Intel Quicksync or Nvidia NVEnc for me in Twitch?

[manikyath]

1 minute ago, Enderman said:

For streaming to twitch the video bitrate should be set exactly at 3.5MBps and nothing higher.

commonsense++ do not piss off the platform you're using

 

while not a hardcoded cap (its actually too hard to bother to code this in) twitch prefers you dont go over 3.5Mbps, and there's actually reasons beyond that not to either way.

 

anyways, on topic: i made a thread not too long ago comparing the different ways, it's probably something worth reading.

i'm still gonna do tests with 1080p30 and 900p60 at 3500kbps with the vareous setting picks as well, but i cant promise when

[zMeul]

@NinjaJc01 ok, I'm back, here are the results

used CQP 23 for the recording at 1080p with both QS and NVENC

game used RotTR, SMAA and rest quality settings at default 

 

1st up QS:

Mountain Peak - avg 60 / min 34.79 / max  90.49

Syria - avg 59.96 / min 37.77 / max 73.33

Geothermal Valley - avg 59.83 / min 33.95 / max 79.49

overall - avg 59.93 FPS

 

NVENC:

Mountain Peak - avg 60 / min 38.94 / max 85.20

Syria - avg 59.96  / min 31.39 / max 85.05

Geothermal Valley - avg 60 / min 50.97 / max 71.93

overall - avg 59.99 FPS

 

if there's something I missed and want to know, ask

[NinjaJc01]

1 minute ago, manikyath said:

commonsense++ do not piss off the platform you're using

 

while not a hardcoded cap (its actually too hard to bother to code this in) twitch prefers you dont go over 3.5Mbps, and there's actually reasons beyond that not to either way.

 

anyways, on topic: i made a thread not too long ago comparing the different ways, it's probably something worth reading.

i'm still gonna do tests with 1080p30 and 900p60 at 3500kbps with the vareous setting picks as well, but i cant promise when

Annoyingly, I care more about my FPS than quality... On a separate note, I just noticed I passed 2k posts. 

[manikyath]

Just now, NinjaJc01 said:

Annoyingly, I care more about my FPS than quality... On a separate note, I just noticed I passed 2k posts. 

well... quality is what draws in the viewers, mate

[NinjaJc01]

1 minute ago, zMeul said:

ok, I'm back, here are the results

used CQP 23 for the recording at 1080p with both QS and NVENC

game used RotTR, SMAA and rest quality settings at default 

 

1st up QS:

Mountain Peak - avg 60 / min 34.79 / max  90.49

Syria - avg 59.96 / min 37.77 / max 73.33

Geothermal Valley - avg 59.83 / min 33.95 / max 79.49

overall - avg 59.93 FPS

 

NVENC:

Mountain Peak - avg 60 / min 38.94 / max 85.20

Syria - avg 59.96  / min 31.39 / max 85.05

Geothermal Valley - avg 60 / min 50.97 / max 71.93

overall - avg 59.99 FPS

 

if there's something I missed and want to know, ask

So there's basically nothing between them?

[NinjaJc01]

Just now, manikyath said:

well... quality is what draws in the viewers, mate

When I get a new CPU or CPU, then I'll go for quality. For some reason, my rig tends to under perform a little.

[manikyath]

Just now, NinjaJc01 said:

When I get a new CPU or CPU, then I'll go for quality. For some reason, my rig tends to under perform a little.

with the right tweaking a 4770 should get you pretty much all you need.

also, a really old quote from slowpoke101: "its better to have a good 480p than a crappy 720p"

 

while streaming in 480p is a bit dated by now, the same story applies between 720p and higher resolutions.

you're better off decreasing the encoding task, than try to force it anyways and end up with a mess.

 

i should refer you specificly to the nvenc and quicksync tests in my thread.

some notes ahead of time:

- ultrafast and veryfast cpu encoding both have neglible cpu usage

- fast is quite high on the cpu usage, but for most games still within limits

- slow is basicly only for dual-pc stream setups or games that dont touch the cpu

- quicksync is a potato, but a very efficient potato (what is that sour cream even doing here...)

- nvenc is a potato as well, but at least this one has firecrackers strapped to it to at leats get it over the big hills.

