Well, if you buy AppleCare, that expires in 3 years. At that point, you may want to pick up a pair of 3.2 Ghz Woodcrests (coming late 2006) or a pair of Clovertowns (4-core), which should be pretty cheap (since we'll have 3.5 GHz 4-8 core chips (non-MCM) by then. So you'd be picking up maybe $500-600 in processors to add another 2 years to the computer's life.
Can someone explain the big attraction of swappable chips, and what timeframe it would be useful for?
At this point, 3.0GHz Xeon (Woodcrest) chips cost just shy of $1000... EACH.
Replacing the 2.66 chips from a standard config Mac Pro to gain a 13% speed advantage would be uneconomic at this time. How long until it becomes so?
I think we'll see higher clocked chips though. It would probably be a viable upgrade of a period of one to four years from now. Sooner than one year or later than four then the cost might not be worth it, but that depends on how much faster compatible chips will be made.
The price for the very top end chips compatible with my existing Xeon systems (2.8GHz was the max, I have 1.7Ghz) had a typical price of $200 each in the last year, but as supply dwindled, the price crept up to closer to $300 assuming they can still be found. One might say that I should have just replaced the computer but I like the overall systems and they've been very reliable.
My point was that it's like if you bought a PentiumD when they came out in a new computer. Now you can grab a serious upgrade (in the form of a higher-clocked PentiumD) for like $150. If you can't afford a good Conroe, that'll buy you another year or two.
It'll be the same way with Mac Pros - you upgrade to extend your purchase cycle from 3 years to 4-5. You don't upgrade every year (at least not the processor), and you don't wait until the parts are all gone.
That means data is being sent four times per clock over the same set of lines, a new set of data is sent at each rise and fall of a quadrature clock. The G4 didn't have that, it sent data over one time per clock, having 1/4th the FSB bandwidth per clock. As such, the multiplier alone (the kind that you used) doesn't tell everything. The multiplier I used was more indicative of the actual data bandwidth the CPU gets. The G4 got 1.3GBps max bandwidth, each core on a Woodrest system gets about 5 GBps FSB bandwidth.
The best article, so far on this, is from Anand. He is discussing the Mac Pro, but as you mentioned Woodcrest, I though it would be relevent.
It's pretty long. You can skip to the memory portion (but you won't )
Can someone explain the big attraction of swappable chips, and what timeframe it would be useful for?
At this point, 3.0GHz Xeon (Woodcrest) chips cost just shy of $1000... EACH.
Replacing the 2.66 chips from a standard config Mac Pro to gain a 13% speed advantage would be uneconomic at this time. How long until it becomes so?
In the Anand article that I posted in the post above, he reccommends getting the 2 GHz version, if you are thinking about upograding, then going for Cloverton four core chips next year.
That would make it worthwhile!. If you can manage to use the *slow* 2 GHz model for six, or so, months.
The only problem I see with buying the 2ghz model is it's a bit expensive.
The difference between the 2.66Ghz MacPro and the 2Ghz model is $300 whereas the difference between two 2.66Ghz and two 2Ghz Xeons is $800 (at most online retailers). Apple are pocketing $500 pure profit there.
In theory, the 2Ghz model should be $1699 not $2199. What's galling for us Brits is our pricing is $2400+ for the 2Ghz, $2700+ for the 2.66 before taxes. Knowing the 2Ghz should be under £900 instead of £1275 takes the shine off it.
The only problem I see with buying the 2ghz model is it's a bit expensive.
The difference between the 2.66Ghz MacPro and the 2Ghz model is $300 whereas the difference between two 2.66Ghz and two 2Ghz Xeons is $800 (at most online retailers). Apple are pocketing $500 pure profit there.
In theory, the 2Ghz model should be $1699 not $2199. What's galling for us Brits is our pricing is $2400+ for the 2Ghz, $2700+ for the 2.66 before taxes. Knowing the 2Ghz should be under £900 instead of £1275 takes the shine off it.
I won't argue with that. It's not my recommendation, but his. His thought is that you might want the machine now, but would rather save, even the $300, for the faster, and even the quad cores later. If the 2's are much faster than what you've got now, perhaps that would be worthwhile. I would still go for the 2.67's if I were to do that, and were somwhat strapped for cash. But, it might pay to get the extra RAM, to bring them up to 4 DIMMS, to get the 4 memory channels instead of the faster chips now.
The only problem I see with buying the 2ghz model is it's a bit expensive.
The difference between the 2.66Ghz MacPro and the 2Ghz model is $300 whereas the difference between two 2.66Ghz and two 2Ghz Xeons is $800 (at most online retailers). Apple are pocketing $500 pure profit there
Or do you think, as is more likely the case in light of comparisons to Dell, that Apple is taking a loss on the 2.00 and making it up with the less-expensive-in-reality jump from 2.66 to 3.00?
