In windows percents are not so high, but anyway there will be some preformance revenue.
When i asked support about plans of releasing 64bit drivers for scope - silence was the answer
Does someone know smth?
Windows-64+Scope
When using AMD64 + FreeBSD some people managed achieve 20% performance boost when using full 64-bit mode (i.e. 64bit os)
In windows percents are not so high, but anyway there will be some preformance revenue.
When i asked support about plans of releasing 64bit drivers for scope - silence was the answer
Does someone know smth?
In windows percents are not so high, but anyway there will be some preformance revenue.
When i asked support about plans of releasing 64bit drivers for scope - silence was the answer
Does someone know smth?
-
Guest
that is NOT due to a 64 bit OS, but a side effect of that type of CPU, which certainly will benefit from an appropriate 'data packaging' (as it's intended for 64 bit).On 2005-06-08 09:12, Vasfed wrote:
...some people managed achieve 20% performance boost when using full 64-bit mode (i.e. 64bit os)...
cheers, Tom
exactly, but it is only possible when os is fully 64 bitOn 2005-06-08 14:03, astroman wrote:that is NOT due to a 64 bit OS, but a side effect of that type of CPU, which certainly will benefit from an appropriate 'data packaging' (as it's intended for 64 bit).On 2005-06-08 09:12, Vasfed wrote:
...some people managed achieve 20% performance boost when using full 64-bit mode (i.e. 64bit os)...
cheers, Tom
for ex. kernel streaming(used for audio too) will go up to two times faster just because of possibility to transfer 8 bytes instead of 4 a time
again this (the width of registers) has nothing to do with a 64 bit OS, which is an adreesing mode in the first place.
Afaik regular graphic processors have (at least ?) twice as broad registers
You may expect some advantages in certain kind of transfers, but the other side of the story is that there will also be a lot of overhead in padding data that just doesn't have the 'numeric size'. I guess it would end somewhat on par with a slight(!) advantage to the 64bit CPU because it has the more 'recent' design.
The only true advantage (I've read of) is the attempt to get rid of the floating point part of the CPU completely, thus avoiding mode switching etc. But this will sacrifice precision and also cause the rewrite of huge amounts of code - you cannot simply recompile such specialized parts of a program.
But since we're mostly concerned with audio and signal processing apps you can safely assume that even if a (general purpose) 64 bit CPU would perform twice as fast, it will still be 5-10 times slower than a specialized device like a DSP or the afforementioned graphic chip in this domain.
All this performance info is benchmarking - there's no real world data available.
There are cases known (and have been evaluated in depth) where a CPU reached only 15 percent of it's estimated performance because the data got stuck between memory and registers - and that happened in a hand optimized piece of assembly code.
How much faster is your machine when making the FSB two times faster ?
To increase performance there are many ways that are more effective than 64bit, but none of them has the ability of generating that amount of cash flow...
cheers, Tom
<font size=-1>[ This Message was edited by: astroman on 2005-06-09 02:51 ]</font>
Afaik regular graphic processors have (at least ?) twice as broad registers
You may expect some advantages in certain kind of transfers, but the other side of the story is that there will also be a lot of overhead in padding data that just doesn't have the 'numeric size'. I guess it would end somewhat on par with a slight(!) advantage to the 64bit CPU because it has the more 'recent' design.
The only true advantage (I've read of) is the attempt to get rid of the floating point part of the CPU completely, thus avoiding mode switching etc. But this will sacrifice precision and also cause the rewrite of huge amounts of code - you cannot simply recompile such specialized parts of a program.
But since we're mostly concerned with audio and signal processing apps you can safely assume that even if a (general purpose) 64 bit CPU would perform twice as fast, it will still be 5-10 times slower than a specialized device like a DSP or the afforementioned graphic chip in this domain.
All this performance info is benchmarking - there's no real world data available.
There are cases known (and have been evaluated in depth) where a CPU reached only 15 percent of it's estimated performance because the data got stuck between memory and registers - and that happened in a hand optimized piece of assembly code.
