I guess I should clarify. I've heard FUD from people who think that even apps that do heavy number crunching and could benefit from more than 4 gigs of ram shouldn't be updated, because there will be a huge performance hit. I have yet to see any real data backing that up, just seems to be paranoid speculation. Obviously, it would be silly to update Textedit to 64 bit, but the notion that every app would take a performance hit seems pretty ridiculous.
That was a stretch. For more than 4 gigs per app, 64 bit should be a lot better. But I forget who it was (emig?), someone here claimed that even their own code ran a bit slower in 64 bit mode just because of the way object oriented code works, I assume this was for smaller code.
Quote:
That's not what the article says. All three models are listed as "early 2008". I have no idea why the 3.0 would lag the 2.8 on any tests, that does seem fishy.
My mistake. Somehow I didn't catch that there was a new 3.0 model.
I guess I should clarify. I've heard FUD from people who think that even apps that do heavy number crunching and could benefit from more than 4 gigs of ram shouldn't be updated, because there will be a huge performance hit. I have yet to see any real data backing that up, just seems to be paranoid speculation. Obviously, it would be silly to update Textedit to 64 bit, but the notion that every app would take a performance hit seems pretty ridiculous.
Technically all apps do take a significant performance hit, but apps that do large-domain number crunching (not just "heavy"... if you only need to deal with integers below 4 billion it doesn't help), access lots of memory, or chug through blocks of data in the CPU make that performance hit back up and more. Basically, if you switch to 64-bits without needing to you're piping huge numbers of 0's around the memory bus for no use whatsoever, hurting performance.
If your app doesn't deal with >4B numbers then you don't get any advantage from registers twice as wide but it may have still taken twice the number of cycles to pump those 0's in. If your app doesn't deal with huge memory but you compile it with that memory model then every "jump" has to read in addresses that are twice as long. ("Relative" addressing, such as tight loops, shouldn't change much, but calling a method in a library could take longer.)
Anyway, as I mentioned on x86 the "64 bit" ISA is actually a lot more than just wider integers and addresses, so yeah, you're going to get a speedup.
I would not assume you can easily add a second quad-core processor at a later date to the single quad-core configuration. Not sure whether Apple will ship those machines with a different heatsink, some sort of thermal recalibration will probably have to be run from an Apple Service Diagnostic disc, and it is likely that the retail price of a 2.8GHz Harpertown processor will greatly exceed the $500 Apple essentially charges for it for quite some time.
Despite that, the new quad-core 2.8GHz Mac Pro is a much better buy as a low-end machine than the old 2.0GHz Mac Pro, which I assume nobody even bought. But I think it's better to buy the octo-core now and just be done with it.
Technically all apps do take a significant performance hit, but apps that do large-domain number crunching (not just "heavy"... if you only need to deal with integers below 4 billion it doesn't help), access lots of memory, or chug through blocks of data in the CPU make that performance hit back up and more.
So you're saying all apps take a hit...but some apps don't.
So what you're really saying is that some apps take a hit and some don't.
If an app has both penalties and gains, and the gains outweigh the penalties, I wouldn't consider that a net loss.
I would not assume you can easily add a second quad-core processor at a later date to the single quad-core configuration. Not sure whether Apple will ship those machines with a different heatsink, some sort of thermal recalibration will probably have to be run from an Apple Service Diagnostic disc, and it is likely that the retail price of a 2.8GHz Harpertown processor will greatly exceed the $500 Apple essentially charges for it for quite some time.
Notice I that I did say presumably? Apple could cripple the single cpu model to keep it from hurting sales of the 2 cpu beast. I don't think that they'll do that but you never know...
As to the pricing of a 2.8 Harpertown cpu, I wouldn't be surprised if that could be bought for $350 in 6 months. Intel are quite aggressive on pricing. If AMD are able to ramp up clock speeds on their Barcelona chips, Intel will drop their prices to in order to continue to hammer AMD.
Man, I was going to get the 3ghz machine, but now i'm not so sure that it's worth the extra $800. I have always gone with the middle of the road machine believing that they are the best quality because of manufacturing processes. The thicker/denser plates getting the hotter/faster speeds, and the thinner being the slower speeds. Guess I transfered this from RAM manufacturing. Not sure if it applies to processors at all?
