audio people NEED the PPC970, without question!
Audio is a world where its often can do or can't do. Not can do or can do 20 seconds later.
I agree there, I think it was the guy doing the audio column on Xlr8yourmac a while ago that decribed audio vs graphics needs when it comes to cpu power.
"Audio performance is always black & white, graphics is always shades of grey" or words to that effect.
Just for comparison, and this is a little biased given the different CPUs involved but here goes :
Photoshop (from barefeats)
P4, 2Ghz : Wait 90 seconds (SP action file)
P4, 2Ghz : Wait 73 seconds (MP action file)
G4, 2 x 1Ghz : Wait 101 seconds (SP action file)
G4, 2 x 1Ghz : Wait 48 seconds (MP action file)
Protools LE, dave c test (from DUC)
All results are realtime performance, extension tweaks and custom RAM allocation is used on the macs and various OS settings are used on the PCs aswell for best performance. The test is not MP aware so only 1 cpu is being used for both Protools LE and the Mac OS. PT 6 under OS X promises a 10-15% improvement over PT 5 under OS 9 on an MP aware mac but even that isn't going to be enough from the results we've had so far :
G4, 2 x 800Mhz : 24 tracks + 8 aux (160 realtime plug-ins)
P4, 2.26 Northwood : 24 tracks + 20 aux (220 realtime plug-ins)
The reason the G4 tested is only a dual 800Mhz model and not a dual 1.25Ghz model for example is because the guy who tested it used a very well optimised extension set and audio drive arangement that hasn't been beaten by any of the newer models other people have run the test on. Also, the PC was tested with far lower buffer setting of 128bytes where as the G4 was using a 1024byte buffer. This is kind of like a cache used to buffer the plug-ins and audio tracks so they don't use as much cpu load.
This proves without any doubt at all that the Northwood pentium 4 with it's 512K L2, 133Mhz bus (Quad pumped) and RDRAM smokes the G4 with it's far lower clockspeed, 256K L2, 133Mhz bus and SDRAM.
The dual 1Ghz DDR macs have already being tested and don't perform any better than the older QS models so the extra benefit of the 167Mhz bus vs the 133Mhz bus on the older models is obviously cancelled out by the smaller L3 cache.
The Dave C test is this (just incase you're interested) :
Sample rate = 44100
Hardware Buffers = 1024
Bitdepth = 16bits
1) Open new PT session
2) Create a new mono audio track
3) Add PT stock effects Compressor, 4B-EQ, Slap delay, medium delay, long delay.
4) Select the newly created track from step 2 above.
5) Duplicate the track.
6) Record enable the track.
7) Repeat step 5 until your CPU goes into the red.
8) If your CPU is in the RED delete the last track created. IF it's still in the red, delete another track until the CPU is back in the green. Click on the CPU meter after every deletion to reset the meter.
9) Hit RECORD, and record for 60 sec. Repeat steps 8 and 9 until you can record 60 secs. without cpu meter peaking.
*** N.B. If you can achieve the full 24 tracks with record enabled and can record for 60 secs. start adding Aux tracks with the same plug-ins and keep going until you hit RED.
)
