Here is how to make a low costs feature rich mini.
1. A high integration 970, that is a chip with an on board memory controller.
2. A new high integration bus interface chip.
3. Move to notebook form factor SATA drives.
4. No sound hardware, rely instead on external USB speakers.
It doesn't have to be this complicated for cheap G5 machines. Apple should use the same exact ASIC chips in the PowerMac G5.
The system ASIC supports dual CPUs, 128 bit dual channel DDR memory, AGP 8x Pro, and 16 bit Hypertransport. In a low-cost system, they only use 1 CPU, a slower CPU bus, single channel DDR memory, AGP 4x. These changes should be nothing but resistor settings to the system ASIC. If they are not they should be.
For the I/O ASIC, it supports too many things to list, but it's the same concept as the system ASIC. For hard drives, use the ATA/100 bus that is already on there. Use a cheaper Firewire PHY layer that only supports Firewire 400, use a cheaper audio device, use the same PCI-based USB 480 device, etc.
Lastly, don't use the PCI/PCI-X Hypertransport tunnel, and just directly connect the system ASIC and I/O ASIC with the Hypertransport buses already on there. If they are not compatible HT buses, they should be made to be. The rest is just using cheaper graphics card, RAM, etc.
Quote:
Look at it this way if a whole computer can be stuffed into a laptop it should be a piece of cake to do a mini desk top. I;m not talking the current PC route whcih has normal motherboards stuffed into mini or mivro enclosures, we are talking a motherboard purposefully designed for a minmalist machine.
Remember my optical-drive-in-the-keyboard idea? Well... how about we go really really old-school.
First off minis are hot in the market place right now. Second the new G5 is much larger than many people are willing ot deal with. As to the dual processor iMacs, it will happen sooner or later SMP is actually a cheaper way to improve performance. I wuold not be surprised if the SMP iMacs are a dual core solution.
Child, there's no logic in assuming people who don't want the huge towers will want small-form factor PCs. Most people will want, *GASP*, mini-tower sized PCs. It's what they are used to.
Well, based on Jobs statement (3 GHz in one year), a 2.5 GHz G5 in January is almost certain. For 2.8 GHz, I don't know, we will see... After June's surprise (2 GHz instead of 1.8 GHz), a January surprise would be equally welcome.
Something very important to place emphasis on is this:
Jobs wouldn't have told us about the speed bump if it wasn't certain!
This is the same guy who had to eat crow over the 500MHz G4 Moto debacle. There is NO way that he is going to promise a speed gain that won't be delivered. So I think that we can assume that your timeline is correct because I profoundly doubt that Apple is going to jump from 2GHz to 3GHz in one single jump. We should see a speed boost announced in January (hopefully along with a second dual CPU machine so that there are 2 dualies on the lineup).
Even though the 7457 has more transistors than the 970, it's die size (98 sq mm) is actually smaller than the 970 (118 sq mm). IBM's floorplan for the 970 seems atrocious considering .....
True, but having more transistors and having them more densly packed has to affect production costs. If I remember correctly, it has been mentioned that IBM utilizes some automated design tools to speed up the actual design. This might explain the less densly packed transistors??
Quote:
If the 970 has much better yield than the 7457, than the 970 may be cheaper.....
No comment here, I've not a clue, well maybe, but no point in appearing to bash Motorola. More densly packed transistors/traces may affect yields??
Quote:
Apple sells an eMac with a 1 GHz 7455 CPU, combo optical, 128 MB RAM and 40 GB disk for 799$....
Kind of my point. The MPC7455 is still on the 0.18µm process and unless Motorola is having a fire sale or their yields are tremedous, the IBM 970 should be price competative with the MPC7455/7445 and the MPC7457/7447.
I just can't see why using the G4's, in and of itself, would allow Apple to produce computers cheaper than using the G5. Unless of course all the other stuff, companion chip, Hypertransport, etc. disproportionately increases the cost over the G4's equivalents.
