Intel's fastest GPUs are are what, 1000x slower than AMD's? Even calling them GPUs is a bit of a stretch.
Hopefully, OCL is the kick in the ass (from Apple) that Apple needs to use at least mainstream, current-generation GPUs.
On a side note, hopefully pretty soon we will be able to just buy a PC graphics card and pop it in the Mac Pro desktop, with EFI replacing BIOS(?) in the PC world. So we won't have to deal with $2500 towers with two year old cards in them with 512MB VRAM.
Graphics cards are the area of greatest innovation the the computer world. <i>Modern</i> cards pack so much power, it is measured in Teraflops. Too bad apple doesn't use them.
/disgruntled Maya user.
Quote:
Hey I just searched around, and it looks like the GT 120 is merely a rebranding of the 9500 GT model. This implies to me that it's actually a downgrade from the 8800 GT that used to be standard in the MacPros.
That may be true but you can't expect developers to embrace your new OCL technology if you make machines unable to use it.
This is very true. Developers need to know that the majority of Apples machines can leverage OpenCL. However fall back to CPU computation should be provided on software that can get away with it.
Quote:
If they use a gpu on the MacBook and mini that isn't OCL compatible then OCL is DOA. Developers will have little reason to adopt it.
For the most part yes that is the case. However I can see some potential OpenCL apps that would run well on these platforms anyways. Hardware is still a limitation with respect to some software, I can see developers excluding the 9400m from thier list of acceptable GPUs simply due to the need for more power.
The key reason for developers to adopt OpenCL isn't anyone hardware model from Apple, rather it is the reality that they can get portions of their software to run much faster. They can do this with tools that are easy to use and work within there normal development environment.
T.... I can see developers excluding the 9400m from thier list of acceptable GPUs simply due to the need for more power...
Why would they do that? IIRC, some OCL benches had the 9400m perform tasks as fast as an 8 core Xenon. ( I'll see if I can find the link later).
In fact I think this plays right into Apple's plans. I imagine Apple leveraging OCL and GCD in order to get acceptable performance out of modest hardware. That keeps them from having to 'keep up with the Jones' WRT hardware components such as gpus and cpus. It puts the emphasis on the software which could distinguish Apple from Windows.
Of course nerds like us will play the 'what if' game, wondering about the performance we could get if Apple would only use the components we'd prefer to have.
Why would they do that? IIRC, some OCL benches had the 9400m perform tasks as fast as an 8 core Xenon. ( I'll see if I can find the link later).
Don't bother as you have missed the point. If a developer has an app that needs acceleration via OCL and that acceleration is only viable on state of the art cards, the developer will not target the 9400M at all. What you target is very much dependent on what you app demands, and the user demands.
In the case of Fermi in is a highly optimized processor that would effectively run rings around a 9400M. Especially 64 bit floating point. If your app requires massive amounts of 64 bit FP you effectively eliminate the 9400M from the line up.
Like it or not I can see apps coming in the future that simply ignore the 9400M. Oh one more thing Bench marks are always interesting but you need to look carefully at what is being tested. many of these bench marks are for 32 bit FP as that is what the current NVida cards do well.
Quote:
In fact I think this plays right into Apple's plans. I imagine Apple leveraging OCL and GCD in order to get acceptable performance out of modest hardware. That keeps them from having to 'keep up with the Jones' WRT hardware components such as gpus and cpus. It puts the emphasis on the software which could distinguish Apple from Windows.
That is a great bit of fanboism if you ask me. First; Windows supports its own acceleration framework and C?UDA runs there too. Second; this bear repeating, but OCL will only accelerate certain sorts of apps. Remember we are talking highly parallel vector like computations here. Even with the improvements seen in Fermi there are still many things that a CPU would be better at.
Quote:
Of course nerds like us will play the 'what if' game, wondering about the performance we could get if Apple would only use the components we'd prefer to have.
Well there is preferred and then there is mandatory.
Features become mandatory when Apple is so far behind the eight ball that you have to recommend that people stay away until a update comes. I'm not talking hardware lust here but rather having an expectation of good value for the money.
If a developer has an app that needs acceleration via OCL and that acceleration is only viable on state of the art cards, the developer will not target the 9400M at all....
In the case of Fermi in is a highly optimized processor that would effectively run rings around a 9400M. ...
