Thinking ahead to the CPU power 2 years from now
-
- KVRer
- Topic Starter
- 17 posts since 21 Jul, 2003
Is there any type of processing that can not be done in real time because it takes too much CPU power?
Not soon, but 2-5 years from now the CPUs will be SO much faster. What types of processing will then become a reality because the systems will be fast enough?
Not soon, but 2-5 years from now the CPUs will be SO much faster. What types of processing will then become a reality because the systems will be fast enough?
-
- KVRian
- 541 posts since 19 Jun, 2002 from London, UK
Very low latency convolution reverbs will be a hot topic.
Plus there's capability for better physical modelling, especially for, say, wind and string instruments. A truly devastating "virtual guitar" could be on the cards.
And that of course opens the door for a physically modelled human voice - none of the current virtual vocalists based on samples, but a truly tweakable voice.
Truly.
Plus there's capability for better physical modelling, especially for, say, wind and string instruments. A truly devastating "virtual guitar" could be on the cards.
And that of course opens the door for a physically modelled human voice - none of the current virtual vocalists based on samples, but a truly tweakable voice.
Truly.
Three shall be the number of the counting
And the number of the counting shall be three.
And the number of the counting shall be three.
-
- KVRAF
- 3139 posts since 6 Sep, 2002 from United Kingdom & Opinions Will Travel :O)
Top notch additive - this is where power is really neededMidigator wrote:Is there any type of processing that can not be done in real time because it takes too much CPU power?
Not soon, but 2-5 years from now the CPUs will be SO much faster. What types of processing will then become a reality because the systems will be fast enough?
And an even better Jurassic Park
Best regards,
Spe3d
:O)
-
- KVRAF
- 6519 posts since 13 Mar, 2002 from UK
That kind of processing power is already with us in the form of dual cpu systems. Unfortunately the attempts to harness it for audio use have been less than impressive. Steiny claim 30-50% improvements using 2 cpu's. Not enough to tempt most users.
It baffles me that some of the smaller host makers have not taken up the challenge. Unlike most areas which could benefit from multi cpu usage, our need is well defined - more plugs. FX Teleport manage close to 200% improvement with 2 cpu's talking over a crude ethernet connection. Surely with the talent around this place someone could at least match that with a well designed multi-processor host?
If (for example) Tracktion could trounce Steiny in this area, a lot of heavy spenders would sit up and take notice.
It baffles me that some of the smaller host makers have not taken up the challenge. Unlike most areas which could benefit from multi cpu usage, our need is well defined - more plugs. FX Teleport manage close to 200% improvement with 2 cpu's talking over a crude ethernet connection. Surely with the talent around this place someone could at least match that with a well designed multi-processor host?
If (for example) Tracktion could trounce Steiny in this area, a lot of heavy spenders would sit up and take notice.
-
- KVRian
- 541 posts since 19 Jun, 2002 from London, UK
And, of course, Doom 4.
Three shall be the number of the counting
And the number of the counting shall be three.
And the number of the counting shall be three.
-
- KVRian
- 541 posts since 19 Jun, 2002 from London, UK
In fact this is where a lot of the work would be well channeled, but for other reasons. Dual core chips are coming out next year from both AMD and Intel, and all developers would be well advised to REALLY make the most of that. The real challenge I guess will be in the compilers and how they handle compiling stuff for legacy single-chip, legacy single-chip HT, legacy 2 processor, dual-core, dual-core HT, dual dual core, dual dual-core HT etc etc...nuffink wrote:That kind of processing power is already with us in the form of dual cpu systems. Unfortunately the attempts to harness it for audio use have been less than impressive.
Terrifying, but potentially fantastic.
Three shall be the number of the counting
And the number of the counting shall be three.
And the number of the counting shall be three.
-
Stupid American Pig Stupid American Pig https://www.kvraudio.com/forum/memberlist.php?mode=viewprofile&u=4753
- KVRAF
- 7065 posts since 25 Nov, 2002 from not sure
I think that the increase in CPU horspower is going to only worsen the problems that we have now- MS seems to think that the more powerful CPUs become, they are entitled to hog more of those resources.
-
- KVRAF
- 1821 posts since 5 Oct, 2003
I eagerly await the day when CPU can fellate on cue...
....did I say that out loud?
...wow...
....did I say that out loud?
...wow...
"Time makes fools of us all. Our only comfort is that greater shall come after us." Eric Temple Bell
http://thetomorrowfile.bandcamp.com/
http://thetomorrowfile.bandcamp.com/
-
- KVRAF
- 4644 posts since 28 Nov, 2002 from Chicago
probably not.jtxx000 wrote:Doesn't it double every 18 months or something like that? If so then we're looking at 6/7 ghz mahines...
More likely you are going to be seeing greatly more efficient 4Ghz machines. Even Intel are admitting publicly now that just pumping up the clock speed is both an unscalable and inefficient way of improving performance.
Someone shot the food. Remember: don't shoot food!
-
- KVRian
- 541 posts since 19 Jun, 2002 from London, UK
True, speed alone ain't everything. Just look at how the Pentium M is crapping all over the Pentium 4 clock-for-clock.
Three shall be the number of the counting
And the number of the counting shall be three.
And the number of the counting shall be three.
-
- KVRAF
- 3139 posts since 6 Sep, 2002 from United Kingdom & Opinions Will Travel :O)
http://www.kvr-vst.com/get.php?mode=res ... mit=SearchMidigator wrote:"Top notch additive - this is where power is really needed"
Can you point me at some links to understand this more?
Thanks.
Start there, it will become clear as most will explain the limitations.
Best regards,
Spe3d
:O)
-
- KVRAF
- 2458 posts since 3 Oct, 2002 from SF CA USA NA Earth
The sort of plugin developers who can invest the effort needed to do detailed hardware emulations (NI, GForce, Arturia) will probably start doing more emulation at the individual component level if they have more CPU power to burn. It will make their stuff more authentic if not any more useful.
Physical modeling of acoustic instruments may take off, but here, the development work, more than the CPU cycles, is the hard part. I think. I could be full of crap.
I'm not sure additive will really get any more exciting. I guess once you can do 50 harmonics on one voice in less than 1% of a CPU it's kinda cool, but I feel like there are additive compromise methods which simultaneously (a) make it easier to edit and (b) make it faster to synthesize; using those techniques you can effectively make a continuum between wavetable and additive synths.
Really, different techniques for implementing oscillators basically don't matter any more. It's the filters, envelopes, distortion, modulation, effects, etc. that you pump the oscillator tones through, that people care about at this point.
Physical modeling of acoustic instruments may take off, but here, the development work, more than the CPU cycles, is the hard part. I think. I could be full of crap.
I'm not sure additive will really get any more exciting. I guess once you can do 50 harmonics on one voice in less than 1% of a CPU it's kinda cool, but I feel like there are additive compromise methods which simultaneously (a) make it easier to edit and (b) make it faster to synthesize; using those techniques you can effectively make a continuum between wavetable and additive synths.
Really, different techniques for implementing oscillators basically don't matter any more. It's the filters, envelopes, distortion, modulation, effects, etc. that you pump the oscillator tones through, that people care about at this point.