Question about streaming

[jewishdolphins]

So when I stream GTA V onto twitch via OBS my cpu usage goes to 100% load (I have a i5 4690k). I wanted to know if I could drastically decrease the usage if I start streaming with a capture card? How could I get around streaming games without having my cpu explode? Would streaming console games work? would using a capture card do anything? I need these answers.

[LAwLz]

Yes you can reduce the load by using a capture card. If you got a fairly modern system you might not need a capture card though.

Which CPU, GPU and motherboard do you have? Modern Intel CPUs (Sandy Bridge or newer with an integrated GPU as well as a motherboard with video out), as well as AMD (Radeon 7000 series or newer) and Nvidia GPUs (650 or new) all support encoding video in dedicated hardware which means it won't use any CPU cycles.

[OAcesync]

So when I stream GTA V onto twitch via OBS my cpu usage goes to 100% load (I have a i5 4690k). I wanted to know if I could drastically decrease the usage if I start streaming with a capture card? How could I get around streaming games without having my cpu explode? Would streaming console games work? would using a capture card do anything? I need these answers.

Streaming console games is fine on the CPU

To lower CPU usage for PC games a capture card would help

[jewishdolphins]

Yes you can reduce the load by using a capture card. If you got a fairly modern system you might not need a capture card though.

Which CPU, GPU and motherboard do you have? Modern Intel CPUs, as well as AMD and Nvidia GPUs all support encoding video in dedicated hardware which means it won't use any CPU cycles.

GPU : EVGA GeForce GTX 960 (Superclocked, 4GB) 

CPU : Intel Core i5 4690k Quad-Core Processor (3.5 Ghz) 

Mobo : Asus Z97-A Motherboard 

Ram : Corsair Vengeance 8 GB (2 x 4 GB) 

HDD : 1TB Western Digital Blue 7200 RPM 

SSD: 240 Gig Intel SSD 

Case : NZXT S340 Black 

OS : Windows 10 

Monitor : Asus VX228H 21.5" Monitor (x2) 

Keyboard : Corsair K65 RGB Mechanical Keyboard 

Mouse : Corsair M65 RGB Mouse 

 

Also, I have no idea what you mean in the last part. If you could explain more about this "dedicated hardware" that would be sweet.

[jewishdolphins]

Yes you can reduce the load by using a capture card. If you got a fairly modern system you might not need a capture card though.

Which CPU, GPU and motherboard do you have? Modern Intel CPUs (Sandy Bridge or newer with an integrated GPU as well as a motherboard with video out), as well as AMD (Radeon 7000 series or newer) and Nvidia GPUs (650 or new) all support encoding video in dedicated hardware which means it won't use any CPU cycles.

what dedicated hardware are u talking about? I know nothing of this so if u can explain more that would be awesome!

[LAwLz]

GPU : EVGA GeForce GTX 960 (Superclocked, 4GB) 

CPU : Intel Core i5 4690k Quad-Core Processor (3.5 Ghz) 

Mobo : Asus Z97-A Motherboard 

Ram : Corsair Vengeance 8 GB (2 x 4 GB) 

HDD : 1TB Western Digital Blue 7200 RPM 

SSD: 240 Gig Intel SSD 

Case : NZXT S340 Black 

OS : Windows 10 

Monitor : Asus VX228H 21.5" Monitor (x2) 

Keyboard : Corsair K65 RGB Mechanical Keyboard 

Mouse : Corsair M65 RGB Mouse 

 

Also, I have no idea what you mean in the last part. If you could explain more about this "dedicated hardware" that would be sweet.

Okay good, you won't need a capture card.

 

1) Open OBS

2) Go into "Settings..."

3) Go to the "Encoding"

4) Where it says "Encoder: x264" CHANGE that to "Nvidia NVENC"

 

I don't know if there is anything else you should do since I don't have an Nvidia card, but hopefully that will work. You should see your CPU usage go down dramatically if it works.

 

 

 

 

what dedicated hardware are u talking about? I know nothing of this so if u can explain more that would be awesome!

What I mean with "dedicated hardware" is that the parts I listed (Intel CPU, Nvidia and Nvidia GPUs) got specialized parts inside of them which can only do video encoding/decoding.

 

It might be a bit hard to grasp without showing pictures, but you can imagine it as your graphics card having another processor inside it. That processor can only handle videos. Right now it is just sitting there doing nothing, but if you configure your software correctly you can let that processor do the video encoding.

Since it is specifically designed to do video encoding, it is really fast and efficient at doing just that.

[jewishdolphins]

Okay good, you won't need a capture card.

