Frequency aside (faster is always better, other factors being equivalent), Maxwell likes more cores, and more modern ones, and does not care about the video card. For a concrete example, I have here two hexacore machines, and one quadcore; one hex is an i7-970 and the other an i7-3930K; both are 3.2GHz, but the 970 is based on the Intel Gulftown architecture, with turbo frequency of 3.46GHz, and the 3930K on Sandy Bridge, with turbo frequency of 3.8GHz. Both are not overclocked, with the 970 rendering the
Benchwell scene (these results are from when I rendered it with Maxwell 2) in 13m 32s, versus 11m 12s for the 3930K. The quadcore machine is a 3.5GHz i7-4771 Haswell, with turbo frequency of 3.9GHz, and though I haven't actually run the benchmark on this particular machine, others on the chart render it roughly in between the two hexacores, at around 12m 30s. I think that's pretty impressive, a quadcore CPU rendering faster than a hexacore one only 3 years older (both were released in Q3, the 970 in 2010 and the 4771 in 2013).
Going by your description, I'd guess your machine is a 2009 Mac Pro with dual Xeon E5520 2.26 GHz "Nehalem" quadcore processors, so judging by results from similar machines, it appears your machine should run the benchmark somewhere between 11m 52s and 13m 33s. To put this into perspective, the new Mac Pro quadcore base model runs a Xeon at 3.7 GHz; its base frequency being slightly faster than my 4771, it seems likely to beat your older dual-quadcore by a small margin. Moving up to six- and eight-core on the new Mac Pro, it is more difficult to say, since base clock speeds diminish as the core count is increased; the six-core has a clock of 3.5GHz, so I should expect it to take in the neighborhood of only 2/3 the time of the 4771, but it is more difficult to judge the eight-core, since its clock speed of 3.0GHz is appreciably lower. The turbo frequencies for these processors are all identical, at 3.9GHz, but it is difficult to say how often those speeds might actually be reached in a long-running process like rendering, once heat has built up in the CPU.
One thing is sure, though, in the case that you're locked into a Mac, and that is that the previous generation
2012 12-core Mac Pro, with dual 3.06GHz Xeon X5675 hexacores, still has very good performance, relatively speaking, turning times in the low 6-minute range. Judging by results from Windows machines using the 12-core 2013 Mac Pro's processor (i.e. the
Xeon E5-2697 v2 2.70GHz, benchmarking in the very low 6-minute range), I don't expect the performance of the new 12-core Mac Pro to be substantially different from that of the old one.
Regarding memory, the most you can use in SketchUp is 4GB, so as long as you have that, plus more to run the operating system & other resident programs, it is all you need. In other words, 8GB should be just fine. Maxwell Render Suite, on the other hand, being 64-bit, can use as much memory as you have, but you are asking about the Maxwell for SketchUp plugin, so this is not a factor.
That is some food for thought, anyway -- I hope it gives you some feeling for how to judge different CPUs, core counts, etc.
- By CDRDA