nVidia and AMD could offer Thunderbolt solutions if they wanted to (knowing their history with Intel, they probably won't though) or maybe someone can hack the drivers of currently supported cards.
I've never been a fan of hacked drivers. I kind of wonder if there's really a market for them. Will they sell on Windows? If not you're unlikely to see it go anywhere on OSX where we've always lived with limited card options. It's just that previously fewer things were gpu reliant. Overall turning a laptop more and more into a crippled device when it isn't docked makes little sense to me. At some point it becomes more practical to own an Air + beefier machine if the enclosure + card cost too much. Another thing is that thunderbolt is slower than a x16 connection. We're migrating toward PCIe 3.0 at this point with 16 lanes allocated to a gpu. This doesn't mean they will be fully saturated. It's just that if it could benefit past 8, 16 is the next step. Considering the moderate choke factor and that I have yet to see an external chassis higher than 150W, you're not going to see anything incredible in terms of price to performance ratio. Even the chassis is expendable. If thunderbolt speeds increase, you will require a new chassis. The transition toward quasi fiber cables this year is more about alleviating length restrictions than overall speed.
Quote:
Originally Posted by Marvin
Would you really use a Thunderbolt RAID controller externally in that case? You'd surely connect Thunderbolt to something like fibre-channel and connect it to a server or storage that had internal hardware RAID. This is server tech and is not primarily Apple's domain any more.
It depends on the goals there. ATTO is a reasonable solution for something like an ROC compatible with a mac pro. It's definitely a better solution than the piece of garbage Apple turned into an option. If they aren't going to do something well, or support it, they should leave such things to third parties rather than offer a cto product that is that bad.
"Quad-core mobile Sandy Bridge, 2.5" SSDs and Thunderbolt together have allowed me to use a notebook as my primary work machine. I get all of the portability benefits of a notebook, but with almost none of the performance sacrifices. The only thing I'm really missing is good, external discrete GPU solution but that's a problem being worked on either via Thunderbolt link aggregation or the second revision of the Thunderbolt spec.
People just assume that because the slots are there and are a certain speed that someone must be using that speed. Data IO isn't high bandwidth as it depends on the source and destination read/write. Real-time graphics is high bandwidth but you can get away with a single high-end card in an x16 slot and many display outputs and they still work ok in x4 slots.
There are compatible solutions for eveything you need to use PCI for. The driver issue probably has something to do with the fact that Thunderbolt is supposed to be plug and play. Windows doesn't have plug and play support yet - you have to have Thunderbolt peripherals plugged in at boot time.
Plug and play is a big advantage of Thunderbolt over PCIe slots. Musicians can take a laptop somewhere and plug it in without rebooting. Unplug and they have everything there to edit on the go.
Some peripherals are expensive. The Sonnet PCI box is cheaper despite the cheap one only having 1 PCI slot but people aren't questioning the $4000 price tag on a capture card. The ideal is not that you'd use an external PCI slot, the ideal is that manufacturers put a Thunderbolt port on their products, the PCI box is a backup.
It will be hard for some to accept but the potential Thunderbolt market size is much bigger than the workstation market. It won't happen overnight but now that some PC manufacturers are getting on board, it will get things moving a bit faster. Manufacturers can make nice products instead of raw boards.
The tech is only about 15 months old. This always happens with new developments. People who want to stick to the old ways try and put it down. It's not just that it's inadequate, people just don't want to face the prospect of towers going away so attempt to pick out essential items that mean it can never happen. PCI slots were developed to give people fast enough expansion on the widest selling machines - demand + volume = profit. Desktop towers are no longer the widest selling machines by a long way and the people who need PCI cards much less than that. The demand + volume is now in the mobile sector and that is the domain of Thunderbolt, that is something people are going to have to come to terms with.
