best pair? 8350 + 770 / 4770k + 760

[edaniux]

the price of an i7 4770k is about $150 more than the fx 8350. And the price of the GTX 770 is also about $150 above the 760.

So which duo is better?

 

[DelightfulDucklings]

I would say out of those two options the fx 8350 and gtx 770 would be much better, however would you be able to go for an i5 4670k and gtx 770? For gaming that would be a better option because games do not utilize all of the 8 cores available in the 8350 so the better clock for clock performance in the i5 really helps in gaming situations. However if you want those extra cores/threads for another reason such as CAD that is entirely understandable

 

[Shneiky]

What about an I5 4670k and 770. Best of both words. Games don't really use that Hyper threading well so it doesn't help much.

 

[anything4this]

i5 4670K + 770, save up for the extra $50 . Do not go with these two as you will regret either the 8350 or the 760 at earlier points than you ever would the 4670k. If you are not encoding, then this is your best bet (by a large margin)!

 

[edaniux]

Yes I had video editing in mind that's why I didn't post the 4670k. Because I need the hyperthreading, or the 8 cores.

 

[DelightfulDucklings]

In that case I would recommend the 8350 + 770 because the graphical difference between the 760 and 770 would be noticeable

 

[ingtar33]

I would get the 8350+770, actually i would get the 8320+770 and use the saved cash to get a better mb and nice cpu cooler

(the 8320 is a 8350 with a slightly lower clock speed... they overclock as well, and behave identically at the same clock speeds, if you look at the top 10 overclocks for amd 8 core cpus you'll see just as many 8320s and 8350s, because as i said, it's the exact same chip and will have a similar max clock speed)

 

[mapesdhs]

Yes I had video editing in mind that's why I didn't post the 4670k. Because I need the hyperthreading, or the 8 cores.

The 8350 is NOT an 8-core CPU. It has 4 modules, each with two int units and a shared fp unit.

Check the video encoding benchmarks in numerous reviews, despite supposedly being "8-core", the
8350 loses out again and again to various Intel options because Intel's IPC is so much better, eg. see:

http://www.tomshardware.com/reviews/core-i7-4960x-ivy-bridge-e-benchmark,3557-8.html

Also, check the application you're using - some video apps can not use more than 4 proper cores,
but they do benefit from the way HT works. It varies. And remember power consumption: the 8350
chews a lot more power, ie. a 4770K will often not only get a task done quicker, it'll whack your
power bill less severely in the process, so even if the chip costs more to buy, in the long run the
total costs will be less.

Lastly, don't forget to investigate possible GPU acceleration of whatever video task you're considering.
This might make the CPU choice moot, depending on what the app can exploit re GPU options.

Ian.

PS. If you're using After Effects, you need big RAM, an SSD for the media and general cache, Intel
6-core rules, and strong GPUs for CUDA are a must.

 

[ingtar33]

every intel in that review is a 4 core with hyperthreading, or a 6core with hyperthreading. the comparable i5 would pace a solid 20%-25% slower then those numbers. And the amd is a 8 core, you're playing with semantics. Its an 8 core that shares parts between core pairs~ of course it's a different discussion all together as to whether this is a good idea or not.

The op clearly has a budget, and an i7 is outside it. since he can't afford an i7 + 770, the previous advice in this thread stands, a 8350+770 is a better combo then a i7+760, if he can only pay a certain amount for his cpu, advising him to spend more money then that in an i7 which is 100 outside his budget makes zero sense. Furthermore, since the i5s are all slower then the fx8350/8320 in video encoding the previous advice in this thread remains, his best bet is to go with the amd.

 

[Shneiky]

Mind I tell you that in Adobe Premier Pro, all that CPU power is used mainly during rendering. So lets say you do 1 hour on cutting and placing the clips together and then render. With an I5 you are gonna render for 1 hour. With I7 you are gonna render in 50 minutes. With PentiumG you are gonna render 1:45. But mind you, with a PentiumG and GT520 (and an edited GPU file in PremierPro to enable the GT520) you will render in 20 minutes.

