Let's suppose they did make 21.5" 4k iMac. What would be the minimum to run it well? We know Apple is absolute shit on Rev. A products.
It wouldn't be rev A any more because Apple is now shipping plenty of high resolution screens.
As for the minimum to run it well, that really can't be answered right now. First my opinion of what is well will vary from everybody else's. The bigger issue is that things like El Capitan demonstrate that Apple has a lot of potential for optimization. Even with the debugging code and real bugs it is still a better performing distribution than Yosemite so that right there gives me warm feelings about what might actually ship. Lastly Apple has a huge number of GPU options to choose from.
The biggest boner for me is adoption by professional apps, I think it will be much harder for Apple to get CAD and other visualization software running on Metal. That industry is heavily weighted towards OpenGL. In fact the players in the professional apps sphere have a lot of pull when it comes to the OpenGL standard. This may be why the committee has not made progress with enhancing OpenGL.
Metal does make you wonder what will happen to OpenGL on the Mac. I'd hate to see Apple drop it or put supporting it on the back burner.
I didn't realize I never submitted this.
Khronos still lists Apple as a contributor for Vulkan, which started a long time before it took on the name Vulkan, and before Metal was announced. Looking at the history there, it seems like they couldn't reach a consensus last time they tried to banish all legacy code. In this case it seems like AMD contributed a large amount of their work from Mantle. If that works out and OSX support materializes, it should draw some amount of CAD developers, especially as Vulkan will in fact support GLSL. Depending on their age, CAD apps are likely to be a combination of GLSL and C. There are also some applications that contain massive amounts of legacy code written by developers that have moved on or now defunct companies. I'm not sure we'll see much from them. For example NVidia bought out Mental Images and basically dismantled it, yet they still maintain Mental Ray. Part of the problem there was the loss of developers, which happened some time before NVidia them out.
It seems like the gpu has outscaled the cpu in performance gains. There are various state validations and things that suck up a lot of cpu overhead in the older OpenGL, and pre-compiling shaders is long overdue. Apple's OpenGL support has been terrible for a long time, so I personally hope Vulkan shapes up well enough to retain their attention, assuming it isn't just a logo tacked on for PR reasons.
Quote:
Originally Posted by wizard69
The bigger issue is that things like El Capitan demonstrate that Apple has a lot of potential for optimization. Even with the debugging code and real bugs it is still a better performing distribution than Yosemite so that right there gives me warm feelings about what might actually ship.
very interesting and corresponds with some stuff I've been hearing which is that key players did not want to give up on the legacy code and design in OpenGL. In other words competing factions probably had a big impact on everybody doing their own thing, thus Mattle, metal,and other new 3D library solutions.
Khronos still lists Apple as a contributor for Vulkan, which started a long time before it took on the name Vulkan, and before Metal was announced. Looking at the history there, it seems like they couldn't reach a consensus last time they tried to banish all legacy code. In this case it seems like AMD contributed a large amount of their work from Mantle. If that works out and OSX support materializes, it should draw some amount of CAD developers, especially as Vulkan will in fact support GLSL. Depending on their age, CAD apps are likely to be a combination of GLSL and C. There are also some applications that contain massive amounts of legacy code written by developers that have moved on or now defunct companies. I'm not sure we'll see much from them. For example NVidia bought out Mental Images and basically dismantled it, yet they still maintain Mental Ray. Part of the problem there was the loss of developers, which happened some time before NVidia them out.
This makes me wonder if metal supports a GLSL layer. I honestly don't research this in depth.
It seems like the gpu has outscaled the cpu in performance gains. There are various state validations and things that suck up a lot of cpu overhead in the older OpenGL, and pre-compiling shaders is long overdue. Apple's OpenGL support has been terrible for a long time, so I personally hope Vulkan shapes up well enough to retain their attention, assuming it isn't just a logo tacked on for PR reasons.
GPUs just keep getting better too. AMD just announce chips with 4000 or so stream processors and incredible floating point performance numbers. The only real problem is that GPUs can't effectively run the branchy code common in most apps.
I'm curious, where did you find this information?
I've been running El Capitan since debut, on a MBP 13". It most definitely feels snappier. There are bugs of course but they have obviously addressed some issues in Yosemite that caused system and GUI performance issues.
