Adventures in 8-core and Magma chassis land

[Top Jimmy]

I want to report on our experiences in getting a new 8-core up and running with an HD-5 in the Magma PE6R4 chassis.

Mac Pro 3.0 GHz Xeon 8-core with 4 GB memory, OS 10.4.10, PT 7.3.1cs4, 1 HD core and 4 HD Accel cards.

When I say in here that PT no worky, I mean I kept getting consistent and repeatable DAE errors with up to four different error numbers. At no time did Digitest give me any errors. Digitest always said things were looking good.

The video card is reported as 16x. The expansion slot utility kept saying that the host card for the expansion chassis needed to be running at 8x, even though it is documented as being a 4x card. Host card in slot 2 at 4x, PT no worky. Host card in slot 3 or 4 at either 8x or 4x PT mostly no worky. The only place where I could get any success was at 8x in slot 4, with the HD core card farthest from the expansion chassis power supply. But, as I added the Accel cards one at a time, the system again no worky once more than two cards were in the chassis.

The final configuration where I got complete and reliable success was with the host card in slot 2, the video card reduced to 8x with the expansion slot cofiguration at 4-4-8-8, and the HD cards are arranged in the chassis with the core card farthest from the power supply in slot 4 and the Accel cards in slots 5-8.

My thanks to Koji Egawa for the time he spent on the phone with me.

[priorytools]

Hello. I found starting with 10.4.9 & PT 7.3.1cs1 I had no problems. . .

I ran 48 tracks at 96k, record & pb no problems. . I then put in cs3 . . Still no problems. .

If you look on the digi support section under the magma PE6R4 there are some guidelines there as to the versions that are fully tested. .

http://www.digidesign.com/index.cfm?...5&itemid=24876


I have stayed with 10.4.9 & cs3.

Maybe you've spent enough time on this, but I have a sneaking suspicion that your OS & cs are your problem. .

Am not at the studio now so can't varify the slot uitil settings I'm on. . I could look see on Monday if you like. .

HHmmm. . . Interesting though. .

[rockridge98]

Question: does the Magma chassis do better on RTAS latency than the HD expansion chassis?
thanks!

[Top Jimmy]

prioritytools,

Thanks for the pointers. The Mac came with 10.4.10 pre-installed and I wasn't interested in rolling back for the only known issue of the prebinding thing since that only occurs if one uses automatic update. We also had previously been using cs4 on the HD3 G5 that this system now replaces so I wasn't concerned about it.

I found success with the system in a configuration not recommended by Digidesign. I think that it's now obvious that the video card at 16x was the major culprit. Having the host card in any slot at either 4x or 8x was not ultimately successful as long as the video card was at 16x. I'm not going to test any other configurations now that the system is reliably working. We've been pushing it hard since I got it up and running and there has been nary a hickup.

[Top Jimmy]

rockridge98,

As I understand it, the HD Expansion has it's own onboard CPU to handle bus traffic. Because of this, track counts are guaranteed, but there is a hit on latency.

The Magma does not have a CPU to handle bus traffic, so there is no additional latency. The downside of it is you do take a hit in track count.

I was told by someone I trust that music mixers prefer the Magma because there's no additional latency while film mixers prefer the HD Expansion because there's no loss in track count.

[rockridge98]

rockridge98,

As I understand it, the HD Expansion has it's own onboard CPU to handle bus traffic. Because of this, track counts are guaranteed, but there is a hit on latency.

The Magma does not have a CPU to handle bus traffic, so there is no additional latency. The downside of it is you do take a hit in track count.

I was told by someone I trust that music mixers prefer the Magma because there's no additional latency while film mixers prefer the HD Expansion because there's no loss in track count.

That is very good news, thanks!

[jeff markham]

It strikes me as a bit odd that we don't have a new generation
of silicon as a solution to this problem (rather than having to
add more of the same in novel ways).

jeff

[bad jitter]

film mixers prefer the HD Expansion because there's no loss in track count.

Sounds strange. Any links / references? I'm using Magma PER-64 and don't see any loss in track count.

BTW, here's my configuration, which works fine:

Slot 1: ATI Radeon X1900 XT
Slot 2: NVIDIA GeForce 7300 GT 256MB
Slot 3: empty
Slot 4: Magma host card

Expancion slot utility set at x8-x8-x1-x8. 10.4.9, 7.3.1cs4 and 4 GB RAM. 1 HD Core, 1 HD Accell and 2 HD Process cards in the chassis.

[Top Jimmy]

No link. The anecdotal info passed to me by the local Digi rep was that it could be in the neighborhood of a 5% loss of the voiced track count.

Don't know about you, but we hardly ever flirt with even half of the voiced tracks our system is supposed to be capable of.

[Top Jimmy]

Update: It turns out that the expansion chassis track limitation is likely a thing of the past. Digidesign has not made any statements specifically addressing the old limitation of the 32 bit expansion chassis and whether or not it has evaporated with the 64 bit Magma chassis available now. My info was based on the published track count limitation guide. It turns out the 5% I quoted off the top of my head was way off. The loss was more like 40%. It was an assumption on my part that the CPU in Digi's Expansion HD had something to do with maximizing the track count and that going to a chassis without one would mean a track count hit. Being that we're now at 64-bit, that might well be a figment of my imagination.