// This comparison is made when I render from Maya on my main rig, and do some PremierPro stuff on my other machine. //

Also the difference in rendering speed between lets say, a gt520, gtx550 or a gtx780 is so laughable compared to the money. Also mind, that if you use Adobe Creative Cloud, the PremierPro there supports OpenCL accelerated Mercury Engine, which means AMD cards come into the picture as well. Since I don't use the CC but have a standard CS6 edition I can't really say how it performs but I would expect the OpenCL version will be faster.

Now how can you improve rendering or working speeds while video editing? Simple - SSDs. Render to an SSD, this decreases rendering times more than any CPU+GPU configuration. A simple PentiumG with GT520 with a SSD outrenders 2700k with a 560TI and an HDD by x2 and up.

That's for PremierPro. Now lets talk about AfterEffects.
AE benefits from RAM. Believe it or not, I5 with 16GB RAM beating 6-core I7s with 8GB is a daily occurrence. And I am not talking about rendering. Anything can render. Even a P4 can render. The problem is your work pipeline. You can really work efficiently if everything stutters in the view port. Also putting your AE tempfile on an SSD will bring further increase in productivity. CUDA in AE only affects certain video effects, the number of which is amazingly small. And CUDA in AE doesn't impact rendering times. (AE CS6 doesn't use the Mercury Engine )

So ladies and gentleman, this is it. My quick guide on how to optimize for PremierPro and After Effects.

//
TL : DR: It is not only MOAR CORES and MOAR GPU. It's not only rendering benchmarks. If you don't use this kind of software and don't know it in and out, don't post advices, misleading other people. AND REMEMBER: Before you start flexing all those cores to render (and compare benchmarks) you have to get to there. In my personal experience the I5s and I7s are a lot snappier in the view port than the AMD FX cpus. Alot of the functions and sub functions for video preview don't scale with the threads the same way rendering does.

 

[edaniux]

I'll be using a lot of Cinema 4D R13 too, which is the software I use for 3D modelling and rendering. And thanks for the PremierPro and AE optimization info, I was planning on getting 16GB RAM anyway but know I'm more convinced that that's the way to go with RAM.
So...to sum it up, you would recommend the 4770k with the gtx 760?

 

[Shneiky]

If your content creation work outweighs gaming then go with the I7 and the 760. If gaming is more important - then go with I5 and 770.

Why I rule the AMD out: Because only in rendering the FX is efficiently being used. While actually working a lot of functions and subfunctions don't scale well enough to saturate the FX's modules as well as rendering does. The less, but more efficient cores and higher per core performance makes the Intel platform snappier in productivity apps. Also the Intel platform has higher GB/s read/write from/to RAM. Which is crucial to performance in AEs video playback. I haven't used much C4D, one of my colleges is dealing with it so I can't really help how to optimize for that software. Good luck.

 

[mapesdhs]

every intel in that review is a 4 core with hyperthreading, ...

Actually I had meant to add, I wouldn't recommend either of them. It's a much better idea
to obtain a used 2600K or 2700K; easier to oc, a lot cheaper, etc.

Ian.

 

[mapesdhs]

Shneiky writes:
> Now how can you improve rendering or working speeds while video
> editing? Simple - SSDs. ...

With respect to rendering, I assume that must be for Premiere
because that's definitely not the case with AE. For interactive
working though, yes it's very important to use SSDs for the
cache locations with AE.

When rendering with AE, and assuming the cache locations are
already set to be on SSD(s), it doesn't seem to matter whether
the render file target is on an SSD or not.


> AE benefits from RAM. Believe it or not, I5 with 16GB RAM beating

That's why my AE system has 64GB running at 2400. 😀


> everything stutters in the view port. Also putting your AE

Having an SSD for the cache locations is essential.


> CUDA in AE only affects certain video effects, ...

You're forgetting it also accelerates some aspects of the
interactive display (Composition, Layer and Footage panels).


> ... And CUDA in AE doesn't impact rendering times.

