DP performance on Mac Pro??

blastula · Post by **blastula** » Wed Dec 06, 2006 7:56 am

Any reports?

Supposedly Logic users are getting enormous plugin counts on the Intel macs.

I'm not hearing anything from DP Intel users. I know you're out there.

Rob

Phil O · Post by **Phil O** » Wed Dec 06, 2006 8:03 am

Can't seem to to choke it with realistic track counts and plugins for audio only. When I first got the machine I pushed it hard just for giggles. I forget the numbers but it was hundreds of tracks.
VI's seems to be where the bottle neck is.

Phil

blastula · Post by **blastula** » Wed Dec 06, 2006 10:07 am

any further elaboration would be appreciated.

i'm currently using a 2 x 2g G5 and salivating over logic users' plugin counts on intel macs.

so, this "VI bottleneck"....has there been any substantial improvement at all with the mac pro? can anyone be a little more specific?

I'm gonna have to take this lack of response and enthusiasm as a definite deterrent.

which is sad, cause folks on other platforms are jumping for joy.

rob

Phil O · Post by **Phil O** » Wed Dec 06, 2006 9:51 pm

Someone had posted a project as a benchmark test a while back. I think it was Aramis (but don't quote me on that). It was a simple sequence using one of the stock DP VI's. I think I got something-teen instances of that VI to run.

The board went down for a few days and I lost track of that thread. You might want to send Aramis a message and ask if it was him and what the results of the survey were.

Sorry I couldn't be more help.

As far as audio plug-ins is concerned - WOW! I did a little experiment for you:

Recorded 30 seconds of mono audio. (24 bit, 44.1KHz)
Placed on that track - 1 MW EQ, 1 MW compressor, and 1 e-verb.
Duplicated track multiple times.
Performance meter hits red on playback for the first time at 244 tracks. That's 732 plug-ins!

Phil

Frodo · Post by **Frodo** » Wed Dec 06, 2006 10:03 pm

Yes-- reports on Intels seem to exceed even Frodo's expectations to the point where the hobbit may be getting an Intel sooner than later-- but not before January announcements from Apple are made. I was going to wait until after Leopard and 64-bit app data were released in April, but I'm now convinced that it may be many months and counter-productive to wait until apps go 64-bit. Intels under Tiger are boasting amazing performance boosts above and beyond expectations.

If you are in the market for a new computer now-- go Intel without hesitation.

matwell · Post by **matwell** » Thu Dec 07, 2006 9:26 am

Phil O wrote:...Recorded 30 seconds of mono audio. (24 bit, 44.1KHz)
Placed on that track - 1 MW EQ, 1 MW compressor, and 1 e-verb.
Duplicated track multiple times.
Performance meter hits red on playback for the first time at 244 tracks. That's 732 plug-ins!

Phil

This is a common mistake when testing system performance. You need to record *multiple* tracks of *new audio*, and not just duplicate one track. The disk access is very different, since the disk has to seek multiple files simultaneously.

If you have a multi-input I/O, then I would record a few passes of the maximum inputs that you have. Maybe 32-48 tracks worth, *generating new audio files*. That will be a much better indicator of system performance.

Jaysplace101 · Post by **Jaysplace101** » Thu Dec 07, 2006 9:44 am

This is a great point Matwell. Also, VI's along with audio is the real issue for me, and there aren't many reports out on this yet. Looking forward to hearing from some users.

j

Phil O · Post by **Phil O** » Thu Dec 07, 2006 9:46 pm

matwell wrote:This is a common mistake when testing system performance. You need to record *multiple* tracks of *new audio*, and not just duplicate one track. The disk access is very different, since the disk has to seek multiple files simultaneously.

If you have a multi-input I/O, then I would record a few passes of the maximum inputs that you have. Maybe 32-48 tracks worth, *generating new audio files*. That will be a much better indicator of system performance.

I see your point and it is well taken. I was just trying to do a down and dirty quickie test for the original poster to give him some idea of what the Mac Pro can do. I was trying to stress the CPU with plug-ins...that's all.

Phil

Frodo · Post by **Frodo** » Fri Dec 08, 2006 2:57 am

Keep in mind that there are several things to test. If you record multiple tracks, such as live concerts or a rhythm section "live" in the studio, then the multitrack audio record test is important. If you do mono or stereo vocal or instrumental recording, then the multitrack record test will not mean much to you-- except perhaps in the way of horsepower overhead.