Something very important to place emphasis on is this:
Jobs wouldn't have told us about the speed bump if it wasn't certain!
This is the same guy who had to eat crow over the 500MHz G4 Moto debacle. There is NO way that he is going to promise a speed gain that won't be delivered. So I think that we can assume that your timeline is correct because I profoundly doubt that Apple is going to jump from 2GHz to 3GHz in one single jump. We should see a speed boost announced in January (hopefully along with a second dual CPU machine so that there are 2 dualies on the lineup).
I guess you are right. Anyway i still wait for my dualie. I expect it will ship before the rev B G5 towers
Even though the 7457 has more transistors than the 970, it's die size (98 sq mm) is actually smaller than the 970 (118 sq mm). IBM's floorplan for the 970 seems atrocious considering how tightly they've packed the 750fx, 35 sq mm for 39M transistors. Since the 7457 is 20% smaller than the 970, and given the same yields, the 7457 should be cheaper.
If the 970 has much better yield than the 7457, than the 970 may be cheaper. Moto seems to have had a lot of yield problems with their 130 nm process, so the 970 may very well have been cheaper for a few months. Who knows.
The 970 is made on larger wafers which tends to improve the cost/die.
I just can't see why using the G4's, in and of itself, would allow Apple to produce computers cheaper than using the G5. Unless of course all the other stuff, companion chip, Hypertransport, etc. disproportionately increases the cost over the G4's equivalents.
If indeed IBM can sell a 1.3 GHz 970 cheaper than Moto can sell a 1.3 GHz 74x7 and associate motherboard costs are about equal, then yeah, Apple would be stupid not to migrate all of its systems to the 970. The real unknown is how much Apple actually pays for each CPU. That's been the question for a long time now.
Also, I think G4 parts are probably a good percentage cheaper than G5 parts right now because they've been in production for such a long time already and Apple can get them cheaper and cheaper as time goes by.
Eugene, is child a term of endearment for you? Maybe I'm out of the infinite loop, but surely I can't be the only one wondering what that's all about. . .
I found it incredibly interesting that the G4 has more transistors than the G5. Doesn't that defy all forms of common sense logic? More complex, powerful chips have more transistors, right?
Anyway, even though all of us would love to see a midrange tower, Apple doesn't want to go that route. It has the eMac covering the low-end, the iMac the <chuckle> "midrange" and the Power Macs at the high end. They're not going to engineer a new case and motherboard for something that will have limited appeal. If the Cube had succeeded, that would be one thing. Since it didn't I doubt we'll see anything similar from Apple for many moons.
Eugene, is child a term of endearment for you? Maybe I'm out of the infinite loop, but surely I can't be the only one wondering what that's all about. . .
I found it incredibly interesting that the G4 has more transistors than the G5. Doesn't that defy all forms of common sense logic? More complex, powerful chips have more transistors, right?
Up to a point. The vast majority of the transistors are in the caches and associated tags.
The G4 has a L3 controller and tags on-board, which probably explains most of the extra transistors. The L3 tags will use a lot of transistors, and very densely, also explaining the higher average density for the G4. The G4 may also use more redundancy in the caches, giving higher yields, increasing transistor count, and increasing average density.
The G5 has more logic transistors, which are less dense than memory transistors.
Eugene, is child a term of endearment for you? Maybe I'm out of the infinite loop, but surely I can't be the only one wondering what that's all about. . .
Eugene, is child a term of endearment for you? Maybe I'm out of the infinite loop, but surely I can't be the only one wondering what that's all about. . .
look child, it's a bit of an inside joke. don't worry about it. it'll get old real quick
What is it with people who make unrealistic predictions?
We're making predictions?
Quote:
Not to single anyone out in particular (I've seen at least a few others), but posts like THT's make me scratch my head in wonderment. Dual-processor iMacs? G4 and G5 "Minis?" I don't think so. While I can't rule it out, I'm even skeptical of dual-1.8 GHz G5s at the low end of the PowerMac lineup.
