Hardware or Software sampling?

[hollowsun]

I have owned many hardware samplers ... i currently have a Yamaha A3000, a Roland S750, and a Boss SP505. They haven't been switched on once this year. I prefer to use Live6 and Sampler ... it's quicker, easier, better sounding and more flexible than the hardware. Cheaper too ...

Your prerogative. I don't think anyone's knocking that choice. I certainly wouldn't. Each to their own... I know which I prefer however!

Steve

[hollowsun]

The problem with hardware samplers is they weren't updated at all.

Yes - because people were buying (or getting cracked ) s/w samplers!!! Had the situation been different, we might be looking at a hardware sampler streaming samples off disk with 1Gb of RAM and so on. But the migration to s/w was inevitable unfortunately and 'hardware' became a dirty word!

I'd venture to suggest that much of the impetus for the switch to s/w samplers was not because they were (or are) necessarily better but the fact that they are cheaper and more convenient (a bit like using big supermarkets over the local high street shops - everything under one roof, cheaper and convenient parking ). S/w samplers are not necessarily easier to use than a h/w sampler either - in fact, often more complicated with the zillion options they offer that, arguably, few people use or understand and/or which can, arguably, cause a loss of focus in the creative process! But the purchasing decision was/is largely based on price and convenience (it's hard for a £1,400 h/w sampler to compete with a £300 s/w sampler .. or worse, a cracked s/w sampler for £0!). S/w samplers won out (right or wrong).

To drag this back on topic, however, the original poster wanted to know whether to get an S6000 or use software. In the end, I suggested both if he could afford it. Hardware die-hards should embrace the possibilities of software and software die-hards should not overlook the benefits and advantages hardware has to offer. The two can actually work together and there is no need for an 'either/or' mentality. They are just tools after all and both can be exploited for their respective benefits in the creative process. Both have advantages and disadvantages.


Steve

[Radek]

I'd like to have A5000 (it has some intriguing capabilities) but I doubt I'd have time to learn it or use. Akai Z4 is very cheap now and it comes even with nice stock sampleset on its hdd. However loading times are terrible... it's not fun waiting 20 minutes to get a piano (great even) loaded.

I don't know why those machines couldn't be improved in memory capacity and i/o perfomance? Those look like trivial issues...

[hollowsun]

However loading times are terrible... it's not fun waiting 20 minutes to get a piano (great even) loaded.

Eh? Dunno where that figure came from. Depending on size of course, just a few minutes - if that... just a few seconds to load ~10Mb

I don't know why those machines couldn't be improved in memory capacity and i/o perfomance? Those look like trivial issues...

They could have been improved .... had there been a commercial incentive (i.e. sales) to do so.


Steve

[snooky]

Hardware is way better than the virus!

I use my tubeamp-sampler all the time to restore dynamics in my VA synths!

[stale bread]

without a doubt I prefer a hybrid system, pair ableton Live with an mpc and you have a set up that becomes more fun the more you use it.

[ckett]

Well,

This has been an interesting discussion. I just bought an Akai S6000 and I must say that it is incredible! The interface design is absolutely brilliant. So easy to use and it sounds teriffic. The filters are glassy, bright and smooth.

I also own Kontakt 2. Kontakt is very good, probably my favorite s/w sampler. The interpolaltion isn't as good as the Akai. I did an interesting test on my PC concerning the interpolation issue.

With the interpolation set to "standard" and the latency set to 256mb buffer (6ms) I was able to get up to 200 stereo voices on my AMD X2 computer.

With the interpolation set to high, I was only able to get about 64 stereo voices. Strangely enough this is what the Akai S6000 offers in terms of stereo voicing with an equal quality interpolation.

So in essence, 1 dual core PC with Kontakt = 1 Akai S6000 in terms of interpolation quality and voice count.

I will be adding an internal HD to the S6000. This should make loading samples much faster. Kontakt is much better in this area, no contest.

So hope you enjoy these observations and thanks for the great discussion!

[kuniklo]

Well,

This has been an interesting discussion. I just bought an Akai S6000 and I must say that it is incredible! The interface design is absolutely brilliant. So easy to use and it sounds teriffic. The filters are glassy, bright and smooth.

How loud is the fan?

[Radek]

However loading times are terrible... it's not fun waiting 20 minutes to get a piano (great even) loaded.

Eh? Dunno where that figure came from. Depending on size of course, just a few minutes - if that... just a few seconds to load ~10Mb

I have been reading the SOS' review:
http://www.soundonsound.com/sos/Jul02/a ... akaiz8.asp
It needs 7 minutes to load the piano what they are talking about in that review. How many megabytes is it?
Anyway several minutes is kind of slow but few second per 10Mb isn't looking so bad.

I don't know why those machines couldn't be improved in memory capacity and i/o perfomance? Those look like trivial issues...

They could have been improved .... had there been a commercial incentive (i.e. sales) to do so.

It's a funny situation imho. Soft samplers could dominate because of higher capacity and i/o quickness. Usually when something better shows competitors try to match (or exceed) it. Nemesys has a patent (or it was conexant?) for their disk streaming in Gigasampler. I'm sure company like Roland could snap it easily but what it could do to their core PCM rom based synths? It's typical if someone get big then starts to be very conservative.

