No, that was benchmarks to disprove your clearly incorrect assertions that Xeons/FB-DIMMs are some tragic mistake for users. Only gamers see any serious degradation and only in a few games.
To date you have yet to show any application where there is more than a minor amount of difference that won't last past the next release of the software that is more multi-core friendly. Which "prosumer" application is horribly impacted? Quicktime I'll give you but that seems more of a QT lack than a Mac Pro problem.
And no...I couldn't care less if you were using macs 15 years or 15 seconds. Lies and untruths are still lies and untruths. For someone that can't even agree on the most neutral of positions to be claiming some kind of matyrdom is just rich.
Vinea
You might wish to try reading that anandtech article you leaked to. In multi-threaded applications the twin dual core processors absolutely sailed through the task with overwhelming horsepower. When it came to single threaded consumer apps, the high latency of FB-DIMMs came into play and applications tended to be about 10-15% slower than their equivalent Core 2 brethren at those single threaded tasks. Of course since Intel never planned for Xeon use in a single threaded environment, it was deemed a more than acceptable trade off.
Except if you offer a single-cpu Mac Pro you can also offer after-market upgrades to make the single-cpu computers dual-cpu computers.
And how many in this segment are going to break Apple's warranty to make use of that? It's not as easy a task as EATX PC workstations. You also have to come up with a heat sink.
Just so we're clear, Intel's 5000X chipset allows for 1 or 2 CPUs, and the Memory Controller requieres FB-DIMMs, but like H said, it may change with the next (or the one after that) generation of chipset.
There is another Xeon chipset (3000 series) that works with single CPUs (Conroe-based Xeons 3000 series) and requieres (only) DDR2 533/667 unbuffered RAM. It only allows for 8GB of RAM (instead of 64GB) and the south bridge doesn't have a lot of PCIe lanes (16). Anyway Intel sees it more as a small (lower-cost) server platform than a workstation chipset.
Unless, a single CPU Mac Pro (2.00/2.66/3.00GHz) is really slow compared to a 2.00/2.16/2.33 iMac, which I seriously doubt, I really think it would be a good way to widen the scope of the Mac Pro, reach new customers, switchers, and those lost ones that onlooker was writing about earlier (creative pros).
I think that $1499 for the 2.00 model and $1799 for the 2.66GHz model are good prices (with 1GB of RAM and at least a 160GB HDD). I don't believe those would compete with the 20" iMac or the 24" iMac (which may seem slower but has a built-in 24" display). But I agree that they may compete with the current dual-dual-core Mac Pro, people may believe that a single dual-core 2.66GHz Mac Pro is enough for their needs. In the other hand, people that would not have bought one Mac Pro may jump sooner because of the price, and small/medium (creative) businesses would be interested, not in one, but in many single CPU computers.
Getting Single CPUs models, as well as introducing quad-core CPUs to the Mac Pro line-up, would also allow Apple to get some of the after-market business if they could offer processor upgrades thru their Apple Stores...
I understand Ben when he is saying that a Conroe based 2.13/2.40/2.66 xMac could be made and would cost less that the Single CPU Mac Pro, but I also understand that it would bring more R&D costs and inventory costs today. The Single CPU Mac Pro solution is a smoother one.
Anyway it's not there yet, but I really hope we'll see something new before or at the WWDC this year.
My mind got ahead of itself. I knew it was capable of being used as a single CPU, but my point is what is the purpose of using a processor that is designed for dual configuration as a single CPU. It's not the best of the processor. That is what I was getting at. If they were going to offer a single CPU they would probably offer a conroe, but then again it's another processor. Not that it matters, but this conversation is way past that now. Apple should be upgrading the PM when Leopard is announced I hope.
You might wish to try reading that anandtech article you leaked to. In multi-threaded applications the twin dual core processors absolutely sailed through the task with overwhelming horsepower. When it came to single threaded consumer apps, the high latency of FB-DIMMs came into play and applications tended to be about 10-15% slower than their equivalent Core 2 brethren at those single threaded tasks. Of course since Intel never planned for Xeon use in a single threaded environment, it was deemed a more than acceptable trade off.