As I said to Eugene. I'm not making a prediction and I'm sure many others aren't either. I'm putting down what Apple should sell, and I'm usually attentive to whether such hardware can be engineered.
Quote:
1. It (usually) costs considerably more to have dual CPUs than a single CPU at a higher clock speed.
From the Apple store:
PowerMac G4
$1,299.00
1.25GHz PowerPC G4
1MB L3 cache
$1,599.00
Dual 1.25GHz PowerPC G4
2MB L3 cache/processor
Here Apple added another CPU and 3 MB of cache for 300$. In my dual G4 iMac proposal, it would be without L3 cache. Even so, the dual 1.25 GHz PowerMac G4 only costs 1599$. For 200$ more at $1800, Apple sells an iMac 17 with 1 1.25 GHz processor without cache, a SuperDrive, a 17" LCD, and a whole lot less expansion.
So take the dual 1.25 GHz PowerMac G4, take out all of the L3 cache, add a SuperDrive and an LCD screen, and I think Apple can charge 1800$ for it and get the same margins.
Quote:
2. The iMac has a small enclosure, and probably can't fit two CPUs (at least, not without addressing heat first).
You didn't see my "Firewire based keyboard with an optical drive" idea from a previous thread. By moving the optical drive to the keyboard, Apple will have a lot more room in the iMac for a quiet cooling system to cool 2 processors. Apple could use a dual processor card employing the same sawtooth socket in Power Mac G4 systems since the PowerMac G4 shipped, thereby saving motherboard space for other components.
Quote:
3. It's not impossible to have a cheap headless box at Apple, but Steve Jobs is loathe to admit that people would want it.
I don't see anything in the first three which is going to bring down costs.
1) Design/test/fabricate new processor - huge cost
2) Design/test/fabricate new controller - huge cost
3) Change to much more expensive discs.
Yes, you cold manage a simpler, cheaper motherboard, and getting rid of the PCI(X) would also reduce costs (although not much), same for the onboard audio.
A whole computer can be shoved into a very small form factor, but it is a lot more expensive than putting one in a big box.
If a new processor with onboard memory controller appears anyway, then it makes good sense to try something like this, but to design one specifically would be ludicrous at the volume it is likely to sell at.
michael
Just wanted to jump in here and say that the onboard memory controller concept will probably be present in the G6 (if Apple chooses to call the Power5 variant a G6).
Look for an expansion of dual machines and little or no change in price.
Thus,
Dual 2.5Ghz 2999
Dual 2Ghz 2499
Single 2 Ghz 1999
Then
G5 CUBES !!!
All single CPU, 1299-1799, 1.6-2.0Ghz G5's depending on price and availability.
And, if Apple really had balls, the intro of a standalone 17" widescreen LCD to be bundled at the 1799 price point and the death of the iMac for all but an extreme low end solution!
I don't see anything in the first three which is going to bring down costs.
1) Design/test/fabricate new processor - huge cost
2) Design/test/fabricate new controller - huge cost
3) Change to much more expensive discs.
More expensive? Compared to what? I haven't seen this as an issue. There certainly is a bit of a premium at the moment, as there is with all new technology, but are we surprised at all here. Beside are we so sure that Apple would be paying those premiums?
Quote:
Yes, you cold manage a simpler, cheaper motherboard, and getting rid of the PCI(X) would also reduce costs (although not much), same for the onboard audio.
A whole computer can be shoved into a very small form factor, but it is a lot more expensive than putting one in a big box.
If a new processor with onboard memory controller appears anyway, then it makes good sense to try something like this, but to design one specifically would be ludicrous at the volume it is likely to sell at.
The whole equation is dependant on Apple / IBM implementing a alternative G5 for lowcosts systems. I just don't see the current G5 implementation migrating to low cost systems anytime soon. I could be completely wrong on that but it is what I precieve right now.
