-
Vegas 10… slow? And other benchmarks
I’ve been hearing a number of questions about whether Vegas 10 is [somewhat | profoundly] slower at rendering than earlier versions. I had not really noticed, until one night, when it was suggesting some crazy-long render, like 20+ hours for something that should have taken less than an hour.
That proved to be a weird and not-yet reproduced system state… it was gone after a reboot. But I figured, while I’m interest, I’d like to see. This also needs to be done using the same CODECs, to avoid 32-bit vs. 64-bit differences, other things out of Vegas’ controls.
So my video is a segment from a wedding video I’m working on. This was shot in a nightclub in San Francisco. I have a segment that’s intended to give a feel of some of the activity, no specific subject. For this, I sped up the video 4x, than ran it through a BCC7 plug-in for temporal ticks. That was serious CPU time… like a day, so I rendered it full, then figured I’d mix the original and see what level of blending I wanted. These tests were just the blend.
The files were two 17:40min, 14.2GiB Avid DNxHD files. 1920×1080/24p. I started with Cineform output, but switched to MPEG-2 HDD 1080/24p 50Mb/s. The output was cropped to 11:00 min. I used Task Manager and Resource Monitor (Windows 7) to look at CPU load. Results:
1: Vegas 9, 64-bit, render to Cineform: cpu = 98%, time = 21:23
2: Vegas 10, 64-bit, render to Cineform: ** CRASH **Yes, Vegas 10, with its new and improved Cineform support, doesn’t like to render to Cineform. This is the latest NeoScene beta, etc. Oh well, so I switched to MXF 4:2:2, which is a pretty reasonable alternative to Cineform or DNxHD and included with Vegas.
3: Vegas 9, 64-bit, render to MXF: cpu = 99%, time = 23:40
4: Vegas 10, 64-bit, render to MXF: cpu = 99%, time = 26:48Ok, that’s a win for Vegas 9. What’s even stranger is when you look at the details:
3: vegas90.exe 46.43% FileIOSurrogate 42.44%
4: vegas100.exe 75.60% FileIOSurrogate 17.05%I don’t quite understand what’s going on there… maybe Sony’s moved some processing formerly done in their FileIOSurrogate process into Vegas?
So, I was done, only, Friday morning, my system wouldn’t boot. Tried it for a few hours, pulled card and DIMMs, tried an alternate power supply, nothing — I got a splash screen, and it rebooted in the BIOS. So I trekked to Microcenter, and last night my system re-emerged with an AMD Phenom II 1090T (on sale for $220) @ 3.2GHz. So just for grins, I re-ran this:
5: Vegas 9, 64-bit, render to MXF: cpu = 98%, time = 16:18
vegas90.exe 44.59% FileIOSurrogate 38.45%6: Vegas 10, 64-bit, render to MXF: cpu = 93%, time = 16:19
vegas100.exe 69.28% FileIOSurrogate 19.13%So it’s a dead heat between the two… even with Vegas 10 not quite so loaded. That’s weird.. did they actually optimize Vegas 10 better for AMD than Intel? I was unhappy seeing the CPU not quite fully loaded in Vegas 10.. that suggests the CPUs are getting a little time off. So I moved one of the DNxHD files to yet another SATA drive, and tried it again:
7: Vegas 9, 64-bit, render to MXF: cpu = 95%, time = 16:30
vegas90.exe 44.56% FileIOSurrogate 39.66%8: Vegas 10, 64-bit, render to MXF: cpu = 97%, time = 15:50
vegas100.exe 71.29% FileIOSurrogate 19.02%It seems Vegas 10 is a bit better at I/O optimization, even if the rendering itself might take longer… sometimes.
Not sure about all conclusions. I do believe that spending ~$400 for a 70% boost in rendering is worth it (my unplanned upgrade). I usually advise people who aren’t spending company money to buy “at the knee of the commodity curve”. In short, a commodity is something that, when you spend twice as much, you get twice as much. Intel always has a $1000-or-so CPU… but that’s way off the scale; probably not that much faster than the next couple down. They can do that because there will always be a select few who’ll pay for the absolute maximum performance… even if that’s a fleeting thing. For example, the Q9550 in my defunct motherboard was a $1000 Intel chip for a few weeks. Six months later, I got it for about $200. Even today, it’s generally priced at more than I paid for the 1090T. Probably because, if you have a classic Core2 system, this remains a decent upgrade. CPU pricing these days is usually about desirability, not just performance. That’s why Intel advertises so much.
I think this definitely suggests that, if you’re getting a much faster CPU (Core i7 980X, the forthcoming “Sandy Bridge” chips.. Intel’s announcing that official at CES in January), you will probably need multiple drives, an SSD, and/or a RAID to keep the CPU well fed, at least from intermediate-class video files (native MPEG-2 or AVC will hit the CPU harder, the disc softer). I think some folks here have already seen high end i7 systems not running at full CPU, depending on the project.
-Dave
-
1 Reply
-
Interesting read. I have done mucho testing… Vegas seems to like a quad core over the 6 core. Clock for clock my AMD phenim II is the same speed as my 980x, that is my AMD is 30% less on the ghz and 30% slower and the renders… Therefore Vegas uses very simple float point calculations. I’ve found that Hyper threads are more like little anchors in Vegas Pro. I actually loose about 2-5% performance.
I have also tested many HD config, raid, ram drive etc. The compression of just about any high end codec limits the HD access to about 15 megs a second. Low bit rate like AVC at 512k is about 80megs a second. Therefore you really only need a solid 7200 RPM Sata HD, for all your needs.
I can only tap out my HD’s (raid 5) running in UNCOMPRESSED at an astounding 575 megs a second for a render. My raid will push just under 800 megs a second, so reading and writing seems to be at it’s peak.
To test that, I’d need to set up my ram drive again and render to the HD’s only with out the source files attached. But then again, I don’t use uncompressed, even at 12TB’s of storage, too much space wasted…
I hope the Buldozers are cool, then I’ll see if 16core – Vegas Max supported cores makes a dif.
This tuesday, I get to test an 8 (16t) core Xeon system…
take it easy…
Reply to this Discussion! Login or Sign Up