|
|
All the news now and in your digital future
New Mac mini OK for FCP6?
Posted by RJ
|
Re: New Mac mini OK for FCP6? March 07, 2009 02:40PM |
Xserves are more cost-effective, believe it or not. I did the math recently.
If you assume the Core 2 in the mini and the Xeon in the Xserve are capable of doing the same amount of work per clock cycle, the mini comes out to about $150 per gigahertz, while the Xserve is $156 per gigahertz for the 2.8 GHz model, so the Xserve is slightly more expensive. But the Core 2 and the Xeon are not equivalent, clock-cycle-for-clock-cycle. Running Compressor, the Xeon outperforms the Core 2 handily. So in the real world, the Xserve is actually more cost effective for building Compressor clusters.
The advantage of the mini, obviously, is that you can buy nodes for $600 a pair, rather than having to shell out $3,500 for eight all at once. But you end up spending more money for the same performance in the long run.
Not to mention the obvious fact that buying Xserves gives you the option of stepping up to Xsan later, which really sends cluster performance through the roof if your transcodes are largely I/O bound.
(I didn't just figure this for fun. I did a cost-benefit analysis recently of building a mini-based Qmaster cluster for our Final Cut Server system. I was as surprised by the result as anybody.)
If you assume the Core 2 in the mini and the Xeon in the Xserve are capable of doing the same amount of work per clock cycle, the mini comes out to about $150 per gigahertz, while the Xserve is $156 per gigahertz for the 2.8 GHz model, so the Xserve is slightly more expensive. But the Core 2 and the Xeon are not equivalent, clock-cycle-for-clock-cycle. Running Compressor, the Xeon outperforms the Core 2 handily. So in the real world, the Xserve is actually more cost effective for building Compressor clusters.
The advantage of the mini, obviously, is that you can buy nodes for $600 a pair, rather than having to shell out $3,500 for eight all at once. But you end up spending more money for the same performance in the long run.
Not to mention the obvious fact that buying Xserves gives you the option of stepping up to Xsan later, which really sends cluster performance through the roof if your transcodes are largely I/O bound.
(I didn't just figure this for fun. I did a cost-benefit analysis recently of building a mini-based Qmaster cluster for our Final Cut Server system. I was as surprised by the result as anybody.)
Sorry, only registered users may post in this forum.