Marvin. I'd suggest you work as a professional audio engineer for over 15 years like I have, before you start commenting on it. Believe it or not I actually started my career when we were still using analogue tapes, so I've seen the changes and felt the changes. I'm not talking of home studios or project studios here, I've seen projects (movies) that had over 700 tracks of audio alone, not including the music tracks. Those were spread over 4 systems using satellites options (2x HD 6 rigs, and 2x native rigs for a total of 14 pci cards). There's isn't a damn way that your supposed thunderbolt can do it all will be able to accommodate the needs. It's fine for small home studios for musicians and all, but you're forgetting the actual professional studios and post production houses.
it's not because a couple of audio cards and one DSP card work in a Thunderbolt environment that the pro audio market is covered.
There's more than one DSP card in the list but you could say the same about devices that are Windows-exclusive. If the Mac Pro only has a 5770 and a 5870, does that mean it's not suitable for high-end graphics work?
There are Thunderbolt solutions targeting everything you use PCI for. While not all the cards have Thunderbolt counterparts, there are solutions for everything with the exception of GPUs on the Mac side but the iMac GPU is fast anyway so that's still a solution.
You try to make the case that Apple couldn't discontinue the Pro because there wouldn't be workable solutions for every scenario but there clearly are solutions that will be available in the timeframe that such a discontinuation would affect.
Wow. Even Intel didn't know that! The domain of Thunderbolt is the mobile sector! You're confusing consumer (and even adding the prosumer) market with the professional market. Nobody in is right mind will use a laptop and a bunch of external boxes in a recording studio (or any professional setting) when a Mac Pro offers a more powerful, elegant, reliable solution. Not everything makes sense/needs to be mobile/portable. Would your first choice for a data server be a laptop?
In any case, my impression was that demand + volume was in the smartphone and tablet sector.
I think you're keeping the professional market unreasonably restricted to what you want the term 'professional' to represent. But hey, don't let these consumers at NAB tell you different:
Notice the optical Thunderbolt cable in there too, connected to a Macbook Pro.
So that's your solution: hacking the drivers... how professional.
Not necessarily hacking the driver binaries but a way to get them to load. The drivers work without modification on Windows, why should it be any different on the Mac side? It must be to do with how the drivers are being loaded.
I've been a huge fan of Thunderbolt since the early days
You say you are behind Thunderbolt but then add 'if only it was better'. If you are truly behind something then you don't try to knock it down at every turn. Your stance is clear, you want the tower and PCI slots and anything else is not for 'professionals'. I think those things will go away soon and professionals will be defined by their work not the size of their box.
No this isn't about EFI or driver issues, apparently if you try to use any GPU, 100% supported by Apple, in a Thunderbolt to PCIe enclosure it will not work.
The Mac GPU drivers won't work because they don't know how to work over Thunderbolt. Apple develops those but AMD/NVidia could write their own if they wanted.
I guess that answers my question, no you wouldn't use a RAID controller in one of those Thunderbolt cases...
But I'm not referring to shared storage, but fast RAID storage for one workstation with a dedicated controller, again there doesn't seem to be a solution available?
The Pegasus RAID is a solution - you manage the hardware controller from the utility:
When you say fast storage, you can put 2x Samsung 830s in a Mini and it will be faster than most hardware RAID drives:
Sure if you need 12TB of fast storage, you go for the Pegasus but it won't be much faster as platter drives are inherently slow and you're not going to put 8 platter drives in RAID0.
I've seen projects (movies) that had over 700 tracks of audio alone not including the music tracks. Those were spread over 4 systems using satellites options (2x HD 6 rigs, and 2x native rigs for a total of 16 pci cards). There's isn't a damn way that your supposed thunderbolt can do it all will be able to accommodate the needs. It's fine for small home studios for musicians and all, but you're forgetting the actual professional studios and post production houses.
If you're hooking up multiple machines, you can hook up multiple Thunderbolt options too like iMacs, Minis or Macbook Pros:
iMacs have two Thunderbolt ports so two HDX each with say 100 tracks per box so 4 iMacs hooked up over ethernet. More maybe or less if you aren't bandwidth-limited and put 3 cards per Magma box? I don't see how it's an impossible usage scenario.
My preferred solution for the Pro is not discontinuation but a smaller box with 6x Thunderbolt ports (obviously one used for the display, possibly chained) and you would be able to connect loads of dedicated hardware to a single machine, in a standard Pro, you get 3 slots free. Although they are higher bandwidth, if you need the cards for processing, you get more with the TB ports.