Like it or not I can see apps coming in the future that simply ignore the 9400M....
Depends on the apps that leverage OCL I guess. That remains to be seen.
Will FCP ( assuming it leverages OCL, but why wouldn't it) require a better gpu than the 9400m? Maybe. Will iMovie (again assuming it leverages OCL)? I doubt it.
Remember there are *way* more Macs with the 9400m than anything else. You target it fro your app and you've got a sizable audience to address.
Quote:
Originally Posted by wizard69
...That is a great bit of fanboism if you ask me. First; Windows supports its own acceleration framework and C?UDA runs there too. Second; this bear repeating, but OCL will only accelerate certain sorts of apps. Remember we are talking highly parallel vector like computations here. Even with the improvements seen in Fermi there are still many things that a CPU would be better at.
Its not fanboism, its reality. When have Apple been aggressive at deploying state of the art gpus? Its been an area of weakness for as long as I can remember. If you think that OCL will be the catalyst for Apple to begin putting high end gpus in all the Macs, I think you're going to be very disappointed. Would I like Apple to use higher performance gpus in Macs? Sure but I'm not holding my breath. I predict that the mini and entry levle iMacs will use integrated graphics. The mid level and best iMacs will use a mid range card adn the Mac Pro will have the BTO option for a high end gpu card. Pretty much like it is now.
Sure Windows has CUDA. But for developers its not much of an attraction. It isn't included with the system os and it only works on NVIDIA gpus. Its not a easy market for developers to target their app at. OCL addresses some of that but even still it remains to be seen if developers will find OCL worth their time to target their apps at.
Depends on the apps that leverage OCL I guess. That remains to be seen.
Exactly, some OpenCL based apps could be very interesting, if the move down from the world of clusters, mainframes and other high performance computers. At the same time run of the mill graphics programs could benefit from OpenCL acceleration. From my perspective high performance, critcal apps are going to target the latest Nvidia GPUs simply because they support the features needed by the developers and users.
Nvidias Fermi for example has drawn gov. lab interest due to features like ECC, vastly improved 64 bit support and other goodies. Now I don't expect people to simulate aging of their stockpiles on their desktop machines, but the latest cards do enable things that have been very difficult to do on a desktop in the past.
So what I'm really saying is that OpenCL opens up the possibility of some very impressive apps moving to the desktop.
Quote:
Will FCP ( assuming it leverages OCL, but why wouldn't it) require a better gpu than the 9400m? Maybe. Will iMovie (again assuming it leverages OCL)? I doubt it.
Yep many exist apps will get a nice boost even on a 9400m. On the other hand Apple could add features to FCP that might require a better GPU than the 9400M. Remember I'm not saying the 9400M is worthless, I'm just saying that there are whole classes of apps that developers won't even try to get to run on the 9400M.
Quote:
Remember there are *way* more Macs with the 9400m than anything else. You target it fro your app and you've got a sizable audience to address.
Yes this is very true but it makes no difference, some apps are just going to require more horse power than the 9400M can deliver. It is a fine point, but people in general have to recognize there are hardware limitations even with the new tech.
Quote:
Its not fanboism, its reality. When have Apple been aggressive at deploying state of the art gpus? Its been an area of weakness for as long as I can remember. If you think that OCL will be the catalyst for Apple to begin putting high end gpus in all the Macs, I think you're going to be very disappointed.
I never said that! Most of Apples product lineup doesn't even provide for that.
Quote:
Would I like Apple to use higher performance gpus in Macs? Sure but I'm not holding my breath. I predict that the mini and entry levle iMacs will use integrated graphics. The mid level and best iMacs will use a mid range card adn the Mac Pro will have the BTO option for a high end gpu card. Pretty much like it is now.
Honestly that depends. It depends on the CPUs choosen some of which don't have provision for integrated GPUs. In a way Intel could end up forcing Apple to use better GPUs or at least not the integrated.
Quote:
Sure Windows has CUDA. But for developers its not much of an attraction. It isn't included with the system os and it only works on NVIDIA gpus. Its not a easy market for developers to target their app at. OCL addresses some of that but even still it remains to be seen if developers will find OCL worth their time to target their apps at.
Comments
....Consumers don't care about OpenCL....
That may be true but you can't expect developers to embrace your new OCL technology if you make machines unable to use it.
