The pricepoints are fine. It's just the speed and option given at particular points that's not so hot. Expand the range of procs and features and you can beef up the lowend units without poaching the upper units in sales. The eMac should dip to $899 or less to cover edu and people needing a second computer.
GPUL can occupy the high end where people have no problem justfying a $3500 machine hopped up with go fast goodies.
The midrange is tough. Apple customers have been brainwashed into thinking that expadability is King when the fact is most tech that used to require PCI cards is integrated. Firewire, Ethernet, ATA Controller all integrated now. Apple realizes this however consumers are slow to come around.
The pricepoints are fine. It's just the speed and option given at particular points that's not so hot. Expand the range of procs and features and you can beef up the lowend units without poaching the upper units in sales. The eMac should dip to $999 or less to cover edu and people needing a second computer.
GPUL can occupy the high end where people have no problem justfying a $3500 machine hopped up with go fast goodies.
The midrange is tough. Apple customers have been brainwashed into thinking that expadability is King when the fact is most tech that used to require PCI cards is integrated. Firewire, Ethernet, ATA Controller all integrated now. Apple realizes this however consumers are slow to come around.</strong><hr></blockquote>
Big news...the Education eMac sells for a base price of $999.
Now back on Topic. I want to know why a processor can grab 8 instruction but only complete 5. Wouldn't natural processor design stipulate that a processor only grab what it can dispatch? Is there any logic which allows it to grab items it knows can "wait" and queue them? There's still much to discuss before Oct 15.
To add another dimension to this discussion of the GPUL, G5 or whatever we shall call it, who else is going to use it beside Apple and IBM? I think there is a good chance SGI might switch to it. Who else? Maybe Sun will use it in some of their equipment? IBM may have a strategy to corner the high end RISC processor market. If IBM makes the price attractive enough, I believe they can do it. The 'desktop and entry level server' chip may be just the tip of the proverbial iceberg.
What is this about the Power4 fetching 8 instruction but only exicuting 5? That doesn't make any sense to me.
I don't think Sun will use GPUL. They have Sparc. SGI may, or are they using Motorola's 7500? I also heard Amiga is developing a PowerPC machine which uses a G4, but will later be upgraded to a G5.
Guess we'll wait and see if IBM's chip is the G5, or whether Moto is shipping the G5, and GPUL is something else. GX maybe?
Now back on Topic. I want to know why a processor can grab 8 instruction but only complete 5. Wouldn't natural processor design stipulate that a processor only grab what it can dispatch? Is there any logic which allows it to grab items it knows can "wait" and queue them? There's still much to discuss before Oct 15.
I have a vague idea that I saw this discussed somewhere already in a reasonably knowledgeable fashion. If memory serves, the idea is that the ability to fetch more instructions than it can retire is essentially a strategy to avoid having too many bubbles in the pipeline. That is, if a bubble opens up, then one of queued instructions can be jammed in.
OK you engineering types: does this sound feasible?
Another idea just occurred to me. Since the POWER4 (and, we presume, the GPUL) cracks instructions to micro-ops, perhaps having a surplus of waiting instructions makes it easier to bundle groups of micro-ops through the execution units in the most efficient arrangement.
I'm not a chip engineer, so apply a sensible level of skepticism....
<strong>To add another dimension to this discussion of the GPUL, G5 or whatever we shall call it, who else is going to use it beside Apple and IBM? I think there is a good chance SGI might switch to it. Who else? Maybe Sun will use it in some of their equipment? IBM may have a strategy to corner the high end RISC processor market. If IBM makes the price attractive enough, I believe they can do it. The 'desktop and entry level server' chip may be just the tip of the proverbial iceberg.</strong><hr></blockquote>
I wouldnt be suprised if IBM were working on this. Linex is mature enough, and has enough of a market share to survive agenst Microsoft. IBM is moving further and further away from Microsoft now. Microsoft burned IBM pretty bad when they stoped codeveloping OS 2, and moved to their Windows NT strategy, directly competing with OS 2. I'm sure this is helping to fuel IBM's move to Linex, and away from the Monopoly of Microsoft. Along the same lines, IBM did have plans for a Mobo that would compete with the Wintel computers, and that was scrapped, at least in the short term. They now have a viable OS for just such a computer, as well as a number of potential customers for a powerfull new processor.
