Question / Help Build a PC with a 5960X. Ill be posting some comparisons between CPU's

Videophile

Elgato
So yea, building a 5960X system tomorrow.

Most likely will stay at stock clock.

I will be doing some tests with OBS in regards to presets, VS a 2600k and a 4770k. Anything you guys want me to look at/for specifically?
 

tiba

Member
So yea, building a 5960X system tomorrow.

Most likely will stay at stock clock.

I will be doing some tests with OBS in regards to presets, VS a 2600k and a 4770k. Anything you guys want me to look at/for specifically?


Check presets please.

Compare

720p@faster@60fps
720p@fast@60fps
720p@medium@60fps
720p@slow@60fps

and

1080p@veryfast@60fps
1080p@faster@30~45fps
1080p@fast@30~45fps
1080p@medium@30~45fps
1080p@medium@30~45fps

all with 3000 bitrate and buffer
 

yanis31

Member
very excited to hear this!
i will check the results,

in the mean time -
maybe someone can do me a favor and give a honest answer at how does a 4790k oc'd to some 4.4 ghz
compare to AMD FX83xx overclocked to 4.5-4.7ghz ?

(keep in mind my intention is not to hijack a post - it's my first on this site actually, and i am interested in this threads main subject just as much - just a quick answer to my question would be appreciated)

my 6350 @ 4.5ghz is doing well at 1080p downscaled to 1536x864 @ 45fps .. tested completely maxed out Tomb raider with a GTX780 for now, with the tressfx and everything - i do have to artificially limit the Actual Game to 45 fps too in order not to have some occasional stutters/high cpu usage in the Cutscenes... in other places it is not necessary...

i am upgrading to a 8 core in the next 2 days since i can trade in my cpu easy so its not going to cost me anything significant,
later when i can afford one of the big boys i will probably jump straight to "2011-3" stuff...
- beacause after 3 nights of tweaking it is apparent that the CPU presets are where it's at.

but at the same time i might have an option to spend a little more and change my mobo and cpu to the 4790k but i will simply not do it in case its barely any better than the 8 core @ streaming.

(and keep in mind i am very educated in the whole intel v.s. amd stuff and read and seen all there is to be seen about it - i just haven't researched streaming in particular that much - so let's keep this professional)

p.s. i believe the 5960x is lowered in clock only to keep it within TDP limits - it really should get watercooled and go 4.5 ish
unless you got a bad chip, (only the motherboard will need to be with real strong VRM circuitry for that...
 
Last edited:

Aeryn

New Member
in the mean time -
maybe someone can do me a favor and give a honest answer at how does a 4790k oc'd to some 4.4 ghz
compare to AMD FX83xx overclocked to 4.5-4.7ghz ?

For offline x264 tests that i've done, Haswell seems about on-par with Piledriver, 1 core to 1 module. So an i5 at 4.6ghz would be about as strong as an fx8320 at 4.6ghz. I actually got roughly faster (fx had to be at 4.8 to match it) but i didn't test a wide range of scenarios, only 1080p with some intense encoding settings not suitable for live encoding. HT increases encoding performance by about 20%, so it'd be about that much faster than an fx8320 that was overclocked to similar standards as it.

The problem is that encoding is a pretty ideal scenario for those chips, and the far stronger singlethreaded performance of Haswell gives some games as much as ~50-60% higher FPS at typical overclocks, so if you're playing those games, going from an fx-63xx to an fx-83xx might not actually improve your FPS at all while streaming


For the purposes of this thread, i think it's pretty hard to test properly but ideally you could do offline x264 encodes with various settings, that way it just encodes as fast as possible instead of you having to awkwardly try to measure CPU usage with HT or second thread in each module messing up the cpu usage% numbers
 

yanis31

Member
Thank you Aeryn for the detailed reply - basically just what i wanted to hear!

EDIT: would i actually be far off the mark by assuming that since videoencoding can use all threads the performance could be quite directly comparable to Cinebench R15 scores? (where the 8 cores are sitting roughly between the i5's and i7's) ?

i will stick with the 8 core for now, because i have no FPS issues in games when the 6core is paired with a GTX780
(pretty crazy combo, that probably could be recommended only to someone wanting to play 1440p on a budget)
and the 8core upgrade costs next to nothing compared to replacing mobo/cpu with intel
- i am not in need of fps improvement just looking for improved encoding performance on a budget,

i have to mention that i am not setting above normal priority for OBS because that will give me some stutters while encoding, - so i do not really lose noticeable fps while streaming it would seem...
in case the intels handle higher priority fine then we can have a completely different conversation, i simply do not know.

but since ive had a 770 in the system before this (long story)
- i know that the cpu isn't bottlenecking nearly as much as many people claim,
780 is still a healthy upgrade... just some of those games like skyrim will have the occasional fps drop from 75 to 48 and a few others that are just poorly optimized, most games run very well and above my monitor's 75 refresh...

ok back to the topic - eager to find out how the haswell-e's compare, and which cpu profiles are they able to pull off!
 
Last edited:

Aeryn

New Member
Cinebench scales differently, for example it gains like 1.3x performance from Hyperthreading while x264 only gets ~1.2x. Cinebench also shows relatively massive gains from RAM performance (equivelant to hundreds of mhz CPU overclock) while x264 can show no performance change at all from it

i wouldn't really recommend upgrading from an fx6300 to an 8320, unless you already have suitable motherboard and cooling to run an 8320 at overclock. If you don't, then it's not as good without them and it's expensive to add them.