The way I see it is the 5.25" bay wastes a lot of space and it costs money to put in there.
The dual-processors obviously offer the best performance so barring some other way of connecting machines up, it would make sense to offer the DP models and that gives the 8 RAM slots.
I'm not sure if Sandy Bridge Mac Pros will support USB 3 but I imagine they could somehow and it's certainly in their best interests to support it.
3rd party manufacturers seem to be putting Thunderbolt on desktop motherboards alongside PCI slots but it's not clear how they are connecting up to GPUs. A poster on the following forum says Apple does it the following way:
"For anyone who cares, Apple routes video signal from the dedicated card then through the port by copying the entirely processed video signal from the dedicated GPU's framebuffer into the integrated framebuffer, which dispatches the signal through the thunderbolt IC and out to the port. Early models of Macs with hybrid graphics used muxer chips that physically switched the port's wirings from one GPU to the other."
This suggests there's some driver-level manipulation to get video out through the Thunderbolt port. While it would allow GPUs to have outwardly facing ports, it means that if you used another card where the drivers don't have this functionality, it would break display output on your Thunderbolt ports. Also, it's not obvious how the system knows which GPU to use for which Thunderbolt port in a multi-GPU system.
The simpler option is the iMac design where only Thunderbolt ports go on the outside. I personally think this is the best way to go in which case, PCI slots are gone for anything but the single GPU. Some will then say that makes the Pro worthless but I think the switch has to come eventually. You can see how small the Boxx RenderPro is without PCI slots and it takes a dual Xeon. The equivalent Mac Pro would be a bit bigger with drives and GPU but not by much.
There's been a leaked photo of some MBP specs (I think dual-core Ivy Bridge arrived today) so we may see the store go down tomorrow. Worst case, it will be WWDC but they've done hardware releases before WWDC in the past.
The simpler option is the iMac design where only Thunderbolt ports go on the outside. I personally think this is the best way to go in which case, PCI slots are gone for anything but the single GPU. Some will then say that makes the Pro worthless but I think the switch has to come eventually. You can see how small the Boxx RenderPro is without PCI slots and it takes a dual Xeon. The equivalent Mac Pro would be a bit bigger with drives and GPU but not by much.
There's been a leaked photo of some MBP specs (I think dual-core Ivy Bridge arrived today) so we may see the store go down tomorrow. Worst case, it will be WWDC but they've done hardware releases before WWDC in the past.
I like reading your posts too, even if we disagree on quite a lot. I think you would need much more from thunderbolt to ditch PCI on that machine. External PCI solutions aren't so great presently, and thunderbolt chips take up a significant amount of logic board space, yet you can only get a maximum of two ports from one chip assuming the regular ones are used. Those Boxx units lack graphics cards too, and I don't see integrated graphics being a viable option for a mac pro successor without real OpenCL/OpenGL performance. In the case of the imac, they have a lot of surface area, and they actually use a fairly expensive gpu relative to its performance. Workstation pricing has gone a little crazy at comparable levels of configuration over the past few year, and I find that a little annoying. The nice thing with the big towers is that you can pretty much always make something work. Even if you need many many ports, displays, storage, dongle keys, etc. you can make it work. With many of Apple's solutions, it can be much more difficult. Ideally I'd like to just make it a laptop upgrade year if Sandy Bridge E mac pros aren't too impressive in performance gains, but I don't know if that will work out.
By the way, one of those shipping slips from the supposed leak would be easy to fake. You could create one, print it, and photograph it with a camera phone in a few minutes. This would be an easy way to make something convincing in appearance. It's also not difficult to make convincing specs. I don't think that photo really means much either way.
The real question is just how many people like that are there? How many sales would Apple lose if they killed off the Mac Pro? Apple knows how many Mac Pros it sells each year and any decision is going to be based on this, not on comments on the Internet.
Apple also knows how much money the Mac Pro makes them... and as long as it's not losing money... I don't see why they would kill the Mac Pro. They still have customers.