This thread is like an old, blind, deaf, diseased dog that should just be shot and put out of its misery. Start a new thread called IBM G5, or GX whatever...
<strong>This thread is like an old, blind, deaf, diseased dog that should just be shot and put out of its misery. Start a new thread called IBM G5, or GX whatever...</strong><hr></blockquote>
Then why post a response? You're preventing it from sinking into oblivion by doing so.
The pricepoints are fine. It's just the speed and option given at particular points that's not so hot. Expand the range of procs and features and you can beef up the lowend units without poaching the upper units in sales. The eMac should dip to $899 or less to cover edu and people needing a second computer.
GPUL can occupy the high end where people have no problem justfying a $3500 machine hopped up with go fast goodies.
The midrange is tough. Apple customers have been brainwashed into thinking that expadability is King when the fact is most tech that used to require PCI cards is integrated. Firewire, Ethernet, ATA Controller all integrated now. Apple realizes this however consumers are slow to come around.
The GPUL (if it ever comes to a Mac) will be in a machine likely priced at LEAST at $5000US and more likely even higher.
this aint no celeron and it isnt going to be produced at the volumes of an x86 so it will remain a high cost processor for some time. Even look at the G4 price compared to almost any x86 CPU.
That's what I was saying earlier. I think it's a possibility that GPUL systems be priced at 5k if they're a Dual Core GPUL running at 4x the speed of a G4. I'm hoping that they're not that expensive but we'll have to see.
I know that this thread has become unwieldy but a couple of pages back I pointed out that with a 300 MM wafer and .9 die shrink, (which will make it the size of a celeron, wookie), IBM intends to be price competitive with this chip. When they start pumping out the eXcaliber blade servers mentioned in this <a href="http://forums.appleinsider.com/cgi-bin/ultimatebb.cgi?ubb=get_topic&f=1&t=002495"; target="_blank">thread</a> they will also have economies of scale as well as keeping the new East Fishkill fab running at capacity.
To repeat my main point: The GPUL will be cheaper than current top-of-the-line Motorola's AMD's or Pentiums, not more expensive as this thread has generally supposed. My signature since 1999 may finally be realized.
<strong>To repeat my main point: The GPUL will be cheaper than current top-of-the-line Motorola's AMD's or Pentiums, not more expensive as this thread has generally supposed. My signature since 1999 may finally be realized.
__________________</strong><hr></blockquote>
Unfortunately the G4 is cheaper than a PIV, but a Mac still costs more than a Windows machine. It's the entire product that will probably/possibly cost more, not the chip itself.
I have a vague idea that I saw this discussed somewhere already in a reasonably knowledgeable fashion. If memory serves, the idea is that the ability to fetch more instructions than it can retire is essentially a strategy to avoid having too many bubbles in the pipeline. That is, if a bubble opens up, then one of queued instructions can be jammed in.
OK you engineering types: does this sound feasible?
Another idea just occurred to me. Since the POWER4 (and, we presume, the GPUL) cracks instructions to micro-ops, perhaps having a surplus of waiting instructions makes it easier to bundle groups of micro-ops through the execution units in the most efficient arrangement.
I'm not a chip engineer, so apply a sensible level of skepticism....</strong><hr></blockquote>
That's pretty accurate to my knowledge. Basically the POWER4 is limited to a sustained rate because it can only dispatch one group at a time, which consists of up to 5 instructions. In the event of a mispridiction though it throws out only the necessary groups, which hopefully leaves some intact reducing the bubble size.
<strong>I know that this thread has become unwieldy but a couple of pages back I pointed out that with a 300 MM wafer and .9 die shrink, (which will make it the size of a celeron, wookie), IBM intends to be price competitive with this chip. When they start pumping out the eXcaliber blade servers mentioned in this <a href="http://forums.appleinsider.com/cgi-bin/ultimatebb.cgi?ubb=get_topic&f=1&t=002495"; target="_blank">thread</a> they will also have economies of scale as well as keeping the new East Fishkill fab running at capacity.
