HOLY ****! Just going to dual-2.66Ghz Xeons ends up with a base price of $5000 without everything left at default... What are they thinking? Thats an outrageous price...
For those making excuses, the new Nehalem Xeons aren't supposed to cost much more than the existing Xeons..
I marvel at the Pro's design esthetics every time I need to open the case for maintenance. It is a piece of art. I'd rather they continue focusing on improving speed than developing a new look.
OpenCL consists of an API for coordinating parallel computation across
heterogeneous processors; and a cross-platform programming language with a well-specified computation environment. The OpenCL standard:
Supports both data- and task-based parallel programming models
Utilizes a subset of ISO C99 with extensions for parallelism
Defines consistent numerical requirements based on IEEE 754
Defines a configuration profile for handheld and embedded devices
Efficiently interoperates with OpenGL, OpenGL ES and other graphics APIs
Tesla uses CUDA: Nvidia's solution for Parallel computations. They of course are adopting OpenCL with CUDA extensions. AMD/ATi 4870 HD uses their Streams solution and offers:
Both ARE NOT AVAILABLE TO APPLE [The Inventor of OpenCL] because we don't have a f'n system with a second dedicated GPU full size PCI Express 2.0 x16 slot.
These cards are support cards for Gaming and Multimedia in the consumer space and Scientific/Engineering in those spaces.
These aren't the GPUs for your system to run your operating system on.
I marvel at the Pro's design esthetics every time I need to open the case for maintenance. It is a piece of art. I'd rather they continue focusing on improving speed than developing a new look.
Nothing in the entire industry comes close for Consumers. Perhaps their are custom casings for scientific clusters and what not, but not in this space for a workstation tower.
And whats up with removing the sole Quadro card?? Apparently professional 3D modeling/animation/CAD/ users don't use the Mac Pro anymore...
To be honest, It has crossed my mind that the price increases and the lack of a Quadro/FirePro option may be Apple giving reason to the pros to jump ship. They don't quite fit the Jobs/Ive plan.
According to Intel pricing at Xbitlabs.com, a 4-core 2.26 Nehalem Xeon (E5520) chip is $373 (in quantities of 1000). The 4-core 2.66 Nehalem Xeon (E5550) chip is $958. The Intel price difference for two chips is $1170. But Apple could be getting a much better price (like $300 or less) on the Mac Pro standard 2.26 chips since they're buying those in much larger quantities than the Mac Pro optional 2.66 chip, moving the price differential over $1300.
So yes Apple is taking a piece, but almost all of the cost difference is attributable to the Intel pricing.
Interesting. Also a thing to consider is that perhaps with the the Nehalem's, the margin between "well performing" and "average/meh performing" chips is much wider. In other words, when Intel makes chips, to my understanding they don't make fundamentally different chips for 2.26 and 2.66 clock speeds. They make one chip. Depending on how each manufactured chip performs, it either gets thrown in the 2.26 bin or the 2.66 bin. Perhaps the chips that are stable/well performing at 2.66 are less prevalent in this architecture.
Quote:
Originally Posted by kim kap sol
It's a little odd that Apple has been removing pro options...however, any self-respecting person would buy the absolute base (and just play around with the processor options) and buy memory and storage from some place that isn't hell-bent on ripping its customers off.
It pains me to see what Apple is charging its customers for HD and RAM upgrades.
Agreed. Any experienced G5/Mac Pro use should be well versed in buying hard drives, ram, opticals, and PCI cards from 3rd party vendors.
Quote:
Originally Posted by MotherBrain
I am really getting a bit tired of hearing that the next model is up to 2X faster. It's getting old and not relative to anything you really do with your computer. I also think it makes Apple look bad. Every model since, who knows when, has been up to 2X faster. If that were REALLY true we would be discussing quantum physics with our HALpro 9000.
I agree with this, I'd like to see a little more retrospective benchmarking, kind of how barefeats will usually throw in a setup thats 2 generations old at the bottom of their comparos.
Quote:
Originally Posted by winterspan
And whats up with removing the sole Quadro card?? Apparently professional 3D modeling/animation/CAD/ users don't use the Mac Pro anymore...
They don't, and many probably never have been. I work in a shop that has all Mac Pro 3.2 Octas doing motion graphics/video production. The guys in our 3d department work on them, although they are always booted into Windows. Lately, for this huge job they've been working on, they brought in rental PCs for themselves and the additional help they brought in, because they're faster. And while I love Macs, and have always used them. I can't help but wonder why the PC I jumped on one day performed better in after effects.