What is notable here is that the MBP was new with only a few months of use on it, so it isn't like there was a lot of cruft on the machine slowing it down. It just feels all around faster, at least for the way I use the machine. A lot of those inexplicable beach balls seem to be gone too. If this version firms up the well I suspect people will be very happy with ElCapitan. I wouldn't suggest wide spread usage yet but it is an interesting early beta for sure.
I really don't know but at this point wouldn't expect it to have a 4 lane PCI-Express controller. If you want that most likely you will have to wait for the next rev. At the moment I can't even recall a decent technical review of the recent iMacs.
A quick edit:
Over on Mac Rumors they are reprint on information suggesting a 4K iMac is coming. The evidence is skimpy but is coming from Beta 2 of ElCapitan. If Apple waits for SkyLake the new version could be sometime off in the future. However there are a couple of Broadwell processors out now that might be suitable.
As it is I really wish that Apple would get off its sealed machine mantra and deliver iMacs with ports for easy internal SSD additions. Well additions and replacements. That is supply us with iMacs with at least two ports for the SSD's and make them easy to get too. Oh and open up the port specs so that additional SSD's can be supplied by third party suppliers. Wishful thinking I know but I'm still a strong believe the only right place for storage expansion is inside the desktops box.
That's a strange way to word it. Usually they're identified by their bandwidth. You could in fact place an SSD that would saturate 4 lanes on a 1 or 2 lane bus with pin compatibility (actually I'm not sure that's the right term here). You just wouldn't obtain the same bandwidth.
Quote:
Originally Posted by wizard69
This makes me wonder if metal supports a GLSL layer. I honestly don't research this in depth.
I am not sure, but Khronos certainly implies that Apple will support Vulkan. I was surprised that they brought Metal to OSX, because the iOS version made it sound like it was designed for their hardware designs without regard for others. Apple basically uses their stuff in all of the idevices, so this seemed reasonable to me.
Quote:
GPUs just keep getting better too. AMD just announce chips with 4000 or so stream processors and incredible floating point performance numbers. The only real problem is that GPUs can't effectively run the branchy code common in most apps.
They're strongest with highly parallel operations. There are various libraries for matrix algebra on gpus, because they initially gained traction for scientific computing. Even ignoring the branchy code, they aren't very effective for anything with a lot of sequential dependencies of any kind. Their strength is merely in the ability to run many simultaneous operations. The reduction in overhead sounds like it's removing a lot of state validation and other checks from the cpu. It's not like that stuff ever worked perfectly though. I haven't tried developing for multiple platforms, but there are many complaints that timing issues and differing implementations can make it problematic to deal with this as it is. The cpu overhead is just an additional tax, and I still don't understand why just in time shader compilation is still a thing. It should have gone away with the fixed function pipeline.
Quote:
I've been running El Capitan since debut, on a MBP 13". It most definitely feels snappier. There are bugs of course but they have obviously addressed some issues in Yosemite that caused system and GUI performance issues.
What is notable here is that the MBP was new with only a few months of use on it, so it isn't like there was a lot of cruft on the machine slowing it down. It just feels all around faster, at least for the way I use the machine. A lot of those inexplicable beach balls seem to be gone too. If this version firms up the well I suspect people will be very happy with ElCapitan. I wouldn't suggest wide spread usage yet but it is an interesting early beta for sure.
I saw quite a few of those in Yosemite, more than previous revisions. I certainly won't miss them. Thanks for the good news.
That's a strange way to word it. Usually they're identified by their bandwidth. You could in fact place an SSD that would saturate 4 lanes on a 1 or 2 lane bus with pin compatibility (actually I'm not sure that's the right term here). You just wouldn't obtain the same bandwidth.
My guess is his goal is some sort of obsolescence prevention. That would be fine if Apple released the port details so that others could market compatible SSD's for the iMacs (really all Apple products). Right now it doesn't make much difference because there is little int he way of third party SSD's for the new Mac PCI Express supported machines.
They're strongest with highly parallel operations. There are various libraries for matrix algebra on gpus, because they initially gained traction for scientific computing. Even ignoring the branchy code, they aren't very effective for anything with a lot of sequential dependencies of any kind. Their strength is merely in the ability to run many simultaneous operations.
It is unfortunate but many see GPU's as extra CPU's available for all sorts of acceleration. This is hardly the case as they at best are special purpose processors.