5%-10%.
You might try reading my posts once in a while. I noted that already. Also noted was the 50-60+% improvement for multiprocessor friendly apps...something you gloss over. Again.
There was little difference for most apps.
SYSmark 2004
General -5%
ICC +2%
Office Productivity -10%
"Overall, based on the SYSMark 2004 results, the Mac Pro at 3.0GHz ends up performing like a similarly clocked Core 2 machine, which isn't bad at all."
Winstone 2004
"The MMCC Winstone 2004 test does have a highly multithreaded component, but the performance benefit from having four cores is less than 4%. Still, the Mac Pro at 3.0GHz ends up being slightly faster than the fastest desktop PC processor."
3D Rendering
"The multithreaded CineBench test shows us similar scaling to what we saw in 3dsmax; the move to four cores gives the Mac Pro a 65% boost in performance"
'nuff said there.
Encoding
Anywhere from -5% to +30%.
"With only two cores enabled the Mac Pro at 3.0GHz is about 5% slower than the Core 2 Extreme X6800, once again thanks to its high latency FB-DIMMs that erase any benefits the faster FSB and clock speeds would have provided. With all four cores enabled however, the Mac Pro manages to perform our encoding test in a mere 28 seconds - an improvement of almost 30%"
It was iTunes that sucked and not QT. My bad.
BenRoethig: 0% correct.
How you turn 5%-10% performance delta into 10%-15% and expect anyone to believe your lies when the article is linked for all to read is beyond me. Some areas the Mac Pro did worse, other areas it did far better but for most things there were minor differences even without 2 of the cores running.
And how many in this segment are going to break Apple's warranty to make use of that? It's not as easy a task as EATX PC workstations. You also have to come up with a heat sink.
The same number that has been popping in upgrades to the G4 Powermacs.
Given that enough sites have done CPU swaps on the Mac Pro its not impossible for folks to do.
Just so we're clear, Intel's 5000X chipset allows for 1 or 2 CPUs, and the Memory Controller requieres FB-DIMMs, but like H said, it may change with the next (or the one after that) generation of chipset.
There is another Xeon chipset (3000 series) that works with single CPUs (Conroe-based Xeons 3000 series) and requieres (only) DDR2 533/667 unbuffered RAM. It only allows for 8GB of RAM (instead of 64GB) and the south bridge doesn't have a lot of PCIe lanes (16). Anyway Intel sees it more as a small (lower-cost) server platform than a workstation chipset.
The 3000P is designed for entry level servers. The reason it has so few PCI-E lanes is that they already have the 975x for the desktop/workstation segment. The only difference between the Core 2 Duo and the Xeon 3000 series CPUs is the inclusion of ECC ram support.
I'm buying into Vinea's single CPU Mac Pro at $1699. That'll fill in the gap pretty nicely without having to redesign a new motherboard for a Conroe box.
The same number that has been popping in upgrades to the G4 Powermacs.
Given that enough sites have done CPU swaps on the Mac Pro its not impossible for folks to do.
Vinea
The G4s used a proprietary cpu card that covered both processors with a heat sink to match. The U3L motherboard used in SP G3s did not have the second sockets so upgrading to dual CPUs was impossible. As for the Mac Pro, it features twin proprietary heat sinks. While the socket would be there on a single cpu model, the heat sink would not.
I'm buying into Vinea's single CPU Mac Pro at $1699. That'll fill in the gap pretty nicely without having to redesign a new motherboard for a Conroe box.
Intel is familiar with the Mac Pro's form factor. It wouldn't be that hard to fit a version of the 965P or 975X to match. besides it wouldn't be the first time Apple used a slightly different chipset for a single processor model.
besides it wouldn't be the first time Apple used a slightly different chipset for a single processor model.
Those situation have normally arisen from Apple using their previous-generation chipset/motherboard in the bottom-of-the range machine. In the case of the Mac Pro, there is no previous generation motherboard.
Are you denying that initiating the offering of a Mac Pro with one CPU instead of two requires vastly less investment than initiating the offering of a Mac Pro with an entirely new motherboard?
