<strong>The FrontSide Bus on the new dualies, can it be changed ? I mean, can it be upgraded by some third parties add-on in the future ?</strong><hr></blockquote>
You can't upgrade motherboards the way you upgrade processors. Open your computer when you get the chance...see that big [green board] with all the little do-hickies soldered onto it (including the RAM slots and all sorts of other stuff)? You're asking to replace that (or a major part of it anyway). It's not possible.
It would be analogous to a surgeon trying to splice together brain parts from two different people. Doesn't happen (except in Egypt under rare circumstances).
If it can't be changed by some third parties add-on, upgrade kit or whatever, then why is Apple putting this damn DDR bottleneck ? Is this just a marketing trick ? The DDR appears to be useless on those machines, I mean it gives no advantages at all over the old SDRAM. There is no speed gain at all on the dual 1 GHz, none ! Nada ! Rien ! Not a single advantage !
I saw some benchmarks today about disk access and video performances, and there is no speed gain whatsoever on the new 1 GHz. I'm pissed off by this marketing racket !
I would prefer to have good old PC133 SDRAM, in place of a crippled Mac just because of some shitty marketing trick ! With DDR, I feel like cheated. Apple, I'm furious at you because of this !
Well, to me its kind of simple. Apple can keep shipping PC133 based systems (since motorola who designs and manufactures the G4 cannot get their teams to design a DDR G4) or at least get some performance benefit by building a system that provides DDR to the rest of the system (like IO - usually the slowest part of a computer) while keeping the 133/167 bus for the CPU.
At the very least, you DO get a performance boost with anything that uses IO a lot - like video, databases, etc...
So it is worthwhile. Would we have preferred full DDR everywhere? Sure - but Motorola or IBM have to give us a G4 or better part that can support it first.
<strong>You can't upgrade motherboards the way you upgrade processors. Open your computer when you get the chance...see that big [green board] with all the little do-hickies soldered onto it (including the RAM slots and all sorts of other stuff)? You're asking to replace that (or a major part of it anyway). It's not possible.
</strong><hr></blockquote>
It's quite possible and common with PCs, and if Apple chose they could make it possible for Macs as well.
First, Apple can do nothing about the system bus speed. That's Motorola's department, and they haven't come out with a G4 that supports anything faster than 167Mhz MPX.
Second, while itÕs not possible to "upgrade" the system bus without replacing the motherboard, it is possible to overclock the system bus. This usually involves tampering with a few registers to make the FSB go faster, but it shouldn't be too hard. It's been done before on previous generation machines.
So you can make the system bus go faster, but it isn't an upgradeable component. Never has been. Not on a Mac. Not on a PC. Moving from MPX to RapidIO or Hyper Transport would require a new motherboard and a new CPU.
Also, the FSB canÕt run much higher than its rated speed as it and your CPU will begin to overheat. Remember increasing the FSB frequency also increases your CPUÕs frequency. Your CPUÕs speed is just a multiple of whatever the FSB is set to. The 1GHz CPUs in AppleÕs latest systems for instance, have a multiplier of 6 and because they run on a 167MHz bus their speed is 6 X 167MHz or 1000MHz. Increasing the FSB frequency to 175MHz, would result in higher clocked CPUs, 1050MHz or 6 X 175MHz. Like the FSB, the multiplier can also be changed to allow you to run the FSB higher without overheating the CPU, or visa versa. However changing the CPUÕs multiplier is VERY difficult if not impossible on a Macintosh.
then why is Apple putting this damn DDR bottleneck <hr></blockquote>
The bottleneck is not caused by the DDR RAM. It was there before DDR ever got here. Apple probably added DDR for one simple reason - people have been screaming for it for months. They did the best they could with the processor specs / bus speeds Motorola gives them.
However, if I were a betting man (and actually had money) I'd wager that we're almost at the end of the proverbial tunnel. IBM looks well-positioned to help Apple out in the next year with some new proecessors / system architectures. No guarantees obviously, but things look bright (meaning we're not looking at Motorola's ass-end for a change).
[quote]It's quite possible and common with PCs, and if Apple chose they could make it possible for Macs as well.<hr></blockquote>
Um, not sure what you mean exactly here. i know you don't mean actually replacing the FSB with a new one, 'cause i've never seen a computer that can do that and i've built dozens of boxes. i could only assume you mean system BIOS'es that allow you to overclock the FSB.
not sure that's the best way to hanlde it. for personal machines i don't mind overclocking, it's my problem when it becomes unstable. i don't see Apple offering the ability to OC their MB's though, as system stability is a key component of the OSX advantage. why shoot yourself in the foot by having a bunch of greenhorns overclocking their machines then getting pissed off when it doesn't work?
