[Irinikus]

I know that this won't be completely accurate, as the versions of Blender used on the Mac's differed, but it does however give you some idea as to how Apple PPC's and SGI's compared performance wise.

[mrthinlysliced]

Interesting stuff Irinikus.

I'm particularly taken by how relatively competitive the Octane2 remained (we know the Tezro is a beast).

Anyone ideas of what the power draw of those are? A kind of "speed/power" adjusted cost thing.

[Irinikus]

Interesting stuff Irinikus.

I'm particularly taken by how relatively competitive the Octane2 remained (we know the Tezro is a beast).

Anyone ideas of what the power draw of those are? A kind of "speed/power" adjusted cost thing.

I must admit that I've been fairly impressed by the Octane2's performance in general and would now personally suggest it as the best option for most users, as machines such as the Tezro will only really pull ahead when it comes to rendering.

The only problem here would be the cost and availability of the dual 600MHz R14K CPU module for the Octane2. (More the availability, as they really don't seem to be very common.)

I currently don't possess the required power meter to give you a power consumption comparison for these machines, but when I get my hands on one I will add the readings for each of these machines to my findings.

One thing that I've noted with my Octane2, having the dual 600MHz R14K's and a V12 is that you can reach the limits of what the power supply can output if you really push it.

This happened to me the other day, while I was running Doom and instances of Quake, Quake2 and Quake3 simultaneously, as I reached the threshold of a resettable fuse on one of the output rails on the power supply.
 
The fuse is the yellow component in this picture:



I would love to try and run this machine with two V12's as Raion wants to do, but I fear that the Cherokee power supplies in these machines won't be able to successfully supply the required current. (this will have to be researched though)

[computron]

Have optimise the render threads setup ??

put more thread than CPU is the way for better perfs..

i have done several test and the results are really better with 4 or 8 threads with a dual cpu octane by examples.

another tips, use a lot of render tiles is better too !!!!

Cheers

Eve

[Irinikus]

Have optimise the render threads setup ??

put more thread than CPU is the way for better perfs..

i have done several test and the results are really better with 4 or 8 threads with a dual cpu octane by examples.

another tips, use a lot of render tiles is better too !!!!

Cheers

Eve

Thanks for the tip!

At some point I can redo the tests to try and get the absolute maximum performance from each of the machines. What I did do for all of these tests, which I carried out on the various SGI's was to use two threads per CPU, to keep things standard, as I simply wanted to gauge their relative performance under a set, set of conditions.

This evening, I'm going to run a comprehensive set of Blender tests on my Tezro, to gauge the absolute performance of this machine is from a Blender test point of view.

EDIT: The absolute best performance that I was able to obtain from this machine was 1 minute 5 seconds and 24 split seconds with 8 threads and the screen split up 16x16.

[Irinikus]

This video shows how the render setup in Blender effects render performance on an SGI Tezro. (Differing thread counts and numbers of tiles.)

[Irinikus]

Here's an updated performance chart, Including the results from my latest SGI:

[Irinikus]

Here's the updated list:

[vishnu]

Thanks for all the effort involved in these comparisons! As SGI owners, patience is a virtue. 😁

[Irinikus]

Thanks for all the effort involved in these comparisons! As SGI owners, patience is a virtue. 😁

It’s a great pleasure! I really enjoy comparing things, so it’s no problem or effort at all!