I won't argue with that. It's not my recommendation, but his. His thought is that you might want the machine now, but would rather save, even the $300, for the faster, and even the quad cores later. If the 2's are much faster than what you've got now, perhaps that would be worthwhile. I would still go for the 2.67's if I were to do that, and were somwhat strapped for cash. But, it might pay to get the extra RAM, to bring them up to 4 DIMMS, to get the 4 memory channels instead of the faster chips now.
I guess it's a tough decision for some.
I think it's an easier decision to downgrade the hard drive to get the 160GB. The difference is $75. Newegg has 250GB SATAII drives for $80, so it's a nearly free backup drive.
I'd probably just keep the standard CPU as given rather than go low in the hopes of a viable upgrade months or years from now, or wait. 2.66GHz chips might be a lot more resellable to make up the $300 difference too.
Did I correctly read that the FB-DIMM bus was lower in bandwidth than what the DDR2-667 indicates? Would dropping to FB-DIMM-533 be detrimental to performance?
I think it's an easier decision to downgrade the hard drive to get the 160GB. The difference is $75. Newegg has 250GB SATAII drives for $80, so it's a nearly free backup drive.
I'd probably just keep the standard CPU as given rather than go low in the hopes of a viable upgrade months or years from now, or wait. 2.66GHz chips might be a lot more resellable to make up the $300 difference too.
Did I correctly read that the FB-DIMM bus was lower in bandwidth than what the DDR2-667 indicates? Would dropping to FB-DIMM-533 be detrimental to performance?
No, it has greater bandwidth. Using four channels, it is greater than 20 GB/s. Also, read and writes are done at the same time without causing contention on the memory bus. Yes dropping to 533 would cause a drop. In fact, I thought it was supposed to support 800. I suppose they are still too expensice. Perhaps in January.
I haven't seen a FB-DIMM-800 anywhere on newegg or any other consumer memory sales site. Whereas they've had FB-DIMMs for a while (a few months at least)
Or do you think, as is more likely the case in light of comparisons to Dell, that Apple is taking a loss on the 2.00 and making it up with the less-expensive-in-reality jump from 2.66 to 3.00?
No. I think the 2.66 seems to be reasonably priced but the Apple are taking an extra $500 either way. Both the 2.0 and 3.0 are $500 too expensive if you just go off the price of the CPUs, which is the only difference after all.
I haven't seen a FB-DIMM-800 anywhere on newegg or any other consumer memory sales site. Whereas they've had FB-DIMMs for a while (a few months at least)
And the ones that are available seem to be suffering in MacPros because they don't have large enough heatsinks. Crucial in particular have had issues.
And the ones that are available seem to be suffering in MacPros because they don't have large enough heatsinks. Crucial in particular have had issues.
Once I heard of the special considerations for memory cooling, I wondered if there will be a premium. Right now, OWC's pricing is a lot closer to Apple's than standard FB-DIMMs on Newegg. The memory configurator on Crucial's and Kingston's site seem to indicate that they will offer such memory but it's not available yet.
Once I heard of the special considerations for memory cooling, I wondered if there will be a premium. Right now, OWC's pricing is a lot closer to Apple's than standard FB-DIMMs on Newegg. The memory configurator on Crucial's and Kingston's site seem to indicate that they will offer such memory but it's not available yet.
It was available but then Crucial took it off again.
Apparently if you put the RAM in the slots furthest away from the Xeons, normal FBDIMMs have a chance of working. Apple use huge heatsinks so that they don't need as much active cooling across the memory and therefore less noise.
Normal FB-DIMMs work, it's a fan issue. Heatsink = less fan.
That's one way of looking at it.
The correct way is that Apple designed the Mac Pro to be quiet. In order to do that, you need more passive cooling (ie. bigger heatsinks)
I suspect once they process shrink the processor on the FB-DIMMs it'll work out so you don't need such big heatsinks. It also almost certainly doesn't help that Apple have put the RAM right next to the CPU.
which slot is furthest from the Xeons? Apparently there's an order you're supposed to put them in...
Normal FB-DIMMs work, it's a fan issue. Heatsink = less fan.
I prefer the concept of heatsinking the DIMMS properly, rather than having to use extra large fans running as high speed.
A passive solution is ALWAYS preferable. The less complexity the better.
I always did my best to design fans out of a system, if at all possible. We were the first company to use heat pipes in high power amplifiers. More reliable than fans. More expensive too.
I prefer the concept of heatsinking the DIMMS properly, rather than having to use extra large fans running as high speed.
A passive solution is ALWAYS preferable. The less complexity the better.