If they use a gpu on the MacBook and mini that isn't OCL compatible then OCL is DOA. Developers will have little reason to adopt it.
Intel's fastest GPUs are are what, 1000x slower than AMD's? Even calling them GPUs is a bit of a stretch.
Hopefully, OCL is the kick in the ass (from Apple) that Apple needs to use at least mainstream, current-generation GPUs.
On a side note, hopefully pretty soon we will be able to just buy a PC graphics card and pop it in the Mac Pro desktop, with EFI replacing BIOS(?) in the PC world. So we won't have to deal with $2500 towers with two year old cards in them with 512MB VRAM.
Graphics cards are the area of greatest innovation the the computer world. <i>Modern</i> cards pack so much power, it is measured in Teraflops. Too bad apple doesn't use them.
/disgruntled Maya user.
Hey I just searched around, and it looks like the GT 120 is merely a rebranding of the 9500 GT model. This implies to me that it's actually a downgrade from the 8800 GT that used to be standard in the MacPros.
http://forums.nvidia.com/index.php?showtopic=91033
That may be true but you can't expect developers to embrace your new OCL technology if you make machines unable to use it.
This is very true. Developers need to know that the majority of Apples machines can leverage OpenCL. However fall back to CPU computation should be provided on software that can get away with it.
If they use a gpu on the MacBook and mini that isn't OCL compatible then OCL is DOA. Developers will have little reason to adopt it.
For the most part yes that is the case. However I can see some potential OpenCL apps that would run well on these platforms anyways. Hardware is still a limitation with respect to some software, I can see developers excluding the 9400m from thier list of acceptable GPUs simply due to the need for more power.
The key reason for developers to adopt OpenCL isn't anyone hardware model from Apple, rather it is the reality that they can get portions of their software to run much faster. They can do this with tools that are easy to use and work within there normal development environment.
Dave
T.... I can see developers excluding the 9400m from thier list of acceptable GPUs simply due to the need for more power...
Why would they do that? IIRC, some OCL benches had the 9400m perform tasks as fast as an 8 core Xenon. ( I'll see if I can find the link later).
In fact I think this plays right into Apple's plans. I imagine Apple leveraging OCL and GCD in order to get acceptable performance out of modest hardware. That keeps them from having to 'keep up with the Jones' WRT hardware components such as gpus and cpus. It puts the emphasis on the software which could distinguish Apple from Windows.
Of course nerds like us will play the 'what if' game, wondering about the performance we could get if Apple would only use the components we'd prefer to have.
Why would they do that? IIRC, some OCL benches had the 9400m perform tasks as fast as an 8 core Xenon. ( I'll see if I can find the link later).
Don't bother as you have missed the point. If a developer has an app that needs acceleration via OCL and that acceleration is only viable on state of the art cards, the developer will not target the 9400M at all. What you target is very much dependent on what you app demands, and the user demands.
In the case of Fermi in is a highly optimized processor that would effectively run rings around a 9400M. Especially 64 bit floating point. If your app requires massive amounts of 64 bit FP you effectively eliminate the 9400M from the line up.
Like it or not I can see apps coming in the future that simply ignore the 9400M. Oh one more thing Bench marks are always interesting but you need to look carefully at what is being tested. many of these bench marks are for 32 bit FP as that is what the current NVida cards do well.
In fact I think this plays right into Apple's plans. I imagine Apple leveraging OCL and GCD in order to get acceptable performance out of modest hardware. That keeps them from having to 'keep up with the Jones' WRT hardware components such as gpus and cpus. It puts the emphasis on the software which could distinguish Apple from Windows.
That is a great bit of fanboism if you ask me. First; Windows supports its own acceleration framework and C?UDA runs there too. Second; this bear repeating, but OCL will only accelerate certain sorts of apps. Remember we are talking highly parallel vector like computations here. Even with the improvements seen in Fermi there are still many things that a CPU would be better at.
Of course nerds like us will play the 'what if' game, wondering about the performance we could get if Apple would only use the components we'd prefer to have.
Well there is preferred and then there is mandatory.
Features become mandatory when Apple is so far behind the eight ball that you have to recommend that people stay away until a update comes. I'm not talking hardware lust here but rather having an expectation of good value for the money.