Only if you're using Classic3D. If you're using the RayTraced
mode, then CUDA absolutely does make a difference because that's
pretty much what is doing all the work.

One of my test scenes takes 70 minutes to render on a 5GHz 2700K,
but done via CUDA with a single GTX 580 it completes in only
69 seconds.


> So ladies and gentleman, this is it. My quick guide on how to
> optimize for PremierPro and After Effects.

Alas your info on CUDA isn't correct. See:

http://help.adobe.com/en_US/aftereffects/cs/using/WS37cb61f8f3397d86-164c9f19127035dd253-8000.html
http://forums.creativecow.net/thread/2/1019120


> personal experience the I5s and I7s are a lot snappier in the
> view port than the AMD FX cpus. ...

Indeed; people forget that numerous individual actions within an
application only execute with one thread. This is definitely true
with Maya, probably AE aswell. Any time this happens, the far
lower IPC of the 8350 will really hurt. It's why my own tests show
such excellent results for pairing a pro card with an oc'd i3 -
single-threaded apps such as ProE speed up a lot.


To the OP, forget both the 8350 and 4770K. Find a used 2600K (or
2700K if you can), a decent used P67 or Z68 board (I got an ASUS
Maximus IV Extreme for a good price), as much RAM as you can,
couple of SSDs, some used GTX 580s (which will leave a 760 or 770
far behind). You'll have a much quicker system at far less cost.


For reference, this is my AE research system:

Coolermaster HAF 932.
ASUS P9X79 WS
i7 3930K C2 @ 4.7GHz (Phanteks PH-TC14PE with 3x 140mm fans)
Thermaltake Toughpower 1475W Modular
GSkill 64GB DDR3/2400 CL10
4x EVGA GTX 580 1.5GB (I'm testing to compare with a Quadro 4000 + 3x 580)
Samsung 840 250GB (system)
OCZ Vertex4 256GB (AE cache)
2x Hitachi 2TB Enterprise SATA


I also have a 5GHz 2700K setup, if you want me to run any
comparative tests. Lots of other AMD/Intel CPUs and AMD/NVIDIA
GPUs aswell if I can be of any assistance with tests.

My system is an evolution of an AE machine I built for a friend
back in February, exploiting used parts where possible to reduce
costs (shortly will be replacing the two 460s in his system with
two 580 3GB cards):

http://www.sgidepot.co.uk/misc/aepc.txt

See the CUDA benchmark discussion thread above on creativecow
for more info on CUDA acceleration with AE.

Ian.

 

[Shneiky]

Respect Ian.

>> Only if you're using Classic3D. If you're using the RayTraced
mode, then CUDA absolutely does make a difference because that's
pretty much what is doing all the work.

That is if you use RayTracing in AE. At least for me, I stick to classic3D most of the time. I get snappier functioning in my workflow with it. But in the end it is how you use AE that makes the difference. After all, I do all those things that are accelerated by the Ray-traced engine not in AE, but in Maya.

 

[mapesdhs]

Shneiky writes:
> Respect Ian.

😀 Not sure that's wise - if my head gets any bigger it'll need its own post code. :}


> That is if you use RayTracing in AE. At least for me, I stick to classic3D most of the time. ...

Must confess that surprises me, given the massive render speedup offered by CUDA in
RT3D mode and the additional effects one can achieve, though I suppose it depends on
what kind of scene one is doing.


> ... But in the end it is how you use AE that makes the difference.

Probably. 😀

I don't actually use AE, I'm just doing CUDA performance research, etc. My main
video setup is a trio of systems running Flame, Smoke and Inferno, though I've
not really done much with them yet, not had the time.

Ian.

 

[anything4this]

Buy a used i7 if possible, and new video card (if you look around you may stumble on amazing deals). Intel will back the cpu and chances are nothing will go wrong with it. The motherboard should be fairly low cost, as well as the ram (buy used as well). Put money into a good power supply though (don't overkill).