Concur 100%. I was just pointing out for anyone on the thread who missed it that you CAN use the regular FB-DIMMs if you're careful (put them in the slot Apple uses now) and willing to deal with some noise. I don't recommend it, but it's possible.
Comments
At this point, 3.0GHz Xeon (Woodcrest) chips cost just shy of $1000... EACH.
Replacing the 2.66 chips from a standard config Mac Pro to gain a 13% speed advantage would be uneconomic at this time. How long until it becomes so?
Can someone explain the big attraction of swappable chips, and what timeframe it would be useful for?
At this point, 3.0GHz Xeon (Woodcrest) chips cost just shy of $1000... EACH.
Replacing the 2.66 chips from a standard config Mac Pro to gain a 13% speed advantage would be uneconomic at this time. How long until it becomes so?
I think we'll see higher clocked chips though. It would probably be a viable upgrade of a period of one to four years from now. Sooner than one year or later than four then the cost might not be worth it, but that depends on how much faster compatible chips will be made.
The price for the very top end chips compatible with my existing Xeon systems (2.8GHz was the max, I have 1.7Ghz) had a typical price of $200 each in the last year, but as supply dwindled, the price crept up to closer to $300 assuming they can still be found. One might say that I should have just replaced the computer but I like the overall systems and they've been very reliable.
It'll be the same way with Mac Pros - you upgrade to extend your purchase cycle from 3 years to 4-5. You don't upgrade every year (at least not the processor), and you don't wait until the parts are all gone.
That means data is being sent four times per clock over the same set of lines, a new set of data is sent at each rise and fall of a quadrature clock. The G4 didn't have that, it sent data over one time per clock, having 1/4th the FSB bandwidth per clock. As such, the multiplier alone (the kind that you used) doesn't tell everything. The multiplier I used was more indicative of the actual data bandwidth the CPU gets. The G4 got 1.3GBps max bandwidth, each core on a Woodrest system gets about 5 GBps FSB bandwidth.
The best article, so far on this, is from Anand. He is discussing the Mac Pro, but as you mentioned Woodcrest, I though it would be relevent.
It's pretty long. You can skip to the memory portion (but you won't
http://www.anandtech.com/mac/showdoc.aspx?i=2811&p=1
Can someone explain the big attraction of swappable chips, and what timeframe it would be useful for?
At this point, 3.0GHz Xeon (Woodcrest) chips cost just shy of $1000... EACH.
Replacing the 2.66 chips from a standard config Mac Pro to gain a 13% speed advantage would be uneconomic at this time. How long until it becomes so?
In the Anand article that I posted in the post above, he reccommends getting the 2 GHz version, if you are thinking about upograding, then going for Cloverton four core chips next year.
That would make it worthwhile!. If you can manage to use the *slow* 2 GHz model for six, or so, months.
The difference between the 2.66Ghz MacPro and the 2Ghz model is $300 whereas the difference between two 2.66Ghz and two 2Ghz Xeons is $800 (at most online retailers). Apple are pocketing $500 pure profit there.
In theory, the 2Ghz model should be $1699 not $2199. What's galling for us Brits is our pricing is $2400+ for the 2Ghz, $2700+ for the 2.66 before taxes. Knowing the 2Ghz should be under £900 instead of £1275 takes the shine off it.
The only problem I see with buying the 2ghz model is it's a bit expensive.
The difference between the 2.66Ghz MacPro and the 2Ghz model is $300 whereas the difference between two 2.66Ghz and two 2Ghz Xeons is $800 (at most online retailers). Apple are pocketing $500 pure profit there.
In theory, the 2Ghz model should be $1699 not $2199. What's galling for us Brits is our pricing is $2400+ for the 2Ghz, $2700+ for the 2.66 before taxes. Knowing the 2Ghz should be under £900 instead of £1275 takes the shine off it.
I won't argue with that. It's not my recommendation, but his. His thought is that you might want the machine now, but would rather save, even the $300, for the faster, and even the quad cores later. If the 2's are much faster than what you've got now, perhaps that would be worthwhile. I would still go for the 2.67's if I were to do that, and were somwhat strapped for cash. But, it might pay to get the extra RAM, to bring them up to 4 DIMMS, to get the 4 memory channels instead of the faster chips now.
I guess it's a tough decision for some.
The only problem I see with buying the 2ghz model is it's a bit expensive.
The difference between the 2.66Ghz MacPro and the 2Ghz model is $300 whereas the difference between two 2.66Ghz and two 2Ghz Xeons is $800 (at most online retailers). Apple are pocketing $500 pure profit there
Or do you think, as is more likely the case in light of comparisons to Dell, that Apple is taking a loss on the 2.00 and making it up with the less-expensive-in-reality jump from 2.66 to 3.00?
