Apple is part of the HYPERTRANSPORT consortium ...
and this was a while back .. seeing as the nForce board ALREADY utilizes hypertransport .. it's not out of the question that Apple/Motorolla could have a working Motherboard/CPU for MWNY utilizing this technology ...
anyone care to comment on hypertransport and wha this might mean to apple?
Switching do a different subject, would parts of the OS need to be rewritten/optimized to use Hyper Transport?</strong><hr></blockquote>
My guess is that it would be a new driver to access HyperTransport calls and to deal with sharing between processors. Programmer might be able to help here, or THT.
Depending on how/where it is used no software changes may be required to use HyperTransport. This thing is an interconnect technology which chips communicate with... the actually communication protocol is implemented in hardware like it is for MPX, PCI, AGP, RapidIO, etc.
The oft-quoted figure of only a 10-15% performance improvement due to DDR comes from general purpose benchmarks which are not completely memory bound. If an algorithm runs out of cache then the performance improvement due to doubling memory bandwidth is going to be small. For tasks which are memory bound, however, the performance improvement can be up to the theoretical 100%... it depends on how good the memory controllers in question are. Apple's current memory controller and the MPX bus are quite good -- hopefully they can keep the quality and simply double the performance. Unfortunately that isn't likely to happen. Most PowerPC read operations are 32-byte bursts and with DDR it'll now take half as many clock cycles which means we'll need to pay the burst overhead twice as often. I don't have enough data on MPX to predict how much of an improvement we're likely to see, but I know it is well pipelined so the overhead might be hidden.
I keep seeing comments that imply that the G4 is in some way substandard in its basic function. This simply isn't true -- the majority of instructions execute with one cycle throughput, and it just doesn't get any better than that. Two things can be done, and they can be done (have been done!) to the G4 -- have more execution units, or run the processor at a higher clock rate (or a subtle variation where you run some execution units at a higher rate). The higher clock rates are attained by lengthening the pipelines and improving the process. The downsides of lengthening the pipe are a higher cost to branch misprediction and more stalls due to data dependencies. Adding more execution units means you need to improve the instruction completion buffers and rename registers... both of which have been done on the G4. By applying the same techniques that they have been applying they would be able to realize continued serious performance improvements. The G4 has scaled from 350 MHz to 1 GHz, and will continue to scale... the major problem has been the rate of developments.
Maybe I should have posted this in another thread, but this one rocks
From MOSR:
"The PowerMac will sport a more dramatic improvement, finally bringing the Xserve's DDR266 memory to the mainstream Mac at a reasonable price point and nearly doubling peak memory bandwidth on that model. A new GeForce4-family graphics card will be standard on all models except the entry-level, which will retain the MX card but with more and faster memory than the iMac's MX accelerator. Processor speeds are not entirely guaranteed just yet thanks to the usual problems at Motorola, but regardless the estimates are impressive -- 1.33GHz at the least, 1.5GHz at the best. 1.4GHz seems a reasonable middle of the road guess at this point. These new G4s will also sport some other improvements, including 512KB of on-chip L2 cache and an improved on-chip memory architecture.
"
Well, if this is true.. 1.33 Ghz and DDR-sdram support the Poweermac line will look decent, at least until the 3 Ghz P4 is released ( or that 64 bit AMD ). If it is a new G4 design w. 512 KB 2-level cache and 4 MB 3-level cache it will sure rock.
"Other internal improvements are to include more room for extra internal drives, a possible third external drive bay, new and improved Airport antennas, and six -- count them six -- full-length PCI slots, three of them 64-bit/66MHz and three 64-bit/33MHz. An 8X AGP graphics slot rounds out a full seven beautiful expansion possibilities just waiting for the professional to drool over.
Apple has always been very conservative in this area so... <img src="graemlins/bugeye.gif" border="0" alt="[Skeptical]" />
"Right now it looks like 800Mbps+ Firewire and USB2 are going to wait for the next revision, due in early 2003 that will also most likely see very new processors that Apple, at least, will call G5s."
