EXS 24 Logic Sampler Users Group

Hey guys,

An old hardware sampler geek here with not yet enough experience in 
software sampler area, so my apologies if the answer to this is obvious.

I'm on the market for an HD to be used exclusively for streaming 
samples in EXS and Kontakt and I'm not sure if the drive with 8 MB 
cache will improve the performance in any way. Can anyone shead some 
light on this? Many thanks in advance.

Best,

Andy

DIGITALMechanics@... wrote: 

>
> I'm on the market for an HD to be used exclusively for streaming 
> samples in EXS and Kontakt and I'm not sure if the drive with 8 MB 
> cache will improve the performance in any way. Can anyone shead some 
> light on this? 


When drives with 2MB of cache are run head to head with  otherwise similar 8MB
cache drives some significant differences are found on some benchmarks. How
well these will translate to EXS24 streaming performance is hard to say --  you
really need to test the drives with the application you have in mind. 

The firmware in the drive will determine what caching strategies are used and
if the right data is cached ahead of a read request (a cache hit) the data will
be transimitted to RAM or the CPU much more quickly than if it has to be read
off the media.

I gather that the caching approaches taken by some drives (eg Seagate's new 10K
rpm SATA Raptor ) are quite different from most enterprise (database) optimised
drives so the performance delta between drives would probably vary from brand
to brand and model to model. 

In the case where you are streaming up a lot of long audio files (eg recorded
audio tracks or sustained piano notes -- think the chord at the end of A day in
the Life) I suspect that the difference would not be great. In the case of
smaller files (eg short percussion samples sitting close together on the drive)
it might make a larger difference. Personally I went with an 8MB drive.

Go to storagereview.com and read some drive reviews if you want more info on
this.

Regards,
Murray

On Monday, August 25, 2003, at 11:30  AM, Murray McDowall wrote:

> The firmware in the drive will determine what caching strategies are 
> used and
> if the right data is cached ahead of a read request (a cache hit) the 
> data will
> be transimitted to RAM or the CPU much more quickly than if it has to 
> be read
> off the media.

But how can it guess what note is going to play next? I seriously doubt 
there is any kind of look ahead function in Logic, EXS or Kontakt, and 
AFAIU without such a function the overall difference would be minimal. 
Am I missing something here?
>
> I gather that the caching approaches taken by some drives (eg 
> Seagate's new 10K
> rpm SATA Raptor ) are quite different from most enterprise (database) 
> optimised
> drives so the performance delta between drives would probably vary 
> from brand
> to brand and model to model.

I'm only talking Seagate here as I'm not considering any other 
manufacturer at this time.
>
> In the case where you are streaming up a lot of long audio files (eg 
> recorded
> audio tracks or sustained piano notes -- think the chord at the end of 
> A day in
> the Life) I suspect that the difference would not be great. In the 
> case of
> smaller files (eg short percussion samples sitting close together on 
> the drive)
> it might make a larger difference. Personally I went with an 8MB drive.

Seems like not that much a difference with piano and orchestral stuff...
>
> Go to storagereview.com and read some drive reviews if you want more 
> info on
> this.

Yeah, thanks for this one as well.

Best,

Andy

DIGITALMechanics@... wrote: 

>
> But how can it guess what note is going to play next? I seriously doubt 
> there is any kind of look ahead function in Logic, EXS or Kontakt, and 
> AFAIU without such a function the overall difference would be minimal. 
> Am I missing something here?


As I'm sure you understand, the audio data for your samples is stored on the
media on linear tracks. The caching is controlled not by the application
(EXS24) but by the drive firmware --that is software stored on ROM on the drive
-- it is the program that runs on the drive's internal microprocessor--
including the cache controller. The caching controller's strategy might be to
read the whole track into cache rather than just the bit that is currently
being requested. If you need the data in subsequent sectors on that track a
short time later they might still be in the cache. 

In my long piano notes example (holding down a chord with the sutain pedal so
that you have to read a bunch of 30+  sec long piano samples off your sample
drive) without caching,  the read head would chase around the media reading a
succession of small segments of those long notes. The result might mean that
the heads return many times to the same file to read its data to produce one
continuous tone. If caching is employed some or even all of the subsequent
returns to that note might be avoided -- the whole note might have been read
into cache on the first partial access. In that case all subsequent requests
for that note's data will be relayed to the EXS24 direct from the cache.
Perhaps 8MB is too small to really make a big difference here -- 128 MB might
really accelerate this sort of application dramatically.

Unless you work for a drive manufacturer and are privy to their testing and
cache controller programming methods this is all guesswork  and the only thing
you can do is test the finished products youself of find a relevant published
review.

Regards,
Murray

> I asked:
>>
>> But how can it guess what note is going to play next? I seriously 
>> doubt
>> there is any kind of look ahead function in Logic, EXS or Kontakt, and
>> AFAIU without such a function the overall difference would be minimal.
>> Am I missing something here?

On Tuesday, August 26, 2003, at 03:38  AM, Murray McDowall explained:
>
>
> As I'm sure you understand, the audio data for your samples is stored 
> on the
> media on linear tracks. The caching is controlled not by the 
> application
> (EXS24) but by the drive firmware --that is software stored on ROM on 
> the drive
> -- it is the program that runs on the drive's internal microprocessor--
> including the cache controller. The caching controller's strategy 
> might be to
> read the whole track into cache rather than just the bit that is 
> currently
> being requested. If you need the data in subsequent sectors on that 
> track a
> short time later they might still be in the cache.
>
> In my long piano notes example (holding down a chord with the sutain 
> pedal so
> that you have to read a bunch of 30+  sec long piano samples off your 
> sample
> drive) without caching,  the read head would chase around the media 
> reading a
> succession of small segments of those long notes. The result might 
> mean that
> the heads return many times to the same file to read its data to 
> produce one
> continuous tone. If caching is employed some or even all of the 
> subsequent
> returns to that note might be avoided -- the whole note might have 
> been read
> into cache on the first partial access. In that case all subsequent 
> requests
> for that note's data will be relayed to the EXS24 direct from the 
> cache.
> Perhaps 8MB is too small to really make a big difference here -- 128 
> MB might
> really accelerate this sort of application dramatically.

OK, now I see. Mucho thanks, Murray.

Best,

Andy
>