OC'ing is a hobby and should be treated as such, not a business solution.
A) The link between the CPU and the Integrated Controller (northbrige in the PC world). The Back Side Bus is the link between the CPU and the cache. Most front side buses are 64-bits (8-bytes) wide. The wider and higher effective MHz the bus, the faster it runs. Now, The front side bus consists of the interface on the CPU, the interface on the IC (which are identical), and the tracers on the motherboard.
Q) Is the FSB upgradable?
A) No, it is part of the motherboard.
Q) Can you overclock the FSB?
A) Yes. The original G4s could run at 100MHz or 133MHz (there may have been settings in between). By changing a set of resisitors on the motherboard of a 100MHz G4, you could run the Front Side Bus at 133MHz. Many PCs allow you to set the FSB to the MHz from the BIOS.
Q) Does overclocking the FSB reduce stability?
A) Yes.
Q) What else has to be done if you overclock the FSB?
A) Because the CPU clock is set as a multiple of the FSB clock, you need to reduce the CPU clock to a stable level. Also, OpenFirmware needs to be modified to give accurate system clock readings.
Q) Does the new G4s support an overclocked FSB?
A) There has been speculation that Apple uses Discovery ICs, but modified to include AGP and FireWire. If this is the case, the new FSB is probably overclockable up to 183MHz, possibly beyond.
Q) Why did Apple include DDR with the bottleneck in the FSB?
A) Because computers are moving away from the old CPU-centric approace (DeCPU, anyone?). Gigabit Ethernet, FireWire (probably), hard drives, optical drives, PCI cards and the graphics card can all access the memory directly. So if multiple devices are accessing the memory at once (eg CPU + graphics card for games, Fiber Channel PCI card + Ethernet for servers), then you will see an improvement over the old systems.
It's quite possible and common with PCs, and if Apple chose they could make it possible for Macs as well.</strong><hr></blockquote>
How is that common with PCs? You might be able to make minor adjustments to the FSB frequency (i.e. switch between 100 and 133MHz), but I don't think any PC motherboard has ever supported switching between SDR and DDR or QDR FSB modes, or a significant frequency boost.
Doesn't the current motherboard know whether it's connected to the dual 867 or dual 1000/1250, and adjust accordingly? What's to say they didn't make it to where it can adjust higher, or go DDR when a certian daughter card is inserted?
<strong>If it can't be changed by some third parties add-on, upgrade kit or whatever, then why is Apple putting this damn DDR bottleneck ? Is this just a marketing trick ? The DDR appears to be useless on those machines, I mean it gives no advantages at all over the old SDRAM. There is no speed gain at all on the dual 1 GHz, none ! Nada ! Rien ! Not a single advantage !
I saw some benchmarks today about disk access and video performances, and there is no speed gain whatsoever on the new 1 GHz. I'm pissed off by this marketing racket !
I would prefer to have good old PC133 SDRAM, in place of a crippled Mac just because of some shitty marketing trick ! With DDR, I feel like cheated. Apple, I'm furious at you because of this !
</strong><hr></blockquote>
Kali! Man, you need to settle down. The new machines are not a hack. They are not a stopgap. They are not crippled. You have looked at comparisons between last months top of the line $3000 machines with 2MB L3 cache (which is as important as FSB speed), with this months middle spec'ed $2400 machine with 1 MB L3 cache. Think of it this way Kali: The FSB separates the CPU from the RAM. As the CPU processes data, it needs to call to the RAM to pick up more data. The L3 cache is like really a small pool of RAM on the CPU side of the FSB. The larger that pool is, the less often the CPU will need to send a request for data across the FSB, and the faster the system can process data.
Now we will have to wait a month to be sure, but I'm willing to bet that the 1.25 top end machine (with 2MB L3 cache! and a 167 system bus) will be more than 25% faster than last months 1 GHz dualie at many bus bound tasks!
Furthermore, you need to think of the machine as an significant evolution. The DDR is there now because Apple phases these technologies into it's products in steps (see the Yikes story as an example; yikes! was not the end of the line for the G3, it was the beginning of the line for the G4). To my mind, the presence of DDR means that Apple has committed to a path which allows a double pumped FSB. That path is not ready yet, but the DDR RAM is in place and ready when it is.
