Page 1 of 5
Speed Test 1.1 - 1.5 32bit - 1.5 64bit
Posted: Tue Jun 05, 2007 6:08 pm
by Left Side
Hi,
Following on from my previous post I have carried out further testing to compare speed between Maxwell 1.1, Maxwell 1.5 32bit and Maxwell 1.5 64 bit.
The test was run on the same machine a dual quad core Dell with 4mb memory running Windows server 2003 x64
The same MXS file was used in each case and set to render to level 18.
The results were:-
Maxwell 1.5 32 bit rendered in 1h42m08s Benchmark 420.36
Maxwell 1.1 rendered in 2h02m17s Benchmark 351.10
Maxwell 1.5 64 bit rendered in 2h35m46s Benchmark 275.62
To me this shows that the 64bit version has a problem.
Anyone else have the same problem with the 64bit version?, is there a special procedure for installing the 64bit version that I have missed?.
Thanks in advance for any help.
Rob
[/img]
Posted: Tue Jun 05, 2007 6:58 pm
by tanguy
OMFG !!!
That really frustrate !
Could you send me your test scene ?
We will make the same here on our blades in 1.5 32 and 64 to compare.
tanguy@pixus.fr
Posted: Tue Jun 05, 2007 7:07 pm
by Left Side
Tanguy,
Thanks for the offer to test. I have sent the file to you.
Let me know how you get on.
Best Regards
Rob
64 vs 32
Posted: Tue Jun 05, 2007 9:47 pm
by inigomontoya
You're doing better than I did.
Quad core q6600 @ 2.4ghz 2GB Ram
w/ the Speedtest 1.1
64 bit Benchmark: 60
32 bit Benchmark: 157
I thought maybe it was because I only have 2GB Ram, but it doesn't seem to be using anywhere near that amount.
Posted: Tue Jun 05, 2007 11:19 pm
by Rickyx
I had some test on Win32 compared to Linux64:
Benchmark is better on 32...
Why?
Posted: Tue Jun 05, 2007 11:21 pm
by Rickyx
I had some test on Win32 compared to Linux64:
Benchmark is better on 32...
Why?
Posted: Wed Jun 06, 2007 12:01 am
by ivox3
Confirmed here...
X64 | actual stop watch times | 8 cores @ 2.33 | no 32 bit 1.1 or 1.5 tested for comparison
Posted: Wed Jun 06, 2007 12:24 am
by Maxer
No I don't beleave it
, why would 1.5 be slower than 1.1 I didn't think they even touched the core in this update.
Posted: Wed Jun 06, 2007 1:26 am
by inigomontoya
1.5 and 1.1 benchmark pretty close if you compare 32 bit versions. I'm only seeing issues with 64 bit.
Re: Speed Test 1.1 - 1.5 32bit - 1.5 64bit
Posted: Wed Jun 06, 2007 1:49 am
by dilbert
Left Side wrote:
Maxwell 1.5 32 bit rendered in 1h42m08s Benchmark 420.36
Maxwell 1.1 rendered in 2h02m17s Benchmark 351.10
Please clarify your test results. From what I see from your results, 1.5 is significantly faster than 1.1 when comparing the two 32 bit versions.
Re: Speed Test 1.1 - 1.5 32bit - 1.5 64bit
Posted: Wed Jun 06, 2007 1:50 am
by dilbert
Sorry, double post.
Posted: Wed Jun 06, 2007 8:30 am
by shen.de
Maxer wrote:No I don't beleave it
, why would 1.5 be slower than 1.1 I didn't think they even touched the core in this update.
sure u can! I have a studio setup with a single chair as a test scene...
on 1.5 it reached SL7 after 10minutes
on 1.1 same scene reached SL 10
thats absolutly anoying!
Posted: Wed Jun 06, 2007 10:02 am
by renbry
for whatever it's worth i did tests on realflow 32bit vs 64bit and 64bit came out slower....i assumed it was using a higher precision solver rather then using the extra bits for speed...
maybe Maxwell is using a similar methodology?
matt
Posted: Wed Jun 06, 2007 10:12 am
by tom
Ivox3, except the difference in speed, I see a serious gamma difference in your comparison. Do you have any idea?
Posted: Wed Jun 06, 2007 10:51 am
by Left Side
Hi,
Dilbert, to clarify:-
Maxwell 1.5 32bit is faster than Maxwell 1.1
but
Maxwell 1.5 64bit is slower than Maxwell 1.1
I would have expected there to be a further speed gain in using the 64bit version of 1.5. But for the time being I will use the 32bit.
Should we post something a 1.5 bug thread?
Thanks for eveyones help.
Rob