I don't think the costs would be as huge as some imagine either, First we are talking hihg volumn. Second there is a great deal of resue of peripherials and standard cells. Since the memory controller is on board the PPC and you intend ot limit the number of ports supported I see this as very doable. We are just talking AGP, USB, Firewire, Networking and SATA here. The SATA could easily be replaced with a Firewire port also. So you're basically talking about a DMA cntroller manageing 4 or 5 types of ports. Seems to be completely doable. It is really not much more than is traditionally supported on a laptop. With a little thought the chip set could support both the laptop and the small form factor market.
Multiple iterations of a chip set happen all the time. Just look at how many Intel produces, though Intel seems to lack the maketing skills to fully explot some of those sets.
Quote:
michael
Obviously I have no idea what Apple s up to. I do know that they don't ignore market trends, small form factor is a reality and in Apples case is important due to the size of the G5. People will want alternatives and while the IMac is nice it has a couple of issues that I personaly don't like. That mainly revolves wanting the CPU unit out of site along with all the wires dangling out the back. The perfect IMac would have a screen arraingement similar to what it currently has with one thing cable running to the CPU assembly else where.
It doesn't have to be this complicated for cheap G5 machines. Apple should use the same exact ASIC chips in the PowerMac G5.
The system ASIC supports dual CPUs, 128 bit dual channel DDR memory, AGP 8x Pro, and 16 bit Hypertransport. In a low-cost system, they only use 1 CPU, a slower CPU bus, single channel DDR memory, AGP 4x. These changes should be nothing but resistor settings to the system ASIC. If they are not they should be.
I'm still leaning towards on chip memory interface. The performance of the G5 is to closely tied to the dual channel arrangement. Maybe if this new PPC had larger caches we would be all set.
Quote:
For the I/O ASIC, it supports too many things to list, but it's the same concept as the system ASIC. For hard drives, use the ATA/100 bus that is already on there. Use a cheaper Firewire PHY layer that only supports Firewire 400, use a cheaper audio device, use the same PCI-based USB 480 device, etc.
So I'm not disagree that the I/O asic supports alot of stuff . The question is how much does it really have to support. My postion is not much especially if one relys on USB and Firewire to take over traditional hardware.
Quote:
Lastly, don't use the PCI/PCI-X Hypertransport tunnel, and just directly connect the system ASIC and I/O ASIC with the Hypertransport buses already on there. If they are not compatible HT buses, they should be made to be. The rest is just using cheaper graphics card, RAM, etc.
This is exactly what I don't want. I want a low cost system with reasonalbly good performance. That means a very good graphics processor, like whatever is #2 at the moment. The rest of the I/O should provide state of the art performance. That is why I want to limit capabilities, give the user what they need but don't scrimp.
The reality of Hypertransport just hit me like a 2x4. If this was implemented onboard the PPC along with the memory controller Apple would have a huge number of options for I/O, as HT has become widely supported. But agian don't add things not needed. Lots of USB, firewire ports, SATA port and thats about it. What would be even neater is seeing a HT based graphics chip implemented - that would be hot.
Quote:
Remember my optical-drive-in-the-keyboard idea? Well... how about we go really really old-school.
If they release a drastically improved lineup before March, I'll poop a brick. I'm about to order one (because I need the horse-power now as opposed to four months from now). I expect an update fairly soon because I doubt they'll jump 1000MHz in less than three updates (total). It just can't be a "big" update.
Such is the nature of this beast. Everything you buy just became obsolete.
Comments
Originally posted by wizard69
Here is how to make a low costs feature rich mini.
1. A high integration 970, that is a chip with an on board memory controller.
2. A new high integration bus interface chip.
3. Move to notebook form factor SATA drives.
4. No sound hardware, rely instead on external USB speakers.
It doesn't have to be this complicated for cheap G5 machines. Apple should use the same exact ASIC chips in the PowerMac G5.