Never the less, my personal machine ( Quad G5, Dec 05) will finally be replaced now. I've been waiting for quite some time.
Interesting. Also a thing to consider is that perhaps with the the Nehalem's, the margin between "well performing" and "average/meh performing" chips is much wider. In other words, when Intel makes chips, to my understanding they don't make fundamentally different chips for 2.26 and 2.66 clock speeds. They make one chip. Depending on how each manufactured chip performs, it either gets thrown in the 2.26 bin or the 2.66 bin. Perhaps the chips that are stable/well performing at 2.66 are less prevalent in this architecture.
Agreed. Any experienced G5/Mac Pro use should be well versed in buying hard drives, ram, opticals, and PCI cards from 3rd party vendors.
I agree with this, I'd like to see a little more retrospective benchmarking, kind of how barefeats will usually throw in a setup thats 2 generations old at the bottom of their comparos.
They don't, and many probably never have been. I work in a shop that has all Mac Pro 3.2 Octas doing motion graphics/video production. The guys in our 3d department work on them, although they are always booted into Windows. Lately, for this huge job they've been working on, they brought in rental PCs for themselves and the additional help they brought in, because they're faster. And while I love Macs, and have always used them. I can't help but wonder why the PC I jumped on one day performed better in after effects.
Never the less, my personal machine ( Quad G5, Dec 05) will finally be replaced now. I've been waiting for quite some time.
The point of OpenGL 3.x and OpenCL is for third parties to leverage OS X and it's advanced toolkits to have those solutions available to the system.
As I've stated, they better either offer a dedicated Fiber connected Tesla/Firestream clone of their own or add another PCI Express 2.0 x16 slot for the Tesla and Firestream.
I realize that. But there is an odd number of sockets. It should be 9.
Each CPU has an independent memory controller on die, so you can't combine them. It isn't 8 DIMM slots as much as 2x 4 DIMM slots. Each controller support up to 9 DIMMs for 18 max.
Each CPU has an independent memory controller and you can't combine them. It isn't 8 DIMM slots as much as 2x 4 DIMM slots. Each controller support up to 9 DIMMs for 18 max.
Each CPU has an independent memory controller and you can't combine them. It isn't 8 DIMM slots as much as 2x 4 DIMM slots. Each controller support up to 9 DIMMs for 18 max.
So does that mean the 8 core machine actually has six memory channels? Who is really going to want to use those last two slots if it is going to hamper performance a bit?
It should have been, but only 3 DIMMs per CPU would leave the pros less than happy and Ive's design team apparently didn't want bothered with any design changes to the case. The engineers did the best they could using the current case. If they would have widened it slightly from 8.1" to the same 8.9" as the el-Capitan cases, there would have been enough room for the extra slots as well as another USB port on the back and perhaps another pair up front.
So does that mean the 8 core machine actually has six memory channels? Who is really going to want to use those last two slots if it is going to hamper performance a bit?
I know it's seems like a minor point, but a lot of people are making the error, and it leads to confusion.
The i7 is Intel's brand for desktop chips.
The Xeon is intel's brand for workstation and server chips.
There is no i7 Xeon.
In terms of silicon, the i7 and these Xeons are exactly the same- in fact, all of the i7 processors Intel sells have a second, disabled, QPI link. A Xeon and an i7 could be cut off the same wafer. They even use the same package, LGA1366. The Xeon 3500-series *is* an i7, renamed, so it's not dual-socket capable but is otherwise the same as the rest.
It should have been, but only 3 DIMMs per CPU would leave the pros less than happy and Ive's design team apparently didn't want bothered with any design changes to the case. The engineers did the best they could using the current case. If they would have widened it slightly from 8.1" to the same 8.9" as the el-Capitan cases, there would have been enough room for the extra slots as well as another USB port on the back and perhaps another pair up front.
Well, I meant 6 for the single cpu model, and 12 for the dual.
I'm sure they could have managed it if they wanted to.
In terms of silicon, the i7 and these Xeons are exactly the same- in fact, all of the i7 processors Intel sells have a second, disabled, QPI link. A Xeon and an i7 could be cut off the same wafer. They even use the same package, LGA1366. The Xeon 3500-series *is* an i7, renamed, so it's not dual-socket capable but is otherwise the same as the rest.