I won't argue with that. It's not my recommendation, but his. His thought is that you might want the machine now, but would rather save, even the $300, for the faster, and even the quad cores later. If the 2's are much faster than what you've got now, perhaps that would be worthwhile. I would still go for the 2.67's if I were to do that, and were somwhat strapped for cash. But, it might pay to get the extra RAM, to bring them up to 4 DIMMS, to get the 4 memory channels instead of the faster chips now.
I guess it's a tough decision for some.
I think it's an easier decision to downgrade the hard drive to get the 160GB. The difference is $75. Newegg has 250GB SATAII drives for $80, so it's a nearly free backup drive.
I'd probably just keep the standard CPU as given rather than go low in the hopes of a viable upgrade months or years from now, or wait. 2.66GHz chips might be a lot more resellable to make up the $300 difference too.
Did I correctly read that the FB-DIMM bus was lower in bandwidth than what the DDR2-667 indicates? Would dropping to FB-DIMM-533 be detrimental to performance?
I think it's an easier decision to downgrade the hard drive to get the 160GB. The difference is $75. Newegg has 250GB SATAII drives for $80, so it's a nearly free backup drive.
I'd probably just keep the standard CPU as given rather than go low in the hopes of a viable upgrade months or years from now, or wait. 2.66GHz chips might be a lot more resellable to make up the $300 difference too.
Did I correctly read that the FB-DIMM bus was lower in bandwidth than what the DDR2-667 indicates? Would dropping to FB-DIMM-533 be detrimental to performance?
No, it has greater bandwidth. Using four channels, it is greater than 20 GB/s. Also, read and writes are done at the same time without causing contention on the memory bus. Yes dropping to 533 would cause a drop. In fact, I thought it was supposed to support 800. I suppose they are still too expensice. Perhaps in January.
Or do you think, as is more likely the case in light of comparisons to Dell, that Apple is taking a loss on the 2.00 and making it up with the less-expensive-in-reality jump from 2.66 to 3.00?
No. I think the 2.66 seems to be reasonably priced but the Apple are taking an extra $500 either way. Both the 2.0 and 3.0 are $500 too expensive if you just go off the price of the CPUs, which is the only difference after all.
I haven't seen a FB-DIMM-800 anywhere on newegg or any other consumer memory sales site. Whereas they've had FB-DIMMs for a while (a few months at least)
And the ones that are available seem to be suffering in MacPros because they don't have large enough heatsinks. Crucial in particular have had issues.
And the ones that are available seem to be suffering in MacPros because they don't have large enough heatsinks. Crucial in particular have had issues.
Once I heard of the special considerations for memory cooling, I wondered if there will be a premium. Right now, OWC's pricing is a lot closer to Apple's than standard FB-DIMMs on Newegg. The memory configurator on Crucial's and Kingston's site seem to indicate that they will offer such memory but it's not available yet.
Once I heard of the special considerations for memory cooling, I wondered if there will be a premium. Right now, OWC's pricing is a lot closer to Apple's than standard FB-DIMMs on Newegg. The memory configurator on Crucial's and Kingston's site seem to indicate that they will offer such memory but it's not available yet.
It was available but then Crucial took it off again.
Apparently if you put the RAM in the slots furthest away from the Xeons, normal FBDIMMs have a chance of working. Apple use huge heatsinks so that they don't need as much active cooling across the memory and therefore less noise.
See http://www.xlr8yourmac.com/index.html#S21538
Normal FB-DIMMs work, it's a fan issue. Heatsink = less fan.
Normal FB-DIMMs work, it's a fan issue. Heatsink = less fan.
That's one way of looking at it.
The correct way is that Apple designed the Mac Pro to be quiet. In order to do that, you need more passive cooling (ie. bigger heatsinks)
I suspect once they process shrink the processor on the FB-DIMMs it'll work out so you don't need such big heatsinks. It also almost certainly doesn't help that Apple have put the RAM right next to the CPU.
which slot is furthest from the Xeons? Apparently there's an order you're supposed to put them in...
Normal FB-DIMMs work, it's a fan issue. Heatsink = less fan.
I prefer the concept of heatsinking the DIMMS properly, rather than having to use extra large fans running as high speed.
A passive solution is ALWAYS preferable. The less complexity the better.
I always did my best to design fans out of a system, if at all possible. We were the first company to use heat pipes in high power amplifiers. More reliable than fans. More expensive too.
I prefer the concept of heatsinking the DIMMS properly, rather than having to use extra large fans running as high speed.
A passive solution is ALWAYS preferable. The less complexity the better.
Concur 100%. I was just pointing out for anyone on the thread who missed it that you CAN use the regular FB-DIMMs if you're careful (put them in the slot Apple uses now) and willing to deal with some noise. I don't recommend it, but it's possible.