Can anyone 'yah' or 'nah' me on this?
Anyways, from what i'm seeing, there isn't much of a performance increase from the 2.8 to the 3, and the only reason for going to a 3 is it sounds better for re-sale in few years.
Man, I was going to get the 3ghz machine, but now i'm not so sure that it's worth the extra $800. I have always gone with the middle of the road machine believing that they are the best quality because of manufacturing processes. The thicker/denser plates getting the hotter/faster speeds, and the thinner being the slower speeds. Guess I transfered this from RAM manufacturing. Not sure if it applies to processors at all?
Can anyone 'yah' or 'nah' me on this?
Anyways, from what i'm seeing, there isn't much of a performance increase from the 2.8 to the 3, and the only reason for going to a 3 is it sounds better for re-sale in few years.
I don't think Apple changes anything about the system with the clock speed except sometimes at the very highest end. It's a higher binned part, but that's a speed rating bin, not a quality rating bin.
I would not assume you can easily add a second quad-core processor at a later date to the single quad-core configuration. Not sure whether Apple will ship those machines with a different heatsink, some sort of thermal recalibration will probably have to be run from an Apple Service Diagnostic disc, and it is likely that the retail price of a 2.8GHz Harpertown processor will greatly exceed the $500 Apple essentially charges for it for quite some time.
Despite that, the new quad-core 2.8GHz Mac Pro is a much better buy as a low-end machine than the old 2.0GHz Mac Pro, which I assume nobody even bought. But I think it's better to buy the octo-core now and just be done with it.
Yeah, a MUCH better deal! If I purchase the quad, I don't plan on trying to install another processor, just trying to save some cash! You make a great point though, for an extra $500 I could get the 2nd processor at a nicely discounted price (vs retail). I just don't think I really need the 8 cores................
The quad core would still be a insane improvement over my 2.0GHz core duo iMac!! So the question now is... buy the quad NOW or save up an extra $500 and be able to keep my computer for another year or two
The quad core would still be a insane improvement over my 2.0GHz core duo iMac!! So the question now is... buy the quad NOW or save up an extra $500 and be able to keep my computer for another year or two
If my quicksilver had a second G4 CPU it'd be running Leopard now. Alas, it is just below the cut point.
It is't actually that unusual when two otherwise identical machines with processors that are so close in speed reverse specs in some areas.
The cpu's may not be running at exactly the speed issues. Memory may be a bit fadter, or slower. The bus may have tiny glitches, etc. It's just a few percent.
So if this is true, it seems like spending an extra $800 to get the 3.0 Ghz is pretty much a waste. But the "few percent" is the same as the difference between the 3.0 Ghz and the 3.2 Ghz, so why are the numbers consistently higher for the 3.2?
Anyone got a more convincing explanation for those numbers on the 3.0? I'm trying to decide whether it's worth the extra money for the 3.0 over the 2.8...
If my quicksilver had a second G4 CPU it'd be running Leopard now. Alas, it is just below the cut point.
If you are talking about the G4 Quicksilver of July 2001, then your ability to run Leopard has little to do with a requirement for a 2nd CPU. You could have opted for higher running single CPU running at 867 MHz. But now that would have costs you an extra $800, but then you could run Leopard. If you would have opted for the high end 2 CPU's, then you would have to spend an extra $1800. And that's in 2001 dollars.
But don't lament. You made the best choice. So now you have all that available money to spend on a new Mac Pro with all kind of cores. And now the Mac Pros don't seem to be expensive at all in comparison to the 2001 prices. Very smart move.
I have a G4 Digital Audio (466) and Mac Pros seem to be an excellent price. Especially considering how long my G4 has been running and is running OX 10.4.11, which a very stable system.
Quote:
The Quicksilver PowerMac was available in three configurations: The 733 MHz model, with 128 MB of RAM, a 40 GB hard drive, and a CD-RW drive, was $1,699, the 867 MHz configuration, with 128 MB of RAM, a 60 GB hard drive and a DVD-R drive, was $2,499, and the high-end dual-800 MHz model, with 256 MB of RAM, an 80 GB hard drive and a DVD-R drive, was $3,499
I've done a lot of 32-bit and 64-bit development and there sure is a lot of FUD out there.