The reduction in overhead sounds like it's removing a lot of state validation and other checks from the cpu. It's not like that stuff ever worked perfectly though. I haven't tried developing for multiple platforms, but there are many complaints that timing issues and differing implementations can make it problematic to deal with this as it is. The cpu overhead is just an additional tax, and I still don't understand why just in time shader compilation is still a thing. It should have gone away with the fixed function pipeline.
My understanding, limited as it is, is the tJIT shaders mean shaders compiled to support a specific GPU.
I saw quite a few of those in Yosemite, more than previous revisions. I certainly won't miss them. Thanks for the good news.
Mac OS seems to be ahead of iOS at this point. IOS is a mess in the current Beta, there are so many glaring GUI faults that I haven't even bothered to write bug reports. I'm now calling Mac OS Swiftier in honor of their new language.
Is Iris 7100 going to be part of Skylake can we assume? Iris Pro 7200 as well.
SkyLake gets a new and improved GPU subsection. What it will be called I don't know off the top of my head. Leaks suggest a significant performance increase.
I really loved the product because of its features and the price range. I had wait a lot for this product arrival in the market and now i own one piece of mac mini.
SkyLake gets a new and improved GPU subsection. What it will be called I don't know off the top of my head. Leaks suggest a significant performance increase.
And Kaby Lake (at least the H-processors) seem to be a huge jump from that!
You may well be waiting forever. Think about how long it took for that damn thing to get haswell.
Very valid point! I was really hoping to make it to SkyLake before my old machine died. So going out and buying a new MBP 13" felt like a bit of a disappointment but after having the machine for awhile I'm pleseantly surprised. The performance is pretty good most of the time though I would love to have quad cores at times. Still overall performance is many times better than the old machine from 2008!
If somebody came up to me and said they are thinking about upgrading an old machine, from 2008 or before, I'd say go for it without hesitation. If nothing else Apples Solid State Drives give Mac OS an entirely different feel over the old laptop drive technologies. The CPUs are also far better performing.
You may well be waiting forever. Think about how long it took for that damn thing to get haswell.
I'm getting either an iMac or Asus Zen AiO. The Mac mini ship has just about sailed for me. I was just making a point about graphics and how well Intel is doing.
Very valid point! I was really hoping to make it to SkyLake before my old machine died. So going out and buying a new MBP 13" felt like a bit of a disappointment but after having the machine for awhile I'm pleseantly surprised. The performance is pretty good most of the time though I would love to have quad cores at times. Still overall performance is many times better than the old machine from 2008!
Well yeah it's to be expected. You have passed over generations of hardware on each component. Lately I'm not always sure what updates Apple will or won't use. I'm a little surprised they skipped over Haswell EP on the mac pro. I don't really have a reason to be concerned about that anymore, but intel switches chipsets on that line once every other cycle. Haswell EP would be one. Broadwell EP will use the same thing. Anything after that will use a different socket again, or at least they won't be electrically compatible. With the mini it has gotten a little silly just how long they hold off on updates. It wasn't really a couple months after the notebooks.
Quote:
Originally Posted by Winter
I'm getting either an iMac or Asus Zen AiO. The Mac mini ship has just about sailed for me. I was just making a point about graphics and how well Intel is doing.
I know you occasionally mention games. I imagine even the latest mini might be a bit disappointing in that regard. If they represented a headless 15" and kept up a bit better, it wouldn't be as bad.
Well yeah it's to be expected. You have passed over generations of hardware on each component.
Without a doubt the SSD is a big factor in the machines usability as is the RAM. I was pleasantly surprised that HomeBrew could saturate the machine with four threads while updating some apps and the machine remained very usable running a video. Not fast or super responsive but at least usable.
Lately I'm not always sure what updates Apple will or won't use. I'm a little surprised they skipped over Haswell EP on the Mac Pro.
I wouldn't say skipped over yet. I don't expect to see a huge update until AMD can supply them with the new HBM GPUs. They might even wait for 14 nm XEONs and probably TB3 support on the new USB interface standard. It could be 2016 before a significantly upgraded Mac Pro is seen.
I don't really have a reason to be concerned about that anymore, but intel switches chipsets on that line once every other cycle. Haswell EP would be one. Broadwell EP will use the same thing. Anything after that will use a different socket again, or at least they won't be electrically compatible.