To repeat my main point: The GPUL will be cheaper than current top-of-the-line Motorola's AMD's or Pentiums, not more expensive as this thread has generally supposed. My signature since 1999 may finally be realized.
__________________</strong><hr></blockquote>
Also, wouldn't a dual core CPU be cheaper than using two CPUs for a dual processor Mac? A single GPUL will compare to two G4s, meaning that the GPUL can cost twice as much as a G4 and end up costing the same amount to Apple for Powermacs.
All this talk of workstation Macs is off the mark. If Apple sold GPUL Macs at $6000 while their Powermac lineup remained as-is with the same old underpowered G4s, that would be a strategy for disaster. Apple needs competitive Powermacs, and no matter how good their workstations are, they NEED the market that their current Powermacs target. Actually they need Powermacs that are even cheaper, not more expensive. A workstation GPUL Mac might sell well for it's price range, but it will never make up for affordable Powermac towers that compete with Wintel towers in the consumer market.
. . . All this talk of workstation Macs is off the mark. If Apple sold GPUL Macs at $6000 while their Powermac lineup remained as-is with the same old underpowered G4s, that would be a strategy for disaster. . . .
</strong><hr></blockquote>
That would also be stupid. The GPUL will likely be in all PowerMacs, since a GPUL chip will likely be cheaper than two G4s. So all PowerMacs will have it, in my opinion. (Which means I agree with you on that point, I think.)
The workstation Mac will also come to pass, since Apple owns Shake. The folks who made Lord of the Rings run Shake on workstations. It may just be a better PowerMac, at the very top end. So, rather than 3 models, we may see 4 or even 5. Or the new workstation Mac could be a new line up, much like the Xserve is sort of a new line up. Those are unimportant details that Apple marketing will figure out.
Apple doesn't need a special processor to create professional workstations.
They might move up to higher end PowerMacs but it won't be the processor type that differentiates these platforms. Unless there is a 32 bit and a 64 bit variant and they differentiate that way. It's getting late enough in the game that I doubt they'd do that though.
Comments
The pricepoints are fine. It's just the speed and option given at particular points that's not so hot. Expand the range of procs and features and you can beef up the lowend units without poaching the upper units in sales. The eMac should dip to $899 or less to cover edu and people needing a second computer.
GPUL can occupy the high end where people have no problem justfying a $3500 machine hopped up with go fast goodies.
The midrange is tough. Apple customers have been brainwashed into thinking that expadability is King when the fact is most tech that used to require PCI cards is integrated. Firewire, Ethernet, ATA Controller all integrated now. Apple realizes this however consumers are slow to come around.
[ 09-24-2002: Message edited by: hmurchison ]</p>
<strong>Matsu-
The pricepoints are fine. It's just the speed and option given at particular points that's not so hot. Expand the range of procs and features and you can beef up the lowend units without poaching the upper units in sales. The eMac should dip to $999 or less to cover edu and people needing a second computer.
GPUL can occupy the high end where people have no problem justfying a $3500 machine hopped up with go fast goodies.
The midrange is tough. Apple customers have been brainwashed into thinking that expadability is King when the fact is most tech that used to require PCI cards is integrated. Firewire, Ethernet, ATA Controller all integrated now. Apple realizes this however consumers are slow to come around.</strong><hr></blockquote>
Big news...the Education eMac sells for a base price of $999.
<strong>
Uh I like Gay Spidy better than that one.
Now back on Topic. I want to know why a processor can grab 8 instruction but only complete 5. Wouldn't natural processor design stipulate that a processor only grab what it can dispatch? Is there any logic which allows it to grab items it knows can "wait" and queue them? There's still much to discuss before Oct 15.
[ 09-24-2002: Message edited by: hmurchison ]</p>
What is this about the Power4 fetching 8 instruction but only exicuting 5? That doesn't make any sense to me.