It is ABSOLUTELY incorrect to say that 64-bit compiled programs will be slower. They will use (slightly) more memory.
I see this all the time - Intel 64 bit compiled code will run 25-30% faster than 32-bit code on the same machine.
In general they are significantly faster , particularly on Intel for the following reasons.
- code compiled for 32-bit Intel is generally compiled for the lowest common denominator Intel 32-bit processor, typically a Pentium III or Pentium 4 for compatibility reasons
- code compiled for 64-bit processors is free to be optimized by the compiler for the all the new 64-bit instructions and capabilities.
A 64-bit program will use more memory, but only because
- pointers are now 64-bit. If you store a lot of pointers, then memory use will go up.
- the C/C++ "long" is now "64-bit". So data stored as "long" will use twice as much space.
But typically, programmers do not use 'long" as such, They will use int32 or int64 or some typedef so that there is no ambiguity about the size of the number. If it will fit into 32-bits , then one would use, say int32 not "long".
So in summary,
C/C++ CODE compiled for 64-bit will run faster
DATA in 64-bit programs may occupy more memory depending on the application.
That was a stretch. For more than 4 gigs per app, 64 bit should be a lot better. But I forget who it was (emig?), someone here claimed that even their own code ran a bit slower in 64 bit mode just because of the way object oriented code works, I assume this was for smaller code.
Some apps will run slower, and some will run faster. It depends on how efficient they are in 64 bit mode, if they can use the wider, deeper registers effectively, etc.
Notice I that I did say presumably? Apple could cripple the single cpu model to keep it from hurting sales of the 2 cpu beast. I don't think that they'll do that but you never know...
As to the pricing of a 2.8 Harpertown cpu, I wouldn't be surprised if that could be bought for $350 in 6 months. Intel are quite aggressive on pricing. If AMD are able to ramp up clock speeds on their Barcelona chips, Intel will drop their prices to in order to continue to hammer AMD.
It's not proper to say that Apple would "cripple" the machine. That's a pejorative term that is entirely out of place.
If Apple supplies a machine with one socket filled, then there would be no reason for them to supply an additional cooling system for the socket that had none. If they did that, then people would have to pay for it.
Considering the fact that very few will ever upgrade the machine to two processors, that would be unfairly charging the big majority. Then those people could rightly calim that Apple deliberately left it in so that they would be charged more.
What really needs to be done, is to convince a third party cpu upgrade company, such as OWC, to come out with chip upgrades. With Intel chips so easy for them to purchase, unlike the G5's, there should be no problem for them. The difference between older, upgradeable machines, and the ones in the Aluminum tower case, is that with these machines, I would never recommend that anyone without experience take it apart so as to upgrade the cpu.
If you remember the old G4's that came with one chip, the heatsink was only for one chip. no one ever could upgrade without buying an entire board, complete with CPU, cache, heatsink, and fan.
Why do you think you should have to anything less now?
It's not proper to say that Apple would "cripple" the machine. That's a pejorative term that is entirely out of place.
If Apple supplies a machine with one socket filled, then there would be no reason for them to supply an additional cooling system for the socket that had none. If they did that, then people would have to pay for it.
I think backtomac might have been thinking of the possibility there wouldn't even be a second socket soldered in. If the socket's not there, then the upgrade path would surely be crippled.
Quote:
Considering the fact that very few will ever upgrade the machine to two processors, that would be unfairly charging the big majority. Then those people could rightly calim that Apple deliberately left it in so that they would be charged more.
I think it's unreasonable to expect a spare heat sink to be included, I don't think anyone else includes a heat sink for unpopulated sockets. The heat sink is included with the company's CPU upgrade kits.
So if this is true, it seems like spending an extra $800 to get the 3.0 Ghz is pretty much a waste. But the "few percent" is the same as the difference between the 3.0 Ghz and the 3.2 Ghz, so why are the numbers consistently higher for the 3.2?
Anyone got a more convincing explanation for those numbers on the 3.0? I'm trying to decide whether it's worth the extra money for the 3.0 over the 2.8...
It depends on how important the few percent is to you.
As far as the difference between the 2.8 and the 3.0 goes, apparently the majority of tests show the 3 to be faster. It's very possible that a few operation's speed is reversed in the different chips.