And finally, it is not wasted in these machines even now. As pointed out above, the rest of the system calls (not CPU related) to RAM will not be constrained by the weak SDR back-side bus. You have seen tests which test this? No you haven't. You have seen a few not very well thought out benchmarks at barefeats. Nothing against Barefeats, but they are trying to test a property of the system that is very hard to test: how that system performs under high load. Traditionally, this is more a property of a good server, rather than a good desktop (and basefeats has not concerned itself with the server market). But peripherals and their use is changing such that it is going to be a desirable property of a desktop machine (for example, AGP needs now--and will certainly need more in the near future--a lot of bandwidth, especially now that compositing the screen is offloaded to the AGP bus, rather than the CPU). Couple that with DMA requests, disk access etc., and you have a system that is more responsive than you have ever had before.
In conclusion, it is my opinion that this is one of the most significant upgrades that we have seen from Apple in the powermacs in years. It is a harbinger of things to come, yes. But it also stands on its own as a solid upgrade.
Just to offer a concrete example of what wormboy is talking about: You know how Quartz Extreme offloads some of the work that Quartz does to the GPU? Well, all that work has to go across the AGP bus to get to the GPU. Now, consider this: In the Quicksilver PowerMacs, the RAM offers about 1GB/s throughput. Both the front side bus and the AGP bus can handle 1GB/s throughput (theoretical - but in practice they're not too far from that), so under a heavy load, the memory controller will end up doling each one about half of what it's capable of carrying. QE will still offer improved performance, but obviously the system is not being used to its utmost.
Now, look at the new PowerMacs. RAM offers 2.1GB/s or 2.7GB/s (theoretical) throughput. The front side bus and the AGP bus both still want about 1GB/s. But now, under heavy load (and, admittedly, under a fairly ideal case), the memory controller can saturate both busses and still have bandwidth left over for hard drive accesses. The ideal case will be much more common for the AGP bus than for the front side bus, because AGP tends to suck huge blocks of contiguous memory out of RAM (especially with QE active), and that happens to be the case when DDR RAM really is twice as fast as SDR RAM. So, even with QE (and maybe a video game) feeding the GPU as much as it can take, the memory controller is still able to feed the CPU as well as the old model could, and if a hard drive wants to read something out of RAM to store it, it can do that too without competing with either the CPU or the GPU. This is actually a nicely balanced system.
1) I've been waiting for someone to come out with a benchmark that captures this somehow. The problem is, what would that test do exactly? Something similar to what barefeats tried on their second attempt? But perhaps a little more challenging? Maybe an Applescript that processes a photoshop file, and also moves two semitransparant terminal windows over a series of complex text filled windows (to stress the AGP/QE bus) while also running data between RAM and disk? The measure would be how long the photoshop script takes, and the idea being that the time needs to be limited by bus access to RAM rather than CPU speed. (Although anyone who does such a benchmark will be accused of platform boosterism, since no such benchmark yet exists).
and 2) Does it matter if under real-world conditions these advantages are never expressed? If you need exotic conditions to observe this advantage, is it an advantage? Not to say that they may not be needed or used under future configurations, just that they may not be appreciated today. I think the answer here is yes, the advantage is real and will be expressed in future systems, with future software demands.
1) I've been waiting for someone to come out with a benchmark that captures this somehow.</strong><hr></blockquote>
Benchmarks to measure overall system performance are kind of tricky, but I'm sure someone could come up with something.
[quote]<strong>and 2) Does it matter if under real-world conditions these advantages are never expressed? If you need exotic conditions to observe this advantage, is it an advantage?</strong><hr></blockquote>
It depends on what the "exotic" conditions are, and when they occur. If they occur under truly random conditions (such as the Photoshop tests at DigitalVideoEditing.com, where they chose some really funky numbers to plug into PShop) then it's not worth much. If they show up when they're really needed - when someone is working in FCP and Photoshop, while capturing video to their HD, say, then it's worth it.
Basically, the new PowerMacs are a step in the direction of a workstation architecture: Bandwidth is everything, and the CPU is not that central. The advantages of this design will show up when the machines are being used as workstations, not so much when you're just running IE.
[quote]<strong>Not to say that they may not be needed or used under future configurations, just that they may not be appreciated today. I think the answer here is yes, the advantage is real and will be expressed in future systems, with future software demands.</strong><hr></blockquote>
I think it's fair to say that the change is needed in the current systems with current software demands. However, much of that software is outside the consumer space for now. So I agree with the implication that this system will actually get better over time.
