Single ci7-5960X/i7-6950X versus dual Xeon E5-2670 for 3D rendering/multi-threaded applications. Which is better/faster?

Eric_89

Commendable
Jul 11, 2016
2
0
1,510
I've been looking to build a new computer for a while. I was waiting for Broadwell-E. But with the prices of their 8 core and 10 core CPUs so crazy high it's got me looking at older dual 8 core Xenon's to get even more cores for a lower price point.


Obviously the Xenon in question is older, and there was the giant flood of them to the market, hence the lower price point. But I would be getting 32 threads with dual Xenon’s instead of 16 threads from the single i7.


I would imagine a lot of people are looking to similar alternatives to getting more cores for cheaper since Broadwell-E is so expensive. But I'm having very little luck finding rendering benchmark results online that compare a single core i7-5960x/6950x to older dual Xenon’s.


I was curious if anyone here has any insight into this? I will not be using this for gaming. Only 3d work and rendering. I don't use any applications that are single threaded. All of my applications are multi-threaded and want as many cores as possible.


From what I've read, a quick and dirty way to compare a xenon to a ci7 is to multiply the GHz by the number of threads. The Xenon I’m looking at is 2.6Ghz x 32 threads since there would be two of them = 83.2. Whereas the core i7 6950x is 3.2Ghz x 16 = 51.2. Is this thought process valid at all?


Thanks :)
 
Solution
Have you found any info on the scalability of the software you are planning on using? Very little software can productively use 32 threads. You will have to benchmark to determine if the hyperthreading will improve your throughput. RAM is faster on the newer platforms. But you get more memory channels with dual socket. You will have to find a power supply that has the necessary dual ATX12V connections to support a dual socket motherboard. You will have a higher power bill with the dual socket build.

The single thread PassMark for the Xeon is 1614 and 1986 for the 5960X. If you can keep the dual CPUs busy then you should have better performance from the Xeon.

kanewolf

Titan
Moderator
Have you found any info on the scalability of the software you are planning on using? Very little software can productively use 32 threads. You will have to benchmark to determine if the hyperthreading will improve your throughput. RAM is faster on the newer platforms. But you get more memory channels with dual socket. You will have to find a power supply that has the necessary dual ATX12V connections to support a dual socket motherboard. You will have a higher power bill with the dual socket build.

The single thread PassMark for the Xeon is 1614 and 1986 for the 5960X. If you can keep the dual CPUs busy then you should have better performance from the Xeon.
 
Solution

Eric_89

Commendable
Jul 11, 2016
2
0
1,510
Thanks for your response!

Most all of the 3d applications that I use will take as many cores/threads that you can throw at them. ZBrush for example can use up to 256 threads. I've also seen benchmarks of Max and Maya using 32 threads.