The system ASIC supports dual CPUs, 128 bit dual channel DDR memory, AGP 8x Pro, and 16 bit Hypertransport. In a low-cost system, they only use 1 CPU, a slower CPU bus, single channel DDR memory, AGP 4x. These changes should be nothing but resistor settings to the system ASIC. If they are not they should be.
For the I/O ASIC, it supports too many things to list, but it's the same concept as the system ASIC. For hard drives, use the ATA/100 bus that is already on there. Use a cheaper Firewire PHY layer that only supports Firewire 400, use a cheaper audio device, use the same PCI-based USB 480 device, etc.
Lastly, don't use the PCI/PCI-X Hypertransport tunnel, and just directly connect the system ASIC and I/O ASIC with the Hypertransport buses already on there. If they are not compatible HT buses, they should be made to be. The rest is just using cheaper graphics card, RAM, etc.
Look at it this way if a whole computer can be stuffed into a laptop it should be a piece of cake to do a mini desk top. I;m not talking the current PC route whcih has normal motherboards stuffed into mini or mivro enclosures, we are talking a motherboard purposefully designed for a minmalist machine.
Remember my optical-drive-in-the-keyboard idea? Well... how about we go really really old-school.
Originally posted by wizard69
First off minis are hot in the market place right now. Second the new G5 is much larger than many people are willing ot deal with. As to the dual processor iMacs, it will happen sooner or later SMP is actually a cheaper way to improve performance. I wuold not be surprised if the SMP iMacs are a dual core solution.
Child, there's no logic in assuming people who don't want the huge towers will want small-form factor PCs. Most people will want, *GASP*, mini-tower sized PCs. It's what they are used to.
Originally posted by PB
Well, based on Jobs statement (3 GHz in one year), a 2.5 GHz G5 in January is almost certain. For 2.8 GHz, I don't know, we will see... After June's surprise (2 GHz instead of 1.8 GHz), a January surprise would be equally welcome.
Something very important to place emphasis on is this:
Jobs wouldn't have told us about the speed bump if it wasn't certain!
This is the same guy who had to eat crow over the 500MHz G4 Moto debacle. There is NO way that he is going to promise a speed gain that won't be delivered. So I think that we can assume that your timeline is correct because I profoundly doubt that Apple is going to jump from 2GHz to 3GHz in one single jump. We should see a speed boost announced in January (hopefully along with a second dual CPU machine so that there are 2 dualies on the lineup).
Originally posted by THT
Even though the 7457 has more transistors than the 970, it's die size (98 sq mm) is actually smaller than the 970 (118 sq mm). IBM's floorplan for the 970 seems atrocious considering .....
True, but having more transistors and having them more densly packed has to affect production costs. If I remember correctly, it has been mentioned that IBM utilizes some automated design tools to speed up the actual design. This might explain the less densly packed transistors??
If the 970 has much better yield than the 7457, than the 970 may be cheaper.....
No comment here, I've not a clue, well maybe, but no point in appearing to bash Motorola. More densly packed transistors/traces may affect yields??
Apple sells an eMac with a 1 GHz 7455 CPU, combo optical, 128 MB RAM and 40 GB disk for 799$....
Kind of my point. The MPC7455 is still on the 0.18µm process and unless Motorola is having a fire sale or their yields are tremedous, the IBM 970 should be price competative with the MPC7455/7445 and the MPC7457/7447.
I just can't see why using the G4's, in and of itself, would allow Apple to produce computers cheaper than using the G5. Unless of course all the other stuff, companion chip, Hypertransport, etc. disproportionately increases the cost over the G4's equivalents.
Originally posted by Yevgeny
Something very important to place emphasis on is this:
Jobs wouldn't have told us about the speed bump if it wasn't certain!
This is the same guy who had to eat crow over the 500MHz G4 Moto debacle. There is NO way that he is going to promise a speed gain that won't be delivered. So I think that we can assume that your timeline is correct because I profoundly doubt that Apple is going to jump from 2GHz to 3GHz in one single jump. We should see a speed boost announced in January (hopefully along with a second dual CPU machine so that there are 2 dualies on the lineup).