Yes, so Apple could sell a "second CPU upgrade kit" that is fitted by your nearest Apple Store, and includes a heat sink.
It's also not like you couldn't build a new heatsink for an upgrade. Crucial added Mac pro compatible FB-DIMM hs in short order. There's currently no market because anyone upgrading a Mac Pro today has 2 working heat sinks.
Not that too many folks are upgrading to Cloverton at the moment anyway...
Yes, so Apple could sell a "second CPU upgrade kit" that is fitted by your nearest Apple Store, and includes a heat sink.
Doesn't seem very Apple like.
Quote:
Originally Posted by Mr. H
Those situation have normally arisen from Apple using their previous-generation chipset/motherboard in the bottom-of-the range machine. In the case of the Mac Pro, there is no previous generation motherboard.
Not quite. Apple used a family of North bridges for the G5 called U3. It came in three different variants. The U3L was used in the low end PowerMac and iMac. It did not feature SMP support or PCI-X support. The general variant was the U3, it featured PCI-X support, but no SMP support. The SMP versions and the G5 xServe used the U3H. While they were related they needed slightly different driver sets. At the same time they also had to deal with the intrepid chipset on Mac Mini, eMac, and noteboooks.
Quote:
Are you denying that initiating the offering of a Mac Pro with one CPU instead of two requires vastly less investment than initiating the offering of a Mac Pro with an entirely new motherboard?
A little less investment for Apple, but a lot more investment for the user. It would be slightly easier, but you can kiss the possibility of higher end desktop switchers goodbye and with them the general users they advise.
Are you denying that initiating the offering of a Mac Pro with one CPU instead of two requires vastly less investment than initiating the offering of a Mac Pro with an entirely new motherboard?
The discussion has been centered on using the Mac Pro to make a cheaper, single CPU tower. However, I don't like any approach that uses the Mac Pro chassis. It might serve as a short term trial run, but it couldn't sell as well as a smaller tower with high performance -- a competitive prosumer Mac that is expandable.
Any approach that uses one Xeon has the added cost of a workstation CPU and more expensive memory, if I have been reading the comments correctly. Its only advantage is the low development cost.
Designing a new motherboard has the advantage of lower cost CPU and RAM, at the expense of higher development cost. However the Mac Pro chassis and other part still keep the parts cost higher than necessary. Apple could replace the power supply, and look for other ways to reduce component cost. Yet at this point it may as well be a whole new design.
Even if Apple took one of these approaches using the Mac Pro chassis, I'm not sure the final product would get good marketing feedback either. I'd say forget the Mac Pro as a test bed, and just design a competitive mini tower in the prosumer price range. It's my opinion.
The discussion has been centered on using the Mac Pro to make a cheaper, single CPU tower. However, I don't like any approach that uses the Mac Pro chassis. It might serve as a short term trial run, but it couldn't sell as well as a smaller tower with high performance -- a competitive prosumer Mac that is expandable.
Any approach that uses one Xeon has the added cost of a workstation CPU and more expensive memory, if I have been reading the comments correctly. Its only advantage is the low development cost.
Designing a new motherboard has the advantage of lower cost CPU and RAM, at the expense of higher development cost. However the Mac Pro chassis and other part still keep the parts cost higher than necessary. Apple could replace the power supply, and look for other ways to reduce component cost. Yet at this point it may as well be a whole new design.
Even if Apple took one of these approaches using the Mac Pro chassis, I'm not sure the final product would get good marketing feedback either. I'd say forget the Mac Pro as a test bed, and just design a competitive mini tower in the prosumer price range. It's my opinion.
I think we're talking about multiple headless segments here. I think you're looking for a lower end mini tower or cube somewhat equivalent to the PC side's MATX towers. Possibly something iMac based, but with a PCI-E x16 slot and maybe and x1. Me, I'm looking for Apple's evolution of the full tower design with a 2.4ghz CPU. Anything like a cube I have to put it on my desk and anything too short I really have to reach for underneath.