If 800 Mbps Firewire is not released this summer, USB2 will continue to win market acceptance. Or let me quote the former Mac-head DKE ( yeah I know, I know):
"FireWire already has 800 Mbps designed (1394b), and will evolve to that speed as we need it (and chips become more available). FireWire was designed to scale in the first place -- and there is also a 1.6 Gbps version or two in the books (not quite formalized, but ready to go if needed). FireWire will go to 800 Mbps this year or next, and 1.6 Gbps probably in another year or so after that. USB 2.0 is not born yet... in fact, it isn't even yet fully conceived -- it is more a dirty thought in someone's mind. "
the new tower design/color will not be black as someone indicated. it wouldn't match any of the displays! it will be white like the iBook i suspect. it will also have a silver shiny apple like the new imacs and the HD CD
Ehh....I'm guessing 50% speed increase, going to 1.4GHz DP, DDR-RAM, and Jaguar. I'm also betting that they stay with the 133MHz bus. We will also get a new SuperDrive, though it probably wont be called SuperDrive 2. Might be called something different, but Not SuperDrive 2.</strong><hr></blockquote>
<strong>Maybe I should have posted this in another thread, but this one rocks
From MOSR:
"The PowerMac will sport a more dramatic improvement, finally bringing the Xserve's DDR266 memory to the mainstream Mac at a reasonable price point and nearly doubling peak memory bandwidth on that model. A new GeForce4-family graphics card will be standard on all models except the entry-level, which will retain the MX card but with more and faster memory than the iMac's MX accelerator. Processor speeds are not entirely guaranteed just yet thanks to the usual problems at Motorola, but regardless the estimates are impressive -- 1.33GHz at the least, 1.5GHz at the best. 1.4GHz seems a reasonable middle of the road guess at this point. These new G4s will also sport some other improvements, including 512KB of on-chip L2 cache and an improved on-chip memory architecture.
"
Well, if this is true.. 1.33 Ghz and DDR-sdram support the Poweermac line will look decent, at least until the 3 Ghz P4 is released ( or that 64 bit AMD ). If it is a new G4 design w. 512 KB 2-level cache and 4 MB 3-level cache it will sure rock.</strong><hr></blockquote>
This is BS...don't listen to Ryan. The G4 can't clock to 1.4GHz or 1.5Ghz. It can, however, clock to 1.466GHz and 1.6GHz. Expect 1.466GHz at the best.
Apple is part of the HYPERTRANSPORT consortium ...
and this was a while back .. seeing as the nForce board ALREADY utilizes hypertransport ....
anyone care to comment on hypertransport and wha this might mean to apple?
QB]<hr></blockquote>
One interesting thing in one of the presentations at the hypertransport site was the claim that hypertransport plays nicer with PCI than RapidIO since it was built with the PCI Spec in Mind. Since PCI isn't going away in the PowerMac anytime soon I thought that was interesting... <img src="graemlins/bugeye.gif" border="0" alt="[Skeptical]" />
<strong>This is BS...don't listen to Ryan. The G4 can't clock to 1.4GHz or 1.5Ghz. It can, however, clock to 1.466GHz and 1.6GHz. Expect 1.466GHz at the best.</strong><hr></blockquote>
Quibbling about these details is pointless... adding clock multipliers or changing the speed of the MPX bus could happen (Moto admitted that a 166 MHz MPX might show up and with a 9x multiplier you'd be close enough to 1.5GHz to call it that).
This is BS...don't listen to Ryan. The G4 can't clock to 1.4GHz or 1.5Ghz. It can, however, clock to 1.466GHz and 1.6GHz. Expect 1.466GHz at the best.</strong><hr></blockquote>
Spart, FYI, Ryan Meader hasn't been involved with MacOSRumors.com for some time -- years, actually.
ob G5, no one believed me when I said not to hold your breath waiting for the G5 a year ago (and then again 6 months ago), so I don't see the point of trying that again.
"Spart, FYI, Ryan Meader hasn't been involved with MacOSRumors.com for some time -- years, actually.
ob G5, no one believed me when I said not to hold your breath waiting for the G5 a year ago (and then again 6 months ago), so I don't see the point of trying that again."
Okay, 'Moki'.
I'll fish then.
Re: the 'G5'.
1. Know anthing about it being 'killed' due to bus/bandwidth issues?
2. Was it killed/delayed because it 'wasn't enough'?
3. Are Apple 'really' involved in its development? Is it that holding it up?
4. Is the G5 'not really' the G5 we thought it was..? Has its tech' mandate changed? HAS the chip supplier of said G5 changed? (IBM...AMD? No teasing now...)
5. When CAN we expect 'it'? (My worse nightmare says we'll be lucky at Mac World New York next year ie mid-2003!)
6. Is the 'G5' the 7500 chip the 'Register' says will launch next January on Rapid Io? Or is the 7500 'merely' a G4 in Rapid Io clothes?
