2009-03-10 03:04 PM
2009-03-10 05:19 PM
2009-03-10 07:34 PM
2009-03-10 08:45 PM
Chazz wrote:Yes, there are extremely brief moments when I can see AC 12 use 8 cores - like for 1 to 2 seconds tops when generating a hatched, shaded elevation. (So, with 4 cores, it would 'cost' you an extra half second to second for those elevations.) Most of the time, only 1 or 2 are being used - sometimes 3+. So, at least with 12 (no idea about the future), 2 to 4 cores are adequate for ArchiCAD.
Sounds like it could be better to get the faster 4 core rather than the slower 8 core. Karl, didn't you say that AC can utilize (at least for certain operations) a full 8 cores?
2009-03-11 06:19 PM
2009-03-11 06:30 PM
2009-03-11 06:50 PM
Matthew wrote:I agree for ArchiCAD use, and for the small Artlantis renders that I do. Like you, Matthew, I'm completely happy with the responsiveness of my prev-gen Mac Pro.
From what I've sen so far the new machines don't represent enough of a bump to justify upgrading after only a year.
2009-03-11 06:56 PM
Chazz wrote:Ouch. I did not check the pricing until you said that! That pretty much took care of my machine-envy issues.
Well this is depressing.... The faster chips are soooooo much more expensive (a US $2600 premium) and anything less does not appear to offer much of a performance benefit over previous generation...
2009-03-14 04:04 AM
2009-03-14 04:18 AM
UPDATE ON MEMORY INSIGHTS
Though the "sweet spot" theory is not supported by the latest After Effects results, we can say with certainty two things:
1. After Effects and other high-end apps do benefit from more memory. Note the difference between going from 3G to 6G in the 4-core Mac Pro and from 6G to 12G in the 8-core Mac Pro.
2. After Effects and other high-end apps do benefit from more cores. Note that the 8-core 2.26GHz beats the higher clocked 4-core 2.93GHz test unit.
But we still contend that 6 sticks are better than 8 (in the case of the 8-core Nehalem). Putting memory in the first three slots of each memory bank produces faster memory transfer speeds than filling all four slots on each bank. We ran a memory stress test with DigLloydTools which does a memmove() to all of unused physical memory. We put 12 GB (6 x 2G) in first. Ran the test. Then installed 16GB (8 x 2G) and ran the test. Here's what we got for max combined read/write throughput: