Render Test results

PeterWright wrote on 11/27/2005, 12:18 AM

Yeah, both my PCs crapped out at 75%. At the time, the screen showed "Motion Blur " & timecode at 00:00:22:21

I must make sure I don't try a project with all these effects ;)

dvboy wrote on 11/27/2005, 3:16 AM

Hmmmm. ... Try as I might with different ram settings, I keep crapping out at 28:03 with low memory problems - and I've got 2G on board with a Pentium D 3.2

This shouldn't be a render test - its a torture test!

Ah well!

edit - at exactly the same spot as Yoyodyne 68,083kb.

Wolfgang S. wrote on 11/27/2005, 4:33 AM

When I open your test project, Vegas does not find files. For example, "c:/windows/media/windows xp balloon.wav" does not exist on my German XP. There are some media files, but I do not know which one to choose.

Which one is that? Could somebody describe me that file?

TheHappyFriar wrote on 11/27/2005, 5:54 AM

doesn't really matter what wav file, any wil do.

I get the same memory error. I had ram-preview set to 64 (a brand spanking new V6c install too!). I'll try setting it to 0.

wierd cuz I just upgraded from 512mb single channel to 1gb duel channel. :)

Udi wrote on 11/27/2005, 6:09 AM

I have the same low-memory error at the same place.

I found that the problem olso occurs if you render between 00:21:00 and 00:22:00.
If I change the transition type on track 5 from 3D fly in/out to something else - like 3D Shuffle - I can render this 1 second section - but it take an hour, on AMD Dual Core X2 4400, 2GB RAM and SATA Raid - using version 6c!

Anothe observation - the CPU is about 60% - so it does not utilize the 2 CPUs effectivly.

But the worst thing - the end result is WRONG!

On some frames, on the created file, I have black background.
The preview on the TL I have the orange Motion Blur text on each frame.

Udi

TheHappyFriar wrote on 11/27/2005, 7:52 AM

strange, i can render the section it stops at no problem.

Spot|DSE wrote on 11/27/2005, 7:57 AM

Hmm...This test completed fine on 2 systems here (both new AMD's) and crashed every other computer in the place. We knew this was nasty, but didn't expect this. Maybe one of the Sony engineers can step in and offer an explanation, because on all machines here we get the same 75% point and crap out. Even if I prerender that segment, or render to new track, I get that result.
It's just generated media, so nothing weird there...Maybe we've uncovered a bug.

fldave wrote on 11/27/2005, 8:27 AM

The render I started last night finished a few minutes ago. My lowly dual Pentium 3 1ghz finished successfully with ver 5d. Yes it took > 10 hours, but no errors. My P4 is running right now.

Ouch, Spot. Seems like more of a paging file test. 1 GB ram on both machines, plus ~1.5 GB page file used is some sections. Couldn't use the CPU enough (P4 bouncing from 100% to mostly 3% around 64% through) because it was waiting for page file access. Page file sits on separate WD SATA Raptor partition

Now I could use that 4GB Ram Disk card to put my swap file on.

Dave

DavidMcKnight wrote on 11/27/2005, 8:38 AM

...because on all machines here we get the same 75% point and crap out...

I'm currently on 76%, nothing crashing yet. But...it's taken 9 hours to get to this point with almost 2 hours left. This is on a P4 2.4 Gh box with 512K ram. Does that seem like an unusually long time, or about right for those specs?

Last night I came home from Fry's with a new Gigabyte 939 mobo, AMD 3500+, and 2 gigs of ram. Gonna gut my current P4 box for a fileserver, and drop these new parts in as my editing box. Sweeeeeet.

(Edit)
DOH! 512 Megs of ram, not K....

(Edit Part Two)
After 18+ hours and 85% complete, I'm shutting this down. It's not locked up, just taking waaaaaaaaaaaaay too long. I'll try again after I've upgraded.

johnmeyer wrote on 11/27/2005, 9:28 AM

Gee, I started complaining in an earlier post, but I'm going to shut up now. I actually feel lucky. No crash, and with Vegas 6.0c and a 2.8 GHz single-CPU, single-core, no-hyperthreading CPU, it finished in a "mere" 5:05:13.

I thought it had hung, but it kept cranking at about one frame a minute and finally made it through the slow spot.

Note to future benchmark developers: Benchmarks should be designed to finish in MINUTES not HOURS. You'll get a lot more results that way.

TheHappyFriar wrote on 11/27/2005, 9:55 AM

david, it might take so long for you because you only have 512k of ram. My Tandy 1000 RT back in '85 had more. :D

if benchs took minutes then all systems would be very simular. more stuff & longer time = better benchmark.

What drive are you guys rendering to? I was rendering to the drive my pagefile was in & it always crashed. I rendered to a different drive & it got farther, but I cancled it because I needed the computer. I'll let it run overnight tonight to see how it goes (it was saying 11 hours left when I stopped it)

rmack350 wrote on 11/27/2005, 9:56 AM

For some strange reason I was able to render the file. 3:26:53 on an Athlon64 3200+ with 1 GB RAM. Preview RAM set to 16 MB.

Rob Mack

Bill Ravens wrote on 11/27/2005, 10:01 AM

Running the test with an Athlon64x2 4600 and 2Gb memory, Vegas 6c crashed(completely shut down Vegas) at the 20 minute point. Predicted render time was ~30 mins.
Interesting data point, however was a max (stable and holding)CPU core temp of 53C with the stock fan on a Gigabyte GA-8KN Ultra 9 mobo.

rmack350 wrote on 11/27/2005, 10:20 AM

My temp hits 41C and then the fan kicks on and the temp starts falling. This is an aftermarket Zalman fan and it's a cool day today, in the low 60s.

Maybe temperature is an issue? I'm kind of surprised I completed the test, considering all the problems.

Another thing. My swap file was already set to over 1.5 GB before I started. I don't think Vegas ever ran it up that high.

Rob Mack

fldave wrote on 11/27/2005, 10:44 AM

Update on P4 render: 2.5 hours thru frame 661, then 1.5 hours on frame 662. No cpu usage, no page file usage for last 30 minutes. Cancelled the render.

1GB ram, ~2.3 GB page file, of which 1.7 GB was used. I had Dynamic ram (mistakenly) set to 512MB. So it appears you need about 2.2GB ram to run the test, any combo of physical and virtual, with minimal Dynamic Ram set in Vegas.

That is about what my dual Pentium 3 used, and it completed the test successfully.

JohnnyRoy wrote on 11/27/2005, 10:49 AM

I just rendered to uncompressed to eliminate any codec issues and I got a dump at 77% with :

Sony Vegas 6.0

Version 6.0c (Build 153)

Exception 0xC0000005 (access violation) READ:0x0 IP:0x7C9111DE

In Module 'ntdll.dll' at Address 0x7C900000 + 0x111DE

Thread: ProgMan ID=0x718 Stack=0x2B7F000-0x2B80000

Registers:

 EAX=00000000 CS=001b EIP=7c9111de EFLGS=00010202

 EBX=00140000 SS=0023 ESP=02b7f49c EBP=02b7f6bc

 ECX=59e10f88 DS=0023 ESI=59e10f80 FS=003b

 EDX=00140178 ES=0023 EDI=001401f8 GS=0000

Bytes at CS:EIP:

 7C9111DE: 8B 10 3B 57 04 0F 85 8C ..;W....

 7C9111E6: 31 02 00 3B D1 0F 85 84 1..;....

Stack Dump:

 02B7F49C: 00000000 

 02B7F4A0: 5AB67608  59BF0000 + F77608 

 02B7F4A4: FFFFFFFF 

 02B7F4A8: 25F39A10  25B10000 + 429A10 

 02B7F4AC: 02B7F6D8  02A80000 + FF6D8 

 02B7F4B0: 7C90EE18  7C900000 + EE18 (ntdll.dll)

 02B7F4B4: 7C910738  7C900000 + 10738 (ntdll.dll)

 02B7F4B8: FFFFFFFF 

 02B7F4BC: 00000FA0 

 02B7F4C0: 7C911596  7C900000 + 11596 (ntdll.dll)

 02B7F4C4: 7C9106EB  7C900000 + 106EB (ntdll.dll)

 02B7F4C8: 00000000 

 02B7F4CC: 5AB6FC10  59BF0000 + F7FC10 

 02B7F4D0: FFFFFFFF 

 02B7F4D4: 25F39A10  25B10000 + 429A10 

 02B7F4D8: 00000000 

> 02B7F4DC: 006678A7  00400000 + 2678A7 (vegas60.exe)

> 02B7F4E0: 006678D7  00400000 + 2678D7 (vegas60.exe)

  02B7F4E4: 0329DD6C  03240000 + 5DD6C 

  02B7F4E8: 25F39A10  25B10000 + 429A10 

  02B7F4EC: 0329DD40  03240000 + 5DD40 

  02B7F4F0: 00000000 

> 02B7F4F4: 0065F5DE  00400000 + 25F5DE (vegas60.exe)

  02B7F4F8: 0329DD40  03240000 + 5DD40 

  02B7F4FC: 00000001 

  02B7F500: 01A32D90  019E0000 + 52D90 

  02B7F504: 001401F0  00140000 + 1F0 

> 02B7F524: 7C910732  7C900000 + 10732 (ntdll.dll)

> 02B7F544: 7C90EE18  7C900000 + EE18 (ntdll.dll)

> 02B7F548: 7C910738  7C900000 + 10738 (ntdll.dll)

> 02B7F554: 7C911596  7C900000 + 11596 (ntdll.dll)

> 02B7F558: 7C9106EB  7C900000 + 106EB (ntdll.dll)

> 02B7F6B0: 7C90EE18  7C900000 + EE18 (ntdll.dll)