[zMeul]

9 minutes ago, NinjaJc01 said:

So there's basically nothing between them?

it depends on your setup, yours might differ

I have a GTX970 G1 and a i5 6500

 

what can I say ... test it out and see what floats your boat

 

to note: there were some areas where I saw a ~20% increase in total CPU usage

[NinjaJc01]

Just now, manikyath said:

with the right tweaking a 4770 should get you pretty much all you need.

also, a really old quote from slowpoke101: "its better to have a good 480p than a crappy 720p"

 

while streaming in 480p is a bit dated by now, the same story applies between 720p and higher resolutions.

you're better off decreasing the encoding task, than try to force it anyways and end up with a mess.

 

i should refer you specificly to the nvenc and quicksync tests in my thread.

some notes ahead of time:

- ultrafast and veryfast cpu encoding both have neglible cpu usage

- fast is quite high on the cpu usage, but for most games still within limits

- slow is basicly only for dual-pc stream setups or games that dont touch the cpu

- quicksync is a potato, but a very efficient potato (what is that sour cream even doing here...)

- nvenc is a potato as well, but at least this one has firecrackers strapped to it to at leats get it over the big hills.

So what should I use for performance in some CPU intensive games, like BeamNG drive? And when I said my PC under performs, that's only in games, even when I'm not streaming. How would I go about setting up a dual PC encoding thing? I have a lot of spare Core 2 Duo PCs not being used that might be ok.

[NinjaJc01]

3 minutes ago, zMeul said:

it depends on your setup, yours might differ

I have a GTX970 G1 and a i5 6500

 

what can I say ... test it out and see what floats your boat

 

to note: there were some areas where I saw a ~20% increase in total CPU usage

20% CPU usage from what? In both tests?

[Zodiark1593]

13 minutes ago, NinjaJc01 said:

Annoyingly, I care more about my FPS than quality... On a separate note, I just noticed I passed 2k posts. 

For video streaming, I'd argue that video quality matters as much as framerate (you really need . I doubt there are many people that want to watch video that's of a poor quality.

 

If your framerate is slow in games, you will need to look at reducing graphics quality settings.

 

CPU encode is also an option that yields good quality for the bandwidth at the expense of needing extra cpu power.

[manikyath]

Just now, NinjaJc01 said:

So what should I use for performance in some CPU intensive games, like BeamNG drive? And when I said my PC under performs, that's only in games, even when I'm not streaming. How would I go about setting up a dual PC encoding thing? I have a lot of spare Core 2 Duo PCs not being used that might be ok.

what the hell are you doing to your poor pc that makes it underperform in gaming...

[zMeul]

Just now, NinjaJc01 said:

20% CPU usage from what? In both tests?

no, in QS vs NVENC

[manikyath]

Just now, Zodiark1593 said:

For video streaming, I'd argue that quality matters as much as framerate (you really need . I doubt there are many people that want to watch video that's of a poor quality.

 

If your framerate is slow in games, you will need to look at reducing graphics quality settings.

its actually often argued that when streaming you want your game to run at a very, very exact mutliple of what you're streaming, because otherwise you're setting your viewers up for tearfest.

[NinjaJc01]

Just now, manikyath said:

what the hell are you doing to your poor pc that makes it underperform in gaming...

I honestly don't know, it's just not performing as well as my friend's PC when he had the 970 I have now. He had a 4690 (non k, at stock). It might be that he has 2400MHz RAM, and I tend to play games that like faster RAM.

 

2 minutes ago, zMeul said:

no, in QS vs NVENC

So which one was worse for CPU usage?

[NinjaJc01]

1 minute ago, manikyath said:

its actually often argued that when streaming you want your game to run at a very, very exact mutliple of what you're streaming, because otherwise you're setting your viewers up for tearfest.

I stream at 30, play at 60. I can't look at 30FPS video anymore without taking a little while to adjust, and anything under 25 now looks ugly and horrible.

[manikyath]

1 minute ago, NinjaJc01 said:

I stream at 30, play at 60. I can't look at 30FPS video anymore without taking a little while to adjust, and anything under 25 now looks ugly and horrible.

when was the last time you watched TV, or went to the cinema?

[zMeul]

3 minutes ago, NinjaJc01 said:

So which one was worse for CPU usage?

QuickSync

 

but, since you have an i7 (4 cores plus hyperthreading), your results will be totally different than mine

[NinjaJc01]

1 minute ago, manikyath said:

when was the last time you watched TV, or went to the cinema?

