How important are cores when gaming/streaming? My findings

craigdt

[H]ard|Gawd
Joined
Oct 27, 2016
Messages
1,075
All this talk about the 9900k/9700k being worth an upgrade if you like to stream your games.

"Sounds like BS, oft-repeated, never tested talking points to me!"

But its a lazy Sunday afternoon, and I had nothing better to do.

I'm sure none of this will new news to most of you.

I ran the canned Far Cry 5 benchmark with my 8700K, with and without streaming.
Also ran the same test with 2 cores disabled.
Since I can't magically add 2 cores... this was the best I could do.

Pretty amazing how much life that streaming sucks out of a pretty potent processor.

I feel bad for all the 4-core gamers :p

Edit- also- It looks like we've got a long ways to go before games start regularly making use of more than 4 cores.

yjetzYx.jpg
 
Last edited:
Do GTA V since it scales with cores really well.

(and has a canned bench)
 
Do GTA V since it scales with cores really well.

(and has a canned bench)

I'd love to.

If someone would send me a Steam key :D

Other than Far Cry 5 and the Final Fantasy demo, I'd don't have a whole lot thats particularly challenging for my PC to run.

Dota, Starcraft 2, Arma 3 (hmm maybe...), Destiny 2...


Edit-
I see that Civ VI has a built in benchmark. I'll install that tonight if anybody is interested.

Also try the World of Tanks Encore too.
 
Last edited:
Alright, here's Civilization VI

20% slower average frame time while streaming

al77qq5.jpg
 
Here's World of Tanks: Encore

Does seem to be nearly as CPU bound.

Probably optimized for crappy machines.

x0vWtLQ.jpg
 
The frame rate that the user receives is the more important metric.

Well that's true I suppose, depending on your goal.

However, there are a ton more factors that would affect that, right?

Having never had more than 1 viewer (my wife, bless her heart), I have no idea what the experience is like :p
 
What resolution is the streaming? I am sure that would have a factor in it as well for both CPU and GPU.
 
I would be curious to see if 720P streaming has more of a CPU impact than 1440P. Sounds silly as I know that 720p can/does leverage CPU more. If it stays about the same in terms of performance hit percentage that truly would be interesting!

Thank you,
 
Kind of curious on OBS settings, that does look like a pretty decent hit on performance. Would be interesting to see if someone could do something similar with a 9900k, if those extra cores actually do much.
 
Thanks for this. Though I never thought of streaming my horrible game play antics, this gives me even more reason to keep it inside and enjoy it for myself.
 
How important are cores when browsing internet? I'm not sure if I wanna upgrade my mom's PC from dual core to >quad core. Kindly advise.
 
The frame rate that the user receives is the more important metric.
Frame time (also rate) and framerate are the same thing. Exactly the same thing. Framerate in common useage refers to an average time of many frames, whereas frame time is exactly how long 1 frame took. But at the end of the day, its exactly the same metric.
 
Frame time (also rate) and framerate are the same thing. Exactly the same thing. Framerate in common useage refers to an average time of many frames, whereas frame time is exactly how long 1 frame took. But at the end of the day, its exactly the same metric.

I was talking about the frame rate of the streaming viewer. This can be a slideshow when created on 6/6 Intel cpus whereas the 6/12 AMDs are perfectly smooth despite having a lower framerates for the STREAMER.
 
I was talking about the frame rate of the streaming viewer. This can be a slideshow when created on 6/6 Intel cpus whereas the 6/12 AMDs are perfectly smooth despite having a lower framerates for the STREAMER.

This is a good point, frames can be lost due to encoding lag if the processor is being overworked. OBS will display a warning if this is the case, but you can also check OBS logs to see if any frames were lost due to this, rendering lag, or dropped due to bitrate/connection issues.

I do some casual streaming on a 5960X and the extra cores definitely help out. I read up on core counts vs encoding efficiency on the OBS forums during the threadripper release wondering what benefits that could have. The impression I got was the x264 encoder will benefit from having more cores up until a point and the rule I repeatedly heard was 1 thread per 40 lines of vertical resolution. I haven't had any cpu with a high enough core count to test that out myself, however if anyone cares to see some tests of a 5960X running on 8 cores vs 6 while streaming, I could mess around and run some tests tomorrow. Obviously the performance won't be the same as a current gen processor, but it should still demonstrate what advantage there may be going from 6 to 8 cores specifically.
 
