1st SAC live, FOH 5 stars, Monitorworld 2.5

[Bob L]

Well... I will definitely experiment with the single chan all monitor mix view as well as a few others I have had planned from the beginning.

Sounds like you pushed your box just a little over the limit.

I would generally feel better about the computer sitting comfortably in the 40-60% load range when handling all the mixes and processing you require.

A duo Core 2 or beyond may be the ticket for you if you are going to push that hard.

Bob L

[Trackzilla]

Well, as I mentioned, this week is planned as 'migration week'. I had planned on re-allocating a pentium D945 based system I have here, but I have an opportunity for a 1U Tyan Tank GT14 with twin xeon 5310 1.6Ghz quad-core processors & a trunkfulla-ram (8gb expandable to 24gb)...are there any multicore optimization or utilization issues or the like that come to mind, or other known issues I should consider before I jump on this puppy in an attempt to permanently silence the question of 'is it enough computer to do that?' For example, are fewer cores & a higher speed preferable for SAC?

[Bob L]

Dual Cores have been tested pretty well... they can be tricky... quad cores could open other issues... more cores introduces much more complexity.

More cores is not always synonymous with more performance... and can most definitely be synonymous with more instability.

You have to experiment.

I have demonstrated less MT load in SAC when forced to a single cpu on the same dual core machine... sometimes the effects of context switching between multiple cores can actually slow down performance... and definitely introduce stability issues.

But... the current SAC engine seems extremely stable on 2 cores.... not sure what will happen on 4.

Bob L

[Trackzilla]

ok, that was my primary question, except this machine would be 8 cores unless my math is wrong, and I assume whatever issues might exist would doubtless be compounded even more effectively. So this box would likely crunch through an army of little tasks, but might well plod along or confuse itself irretrievably with one big complex one eh?
And I'm assuming that the tricks used for hyperthreading cpus to make them act as a single core won't work on true multicore units and your best hope if/when they are unhappy is to hope you can create harmony by setting affinity to certain cores...but that is also partially dependent on how your drivers interact as well...am I basically correct?

Oh, and BTW, thanks for following this thread through and giving serious consideration to the case presented for an alternate monitor view, you are amazing both in the exquisite quality of work you do AND the attention you pay to issues raised by your community of users!

[Bob L]

Hyperthreading is a complete mess, in my experience... plays havoc with soundcard driver threads.

If you have a hyperthreaded cpu, I highly recommend you set the box to one cpu as mentioned in my Windows Tweaks document found in the misc downloads at the www.SAWStudio.com site.

Bob L

[Naturally Digital]

Dual Cores have been tested pretty well... they can be tricky... quad cores could open other issues... more cores introduces much more complexity.

More cores is not always synonymous with more performance... and can most definitely be synonymous with more instability.

I have demonstrated less MT load in SAC when forced to a single cpu on the same dual core machine... sometimes the effects of context switching between multiple cores can actually slow down performance... and definitely introduce stability issues.

Bob, I know we've discussed this before (as regards SAW)... and I'm not wanting to re-hash the conversation but...

Do you think it might someday be possible to run the processing (plugins and channel strip processing) for each SAC console layer (FOH, Mon 1, Mon2 etc) on seperate cores?

[Bob L]

Possible... but splitting the engine in that manner opens an amazing amount of possible instability and trouble.

As the engine continues to mature... perhaps this will be part of the design.

At the moment, my primary concern is stability so people can start to become comfortable with the SAC idea and feel safe that it will not down their show performance with random crashes.

Bob L

[Naturally Digital]

Well, as I mentioned, this week is planned as 'migration week'. I had planned on re-allocating a pentium D945 based system I have here, but I have an opportunity for a 1U Tyan Tank GT14 with twin xeon 5310 1.6Ghz quad-core processors & a trunkfulla-ram (8gb expandable to 24gb)...are there any multicore optimization or utilization issues or the like that come to mind, or other known issues I should consider before I jump on this puppy in an attempt to permanently silence the question of 'is it enough computer to do that?' For example, are fewer cores & a higher speed preferable for SAC?

Based on personal experience with my quad-core 2.4Ghz (Q6600) I think this should be very stable.

My concern (and maybe someone else will confirm or not) is that the absolute performance (number of plugins, number of channel dynamics, eq etc) might be limited by the Ghz of the processor and since the main processing loads up one core only, a 1.6Ghz proc (no matter how many cores, dual cpu etc) might not give as much horsepower as a cpu with higher clock speeds (even a P4 3.4 or 3.6Ghz).

The multicore cpu machines definitely seem quite stable to me.

I'm in the process of putting together a (old - :rolleyes dual 3.4Ghz xeon (not multicore) machine so I can compare the performance to my quad-core machine.

Benefits of multicore cpus definitely become apparent when running SAW and SAC together on the same machine and forcing their processes to separate cores using the affinity tool in Taskmanager.

[Naturally Digital]

At the moment, my primary concern is stability so people can start to become comfortable with the SAC idea and feel safe that it will not down their show performance with random crashes.

Good to hear.