Of course, if the rumors are right about this forthcoming IBM chip, then all of a sudden it'll be the RAM hustling to keep up with the bus to the CPU, and we'll be back to square one.
[quote] 1. The FSB clock looks software controlled. <hr></blockquote>
Now, I know squat about chip or board design, but doesn't this suggest the possibility of altering the FSB using some sort of software utility (perhaps to be released later, along with third party CPU upgrades)?
Comments
<strong>The FrontSide Bus on the new dualies, can it be changed ? I mean, can it be upgraded by some third parties add-on in the future ?</strong><hr></blockquote>
Nope.
But I know it's the bottleneck on these machines. So that's why I'm asking the question !
So is it upgradable or not ? Is it stocked on the motherboard or on the chip ?
[ 08-20-2002: Message edited by: Kali ]</p>
It would be analogous to a surgeon trying to splice together brain parts from two different people. Doesn't happen (except in Egypt under rare circumstances).
[ 08-20-2002: Message edited by: Moogs ]</p>
I saw some benchmarks today about disk access and video performances, and there is no speed gain whatsoever on the new 1 GHz. I'm pissed off by this marketing racket !
I would prefer to have good old PC133 SDRAM, in place of a crippled Mac just because of some shitty marketing trick ! With DDR, I feel like cheated. Apple, I'm furious at you because of this !
At the very least, you DO get a performance boost with anything that uses IO a lot - like video, databases, etc...
So it is worthwhile. Would we have preferred full DDR everywhere? Sure - but Motorola or IBM have to give us a G4 or better part that can support it first.
<strong>You can't upgrade motherboards the way you upgrade processors. Open your computer when you get the chance...see that big [green board] with all the little do-hickies soldered onto it (including the RAM slots and all sorts of other stuff)? You're asking to replace that (or a major part of it anyway). It's not possible.
</strong><hr></blockquote>
It's quite possible and common with PCs, and if Apple chose they could make it possible for Macs as well.
Second, while itÕs not possible to "upgrade" the system bus without replacing the motherboard, it is possible to overclock the system bus. This usually involves tampering with a few registers to make the FSB go faster, but it shouldn't be too hard. It's been done before on previous generation machines.
So you can make the system bus go faster, but it isn't an upgradeable component. Never has been. Not on a Mac. Not on a PC. Moving from MPX to RapidIO or Hyper Transport would require a new motherboard and a new CPU.
Also, the FSB canÕt run much higher than its rated speed as it and your CPU will begin to overheat. Remember increasing the FSB frequency also increases your CPUÕs frequency. Your CPUÕs speed is just a multiple of whatever the FSB is set to. The 1GHz CPUs in AppleÕs latest systems for instance, have a multiplier of 6 and because they run on a 167MHz bus their speed is 6 X 167MHz or 1000MHz. Increasing the FSB frequency to 175MHz, would result in higher clocked CPUs, 1050MHz or 6 X 175MHz. Like the FSB, the multiplier can also be changed to allow you to run the FSB higher without overheating the CPU, or visa versa. However changing the CPUÕs multiplier is VERY difficult if not impossible on a Macintosh.
then why is Apple putting this damn DDR bottleneck <hr></blockquote>
The bottleneck is not caused by the DDR RAM. It was there before DDR ever got here. Apple probably added DDR for one simple reason - people have been screaming for it for months. They did the best they could with the processor specs / bus speeds Motorola gives them.
However, if I were a betting man (and actually had money) I'd wager that we're almost at the end of the proverbial tunnel. IBM looks well-positioned to help Apple out in the next year with some new proecessors / system architectures. No guarantees obviously, but things look bright (meaning we're not looking at Motorola's ass-end for a change).
[ 08-24-2002: Message edited by: Moogs ]</p>
Um, not sure what you mean exactly here. i know you don't mean actually replacing the FSB with a new one, 'cause i've never seen a computer that can do that and i've built dozens of boxes. i could only assume you mean system BIOS'es that allow you to overclock the FSB.
not sure that's the best way to hanlde it. for personal machines i don't mind overclocking, it's my problem when it becomes unstable. i don't see Apple offering the ability to OC their MB's though, as system stability is a key component of the OSX advantage. why shoot yourself in the foot by having a bunch of greenhorns overclocking their machines then getting pissed off when it doesn't work?