Something about bin speeds. Remember that these are supposed to be IDENTICAL chips. What that means, is that in theory, they should all be the SAME speed. There should be NO bin speed differences.
In practice, manufacturing isn't perfect. Ever so slight differences in line widths, depths, and spacing, add up to minute differences in capacitance. The higher the capacitance, the slower the speed.
BUT, these chips are hellishly complex. That means that on the same chip, there will be these variations. Where they are on the chip will decide how slow the chip is, and what part of the chip is affected the most.
So, one chip can have a silghtly higher bin speed overall, but still be slightly slower in some operations.
Each chip will be affected differently. Two chips with the same bin speed can have slight variations in operational "evenness". Some may have some operations a fair amount faster, and some a fair amount slower, and others may hover closely around the speed number given.
So, these tests we've read are valid for those machines. but, as we are talking about minute differences, too small to notice in most cases, they are not valid for a different set of machines because of the varibility in manufacturing.
This also holds true for the motherboard, as well as for the rest of the machine.
On average, each machine will be approximately spaced as they should be. It's possible that the 3.2 machine here was fast, but the 3.0 machine was slow, and that the 2.8 was also fast.
This is always going to happen when the speed differences between machines are so small. If we had a 2.6 a 2.9, and a 3.2, we wouldn't see this happening as much.
I think backtomac might have been thinking of the possibility there wouldn't even be a second socket soldered in. If the socket's not there, then the upgrade path would surely be crippled.
You might remember that I was saying that that could be possible as well.
But, that still isn't automatically crippling the machine. If it could be shown that it costs no more to have a complete two socket mobo, then Apple could be accused of "crippling" the machine. But, if it could be shown that it was even "slightly" cheaper to go with a single socket board, then the criticism would be unfair.
Quote:
I think it's unreasonable to expect a spare heat sink to be included, I don't think anyone else includes a heat sink for unpopulated sockets. The heat sink is included with the company's CPU upgrade kits.
I agree. And that's why this is a great opportunity for a third party to come out with an upgrade. Assuming it would be cheap enough.
But, as I also said, upgrading these machines is VERY difficult. I don't recommend that anyone who isn't experienced in electronics do it themselves.
Comments
I guess I should clarify. I've heard FUD from people who think that even apps that do heavy number crunching and could benefit from more than 4 gigs of ram shouldn't be updated, because there will be a huge performance hit. I have yet to see any real data backing that up, just seems to be paranoid speculation. Obviously, it would be silly to update Textedit to 64 bit, but the notion that every app would take a performance hit seems pretty ridiculous.
That was a stretch. For more than 4 gigs per app, 64 bit should be a lot better. But I forget who it was (emig?), someone here claimed that even their own code ran a bit slower in 64 bit mode just because of the way object oriented code works, I assume this was for smaller code.
That's not what the article says. All three models are listed as "early 2008". I have no idea why the 3.0 would lag the 2.8 on any tests, that does seem fishy.
My mistake. Somehow I didn't catch that there was a new 3.0 model.
I guess I should clarify. I've heard FUD from people who think that even apps that do heavy number crunching and could benefit from more than 4 gigs of ram shouldn't be updated, because there will be a huge performance hit. I have yet to see any real data backing that up, just seems to be paranoid speculation. Obviously, it would be silly to update Textedit to 64 bit, but the notion that every app would take a performance hit seems pretty ridiculous.
Technically all apps do take a significant performance hit, but apps that do large-domain number crunching (not just "heavy"... if you only need to deal with integers below 4 billion it doesn't help), access lots of memory, or chug through blocks of data in the CPU make that performance hit back up and more. Basically, if you switch to 64-bits without needing to you're piping huge numbers of 0's around the memory bus for no use whatsoever, hurting performance.
If your app doesn't deal with >4B numbers then you don't get any advantage from registers twice as wide but it may have still taken twice the number of cycles to pump those 0's in. If your app doesn't deal with huge memory but you compile it with that memory model then every "jump" has to read in addresses that are twice as long. ("Relative" addressing, such as tight loops, shouldn't change much, but calling a method in a library could take longer.)
Anyway, as I mentioned on x86 the "64 bit" ISA is actually a lot more than just wider integers and addresses, so yeah, you're going to get a speedup.
