X4 Ray Tracing - Optimal CPU

[Ian Pellant]

Hi Marty,

Darn it.... oc the i7...

I'm using TechBuy (www.techbuy.com.au) as my supplier... they added some new i7 processors yesterday. But they all need really fast RAM to get the most out of them. RAM over 1600 is uncommon and expensive, in Aus. The new Intels seem to be asking for 2100 or even 2400 RAM.... If the i7 needs fast RAM to overclock then it really could be pricey... I dunno.

Web reviews indicated that the AMD with memory faster than 1600 will overclock fine (spent a lot of time parsing the motherboard, memory web sites for the best choices, then trying to find them on TechBuy was tedious).

I just did a 2 session Chief test, running the test plan ray trace simultaneously. Ran fine... but each ray trace took exactly twice as long as running a single session. In other words - no time improvement over running a single session queueing the ray traces conseqeutively.

I expected that since it ran on an old Intel single core, which hyperthreads to pretend it is 2 cores. (Intel: the core count pretender?)

So: now the big question is whether the AMD FX can run multiple sessions of Chief and the ray tracers simultaneously, at no speed loss in each session.... or can Intel do it better?

From what I read on the web reviews, they didn't load up 4 copies of an application and run those simultaneously. Possibly the AMD FX will run applications in its' modules (4 modules, 2 physical cores per module... thread sharing between those 2 cores... 4 virtual computers at the same time?).

If Chief has not been compiled to support multiple instances so each can run in a separate core... Doug: why not?

Cheers.

[marty]

I had a look on line and it seems 4.5Ghz to 5Ghz is easily achievable with the i7 2600k
http://benchmarkreviews.com/index.ph...&limitstart=14

[Ian Pellant]

Marty,

Dare I ask you to run the test plan ray trace on your computer; then overclock it and see what the difference is?

Thanks.

[jpremodeler]

Marty,

Dare I ask you to run the test plan ray trace on your computer; then overclock it and see what the difference is?

Thanks.

I had a tough time at 5.0 keeping stable. I am air cooled, and my ram config only let me use 1 fan on my heat sink. If you can keep the 2600k stable at 5.0 with air, hat's off.

I did not benchmark my pre/post overclock, but it was a decent bump in speed. If I burn the cpu in a year, the time saved is worth it.

[Ian Pellant]

Overclocking....

I'm applying OC to my old P4 3.6GHZ.

It lives in a Shuttle that has a liquid cooled heat pipe system.

Results so far:
No OC 200: 17:16 CPU temp + 0C
OC 210: 16:27 CPU temp +1C
OC 215: 16:03 CPU temp +1C

Becoming interesting. CPU is running at around 3.9GHz and the performance boost is noticeable.

Currently running a test at OC 220 (frequency boost by 10%)... fly or fry?

The CPU normally runs flat out at 71C. Intel throttles it down at around 90C (Those old Prescott processors were designed to run hot). Seems I have a lot more OC headroom.

Cheers.

[jpremodeler]

I run at 71-73 C during raytracing with air

[Ian Pellant]

... went back to the Intel web site and found the white paper for the P4 660 series. (It is 6 years old, after all). It was designed to cut in throttling at over 73C. I'm on the edge.

OC over 4GHz and Chief ray tracer starts failing. It may be CPU temp throttling.

Seems stable at 220 (3.94GHz) which cuts the test ray trace to 15:43. Shaving almost 2 minutes off the time.

Studying the fan speed settings to see if the CPU temp can be brought down a bit.

Cheers