Bit the bullet and did a CPU transplant

[mp2017]

1. Me too. Just traded in my 2010 6-Core 3.33 GHz, for 2012 12-core 3.46 My render speed is slightly more than doubled. Renders that took an hour, now cooking in 27 minutes.

The question is: Did the GeekBench scores reflect the observed rendering increase?

As an example: I've performed some power measurements for another discussion in the PPC thread. As part of that I have both Geekbench and Handbrake times for a number of systems. Here are a couple of metrics from those results:

• MacBook 2,1: 2.16GHz Core 2 Duo, 2GB RAM, GMA950 graphics, OS X 10.6.8, Geekbench 3, Handbrake 0.15.5
  • Geekbench scores: 592 single, 1,030 multi
  • Handbrake transcode time (in seconds): 1,569
• iMac 11,2: 3.06GHz Core i3, 8GB RAM, Radeon HD 4670, macOS High Sierra 10.13, Geekbench 3, Handbrake 0.15.5
  • Geekbench scores: 1,966 single, 4,134 multi
  • Handbrake transcode time (in seconds): 644

So what do these numbers tell us? According to Geekbench the iMac single thread score is 3.3 times faster than the MacBook score and the multithread score is 4.0 times faster than the MacBook. However the Handbrake transcode time for the iMac was only 2.4 times faster than the MacBook. Given Handbrake is a multithread aware application Geekbench significantly over estimated the speed increase of the iMac over the MacBook. IOW in this particular instance Geekbench did not accurately reflect the speed gains of the iMac over the MacBook.

This is why Geekbench scores should be taken with a grain of salt. One can only rely on them to be indicative of how a systems performs running Geekbench. Furthermore the Handbrake transcode times I've provided should not be taken as absolute. Under different circumstances, such as a different source file or different transcode settings, the results are likely to vary.

2. The cool thing about the cMP is I'm still far from maxing it out. Because my CGI software can also go GPU render (with the addition of Octane as a plugin to Modo) I can add 1 or even 2 1080ti cards and triple or quadruple my render speeds yet again.

3. In 2012, a maxed out 12 core had a price tag of $12,000. I wanted it, but didn't need it. When the modular Mac Pro emerges, I will likely be in the same boat. Perhaps I will be buying the 2018 Mac pro in 2022 LOL.

4. Another giggle. When I traded in my beige G3 for the silver tower G5, I thought this might be the last computer I'd ever need. My perception of what "fast" is keeps changing!

This is because the problems systems are used to solve become larger. Years ago NTSC / PAL video was the norm. Then we moved to HD. Now we're up to 4k. At some point 6k or 8k may become the norm. The surprising thing to me is the resources needed to view the modern web. The web has really become bloated!

 

[h9826790]

The question is: Did the GeekBench scores reflect the observed rendering increase?

As an example: I've performed some power measurements for another discussion in the PPC thread. As part of that I have both Geekbench and Handbrake times for a number of systems. Here are a couple of metrics from those results:

• MacBook 2,1: 2.16GHz Core 2 Duo, 2GB RAM, GMA950 graphics, OS X 10.6.8, Geekbench 3, Handbrake 0.15.5
  • Geekbench scores: 592 single, 1,030 multi
  • Handbrake transcode time (in seconds): 1,569
• iMac 11,2: 3.06GHz Core i3, 8GB RAM, Radeon HD 4670, macOS High Sierra 10.13, Geekbench 3, Handbrake 0.15.5
  • Geekbench scores: 1,966 single, 4,134 multi
  • Handbrake transcode time (in seconds): 644

So what do these numbers tell us? According to Geekbench the iMac single thread score is 3.3 times faster than the MacBook score and the multithread score is 4.0 times faster than the MacBook. However the Handbrake transcode time for the iMac was only 2.4 times faster than the MacBook. Given Handbrake is a multithread aware application Geekbench significantly over estimated the speed increase of the iMac over the MacBook. IOW in this particular instance Geekbench did not accurately reflect the speed gains of the iMac over the MacBook.

This is why Geekbench scores should be taken with a grain of salt. One can only rely on them to be indicative of how a systems performs running Geekbench. Furthermore the Handbrake transcode times I've provided should not be taken as absolute. Under different circumstances, such as a different source file or different transcode settings, the results are likely to vary.


This is because the problems systems are used to solve become larger. Years ago NTSC / PAL video was the norm. Then we moved to HD. Now we're up to 4k. At some point 6k or 8k may become the norm. The surprising thing to me is the resources needed to view the modern web. The web has really become bloated!

Geekbench 3 or 4? Geekbench 4 score is non linear scale.

 

[mp2017]

Geekbench 3 or 4? Geekbench 4 score is non linear scale.

Version 3. Specifically version 3.4.2. Version 4 does not run on Snow Leopard (i.e. the MacBook). Can you elaborate on the non-linear scaling in GB4?