OC'ing is a hobby and should be treated as such, not a business solution.
<strong>Um, not sure what you mean exactly here. i know you don't mean actually replacing the FSB with a new one</strong><hr></blockquote>
nah, he's talking about the motherboard (FSB included).
A) The link between the CPU and the Integrated Controller (northbrige in the PC world). The Back Side Bus is the link between the CPU and the cache. Most front side buses are 64-bits (8-bytes) wide. The wider and higher effective MHz the bus, the faster it runs. Now, The front side bus consists of the interface on the CPU, the interface on the IC (which are identical), and the tracers on the motherboard.
Q) Is the FSB upgradable?
A) No, it is part of the motherboard.
Q) Can you overclock the FSB?
A) Yes. The original G4s could run at 100MHz or 133MHz (there may have been settings in between). By changing a set of resisitors on the motherboard of a 100MHz G4, you could run the Front Side Bus at 133MHz. Many PCs allow you to set the FSB to the MHz from the BIOS.
Q) Does overclocking the FSB reduce stability?
A) Yes.
Q) What else has to be done if you overclock the FSB?
A) Because the CPU clock is set as a multiple of the FSB clock, you need to reduce the CPU clock to a stable level. Also, OpenFirmware needs to be modified to give accurate system clock readings.
Q) Does the new G4s support an overclocked FSB?
A) There has been speculation that Apple uses Discovery ICs, but modified to include AGP and FireWire. If this is the case, the new FSB is probably overclockable up to 183MHz, possibly beyond.
Q) Why did Apple include DDR with the bottleneck in the FSB?
A) Because computers are moving away from the old CPU-centric approace (DeCPU, anyone?). Gigabit Ethernet, FireWire (probably), hard drives, optical drives, PCI cards and the graphics card can all access the memory directly. So if multiple devices are accessing the memory at once (eg CPU + graphics card for games, Fiber Channel PCI card + Ethernet for servers), then you will see an improvement over the old systems.
Barto
<strong>
It's quite possible and common with PCs, and if Apple chose they could make it possible for Macs as well.</strong><hr></blockquote>
How is that common with PCs? You might be able to make minor adjustments to the FSB frequency (i.e. switch between 100 and 133MHz), but I don't think any PC motherboard has ever supported switching between SDR and DDR or QDR FSB modes, or a significant frequency boost.
Bye,
RazzFazz
There is a PowerPC motherboard which can switch between 370, MPX and 60x buses.
Barto
<strong>If it can't be changed by some third parties add-on, upgrade kit or whatever, then why is Apple putting this damn DDR bottleneck ? Is this just a marketing trick ? The DDR appears to be useless on those machines, I mean it gives no advantages at all over the old SDRAM. There is no speed gain at all on the dual 1 GHz, none ! Nada ! Rien ! Not a single advantage !
I saw some benchmarks today about disk access and video performances, and there is no speed gain whatsoever on the new 1 GHz. I'm pissed off by this marketing racket !
I would prefer to have good old PC133 SDRAM, in place of a crippled Mac just because of some shitty marketing trick ! With DDR, I feel like cheated. Apple, I'm furious at you because of this !
Kali! Man, you need to settle down. The new machines are not a hack. They are not a stopgap. They are not crippled. You have looked at comparisons between last months top of the line $3000 machines with 2MB L3 cache (which is as important as FSB speed), with this months middle spec'ed $2400 machine with 1 MB L3 cache. Think of it this way Kali: The FSB separates the CPU from the RAM. As the CPU processes data, it needs to call to the RAM to pick up more data. The L3 cache is like really a small pool of RAM on the CPU side of the FSB. The larger that pool is, the less often the CPU will need to send a request for data across the FSB, and the faster the system can process data.
Now we will have to wait a month to be sure, but I'm willing to bet that the 1.25 top end machine (with 2MB L3 cache! and a 167 system bus) will be more than 25% faster than last months 1 GHz dualie at many bus bound tasks!
Furthermore, you need to think of the machine as an significant evolution. The DDR is there now because Apple phases these technologies into it's products in steps (see the Yikes story as an example; yikes! was not the end of the line for the G3, it was the beginning of the line for the G4). To my mind, the presence of DDR means that Apple has committed to a path which allows a double pumped FSB. That path is not ready yet, but the DDR RAM is in place and ready when it is.
