What I find perplexing is that we did not see such negativity when the industry changed form 16 bit to 32 bit hardware.
A minor point perhaps, but this is simply not so. I have been researching some computing history for my senior thesis, and spent a while reading old PC Weekly mags. To put it mildly, the 16-32 bit shift was hugely controversial, especially concerning compatibility. A lot of people were screaming bloody murder in the letters pages when it emerged that they would be left behind.
And remember, there were far better reasons for that switch, at least in terms of computing power. There aren't any such reasons yet for the 32-64 bit switch, but that sure as heck isn't gonna stop the whining.....
As far as consumer applications go I think it is only a matter of developers seeing enough 64 bit hardware in the wild for them to target 64 bit hardware. Games are one item that could make immediate use of the extended address space 64 bits offer. Media editing programms are not far behind. I would have to say that there are actually a number of potential consumer applications that could take advantage of 64 bits.
The only hold up we have is the adoption of 64 bit hardware. There has to be enough hardware out there to enable profitable sales of the software. This is where Apple has the potential to lead if they can transition their consumer lines quickly to 64 bit.
Few (if any) games need a 64-bit address space, and the trade-off of having 8 byte pointers would typically make this a poor design choice. There might be a few data structures that would benefit from a sparse 64-bit address space, but in most cases there are better alternatives that work more efficiently in a 32-bit space. The same is true for most consumer applications. Media editing programs (at least for video) are an exception that could benefit from the huge address space, but solutions are already in place to support data streaming so the payoff isn't too dramatic.
Notice that "32/64 bit" mealy-mouthedness. Prescott is a 32-bit processor PERIOD. It may have some small 64-bit aspects, but they're nothing new.
Intel is spreading FUD about 64-bit like a thick layer of manure.
You really don't know what you're talking about. Prescott, using Socket 775 from memory, is a full 64 bit chip with 32 bit backward compatibility just like the Opteron is. The current Prescott has the same functionality it is just unusable like HT was in early PIVs. If you've read Intel's programmer notes you will even notice some very distinct similarities to AMD's notes for the Opterons so if Intel is producing 32 bit chips then so is AMD, and if you think that you need to go back and read up a bit.
Quote:
Originally posted by G-News
We're talking about Nocoma, not Prescott, in case you still haven't noticed.
Nocona is the same core as Prescott. Only differences really are MP support, bus speeds and cache sizes.
Few (if any) games need a 64-bit address space, and the trade-off of having 8 byte pointers would typically make this a poor design choice. There might be a few data structures that would benefit from a sparse 64-bit address space, but in most cases there are better alternatives that work more efficiently in a 32-bit space. The same is true for most consumer applications. Media editing programs (at least for video) are an exception that could benefit from the huge address space, but solutions are already in place to support data streaming so the payoff isn't too dramatic.
Actually the 32/64 in consumer desktops has only one function...to sell more computers/ Let'
s be honest, if you have a fairly recent vintage computer (about 3-4 years old now) and you go online, play solitare, send email, do spreadsheets and play MP#s, any computer will have more than enough power for you. The tech industry is just making up excuses to get you to buy a new one. Seriously, in those applications you probablly wouldn't even notice the speed increase from a 1ghz pentium to a 3.5 pentium.
By convincing people to edit video on their computer and play more complex games, it is the only way to get people to buy a new computer. The coputer indistry can't live on $500 computers alone.
Actually the 32/64 in consumer desktops has only one function...to sell more computers/ Let'
s be honest, if you have a fairly recent vintage computer (about 3-4 years old now) and you go online, play solitare, send email, do spreadsheets and play MP#s, any computer will have more than enough power for you. The tech industry is just making up excuses to get you to buy a new one. Seriously, in those applications you probablly wouldn't even notice the speed increase from a 1ghz pentium to a 3.5 pentium.
By convincing people to edit video on their computer and play more complex games, it is the only way to get people to buy a new computer. The coputer indistry can't live on $500 computers alone.
The same argument was used when desktops transitioned from 16 to 32 bit. Look how that turned out.
you have to be honest enough to admit that 16bit is more than enough for word processing.
Not really. I have a document that's 1100+ pages long and has 400 000+ words, or around 2 million characters. That in fact isn't even the full document. I have that again in a separate file. A 16 bit computer simply couldn't handle that.
The same argument was used when desktops transitioned from 16 to 32 bit. Look how that turned out.