[DominicPerry]

If you want the fastest Intel chip to run as a single core, you are better off choosing an E8500 (or E8400) Dual Core (over a Quad 6600 or the like) as they are not just faster in raw clock speed but the Front Side Bus is faster too.

Dominic

[Trackzilla]

Yeah, these comments are all making sense to me, and confirming my gut feelings about this prospective box. It would likely bottleneck at the clock speed of the core that ended up thinking about SAC. The things that attracted me most about it were the FSB speeds and amount/speed of the RAM. I do plan to grow to running SAC and SAW to offer live remote recording as I have historically done with physical splitters, interfaces, and Vegas. Based on that it seems that I may be best served with two processors with as high a clock speed each as I can scrape together. If I can get the 'lil mrs' to put up with a bit of reorganizing which boxes do what around here for a bit I would have access to several different architectures to experiment with before any additional outlay, but she is NOT fond of disruption & change, so that'll be it's own battle.
Perhaps I can present it as an attempt to allocate a better box for her video processing needs and quietly find out what I need to know at the same time

Regarding your last comment about stability being the primary concern Bob, that is exactly why I took the known marginal box I did to the show I took it to. It had tested as basically stable, but too consumptive of it's limited resources, and I wanted to put SAC in a scenario where I had conventional redundancy but a marginal box and see what it took to force it to fly. I took our Yamaha D1000 as I could preset a default recovery scene and simply swap the ADAT ins from the MOTU to the Yammy if needed...it wasn't. Now I can feel confident about recognizing if a more capable box begins to display resource difficulties in a more high pressure situation and what steps to take to quickly resolve them. It DID run out of resources, it coughed & sputtered a few times, but I was intentionally abusing the concept in a semi-controlled environment. More to the point, it RAN that way, once I increased buffering to the point it could keep up it remained stable, just a bit slow to refresh the GUI. The only time it actually locked up was when I tried to add a second VST reverb during the first band changeover and it took all of maybe 25 seconds to fix with a quick restart of SAC and a reload of the saved stuff from just before that add...WOW it is great that your apps load so fast! The festival in question has been a 'test bed' for new technology for several years and they actually enjoy that concept as it is a tiny town with an unnaturally high concentration of talent & techs. 'Where's the console?....that's amazing...sounds fantastic!' was the topic of the weekend.

[gregangst]

I too had great success with SAC this weekend. My setup used SAC for FOH only, however in the near future the console will be gone and SAC will do all of the honors. This show was a three hour event, live drums, bass, guitars, keyboards, vocals, and tracks.

During the whole show Sawstudio was linked to SAC. Tracks played perfectly out of Sawstudio into SAC.

My setup was my LAPTOP dual core 2.4 GHz Centrino with 2 gig of ram,
one RME FIREFACE 800, and two RME Octamic modules. We used firewire 400 to the Laptop. We used a second to play MP3 files thru the RME SPDIF input.

CPU load showed only about 32 % thru the whole show. I had 4 monitor consoles source linked at the same time.

This is very exciting.

[Warren]

I too had great success with SAC this weekend. My setup used SAC for FOH only, however in the near future the console will be gone and SAC will do all of the honors. This show was a three hour event, live drums, bass, guitars, keyboards, vocals, and tracks.

During the whole show Sawstudio was linked to SAC. Tracks played perfectly out of Sawstudio into SAC.

My setup was my LAPTOP dual core 2.4 GHz Centrino with 2 gig of ram,
one RME FIREFACE 800, and two RME Octamic modules. We used firewire 400 to the Laptop. We used a second to play MP3 files thru the RME SPDIF input.

CPU load showed only about 32 % thru the whole show. I had 4 monitor consoles source linked at the same time.

This is very exciting.

Great news!

What effects where you using? Comp, EQ, Reverbs
I have found some to be fine at first then cpu cycles will runaway up to 100% then crash the machine. Doesn't sound as you where using any of those.

[Bob L]

Be careful with certain DX or VST plugs that are prone to the dreaded floating point silence problem... these cannot be used.

I recommend using the built-in console processing for most chan basic adjustments.

You should test each plugin that you want to use ahead of time and remove any that cause latency or will cause the floating point runanway during silence problem.

Bob L

[Warren]

Be careful with certain DX or VST plugs that are prone to the dreaded floating point silence problem... these cannot be used.

I recommend using the built-in console processing for most chan basic adjustments.

You should test each plugin that you want to use ahead of time and remove any that cause latency or will cause the floating point runanway during silence problem.

Bob L

I've had a lot of dealings with them (pourly written plugs), since I have quite a few. Sad thing is I like some. Native is surely the way to go to keep things stable. It would be nice to see more vendors write native SAW plugs. But with the native plugs SAC comes with I can do most things I ever would need too do, in a Live Sound situation.

Thanks

Warren (Chris)

[gregangst]

I was using Bob's built-in effects, reverbater, frequency analyzer on the output. I was using two reverbs, along with most of the built-in eq's on each channel. I also used compressors when needed.

The system was very stable. I have had other occurances where we had problems. Now that the RME's are word locked, everything worked great, no pops, no buffer slips,

It just plain works and works very well.
Great product Bob. I have been using Bob's products since the early 1990's and have been bery satisfied with how well they perform and keep on performing.