 

1) Open OBS

2) Go into "Settings..."

3) Go to the "Encoding"

4) Where it says "Encoder: x264" CHANGE that to "Nvidia NVENC"

 

I don't know if there is anything else you should do since I don't have an Nvidia card, but hopefully that will work. You should see your CPU usage go down dramatically if it works.

If this works I will love you..I'll update you in approximately 10-15 mins-

[Samfisher]

 

 

I don't know if there is anything else you should do since I don't have an Nvidia card, but hopefully that will work. You should see your CPU usage go down dramatically if it works.

 

If this works I will love you..I'll update you in approximately 10-15 mins-

 

And also see a drastic drop in quality.

[jewishdolphins]

 

 

 

And also see a drastic drop in quality.

 

so is a capture card the best way to go for me?

[LAwLz]

And also see a drastic drop in quality.

Depends on the rest of the settings. The standard settings for x264 in OBS are pretty terrible.

I don't have an Nvidia GPU so I can't test the quality, but Intel QuickSync can get fairly close to x264 in terms of quality (and runs circles around it in terms of speed).

[Samfisher]

Depends on the rest of the settings. The standard settings for x264 in OBS are pretty terrible.

I don't have an Nvidia GPU so I can't test the quality, but Intel QuickSync can get fairly close to x264 in terms of quality (and runs circles around it in terms of speed).

NVENC and QuickSync will never come close to X264 quality at the same bitrates.

[LAwLz]

NVENC and QuickSync will never come close to X264 quality at the same bitrates.

1) They most certainly can.

2) There are far more things related to the quality than just bit rate. x264 at veryfast for example (the default for obs) looks really bad.

3) It doesn't really matter what quality he might have theoretically been able to get with x264 if his computer can't handle it.

[Samfisher]

1) They most certainly can.

2) There are far more things related to the quality than just bit rate. x264 at veryfast for example (the default for obs) looks really bad.

3) It doesn't really matter what quality he might have theoretically been able to get with x264 if his computer can't handle it.

They actually cannot.  I have done extensive testing at all kinds of bitrates at the same x264 preset.  Veryfast x264 at 3500kbps is more than fine for most things.  Fast and medium destroy your CPU more than any other settings.

[LAwLz]

They actually cannot.  I have done extensive testing at all kinds of bitrates at the same x264 preset.  Veryfast x264 at 3500kbps is more than fine for most things.  Fast and medium destroy your CPU more than any other settings.

They only explanation I got for those results is that you are not that experienced when it comes to encoding and are using the wrong settings.

You also keep ignoring that his computer can't handle doing it on the CPU.

[LAwLz]

Here you go. Since I don't have an Nvidia GPU I used QuickSync instead. This is a comparison of QuickSync vs x264 veryfast at 3500 Kbps.

Screenshot Comparison 1

Screenshot Comparison 2

Screenshot Comparison 3

 

The files where the screenshots were taken from can be found here (QuickSync) and here (x264).

 

 

How I tested it:

1) Record a little video with FRAPS. I used that because the video is completely uncompressed and I can do a fair comparison by comparing the exact same frames.

2) Play the video in MPC-HC and record the screen with QuickSync and x264, using the default settings for both (except setting it to CBR and set the bit rate to 3500Kbps). This is why the videos are slightly different length and start/stop at different times. I also covered my username.

3) Find the exact same frame in both videos and take a lossless screenshot. Upload them to Screenshot Comparison.

 

 

I will give x264 the edge in terms of quality, but I would say QuickSync is close. For archival purposes I would go with x264 (and placebo instead of veryfast) any day, but we are talking about live streaming here... The huge performance boost is well worth the slight quality degradation.

[Samfisher]

Here you go. Since I don't have an Nvidia GPU I used QuickSync instead. This is a comparison of QuickSync vs x264 veryfast at 3500 Kbps.

Screenshot Comparison 1

Screenshot Comparison 2

Screenshot Comparison 3

 

The files where the screenshots were taken from can be found here (QuickSync) and here (x264).

 

 

How I tested it:

1) Record a little video with FRAPS. I used that because the video is completely uncompressed and I can do a fair comparison by comparing the exact same frames.

2) Play the video in MPC-HC and record the screen with QuickSync and x264, using the default settings for both (except setting it to CBR and set the bit rate to 3500Kbps). This is why the videos are slightly different length and start/stop at different times. I also covered my username.

3) Find the exact same frame in both videos and take a lossless screenshot. Upload them to Screenshot Comparison.