People keep bringing that up, but what isn't understood is that the change in magnitude of the 16->32 vs 32->64 shifts is vastly different. Yes there are good uses for 64-bit address spaces, but most software doesn't need it at all and will actually lose some performance on identical hardware by using 8-byte pointers. Not to mention losing the ability to run on 32-bit hardware.
Furthermore, the ability to do 32-bit math efficiently was an enormous boon when moving from 16-bit machines. Thanks to the ubiquitous floating point hardware, with support for 32-bit and 64-bit float types, most of the math needs are already covered. Yes, some software will benefit from native 64-bit integer math... but the amount of software where that is actually the performance bottleneck is tiny.
For most users the reality of the arrival of 64-bit hardware is something of a yawner. The boys in marketing will try to dance around that, of course.
64 bit's can also help 3D rendering, Movie Editing, and if implemented correctly most apps that people complain about being to slow. I don't see it as a yawner,. More like an opportunity.
you know, 64bit still doesn't fix the fact that most systems have one or multiple bottlenecks, that software is poorly written or that harddrives happen to be slow, which seems more like the prime 3 reasons for things being slow, if you ask me.
64 bit's can also help 3D rendering, Movie Editing, and if implemented correctly most apps that people complain about being to slow. I don't see it as a yawner,. More like an opportunity.
How are you thinking that it helps...
3D rendering? Since most of the calculations are in 64-bit floating point, the 64-bit integer and 64-bit addressing isn't required. If you mean that the data sets are 2+ GB in size, then 64-bit addressing will certainly help, but the number of people doing that is tiny and the performance of processing that much data is still going to be painfully slow (or could be handled by streaming the data instead).
Movie editing? Since individual frames are much much less than 4 GB and streaming solutions are already in place the benefit of 64-bits here is more due to programmer laziness than any real performance improvement... unless you actually have >2 GB of RAM. For the pro video editors in the crowd 64-bit addressing becomes a big improvement, but this is a pretty small set of users. Important and vocal, but certainly not most of Apple's users. With a data set of this size, performance isn't going to be stellar and probably not much different than using a streamed mechanism.
64-bit hardware gives you very little in terms of speed. It does give new capability, but the number of users and software that need this capability is very small. There are definitely some places where 64-bit can make a big difference, but (I repeat) they are not common and they are not what most users will experience. The big ones for Apple are probably in scientific computing since that is an important area for Apple to gain mindshare and leverage their Unix basis.
64-bit hardware gives you very little in terms of speed.
I disagree. As the majority of current 64bit systems happen to be in the mid-highend server.
These machines are generally used for highend apps that couldn't run effiently on a 32bit server.
Most of these machines will required 4GB memory per CPU and have more than 4 CPU's.
Most of these machines will have database functionallity and will be more than 4GB databases.
Imagine googles performance with a 32-bit machine.
A webserver that loads static content to memory had major gains over those loading from hard drives.
Encyption has huge performance advantages especially with 256, 512 or 1024bit encrption.
Using a 64bit system for surfing the net, reading e-mail an d doing basic word/excel stuff is a waste of time. In saying this you could it is also valid for a 32bit system.
Just use the right tool for the right job.
64Bit is old hat and Intel just has its marketing cap on.
I disagree. As the majority of current 64bit systems happen to be in the mid-highend server.
These machines are generally used for highend apps that couldn't run effiently on a 32bit server.
Most of these machines will required 4GB memory per CPU and have more than 4 CPU's.
Most of these machines will have database functionallity and will be more than 4GB databases.
Imagine googles performance with a 32-bit machine.
A webserver that loads static content to memory had major gains over those loading from hard drives.
Encyption has huge performance advantages especially with 256, 512 or 1024bit encrption.
Using a 64bit system for surfing the net, reading e-mail an d doing basic word/excel stuff is a waste of time. In saying this you could it is also valid for a 32bit system.
Just use the right tool for the right job.
64Bit is old hat and Intel just has its marketing cap on.
Dobby.
Uhh... okay, but when was I talking about servers? I was saying most users have no need for 64-bit's capabilities.
If you are talking about servers running huge databases then absolutely 64-bit systems are a tremendous benefit. The main win is the massive address space so that the entire database can live in RAM (or at least large portions of it).
For encryption SIMD units are usually a bigger win than 64-bit integer math.
Comments
Originally posted by wizard69
What I find perplexing is that we did not see such negativity when the industry changed form 16 bit to 32 bit hardware.
A minor point perhaps, but this is simply not so. I have been researching some computing history for my senior thesis, and spent a while reading old PC Weekly mags. To put it mildly, the 16-32 bit shift was hugely controversial, especially concerning compatibility. A lot of people were screaming bloody murder in the letters pages when it emerged that they would be left behind.
