7, 0x26 } };
// {C5F733B9-EA97-4cf9-BEC2-BF78A74FD105}
static const GUID NV_ENC_PRESET_LOW_LATENCY_HQ_GUID =
{ 0xc5f733b9, 0xea97, 0x4cf9, { 0xbe, 0xc2, 0xbf, 0x78, 0xa7, 0x4f, 0xd1, 0x5 } };
// {67082A44-4BAD-48FA-98EA-93056D150A58}
static const GUID NV_ENC_PRESET_LOW_LATENCY_HP_GUID =
{ 0x67082a44, 0x4bad, 0x48fa, { 0x98, 0xea, 0x93, 0x5, 0x6d, 0x15, 0xa, 0x58 } };
// {D5BFB716-C604-44e7-9BB8-DEA5510FC3AC}
static const GUID NV_ENC_PRESET_LOSSLESS_DEFAULT_GUID =
{ 0xd5bfb716, 0xc604, 0x44e7, { 0x9b, 0xb8, 0xde, 0xa5, 0x51, 0xf, 0xc3, 0xac } };
// {149998E7-2364-411d-82EF-179888093409}
static const GUID NV_ENC_PRESET_LOSSLESS_HP_GUID =
{ 0x149998e7, 0x2364, 0x411d, { 0x82, 0xef, 0x17, 0x98, 0x88, 0x9, 0x34, 0x9 } };
And the preset Max Quality of which of these presets applies?That's not how it works, there is no slow, medium or fast preset defined in the API, changing the header file won't change that...
It turns out that Max Quality is the same as the Slow preset.It doesn't have a corresponding preset, it is the NVENC HQ preset and 2-pass enabled, but that is something done outside of the NVENC API.
Quality = NVENC HQ preset and 2-pass disabled
Max Performance = NVENC HP preset and 2-pass disabled
So going off of what you pasted, there is no need for a slow, medium or fast option, since those combination of settings is already there.
Here is the source of nvenc, which was once found on this forum.-added bframes as a reference (enabled automatically if bframes > 0)
-split out spatial aq and temporal aq
-added spatial aq strength
-added lookahead depth