EXS 24 Logic Sampler Users Group

Pev  wrote: 
>
> > (although noise issues have been seen to). 
> They have? This is something there is no mention of in the specs. I would
> assume it is no quieter until I hear otherwise from someone thats actually
> received one.



Probably wise but there was specific mention of it from Apple -- a user of the
current "leaf blower/wind tunnel" models was saying "What about us" on the
forums.

>
> > If this is so it would appear that the revised G4 with the bandwidth to
> > allow the PC 2700 DDR to earn its keep is still not available. It looks 
> > like the L3 cache is all that is happening to speed memory access -- not 
> > really an adequate solution for DAWs where you are running lots of 
> > plugs simultaneously. 




>
> Mmm. I'd disagree. As I mentioned, most memory bandwidth is used either
> GFX>mem or disk>mem.  IIRC the CPU->Mem bandwidth in the MDD machines is
> approximately 150MB/s. To put that in context, a 24bit 96K stereo audio
> stream uses approx 0.56MB/s of bandwidth. Being pessimistic and assuming that
> 50% of the cpu->mem bandwidth is used up by GFX and IO device access, I dont
> think the lack of *extra* bandwidth is a serious constraint for most DAW
> users.


A lot of plugins are going to use some memory space for their processing of the
audio stream. Reverb particularly uses a lot of memory space and a lot or
memory I/O. So if you have on one channel a soft synth/sampler playing
streaming audio files from HD,  a compressor and some other processing like
modulation and a reverb you could easily have a lot of traffic to and from
memory.

You seem to be approaching this from first principles which is a good policy in
my view. However, if you look at what is happening with x86 hardware, most
forthcoming chipsets (Intel/VIA/SIS)  are doubling bandwidth to memory (dual
channel DDR) and are shooting for 6.4 Gb/sec (two channels of DDR 400) and even
with these designs there is some commentry and analysis which is saying that
when the P4 gets up to 3GHz RDRAM is really the best solution to keep processor
performance scaling with clockspeed. 

Fast CPUs spend a lot of time spinning their wheels waiting for the next
instruction or a data read/write to finish. To state the obvious -- the faster
these tasks are dealt with the more performance you get out of your system. The
G4 revision which offers comparable CPU to RAM bandwidth to other desktop
processors has apparently been held up. If it were not -- it would be in these
machines and Apple would be trumpetting the vastly improved memory bandwidth. I
think we should wait and read some informed commentry on these issues on the
Mac hardware sites -- what CPU revision etc. 

Regards,
M