And remember, there were far better reasons for that switch, at least in terms of computing power. There aren't any such reasons yet for the 32-64 bit switch, but that sure as heck isn't gonna stop the whining.....
Originally posted by wizard69
As far as consumer applications go I think it is only a matter of developers seeing enough 64 bit hardware in the wild for them to target 64 bit hardware. Games are one item that could make immediate use of the extended address space 64 bits offer. Media editing programms are not far behind. I would have to say that there are actually a number of potential consumer applications that could take advantage of 64 bits.
The only hold up we have is the adoption of 64 bit hardware. There has to be enough hardware out there to enable profitable sales of the software. This is where Apple has the potential to lead if they can transition their consumer lines quickly to 64 bit.
Few (if any) games need a 64-bit address space, and the trade-off of having 8 byte pointers would typically make this a poor design choice. There might be a few data structures that would benefit from a sparse 64-bit address space, but in most cases there are better alternatives that work more efficiently in a 32-bit space. The same is true for most consumer applications. Media editing programs (at least for video) are an exception that could benefit from the huge address space, but solutions are already in place to support data streaming so the payoff isn't too dramatic.
Originally posted by onlooker
You must have missed the presentation.
Notice that "32/64 bit" mealy-mouthedness. Prescott is a 32-bit processor PERIOD. It may have some small 64-bit aspects, but they're nothing new.
Intel is spreading FUD about 64-bit like a thick layer of manure.
Originally posted by G-News
We're talking about Nocoma, not Prescott, in case you still haven't noticed.
What he said.
Originally posted by cubist
Notice that "32/64 bit" mealy-mouthedness. Prescott is a 32-bit processor PERIOD. It may have some small 64-bit aspects, but they're nothing new.
Intel is spreading FUD about 64-bit like a thick layer of manure.
You really don't know what you're talking about. Prescott, using Socket 775 from memory, is a full 64 bit chip with 32 bit backward compatibility just like the Opteron is. The current Prescott has the same functionality it is just unusable like HT was in early PIVs. If you've read Intel's programmer notes you will even notice some very distinct similarities to AMD's notes for the Opterons so if Intel is producing 32 bit chips then so is AMD, and if you think that you need to go back and read up a bit.
Originally posted by G-News
We're talking about Nocoma, not Prescott, in case you still haven't noticed.
Nocona is the same core as Prescott. Only differences really are MP support, bus speeds and cache sizes.
Originally posted by Programmer
Few (if any) games need a 64-bit address space, and the trade-off of having 8 byte pointers would typically make this a poor design choice. There might be a few data structures that would benefit from a sparse 64-bit address space, but in most cases there are better alternatives that work more efficiently in a 32-bit space. The same is true for most consumer applications. Media editing programs (at least for video) are an exception that could benefit from the huge address space, but solutions are already in place to support data streaming so the payoff isn't too dramatic.
Actually the 32/64 in consumer desktops has only one function...to sell more computers/ Let'
s be honest, if you have a fairly recent vintage computer (about 3-4 years old now) and you go online, play solitare, send email, do spreadsheets and play MP#s, any computer will have more than enough power for you. The tech industry is just making up excuses to get you to buy a new one. Seriously, in those applications you probablly wouldn't even notice the speed increase from a 1ghz pentium to a 3.5 pentium.
By convincing people to edit video on their computer and play more complex games, it is the only way to get people to buy a new computer. The coputer indistry can't live on $500 computers alone.
Originally posted by jade
Actually the 32/64 in consumer desktops has only one function...to sell more computers/ Let'
s be honest, if you have a fairly recent vintage computer (about 3-4 years old now) and you go online, play solitare, send email, do spreadsheets and play MP#s, any computer will have more than enough power for you. The tech industry is just making up excuses to get you to buy a new one. Seriously, in those applications you probablly wouldn't even notice the speed increase from a 1ghz pentium to a 3.5 pentium.
By convincing people to edit video on their computer and play more complex games, it is the only way to get people to buy a new computer. The coputer indistry can't live on $500 computers alone.
The same argument was used when desktops transitioned from 16 to 32 bit. Look how that turned out.
Originally posted by G-News
you have to be honest enough to admit that 16bit is more than enough for word processing.
Not really. I have a document that's 1100+ pages long and has 400 000+ words, or around 2 million characters. That in fact isn't even the full document. I have that again in a separate file. A 16 bit computer simply couldn't handle that.