If Apple does kill the Mac Pro... all those customers will run into the arms of HP or Dell.
Yeah... that's right... there are other companies who still make workstation-class machines.
HP makes the Z-series workstations... Dell makes the Precision and Optiplex workstations... and other companies like Boxx do too.
If it comes down to money... I bet the Mac Pro still turns a profit... albeit not as big a percentage as the iPhone and iPad. But that shouldn't matter. A profitable product is a profitable product... no matter the volume.
I bet Dell doesn't sell nearly as many Precision workstations as they do Inspiron laptops... yet they still offer those high-end machines. What's their secret?
Besides... wouldn't Apple's own software engineers prefer using a 12-core Mac Pro? Or are they gonna be happy moving from a 12-core workstation down to a quad-core iMac?
Besides... wouldn't Apple's own software engineers prefer using a 12-core Mac Pro? Or are they gonna be happy moving from a 12-core workstation down to a quad-core iMac?
Of course not, it's written in linux on a Dell workstation
Apple also knows how much money the Mac Pro makes them... and as long as it's not losing money... I don't see why they would kill the Mac Pro. They still have customers.
We can put an upper bound on the amount:
5% of all Mac models = 0.05 x 5 million per quarter = 250k per quarter @ $1000 margins = $250m
Apple's profit last quarter = $11.6b so the Mac Pro contributes at most 2% to Apple's profits now
The same reasoning applies to all the Macs though:
If it's still making money, it sustains those jobs and the Mac OS feeds into iOS - iOS developers aren't using PCs or iOS devices.
However, a lot of tasks that were previously thought to be exclusive to the Pro no longer are. The Pro fans talk about the lower Mac models getting too hot and 'burning out' or shutting down but it's a load of baloney/FUD. A great number of people use the cheaper models for high-value and high-resource (i.e professional) tasks and this is evident from their sales volume.
The Mac Pro has unquestionable design benefits such as running cool, having expansion, offering up to 3x the performance of the iMac but the lower models have the benefit of costing much less money. While high-value industry workers can sometimes afford the expense, it's not always the case nor is it essential. Some of the smartest and most talented people in the world have to work on shoe-string budgets and this can contribute to some of their best achievements. Educational funding is rarely generous enough that researchers get the best equipment. R&D happens before you start making money.
Only Apple knows what's selling and their route ahead. You can see the graph above though and the rising vectors. Like it or not, along the way, there will be casualties in Apple's line, the only question is the timing. The Xserve was the first but it won't be the last.
If Apple does kill the Mac Pro... all those customers will run into the arms of HP or Dell.
I don't think they'd all jump ship. Many quad and maybe even 6-core owners would go to the iMacs.
Instead of killing it like the XServe, I suspect they'd ramp it down. Cull the models that people aren't buying - I personally suspect the Server model is for the chopping block - and just do that as time goes on.
Only Apple knows what's selling and their route ahead. You can see the graph above though and the rising vectors. Like it or not, along the way, there will be casualties in Apple's line, the only question is the timing. The Xserve was the first but it won't be the last.
I don't think they'd all jump ship. Many quad and maybe even 6-core owners would go to the iMacs.
Instead of killing it like the XServe, I suspect they'd ramp it down. Cull the models that people aren't buying - I personally suspect the Server model is for the chopping block - and just do that as time goes on.
Right now they have a lot of odd configurations. The 8 core as an example is slower than the 6 in many workloads. If they consolidated but used the streamlined setup to keep the offered configurations competitive, that would work well. There is no true sever model. It's basically OSX server + second hard drive and 8GB of ram. You can configure that from the normal one, so cutting it really doesn't affect anyone as you can still buy the same configuration without the moniker. You can check for yourself. There's nothing unique about it beyond what I mentioned.
By the way, that's kind of a troll article as it only considers direct contributions to the bottom line. It doesn't consider internal use by Apple employees, IOS development, or contribution to their ecosystem. Saying you'd only lose those sales by dropping it is a fallacy, and they'd incur additional costs supporting Windows or Linux based IOS developer tools. I'm not even painting this as something bigger than it is. I dislike the way the article is presented.
