 |
|
|
|
|
|
| |
| |
|
|
|
|
| |
| |
|
|
I am comparing the "basic" performance between the 3.7 b35 (sse2 and none sse2)
to 3.6.2
I have used one file, phot_met_glass.pov, and I get the following results;
3.6.2 512x384 No AA 9.13
3.6.2 512x384 AA .3 15.55
3.7.b35 512x384 No AA 14.8
3.7.b35 512x384 AA .3 24.32
3.7.b35 SSE2 512x384 No AA 13.43
3.7.b35 SSE2 512x384 AA .3 23.29
SSE gives little improvement over none SSE but that might be due the the
features used in the test file.
The bigger question is the apparent degraded performance to 2.6.2! I am using a
1st generation 4 core Phenom 9500 processor at 2.2GHz (TLB workaround enabled),
Vista Home Premium and 2Gb ram.
Am I missing something or is the test file unsuitable? This is the first quick
comparison so I will try some more complex scenes to see if the results compare.
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
On 14.02.10 21:40, GraemeM wrote:
> Am I missing something or is the test file unsuitable? This is the first quick
> comparison so I will try some more complex scenes to see if the results compare.
I assume the numbers you provide are in seconds? Where is most time spend,
in the rendering pass or in the photon pass?
You should measure with a scene that takes no less then a few minutes to
render, everything else will measure a lot of things not related to
rendering speed.
Thorsten
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
I only have the scenes provided in 3.6.2 at the moment, however the next file
shows a definite performance improvement with 3.7 and even better with SSE2.
patio-radio.pov
3.6.2 512x384 No AA 90.87
3.6.2 512x384 AA .3 206.25
3.7.b35 512x384 No AA 80.9
3.7.b35 512x384 AA .3 189.39
3.7.b35 SSE2 512x384 No AA 71.79
3.7.b35 SSE2 512x384 AA .3 166.8
Thorsten, yes the last numbers are seconds (It didn't occur to me to put that in
and I should know better!). In 3.6.2 the time is nearly all rendering time,
I'll check with 3.7 and find a more complex scene to give a more robust test.
Thanks, Graeme.
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
On 14.02.10 22:11, GraemeM wrote:
> I only have the scenes provided in 3.6.2 at the moment, however the next file
> shows a definite performance improvement with 3.7 and even better with SSE2.
>
> patio-radio.pov
LOL, I have to say you really have a lot of bad luck with the sample scenes
you pick :-)
This scene uses radiosity, an experimental feature changed significantly in
3.7, and is not suitable for comparing 3.6 and 3.7 render times. Best try
some of the more "simple" scenes (without radiosity and photons), though
most don't render that long any more on modern systems.
Thorsten
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
With balcony.pov, yet another file with radiosity. I'm doing this between
reading messages.
3.6.2 512x384 No AA 232.6
3.6.2 512x384 AA .3 ?
3.7.b35 512x384 No AA 330.0
3.7.b35 512x384 AA .3 ?
3.7.b35 SSE2 512x384 No AA 320.64
3.7.b35 SSE2 512x384 AA .3 705.87
I'm not sure how to read this information. 3.6.2 single thread is simple enough
but I don't understand the 3.7 multi thread numbers yet.
3.6.2 512x384 No AA
Total Scene Processing Times
Parse Time: 0 hours 0 minutes 1 seconds (1 seconds)
Photon Time: 0 hours 0 minutes 1 seconds (1 seconds)
Render Time: 0 hours 3 minutes 51 seconds (231 seconds)
Total Time: 0 hours 3 minutes 53 seconds (233 seconds)
CPU time used: kernel 3.23 seconds, user 229.37 seconds, total 232.60 seconds
Render averaged 845.27 PPS over 196608 pixels
3.7.b35 SSE2 512x384 No AA
Render Time:
Photon Time: 0 hours 0 minutes 0 seconds (0.506 seconds)
using 7 thread(s) with 1.480 CPU-seconds total
Radiosity Time: 0 hours 0 minutes 8 seconds (8.537 seconds)
using 16 thread(s) with 44.720 CPU-seconds total
Trace Time: 0 hours 1 minutes 10 seconds (70.321 seconds)
using 4 thread(s) with 270.940 CPU-seconds total
POV-Ray finished
-
CPU time used: kernel 1.45 seconds, user 319.19 seconds, total 320.64 seconds.