I don't think Sun will use GPUL. They have Sparc. SGI may, or are they using Motorola's 7500? I also heard Amiga is developing a PowerPC machine which uses a G4, but will later be upgraded to a G5.
Guess we'll wait and see if IBM's chip is the G5, or whether Moto is shipping the G5, and GPUL is something else. GX maybe?
<strong>
Now back on Topic. I want to know why a processor can grab 8 instruction but only complete 5. Wouldn't natural processor design stipulate that a processor only grab what it can dispatch? Is there any logic which allows it to grab items it knows can "wait" and queue them? There's still much to discuss before Oct 15.
[ 09-24-2002: Message edited by: hmurchison ]</strong><hr></blockquote>
I have a vague idea that I saw this discussed somewhere already in a reasonably knowledgeable fashion. If memory serves, the idea is that the ability to fetch more instructions than it can retire is essentially a strategy to avoid having too many bubbles in the pipeline. That is, if a bubble opens up, then one of queued instructions can be jammed in.
OK you engineering types: does this sound feasible?
Another idea just occurred to me. Since the POWER4 (and, we presume, the GPUL) cracks instructions to micro-ops, perhaps having a surplus of waiting instructions makes it easier to bundle groups of micro-ops through the execution units in the most efficient arrangement.
I'm not a chip engineer, so apply a sensible level of skepticism....
<strong>To add another dimension to this discussion of the GPUL, G5 or whatever we shall call it, who else is going to use it beside Apple and IBM? I think there is a good chance SGI might switch to it. Who else? Maybe Sun will use it in some of their equipment? IBM may have a strategy to corner the high end RISC processor market. If IBM makes the price attractive enough, I believe they can do it. The 'desktop and entry level server' chip may be just the tip of the proverbial iceberg.</strong><hr></blockquote>
I wouldnt be suprised if IBM were working on this. Linex is mature enough, and has enough of a market share to survive agenst Microsoft. IBM is moving further and further away from Microsoft now. Microsoft burned IBM pretty bad when they stoped codeveloping OS 2, and moved to their Windows NT strategy, directly competing with OS 2. I'm sure this is helping to fuel IBM's move to Linex, and away from the Monopoly of Microsoft. Along the same lines, IBM did have plans for a Mobo that would compete with the Wintel computers, and that was scrapped, at least in the short term. They now have a viable OS for just such a computer, as well as a number of potential customers for a powerfull new processor.
<strong>This thread is like an old, blind, deaf, diseased dog that should just be shot and put out of its misery. Start a new thread called IBM G5, or GX whatever...</strong><hr></blockquote>
Then why post a response? You're preventing it from sinking into oblivion by doing so.
<strong>Matsu-
The pricepoints are fine. It's just the speed and option given at particular points that's not so hot. Expand the range of procs and features and you can beef up the lowend units without poaching the upper units in sales. The eMac should dip to $899 or less to cover edu and people needing a second computer.
GPUL can occupy the high end where people have no problem justfying a $3500 machine hopped up with go fast goodies.
The midrange is tough. Apple customers have been brainwashed into thinking that expadability is King when the fact is most tech that used to require PCI cards is integrated. Firewire, Ethernet, ATA Controller all integrated now. Apple realizes this however consumers are slow to come around.
[ 09-24-2002: Message edited by: hmurchison ]</strong><hr></blockquote>
$3500??
The GPUL (if it ever comes to a Mac) will be in a machine likely priced at LEAST at $5000US and more likely even higher.
this aint no celeron and it isnt going to be produced at the volumes of an x86 so it will remain a high cost processor for some time. Even look at the G4 price compared to almost any x86 CPU.
That's what I was saying earlier. I think it's a possibility that GPUL systems be priced at 5k if they're a Dual Core GPUL running at 4x the speed of a G4. I'm hoping that they're not that expensive but we'll have to see.
To repeat my main point: The GPUL will be cheaper than current top-of-the-line Motorola's AMD's or Pentiums, not more expensive as this thread has generally supposed. My signature since 1999 may finally be realized.