Comments
Quote:
Originally Posted by Marvin
The Mac needs specific drivers as well as EFI GPUs. If you plug a 7970 into a Mac Pro, you can't use it on the Mac side either:
http://www.tonymacx86.com/viewtopic.php?f=15&t=45671&sid=678afd2fbd8e65a9d073d2ff554feee4&start=30
nVidia and AMD could offer Thunderbolt solutions if they wanted to (knowing their history with Intel, they probably won't though) or maybe someone can hack the drivers of currently supported cards.
I've never been a fan of hacked drivers. I kind of wonder if there's really a market for them. Will they sell on Windows? If not you're unlikely to see it go anywhere on OSX where we've always lived with limited card options. It's just that previously fewer things were gpu reliant. Overall turning a laptop more and more into a crippled device when it isn't docked makes little sense to me. At some point it becomes more practical to own an Air + beefier machine if the enclosure + card cost too much. Another thing is that thunderbolt is slower than a x16 connection. We're migrating toward PCIe 3.0 at this point with 16 lanes allocated to a gpu. This doesn't mean they will be fully saturated. It's just that if it could benefit past 8, 16 is the next step. Considering the moderate choke factor and that I have yet to see an external chassis higher than 150W, you're not going to see anything incredible in terms of price to performance ratio. Even the chassis is expendable. If thunderbolt speeds increase, you will require a new chassis. The transition toward quasi fiber cables this year is more about alleviating length restrictions than overall speed.
Quote:
Originally Posted by Marvin
Would you really use a Thunderbolt RAID controller externally in that case? You'd surely connect Thunderbolt to something like fibre-channel and connect it to a server or storage that had internal hardware RAID. This is server tech and is not primarily Apple's domain any more.
It depends on the goals there. ATTO is a reasonable solution for something like an ROC compatible with a mac pro. It's definitely a better solution than the piece of garbage Apple turned into an option. If they aren't going to do something well, or support it, they should leave such things to third parties rather than offer a cto product that is that bad.
[...]
[URL=http://images.anandtech.com/reviews/motherboards/thunderbolt/traces.jpg][IMG]http://forums.appleinsider.com/image/id/166187/width/600/height/453[/IMG][/URL]
There's an impressive amount of engineering that has to go into bringing Thunderbolt support to a motherboard."
• http://www.anandtech.com/show/5884/thunderbolt-on-windows-part-2-intels-dz77rek75-asus-p8z77v-premium
Quote:
Originally Posted by Marvin
You mean using audio cards that need a single PCI lane (2.5Gbps) like this one that works in the Thunderbolt box:
http://www.rme-audio.de/en_products_hdspe_aes.php
or this x1 $5000+ DSP card:
http://www.uaudio.com/uad-plug-ins/pcie/uad-2-quad-omni-6.html
People just assume that because the slots are there and are a certain speed that someone must be using that speed. Data IO isn't high bandwidth as it depends on the source and destination read/write. Real-time graphics is high bandwidth but you can get away with a single high-end card in an x16 slot and many display outputs and they still work ok in x4 slots.
There are compatible solutions for eveything you need to use PCI for. The driver issue probably has something to do with the fact that Thunderbolt is supposed to be plug and play. Windows doesn't have plug and play support yet - you have to have Thunderbolt peripherals plugged in at boot time.
Plug and play is a big advantage of Thunderbolt over PCIe slots. Musicians can take a laptop somewhere and plug it in without rebooting. Unplug and they have everything there to edit on the go.
Some peripherals are expensive. The Sonnet PCI box is cheaper despite the cheap one only having 1 PCI slot but people aren't questioning the $4000 price tag on a capture card. The ideal is not that you'd use an external PCI slot, the ideal is that manufacturers put a Thunderbolt port on their products, the PCI box is a backup.
It will be hard for some to accept but the potential Thunderbolt market size is much bigger than the workstation market. It won't happen overnight but now that some PC manufacturers are getting on board, it will get things moving a bit faster. Manufacturers can make nice products instead of raw boards.