I guess you are right. Anyway i still wait for my dualie. I expect it will ship before the rev B G5 towers
Originally posted by THT
Even though the 7457 has more transistors than the 970, it's die size (98 sq mm) is actually smaller than the 970 (118 sq mm). IBM's floorplan for the 970 seems atrocious considering how tightly they've packed the 750fx, 35 sq mm for 39M transistors. Since the 7457 is 20% smaller than the 970, and given the same yields, the 7457 should be cheaper.
If the 970 has much better yield than the 7457, than the 970 may be cheaper. Moto seems to have had a lot of yield problems with their 130 nm process, so the 970 may very well have been cheaper for a few months. Who knows.
The 970 is made on larger wafers which tends to improve the cost/die.
Originally posted by rickag
I just can't see why using the G4's, in and of itself, would allow Apple to produce computers cheaper than using the G5. Unless of course all the other stuff, companion chip, Hypertransport, etc. disproportionately increases the cost over the G4's equivalents.
If indeed IBM can sell a 1.3 GHz 970 cheaper than Moto can sell a 1.3 GHz 74x7 and associate motherboard costs are about equal, then yeah, Apple would be stupid not to migrate all of its systems to the 970. The real unknown is how much Apple actually pays for each CPU. That's been the question for a long time now.
Also, I think G4 parts are probably a good percentage cheaper than G5 parts right now because they've been in production for such a long time already and Apple can get them cheaper and cheaper as time goes by.
I found it incredibly interesting that the G4 has more transistors than the G5. Doesn't that defy all forms of common sense logic? More complex, powerful chips have more transistors, right?
Anyway, even though all of us would love to see a midrange tower, Apple doesn't want to go that route. It has the eMac covering the low-end, the iMac the <chuckle> "midrange" and the Power Macs at the high end. They're not going to engineer a new case and motherboard for something that will have limited appeal. If the Cube had succeeded, that would be one thing. Since it didn't I doubt we'll see anything similar from Apple for many moons.
Originally posted by Big Mac
Eugene, is child a term of endearment for you? Maybe I'm out of the infinite loop, but surely I can't be the only one wondering what that's all about. . .
I found it incredibly interesting that the G4 has more transistors than the G5. Doesn't that defy all forms of common sense logic? More complex, powerful chips have more transistors, right?
Up to a point. The vast majority of the transistors are in the caches and associated tags.
The G4 has a L3 controller and tags on-board, which probably explains most of the extra transistors. The L3 tags will use a lot of transistors, and very densely, also explaining the higher average density for the G4. The G4 may also use more redundancy in the caches, giving higher yields, increasing transistor count, and increasing average density.
The G5 has more logic transistors, which are less dense than memory transistors.
michael
Originally posted by Big Mac
Eugene, is child a term of endearment for you? Maybe I'm out of the infinite loop, but surely I can't be the only one wondering what that's all about. . .
Child, if you have to ask you don't know!
Originally posted by Big Mac
Eugene, is child a term of endearment for you? Maybe I'm out of the infinite loop, but surely I can't be the only one wondering what that's all about. . .
look child, it's a bit of an inside joke. don't worry about it. it'll get old real quick
Originally posted by Eugene
The current high-end becomes the mid-end. The low-end remains a single CPU machine. In January:
1x1800 MHz - $1800
2x2000 MHz - $2400
2x2400 MHz - $3000
Hasn't the high end normally become the low end on each new rev?
(Except for the jump to G5)
Originally posted by Commodus
What is it with people who make unrealistic predictions?
We're making predictions?
Not to single anyone out in particular (I've seen at least a few others), but posts like THT's make me scratch my head in wonderment. Dual-processor iMacs? G4 and G5 "Minis?" I don't think so. While I can't rule it out, I'm even skeptical of dual-1.8 GHz G5s at the low end of the PowerMac lineup.