This is a good point, frames can be lost due to encoding lag if the processor is being overworked. OBS will display a warning if this is the case, but you can also check OBS logs to see if any frames were lost due to this, rendering lag, or dropped due to bitrate/connection issues.

I do some casual streaming on a 5960X and the extra cores definitely help out. I read up on core counts vs encoding efficiency on the OBS forums during the threadripper release wondering what benefits that could have. The impression I got was the x264 encoder will benefit from having more cores up until a point and the rule I repeatedly heard was 1 thread per 40 lines of vertical resolution. I haven't had any cpu with a high enough core count to test that out myself, however if anyone cares to see some tests of a 5960X running on 8 cores vs 6 while streaming, I could mess around and run some tests tomorrow. Obviously the performance won't be the same as a current gen processor, but it should still demonstrate what advantage there may be going from 6 to 8 cores specifically.

Sounds interesting. I guess you could stream it to yourself on a 2nd screen to see how smooth it is for the viewer.

Also, it would be nice to see how 8 cores w/o ht compares to 6 cores with ht as that is the difference between the 8700k and 9700k.
 
How important are cores when browsing internet? I'm not sure if I wanna upgrade my mom's PC from dual core to >quad core. Kindly advise.
This is a great question. For web browsing and basic office apps like Word and Excel 2 cores is more than plenty. Where it really counts is how fast those cores are. Websites are becoming busier and busier with ad's that is usually what takes up most of the processing of a web page.
 
I completed some tests playing around with my 5960X. Decided to test it out with GTA V since it was requested and so I could beat craigdt to the punch! :p
Good call by Nightfire to suggest throwing 8 core w/o HT into the mix. It was certainly interesting.

gta v encoding tests.JPG


The top results show what performance we get in game while streaming. I chose a rather aggressive CPU preset to push things to their limit and see how each configuration fares in terms of frames lost during the encoding process. If we're actually streaming we'd ideally want 0 frames lost so I'd keep things on a faster preset, but we need to actually lose some frames to get results we can use for a comparison!

Next I decided it'd be interesting to see how OBS fares in encoding between these configurations when using a stable video source. Since the benchmark varies in performance between the configurations OBS technically has a lower workload in the 8 Core w/o HT and 6 Core configurations since they have a lower overall frame rate. OBS will try to encode at 60FPS regardless but it is also receiving and encoding duplicate frames when the ingame frame rate isn't as high.
I recorded the GTA V benchmark using Bandicam, which has a much lower CPU usage. This allowed me to have a video file of the benchmark recorded at 60FPS with the benchmark running consistently over that. Now each configuration actually has to process the same workload. We see the number of lost frames goes way up at our Medium preset with the increased number of unique frames to encode. Dropped frames goes down on the 8 Core HT config, likely because the in game performance was generally already over 60FPS and without the game running there were more CPU resources available.
I tried using a few different CPU presets to see how results vary. Using Slow we lose way too many frames even using 8 core HT to make for a good stream. However, notice how much of an improvement there is between 8 cores with HT versus 8 cores without, and how little the difference is between 8 cores without HT and 6 cores with HT.

I included MEGUI encoding results to get an idea of the pure encoding performance differences between configurations. Going from 6 to 8 cores results in performance improving by about a third, which is consistent with what you'd expect. What is interesting is how the benefits of hyperthreading are greater at the more aggressive encoding presets. I don't know enough about the intricacies of video encoding to explain why that may be, but if you want to use the more aggressive settings you really are going to want hyper threading available on your CPU.

So looking this over, at least from the results with this game, it seems like going from 6 cores with HT to 8 cores without you won't suffer a performance hit while streaming. It should be a bit of an improvement. However, you may be missing out on a great deal of the benefit you'd get from those additional cores with HT, especially on those more aggressive CPU presets you'd want to use on a high end CPU. The 8th & 9th gen series will obviously have higher clocks and better IPC than my 5960X, but I hope this info is of use to some out there. I may do some additional tests so if anyone has some criticisms or suggestions to improve my testing, please let me know!
 
Playing BF1 in a full server with a 4 core and 6 core is a night day difference, and that is without streaming.

I wish reviewers would test CPU's inside full MP games, and not some scripted single player benchmarks with nothing going on, it is meaningless for the CPU.
 
So looking this over, at least from the results with this game, it seems like going from 6 cores with HT to 8 cores without you won't suffer a performance hit while streaming.
Doesn't your data show that 8 wo/HT is significantly faster than 6 w/HT?
 