7. Is the 'new' Apollo gonna add a fpu to that solitary fpu in the current G4? (Seems lonely and out gunned compared to the Athlon's three?)
8. Er...where were you on the night in question...
"ob G5, no one believed me when I said not to hold your breath waiting for the G5 a year ago (and then again 6 months ago), so I don't see the point of trying that again. "
I'll believe you if you wax more cryptically around my questions...
Lemon, there is little point in questioning Moki about this as there are two possibilities:
1) He isn't covered by NDA, and thus doesn't know anything.
2) He is covered by NDA and can only say this about the G5 because things which don't exist typically aren't covered by NDA.
This still applies even if friends of his have NDA, since Apple seems pretty aggressive about culling leaks and I'm sure Moki wouldn't sell his friends out just to post on AI... especially if that would hurt Apple, since hurting Apple will hurt Ambrosia.
We need our sources to be anonymous, secretive, and untrustworthy so that we can trust that they are potentially legit.
Spart, FYI, Ryan Meader hasn't been involved with MacOSRumors.com for some time -- years, actually.
ob G5, no one believed me when I said not to hold your breath waiting for the G5 a year ago (and then again 6 months ago), so I don't see the point of trying that again. </strong><hr></blockquote>
[quote] ob G5, no one believed me when I said not to hold your breath waiting for the G5 a year ago (and then again 6
months ago), so I don't see the point of trying that again.
<hr></blockquote>
So maybe we should start a thread entitled, "G4 Forever". (Of course with the current state of the world, forever might not be so long.)
I would, though, like to know whether there is a next generation processor, and if not, how does Apple (and all of you whose livlihoods depend on Apple) plan to survive?
Comments
Apple is part of the HYPERTRANSPORT consortium ...
and this was a while back .. seeing as the nForce board ALREADY utilizes hypertransport .. it's not out of the question that Apple/Motorolla could have a working Motherboard/CPU for MWNY utilizing this technology ...
anyone care to comment on hypertransport and wha this might mean to apple?
thx
ps. im new to this forum .. hello to all !!!
<strong>
MOSR posted it then pulled it a day or two later</strong><hr></blockquote>
Maybe this means Ryan finnally figured out how stupid what he was saying is...
Switching do a different subject, would parts of the OS need to be rewritten/optimized to use Hyper Transport?
<strong>
Switching do a different subject, would parts of the OS need to be rewritten/optimized to use Hyper Transport?</strong><hr></blockquote>
My guess is that it would be a new driver to access HyperTransport calls and to deal with sharing between processors. Programmer might be able to help here, or THT.
The oft-quoted figure of only a 10-15% performance improvement due to DDR comes from general purpose benchmarks which are not completely memory bound. If an algorithm runs out of cache then the performance improvement due to doubling memory bandwidth is going to be small. For tasks which are memory bound, however, the performance improvement can be up to the theoretical 100%... it depends on how good the memory controllers in question are. Apple's current memory controller and the MPX bus are quite good -- hopefully they can keep the quality and simply double the performance. Unfortunately that isn't likely to happen. Most PowerPC read operations are 32-byte bursts and with DDR it'll now take half as many clock cycles which means we'll need to pay the burst overhead twice as often. I don't have enough data on MPX to predict how much of an improvement we're likely to see, but I know it is well pipelined so the overhead might be hidden.
I keep seeing comments that imply that the G4 is in some way substandard in its basic function. This simply isn't true -- the majority of instructions execute with one cycle throughput, and it just doesn't get any better than that. Two things can be done, and they can be done (have been done!) to the G4 -- have more execution units, or run the processor at a higher clock rate (or a subtle variation where you run some execution units at a higher rate). The higher clock rates are attained by lengthening the pipelines and improving the process. The downsides of lengthening the pipe are a higher cost to branch misprediction and more stalls due to data dependencies. Adding more execution units means you need to improve the instruction completion buffers and rename registers... both of which have been done on the G4. By applying the same techniques that they have been applying they would be able to realize continued serious performance improvements. The G4 has scaled from 350 MHz to 1 GHz, and will continue to scale... the major problem has been the rate of developments.
Fair comment.
To me. A dual gig last Summer would have seen AIM getting back on track.
Now. It's lucky if we hit 1.2 being really optimistic.
Still.