Intel is really taking advantage of its leadership position in server chips. This has resulted in very high prices and dragging out new chip releases.
I often wonder if Apple would have Intel do a custom XEON like they have for others. Last I heard Intel had 34 custom spins of XEON in the works for different companies. Of course I'm not sure what Apple would ask for here but it might land them better suited chips or chipsets. Ideally the custom chip would offer a few more PCI Express lanes to allow for dua ,SSDs and the controllers for the new style combo USB ports.
With the mini it has gotten a little silly just how long they hold off on updates. It wasn't really a couple months after the notebooks.
Apples behavior with the Mini is hard to understand but part of it is Intel not providing the optimal hardware.
I know you occasionally mention games. I imagine even the latest mini might be a bit disappointing in that regard. If they represented a headless 15" and kept up a bit better, it wouldn't be as bad.
Apples behavior with the Mini is hard to understand but part of it is Intel not providing the optimal hardware.
No, it's not hard to understand. Not at all.
Apple hates the Mini. Or, at the very least, Cupertino isn't really fond of it.
Apple's culture believes that not controlling the entire user experience almost killed the company. They see headless desktops as a legacy of the pre-iMac days where the beige LC, Performa and Power Macs made the desktop computer a style-less commodity. A Mac desktop user without a built-in screen only contributes a fraction of the margin to Apple's bottom line, and then does so again when they upgrade their half-a-machine four years later.
This isn't about available chips. It never has been. Apple has all the components it needs right now to build a Mini that rivals or surpasses the iMac. They simply refuse to do so, and their business logic is sound. The availability of the Mini was primarily about switchers to the platform at the expense of Windows, and offering a price point that gets cost conscious buyers into an Apple Store.
Once you're in the Store, Apple goes all out to convince you not to buy the Mini. Sales people point out it's underpowered. The lack of an Apple monitor to accompany it (despite the fact that Apple has easy access to 21" panels) means that Mini buyers will need a third party monitor, PLUS FaceTime camera PLUS good speakers PLUS keyboard and mouse. The salesperson will point out that all this puts the 21" iMac within striking distance, and most buyers will accept the option. In a world where the Mini is no longer upgradable, why would you even object?
This is why I've focused on asking that the iMacs gain some upgradability with respect to RAM and HD/SSD. The Mini has been out for more than ten years now and to keep waiting for it to suddenly become a respectable midrange Mac is a fool's errand.
I believe it has served its purpose, and Apple should terminate the Mac Mini line (or shrink it further for embedded applications only) and drop the entry level iMac to an appropriate level to compensate.
Comments
It wouldn't be rev A any more because Apple is now shipping plenty of high resolution screens.
As for the minimum to run it well, that really can't be answered right now. First my opinion of what is well will vary from everybody else's. The bigger issue is that things like El Capitan demonstrate that Apple has a lot of potential for optimization. Even with the debugging code and real bugs it is still a better performing distribution than Yosemite so that right there gives me warm feelings about what might actually ship. Lastly Apple has a huge number of GPU options to choose from.
The biggest boner for me is adoption by professional apps, I think it will be much harder for Apple to get CAD and other visualization software running on Metal. That industry is heavily weighted towards OpenGL. In fact the players in the professional apps sphere have a lot of pull when it comes to the OpenGL standard. This may be why the committee has not made progress with enhancing OpenGL.
Metal does make you wonder what will happen to OpenGL on the Mac. I'd hate to see Apple drop it or put supporting it on the back burner.
I didn't realize I never submitted this.
Khronos still lists Apple as a contributor for Vulkan, which started a long time before it took on the name Vulkan, and before Metal was announced. Looking at the history there, it seems like they couldn't reach a consensus last time they tried to banish all legacy code. In this case it seems like AMD contributed a large amount of their work from Mantle. If that works out and OSX support materializes, it should draw some amount of CAD developers, especially as Vulkan will in fact support GLSL. Depending on their age, CAD apps are likely to be a combination of GLSL and C. There are also some applications that contain massive amounts of legacy code written by developers that have moved on or now defunct companies. I'm not sure we'll see much from them. For example NVidia bought out Mental Images and basically dismantled it, yet they still maintain Mental Ray. Part of the problem there was the loss of developers, which happened some time before NVidia them out.