Dave
If a developer has an app that needs acceleration via OCL and that acceleration is only viable on state of the art cards, the developer will not target the 9400M at all....
In the case of Fermi in is a highly optimized processor that would effectively run rings around a 9400M. ...
Like it or not I can see apps coming in the future that simply ignore the 9400M....
Depends on the apps that leverage OCL I guess. That remains to be seen.
Will FCP ( assuming it leverages OCL, but why wouldn't it) require a better gpu than the 9400m? Maybe. Will iMovie (again assuming it leverages OCL)? I doubt it.
Remember there are *way* more Macs with the 9400m than anything else. You target it fro your app and you've got a sizable audience to address.
...That is a great bit of fanboism if you ask me. First; Windows supports its own acceleration framework and C?UDA runs there too. Second; this bear repeating, but OCL will only accelerate certain sorts of apps. Remember we are talking highly parallel vector like computations here. Even with the improvements seen in Fermi there are still many things that a CPU would be better at.
Its not fanboism, its reality. When have Apple been aggressive at deploying state of the art gpus? Its been an area of weakness for as long as I can remember. If you think that OCL will be the catalyst for Apple to begin putting high end gpus in all the Macs, I think you're going to be very disappointed. Would I like Apple to use higher performance gpus in Macs? Sure but I'm not holding my breath. I predict that the mini and entry levle iMacs will use integrated graphics. The mid level and best iMacs will use a mid range card adn the Mac Pro will have the BTO option for a high end gpu card. Pretty much like it is now.
Sure Windows has CUDA. But for developers its not much of an attraction. It isn't included with the system os and it only works on NVIDIA gpus. Its not a easy market for developers to target their app at. OCL addresses some of that but even still it remains to be seen if developers will find OCL worth their time to target their apps at.
Depends on the apps that leverage OCL I guess. That remains to be seen.
Exactly, some OpenCL based apps could be very interesting, if the move down from the world of clusters, mainframes and other high performance computers. At the same time run of the mill graphics programs could benefit from OpenCL acceleration. From my perspective high performance, critcal apps are going to target the latest Nvidia GPUs simply because they support the features needed by the developers and users.
Nvidias Fermi for example has drawn gov. lab interest due to features like ECC, vastly improved 64 bit support and other goodies. Now I don't expect people to simulate aging of their stockpiles on their desktop machines, but the latest cards do enable things that have been very difficult to do on a desktop in the past.
So what I'm really saying is that OpenCL opens up the possibility of some very impressive apps moving to the desktop.
Will FCP ( assuming it leverages OCL, but why wouldn't it) require a better gpu than the 9400m? Maybe. Will iMovie (again assuming it leverages OCL)? I doubt it.
Yep many exist apps will get a nice boost even on a 9400m. On the other hand Apple could add features to FCP that might require a better GPU than the 9400M. Remember I'm not saying the 9400M is worthless, I'm just saying that there are whole classes of apps that developers won't even try to get to run on the 9400M.
Remember there are *way* more Macs with the 9400m than anything else. You target it fro your app and you've got a sizable audience to address.
Yes this is very true but it makes no difference, some apps are just going to require more horse power than the 9400M can deliver. It is a fine point, but people in general have to recognize there are hardware limitations even with the new tech.
Its not fanboism, its reality. When have Apple been aggressive at deploying state of the art gpus? Its been an area of weakness for as long as I can remember. If you think that OCL will be the catalyst for Apple to begin putting high end gpus in all the Macs, I think you're going to be very disappointed.
I never said that! Most of Apples product lineup doesn't even provide for that.
Would I like Apple to use higher performance gpus in Macs? Sure but I'm not holding my breath. I predict that the mini and entry levle iMacs will use integrated graphics. The mid level and best iMacs will use a mid range card adn the Mac Pro will have the BTO option for a high end gpu card. Pretty much like it is now.
Honestly that depends. It depends on the CPUs choosen some of which don't have provision for integrated GPUs. In a way Intel could end up forcing Apple to use better GPUs or at least not the integrated.
Sure Windows has CUDA. But for developers its not much of an attraction. It isn't included with the system os and it only works on NVIDIA gpus. Its not a easy market for developers to target their app at. OCL addresses some of that but even still it remains to be seen if developers will find OCL worth their time to target their apps at.
Dave