| ||||||||||
| Follow up… I am getting a matched set of (8x8GB) Hynix chips to test again. However, I did want to post a follow up observation. I set up a render with the project file to be rendered 10 times in a row. What was interesting was how the reported render times had roughly increased 20-30 seconds with each subsequent render, such that the 1st render was 8:35 and the last render was 12:24. So My next step was to open up Thermal Monitor to see what was going on with temperatures. Sure enough there seems to be a relationship between temps in the MacPro and render times. If I leave fans in normal mode, render times appear to increase as the temps inside the Mac Increase. If I use smcFAN to manually program higher fan speeds, render times remain consistent. So, when the computer memory temps are around 40-50*C, I get my fastest render time of 8:35. If memory temps are up around 65*C the render time increases to 9:30. I did not observe what the temps were for the 12:24 render time, but maybe I should try to check that. So, I am now wondering… which internal components, when internal temps become elevated, might reduce rendering efficiency? The only areas which seem to steadily rise are the memory slots 1-8 (1-4 being the hottest), CPU A heatsink, CPU A Temp diode and Northbridge chip. I had heard as long as temps stay below 95*C everything should be fine, and higher temps would trigger a shutdown. Does any of this make sense or ring a bell? I should have the replacement 8GB chips in a day or so and I am looking forward to seeing if these observations persist. Thanks for any comments you would care to make. RJ ... On Aug 11, 2014, at 11:59 PM, email blanca wrote: > Greetings… > > I have been test driving a replacement Mac Pro and have observed some dramatic variations in rendering speed. I understand to expect variations when dialing in the rendering sweet spot as far as # of processors and ram dedicated to those processors. I speaking about dramatic rendering speed variations within a given setup. > > The test set up is: > > 2009 4,1 Mac Pro with the 5,1 firmware upgrade > 2x X5670 2.93ghz 12 core tray > 64GB of ram, 4x from one brand, 3x from another brand and a solo chip from a 3rd brand. The rig supplier says the ram is identical in specs but I have no way to confirm, although they all report as 1333mhz ddr3 chips. I am trying to get 8 matched 8GB chips so I can to test further. > nVidia GTX 680 2GB for Mac > Crucial M4 512GB internal SSD for AE Cache and project rendering > External storage: Sonnet eSATA 3.0gbs via an E4P PCIe card and external D500P enclosures with WD 3TB HDDs > AE CS6 and AE CC 2014 (most recent updates) > Mac OS 10.7.5 and 10.9.4 on 1TB and 1.5TB seagate drives > > *** > > I have a test AE project, with render times on a 2007 3.0ghz 8 core 32GB ram Mac Pro under 10.7.5 and CS6, required 40min in non multi processor mode and 23 minutes under multi processor mode. > > I exhausted testing in CS6 to find the # processors and Ram/processor to find the sweet spot for 64GB of system memory with 12/24 cores. The project mentioned above renders in 8.5 -10 minutes under multi processor mode in CS6 on this faster Mac Pro. The observation I have questions about, is re-rendering this AE project and the render time being as fast as 8min 15sec or as slow as 9min 56sec, while I have made no changes to the project, AE settings or hardware. That seems like a very big swing in rendering efficiency. > > I have wondered if the mixed bag of memory modules may have some inherent issues which might intermittently contribute to the rendering process. I have tested the ram with Rember and Tech Tool Pro and the memory reported OK. I hope to shortly run the AHT on the MacPro to see if it finds any issues with the memory or the computer. I have tried deleting the rendered files from the SSD in case those files might have contributed to a slowing down of the SSD read/write speed. > > I won't get into trying to test AE CC, as that appears to be another can of worms, as far as AE CC being "render crippled" compared to CS6. > > I would welcome comments or suggestions to help me understand why some CS6 renders of the same project file are more than 20% better with no changes in the set up. > > Thanks > > RJ MIles > MacTunes > ... |