The tech is only about 15 months old. This always happens with new developments. People who want to stick to the old ways try and put it down. It's not just that it's inadequate, people just don't want to face the prospect of towers going away so attempt to pick out essential items that mean it can never happen. PCI slots were developed to give people fast enough expansion on the widest selling machines - demand + volume = profit. Desktop towers are no longer the widest selling machines by a long way and the people who need PCI cards much less than that. The demand + volume is now in the mobile sector and that is the domain of Thunderbolt, that is something people are going to have to come to terms with.
Marvin. I'd suggest you work as a professional audio engineer for over 15 years like I have, before you start commenting on it. Believe it or not I actually started my career when we were still using analogue tapes, so I've seen the changes and felt the changes. I'm not talking of home studios or project studios here, I've seen projects (movies) that had over 700 tracks of audio alone, not including the music tracks. Those were spread over 4 systems using satellites options (2x HD 6 rigs, and 2x native rigs for a total of 14 pci cards). There's isn't a damn way that your supposed thunderbolt can do it all will be able to accommodate the needs. It's fine for small home studios for musicians and all, but you're forgetting the actual professional studios and post production houses.
There's more than one DSP card in the list but you could say the same about devices that are Windows-exclusive. If the Mac Pro only has a 5770 and a 5870, does that mean it's not suitable for high-end graphics work?
There are Thunderbolt solutions targeting everything you use PCI for. While not all the cards have Thunderbolt counterparts, there are solutions for everything with the exception of GPUs on the Mac side but the iMac GPU is fast anyway so that's still a solution.
You try to make the case that Apple couldn't discontinue the Pro because there wouldn't be workable solutions for every scenario but there clearly are solutions that will be available in the timeframe that such a discontinuation would affect.
I think you're keeping the professional market unreasonably restricted to what you want the term 'professional' to represent. But hey, don't let these consumers at NAB tell you different:
Notice the optical Thunderbolt cable in there too, connected to a Macbook Pro.
Not necessarily hacking the driver binaries but a way to get them to load. The drivers work without modification on Windows, why should it be any different on the Mac side? It must be to do with how the drivers are being loaded.
You say you are behind Thunderbolt but then add 'if only it was better'. If you are truly behind something then you don't try to knock it down at every turn. Your stance is clear, you want the tower and PCI slots and anything else is not for 'professionals'. I think those things will go away soon and professionals will be defined by their work not the size of their box.
The Mac GPU drivers won't work because they don't know how to work over Thunderbolt. Apple develops those but AMD/NVidia could write their own if they wanted.
The Pegasus RAID is a solution - you manage the hardware controller from the utility:
http://www.anandtech.com/show/4489/promise-pegasus-r6-mac-thunderbolt-review/4
When you say fast storage, you can put 2x Samsung 830s in a Mini and it will be faster than most hardware RAID drives:
Sure if you need 12TB of fast storage, you go for the Pegasus but it won't be much faster as platter drives are inherently slow and you're not going to put 8 platter drives in RAID0.
If you're hooking up multiple machines, you can hook up multiple Thunderbolt options too like iMacs, Minis or Macbook Pros:
http://duc.avid.com/showthread.php?t=323050
iMacs have two Thunderbolt ports so two HDX each with say 100 tracks per box so 4 iMacs hooked up over ethernet. More maybe or less if you aren't bandwidth-limited and put 3 cards per Magma box? I don't see how it's an impossible usage scenario.
My preferred solution for the Pro is not discontinuation but a smaller box with 6x Thunderbolt ports (obviously one used for the display, possibly chained) and you would be able to connect loads of dedicated hardware to a single machine, in a standard Pro, you get 3 slots free. Although they are higher bandwidth, if you need the cards for processing, you get more with the TB ports.
Marvin, you're obviously very bright and resourceful, but some real world professional experience would do you wonders.
I stopped reading after this. : D
Seriously though, what do you all want to see from Apple in an update? Out of the following:
- dual Xeon options
- 8 RAM slots
- 4 PCIe 3 slots
- USB 3
- FW800
- dual-ethernet
- 5.25" drive bay
- 4 HDD bays
- Thunderbolt ports (how many?)