It seems like the gpu has outscaled the cpu in performance gains. There are various state validations and things that suck up a lot of cpu overhead in the older OpenGL, and pre-compiling shaders is long overdue. Apple's OpenGL support has been terrible for a long time, so I personally hope Vulkan shapes up well enough to retain their attention, assuming it isn't just a logo tacked on for PR reasons.
The bigger issue is that things like El Capitan demonstrate that Apple has a lot of potential for optimization. Even with the debugging code and real bugs it is still a better performing distribution than Yosemite so that right there gives me warm feelings about what might actually ship.
I'm curious, where did you find this information?
I've been running El Capitan since debut, on a MBP 13". It most definitely feels snappier. There are bugs of course but they have obviously addressed some issues in Yosemite that caused system and GUI performance issues.
What is notable here is that the MBP was new with only a few months of use on it, so it isn't like there was a lot of cruft on the machine slowing it down. It just feels all around faster, at least for the way I use the machine. A lot of those inexplicable beach balls seem to be gone too. If this version firms up the well I suspect people will be very happy with ElCapitan. I wouldn't suggest wide spread usage yet but it is an interesting early beta for sure.
I really don't know but at this point wouldn't expect it to have a 4 lane PCI-Express controller. If you want that most likely you will have to wait for the next rev. At the moment I can't even recall a decent technical review of the recent iMacs.
A quick edit:
Over on Mac Rumors they are reprint on information suggesting a 4K iMac is coming. The evidence is skimpy but is coming from Beta 2 of ElCapitan. If Apple waits for SkyLake the new version could be sometime off in the future. However there are a couple of Broadwell processors out now that might be suitable.
As it is I really wish that Apple would get off its sealed machine mantra and deliver iMacs with ports for easy internal SSD additions. Well additions and replacements. That is supply us with iMacs with at least two ports for the SSD's and make them easy to get too. Oh and open up the port specs so that additional SSD's can be supplied by third party suppliers. Wishful thinking I know but I'm still a strong believe the only right place for storage expansion is inside the desktops box.
Does the retina iMac have 4 lane PCIe SSD?
That's a strange way to word it. Usually they're identified by their bandwidth. You could in fact place an SSD that would saturate 4 lanes on a 1 or 2 lane bus with pin compatibility (actually I'm not sure that's the right term here). You just wouldn't obtain the same bandwidth.
This makes me wonder if metal supports a GLSL layer. I honestly don't research this in depth.
I am not sure, but Khronos certainly implies that Apple will support Vulkan. I was surprised that they brought Metal to OSX, because the iOS version made it sound like it was designed for their hardware designs without regard for others. Apple basically uses their stuff in all of the idevices, so this seemed reasonable to me.
Quote:
They're strongest with highly parallel operations. There are various libraries for matrix algebra on gpus, because they initially gained traction for scientific computing. Even ignoring the branchy code, they aren't very effective for anything with a lot of sequential dependencies of any kind. Their strength is merely in the ability to run many simultaneous operations. The reduction in overhead sounds like it's removing a lot of state validation and other checks from the cpu. It's not like that stuff ever worked perfectly though. I haven't tried developing for multiple platforms, but there are many complaints that timing issues and differing implementations can make it problematic to deal with this as it is. The cpu overhead is just an additional tax, and I still don't understand why just in time shader compilation is still a thing. It should have gone away with the fixed function pipeline.
I've been running El Capitan since debut, on a MBP 13". It most definitely feels snappier. There are bugs of course but they have obviously addressed some issues in Yosemite that caused system and GUI performance issues.
What is notable here is that the MBP was new with only a few months of use on it, so it isn't like there was a lot of cruft on the machine slowing it down. It just feels all around faster, at least for the way I use the machine. A lot of those inexplicable beach balls seem to be gone too. If this version firms up the well I suspect people will be very happy with ElCapitan. I wouldn't suggest wide spread usage yet but it is an interesting early beta for sure.
I saw quite a few of those in Yosemite, more than previous revisions. I certainly won't miss them. Thanks for the good news.
Mac OS seems to be ahead of iOS at this point. IOS is a mess in the current Beta, there are so many glaring GUI faults that I haven't even bothered to write bug reports. I'm now calling Mac OS Swiftier in honor of their new language.
SkyLake gets a new and improved GPU subsection. What it will be called I don't know off the top of my head. Leaks suggest a significant performance increase.