 

 

I will give x264 the edge in terms of quality, but I would say QuickSync is close. For archival purposes I would go with x264 (and placebo instead of veryfast) any day, but we are talking about live streaming here... The huge performance boost is well worth the slight quality degradation.

You're comparing with hearthstone...where the only elements that ever move is the 10 cards in your hand and the card animations... The vast majority of your screen is static and hence need way less bitrate to look "ok".

[LAwLz]

You're comparing with hearthstone...where the only elements that ever move is the 10 cards in your hand and the card animations... The vast majority of your screen is static and hence need way less bitrate to look "ok".

What does bitrate have to do with anything? I used the same bit rate for both x264 and QuickSync, so the whole "you need less bit rate to make it look ok" is completely irrelevant here.

 

But okay, then let me bring up these examples from Anandtech's review then.

x86 vs QuickSync 1

x86 vs QuickSync 2

x86 vs QuickSync 3

 

 

I'll be back with another comparison then... Just so that you can get proven wrong again.

 

---

 

Edit:

Here you go. When I used a video with far more movement then it turns out QuickSync is actually superior to veryfast x264. So that means there is even less reason of using x264 over QuickSync for streaming. You are just wasting resources and the end result is a lower quality video.

The test was done the same way as before, except now I used one of the logo intros from a BDMV instead of a FRAPS video.

You can find the QuickSync video here, and the x264 video here.

 

Here are the comparison screenshots:

First

Second

Third

 

I would appropriate it if you could stop confusing OP with misinformation. I am trying to help and you are making it far more difficult than it has to be.

[Samfisher]

What does bitrate have to do with anything? I used the same bit rate for both x264 and QuickSync, so the whole "you need less bit rate to make it look ok" is completely irrelevant here.

 

Bitrate for streaming matters a whole lot because you can pack more detail and fidelity into the finished output than you can at a lower bitrate.  Converting video is an entirely different thing, and if anything the Anand article concludes that QS is worse than x264.

 

And I also suppose every streamer who has ever streamed is wrong as well, when limited to the 3500kbps of Twitch.

 

https://www.reddit.com/r/Twitch/comments/22isi8/obs_x264_vs_intel_quick_sync/

 

http://www.teamfortress.tv/17152/quick-sync-on-your-intel-integrated-gpu-in-obs

 

And this is my own test from a few years ago :

 

 

QS at 3500kbps

 

 

 

x264 at 3500kbps

 

[LAwLz]

Bitrate for streaming matters a whole lot because you can pack more detail and fidelity into the finished output than you can at a lower bitrate.

What the hell does that have to do with our conversation? I know what bitrate is, you don't have to explain it. What I don't understand is why you are talking about it at all since all the files I have uploaded have the exact same bit rate. Do you understand that? I am using CBR QuickSync and CBR x264 with the exact same target bitrate. It's like you are throwing around "technical terms" just to sound smart but they are completely irrelevant to the conversation. Are you going to bring up color depth next despite both of them having the exact same as well? Color depth is important for the image quality and compression rate as well.

 

 

Converting video is an entirely different thing, and if anything the Anand article concludes that QS is worse than x264.

Are we reading the same article? The conclusion was "Quick Sync: The best way to transcode", and that was 4 years ago. Quick Sync has improved a lot since then too. The conclusion was that the quality was almost as good as x264 for dramatically higher performance. That's widely different from your "nowhere near x264 quality" you said earlier.

 

 

And I also suppose every streamer who has ever streamed is wrong as well, when limited to the 3500kbps of Twitch.

What does that even mean? Again, why do you keep talking about bitrate? Where did I say something that goes against "every streamer who has ever streamed"?

I don't understand why you bring up some 3500Kbps limit on Twitch either because my files are 3000Kbps (video) plus about 100Kbps audio. That's fairly close to but not above that 3500Kbps limit you are speaking of.

So again, more irrelevant stuff that has nothing to do with the discussion.

 

 

 

https://www.reddit.com/r/Twitch/comments/22isi8/obs_x264_vs_intel_quick_sync/

 

http://www.teamfortress.tv/17152/quick-sync-on-your-intel-integrated-gpu-in-obs

You can't seriously think that 3 random posters on reddit who are just grabbing numbers out of thin air got any credibility whatsoever.

If I logged in there and said "x264 is crap" would you take that as a fact as well? Or do you only trust evidence that supports your predefined conclusion?

And before you say anything, yes the average forum poster is in fact a clueless idiot who does not understand what they are talking about. They just parrot what they have heard other (possibly also idiots) have said.