 

[edaniux]

Thanks to everyone, specially Shneiky and Mape who took their time to explain. I've decided to go for the i7 4770k 16gbRam and an MSI Mobo, and save up some money to get the 770, in the meantime I'll use my old GPU. My current rig is really outdated so any upgrading will improve performance a lot anyway.

If only I could choose 2 best answers...

I'll just toss a coin...

 

[mapesdhs]

Sneiky writes:
> That is if you use RayTracing in AE. ...

Yes, that's specifically what I was referring to. Two orders of magnitude faster. 8)


> At least for me, I stick to classic3D most of the time. ...

Something confuses me though - sometimes when I've tried to render a scene in that mode
(such as the benchmark file created by Teddy Gage), the visual result is completely different,
a if Classic3D mode isn't able to produce the effects one expects. The AE guy I've been helping
told me he can't get the effects he wants in Classic3D mode; his work seems to involve creating
a lot of reflective surfaces, the natural hunting ground for ray tracing I suppose.


> ... I get snappier functioning in my workflow with it. ...

I guess for the systems I've built, this isn't an issue. 😉


> ... Ray-traced engine not in AE, but in Maya.

That's where my knowledge ends. 😀 I've not examined Maya performance on PCs at all yet,
so far I've only been investigating Maya on old SGIs, mainly because review sites seem to
cover it reasonably well, and Viewperf 11 does include a Maya test.

Ian.

 

[mapesdhs]

anything4this writes:
> Buy a used i7 if possible, ...

That's certainly what I would do, 2700K if possible.


> ... and new video card (if you look around you may stumble on amazing deals). ...

That would at least give him a full warranty, though I've done well with used cards.
My last GTX 580 only cost 98 UKP total. Couple of those will easily outperform a
780 for CUDA and be less than half the cost.


> ... The motherboard should be fairly low cost, ...

True, new mbd prices are pretty good these days. I've bagged some good deals
on used boards (M4E P67 for only 87!) but buying a used mbd on eBay is probably
the riskiest of any computer item purchase I'd say; only buy from someone who
has a listing that says they accept returns, and make sure any board ships with
the protective CPU socket cover in place. And of course ensure it comes with the
I/O shield.


> ... as well as the ram (buy used as well). ...

Not sure about the US but used prices on RAM have been really ramping up here
recently. I won a used 8GB DDR3/1600 kit last week for not much less than what
the price was a year ago for a new 1866 or 2133 kit, while the latter new have
gone up by 50% in some cases.

As for new kits, I'd recommend the GSkill 32GB (4x8GB) DDR3/2400 CL10 kit,
but it's probably not cheap (still, saving on the gfx would then provide the budget
to cover this).


> ... Put money into a good power supply though (don't overkill).

Another area where I've done well. I've bought numerous Thermaltake Toughpower
units, mostly 750W/850W units for lesser builds, several 1kW and a 1200W for more
substantial setups. If he can get a 1Kw unit, that would be good. For my 3930K though,
and unusually for me, I did decide to buy a new 1475W unit so that I could be certain
it would handle four 580s safely.

Ian.

 

[mapesdhs]

edaniux writes:
> Thanks to everyone, specially Shneiky and Mape who took their time to explain. ...

Most welcome!!


> ... I've decided to go for the i7 4770k ...

In that case get the best cooler you can, top-end air cooler like the Phanteks I mentioned, or
an H100i. From the toms review, an oc'd 4770K runs hot. That's why I think a used 2700K is
a better idea (huge cost saving easily covers a good cooler and offsets good RAM cost), but
never mind.


> 16gbRam and an MSI Mobo, and save up some money to get the 770, ...

If what you're plannig to do can exploit CUDA, get two used 3GB 580s instead,
they'll leave a 770 in the dust and cost about the same. One 580 would be much
cheaper and might save enough to allow you get 32GB RAM instead.


> ... I'll use my old GPU. My current rig is really outdated so any upgrading will improve performance a lot anyway.

Best wishes whatever the outcome! 8)


Btw, if it's of any help, my CPU benchmarks page is here at the moment:

http://www.sgidepot.co.uk/misc/tests-jj.txt


Ian.