Comments
No, that was benchmarks to disprove your clearly incorrect assertions that Xeons/FB-DIMMs are some tragic mistake for users. Only gamers see any serious degradation and only in a few games.
To date you have yet to show any application where there is more than a minor amount of difference that won't last past the next release of the software that is more multi-core friendly. Which "prosumer" application is horribly impacted? Quicktime I'll give you but that seems more of a QT lack than a Mac Pro problem.
And no...I couldn't care less if you were using macs 15 years or 15 seconds. Lies and untruths are still lies and untruths. For someone that can't even agree on the most neutral of positions to be claiming some kind of matyrdom is just rich.
Vinea
You might wish to try reading that anandtech article you leaked to. In multi-threaded applications the twin dual core processors absolutely sailed through the task with overwhelming horsepower. When it came to single threaded consumer apps, the high latency of FB-DIMMs came into play and applications tended to be about 10-15% slower than their equivalent Core 2 brethren at those single threaded tasks. Of course since Intel never planned for Xeon use in a single threaded environment, it was deemed a more than acceptable trade off.
Except if you offer a single-cpu Mac Pro you can also offer after-market upgrades to make the single-cpu computers dual-cpu computers.
And how many in this segment are going to break Apple's warranty to make use of that? It's not as easy a task as EATX PC workstations. You also have to come up with a heat sink.
There is another Xeon chipset (3000 series) that works with single CPUs (Conroe-based Xeons 3000 series) and requieres (only) DDR2 533/667 unbuffered RAM. It only allows for 8GB of RAM (instead of 64GB) and the south bridge doesn't have a lot of PCIe lanes (16). Anyway Intel sees it more as a small (lower-cost) server platform than a workstation chipset.
Unless, a single CPU Mac Pro (2.00/2.66/3.00GHz) is really slow compared to a 2.00/2.16/2.33 iMac, which I seriously doubt, I really think it would be a good way to widen the scope of the Mac Pro, reach new customers, switchers, and those lost ones that onlooker was writing about earlier (creative pros).
I think that $1499 for the 2.00 model and $1799 for the 2.66GHz model are good prices (with 1GB of RAM and at least a 160GB HDD). I don't believe those would compete with the 20" iMac or the 24" iMac (which may seem slower but has a built-in 24" display). But I agree that they may compete with the current dual-dual-core Mac Pro, people may believe that a single dual-core 2.66GHz Mac Pro is enough for their needs. In the other hand, people that would not have bought one Mac Pro may jump sooner because of the price, and small/medium (creative) businesses would be interested, not in one, but in many single CPU computers.
Getting Single CPUs models, as well as introducing quad-core CPUs to the Mac Pro line-up, would also allow Apple to get some of the after-market business if they could offer processor upgrades thru their Apple Stores...
I understand Ben when he is saying that a Conroe based 2.13/2.40/2.66 xMac could be made and would cost less that the Single CPU Mac Pro, but I also understand that it would bring more R&D costs and inventory costs today. The Single CPU Mac Pro solution is a smoother one.
Anyway it's not there yet, but I really hope we'll see something new before or at the WWDC this year.
You might wish to try reading that anandtech article you leaked to. In multi-threaded applications the twin dual core processors absolutely sailed through the task with overwhelming horsepower. When it came to single threaded consumer apps, the high latency of FB-DIMMs came into play and applications tended to be about 10-15% slower than their equivalent Core 2 brethren at those single threaded tasks. Of course since Intel never planned for Xeon use in a single threaded environment, it was deemed a more than acceptable trade off.
5%-10%.
You might try reading my posts once in a while. I noted that already. Also noted was the 50-60+% improvement for multiprocessor friendly apps...something you gloss over. Again.
There was little difference for most apps.
SYSmark 2004
General -5%
ICC +2%
Office Productivity -10%
"Overall, based on the SYSMark 2004 results, the Mac Pro at 3.0GHz ends up performing like a similarly clocked Core 2 machine, which isn't bad at all."