You can buy a Mac Pro with only one cpu and presumably the other socket ready for another cpu if you wish.
See here.
I would not assume you can easily add a second quad-core processor at a later date to the single quad-core configuration. Not sure whether Apple will ship those machines with a different heatsink, some sort of thermal recalibration will probably have to be run from an Apple Service Diagnostic disc, and it is likely that the retail price of a 2.8GHz Harpertown processor will greatly exceed the $500 Apple essentially charges for it for quite some time.
Despite that, the new quad-core 2.8GHz Mac Pro is a much better buy as a low-end machine than the old 2.0GHz Mac Pro, which I assume nobody even bought.
Technically all apps do take a significant performance hit, but apps that do large-domain number crunching (not just "heavy"... if you only need to deal with integers below 4 billion it doesn't help), access lots of memory, or chug through blocks of data in the CPU make that performance hit back up and more.
So you're saying all apps take a hit...but some apps don't.
So what you're really saying is that some apps take a hit and some don't.
If an app has both penalties and gains, and the gains outweigh the penalties, I wouldn't consider that a net loss.
I would not assume you can easily add a second quad-core processor at a later date to the single quad-core configuration. Not sure whether Apple will ship those machines with a different heatsink, some sort of thermal recalibration will probably have to be run from an Apple Service Diagnostic disc, and it is likely that the retail price of a 2.8GHz Harpertown processor will greatly exceed the $500 Apple essentially charges for it for quite some time.
Notice I that I did say presumably? Apple could cripple the single cpu model to keep it from hurting sales of the 2 cpu beast. I don't think that they'll do that but you never know...
As to the pricing of a 2.8 Harpertown cpu, I wouldn't be surprised if that could be bought for $350 in 6 months. Intel are quite aggressive on pricing. If AMD are able to ramp up clock speeds on their Barcelona chips, Intel will drop their prices to in order to continue to hammer AMD.
Can anyone 'yah' or 'nah' me on this?
Anyways, from what i'm seeing, there isn't much of a performance increase from the 2.8 to the 3, and the only reason for going to a 3 is it sounds better for re-sale in few years.
Man, I was going to get the 3ghz machine, but now i'm not so sure that it's worth the extra $800. I have always gone with the middle of the road machine believing that they are the best quality because of manufacturing processes. The thicker/denser plates getting the hotter/faster speeds, and the thinner being the slower speeds. Guess I transfered this from RAM manufacturing. Not sure if it applies to processors at all?
Can anyone 'yah' or 'nah' me on this?
Anyways, from what i'm seeing, there isn't much of a performance increase from the 2.8 to the 3, and the only reason for going to a 3 is it sounds better for re-sale in few years.
I don't think Apple changes anything about the system with the clock speed except sometimes at the very highest end. It's a higher binned part, but that's a speed rating bin, not a quality rating bin.
I would not assume you can easily add a second quad-core processor at a later date to the single quad-core configuration. Not sure whether Apple will ship those machines with a different heatsink, some sort of thermal recalibration will probably have to be run from an Apple Service Diagnostic disc, and it is likely that the retail price of a 2.8GHz Harpertown processor will greatly exceed the $500 Apple essentially charges for it for quite some time.
Despite that, the new quad-core 2.8GHz Mac Pro is a much better buy as a low-end machine than the old 2.0GHz Mac Pro, which I assume nobody even bought.
Yeah, a MUCH better deal! If I purchase the quad, I don't plan on trying to install another processor, just trying to save some cash! You make a great point though, for an extra $500 I could get the 2nd processor at a nicely discounted price (vs retail). I just don't think I really need the 8 cores................
The quad core would still be a insane improvement over my 2.0GHz core duo iMac!! So the question now is... buy the quad NOW or save up an extra $500 and be able to keep my computer for another year or two
The quad core would still be a insane improvement over my 2.0GHz core duo iMac!! So the question now is... buy the quad NOW or save up an extra $500 and be able to keep my computer for another year or two
If my quicksilver had a second G4 CPU it'd be running Leopard now. Alas, it is just below the cut point.
It is't actually that unusual when two otherwise identical machines with processors that are so close in speed reverse specs in some areas.
The cpu's may not be running at exactly the speed issues. Memory may be a bit fadter, or slower. The bus may have tiny glitches, etc. It's just a few percent.