And finally, it is not wasted in these machines even now. As pointed out above, the rest of the system calls (not CPU related) to RAM will not be constrained by the weak SDR back-side bus. You have seen tests which test this? No you haven't. You have seen a few not very well thought out benchmarks at barefeats. Nothing against Barefeats, but they are trying to test a property of the system that is very hard to test: how that system performs under high load. Traditionally, this is more a property of a good server, rather than a good desktop (and basefeats has not concerned itself with the server market). But peripherals and their use is changing such that it is going to be a desirable property of a desktop machine (for example, AGP needs now--and will certainly need more in the near future--a lot of bandwidth, especially now that compositing the screen is offloaded to the AGP bus, rather than the CPU). Couple that with DMA requests, disk access etc., and you have a system that is more responsive than you have ever had before.
In conclusion, it is my opinion that this is one of the most significant upgrades that we have seen from Apple in the powermacs in years. It is a harbinger of things to come, yes. But it also stands on its own as a solid upgrade.
[ 08-25-2002: Message edited by: wormboy ]</p>
Now, look at the new PowerMacs. RAM offers 2.1GB/s or 2.7GB/s (theoretical) throughput. The front side bus and the AGP bus both still want about 1GB/s. But now, under heavy load (and, admittedly, under a fairly ideal case), the memory controller can saturate both busses and still have bandwidth left over for hard drive accesses. The ideal case will be much more common for the AGP bus than for the front side bus, because AGP tends to suck huge blocks of contiguous memory out of RAM (especially with QE active), and that happens to be the case when DDR RAM really is twice as fast as SDR RAM. So, even with QE (and maybe a video game) feeding the GPU as much as it can take, the memory controller is still able to feed the CPU as well as the old model could, and if a hard drive wants to read something out of RAM to store it, it can do that too without competing with either the CPU or the GPU. This is actually a nicely balanced system.
[ 08-25-2002: Message edited by: Amorph ]</p>
1) I've been waiting for someone to come out with a benchmark that captures this somehow. The problem is, what would that test do exactly? Something similar to what barefeats tried on their second attempt? But perhaps a little more challenging? Maybe an Applescript that processes a photoshop file, and also moves two semitransparant terminal windows over a series of complex text filled windows (to stress the AGP/QE bus) while also running data between RAM and disk? The measure would be how long the photoshop script takes, and the idea being that the time needs to be limited by bus access to RAM rather than CPU speed. (Although anyone who does such a benchmark will be accused of platform boosterism, since no such benchmark yet exists).
and 2) Does it matter if under real-world conditions these advantages are never expressed? If you need exotic conditions to observe this advantage, is it an advantage? Not to say that they may not be needed or used under future configurations, just that they may not be appreciated today. I think the answer here is yes, the advantage is real and will be expressed in future systems, with future software demands.
[ 08-25-2002: Message edited by: wormboy ]</p>
<strong>Agreed. Two points though...
1) I've been waiting for someone to come out with a benchmark that captures this somehow.</strong><hr></blockquote>
Benchmarks to measure overall system performance are kind of tricky, but I'm sure someone could come up with something.
[quote]<strong>and 2) Does it matter if under real-world conditions these advantages are never expressed? If you need exotic conditions to observe this advantage, is it an advantage?</strong><hr></blockquote>
It depends on what the "exotic" conditions are, and when they occur. If they occur under truly random conditions (such as the Photoshop tests at DigitalVideoEditing.com, where they chose some really funky numbers to plug into PShop) then it's not worth much. If they show up when they're really needed - when someone is working in FCP and Photoshop, while capturing video to their HD, say, then it's worth it.
Basically, the new PowerMacs are a step in the direction of a workstation architecture: Bandwidth is everything, and the CPU is not that central. The advantages of this design will show up when the machines are being used as workstations, not so much when you're just running IE.
[quote]<strong>Not to say that they may not be needed or used under future configurations, just that they may not be appreciated today. I think the answer here is yes, the advantage is real and will be expressed in future systems, with future software demands.</strong><hr></blockquote>
I think it's fair to say that the change is needed in the current systems with current software demands. However, much of that software is outside the consumer space for now. So I agree with the implication that this system will actually get better over time.
Of course, if the rumors are right about this forthcoming IBM chip, then all of a sudden it'll be the RAM hustling to keep up with the bus to the CPU, and we'll be back to square one.
[quote] 1. The FSB clock looks software controlled. <hr></blockquote>
Now, I know squat about chip or board design, but doesn't this suggest the possibility of altering the FSB using some sort of software utility (perhaps to be released later, along with third party CPU upgrades)?