As I said to Eugene. I'm not making a prediction and I'm sure many others aren't either. I'm putting down what Apple should sell, and I'm usually attentive to whether such hardware can be engineered.
1. It (usually) costs considerably more to have dual CPUs than a single CPU at a higher clock speed.
From the Apple store:
PowerMac G4
$1,299.00
1.25GHz PowerPC G4
1MB L3 cache
$1,599.00
Dual 1.25GHz PowerPC G4
2MB L3 cache/processor
Here Apple added another CPU and 3 MB of cache for 300$. In my dual G4 iMac proposal, it would be without L3 cache. Even so, the dual 1.25 GHz PowerMac G4 only costs 1599$. For 200$ more at $1800, Apple sells an iMac 17 with 1 1.25 GHz processor without cache, a SuperDrive, a 17" LCD, and a whole lot less expansion.
So take the dual 1.25 GHz PowerMac G4, take out all of the L3 cache, add a SuperDrive and an LCD screen, and I think Apple can charge 1800$ for it and get the same margins.
2. The iMac has a small enclosure, and probably can't fit two CPUs (at least, not without addressing heat first).
You didn't see my "Firewire based keyboard with an optical drive" idea from a previous thread. By moving the optical drive to the keyboard, Apple will have a lot more room in the iMac for a quiet cooling system to cool 2 processors. Apple could use a dual processor card employing the same sawtooth socket in Power Mac G4 systems since the PowerMac G4 shipped, thereby saving motherboard space for other components.
3. It's not impossible to have a cheap headless box at Apple, but Steve Jobs is loathe to admit that people would want it.
Well, I can't say anything about this.
Originally posted by Derrick 61
Hasn't the high end normally become the low end on each new rev?
(Except for the jump to G5)
When I bought my 2x1000 MHz QuickSilver 2002, the low-end was a single 800 MHz model. The previous high-end was a 2x800 MHz model.
The next revision's low-end model was a 2x867 MHz G4.
Also, the 2x1250 MHz was the high end of that revision, and the low-end of the next revision was a single 1000 MHz G4.
Originally posted by mmicist
I don't see anything in the first three which is going to bring down costs.
1) Design/test/fabricate new processor - huge cost
2) Design/test/fabricate new controller - huge cost
3) Change to much more expensive discs.
Yes, you cold manage a simpler, cheaper motherboard, and getting rid of the PCI(X) would also reduce costs (although not much), same for the onboard audio.
A whole computer can be shoved into a very small form factor, but it is a lot more expensive than putting one in a big box.
If a new processor with onboard memory controller appears anyway, then it makes good sense to try something like this, but to design one specifically would be ludicrous at the volume it is likely to sell at.
michael
Just wanted to jump in here and say that the onboard memory controller concept will probably be present in the G6 (if Apple chooses to call the Power5 variant a G6).
Thus,
Dual 2.5Ghz 2999
Dual 2Ghz 2499
Single 2 Ghz 1999
Then
G5 CUBES !!!
All single CPU, 1299-1799, 1.6-2.0Ghz G5's depending on price and availability.
And, if Apple really had balls, the intro of a standalone 17" widescreen LCD to be bundled at the 1799 price point and the death of the iMac for all but an extreme low end solution!
Originally posted by mmicist
I don't see anything in the first three which is going to bring down costs.
1) Design/test/fabricate new processor - huge cost
2) Design/test/fabricate new controller - huge cost
3) Change to much more expensive discs.
More expensive? Compared to what? I haven't seen this as an issue. There certainly is a bit of a premium at the moment, as there is with all new technology, but are we surprised at all here. Beside are we so sure that Apple would be paying those premiums?
Quote:
Yes, you cold manage a simpler, cheaper motherboard, and getting rid of the PCI(X) would also reduce costs (although not much), same for the onboard audio.
