X4 Ray Tracing - Optimal CPU

[marty]

Last time - I am now going to do some work!!

Re set to 4.2Ghz and got 57, 55, 58 seconds so the 47 seconds first time up was a one off? Is it the clock thats a problem?

I think I will leave it at 4.2 as the small improvement going to 4.5 doesnt seem worth the potential problems that could come with pushing the system. I have an extra big fan but no other special cooling features.

[Ian Pellant]

marty,

Congratulations! You broke the minute barrier.

I am in the office running the test on a 3 year old HP workstation. It has an Intel Zeon W3505 CPU @ 2.53GHz. It's same die as the i7-9xx series. 2 physical cores; 2 thread per core. Win XP 32.

Very interesting results:

Single session: 12:54

Now the interesting bit:

2 sessions of Chief at the same time:
1 12:54
2 12:54
that is: each completed in 12:54 - same as a single session.
that is: 2 instances of Chief running on separate copies of the test plan completed the ray trace exactly as if it was one normal session of Chief. This is a doubling of productivity when many plans need to be ray traced. The computer is behaving as if it is two computers. Chief is raytracing completely within each core using 2 threads per core.

Now the decider:
4 sessions of Chief at the same time:
1 24:28
2 25:48
3 24:48
4 24:54
This indicates that each of the 2 physical CPU are sharing the 4 threads across their 2 thread capacity... so the ray tracing time per instance has doubled. Same sort of result as my old single core P4.

I closed 2 sessions and ran the remaining two simultaneously. Results:
1 12:38
2 12:36
Test confirmed. A 2 physical core Xeon processor will run the ray tracer contained within each core. (On Win XP 32bit)

However:
Tests on an Intel i7 running Win 7 64bit did not behave this way... it was back to doubling the session time per ray trace.

For rhe best performance, one may have to step up to an HP or Lenova workstation fitted with an Intel Xeon processor.... (unless the Amd FX-8150 has core behavior similar to the Xeon)

Cheers.

[marty]

Ok I said I was working but thought I would run two on my system and got 1:55 for both which is double the time.
The times you are getting are so slow I wonder why you are bothering with testing them. Its good to see the higher specs returning a noticeable improvement - so often upgrades fail to deliver.