i haven't seen any benchmarks of the 1.25Ghz but the 1Ghz-dual doesn't perform much better than the 133mhz-bus-1Ghz-dual because it has only 1MB L3 cache instead of the 2MBL3 of the older 1Gig-dual. however - the 1.25Ghz G4with 166Mhz FSB also has 2MB of L3 cache so i would be very interested in any benchmarks of this machine ....</strong><hr></blockquote>
Is the lack of truly separate vector/scalar units why 128 bit SSE has some operations on two 64 bit doubles? (Better stream loading instructions? Less vector/scalar switches?)
[quote]Will anyone care that a 1.8 GHz 970 is slower than a 3.6 Ghz p4 if you've got the ability to string 970s together like popcorn?<hr></blockquote>
Actually, based on the articles and speculative benchmarks I've seen, it *won't* be slower. Not significantly so anyway.
We'll still have to deal with the marketing aspects because stupid people will still be stupid (i.e. they'll think the clock speed matters), but otherwise we'll be using the most powerful desktop machines on the market it sounds like - and with no RDF to color our perceptions as yet.
Looks like the 970 will be a real powerhouse for multimedia.</strong><hr></blockquote>
yes, that's exactly right. To paraphase one commentator "If I were designing a chip meant to tear through media processing, this would be it"
Also, there is more to this whole 970 thing than has been announced (and what has been announced is much more impressive than some naysayers would believe).
Well, I'd like to see the 970 begin to deliver on the original(?) promise of PPC to be twice as powerful and twice as cheap as the opposition?
I like the idea that the 970 holds hereto undisclosed possibilities.
If Apple has had any hand in the design of this chip ala VMX or perhaps more pertinant to Apple, the bus design or 'controller' 'applePi' then...I'm hopeful that we will get 'little' Apple-isms that will give the Mac platform 'one over' the x86 opposition. Especially in terms of multimedia and the digital hub strategy.
What form the big picture regarding the 970 takes is guesswork for me.
But Apple like their 'integrated' designs and I'd like to see the 970 become form the bedrock of a multimedia powerhouse POWERMac system. Parallelism and massive throughput? If you look at the way 'X' is designed...if Apple could put together a hardware equivalent... Currently, I don't think Apple's hardware is where 'X' is trying to be principle.
I remember seeing those old Amiga demos and seeing stuff that was being done on the desktop for the first time. I'd like to see some kind of break through in the way 970, the graphics card and bandwidth are more tightly integrated to provide realtime previews of 3D renders or something.
I'm already drooling about the prospect of running Lightwave on the 970. But I wonder what delights this rumoured 'Apple Pi' will hold.
One wonders...could Apple Pi merely be a means by which Apple hooks up to different processors? PPC or even x86? (Hey, it was just a thought...)
Having looked at the arse article, the 970 looks like it would tear through Lightwave renders as is...and then there's the prospect of it being a better multiprocessor and VMX player than G4e! The VMX unit on that kinda throughput should really be able to cut loose!
Lemon Bon Bon
Another thought...why didn't IBM/Apple use the 970 to allow the VMX unit to do double precision fp? Too complicated? I'm not clear on whether the VMX unit is merely a compatible port or whether (other than bandwidth improvements...) there is some improvement over the altivec unit in the G4?
The thing I want to know is, the 970 has great scaling possibilities from what I've heard, much better than the G4. So, is IBM known to update these kind of processors frequently? If so, will we see PowerMacs with processor speed bumps every three months, or every six monthsa at witness massive boosts?
<strong>The thing I want to know is, the 970 has great scaling possibilities from what I've heard, much better than the G4. So, is IBM known to update these kind of processors frequently? If so, will we see PowerMacs with processor speed bumps every three months, or every six monthsa at witness massive boosts?
What do you think Apple would go for?</strong><hr></blockquote>
There is one principle reason for going with a deeper pipeline: it is much, much easier to crank up the clockspeed on the chip -- witness the design decisions that Intel made with the Pentium IV. Initial runs were slower than existing Pentium III's, but they were able to ramp up the clockspeed quite quickly because of the pipeline depth.
The IBM chip seems to follow a somewhat similar design philosophy, so I think it is safe to assume that this baby has legs.
The PowerPC 970 has 2 Velocity Engine units, like the G4. The G4e has 4 units, btw. The G4e splits the Simple Int, Complex Int and FP vector unit up into three seperate units.