 

 

 

And this is my own test from a few years ago :

QS at 3500kbps

<video>

 

 

x264 at 3500kbps

<video>

Wow great test bro. I especially like how you did not use the same video which means we have no way of comparing it on a frame by frame basis, and that you also uploaded it to a website which adds additional compression which means even the same video uploaded twice will look slightly different.

That makes your test, which was also using now outdated software, very relevant and compelling.

/sarcasm

 

 

There were very good reasons why I did my tests exactly the way I did, with lossless videos as the source and PNG images. I did that because it is the only way you can get an accurate comparison. Your way of doing it is completely meaningless. It's like testing which car is the fastest, but you drive one car on the top gear test track and the other one on the Nurburgring, and not using this year's model of the cars either...

[Samfisher]

-snip

The conversion is about STREAMING TO OBS like the OP said, NOT converting video which takes care of your first 2 points.

 

3 random posters sure, I could find way more if you want, just type x264 vs Quick Sync stream into Google yourself.

 

Not the same video sure, but in every instance the x264 video looks better while the QS version is bitrate starved and keeps almost no detail when there's anything happening on screen.  And of course I uploaded it to YouTube, what sane person uploads to another platform that has no gaming community?  Twitch VOD? Hah!  And it's as if I never compared both source videos before uploading them to YouTube, plus I didn't know about the -multi switch to OBS at the time, was like 20 minutes into testing if anything.

 

Both videos were uploaded for a friend of mine overseas to show him the difference.  When I started streaming, I tried everything, and had a live audience commenting on quality.  QuickSync loses 100% of the time regardless of settings.

[LAwLz]

The conversion is about STREAMING TO OBS like the OP said, NOT converting video which takes care of your first 2 points.

This comment just shows you ignorant you are about this... I would appropriate it if you could just stop posting because you are spreading misinformation and confusing people.

Maybe you don't understand why I did the test the way I did, but I can ensure you that it is a perfectly valid test, and the only way you can get an apples to apples comparison.

 

 

3 random posters sure, I could find way more if you want, just type x264 vs Quick Sync stream into Google yourself.

Reminds me of this:

 

It doesn't matter if you find 3 or 30 random forum posts which are just making incorrect claims without evidence to back them up with.

 

 

Not the same video sure, but in every instance the x264 video looks better while the QS version is bitrate starved and keeps almost no detail when there's anything happening on screen.  And of course I uploaded it to YouTube, what sane person uploads to another platform that has no gaming community?  Twitch VOD? Hah!  And it's as if I never compared both source videos before uploading them to YouTube, plus I didn't know about the -multi switch to OBS at the time, was like 20 minutes into testing if anything.

Did you not read my post or are you willfully ignorant? Do you really don't understand why uploading it to YouTube invalidates your test, among with all the other reasons why your test is invalid? Do you not understand why I used PNG screenshots instead of for example JPEG, or uploaded the original files to Dropbox? If you don't even understand such a simple thing then I don't know what to say... It's hard to have a discussion with someone who does not understand the fundamentals of the subject.

 

 

You haven't commented on QuickSync looking far better in my second test, which had lots of colors, gradient and movements. Did you miss that or did you just ignore it because it does not agree with your predefined conclusion?

[Samfisher]

This comment just shows you ignorant you are about this... I would appropriate it if you could just stop posting because you are spreading misinformation and confusing people.

Maybe you don't understand why I did the test the way I did, but I can ensure you that it is a perfectly valid test, and the only way you can get an apples to apples comparison.

 

 

Reminds me of this:

You.png

 

It doesn't matter if you find 3 or 30 random forum posts which are just making incorrect claims without evidence to back them up with.

You mean like this post from an OBS moderator here, and an Admin a few posts down : I'm pretty sure NONE of these posters ever tested on QS and just like talking out their ass, as you seem to imply I do.

 

https://obsproject.com/forum/threads/intel-quick-sync-is-beautiful-and-the-future-of-streaming.7597/

 

And this poster who obviously tested it on his own as he switched between the two :

 

https://obsproject.com/forum/threads/obs-quicksync-settings.37111/

 

And here :

 

https://obsproject.com/forum/threads/question-stream-with-quicksync-and-or-nvenc-in-youtube.36344/

 

The same admin here :

 

https://obsproject.com/forum/threads/how-to-use-quicksync.16664/

 

And JIM the developer of OBS here :

 

https://obsproject.com/forum/threads/which-encoder-for-best-quality.19986/

 

And another moderator here :

 

https://obsproject.com/forum/threads/quicksync-bad-quality.13527/

 

And more posters here :

 

https://obsproject.com/forum/threads/gtx-970-nvenc-quick-sync-bitrates-x264-and-more-probably.35251/

 

And official guide on the OBS forums here :

 

https://obsproject.com/forum/threads/high-cpu-usage-high-encoding-taking-too-long-to-encode-read-this-first.23334/

 

Did you not read my post or are you willfully ignorant? Do you really don't understand why uploading it to YouTube invalidates your test, among with all the other reasons why your test is invalid? Do you not understand why I used PNG screenshots instead of for example JPEG, or uploaded the original files to Dropbox? If you don't even understand such a simple thing then I don't know what to say... It's hard to have a discussion with someone who does not understand the fundamentals of the subject.