So if this is true, it seems like spending an extra $800 to get the 3.0 Ghz is pretty much a waste. But the "few percent" is the same as the difference between the 3.0 Ghz and the 3.2 Ghz, so why are the numbers consistently higher for the 3.2?
Anyone got a more convincing explanation for those numbers on the 3.0? I'm trying to decide whether it's worth the extra money for the 3.0 over the 2.8...
It's kind of lame. Why not just put up a link and let people support the guys who actually did the benchmarks.
Do you guys get permission to redistribute this story? Or do you just make it into quotes and pretend like its yours.
It's kind of lame. Why not just put up a link and let people support the guys who actually did the benchmarks.
Uhhhhhhhhhhhh....... if it wasn't for AI, I wouldn't know of Primate Labs OR geekbench... get real!
If my quicksilver had a second G4 CPU it'd be running Leopard now. Alas, it is just below the cut point.
If you are talking about the G4 Quicksilver of July 2001, then your ability to run Leopard has little to do with a requirement for a 2nd CPU. You could have opted for higher running single CPU running at 867 MHz. But now that would have costs you an extra $800, but then you could run Leopard. If you would have opted for the high end 2 CPU's, then you would have to spend an extra $1800. And that's in 2001 dollars.
But don't lament. You made the best choice. So now you have all that available money to spend on a new Mac Pro with all kind of cores. And now the Mac Pros don't seem to be expensive at all in comparison to the 2001 prices. Very smart move.
I have a G4 Digital Audio (466) and Mac Pros seem to be an excellent price. Especially considering how long my G4 has been running and is running OX 10.4.11, which a very stable system.
The Quicksilver PowerMac was available in three configurations: The 733 MHz model, with 128 MB of RAM, a 40 GB hard drive, and a CD-RW drive, was $1,699, the 867 MHz configuration, with 128 MB of RAM, a 60 GB hard drive and a DVD-R drive, was $2,499, and the high-end dual-800 MHz model, with 256 MB of RAM, an 80 GB hard drive and a DVD-R drive, was $3,499
It is ABSOLUTELY incorrect to say that 64-bit compiled programs will be slower. They will use (slightly) more memory.
I see this all the time - Intel 64 bit compiled code will run 25-30% faster than 32-bit code on the same machine.
In general they are significantly faster , particularly on Intel for the following reasons.
- code compiled for 32-bit Intel is generally compiled for the lowest common denominator Intel 32-bit processor, typically a Pentium III or Pentium 4 for compatibility reasons
- code compiled for 64-bit processors is free to be optimized by the compiler for the all the new 64-bit instructions and capabilities.
A 64-bit program will use more memory, but only because
- pointers are now 64-bit. If you store a lot of pointers, then memory use will go up.
- the C/C++ "long" is now "64-bit". So data stored as "long" will use twice as much space.
But typically, programmers do not use 'long" as such, They will use int32 or int64 or some typedef so that there is no ambiguity about the size of the number. If it will fit into 32-bits , then one would use, say int32 not "long".
So in summary,
C/C++ CODE compiled for 64-bit will run faster
DATA in 64-bit programs may occupy more memory depending on the application.
That was a stretch. For more than 4 gigs per app, 64 bit should be a lot better. But I forget who it was (emig?), someone here claimed that even their own code ran a bit slower in 64 bit mode just because of the way object oriented code works, I assume this was for smaller code.
Some apps will run slower, and some will run faster. It depends on how efficient they are in 64 bit mode, if they can use the wider, deeper registers effectively, etc.
Notice I that I did say presumably? Apple could cripple the single cpu model to keep it from hurting sales of the 2 cpu beast. I don't think that they'll do that but you never know...
As to the pricing of a 2.8 Harpertown cpu, I wouldn't be surprised if that could be bought for $350 in 6 months. Intel are quite aggressive on pricing. If AMD are able to ramp up clock speeds on their Barcelona chips, Intel will drop their prices to in order to continue to hammer AMD.
It's not proper to say that Apple would "cripple" the machine. That's a pejorative term that is entirely out of place.
If Apple supplies a machine with one socket filled, then there would be no reason for them to supply an additional cooling system for the socket that had none. If they did that, then people would have to pay for it.