Makes no difference. Branding is branding. The extra link does matter though. The second socket ability does matter too.
Look, Intel has the right to brand their products they way they want to, and we should stick to that when discussing those products.
We have to use the correct names for products for us to make sense of what we read. I'm reading too many times someone simply calling a Xeon an i7.
I agree that the price is high, but the price of Mac pros is always pretty high anyway and mostly inconsequential to the purchase decision.
My biggest gripe about this machine is that if the picture is accurate, there seems to be a new motherboard, and new configuration of the entire lower half of the insides. So again, while masquerading as an "upgradeable" Mac, this machine is essentially completely different from the previous model.
I'm not really sure why they even bother to let us open the case since the odds of upgrading a Mac pro past the stuff that you generally get when you purchase it, are something like slim to none. Anyone know if the new processors would pop into the older MacPros? It sure looks like the answer is no yet again.
Nehalem is a completely new architecture that is incompatible with older architectures. This is why you cannot simply "upgrade the processor." If you spent your time researching why this was such a radical change instead of making ill-informed forum posts, you might begin to know what you are talking about.
For example, the Penryn (Early 2008) Mac Pros use dual channel DDR2 memory. These Mac Pros use triple channel DDR3 RAM. These two technologies are completely incompatible with each other. It is IMPOSSIBLE for Apple to make it "easy to upgrade."
Nehalem moves the memory controller onto the CPU. This has never been done before. It is not possible to make this compatible with previous generations of hardware.
Well, I meant 6 for the single cpu model, and 12 for the dual.
I'm sure they could have managed it if they wanted to.
They could have, but unfortunately this is just another example of the components having to conform to the case instead of the case conforming to the components.
Quote:
Originally Posted by uni
Like re-design the inner case?
Oh wait..
No, more like use a case that is designed for dual Nehalem cpus and 12 DIMM slots instead of trying to cram it all into a case designed six years ago for the G5 and modified for socket 771 core-based xeons.
Comments
For those making excuses, the new Nehalem Xeons aren't supposed to cost much more than the existing Xeons..
Its hard to perfect...."perfection".
Still a sexy looking cabinet.
I marvel at the Pro's design esthetics every time I need to open the case for maintenance. It is a piece of art. I'd rather they continue focusing on improving speed than developing a new look.
No SLI, yes, but the rest?
How do you come to that conclusion?
Though I don't seem to know what Tesla is suppose to be.
http://www.nvidia.com/object/product..._c1060_us.html
OpenCL deals strictly with the following:
IEEE 754 single & double operations.
OpenCL consists of an API for coordinating parallel computation across
heterogeneous processors; and a cross-platform programming language with a well-specified computation environment. The OpenCL standard:
Tesla uses CUDA: Nvidia's solution for Parallel computations. They of course are adopting OpenCL with CUDA extensions. AMD/ATi 4870 HD uses their Streams solution and offers:
ATi Firestream [Tesla competitor]
http://ati.amd.com/technology/stream...ream_9270.html
Both systems are OpenCL compliant.
Both ARE NOT AVAILABLE TO APPLE [The Inventor of OpenCL] because we don't have a f'n system with a second dedicated GPU full size PCI Express 2.0 x16 slot.
These cards are support cards for Gaming and Multimedia in the consumer space and Scientific/Engineering in those spaces.
These aren't the GPUs for your system to run your operating system on.
I marvel at the Pro's design esthetics every time I need to open the case for maintenance. It is a piece of art. I'd rather they continue focusing on improving speed than developing a new look.
Nothing in the entire industry comes close for Consumers. Perhaps their are custom casings for scientific clusters and what not, but not in this space for a workstation tower.
If there are 3 or 6 DIMMs (per socket), the memory controller operates in triple channel, if there are 2 or 4 DIMMs, it operates in dual channel.
I realize that. But there is an odd number of sockets. It should be 9.
And whats up with removing the sole Quadro card?? Apparently professional 3D modeling/animation/CAD/ users don't use the Mac Pro anymore...
To be honest, It has crossed my mind that the price increases and the lack of a Quadro/FirePro option may be Apple giving reason to the pros to jump ship. They don't quite fit the Jobs/Ive plan.
