Can somebody else please check my findings? Loading Jim Burton's P5 CPU test (it's in free stuff) into P6, it initially rendered at 400x400 (size of the preview window), shadow mapping seemed a bit slow but not TOO bad. Changed render size to 800 x 800 to match P5 test and was surprised by how long it took. Repeated test several times to be sure: P5 - 118 to 121 sec P6 - 209 to 215 sec Other than having to set the render size, all the other render settings loaded same as in P5. (AMD A64 FX 53, 2GB RAM, ATI Radeon X800 Pro) On the plus side, the parameters palette doesn't seem to steel focus from keyboard shortcuts any longer and Delete followed by enter actually removes the object rather than starting an animation preview. -- Mark

Mark

Interesting numbers. Were both using the firefly renderer? How do the images compare in quality? Lighting? Etc. Just curious.

That's...unnerving that an FX53 would see that large an increase in time. That's really a powerhouse of a system you have there.

The end result looked identical as far as I could see. Comparing back and forward I couldn't spot any differences in the settings. Both are 800 x 800 renders.

Mark

second test. Simple 3 figure group, default lights, 1000x1000, raytraced. P5 - 54 seconds P6 - 136 seconds.

Mark

hi yggdrasil , You still have the best Jim Burton times, glad to see you are posting. I don't have p6 yet, but your information is upsetting. I wonder if there is something defaulting into play in P6 that is not obvious? Yet, your comparison window seems like apples are being matched with apples. When the test loads into P6, it doesn't add/change anything in the lights, does it? ::::: Opera :::::

Try Changing the max bucket size. That helped my render times a LOT

batesd, we are worried about render speeds on the SAME settings as P5 and trying to figure out what could contribute to a slowdown with NO changes on the benchmark. ::::: Opera :::::

Hi, sorry for the delay in replying, but unimportant annoying little things like sleep and work just got in the way... I checked the lights. The settings appear identical, and final rendered images are the same as far as I can tell. As Opera has said, purpose of test is to compare identical settings, however, just in case P6 had some particular problem with the small bucket size, I reran the test with a bucket size of 128 P5 - 93 sec P6 - 152 sec

Mark

okay, that is a red flag. yggdrasil, I won't have my P6 for a week, I'm sure by the time the box gets here etc. I of course will run the Burton test against my current 159 seconds for P5 on my AMD 3500. I will also run some of my own files, as you did. If P6 FireFly has been hit with some sort of setback given no change in settings, that is a serious serious reason for me (and many) to either voice concern or not upgrade. I can put up with most of the new features being not useful, or models not right for me, but I can't upgrade if the basic render core execution has taken a hit for no reason. ::::: Opera :::::

Attached Link: http://www.renderosity.com/messages.ez?ForumID=12356&Form.ShowMessage=2174088

Uh-Oh, Official? report of memory management bug(s) not found until near release that could be affecting the render speeds, inclucing Burton time trial. See post 15 of thread at attached link. ::::: Opera :::::

P5 : 110s P6 : 230s P6 with file saved from P5 : 180s Took a while to find out why I was having trouble loading the test file, but once cured, P6 is running well.

[hardware sidebar, not directly on topic of p5 vs p6 issue: nruddock, can you give your hardware, please? These fast times (110) on P5 are both exiting and driving me nuts because I though my new rig was tops at 159 but I am being destroyed by other rigs left and right. I am on AMD Athlon 3500+ XP GIGABYTE KSTriton series, NS Ultra-939 4 GIG Mushkin RAM (yeild 3.62 Gig) 2 37-Gig Maxtor Raptor Hard Drives Raid-0 Array Windows XP Pro, Page File 4.2 Gig Nvidia GForce 6800 video card ] 110 vx 230 is serious drop off. ::::: Opera :::::

Note the difference when using the file saved out of P5.
Loading the original straight into P6 causes it to use different Firefly settings.
Loading the file ssaved from P5 makes sure the render settings are the same and knocks a chunk of time off.

Time for the P5 varies between 110s and 130s.
ViaTech VT8386
AMD 1.4GHz
768Mb
1 20GB Maxtor, 1 60Gb Maxtor, 1 120GB WD
Win2k
Nvidia TNT Riva 64

I actually considering myself lucky that it's working at all, as I've had trouble the my graphics card in the past.

Message edited on: 03/22/2005 18:33

Resaving the "CPU test - p5.pz3" from P5 before loading into P6 made absolutely no diference for me. Still nearly 3.5 mins to render (using 800x800 render size and bucket size of 32)

Mark

Note: make sure "Keep textures loaded" and "Reuse shadow maps" are turned off, otherwise 2nd and subsequent renders will be significantly faster than the first (about 160sec vs 210sec for my machine). Clearing these settings restores original render times.

Mark

nruddock, I am not trying to be a pest, just trying seriously to track down the "Why" of real fast Burton times....you give the MHz of your processor, but can you say the name of your processor? I need to find out the size of the L1/L2 cache. is 768MB the L2 Cache? AMD????? Thanks ::::: Opera :::::

nruddock >> Loading the original straight into P6 causes it to use different Firefly settings. << whereas yggdrasil's post #3 above shows no differences. nruddock, can you pinpoint how loading the benchmark into P6 changes Firefly settings? ::::: Opera :::::

The processor is at least a couple of years old, and I don't remember which model it is (not taking the fan off to look). The 768Mb is my main memory. yggdrasil's picture looks different to the one I've got, so I'm wondering if their is a newer version specifically for P5 (there was only one in FreeStuff), but if that were the case it would most likely have the Firefly settings saved in it. The load straight into P6 case uses P6's default settings, in the saved from P5 case it picks up the P5 default's.

Attached Link: http://www.renderosity.com/freestuff.ez?Form.Contrib=Jim+Burton&Topsectionid=0

I'm using the 'CPU Test - P5', Jim also produced a P4 test, which renders much faster. Both are in Free stuff here, at the above link. -- Mark

Mark

Got the right file now.
P5 310s
P6 415s
These are in line with my clock speed relative to other peoples.

yes, in line. also, the p5/p6 differential seems like other results reported here. This is not good...will need to be fixed and hopefully can be fixed. We can't have render speeds headed in the wrong direction! ::::: Opera :::::