I am currently running a dual core AMD 2.8ghz machine over clocked to 3ghz and I am in desperate need of an upgrade. I have searched high and low and cannot get a definitive answer for the question above.
Assuming that the rest of the component specs are identical in both machine builds (I plan on building this setup as soon as I can make up my mind as to what CPU config I use), what would render a 3D scene created in blender faster:
a dual 4 core xeon setup with the processors running at 2ghz
or
a single over clocked i7 quad core running at 4ghz.
Please keep the answers to the blender render scenario as introducing any variables like multi tasking into the equation will just confuse matters and probably turn this thread into a mirror of all the other confusing threads out there.
You will need to physically run benchmarks on both systems yourself as guessing with so few specs (motherboard, RAM, OS etc.) is impossible as you pointed out. Perhaps you could contact a vendor and find out if they are willing. The forums at AnandTech would be a better place to posit this question as this is a hardware site and they use Blender for some of their benchmarks.
Thanks for the suggestion I will post it over at anandtech and see what help I can find. As for the other specs I have not listed them as what ever I build will have roughly the same hardware apart from the processors and motherboard. I was just curious to see if anyone had already done the leg work for me. If I do find a good answer I will post it here.
The 4 core will render always faster.
Because not absolutely all tasks are multithreaded so you end sometimes with only 1 task running and so the 4 core with faster speed will beat always the 8 core at half speed.
I am at the mercy of others when purchasing hardware and other factors like the motherboard can make a huge difference to performance, if some of the components are badly matched or under specced. I have two machines with identical cpu, gpu, monitor and RAM configurations (e6600, Radeon X1550, 2GB RAM, 22" LCD monitor). The one runs like treacle the other turned out to be quicker with Adobe CS4 doing certain tasks like exporting and previewing pdfs faster than our iMacs which ironically have the almost exactly the same specs. They have a little more RAM. The maths looks good, but it really is not as simple as that. A real life test really is the best way to test performance, especially if your hardware meant for a specific task. Have a look at some GPU tests and the sometimes marked performance differences between vendors’ cards that are based on the identical reference boards. You can run Blender off a flash drive on any OS, so I would run a few tests of your own on the demo models at your hardware vendor before purchasing.
The problem is that performance does not always scale linearly with MHz.
Higher MHz obviously means better single-threaded performance, which speeds up the render pre-processing tasks. I’d say the dual xeon setup is safer because it obviously runs much cooler and the performance scales nicely when rendering large scenes.
unless you have good cooling i would go for more vores over faster cores. the reason computers went to multi cores rather than just sticking with faster single cores was was heat. you can only make a cpu go so fast before it literally melts and dies. the more cores would should be a much more stable machine. sincee they are both intel products why dont you call intel or send them an email and ask them which cpu is going to be better for your needs. they are gonna get your money either way so they can aford to be totally honest, and will probably know the strengths and weaknesses of both chip better than anyone else.
speaking of cooling, one of the strengths of the i7 is that if you have the cooling it overclocks atleast 50% on air (cooled with just fans) with little effort on the user’s end. Whatever you buy, though, blender will try to load each core it is using to 100%! Get good cooling, no, excellent cooling no matter what you buy if you plan on rendering a lot.
That would be a great resource. I am now thinking about a wiki with a link to a “standard” scene. People could render thhe scene, enter thier system specs and render times!
I dont think Intel is the sort of company that would admit to customers that a cheaper overclocked system is just as fast as thier much more expensive server CPUs. They would almost certainly recommend the Xeon system. I know the Xeons would be much more stable and use much less power and have a longer CPU life but I just need a speed comparison at this stage.
For cooling I have opted for a water cooled system that I know works as a friend of mine runs one for gaming. The system that I intend to build is a modified high end gaming machine (similare to a friend’s machine) with a quadro card instead of the gaming card and a few other tweaks to help with video and graphics.
Yes Definatly! I intend to run a closed loop water cooling system to keep temps down. I know this setup works as a friend of mine is using a similar system to cool his gaming rig including his GFX card.
Could you please elaborate on that Felix. I would have thought that by adding more tiles you just render more of the frame at the same time, but at the same speed. So if you rendered say half the amount of tiles at double the speed you would get the same result in regards to render times. Is this right?