Elapsed time 85.30 seconds, CPU vs elapsed time ratio 3.76.
Render averaged 2304.95 PPS (613.17 PPS CPU time) over 196608 pixels.
----------------------------------------------------------------------------
One thing that stands out is that 3.6.2 states 845.27 PPS (pixels per second?)
and 3.7 states 2304.95 PPS, three time as many but 87 seconds slower.
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
>
> 3.6.2 512x384 No AA
>
> Total Scene Processing Times
> Parse Time: 0 hours 0 minutes 1 seconds (1 seconds)
> Photon Time: 0 hours 0 minutes 1 seconds (1 seconds)
> Render Time: 0 hours 3 minutes 51 seconds (231 seconds)
> Total Time: 0 hours 3 minutes 53 seconds (233 seconds)
> CPU time used: kernel 3.23 seconds, user 229.37 seconds, total 232.60 seconds
> Render averaged 845.27 PPS over 196608 pixels
>
> 3.7.b35 SSE2 512x384 No AA
>
> Render Time:
> Photon Time: 0 hours 0 minutes 0 seconds (0.506 seconds)
> using 7 thread(s) with 1.480 CPU-seconds total
> Radiosity Time: 0 hours 0 minutes 8 seconds (8.537 seconds)
> using 16 thread(s) with 44.720 CPU-seconds total
> Trace Time: 0 hours 1 minutes 10 seconds (70.321 seconds)
> using 4 thread(s) with 270.940 CPU-seconds total
> POV-Ray finished
> -
> CPU time used: kernel 1.45 seconds, user 319.19 seconds, total 320.64 seconds.
Whats this? vvvvvvvvvvvvv
> Elapsed time 85.30 seconds, CPU vs elapsed time ratio 3.76.
> Render averaged 2304.95 PPS (613.17 PPS CPU time) over 196608 pixels.
> ----------------------------------------------------------------------------
>
> One thing that stands out is that 3.6.2 states 845.27 PPS (pixels per second?)
> and 3.7 states 2304.95 PPS, three time as many but 87 seconds slower.
I think I have been misreading the information, I think I should have been
looking at the 3.6.2 total time (232.6 seconds) and the 3.7 elapsed time! (85.3
seconds)
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
On 14.02.10 22:48, GraemeM wrote:
> Render Time:
> Photon Time: 0 hours 0 minutes 0 seconds (0.506 seconds)
> using 7 thread(s) with 1.480 CPU-seconds total
> Radiosity Time: 0 hours 0 minutes 8 seconds (8.537 seconds)
> using 16 thread(s) with 44.720 CPU-seconds total
> Trace Time: 0 hours 1 minutes 10 seconds (70.321 seconds)
> using 4 thread(s) with 270.940 CPU-seconds total
> POV-Ray finished
> -
> CPU time used: kernel 1.45 seconds, user 319.19 seconds, total 320.64 seconds.
> Elapsed time 85.30 seconds, CPU vs elapsed time ratio 3.76.
> Render averaged 2304.95 PPS (613.17 PPS CPU time) over 196608 pixels.
> ----------------------------------------------------------------------------
>
> One thing that stands out is that 3.6.2 states 845.27 PPS (pixels per second?)
> and 3.7 states 2304.95 PPS, three time as many but 87 seconds slower.
You are getting confused by very poor terminology in the output: The
CPU-seconds are not what is relevant, but the actual seconds it one line
shows above are the time it took to render the scene.
I had not realized there was output like this in the current beta (and
apparently many recent betas before). This is unintentional, so thanks for
pointing out this confusing output.
Thorsten, POV-Team
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
| |
|
|
Thorsten
Now I have spotted my mistake I think that the terminology in the Beta is clear.
(total time = sum of core times, elapsed time = overall run time)
The terminology is also clear in 3.6. (elapsed time = total time)
What is confusing is when they are compared as total time <> elapsed time.
Its not the program its just a change in users understanding what's actually
going on. This question would never have risen if I'd started more rigorously
and timed the versions with a totally independent clock, buts going too far for
a casual user playing at home.
Thanks, Graeme.
Post a reply to this message
|
|
| |
| |
|
|
|
|
| |