- desktop 7970 or GTX 680 GPU
- same chassis/new chassis?
The way I see it is the 5.25" bay wastes a lot of space and it costs money to put in there.
The dual-processors obviously offer the best performance so barring some other way of connecting machines up, it would make sense to offer the DP models and that gives the 8 RAM slots.
I'm not sure if Sandy Bridge Mac Pros will support USB 3 but I imagine they could somehow and it's certainly in their best interests to support it.
3rd party manufacturers seem to be putting Thunderbolt on desktop motherboards alongside PCI slots but it's not clear how they are connecting up to GPUs. A poster on the following forum says Apple does it the following way:
"For anyone who cares, Apple routes video signal from the dedicated card then through the port by copying the entirely processed video signal from the dedicated GPU's framebuffer into the integrated framebuffer, which dispatches the signal through the thunderbolt IC and out to the port. Early models of Macs with hybrid graphics used muxer chips that physically switched the port's wirings from one GPU to the other."
http://www.tonymacx86.com/viewtopic.php?f=277&t=61290&start=10
This suggests there's some driver-level manipulation to get video out through the Thunderbolt port. While it would allow GPUs to have outwardly facing ports, it means that if you used another card where the drivers don't have this functionality, it would break display output on your Thunderbolt ports. Also, it's not obvious how the system knows which GPU to use for which Thunderbolt port in a multi-GPU system.
The simpler option is the iMac design where only Thunderbolt ports go on the outside. I personally think this is the best way to go in which case, PCI slots are gone for anything but the single GPU. Some will then say that makes the Pro worthless but I think the switch has to come eventually. You can see how small the Boxx RenderPro is without PCI slots and it takes a dual Xeon. The equivalent Mac Pro would be a bit bigger with drives and GPU but not by much.
There's been a leaked photo of some MBP specs (I think dual-core Ivy Bridge arrived today) so we may see the store go down tomorrow. Worst case, it will be WWDC but they've done hardware releases before WWDC in the past.
Quote:
Originally Posted by Marvin
The simpler option is the iMac design where only Thunderbolt ports go on the outside. I personally think this is the best way to go in which case, PCI slots are gone for anything but the single GPU. Some will then say that makes the Pro worthless but I think the switch has to come eventually. You can see how small the Boxx RenderPro is without PCI slots and it takes a dual Xeon. The equivalent Mac Pro would be a bit bigger with drives and GPU but not by much.
There's been a leaked photo of some MBP specs (I think dual-core Ivy Bridge arrived today) so we may see the store go down tomorrow. Worst case, it will be WWDC but they've done hardware releases before WWDC in the past.
I like reading your posts too, even if we disagree on quite a lot. I think you would need much more from thunderbolt to ditch PCI on that machine. External PCI solutions aren't so great presently, and thunderbolt chips take up a significant amount of logic board space, yet you can only get a maximum of two ports from one chip assuming the regular ones are used. Those Boxx units lack graphics cards too, and I don't see integrated graphics being a viable option for a mac pro successor without real OpenCL/OpenGL performance. In the case of the imac, they have a lot of surface area, and they actually use a fairly expensive gpu relative to its performance. Workstation pricing has gone a little crazy at comparable levels of configuration over the past few year, and I find that a little annoying. The nice thing with the big towers is that you can pretty much always make something work. Even if you need many many ports, displays, storage, dongle keys, etc. you can make it work. With many of Apple's solutions, it can be much more difficult. Ideally I'd like to just make it a laptop upgrade year if Sandy Bridge E mac pros aren't too impressive in performance gains, but I don't know if that will work out.
By the way, one of those shipping slips from the supposed leak would be easy to fake. You could create one, print it, and photograph it with a camera phone in a few minutes. This would be an easy way to make something convincing in appearance. It's also not difficult to make convincing specs. I don't think that photo really means much either way.
Apple also knows how much money the Mac Pro makes them... and as long as it's not losing money... I don't see why they would kill the Mac Pro. They still have customers.