The G4 and the PPC970 have similar Altivec units, each with one vector permute unit, two integer units, and one floating point unit. I don't see how the G4e has 4 units and the PPC970 has only 2.
Because of SMP potential, the PPC970 only has to be nearly as fast as the Pentium 4 for Apple to build Towers that mop the floor with Wintel's best. Imagine:
Fast: single PPC 970
Faster: Dual PPC 970
Fastest: Quad PPC 970
Unfortunately, it's much more likely that Apple will use single 970s for the fast and faster Powermacs, and a dual configuration for the Fastest Powermac. But still, a dual 970 would be considerably faster than a Pentium 4. Maybe 150% faster!
What I really want to know is how much this chip is going to cost. If its cheap for Apple to put 2 or 4 of these in a machine, then how much will it matter that an expensive P4 (P5) out performs it? Hmmm.... The current Wind-Tunnel G4s raised a few eyebrows when it first came out do to the new case design. These things were designed to disapate heat! A HUGE (7 lbs) heat sink w/ matching fan, a small case fan, 2 fans on the power supply, and a ton of ventalation in the back. WAY more cooling that those 2 little G4s require. I think Apple is trying to avoid the fiasco it had with the Sawtooth (1st gen) G4s where they just slapped a G4 onto a G3 mobo. This time around, I believe they're releasing a new mobo first and then put a new proc in it down the road. I've also read stuff in forums suggesting that the power supply for the Wind-Tunnel had way more juice than the system currently demands. Can anyone out there do the math on this? We know how much power the PPC 970 eats. Can we figure out how much heat the Wind-Tunnel case is designed to disapate? What about how much power the power supply is putting? With these numbers, can we figure out how many PPC 970 the Wind-Tunnel case could power and cool? I've been suffering with a 266MHz G3 iMac, and I refuse to upgrade until Apple comes out with a system that really is worth that premium they charge, and a G4 is not it.
Moki if you say the 970 has legs....how you compare that to the design of the Hammer and its potentional for scaling? Also what is your view on the Pentium Iv' 3 ghz's Hyperthreading? How much of a gain in reality will that give? Also IBM claims to have a similar technology coming up but says it is implimented much better...you have probably read IBM's claims so do you think they are possible?
The G4 and the PPC970 have similar Altivec units, each with one vector permute unit, two integer units, and one floating point unit. I don't see how the G4e has 4 units and the PPC970 has only 2.
</strong><hr></blockquote>
Look again. The G4 diagram has 4 separate units inside the vector unit - one dedicated unit each for permute, simple int, complex int and float. The PPC970 on the other hand seems to have one permute unit, and one big vector math unit that handles simple & complex ints and floats (note that the boxes for those are not "separated", unlike all the other execution units).
[quote]Another thought...why didn't IBM/Apple use the 970 to allow the VMX unit to do double precision fp? <hr></blockquote>
Double precision (64 bit) only gives you two items per 128bit vector, which doesn't have any advantages over using scalar code/units (for VMX/Altivec anyway). 256 bit vectors in VMX/Altivec may happen eventually (four years?) but it would be a major undertaking: doubling the size of the vector registers and more operations (operations on arrays of 16 bytes? :cool
Comments
<strong>
i haven't seen any benchmarks of the 1.25Ghz but the 1Ghz-dual doesn't perform much better than the 133mhz-bus-1Ghz-dual because it has only 1MB L3 cache instead of the 2MBL3 of the older 1Gig-dual. however - the 1.25Ghz G4with 166Mhz FSB also has 2MB of L3 cache so i would be very interested in any benchmarks of this machine ....</strong><hr></blockquote>
FWIW <a href="http://www.barefeats.com/pmddr.html#memory"; target="_blank">This</a> addresses memory speed for the 1 GHz DDR versus SDR
Actually, based on the articles and speculative benchmarks I've seen, it *won't* be slower. Not significantly so anyway.
We'll still have to deal with the marketing aspects because stupid people will still be stupid (i.e. they'll think the clock speed matters), but otherwise we'll be using the most powerful desktop machines on the market it sounds like - and with no RDF to color our perceptions as yet.
<strong>I liked the arse article.
Looks like the 970 will be a real powerhouse for multimedia.</strong><hr></blockquote>
yes, that's exactly right. To paraphase one commentator "If I were designing a chip meant to tear through media processing, this would be it"
Also, there is more to this whole 970 thing than has been announced (and what has been announced is much more impressive than some naysayers would believe).