 

 

You haven't commented on QuickSync looking far better in my second test, which had lots of colors, gradient and movements. Did you miss that or did you just ignore it because it does not agree with your predefined conclusion?

I DID read your post, but uploading a poorer quality to Youtube will also yield a poor quality Youtube video whereas a better quality upload will also show an equally better video.  And like I said I have done my own viewing of the source files PRIOR to uploading to YT.  It seems YOU didn't read my post.  It's the second time I'm posting this.

 

I did not miss your 2nd test but you know what? To hell with your test.  Your test is not indicative of the quality available to streamers on any platform.  The whole point of this thread's existence is OBS/streaming performance.  There is NO data that shows QS is as good as X264 at the same streaming bitrate.

[.spider.]

A capture card with built in hardware encoder would give you much worse quality.

All these hardware encoder are abusing h264 as an intermediate codec.

At lower bitrates there's no way around x264.

Thus you would need a second PC with capture card.

[LAwLz]

You mean like this post from an OBS moderator here, and an Admin a few posts down : I'm pretty sure NONE of these posters ever tested on QS and just like talking out their ass, as you seem to imply I do.

<links>

Oh good, more legitimate links. Let's see what they are saying.

"You can offload encoding load to those hardware encoders at the cost of a somewhat noticeable decrease in quality at the same bit rate."

 

Would you look at that. The official OBS guide agrees with me and not you.

Their claim: "somewhat noticeable decrease in quality"

My claim: "QuickSync can get fairly close to x264 in terms of quality"

Your claim: "QuickSync will never come close to x264 quality"

 

 

And hey would you look at that, a lot of the posts are from when QS was just implemented and in the testing stage. But let's ignore the tests we already have in the thread, using the most up to date software. Let's just go with what old posts using now outdated versions of the software says, back when QS support was brand new in OBS. That sounds like a much better idea, right?

 

 

I DID read your post, but uploading a poorer quality to Youtube will also yield a poor quality Youtube video whereas a better quality upload will also show an equally better video.  And like I said I have done my own viewing of the source files PRIOR to uploading to YT.  It seems YOU didn't read my post.  It's the second time I'm posting this.

No, because that's not how it works. You could upload the exact same file twice to YouTube and it can be slightly different. It is also incredibly hard to determine the quality from your video because it is not the same clip. There is no way to compare them on a frame by frame basis.

You might as well test two headphones using different songs. It's a pointless test.

 

 

I did not miss your 2nd test but you know what? To hell with your test.  Your test is not indicative of the quality available to streamers on any platform.  The whole point of this thread's existence is OBS/streaming performance.  There is NO data that shows QS is as good as X264 at the same streaming bitrate.

How is my test invalid? It is as scientific as you can get. It shows OBS streaming performance perfectly.

I have not said that QS is as good in terms of quality as x264. What I have been saying all this time is that QS can get close to x264 veryfast quality, at the same bitrate, but with much higher performance. You said it can't and I think I have proven you wrote over and over again.

 

You are lacking knowledge about video encoding as well as how to do proper tests. Please stop spreading misinformation.

[Samfisher]

Oh good, more legitimate links. Let's see what they are saying.

"You can offload encoding load to those hardware encoders at the cost of a somewhat noticeable decrease in quality at the same bit rate."

 

Would you look at that. The official OBS guide agrees with me and not you.

Their claim: "QuickSync trades quality for performance. 2800 bitrate isn't even enough for a game like STALKER even with x264, a superior quality software encoder. The only thing that makes QuickSync better is more bitrate.(2014)

 

Their claim : Quality wise x264 is going to win outright, I can tell you that with 100% confidence. If you can manage to handle the Fast x264 preset, even better. (2014)

 

Their claim : Basically everything in this thread, late 2015

 

And the OBS guide, early 2015 : https://obsproject.com/forum/threads/high-cpu-usage-high-encoding-taking-too-long-to-encode-read-this-first.23334/