[Christer Molander]

Seek time – I mean the discs seeking time. The hardware. But now this has taken a new turn. After two loong conversations with Sonnettech support (they really try, and knows a lot, but not all about this) we now get a troughput of 8 ProResHQ 720p50 and 8 streams if 24/48 audio. Pretty impressive.

We found out that trying to bypass the cards caching will let the discs use it´s internal caching.
So what we did was setting up the ATTO R380 card like this. Everything as recommended in the manual except;

SpeedRead = adaptive (SpeedRead specifies the cache policy to be used during read operations. Once a read command is given, the ExpressSAS RAID storage controller retrieves the next set of sequential data from the RAID group and caches it in internal memory. If you select Never, read caching is never performed. If you select Always, read caching is always performed. If you select Adaptive, the default, SpeedRead is enabled or disabled depending on the sequential patterns detected in I/O requests.)

Prefetch = 0. (Prefetch specifies the number of stripes that are read when SpeedRead is enabled or adaptive. The valid values for Prefetch are 0, 1, 2, 3, 4, 5 and 6, and the default value is 1. This property can only be changed after the RAID group is created. To access this property, select the RAID group and view its properties.)

Hope this help
/Christer

[Christer Molander]

Just to sum up. On a project that we mastered out before summer and was aired a week ago. 4 streams of ProRes HQ. It´s not about anything else than the discs seeking time.
We divided the material evenly over two RAID groups on the same DX800RAID and voliá – everything worked like a charm.
So my saying is; it´s not the CPUs, It´s not the throughput, it´s the discs seeking time. And that´s why AVID is working. They know about this from ProTools were you can work with many tens of streams that are cached up before playing out.
Well that´s my guess anyway.

[Christer Molander]

This is were I think it is heading. Our studios have 4 NLE´s. All of these has been FCP7. Now two are Avid MC5.5. When the editors have learned MC it´s hard to see why switch to FCPX.

[Christer Molander]

Snow Leopard. And nothing changed with the very recent update to 10.6.7.
Even though it was said to have e few fixes in QuickTime.

[Christer Molander]

Got maybe an answer this afternoon. It can be the ATTO-card. The supportpeople agreed that the card shows too much troughput fluctuations (see image) to handle the four ProResHQ-streams. They will send me a new one to test.
Hope this works. Maybe I can get to work without hickkups and annoying “lost frames”-messages.

[Christer Molander]

Jesper is dead-on right.
Now when we are in 7.0.2 the same problems in multicam-cam editing are still there. FCP crashes, hangs and won´t do what I tell it to do.
I´ve tried two different approaches to my latest project.
Four stream with the original files, XDCAM HD 422 – not converted, all 4 streams the same.
The other one is with same material converted to ProRes422 HQ.
Tried on MP late 2009, 6GB RAM, Aja IoHD and Sonnet Fusion DX800RAID connected with a ATTO R380 2-channel SAS, Aja System Test measured to about 620MB/s.

FCP is barely getting me thru. It hangs, behaves strange and does not at all works as it did with DV-material a few years back.

Me also thinks fishy code somewhele….