According to Intel pricing at Xbitlabs.com, a 4-core 2.26 Nehalem Xeon (E5520) chip is $373 (in quantities of 1000). The 4-core 2.66 Nehalem Xeon (E5550) chip is $958. The Intel price difference for two chips is $1170. But Apple could be getting a much better price (like $300 or less) on the Mac Pro standard 2.26 chips since they're buying those in much larger quantities than the Mac Pro optional 2.66 chip, moving the price differential over $1300.
So yes Apple is taking a piece, but almost all of the cost difference is attributable to the Intel pricing.
Interesting. Also a thing to consider is that perhaps with the the Nehalem's, the margin between "well performing" and "average/meh performing" chips is much wider. In other words, when Intel makes chips, to my understanding they don't make fundamentally different chips for 2.26 and 2.66 clock speeds. They make one chip. Depending on how each manufactured chip performs, it either gets thrown in the 2.26 bin or the 2.66 bin. Perhaps the chips that are stable/well performing at 2.66 are less prevalent in this architecture.
It's a little odd that Apple has been removing pro options...however, any self-respecting person would buy the absolute base (and just play around with the processor options) and buy memory and storage from some place that isn't hell-bent on ripping its customers off.
It pains me to see what Apple is charging its customers for HD and RAM upgrades.
Agreed. Any experienced G5/Mac Pro use should be well versed in buying hard drives, ram, opticals, and PCI cards from 3rd party vendors.
I am really getting a bit tired of hearing that the next model is up to 2X faster. It's getting old and not relative to anything you really do with your computer. I also think it makes Apple look bad. Every model since, who knows when, has been up to 2X faster. If that were REALLY true we would be discussing quantum physics with our HALpro 9000.
I agree with this, I'd like to see a little more retrospective benchmarking, kind of how barefeats will usually throw in a setup thats 2 generations old at the bottom of their comparos.
And whats up with removing the sole Quadro card?? Apparently professional 3D modeling/animation/CAD/ users don't use the Mac Pro anymore...
They don't, and many probably never have been. I work in a shop that has all Mac Pro 3.2 Octas doing motion graphics/video production. The guys in our 3d department work on them, although they are always booted into Windows. Lately, for this huge job they've been working on, they brought in rental PCs for themselves and the additional help they brought in, because they're faster. And while I love Macs, and have always used them. I can't help but wonder why the PC I jumped on one day performed better in after effects.
Never the less, my personal machine ( Quad G5, Dec 05) will finally be replaced now. I've been waiting for quite some time.
Interesting. Also a thing to consider is that perhaps with the the Nehalem's, the margin between "well performing" and "average/meh performing" chips is much wider. In other words, when Intel makes chips, to my understanding they don't make fundamentally different chips for 2.26 and 2.66 clock speeds. They make one chip. Depending on how each manufactured chip performs, it either gets thrown in the 2.26 bin or the 2.66 bin. Perhaps the chips that are stable/well performing at 2.66 are less prevalent in this architecture.
Agreed. Any experienced G5/Mac Pro use should be well versed in buying hard drives, ram, opticals, and PCI cards from 3rd party vendors.
I agree with this, I'd like to see a little more retrospective benchmarking, kind of how barefeats will usually throw in a setup thats 2 generations old at the bottom of their comparos.
They don't, and many probably never have been. I work in a shop that has all Mac Pro 3.2 Octas doing motion graphics/video production. The guys in our 3d department work on them, although they are always booted into Windows. Lately, for this huge job they've been working on, they brought in rental PCs for themselves and the additional help they brought in, because they're faster. And while I love Macs, and have always used them. I can't help but wonder why the PC I jumped on one day performed better in after effects.
Never the less, my personal machine ( Quad G5, Dec 05) will finally be replaced now. I've been waiting for quite some time.
The point of OpenGL 3.x and OpenCL is for third parties to leverage OS X and it's advanced toolkits to have those solutions available to the system.
As I've stated, they better either offer a dedicated Fiber connected Tesla/Firestream clone of their own or add another PCI Express 2.0 x16 slot for the Tesla and Firestream.
I realize that. But there is an odd number of sockets. It should be 9.
Each CPU has an independent memory controller on die, so you can't combine them. It isn't 8 DIMM slots as much as 2x 4 DIMM slots. Each controller support up to 9 DIMMs for 18 max.
Each CPU has an independent memory controller and you can't combine them. It isn't 8 DIMM slots as much as 2x 4 DIMM slots. Each controller support up to 9 DIMMs for 18 max.