Biggest issue with FX is just lacking the singlethreaded performance, because in some applications and some games they are neck and neck with some Haswell CPU's that cost more, but in others they are not even close to competing because they need the extra threads to keep up. Sometimes its not an issue at all but other times it's like using a CPU from 2007, which can be very problematic, so i didn't even really look at those upgrading from first gen i7 system. It's very much a sidegrade for anyone with a modern CPU (phenom II x6, first gen core i5/i7)

you should be able to estimate Haswell-E performance with any Haswell CPU quite easily, the 8c/16t CPU performs pretty much exactly like two 4c/8t CPU's strapped together (so ~2x a 4790k, 2.4x a 4690k at similar clocks)
 

yanis31

Member
I hear ya! - but to me it made sense since i have a full 6350 system running on a 8+2 power phase 970 chipset board, - cooler is Zalman's CNPS12X
a much cheaper upgrade since it's easy to sell the 6core... and i would probably still be looking for an upgrade
even if i got a 4790k ... that is of course unless it turns out OBS is not multithreaded enough to use all the 8 cores on
the 5960x ... but since it should be this stays true...

- not too sure what to say about the cinebench ram performance effects..
i had a 4 gig kit of kingston hyperx genesis grey 1600 9-9-9-24 in there when i built the pc...
i upgraded to 8gb corsair's Vengeance LP 1866 9-10-9-27 kit later and cinebench performance
almost went down 2-3 points at same processor speed... maybe the kingston was 1t and this perhaps is 2t
not sure... or maybe the amd mem controller can't squeeze more out of itself even tho it officially supports it,
but my initial thoughts were that cinebench prefers lower latency more than mhz, i believe i didn't see a performance decrease when i set my corsair's to 1600 9-9-9 speeds...

anyway - fast forward to today - i have this sitting on my desk as a write this, waiting to be plugged in :)
34d0vy2.jpg

why it's interesting and why would anyone even bother to do this?
check the posts by "the stilt" on this forum
http://www.xtremesystems.org/forums/showthread.php?289461-A-little-birdie-told-me/page3

and now my sicere apologies to the OP for taking this way off track....
 
Last edited:

Videophile

Elgato
I hear ya! - but to me it made sense since i have a full 6350 system running on a 8+2 power phase 970 chipset board, - cooler is Zalman's CNPS12X
a much cheaper upgrade since it's easy to sell the 6core... and i would probably still be looking for an upgrade
even if i got a 4790k ... that is of course unless it turns out OBS is not multithreaded enough to use all the 8 cores on
the 5960x ... but since it should be this stays true...

- not too sure what to say about the cinebench ram performance effects..
i had a 4 gig kit of kingston hyperx genesis grey 1600 9-9-9-24 in there when i built the pc...
i upgraded to 8gb corsair's Vengeance LP 1866 9-10-9-27 kit later and cinebench performance
almost went down 2-3 points at same processor speed... maybe the kingston was 1t and this perhaps is 2t
not sure... or maybe the amd mem controller can't squeeze more out of itself even tho it officially supports it,
but my initial thoughts were that cinebench prefers lower latency more than mhz, i believe i didn't see a performance decrease when i set my corsair's to 1600 9-9-9 speeds...

anyway - fast forward to today - i have this sitting on my desk as a write this, waiting to be plugged in :)
34d0vy2.jpg

why it's interesting and why would anyone even bother to do this?
check the posts by "the stilt" on this forum
http://www.xtremesystems.org/forums/showthread.php?289461-A-little-birdie-told-me/page3

and now my sicere apologies to the OP for taking this way off track....
No problem. :)
 

yanis31

Member
BTW doing fine with the 8370e ... haven't spent a huge time tweaking but for now have settled at a 24/7stable 4.66ghz @ 1.4 volts that passes all stresstests ... prime95 smallFFTs usually ends up max 57-58 degees celsius running on all cores
(the 6350 was eating 1.5v @ 4.5ghz to keep prime stable, and temps were higher, and reported VRM wattages almost the same as with this 8 core due to the excesive volts)
currently running 720p streams @45fps 3500kb/s with cpu preset "high" and looks like i have plenty of cpu headroom left...
 

CritVV

Member
You guys should look at this guy on twitch, and read his description, his twitch name is: Miley_gives_Becky.

He has a complete gaming/streaming setup, that costed him around 10000 dollars, completely liquid cooled. He has a rig that is being used for streaming, which includes an i7 5960x @ 4.4 Ghz and he can stream high quality 1440p @ 60 FPS at a medium preset with that beast. Quality is amazing, but he does exceed twitch's max bitrate by quite a lot (6000 bitrate). But still, his pc setup is admirable.
 
Yes, the 4770k at 1080p @ 30fps and 720p @ 60fps along with cpu usage and fps Very fast preset or any preset you would recommend.
think he was referring to the 5960X nevermind, I'm a moron.
anyways....
I own a 4770K and can do both of those with fast, faster and everything above no problem with minimal ware on cpu resources. I have never had a issue with FPS yet. But even with all the cpu power a i7 4770k there is no reason to do 1920x1080 @ 30/60fps.. I stream at 1280x720@30fps with 2000-2500 bitrate.
 
Last edited:
Top