> 02B7F6B4: 7C9106F0  7C900000 + 106F0 (ntdll.dll)

> 02B7F6C0: 7C81006F  7C800000 + 1006F (kernel32.dll)

> 02B7F6FC: 7C8399F3  7C800000 + 399F3 (kernel32.dll)

> 02B7F700: 7C810028  7C800000 + 10028 (kernel32.dll)

> 02B7F70C: 00655A43  00400000 + 255A43 (vegas60.exe)

  02B7F710: 00000040 

  02B7F714: 00000070 

  02B7F718: 00000000 

  02B7F71C: 5AB67608  59BF0000 + F77608 

> 02B7F724: 0065D600  00400000 + 25D600 (vegas60.exe)

  02B7F728: 25FCE008  25B10000 + 4BE008 

  02B7F72C: 5AB67608  59BF0000 + F77608 

> 02B7F730: 0065D62D  00400000 + 25D62D (vegas60.exe)

- - -

 02B7FFF0: 00000000 

 02B7FFF4: 00501F10  00400000 + 101F10 (vegas60.exe)

 02B7FFF8: 00987688  00400000 + 587688 (vegas60.exe)

 02B7FFFC: 00000000

Not sure what's going on but it’s not good. Vegas just disappeared after that. My proc temps were steady at 52C so I don’t think it’s heat related.

~jr

Wes C. Attle wrote on 11/27/2005, 4:01 PM

Spot, are you sure there are no default or quality settings that need to be changed? You said at the top of this thread that the fastest CPU's can do this test in 15 minutes. I don't think it's true unless there is a difference in Vegas quality settings or motion blur or something else we might not know about.

My score: 02:16:24 on Dual Opteron 252 K8WE 4GB RAM
Where do you find your playback frame-rate?

gordyboy wrote on 11/27/2005, 4:44 PM

Yep - crashes out at frame 637 on mine - admittedly a Pentium 4 3.6ghz - 2gb RAM.

Changing RAM settings made no difference.

gb

CLamCam wrote on 11/27/2005, 4:54 PM

Well, thank goodness! Most everyone is have the "not enough memory" problem.

I was afraid my 3.6GH P4 w/1GB ram was broke. 8-)

OK, Spot, you've shown you can create a project that none of us can render. How about an easier benchmark test..

Don't know if anyone was predicitng a 15 min render but at the time my render crapped out it was looking like an hour or so.

Love the benchmark concept. Just don't make us have to buy your hardware setup.

Carroll Lam

Spot|DSE wrote on 11/27/2005, 4:57 PM

yeah, we're revamping this right now, simply because we've only got two systems that can render it, and even the dual AMD (looked initially like 15 mins, but ended up being nearly 30) ran slow when it hit the same spot as everyone else.
While we needed a new in-house benchmark....and this one should work fine, we have realized there is some kind of problem with Vegas rendering this to most any codec. It's inexplicable, since it's almost all generated media, and just a small bit of motion blur/supersampling.
We're test rendering a new project right now, hoping it won't crash our older systems. On the other side, I hope to hear from the engineering team as to why this is crashing some and not others. I've reset paging files, turned up/down RAM, most everything on the slower systems, and still lose it.

johnmeyer wrote on 11/27/2005, 5:03 PM

You should also make the render test shorter, so that it renders in less than 15 minutes on a baseline computer (e.g., 2.5GHz P4). That will ensure that even really fast computers will still take long enough to be able to measure differences between them (i.e., you don't want to have a benchmark that takes five seconds on a slow computer, and then is sub- one second on faster computers, because you won't be able to discern meaningful differences).

Spot|DSE wrote on 11/27/2005, 5:07 PM

Well, that's sorta the whole point. The original render test has gone from being a 10 minute render for many, to being sub 20 seconds for new systems. It's nearly 3 years old now, with a slight modification for Vegas 5...We wanted to see this come in about 15 minutes tops. We'll get it, just a question of what's killing the systems. Very strange. Initially thought it was the MPEG encoder, but John Rofrano crashed on uncompressed, too.

TheHappyFriar wrote on 11/27/2005, 7:21 PM

hey, how many threads are people rendering with? The default is 4 but single CPU/core PC's & non HT ones can only do 1. Is it possible that some effects are attempting to use the extra threads & can't? Thus crashing?

fldave wrote on 11/27/2005, 8:32 PM

Important point is: Threads are not equal to CPUs.

During the rendertest today, when it was going very slow, I checked XP's task manager. There were 312 threads being processed on my machine at that time. My Vegas option is set at 4, and my task manager showed both "cpu"s as peaked at 100% each.

A "smart" multi-threaded program can process multiple threads at once quicker than running commands through a single "pipe" or thread. If one thread takes up 100% of the CPU, then it makes no sense to allow more than that one thread to run at a time.

I haven't done any tests for Vegas's thread option, but I would at least make sure it is equal to or more than the number of cores*number of cpus you have.