Winstone 2004
"The MMCC Winstone 2004 test does have a highly multithreaded component, but the performance benefit from having four cores is less than 4%. Still, the Mac Pro at 3.0GHz ends up being slightly faster than the fastest desktop PC processor."
3D Rendering
"The multithreaded CineBench test shows us similar scaling to what we saw in 3dsmax; the move to four cores gives the Mac Pro a 65% boost in performance"
'nuff said there.
Encoding
Anywhere from -5% to +30%.
"With only two cores enabled the Mac Pro at 3.0GHz is about 5% slower than the Core 2 Extreme X6800, once again thanks to its high latency FB-DIMMs that erase any benefits the faster FSB and clock speeds would have provided. With all four cores enabled however, the Mac Pro manages to perform our encoding test in a mere 28 seconds - an improvement of almost 30%"
It was iTunes that sucked and not QT. My bad.
BenRoethig: 0% correct.
How you turn 5%-10% performance delta into 10%-15% and expect anyone to believe your lies when the article is linked for all to read is beyond me. Some areas the Mac Pro did worse, other areas it did far better but for most things there were minor differences even without 2 of the cores running.
Vinea
And how many in this segment are going to break Apple's warranty to make use of that? It's not as easy a task as EATX PC workstations. You also have to come up with a heat sink.
The same number that has been popping in upgrades to the G4 Powermacs.
Given that enough sites have done CPU swaps on the Mac Pro its not impossible for folks to do.
Vinea
Apple i'm warning you, you better come out with a mac pro soon or i'm gonna /throat
Maybe in June....He Said in Season Spring
Quarters are often like this:
Jan-Mar - Winter
April-Jun - Spring
Jul-Sept - Summer
Oct-Dec - Fall
then...........
Just so we're clear, Intel's 5000X chipset allows for 1 or 2 CPUs, and the Memory Controller requieres FB-DIMMs, but like H said, it may change with the next (or the one after that) generation of chipset.
There is another Xeon chipset (3000 series) that works with single CPUs (Conroe-based Xeons 3000 series) and requieres (only) DDR2 533/667 unbuffered RAM. It only allows for 8GB of RAM (instead of 64GB) and the south bridge doesn't have a lot of PCIe lanes (16). Anyway Intel sees it more as a small (lower-cost) server platform than a workstation chipset.
The 3000P is designed for entry level servers. The reason it has so few PCI-E lanes is that they already have the 975x for the desktop/workstation segment. The only difference between the Core 2 Duo and the Xeon 3000 series CPUs is the inclusion of ECC ram support.
The same number that has been popping in upgrades to the G4 Powermacs.
Given that enough sites have done CPU swaps on the Mac Pro its not impossible for folks to do.
Vinea
The G4s used a proprietary cpu card that covered both processors with a heat sink to match. The U3L motherboard used in SP G3s did not have the second sockets so upgrading to dual CPUs was impossible. As for the Mac Pro, it features twin proprietary heat sinks. While the socket would be there on a single cpu model, the heat sink would not.
I'm buying into Vinea's single CPU Mac Pro at $1699. That'll fill in the gap pretty nicely without having to redesign a new motherboard for a Conroe box.
Intel is familiar with the Mac Pro's form factor. It wouldn't be that hard to fit a version of the 965P or 975X to match. besides it wouldn't be the first time Apple used a slightly different chipset for a single processor model.
While the socket would be there on a single cpu model, the heat sink would not.
Yes, so Apple could sell a "second CPU upgrade kit" that is fitted by your nearest Apple Store, and includes a heat sink.
besides it wouldn't be the first time Apple used a slightly different chipset for a single processor model.
Those situation have normally arisen from Apple using their previous-generation chipset/motherboard in the bottom-of-the range machine. In the case of the Mac Pro, there is no previous generation motherboard.
Are you denying that initiating the offering of a Mac Pro with one CPU instead of two requires vastly less investment than initiating the offering of a Mac Pro with an entirely new motherboard?
Yes, so Apple could sell a "second CPU upgrade kit" that is fitted by your nearest Apple Store, and includes a heat sink.