I really loved the product because of its features and the price range. I had wait a lot for this product arrival in the market and now i own one piece of mac mini.
And Kaby Lake (at least the H-processors) seem to be a huge jump from that!
And Kaby Lake (at least the H-processors) seem to be a huge jump from that!
You may well be waiting forever. Think about how long it took for that damn thing to get haswell.
Very valid point! I was really hoping to make it to SkyLake before my old machine died. So going out and buying a new MBP 13" felt like a bit of a disappointment but after having the machine for awhile I'm pleseantly surprised. The performance is pretty good most of the time though I would love to have quad cores at times. Still overall performance is many times better than the old machine from 2008!
If somebody came up to me and said they are thinking about upgrading an old machine, from 2008 or before, I'd say go for it without hesitation. If nothing else Apples Solid State Drives give Mac OS an entirely different feel over the old laptop drive technologies. The CPUs are also far better performing.
I'm getting either an iMac or Asus Zen AiO. The Mac mini ship has just about sailed for me. I was just making a point about graphics and how well Intel is doing.
Very valid point! I was really hoping to make it to SkyLake before my old machine died. So going out and buying a new MBP 13" felt like a bit of a disappointment but after having the machine for awhile I'm pleseantly surprised. The performance is pretty good most of the time though I would love to have quad cores at times. Still overall performance is many times better than the old machine from 2008!
Well yeah it's to be expected. You have passed over generations of hardware on each component. Lately I'm not always sure what updates Apple will or won't use. I'm a little surprised they skipped over Haswell EP on the mac pro. I don't really have a reason to be concerned about that anymore, but intel switches chipsets on that line once every other cycle. Haswell EP would be one. Broadwell EP will use the same thing. Anything after that will use a different socket again, or at least they won't be electrically compatible. With the mini it has gotten a little silly just how long they hold off on updates. It wasn't really a couple months after the notebooks.
I'm getting either an iMac or Asus Zen AiO. The Mac mini ship has just about sailed for me. I was just making a point about graphics and how well Intel is doing.
I know you occasionally mention games. I imagine even the latest mini might be a bit disappointing in that regard. If they represented a headless 15" and kept up a bit better, it wouldn't be as bad.
I often wonder if Apple would have Intel do a custom XEON like they have for others. Last I heard Intel had 34 custom spins of XEON in the works for different companies. Of course I'm not sure what Apple would ask for here but it might land them better suited chips or chipsets. Ideally the custom chip would offer a few more PCI Express lanes to allow for dua ,SSDs and the controllers for the new style combo USB ports. Apples behavior with the Mini is hard to understand but part of it is Intel not providing the optimal hardware.
Apples behavior with the Mini is hard to understand but part of it is Intel not providing the optimal hardware.
No, it's not hard to understand. Not at all.
Apple hates the Mini. Or, at the very least, Cupertino isn't really fond of it.
Apple's culture believes that not controlling the entire user experience almost killed the company. They see headless desktops as a legacy of the pre-iMac days where the beige LC, Performa and Power Macs made the desktop computer a style-less commodity. A Mac desktop user without a built-in screen only contributes a fraction of the margin to Apple's bottom line, and then does so again when they upgrade their half-a-machine four years later.
This isn't about available chips. It never has been. Apple has all the components it needs right now to build a Mini that rivals or surpasses the iMac. They simply refuse to do so, and their business logic is sound. The availability of the Mini was primarily about switchers to the platform at the expense of Windows, and offering a price point that gets cost conscious buyers into an Apple Store.
Once you're in the Store, Apple goes all out to convince you not to buy the Mini. Sales people point out it's underpowered. The lack of an Apple monitor to accompany it (despite the fact that Apple has easy access to 21" panels) means that Mini buyers will need a third party monitor, PLUS FaceTime camera PLUS good speakers PLUS keyboard and mouse. The salesperson will point out that all this puts the 21" iMac within striking distance, and most buyers will accept the option. In a world where the Mini is no longer upgradable, why would you even object?
This is why I've focused on asking that the iMacs gain some upgradability with respect to RAM and HD/SSD. The Mini has been out for more than ten years now and to keep waiting for it to suddenly become a respectable midrange Mac is a fool's errand.
I believe it has served its purpose, and Apple should terminate the Mac Mini line (or shrink it further for embedded applications only) and drop the entry level iMac to an appropriate level to compensate.