Anyway - are there benchmarks of load/save perfomance for various hardware samplers?

[Radek]

Well,

This has been an interesting discussion. I just bought an Akai S6000 and I must say that it is incredible! The interface design is absolutely brilliant. So easy to use and it sounds teriffic. The filters are glassy, bright and smooth.

How loud is the fan?

You might consider the Z4/8 instead. Those don't have fans (at least not according to manuals).

[hollowsun]

How loud is the fan?

S5/6000 don't have fans.

[jupiter8]

With the interpolation set to high, I was only able to get about 64 stereo voices. Strangely enough this is what the Akai S6000 offers in terms of stereo voicing with an equal quality interpolation.

If you have memory to spare try the "load as 32 bit" option in Kontakt and try again. IIRC that roughly doubled the polyphony on a Mac G5 in a test i saw. And turn of the filters if you're not using them. Of course that would'nt be a fair comparison to the S6000 since you can't turn them off. OTOH with Kontakt you don't vaste CPU on unnescessary things.

SOS did a polyphony test on a Mac G5 (don't remember the model) and they got like 3500 voices out of the EXS 24 in 32 bit float mode and without filters.
OTOH with the filters on and no 32 bit mode it was something like 200.

[hollowsun]

It needs 7 minutes to load the piano what they are talking about in that review. How many megabytes is it?

That particular one was 256Mb which is pretty much an exception. Being an 'old school' sampler using 'old school' sample editing/looping/programming techniques (as opposed to 'new school' techniques of overlong, unlooped samples on every note with XX velocity layers!), most sounds are rarely above 10Mb or so (some more, some less) and take just seconds to load. That comment in the SOS review (especially as a listed 'con') was a little unfair as most sounds take nowhere as long.

Anyway several minutes is kind of slow but few second per 10Mb isn't looking so bad.

Those were the results of quick test yesterday loading some of my stuff here which was typically 10Mb or so each... some more, some less but on average, just a few seconds load.

Usually when something better shows competitors try to match (or exceed) it.

Yes but at the time, hardware sampler sales were declining rapidly and there was little incentive to invest heavily in something that may not have any financial return given the widespread acceptance of s/w samplers. Of course, it's a 'chicken and egg' situation - had h/w manufacturers come out with a killer 1Gb streaming sampler, maybe s/w samplers wouldn't have taken such a hold. But, as you say....

It's typical if someone get big then starts to be very conservative.

Yep! That said, it's very different for a hardware manufacturer to commit to - say - £500,000 of R+D with no guarantee of recouping those costs. S/w manufacturers, on the other hands, merely have to paste in some code, add a few extra UI bitmaps and keep their fingers crossed that the host's CPU can cope!! Well... maybe not as simple as that but you get the gist They certainly don't have to consider circuit design, PCB layout, component purchase/procurement, tooling costs, manufacturing costs, international safety testing/compliance, etc..

Anyway - are there benchmarks of load/save perfomance for various hardware samplers?

Not to my knowledge.


Steve

[Sascha Franck]

With the filter (12dB LP) switched on, I'm getting around 180 voices (mono, 44.1kHz, 16bit samples, internally processed at 32bit, virtual memory off) out of the EXS on my Pentium M 1.86 laptop at 128 samples buffersize (using an Indigo I/O). The samples are all "unpitched" and the EXS is loading them into RAM entirely.
On the Macbook (2.0GHz, 2GB RAM) I'm getting easily over 600 voices (same settings for the EXS, 64 samples buffersize with an M-Audio FW410).
As said, all this is with the filter switched on.

CPU useage increases as soon as the EXS has to do some pitching (both true for pitchbend and patches using only a few zones/samples).
In addition, the EXS' interpolation quality is best to be called "fair", in other words: you'll experience quite some aliasing, once it has to do any pitch jobs. Yet, for playback only things (which almost any, say, orchestra library would qualify for), it's incredibly more efficient than any hardware sampler.

I've yet to do some Kontakt 2 tests on the Macbook.

[jupiter8]

I just found the article. They benchmark a 2.7 gHz dual G5. And the AMD X2 should be a good deal faster i believe. This sample is from the VSL library so it is possible that it does'nt pitch shift at all. IE one sample for each key.

Moving on to EXS24, I loaded the usual 'HA_ES' Harp patch from the Vienna Symphonic Library and disabled the Virtual Memory option for disk streaming, to concentrate the results on processor and memory performance. Using the sample's original 16-bit storage format with the filter disabled, I could play back 844 voices (13 instances playing 64 voices and one instance playing 12 voices) with 95 percent CPU usage, compared to 752 and 704 voices for the dual-2.5GHz and dual-2GHz models respectively. Enabling the filter brought these figures down to 384 voices (6 x 64) with around 85 percent User CPU usage, compared to 240 and 192 for the 2.5GHz and 2GHz models respectively.

Enabling 32-bit floating point storage for samples always gives a significant performance boost, in addition to increasing the memory footprint required for the samples, and here the dual-2.7GHz excelled, being capable of playing 4224 (66 x 64) voices with 93 percent User CPU usage, over the 3648 and 3264 voices attained by the 2.5GHz and 2GHz models. So, overall, while the new high-end PowerMac doesn't offer any significant features over the previous generation, the extra speed is definitely going to be welcomed by users requiring host-based processing.