G5 with Mac Mini Slave Audio Drop Out

[kevin_9944]

We have two mirrored rigs running Pro Tools on the master and VEPro on the slaves. The rigs have minor, but perhaps significant differences. One (the PCIx rig with the mid 2007 mac mini slave), I get audio drop outs. The other, a PCIe rig plays back fine. Any pointers as to who or what could be the culprit? I do notice that the cpu % meter in the bottom right hand corner of the VEPro screen will dive to zero when I get the audio drop out. The drop out is only momentary.  Happening randomly. I can lay down a one minute mix with three drop outs and the same pass again straight away will perform with no drop outs.

rig 1

master: g5 2.3dp PCIe

slave: mac mini (early 2009 2ghz intel core 2 duo)

rig 2

master: g5 2.3dp PCIx 

slave: mac mini (mid 2007 2ghz intel core 2 duo)

Both masters have

hd3

OS X 10.5.8

PT 7.4.2cs4

Both slaves have

VEPro 4.1.7043 (downloading 4.1.7384 but just delivering on a gig so not in a rush to install)

VEPro plug in, set to "two buffers"

VEPro (32bit) server: 

Multiprocessing - 2 threads

Audio ports - 32

Midi ports - 16

Both rigs are connected via their own Netgear GS605 gigabit switch.

The obvious difference between the two set ups is that the g5 PCIx doesn't do jumbo frames and the g5 PCIe has two gigabit ethernet ports.  

Pro Tools Playback engine

H/W buffer size 512, 1024 or 2048 (tried them all)

Cache Size - Large

Any thing that would point me in a new area of investigation would be much appreciated.

[Cyril Blanc]

Why do you have to rigs ?

Why do you use two Giga switch ?

Have you try :

rig 1

master: g5 2.3dp PCIe

slave: mac mini (mid 2007 2ghz intel core 2 duo)

rig 2

master: g5 2.3dp PCIx 

slave: mac mini (early 2009 2ghz intel core 2 duo)

[kevin_9944]

@Another User said:

Have you try :

rig 1

master: g5 2.3dp PCIe

slave: mac mini (mid 2007 2ghz intel core 2 duo)

rig 2

master: g5 2.3dp PCIx 

slave: mac mini (early 2009 2ghz intel core 2 duo)

Nice suggestion. I will try it over the coming days, when the current delivery schedule permits and report back.

Many thanks. 

[civilization 3]

have you tried medium cache size?

I'm assuming you are using a fixed IP and not dynamic. The latter situation will cause significant performance loss.

I think jumbo frames improved my performance a bit, so I guess that may be somewhat of a factor. anyway, I think your bottleneck is probably the network and not the difference in PCI cards.

[kevin_9944]

@civilization 3 said:

have you tried medium cache size?

Do you mean in Pro Tools? I've tried Normal and Large (Improves Performance). Is there a VEPro cache I've overlooked? Would be great if I had overlooked something rudimentary and fundamental! 

We will swap Mac Minis in the coming days to see if the jumbo frames and network are the true bottle neck.

Thanks for the input.

[civilization 3]

in Pro Tools... 'normal' I mean. I have no theory for it, just wondering, 'large' seems the intuitive move but who knows what happens. 

I had dropouts with VE Pro with dynamic IP kind of bad at times until I got a clue.

is the PCI-x a northbridge slot?

[kevin_9944]

Normal is our "default". When you posted, I was laying down something, with a dropout every other pass. So I switched to "Large (Improves Performance)"... but it didn't improve performance. So I switched back.

Interesting that the machine still stuck on "dynamic" is the one not having the issues. 

Regarding north bridge, a term I was unfamiliar with, I couldn't find any documentation in a search, that said one way or the other in regard to the PCI-x G5. 

Again, thanks for the thoughts.

[civilization 3]

the north bridge is right by the CPU in the logic board. I imagine any modern mac places the thing there.

In fact I couldn't tell you the difference between PCIx and PCIe, but I know I have a PCIe and the slot is north bridge. South bridge = less performance in any case. Before PCIx it was typical for a PCI slot to be on the south bridge. I doubt this is the issue, I'm just trying to be thorough

I think the VE Pro manual tells us to strictly use fixed IP. I know that's what they'll say here. I wouldn't take the situation that the machine that uses dynamic isn't the one having the bad issues without a grain of salt, that is if they are on the same network [My thinking is, it's the network that's suffering.]. 

I had issues, dropouts and stutters, and waiting for the system to catch up - for like a whole minute, beach ball o' doom - that just mystified me, given my setup, when I was using dynamic IP. At this point VE Pro's performance is just excellent here.

[kevin_9944]

In my short search I couldn't determine if the PCI was connected to north bridge on the PCI-x G5. I've assumed that PCIx is older than PCIe therefore likely to be my bottleneck. Other people seem to have had very good success with their G5s; PCIx and PCIe, so I'm going to continue on the understanding that it's an error in my set up, not a shortcoming of my hardware, unless I can determine otherwise.

The reason for each pair of macs having their own gigabit switch is so that I can easily isolate them from the LAN by yanking the physical cable out of the back, disconnecting completely. If I do physically isolate the pair with issues, I still have the problem, so I can say, in my case, that as the PCIe G5 pair are not physically connected via the LAN, they are not a contributing factor. But, as you rightly remind me, I am intending to switch the properly performing G5 pair to static IPs, once this gig is delivered.

Thanks.