If Apple does kill the Mac Pro... all those customers will run into the arms of HP or Dell.
Yeah... that's right... there are other companies who still make workstation-class machines.
HP makes the Z-series workstations... Dell makes the Precision and Optiplex workstations... and other companies like Boxx do too.
If it comes down to money... I bet the Mac Pro still turns a profit... albeit not as big a percentage as the iPhone and iPad. But that shouldn't matter. A profitable product is a profitable product... no matter the volume.
I bet Dell doesn't sell nearly as many Precision workstations as they do Inspiron laptops... yet they still offer those high-end machines. What's their secret?
Besides... wouldn't Apple's own software engineers prefer using a 12-core Mac Pro? Or are they gonna be happy moving from a 12-core workstation down to a quad-core iMac?
Quote:
Originally Posted by Michael Scrip
Besides... wouldn't Apple's own software engineers prefer using a 12-core Mac Pro? Or are they gonna be happy moving from a 12-core workstation down to a quad-core iMac?
Of course not, it's written in linux on a Dell workstation
We can put an upper bound on the amount:
5% of all Mac models = 0.05 x 5 million per quarter = 250k per quarter @ $1000 margins = $250m
Apple's profit last quarter = $11.6b so the Mac Pro contributes at most 2% to Apple's profits now
The same reasoning applies to all the Macs though:
http://www.computerworld.com/s/article/9226579/Macs_contribute_record_low_13_to_Apple_s_revenue
If it's still making money, it sustains those jobs and the Mac OS feeds into iOS - iOS developers aren't using PCs or iOS devices.
However, a lot of tasks that were previously thought to be exclusive to the Pro no longer are. The Pro fans talk about the lower Mac models getting too hot and 'burning out' or shutting down but it's a load of baloney/FUD. A great number of people use the cheaper models for high-value and high-resource (i.e professional) tasks and this is evident from their sales volume.
The Mac Pro has unquestionable design benefits such as running cool, having expansion, offering up to 3x the performance of the iMac but the lower models have the benefit of costing much less money. While high-value industry workers can sometimes afford the expense, it's not always the case nor is it essential. Some of the smartest and most talented people in the world have to work on shoe-string budgets and this can contribute to some of their best achievements. Educational funding is rarely generous enough that researchers get the best equipment. R&D happens before you start making money.
Only Apple knows what's selling and their route ahead. You can see the graph above though and the rising vectors. Like it or not, along the way, there will be casualties in Apple's line, the only question is the timing. The Xserve was the first but it won't be the last.
I don't think they'd all jump ship. Many quad and maybe even 6-core owners would go to the iMacs.
Instead of killing it like the XServe, I suspect they'd ramp it down. Cull the models that people aren't buying - I personally suspect the Server model is for the chopping block - and just do that as time goes on.
Quote:
Originally Posted by Marvin
Only Apple knows what's selling and their route ahead. You can see the graph above though and the rising vectors. Like it or not, along the way, there will be casualties in Apple's line, the only question is the timing. The Xserve was the first but it won't be the last.
I don't think they'd all jump ship. Many quad and maybe even 6-core owners would go to the iMacs.
Instead of killing it like the XServe, I suspect they'd ramp it down. Cull the models that people aren't buying - I personally suspect the Server model is for the chopping block - and just do that as time goes on.
Right now they have a lot of odd configurations. The 8 core as an example is slower than the 6 in many workloads. If they consolidated but used the streamlined setup to keep the offered configurations competitive, that would work well. There is no true sever model. It's basically OSX server + second hard drive and 8GB of ram. You can configure that from the normal one, so cutting it really doesn't affect anyone as you can still buy the same configuration without the moniker. You can check for yourself. There's nothing unique about it beyond what I mentioned.
By the way, that's kind of a troll article as it only considers direct contributions to the bottom line. It doesn't consider internal use by Apple employees, IOS development, or contribution to their ecosystem. Saying you'd only lose those sales by dropping it is a fallacy, and they'd incur additional costs supporting Windows or Linux based IOS developer tools. I'm not even painting this as something bigger than it is. I dislike the way the article is presented.