[numerobis]

...My MoBo's have the option of changing the RAM multiplier so it runs at 333x3=1000mHz or 333x4=1333mHz. I actually tried at 333x2.5 for 800mHz and it was considerably slower due to being out of sync. Was that what you were asking?

oh, i see!... i didn't know that you can set the fsb multiplier, or let's say i didn't looked for it - all actual boards should be able to do at least 400x4 ...your board or an 965 should be good for ~500x4

so, if you have it at 2.5-3x i think that should be the reason for a 5% lower speed...

About the 64bit slowdowns, is anybody on a pure PC 64bit machine experiencing these slowdowns?

yes, i postet the results in the speed comparison thread some weeks ago:
http://www.maxwellrender.com/forum/view ... torder=asc

with the old 1.1 testscene i got these values:

1.5 32bit bench ~180
1.5 64bit bench ~75

(q6600@default winxpx64)

[-Adrian]

About the 64bit slowdowns, is anybody on a pure PC 64bit machine experiencing these slowdowns?

For me 64bit was like 3 minutes faster. I believe i see a pattern of machines with more than 2 cores experiencing this slowdown.

[GM5]

With 8 cores 64bit Maxwell is about 4 times slower. 32bit was just over 31 minutes, 64bit was heading for over 2 hours when I canceled the render. Speedtest 1.1 had the same problem with 64bit Maxwell. Hope this is fixed soon...

-Greg

[Mihai]

But Greg, you're running 64 XP on a Mac right? Trying to see if it's a problem on all 64bit Windows, or just those running it on Macs.

inigomontoya, thanks for your feedback, trying to check this now.....

[GM5]

But Greg, you're running 64 XP on a Mac right? Trying to see if it's a problem on all 64bit Windows, or just those running it on Macs.

inigomontoya, thanks for your feedback, trying to check this now.....

I'm using a MacPro at home but I'm not running it in emulation. As far as Win XP 64 is concerned, it is a PC. I have run benchmarks using 64bit Vray on this system and it is faster than the 32bit version.

BTW, I noticed the same 64bit Maxwell slow down on my system at the studio - a Boxx dual quad 2.66 PC. I just don't have time to run these tests at work so I can't give you render time comparisons.

[ivox3]

Mihai, .... maybe you didn't see my results/posts .. ??

As you know, ... a pure 8 core 64bit PC over here, ...I attest for one confirmed slowdown.



btw: Nice bit in 3D W.....

[-Adrian]

Ok guys, as requested we're starting over with a much quicker scene (<50%). I added a column for 32/64bit so the two are in the same list now. Due to the liberty of having a second chance i also figured out how to add just a single table for each coming Maxwell version and keep the rest of the db untouched and cross referenced. So in the future when 1.6, 1.7 and 2.0 come around, you can pick which list to contribute to or let it default to the current one. The old lists will be selectable at the top for nostalgia/comparison (pfff, 8 cores!?).

Hopefully the void on the frontpage doesn't come as too much of a shock, i've backed up your entries just in case

I took over some of your CPUs, but mainly those i was familiar with and had a logo for (intel core family) and the IBM Apples. The rest will have to be added again, sorry for that.

I came to think that page could be be broken on IE6, and maybe that caused some of the strange entries in the past. Since the IE7 upgrade replaced everything here, i can't check that. So better use Firefox, IE7, Safari, Opera etc.

benchwell.com

[GM5]

Adrian,

Just posted my info and it worked fine (used Firefox). Nice job!

-Greg

[tom]

Ok guys, as requested we're starting over

Oh ship!

[dkessler]

Adrian,

The Mac OS X .mxs file I downloaded was set to 2 CPU Threads. I had to change it to 0 in Studio. This may be why Q! Berlin's time is over an hour, and mine is 24 minutes (both 8-core Macs).

Dan

[mgroeteke]

Adrian,

The Mac OS X .mxs file I downloaded was set to 2 CPU Threads. I had to change it to 0 in Studio. This may be why Q! Berlin's time is over an hour, and mine is 24 minutes (both 8-core Macs).

Dan

confirmed. are we supposed to set it to the optimum setting or not?

markus

[jdp]

Just for the record, I packed the mac version for -Adrian, and I noticed the thread problem only afterward. As far as I can tell you, the PC scene is set up exactly the same, just with different paths. I can re-save and pack the scene with thread set to 0.

[-Adrian]

jpd, we should try that. However i noticed that when i reopen the scene it's on 2 threads again right away and i can't say if that's auto-detection or if i simply can't safe the file with 0. In Maxwell 1.1 the setting remained at 0 no matter what.

confirmed. are we supposed to set it to the optimum setting or not?

Yes, use all power you got, 8 threads for 8 cores

This may be why Q! Berlin's time is over an hour, and mine is 24 minutes (both 8-core Macs).

I agree.

Oh ship!

Definitely the last start-over, not gonna happen again. I know you contributed the most of all. See the sunny side, your slowest machine is gonna take like 3 hours instead of 7

[-Adrian]

If someone with more or less than 2 cores could open this scene and tell me if threads are correct now. Exported from Studio with 0, but 2 again when i open it.

[Q2]

Adrian

Just did, but it still only uses 2 threads!

Q! Berlin