Fine, then it should have been 6 or 12.
Each CPU has an independent memory controller and you can't combine them. It isn't 8 DIMM slots as much as 2x 4 DIMM slots. Each controller support up to 9 DIMMs for 18 max.
So does that mean the 8 core machine actually has six memory channels? Who is really going to want to use those last two slots if it is going to hamper performance a bit?
Fine, then it should have been 6 or 12.
It should have been, but only 3 DIMMs per CPU would leave the pros less than happy and Ive's design team apparently didn't want bothered with any design changes to the case. The engineers did the best they could using the current case. If they would have widened it slightly from 8.1" to the same 8.9" as the el-Capitan cases, there would have been enough room for the extra slots as well as another USB port on the back and perhaps another pair up front.
So does that mean the 8 core machine actually has six memory channels? Who is really going to want to use those last two slots if it is going to hamper performance a bit?
2x triple-channel or 4x dual channel.
I know it's seems like a minor point, but a lot of people are making the error, and it leads to confusion.
The i7 is Intel's brand for desktop chips.
The Xeon is intel's brand for workstation and server chips.
There is no i7 Xeon.
In terms of silicon, the i7 and these Xeons are exactly the same- in fact, all of the i7 processors Intel sells have a second, disabled, QPI link. A Xeon and an i7 could be cut off the same wafer. They even use the same package, LGA1366. The Xeon 3500-series *is* an i7, renamed, so it's not dual-socket capable but is otherwise the same as the rest.
It should have been, but only 3 DIMMs per CPU would leave the pros less than happy and Ive's design team apparently didn't want bothered with any design changes to the case. The engineers did the best they could using the current case. If they would have widened it slightly from 8.1" to the same 8.9" as the el-Capitan cases, there would have been enough room for the extra slots as well as another USB port on the back and perhaps another pair up front.
Well, I meant 6 for the single cpu model, and 12 for the dual.
I'm sure they could have managed it if they wanted to.
In terms of silicon, the i7 and these Xeons are exactly the same- in fact, all of the i7 processors Intel sells have a second, disabled, QPI link. A Xeon and an i7 could be cut off the same wafer. They even use the same package, LGA1366. The Xeon 3500-series *is* an i7, renamed, so it's not dual-socket capable but is otherwise the same as the rest.
Makes no difference. Branding is branding. The extra link does matter though. The second socket ability does matter too.
Look, Intel has the right to brand their products they way they want to, and we should stick to that when discussing those products.
We have to use the correct names for products for us to make sense of what we read. I'm reading too many times someone simply calling a Xeon an i7.
I agree that the price is high, but the price of Mac pros is always pretty high anyway and mostly inconsequential to the purchase decision.
My biggest gripe about this machine is that if the picture is accurate, there seems to be a new motherboard, and new configuration of the entire lower half of the insides. So again, while masquerading as an "upgradeable" Mac, this machine is essentially completely different from the previous model.
I'm not really sure why they even bother to let us open the case since the odds of upgrading a Mac pro past the stuff that you generally get when you purchase it, are something like slim to none. Anyone know if the new processors would pop into the older MacPros? It sure looks like the answer is no yet again.
Nehalem is a completely new architecture that is incompatible with older architectures. This is why you cannot simply "upgrade the processor." If you spent your time researching why this was such a radical change instead of making ill-informed forum posts, you might begin to know what you are talking about.
For example, the Penryn (Early 2008) Mac Pros use dual channel DDR2 memory. These Mac Pros use triple channel DDR3 RAM. These two technologies are completely incompatible with each other. It is IMPOSSIBLE for Apple to make it "easy to upgrade."
Nehalem moves the memory controller onto the CPU. This has never been done before. It is not possible to make this compatible with previous generations of hardware.
Couldn't they have done something, anything to distinguish it from it's 6 year old design?
Like re-design the inner case?
Oh wait..
Well, I meant 6 for the single cpu model, and 12 for the dual.
I'm sure they could have managed it if they wanted to.
They could have, but unfortunately this is just another example of the components having to conform to the case instead of the case conforming to the components.
Like re-design the inner case?
Oh wait..
No, more like use a case that is designed for dual Nehalem cpus and 12 DIMM slots instead of trying to cram it all into a case designed six years ago for the G5 and modified for socket 771 core-based xeons.