__________________
<strong>To repeat my main point: The GPUL will be cheaper than current top-of-the-line Motorola's AMD's or Pentiums, not more expensive as this thread has generally supposed. My signature since 1999 may finally be realized.
__________________</strong><hr></blockquote>
Unfortunately the G4 is cheaper than a PIV, but a Mac still costs more than a Windows machine. It's the entire product that will probably/possibly cost more, not the chip itself.
<strong>
I have a vague idea that I saw this discussed somewhere already in a reasonably knowledgeable fashion. If memory serves, the idea is that the ability to fetch more instructions than it can retire is essentially a strategy to avoid having too many bubbles in the pipeline. That is, if a bubble opens up, then one of queued instructions can be jammed in.
OK you engineering types: does this sound feasible?
Another idea just occurred to me. Since the POWER4 (and, we presume, the GPUL) cracks instructions to micro-ops, perhaps having a surplus of waiting instructions makes it easier to bundle groups of micro-ops through the execution units in the most efficient arrangement.
I'm not a chip engineer, so apply a sensible level of skepticism....</strong><hr></blockquote>
That's pretty accurate to my knowledge. Basically the POWER4 is limited to a sustained rate because it can only dispatch one group at a time, which consists of up to 5 instructions. In the event of a mispridiction though it throws out only the necessary groups, which hopefully leaves some intact reducing the bubble size.
[ 09-25-2002: Message edited by: Telomar ]</p>
<strong>I know that this thread has become unwieldy but a couple of pages back I pointed out that with a 300 MM wafer and .9 die shrink, (which will make it the size of a celeron, wookie), IBM intends to be price competitive with this chip. When they start pumping out the eXcaliber blade servers mentioned in this <a href="http://forums.appleinsider.com/cgi-bin/ultimatebb.cgi?ubb=get_topic&f=1&t=002495"; target="_blank">thread</a> they will also have economies of scale as well as keeping the new East Fishkill fab running at capacity.
To repeat my main point: The GPUL will be cheaper than current top-of-the-line Motorola's AMD's or Pentiums, not more expensive as this thread has generally supposed. My signature since 1999 may finally be realized.
__________________</strong><hr></blockquote>
Also, wouldn't a dual core CPU be cheaper than using two CPUs for a dual processor Mac? A single GPUL will compare to two G4s, meaning that the GPUL can cost twice as much as a G4 and end up costing the same amount to Apple for Powermacs.
All this talk of workstation Macs is off the mark. If Apple sold GPUL Macs at $6000 while their Powermac lineup remained as-is with the same old underpowered G4s, that would be a strategy for disaster. Apple needs competitive Powermacs, and no matter how good their workstations are, they NEED the market that their current Powermacs target. Actually they need Powermacs that are even cheaper, not more expensive. A workstation GPUL Mac might sell well for it's price range, but it will never make up for affordable Powermac towers that compete with Wintel towers in the consumer market.
If they had lowered the price, as many have suggested they should, when the GP-UL comes, the low price supporters would say "Here we go again!"
<strong>
. . . All this talk of workstation Macs is off the mark. If Apple sold GPUL Macs at $6000 while their Powermac lineup remained as-is with the same old underpowered G4s, that would be a strategy for disaster. . . .
</strong><hr></blockquote>
That would also be stupid. The GPUL will likely be in all PowerMacs, since a GPUL chip will likely be cheaper than two G4s. So all PowerMacs will have it, in my opinion. (Which means I agree with you on that point, I think.)
The workstation Mac will also come to pass, since Apple owns Shake. The folks who made Lord of the Rings run Shake on workstations. It may just be a better PowerMac, at the very top end. So, rather than 3 models, we may see 4 or even 5. Or the new workstation Mac could be a new line up, much like the Xserve is sort of a new line up. Those are unimportant details that Apple marketing will figure out.
They might move up to higher end PowerMacs but it won't be the processor type that differentiates these platforms. Unless there is a 32 bit and a 64 bit variant and they differentiate that way. It's getting late enough in the game that I doubt they'd do that though.
[ 09-26-2002: Message edited by: Telomar ]</p>