Originally posted by onlooker
The same argument was used when desktops transitioned from 16 to 32 bit. Look how that turned out.
People keep bringing that up, but what isn't understood is that the change in magnitude of the 16->32 vs 32->64 shifts is vastly different. Yes there are good uses for 64-bit address spaces, but most software doesn't need it at all and will actually lose some performance on identical hardware by using 8-byte pointers. Not to mention losing the ability to run on 32-bit hardware.
Furthermore, the ability to do 32-bit math efficiently was an enormous boon when moving from 16-bit machines. Thanks to the ubiquitous floating point hardware, with support for 32-bit and 64-bit float types, most of the math needs are already covered. Yes, some software will benefit from native 64-bit integer math... but the amount of software where that is actually the performance bottleneck is tiny.
For most users the reality of the arrival of 64-bit hardware is something of a yawner. The boys in marketing will try to dance around that, of course.
Originally posted by Programmer
.
For most users the reality of the arrival of 64-bit hardware is something of a yawner. The boys in marketing will try to dance around that, of course.
Yup the only people who think their computers are too slow are editing video, have computers from 1995, want more fps in video games or using dial up.
The average user really has no use for 64 bit.
And I will fully admit I have no use for 64 bit but that will not stop me from getting a g5 powerbook.
Originally posted by onlooker
64 bit's can also help 3D rendering, Movie Editing, and if implemented correctly most apps that people complain about being to slow. I don't see it as a yawner,. More like an opportunity.
How are you thinking that it helps...
3D rendering? Since most of the calculations are in 64-bit floating point, the 64-bit integer and 64-bit addressing isn't required. If you mean that the data sets are 2+ GB in size, then 64-bit addressing will certainly help, but the number of people doing that is tiny and the performance of processing that much data is still going to be painfully slow (or could be handled by streaming the data instead).
Movie editing? Since individual frames are much much less than 4 GB and streaming solutions are already in place the benefit of 64-bits here is more due to programmer laziness than any real performance improvement... unless you actually have >2 GB of RAM. For the pro video editors in the crowd 64-bit addressing becomes a big improvement, but this is a pretty small set of users. Important and vocal, but certainly not most of Apple's users. With a data set of this size, performance isn't going to be stellar and probably not much different than using a streamed mechanism.
64-bit hardware gives you very little in terms of speed. It does give new capability, but the number of users and software that need this capability is very small. There are definitely some places where 64-bit can make a big difference, but (I repeat) they are not common and they are not what most users will experience. The big ones for Apple are probably in scientific computing since that is an important area for Apple to gain mindshare and leverage their Unix basis.
Originally posted by Programmer
How are you thinking that it helps...
64-bit hardware gives you very little in terms of speed.
I disagree. As the majority of current 64bit systems happen to be in the mid-highend server.
These machines are generally used for highend apps that couldn't run effiently on a 32bit server.
Most of these machines will required 4GB memory per CPU and have more than 4 CPU's.
Most of these machines will have database functionallity and will be more than 4GB databases.
Imagine googles performance with a 32-bit machine.
A webserver that loads static content to memory had major gains over those loading from hard drives.
Encyption has huge performance advantages especially with 256, 512 or 1024bit encrption.
Using a 64bit system for surfing the net, reading e-mail an d doing basic word/excel stuff is a waste of time. In saying this you could it is also valid for a 32bit system.
Just use the right tool for the right job.
64Bit is old hat and Intel just has its marketing cap on.
Dobby.
Originally posted by dobby
I disagree. As the majority of current 64bit systems happen to be in the mid-highend server.
These machines are generally used for highend apps that couldn't run effiently on a 32bit server.
Most of these machines will required 4GB memory per CPU and have more than 4 CPU's.
Most of these machines will have database functionallity and will be more than 4GB databases.
Imagine googles performance with a 32-bit machine.
A webserver that loads static content to memory had major gains over those loading from hard drives.
Encyption has huge performance advantages especially with 256, 512 or 1024bit encrption.
Using a 64bit system for surfing the net, reading e-mail an d doing basic word/excel stuff is a waste of time. In saying this you could it is also valid for a 32bit system.
Just use the right tool for the right job.
64Bit is old hat and Intel just has its marketing cap on.
Dobby.
Uhh... okay, but when was I talking about servers? I was saying most users have no need for 64-bit's capabilities.
If you are talking about servers running huge databases then absolutely 64-bit systems are a tremendous benefit. The main win is the massive address space so that the entire database can live in RAM (or at least large portions of it).
For encryption SIMD units are usually a bigger win than 64-bit integer math.