It's also not like you couldn't build a new heatsink for an upgrade. Crucial added Mac pro compatible FB-DIMM hs in short order. There's currently no market because anyone upgrading a Mac Pro today has 2 working heat sinks.
Not that too many folks are upgrading to Cloverton at the moment anyway...
Vinea
Yes, so Apple could sell a "second CPU upgrade kit" that is fitted by your nearest Apple Store, and includes a heat sink.
Doesn't seem very Apple like.
Those situation have normally arisen from Apple using their previous-generation chipset/motherboard in the bottom-of-the range machine. In the case of the Mac Pro, there is no previous generation motherboard.
Not quite. Apple used a family of North bridges for the G5 called U3. It came in three different variants. The U3L was used in the low end PowerMac and iMac. It did not feature SMP support or PCI-X support. The general variant was the U3, it featured PCI-X support, but no SMP support. The SMP versions and the G5 xServe used the U3H. While they were related they needed slightly different driver sets. At the same time they also had to deal with the intrepid chipset on Mac Mini, eMac, and noteboooks.
Are you denying that initiating the offering of a Mac Pro with one CPU instead of two requires vastly less investment than initiating the offering of a Mac Pro with an entirely new motherboard?
A little less investment for Apple, but a lot more investment for the user. It would be slightly easier, but you can kiss the possibility of higher end desktop switchers goodbye and with them the general users they advise.
Maybe in June....He Said in Season Spring
Quarters are often like this:
Jan-Mar - Winter
April-Jun - Spring
Jul-Sept - Summer
Oct-Dec - Fall
then...........
I thought that quarter 1 included xmas shopping?
Are you denying that initiating the offering of a Mac Pro with one CPU instead of two requires vastly less investment than initiating the offering of a Mac Pro with an entirely new motherboard?
The discussion has been centered on using the Mac Pro to make a cheaper, single CPU tower. However, I don't like any approach that uses the Mac Pro chassis. It might serve as a short term trial run, but it couldn't sell as well as a smaller tower with high performance -- a competitive prosumer Mac that is expandable.
Any approach that uses one Xeon has the added cost of a workstation CPU and more expensive memory, if I have been reading the comments correctly. Its only advantage is the low development cost.
Designing a new motherboard has the advantage of lower cost CPU and RAM, at the expense of higher development cost. However the Mac Pro chassis and other part still keep the parts cost higher than necessary. Apple could replace the power supply, and look for other ways to reduce component cost. Yet at this point it may as well be a whole new design.
Even if Apple took one of these approaches using the Mac Pro chassis, I'm not sure the final product would get good marketing feedback either. I'd say forget the Mac Pro as a test bed, and just design a competitive mini tower in the prosumer price range. It's my opinion.
Bingo ! Hell, yes - something slimline 1RU a bit like the xServe. Along those lines.
Or a tower.
The discussion has been centered on using the Mac Pro to make a cheaper, single CPU tower. However, I don't like any approach that uses the Mac Pro chassis. It might serve as a short term trial run, but it couldn't sell as well as a smaller tower with high performance -- a competitive prosumer Mac that is expandable.
Any approach that uses one Xeon has the added cost of a workstation CPU and more expensive memory, if I have been reading the comments correctly. Its only advantage is the low development cost.
Designing a new motherboard has the advantage of lower cost CPU and RAM, at the expense of higher development cost. However the Mac Pro chassis and other part still keep the parts cost higher than necessary. Apple could replace the power supply, and look for other ways to reduce component cost. Yet at this point it may as well be a whole new design.
Even if Apple took one of these approaches using the Mac Pro chassis, I'm not sure the final product would get good marketing feedback either. I'd say forget the Mac Pro as a test bed, and just design a competitive mini tower in the prosumer price range. It's my opinion.
I think we're talking about multiple headless segments here. I think you're looking for a lower end mini tower or cube somewhat equivalent to the PC side's MATX towers. Possibly something iMac based, but with a PCI-E x16 slot and maybe and x1. Me, I'm looking for Apple's evolution of the full tower design with a 2.4ghz CPU. Anything like a cube I have to put it on my desk and anything too short I really have to reach for underneath.