hardware question: dual quad core vs core i7

i am working with a dual xeon quad core system at work. that’s 8 cores … it’s quite fast … and i cannot work with my old dual core (2 cores) system at home anymore :smiley: it just drives me nuts!

so i am wondering what my next system should be. now the office pc setup is very expensive. the new intel core i7 cpu’s are cheaper, have 4 cores that - due to hyperthreading - show as 8 cores in the task manager. so it could be close to the work setup. but … is it?

does anybody have theoretical or practical info how these two setups would compare in terms of rendering speed?

thanks in advance

Not quite.
That hyper threading trick splits up one core to look like 2 but does not add any processing power. To me HT looks like CPU internal task switching. The only use for that would be if one thread was waiting for some external event ( e.g. a database server to answer ) the other thread is on the processor pipeline too so it can take over the the CPU core without delay. But for rendering it does not make too much difference if you render on 1 core in 1 thread or on 2 hyper thread created “virtual” CPUs in 2 threads. From what i have seen on HT (Xeon) systems it is even a bit slower than turning off HT in BIOS.

not exactly what you asked for, but maybe close enough - comparison between i7 and q6600 and how they do in relation to the phenom ii: http://www.tomshardware.com/reviews/phenom-ii-940,2114-10.html

if you are talking specifically about blender, then the blender benchmark page should have valuable information: http://www.eofw.org/bench/

I have another way you can speed up your cpu but i wouldnt recommend unless youve worked with servers before! I have a quad core cpu and a dual core server that i use as a personal server that i put uneeded stuff on my laptop on it! Not only can it kimit the powerand stuff you use it cool to! Or if your willing to void your waranty you can always use that server,or another one, and make it like quad core laptop with an extra dual core hardcore! But i wouldnt do that unless you dont care to void your warranty or you sure you know how to do it! Hope that helped!

The Creator

one of the cineform guys has been running benchmarks for their encoder (probably as close to a render benchmark as you’ll find in terms of raw CPU abuse), he’s finding the i7 considerably faster. Have a read:

There are some Blender benchmarks for the I7 here - http://www.blendernation.com/2009/01/17/blender-and-intel-benchmark/

and some others here - http://www.eofw.org/bench/

While the i7 is clearly faster than a current quad core system, a dual quad is still going to blow it away vis a vis Blender render times.

They say you have to use 3GB of RAM with core i7. If you use 4GB it will be slow. Is this correct? and why?


I think it comes from the triple-channel system bus: for most of the current motherboards (dual channel ones), you have to install memory sticks by two for maximum performance versus by three for i7 system…

This should mean RAM can be increased to 6GB with core i7
We don’t know where we stand from Intel, they keep changing stuff, who knows what they’ll do next.

I’d personally go for a set-up of dual AMD Phenom X4 black edition, and do some clocking.

Yeah! Support the underdog! Keep competition alive! Otherwise we’d all be stuck on Single threaded P4’s right now. :wink:

well, my view point was a bit more practical than that, really…

The black edition has got all the multipliers(not sure if thats what their called in english) open, so its damn clockable,
the problem with first gen. phenom’s though is that you’ll have to seriously clock to get the heat up if your using something extreme for cooling(such as dry ice) becaused of a design bug.

read the wikipedia article to learn more about that if interested.

The phenom II is fixed & looks damn promising though!

I’m an AMD fanboy and have been eagerly awaiting these. I run all of my machines heavily overclocked and have had some great success (I have my Athlon X2 3600+ Brisbane overclocked to 2.8 from 1.9 ghz) What’s funny is that my 3600+ overclocked FEELS faster than my X2 5000+ BE overclocked from 2.6 Ghz to 3.0 Ghz.

Read this comparison by Toms Hardware

You’ll see that in 3D max (the closest benchmark they do for Blender rendering) the i920 smokes the Phenom II. Read the article to get into Tom’s ability to overclock each chip.

But, for most people on an AM2+ MB, you can easily pop one of the new Phenom II’s in and you are good to go. Sure, you are still using DDR2 instead of DD3. But to go to an Intel i920 you have to pop for a MB, ram and proc.

well, when i7 was released intel stated that there wont be any dual processor mobo’s coming for it,
witch is why I’d still go for a black edition, though I just decided to see if I can find some new info and there ya have it:


INDEED! More $$$$ than I could afford.

yeah but imagine sculpting in blender on a 64bit system with this set-up and a few good quadros in sli. :eek:
oh and while using glsl viewport, duh. :stuck_out_tongue:

edit: yeah, so I like to dream of being THAT rich…

Don’t all of us wish we had rigs like that! I’d love to throw some intense fluid sims at it. Oh…and throw in some uber cool unbiased renders that converge in minutes.

Is this really possible? If so, please point me to where, please :D.

That’s not quite true, simultaneous multithreading does add processing power with the proper core design (and no, Netburst was not quite a beneficial core design).
Most modern CPU cores have many different execution units, and a single thread can never constantly occupy them all, even if everything is in the caches and registers, because of data dependency or simply because the code doesn’t have enough instructions of each type in the decoding window.

If you look at Cinebench results for example, Core i7 scores in the multithreaded test quite a bit higher than 4 times the single threaded score (anandtech states 18810 points vs. 4475 for Core i7-965, that 4.2 times higher, while Core2Quad scores only 3.57 times higher with multithreading…asuming the scores really are inverse proportional to execution time).
But it requires that your application scales well with threads in the first place…and of course it’s not a magical doubling of processing power.