galexander
Well-known member
- Joined
- Jun 24, 2007
- Messages
- 192
- Reaction score
- 0
- Points
- 0
I Downloaded Iometer and see that there's a lot of complexity in the choices and I have no reason to learn up on it all. Feel free to feed me a set of parameters that make sense from your end and I'll run them.
I use a nice utility I learned about from Axel years ago made by Matrox called filexfer. It can test your drives using very realistic video editing settings for file I/O, and can even read/write multiple streams at once.
For instance, the raw WD 500 gig SATA's I have on hand will do approx 84 MB/sec read/writes when empty when tested on a single video stream. Run two streams and the performance degrades down to the lower 30's. In comparison, my 4 WD 250 gig RAID 0 will still read/write over 200 MB/sec with only 9 gigs free space!
I've been doing some tests with REDCINE and my system. Using JJ's 4K 235 frame funny car clip for testing and reading from one nearly empty SATA and saving to another. My project settings are 1080p, the shot is scaled to fit height and an s-curve applied for grading. Processing output is set to "standard" and scaling is cubic, and the results when saving to a DPX stream is 2.4 frames/sec. Increasing the processing to "full", which then scales from the full 4K information, the result is 2.46 seconds per frame.
Yikes! I'm pretty sure I've got all REDCINE settings optimized, the window is the lowest quality setting and I'm careful not to rescale in output.
please, take this in the most positive manner you can.
these numbers to me are really not much use. what codec is it simulating? are you simulating a stream directly from the camera so the whole file is buffered initially?
excuse me but what the hell is "standard" processing or "full"? does this simulate rendering? or color corrections? pixel masking? noise removal? dark current calibration? calculating motion vectors? are you simulating a log or linear? gamma corrections? or somewhere in the post workflow?
why are you scaling the image? why aren't you working directly on a 2k or 4k frame? i don't have a redcine so i don't have an understanding of what you are doing but i have Flaskmpeg, Avisynth, and VirtualDub which are brutually and exceedingly fast.
does this simulate a convert to a raw footage first? use it compressed?
IOmeter doesn't care about any of this, you can benchmark ANY system, it is a tool worth knowing how to use IMHO. i'd rather have the old swiss army knife than a specialized tool.
all i need to know is the following,
what is the biggest byte size (jumbo frame) of the data that Red supports over the coppper GiGe? if it's intel chipset it might be around 9k, broadcom maybe 7/8k.
why are you seeing such a significant drop when you switch to a 4k frame? a 4k frame(4096x2160) is 4.3 times a 1080p (1900x1080) frame, but your processing decreased by a factor nearly 6 from 2.4 fps to .406 fps. whatever you did has a ton of overhead.