I wouldn't mind the G4 being at 1.2 mhz if it had 3 fpu like the Athlon(!), a 4 megs of level 3 cache and DDR and faster bus.
No. I'm kidding myself.
I want the g5 with rapid io and gravy on top.
Lemon Bon Bon
From MOSR:
"The PowerMac will sport a more dramatic improvement, finally bringing the Xserve's DDR266 memory to the mainstream Mac at a reasonable price point and nearly doubling peak memory bandwidth on that model. A new GeForce4-family graphics card will be standard on all models except the entry-level, which will retain the MX card but with more and faster memory than the iMac's MX accelerator. Processor speeds are not entirely guaranteed just yet thanks to the usual problems at Motorola, but regardless the estimates are impressive -- 1.33GHz at the least, 1.5GHz at the best. 1.4GHz seems a reasonable middle of the road guess at this point. These new G4s will also sport some other improvements, including 512KB of on-chip L2 cache and an improved on-chip memory architecture.
"
Well, if this is true.. 1.33 Ghz and DDR-sdram support the Poweermac line will look decent, at least until the 3 Ghz P4 is released ( or that 64 bit AMD ). If it is a new G4 design w. 512 KB 2-level cache and 4 MB 3-level cache it will sure rock.
"Other internal improvements are to include more room for extra internal drives, a possible third external drive bay, new and improved Airport antennas, and six -- count them six -- full-length PCI slots, three of them 64-bit/66MHz and three 64-bit/33MHz. An 8X AGP graphics slot rounds out a full seven beautiful expansion possibilities just waiting for the professional to drool over.
"
Hmmm 8X AGP <img src="graemlins/bugeye.gif" border="0" alt="[Skeptical]" />
Apple has always been very conservative in this area so... <img src="graemlins/bugeye.gif" border="0" alt="[Skeptical]" />
"Right now it looks like 800Mbps+ Firewire and USB2 are going to wait for the next revision, due in early 2003 that will also most likely see very new processors that Apple, at least, will call G5s."
If 800 Mbps Firewire is not released this summer, USB2 will continue to win market acceptance. Or let me quote the former Mac-head DKE ( yeah I know, I know):
"FireWire already has 800 Mbps designed (1394b), and will evolve to that speed as we need it (and chips become more available). FireWire was designed to scale in the first place -- and there is also a 1.6 Gbps version or two in the books (not quite formalized, but ready to go if needed). FireWire will go to 800 Mbps this year or next, and 1.6 Gbps probably in another year or so after that. USB 2.0 is not born yet... in fact, it isn't even yet fully conceived -- it is more a dirty thought in someone's mind. "
DKE, 1999
<a href="http://www.mackido.com/Hardware/USB2.html" target="_blank">http://www.mackido.com/Hardware/USB2.html</a>
I'm still waiting for 800 Mbps..
Anyway, there is a buzz about a 7470 chip......Anyone still think there is a G5 in 2002?
Jet, aka, TING5, aka SdC, aka Al T. Veck, aka a buncha other names
<strong>
Ehh....I'm guessing 50% speed increase, going to 1.4GHz DP, DDR-RAM, and Jaguar. I'm also betting that they stay with the 133MHz bus. We will also get a new SuperDrive, though it probably wont be called SuperDrive 2. Might be called something different, but Not SuperDrive 2.</strong><hr></blockquote>
uberdrive
<strong>Maybe I should have posted this in another thread, but this one rocks
From MOSR:
"The PowerMac will sport a more dramatic improvement, finally bringing the Xserve's DDR266 memory to the mainstream Mac at a reasonable price point and nearly doubling peak memory bandwidth on that model. A new GeForce4-family graphics card will be standard on all models except the entry-level, which will retain the MX card but with more and faster memory than the iMac's MX accelerator. Processor speeds are not entirely guaranteed just yet thanks to the usual problems at Motorola, but regardless the estimates are impressive -- 1.33GHz at the least, 1.5GHz at the best. 1.4GHz seems a reasonable middle of the road guess at this point. These new G4s will also sport some other improvements, including 512KB of on-chip L2 cache and an improved on-chip memory architecture.
"
Well, if this is true.. 1.33 Ghz and DDR-sdram support the Poweermac line will look decent, at least until the 3 Ghz P4 is released ( or that 64 bit AMD ). If it is a new G4 design w. 512 KB 2-level cache and 4 MB 3-level cache it will sure rock.</strong><hr></blockquote>
This is BS...don't listen to Ryan. The G4 can't clock to 1.4GHz or 1.5Ghz. It can, however, clock to 1.466GHz and 1.6GHz. Expect 1.466GHz at the best.