A whole computer can be shoved into a very small form factor, but it is a lot more expensive than putting one in a big box.
If a new processor with onboard memory controller appears anyway, then it makes good sense to try something like this, but to design one specifically would be ludicrous at the volume it is likely to sell at.
The whole equation is dependant on Apple / IBM implementing a alternative G5 for lowcosts systems. I just don't see the current G5 implementation migrating to low cost systems anytime soon. I could be completely wrong on that but it is what I precieve right now.
I don't think the costs would be as huge as some imagine either, First we are talking hihg volumn. Second there is a great deal of resue of peripherials and standard cells. Since the memory controller is on board the PPC and you intend ot limit the number of ports supported I see this as very doable. We are just talking AGP, USB, Firewire, Networking and SATA here. The SATA could easily be replaced with a Firewire port also. So you're basically talking about a DMA cntroller manageing 4 or 5 types of ports. Seems to be completely doable. It is really not much more than is traditionally supported on a laptop. With a little thought the chip set could support both the laptop and the small form factor market.
Multiple iterations of a chip set happen all the time. Just look at how many Intel produces, though Intel seems to lack the maketing skills to fully explot some of those sets.
Quote:
michael
Obviously I have no idea what Apple s up to. I do know that they don't ignore market trends, small form factor is a reality and in Apples case is important due to the size of the G5. People will want alternatives and while the IMac is nice it has a couple of issues that I personaly don't like. That mainly revolves wanting the CPU unit out of site along with all the wires dangling out the back. The perfect IMac would have a screen arraingement similar to what it currently has with one thing cable running to the CPU assembly else where.
Originally posted by THT
It doesn't have to be this complicated for cheap G5 machines. Apple should use the same exact ASIC chips in the PowerMac G5.
The system ASIC supports dual CPUs, 128 bit dual channel DDR memory, AGP 8x Pro, and 16 bit Hypertransport. In a low-cost system, they only use 1 CPU, a slower CPU bus, single channel DDR memory, AGP 4x. These changes should be nothing but resistor settings to the system ASIC. If they are not they should be.
I'm still leaning towards on chip memory interface. The performance of the G5 is to closely tied to the dual channel arrangement. Maybe if this new PPC had larger caches we would be all set.
Quote:
For the I/O ASIC, it supports too many things to list, but it's the same concept as the system ASIC. For hard drives, use the ATA/100 bus that is already on there. Use a cheaper Firewire PHY layer that only supports Firewire 400, use a cheaper audio device, use the same PCI-based USB 480 device, etc.
So I'm not disagree that the I/O asic supports alot of stuff . The question is how much does it really have to support. My postion is not much especially if one relys on USB and Firewire to take over traditional hardware.
Quote:
Lastly, don't use the PCI/PCI-X Hypertransport tunnel, and just directly connect the system ASIC and I/O ASIC with the Hypertransport buses already on there. If they are not compatible HT buses, they should be made to be. The rest is just using cheaper graphics card, RAM, etc.
This is exactly what I don't want. I want a low cost system with reasonalbly good performance. That means a very good graphics processor, like whatever is #2 at the moment. The rest of the I/O should provide state of the art performance. That is why I want to limit capabilities, give the user what they need but don't scrimp.
The reality of Hypertransport just hit me like a 2x4. If this was implemented onboard the PPC along with the memory controller Apple would have a huge number of options for I/O, as HT has become widely supported. But agian don't add things not needed. Lots of USB, firewire ports, SATA port and thats about it. What would be even neater is seeing a HT based graphics chip implemented - that would be hot.
Quote:
Remember my optical-drive-in-the-keyboard idea? Well... how about we go really really old-school.
Such is the nature of this beast. Everything you buy just became obsolete.
Originally posted by Moogs
Such is the nature of this beast. Everything you buy just became obsolete.
It hardly becomes obsolete. Not leading-edge, yes... but not obsolete and certainly not any less capable than when you bought it.