Question / Help Streaming with Intel i7-5960X

yanis31

Member
Measuring CPU usage is the wrong way to benchmark this. Run your game and then run OBS at various presets with the same resolution, downscale, FPS, etc. And then look at the log file and measure the duplicated frames. If duplicate frames are less than 1%, it passed, if they're more than that, it failed. Make sure to give it at least five minutes of encoding time to make it statistically valid. And try to run a timedemo or something that is a consistent CPU effort on the game-side.

i did that because my rig is not even close to being able to stream 1080p60fps on anything cpu intensive and this shows approximately how much cpu usage streaming itself does. And i have my doubts about 5820k... its more or less 50% more than a 4770k/4790k + a quad channel memory controller... and those two aren't close to being able to handle that either.
i have spent days... almost weeks in fact testing different obs settings on different games - that data is enough for me to understand approximately what to expect - plus the differences in cpu usage between "veryfast"and "fast" cpu presets... not to even mention slower ones.... that is why i think even if it will be able to do it it will be struggling ALOT. and that's not what you want when running such high overclocks.
 
Last edited:

yanis31

Member
thread is dead?
- We could add something new to this fun:
- Assassin's Creed Unity is insane! those crowds kill the stream @ 720p 30fps ... it's absolutely ridiculous! :)
 

yanis31

Member
Measuring CPU usage is the wrong way to benchmark this. Run your game and then run OBS at various presets with the same resolution, downscale, FPS, etc. And then look at the log file and measure the duplicated frames. If duplicate frames are less than 1%, it passed, if they're more than that, it failed. Make sure to give it at least five minutes of encoding time to make it statistically valid. And try to run a timedemo or something that is a consistent CPU effort on the game-side.

i started checking this a while ago and yes i can confirm - the tests that "pass" just by looking at the task manager / checking if no "high cpu encoding" messages appear coincide well with that 1% stuff... all the log files reported less duplicated frames, the ones that did more also had the message popup and read cpu usage almost at 100% at times BUT:
i recently got acess to a stock clocked 4/4.4 i7 -4790k - it would seem that on this cpu the usage readout didn't go as high when it started to fail the stream... probably has something to do with the hyperthreading architecture cause the whole cpu graph looks more detailed on amd somehow...
also as a sidenote - the 4790k stock is barely any better at encoding than my 8370@4.7 ... it seemed a little better @ 1080_60_fast and could almost swallow medium but with something like 5% duplicated frames while mine was hitting some 20% at that setting... (results will vary depending on the source video) -i didn't notice issues at "fast" while mine was near it's limits... so with a bit of overclock to some 4.5 ish ghz or so the intel would be nice but still ... not the huge difference people expect from intel...
 
Last edited:

Boildown

Active Member
I'd recommend waiting for Broadwell-E to be released. Should be soon, we'll know by the end of the month.
 
Top