Considering the fact that very few will ever upgrade the machine to two processors, that would be unfairly charging the big majority. Then those people could rightly calim that Apple deliberately left it in so that they would be charged more.
What really needs to be done, is to convince a third party cpu upgrade company, such as OWC, to come out with chip upgrades. With Intel chips so easy for them to purchase, unlike the G5's, there should be no problem for them. The difference between older, upgradeable machines, and the ones in the Aluminum tower case, is that with these machines, I would never recommend that anyone without experience take it apart so as to upgrade the cpu.
If you remember the old G4's that came with one chip, the heatsink was only for one chip. no one ever could upgrade without buying an entire board, complete with CPU, cache, heatsink, and fan.
Why do you think you should have to anything less now?
It's not proper to say that Apple would "cripple" the machine. That's a pejorative term that is entirely out of place.
If Apple supplies a machine with one socket filled, then there would be no reason for them to supply an additional cooling system for the socket that had none. If they did that, then people would have to pay for it.
I think backtomac might have been thinking of the possibility there wouldn't even be a second socket soldered in. If the socket's not there, then the upgrade path would surely be crippled.
Considering the fact that very few will ever upgrade the machine to two processors, that would be unfairly charging the big majority. Then those people could rightly calim that Apple deliberately left it in so that they would be charged more.
I think it's unreasonable to expect a spare heat sink to be included, I don't think anyone else includes a heat sink for unpopulated sockets. The heat sink is included with the company's CPU upgrade kits.
So if this is true, it seems like spending an extra $800 to get the 3.0 Ghz is pretty much a waste. But the "few percent" is the same as the difference between the 3.0 Ghz and the 3.2 Ghz, so why are the numbers consistently higher for the 3.2?
Anyone got a more convincing explanation for those numbers on the 3.0? I'm trying to decide whether it's worth the extra money for the 3.0 over the 2.8...
It depends on how important the few percent is to you.
As far as the difference between the 2.8 and the 3.0 goes, apparently the majority of tests show the 3 to be faster. It's very possible that a few operation's speed is reversed in the different chips.
Something about bin speeds. Remember that these are supposed to be IDENTICAL chips. What that means, is that in theory, they should all be the SAME speed. There should be NO bin speed differences.
In practice, manufacturing isn't perfect. Ever so slight differences in line widths, depths, and spacing, add up to minute differences in capacitance. The higher the capacitance, the slower the speed.
BUT, these chips are hellishly complex. That means that on the same chip, there will be these variations. Where they are on the chip will decide how slow the chip is, and what part of the chip is affected the most.
So, one chip can have a silghtly higher bin speed overall, but still be slightly slower in some operations.
Each chip will be affected differently. Two chips with the same bin speed can have slight variations in operational "evenness". Some may have some operations a fair amount faster, and some a fair amount slower, and others may hover closely around the speed number given.
So, these tests we've read are valid for those machines. but, as we are talking about minute differences, too small to notice in most cases, they are not valid for a different set of machines because of the varibility in manufacturing.
This also holds true for the motherboard, as well as for the rest of the machine.
On average, each machine will be approximately spaced as they should be. It's possible that the 3.2 machine here was fast, but the 3.0 machine was slow, and that the 2.8 was also fast.
This is always going to happen when the speed differences between machines are so small. If we had a 2.6 a 2.9, and a 3.2, we wouldn't see this happening as much.
I think backtomac might have been thinking of the possibility there wouldn't even be a second socket soldered in. If the socket's not there, then the upgrade path would surely be crippled.
You might remember that I was saying that that could be possible as well.
But, that still isn't automatically crippling the machine. If it could be shown that it costs no more to have a complete two socket mobo, then Apple could be accused of "crippling" the machine. But, if it could be shown that it was even "slightly" cheaper to go with a single socket board, then the criticism would be unfair.
I think it's unreasonable to expect a spare heat sink to be included, I don't think anyone else includes a heat sink for unpopulated sockets. The heat sink is included with the company's CPU upgrade kits.
I agree. And that's why this is a great opportunity for a third party to come out with an upgrade. Assuming it would be cheap enough.
But, as I also said, upgrading these machines is VERY difficult. I don't recommend that anyone who isn't experienced in electronics do it themselves.