Also, the multitrack record test has much to do with bus speed and bus bandwidth as data goes to the HD-- and that depends entirely on what HD is being used, what its transfer rate is, read/write time, etc. That's quite a different test than CPU alone. It's safe to say that the Intel's bus speeds overall are substantially improved over their PPC elders, and PCI-e offers what could arguably be the most important performance boost in this regard. FW 400 and 800 performance remain the same for good or ill.

The most realistic test is one that reflects *your* way of working, but always understand that everyone works differently.

As for VIs, the Intels offer some boosts in performance, but various sources confirm a marginal VI track count increase for the most intensive libraries like EWQLSO or VSL Cube. Computer networks remain the best solution for such things. This is because of the limitations of the software rather than the hardware. If what you use works well inside 2.5GB, then you're in good shape. The bigger orchestral VIs tend to take 8-10GB for granted, which is why 3-4 computers in a network remain unvoidable for substantial productions. This will be the case at least until 64-bit addressing is fully embraced by software developers.

I think it's important to keep realistic tabs on what performance boosts the Intels offer, and this thread serves a good purpose. Audio files and dynamics processor plugins are reaping the greatest benefits with track counts. Virtual instruments and convolution reverbs benefit as well, but to a lesser degree.

Phil O · Post by **Phil O** » Fri Dec 08, 2006 4:35 am

Thank you, Mr. Frodo. I don't really think I was being much help on the subject.

Phil

Frodo · Post by **Frodo** » Fri Dec 08, 2006 1:45 pm

Phil O wrote:Thank you, Mr. Frodo. I don't really think I was being much help on the subject.

Phil

Mister Phil:

Ya done good, sir! Actually, your test is just fine especially when one is comparing the same test to an older machine. It's a perfectly legit and important benchamark to note.

Once people sort just how it is they work or where their weakest links are in the recording process, it becomes much easier for them to design their tests for their specific needs.

For example, someone may be quite unhappy with HD recording performance on their G5 Quad, and wonder if they should buy an Intel. Since so much of HD recording has to do wth the HD itself, it could be that a gaggle of FW400 drives are simply choking themselves trying to keep up with data transfer amounts and rates. The solution here in this case would be NOT to buy an Intel as a first consideration, but maybe to take advantage of the G5 Quad's PCI-e slot with a 64-bit host of some sort. Getting faster drives in this case will also offer a very cost effective solution for lighting a fire under HD performance where FW drives' performance remains the same-- and they may compete for bandwidth should there be an FW audio interface involved.

Still, I think there is much to recommend for testing the effectiveness of recording a single mono track at the lowest possible buffer setting. If nothing else, it will offer a bottom line in advance of doing stress tests.

My only argument for not buying an Intel hinges on what software is available for it or not and which apps are working well (or not) according to spec.

blastula · Post by **blastula** » Fri Dec 08, 2006 5:50 pm

but maybe to take advantage of the G5 Quad's PCI-e slot with a 64-bit host of some sort

frodo,
what does this mean?

thanks,
rob

Frodo · Post by **Frodo** » Sat Dec 09, 2006 2:41 am

blastula wrote:but maybe to take advantage of the G5 Quad's PCI-e slot with a 64-bit host of some sort

frodo,
what does this mean?

thanks,
rob

PCIexpress offers performance advantages over PCIx slots. With a PCI-express host card (firewire or eSATA, for example), one could make use of this performance boost with their external hard drives. It's a much less expensive upgrade than buying a whole new Intel.

PCI-express is the new Mac PCI standard on Intels, but those who still have Quad PPCs and, I believe, 2.7 dual PPCs can use PCI-express cards. The wider bandwidth of PCI-express means smoother data throughput.

Phil O · Post by **Phil O** » Sun Dec 10, 2006 8:15 am

matwell wrote:This is a common mistake when testing system performance. You need to record *multiple* tracks of *new audio*, and not just duplicate one track. The disk access is very different, since the disk has to seek multiple files simultaneously.

If you have a multi-input I/O, then I would record a few passes of the maximum inputs that you have. Maybe 32-48 tracks worth, *generating new audio files*. That will be a much better indicator of system performance.

Done.

Test results:
132 tracks (90 seconds of audio/track, 132 audio files) with same three plug-ins as previous test. (396 plugins total)

Operating system and DP are on an internal SATA Samsung HD160JJ. Test project is on an internal SATA Western Digital WD2500JS.

The disk IO seems to be the limiting factor here (playback meter spikes at 133 tracks). CPU meter barely hits 1/2 FS and Activity Monitor show DP CPU usage at around 60%.

Phil

corbo-billy · Post by **corbo-billy** » Sun Dec 10, 2006 10:33 am

Hi _

That would have been to have the adjustments well as the Buffer in Configure Hardware Driver _