[QB]If i i do recall ...
Apple is part of the HYPERTRANSPORT consortium ...
and this was a while back .. seeing as the nForce board ALREADY utilizes hypertransport ....
anyone care to comment on hypertransport and wha this might mean to apple?
QB]<hr></blockquote>
One interesting thing in one of the presentations at the hypertransport site was the claim that hypertransport plays nicer with PCI than RapidIO since it was built with the PCI Spec in Mind. Since PCI isn't going away in the PowerMac anytime soon I thought that was interesting... <img src="graemlins/bugeye.gif" border="0" alt="[Skeptical]" />
<a href="http://www.hypertransport.org/docs/HT_Comdex_Integrated.pdf" target="_blank">linkage (PDF)</a>
<strong>This is BS...don't listen to Ryan. The G4 can't clock to 1.4GHz or 1.5Ghz. It can, however, clock to 1.466GHz and 1.6GHz. Expect 1.466GHz at the best.</strong><hr></blockquote>
Quibbling about these details is pointless... adding clock multipliers or changing the speed of the MPX bus could happen (Moto admitted that a 166 MHz MPX might show up and with a 9x multiplier you'd be close enough to 1.5GHz to call it that).
<strong>
This is BS...don't listen to Ryan. The G4 can't clock to 1.4GHz or 1.5Ghz. It can, however, clock to 1.466GHz and 1.6GHz. Expect 1.466GHz at the best.</strong><hr></blockquote>
Spart, FYI, Ryan Meader hasn't been involved with MacOSRumors.com for some time -- years, actually.
ob G5, no one believed me when I said not to hold your breath waiting for the G5 a year ago (and then again 6 months ago), so I don't see the point of trying that again.
ob G5, no one believed me when I said not to hold your breath waiting for the G5 a year ago (and then again 6 months ago), so I don't see the point of trying that again."
Okay, 'Moki'.
I'll fish then.
Re: the 'G5'.
1. Know anthing about it being 'killed' due to bus/bandwidth issues?
2. Was it killed/delayed because it 'wasn't enough'?
3. Are Apple 'really' involved in its development? Is it that holding it up?
4. Is the G5 'not really' the G5 we thought it was..? Has its tech' mandate changed? HAS the chip supplier of said G5 changed? (IBM...AMD? No teasing now...)
5. When CAN we expect 'it'? (My worse nightmare says we'll be lucky at Mac World New York next year ie mid-2003!)
6. Is the 'G5' the 7500 chip the 'Register' says will launch next January on Rapid Io? Or is the 7500 'merely' a G4 in Rapid Io clothes?
7. Is the 'new' Apollo gonna add a fpu to that solitary fpu in the current G4? (Seems lonely and out gunned compared to the Athlon's three?)
8. Er...where were you on the night in question...
'G5' true believer.
Lemon Bon Bon
[ 06-04-2002: Message edited by: Lemon Bon Bon ]</p>
I'll believe you if you wax more cryptically around my questions...
Lemon Bon Bon
1) He isn't covered by NDA, and thus doesn't know anything.
2) He is covered by NDA and can only say this about the G5 because things which don't exist typically aren't covered by NDA.
This still applies even if friends of his have NDA, since Apple seems pretty aggressive about culling leaks and I'm sure Moki wouldn't sell his friends out just to post on AI... especially if that would hurt Apple, since hurting Apple will hurt Ambrosia.
We need our sources to be anonymous, secretive, and untrustworthy so that we can trust that they are potentially legit.
<strong>
Spart, FYI, Ryan Meader hasn't been involved with MacOSRumors.com for some time -- years, actually.
ob G5, no one believed me when I said not to hold your breath waiting for the G5 a year ago (and then again 6 months ago), so I don't see the point of trying that again. </strong><hr></blockquote>
I believed ya.
Okay. Moki, can you post answers to those questions anonymously?
Lemon Bon Bon
Always good to have someone reasonable on the boards.
months ago), so I don't see the point of trying that again.
<hr></blockquote>
So maybe we should start a thread entitled, "G4 Forever". (Of course with the current state of the world, forever might not be so long.)
I would, though, like to know whether there is a next generation processor, and if not, how does Apple (and all of you whose livlihoods depend on Apple) plan to survive?