Alright, some rudimentary follow-up info regarding Grand Theft Auto 5

Basically all the settings cranked up, 1440p:

Run 1, no streaming:
Average framerate : 98.4 FPS
Minimum framerate : 39.9 FPS

Run 2, streaming via OBS:
Average framerate : 100.9 FPS
Minimum framerate : 39.3 FPS

Have no idea why there was a slight increase with streaming, and why my findings were different than tryin49 above.
I did verify I was streaming.

It does seem that its not a super-optimzed game for PC Master Race.
Just a thought. I'll do some more testing later.

Edit-
Shout out to cybereality
He hooked me up with the Steam key that made all this mind-blowing analysis possible
 
Last edited:
Doesn't your data show that 8 wo/HT is significantly faster than 6 w/HT?
Yes, that is what I wrote. Apologies if it was worded awkwardly.

Craigdt, I meant to ask you earlier what settings you're using in OBS? The performance difference in our benchmarks are likely down to that as the more intense CPU presets really start beating the hell out of the processor and I was using medium. I'm guessing you must be on a super fast preset or something like that since the framerate actually went up for you and those lighter presets will barely touch the CPU at all relatively speaking.
 
Y

Craigdt, I meant to ask you earlier what settings you're using in OBS? The performance difference in our benchmarks are likely down to that as the more intense CPU presets really start beating the hell out of the processor and I was using medium. I'm guessing you must be on a super fast preset or something like that since the framerate actually went up for you and those lighter presets will barely touch the CPU at all relatively speaking.

"veryfast"

I'll run some new tests at "slower" and see what happens

Edit:
Changed my OBS settings to "slower" which I guess means higher quality?
In this setting, the CPU usage was at 100% most of the time and I experienced huge FPS drops.
Basically made the game unplayable.

Average framerate : 75.3 FPS
Minimum framerate : 1.4 FPS
 
Last edited:
"veryfast"

I'll run some new tests at "slower" and see what happens

Edit:
Changed my OBS settings to "slower" which I guess means higher quality?
In this setting, the CPU usage was at 100% most of the time and I experienced huge FPS drops.
Basically made the game unplayable.

Average framerate : 75.3 FPS
Minimum framerate : 1.4 FPS

Yes, the slower the settings are better quality. Even medium setting us mostly academic, and slower is a bit ridiculous.
 
Yes, the slower the settings are better quality. Even medium setting us mostly academic, and slower is a bit ridiculous.

Not that I ever have any Twitch viewers, but it'd be interesting to see if there is a visible difference.

Don't half the Twitch streamers use their console anyways?
 
So, are they important or not? Does anyone have a Cole's Notes (summarized) version? ;-)

I'm kidding. It looks like there's a benefit with 2 extra cores? I was wondering if there's any considerable difference if it was an i7-8086K vs i7-9700K - I guess the 6C/12T w/ hyperthreading vs 8C/8T w/o - was in there or?
 
Not that I ever have any Twitch viewers, but it'd be interesting to see if there is a visible difference.

Don't half the Twitch streamers use their console anyways?
I will ask my son if he has info about it.

He said that you stream whatever rate you are at and it is static, no matter how many viewers. The viewers can pick their own res.
 
Last edited:
So, my opinion, strongly supported by my and
So, are they important or not? Does anyone have a Cole's Notes (summarized) version? ;-)

I'm kidding. It looks like there's a benefit with 2 extra cores? I was wondering if there's any considerable difference if it was an i7-8086K vs i7-9700K - I guess the 6C/12T w/ hyperthreading vs 8C/8T w/o - was in there or?

If you are streaming, it appears that more cores are critical.
It would be fun to test this vs a higher core, slower clock speed processor like 2700x, to see while streaming if fps are higher.

I think I'll disable 2 cores and run the tests while not streaming.
This would simulate something like a 7700k vs 8700k.

Although, the data surely already exists. Would be fun to recreate though.
 
If someone tells me how I can disable hyperthreading in my Asus Strix z370 bios, I'll do that too
 
I think I'll disable 2 cores and run the tests while not streaming.
This would simulate something like a 7700k vs 8700k.

Derp, I basically already did this in the original post.

New Data, for Assassin's Creed Odyssey.
Looks 6c6t is better than 4c8t.

With 6 cores and 12 threads, the lows are much higher, making a more smooth experience.
Odd how 6c12t has a slightly lower FPS. Probably within the margin of error.

KJ2mobU.jpg
 
Back
Top