How much faster is your machine when making the FSB two times faster ?
To increase performance there are many ways that are more effective than 64bit, but none of them has the ability of generating that amount of cash flow...
cheers, Tom
<font size=-1>[ This Message was edited by: astroman on 2005-06-09 02:51 ]</font>
marketing.....
this is bullsh@t
if they spent the time optimizing the current hardware, it's be more than powerful enough for most users already. when everything is 64bit next week, i'm sure it'll be great and i'll use it, but currently, i won't be able to do any more work with 64bit and i'd have to spend a lot of extra money for little to no improvement and maybe even a slight loss of performance in some places(no true 64bit software!)....
this is bullsh@t
if they spent the time optimizing the current hardware, it's be more than powerful enough for most users already. when everything is 64bit next week, i'm sure it'll be great and i'll use it, but currently, i won't be able to do any more work with 64bit and i'd have to spend a lot of extra money for little to no improvement and maybe even a slight loss of performance in some places(no true 64bit software!)....
no doubt.
and greed was a good call.
i said marketing because that's what's making people exited about products that the industry sez are going to be supplanted in months, and that don't work any better than the products currently out there. the cutting edge being sold at the moment is not the next big thing, it's a stopgap product to have something new to sell until the REAL next big thing....at least the asb synths won't be useless all the sudden in six months. can you imagine the uproar? and yet a new cutting edge 64bit computer will cost more than an asb...after witnessing rambus(sold as the ultimate future of memory technology just months before ddr) and other foolishness, i'm amazed at how quickly people will line up to be fleeced.
most likely, eventually, everything will HAVE to be 64bit compatible, and maybe at that point, there'll be a real advantage in it. that time could be days months or years in the future, but until then, ignore it. until then it's just a bragging point or an interesting oddity, not a good reliable tool.
and greed was a good call.
i said marketing because that's what's making people exited about products that the industry sez are going to be supplanted in months, and that don't work any better than the products currently out there. the cutting edge being sold at the moment is not the next big thing, it's a stopgap product to have something new to sell until the REAL next big thing....at least the asb synths won't be useless all the sudden in six months. can you imagine the uproar? and yet a new cutting edge 64bit computer will cost more than an asb...after witnessing rambus(sold as the ultimate future of memory technology just months before ddr) and other foolishness, i'm amazed at how quickly people will line up to be fleeced.
most likely, eventually, everything will HAVE to be 64bit compatible, and maybe at that point, there'll be a real advantage in it. that time could be days months or years in the future, but until then, ignore it. until then it's just a bragging point or an interesting oddity, not a good reliable tool.
thanks for passing that clear information 
to come back to the beginning of the thread:
a 20% improvement of native processing is a fairly low motivation to write a new driver for a still experimental (imho) OS.
20% is similiar to 2 versus 2.5 GHZ - I wouldn't even open my box for that
since we're more or less interested in signal processing, this should make clear that there's nothing to expect from the main suppliers. Nothing worth the effort...
but it has been mentioned that the 'abuse' of graphic CPUs is much more interesting in this context.
If there is high audio quality to expect is a completely different question, as that's depending on a very high level of sophistication to write processing algorithms - definetely beyond the 'regular' VSTI developer
cheers, tom
<font size=-1>[ This Message was edited by: astroman on 2005-07-18 02:52 ]</font>
to come back to the beginning of the thread:
a 20% improvement of native processing is a fairly low motivation to write a new driver for a still experimental (imho) OS.
20% is similiar to 2 versus 2.5 GHZ - I wouldn't even open my box for that
since we're more or less interested in signal processing, this should make clear that there's nothing to expect from the main suppliers. Nothing worth the effort...
but it has been mentioned that the 'abuse' of graphic CPUs is much more interesting in this context.
If there is high audio quality to expect is a completely different question, as that's depending on a very high level of sophistication to write processing algorithms - definetely beyond the 'regular' VSTI developer
cheers, tom
<font size=-1>[ This Message was edited by: astroman on 2005-07-18 02:52 ]</font>