I went to the cinema today, but that took me a few mins to adjust, and it was at glorious 4K. It's mainly action sequences or camera pans that I feel suffer. I don't tend to watch TV, as my current TV box has a menu system that runs at like 15FPS.

[Zodiark1593]

8 minutes ago, manikyath said:

its actually often argued that when streaming you want your game to run at a very, very exact mutliple of what you're streaming, because otherwise you're setting your viewers up for tearfest.

My sentiments exactly. I'm not downplaying the effects of a poorly synced framerate by any means. My point being, if one cannot achieve both the desired framerate and quality, the streaming setup should be reconsidered.

 

If a steady 60 fps can be achieved regardless of whether NVENC or Quicksync (or even x264) is used, obviously whichever yields the best quality should be chosen. If 60 fps cannot be achieved without heavy quality sacrifice, or vice versa, lower the target framerate (30 fps), visual settings, or consider an upgrade.

[NinjaJc01]

2 minutes ago, Zodiark1593 said:

My sentiments exactly. I'm not downplaying the effects of a poorly synced framerate by any means. My point being, if one cannot achieve both the desired framerate and quality, the streaming setup should be reconsidered.

 

If a steady 60 fps can be achieved regardless of whether NVENC or Quicksync (or even x264) is used, obviously whichever yields the best quality should be chosen. If 60 fps cannot be achieved without heavy quality sacrifice, or vice versa, lower the target framerate (30 fps), visual settings, or consider an upgrade.

I tend to get 60FPS in most games, but my stream framerate never seems to go above 30ish.

[manikyath]

1 minute ago, Zodiark1593 said:

My sentiments exactly. I'm not downplaying the effects of a poorly synced framerate by any means. My point being, if one cannot achieve both the desired framerate and quality, the streaming setup should be reconsidered.

 

If a steady 60 fps can be achieved regardless of whether NVENC or Quicksync (or even x264) is used, obviously whichever yields the best quality should be chosen. If 60 fps cannot be achieved without heavy quality sacrifice, or vice versa, lower the target framerate (30 fps), visual settings, or consider an upgrade.

to be honest, quicksync is pretty potato for doing pretty much anything, the algorithm impressed me, but the igpu just does NOT have the performance to back it up.

 

nvenc uses a very similar algorithm, but is still rather disappointing at times, at least nvidia cards supporting nvenc also have some "brute force" behind them for the bad cases.

 

as for x264.. its just brute force all the way

[NinjaJc01]

Just now, manikyath said:

to be honest, quicksync is pretty potato for doing pretty much anything, the algorithm impressed me, but the igpu just does NOT have the performance to back it up.

 

nvenc uses a very similar algorithm, but is still rather disappointing at times, at least nvidia cards supporting nvenc also have some "brute force" behind them for the bad cases.

 

as for x264.. its just brute force all the way

So for the sake of my stream speed/quality and my framerate, nvenc would be better? Since I am playing CPU intensive games?

[manikyath]

Just now, NinjaJc01 said:

So for the sake of my stream speed/quality and my framerate, nvenc would be better? Since I am playing CPU intensive games?

it depends, all 3 have very specific weaknesses:

- quicksync dunks out completely when there's lots of moving objects, if most of the screen is stationary or slow moving it's pretty amazing, but in the tougher scenarios its a joke.

- nvenc is closer to the x264 cpu brute force way, but has some form of intelligent algorithm on where to focus its power, sadly the performance is still quite often lacking, and the algorithm isnt as good as quick sync

- x264 is simply more cpu power = better stream. although i'd dare to argue that in the majority of cases it's far superior of both other options when you have an i7 at your disposal.

[NinjaJc01]

1 minute ago, manikyath said:

it depends, all 3 have very specific weaknesses:

- quicksync dunks out completely when there's lots of moving objects, if most of the screen is stationary or slow moving it's pretty amazing, but in the tougher scenarios its a joke.

- nvenc is closer to the x264 cpu brute force way, but has some form of intelligent algorithm on where to focus its power, sadly the performance is still quite often lacking, and the algorithm isnt as good as quick sync

- x264 is simply more cpu power = better stream. although i'd dare to argue that in the majority of cases it's far superior of both other options when you have an i7 at your disposal.

I'll try NVEnc and x264 when I next have access to my gaming PC, since I currently use Quicksync.