Part 2 "coming soon"
Part 2 coming soon!
<strong>
Also, there is more to this whole 970 thing than has been announced...</strong><hr></blockquote>
Like how quickly IBM will ramp up production and move past .13 microns?
<strong>Like how quickly IBM will ramp up production and move past .13 microns?</strong><hr></blockquote>
I would rather have IBM not disclosing a few technical details and a possible ultra-low price than the process size
I like the idea that the 970 holds hereto undisclosed possibilities.
If Apple has had any hand in the design of this chip ala VMX or perhaps more pertinant to Apple, the bus design or 'controller' 'applePi' then...I'm hopeful that we will get 'little' Apple-isms that will give the Mac platform 'one over' the x86 opposition. Especially in terms of multimedia and the digital hub strategy.
What form the big picture regarding the 970 takes is guesswork for me.
But Apple like their 'integrated' designs and I'd like to see the 970 become form the bedrock of a multimedia powerhouse POWERMac system. Parallelism and massive throughput? If you look at the way 'X' is designed...if Apple could put together a hardware equivalent... Currently, I don't think Apple's hardware is where 'X' is trying to be principle.
I remember seeing those old Amiga demos and seeing stuff that was being done on the desktop for the first time. I'd like to see some kind of break through in the way 970, the graphics card and bandwidth are more tightly integrated to provide realtime previews of 3D renders or something.
I'm already drooling about the prospect of running Lightwave on the 970. But I wonder what delights this rumoured 'Apple Pi' will hold.
One wonders...could Apple Pi merely be a means by which Apple hooks up to different processors? PPC or even x86? (Hey, it was just a thought...)
Having looked at the arse article, the 970 looks like it would tear through Lightwave renders as is...and then there's the prospect of it being a better multiprocessor and VMX player than G4e! The VMX unit on that kinda throughput should really be able to cut loose!
Lemon Bon Bon
Another thought...why didn't IBM/Apple use the 970 to allow the VMX unit to do double precision fp? Too complicated? I'm not clear on whether the VMX unit is merely a compatible port or whether (other than bandwidth improvements...) there is some improvement over the altivec unit in the G4?
[ 10-29-2002: Message edited by: Lemon Bon Bon ]
[ 10-29-2002: Message edited by: Lemon Bon Bon ]</p>
What do you think Apple would go for?
<strong>The thing I want to know is, the 970 has great scaling possibilities from what I've heard, much better than the G4. So, is IBM known to update these kind of processors frequently? If so, will we see PowerMacs with processor speed bumps every three months, or every six monthsa at witness massive boosts?
What do you think Apple would go for?</strong><hr></blockquote>
There is one principle reason for going with a deeper pipeline: it is much, much easier to crank up the clockspeed on the chip -- witness the design decisions that Intel made with the Pentium IV. Initial runs were slower than existing Pentium III's, but they were able to ramp up the clockspeed quite quickly because of the pipeline depth.
The IBM chip seems to follow a somewhat similar design philosophy, so I think it is safe to assume that this baby has legs.
<strong>
The PowerPC 970 has 2 Velocity Engine units, like the G4. The G4e has 4 units, btw. The G4e splits the Simple Int, Complex Int and FP vector unit up into three seperate units.
Barto</strong><hr></blockquote>
According to these diagrams:
<a href="http://arstechnica.com/cpu/02q2/ppc970/screenshot-1.html"; target="_blank">http://arstechnica.com/cpu/02q2/ppc970/screenshot-1.html</a>;
<a href="http://arstechnica.com/cpu/01q2/p4andg4e/figure3.jpg"; target="_blank">http://arstechnica.com/cpu/01q2/p4andg4e/figure3.jpg</a>;
The G4 and the PPC970 have similar Altivec units, each with one vector permute unit, two integer units, and one floating point unit. I don't see how the G4e has 4 units and the PPC970 has only 2.
Because of SMP potential, the PPC970 only has to be nearly as fast as the Pentium 4 for Apple to build Towers that mop the floor with Wintel's best. Imagine:
Fast: single PPC 970
Faster: Dual PPC 970
Fastest: Quad PPC 970
Unfortunately, it's much more likely that Apple will use single 970s for the fast and faster Powermacs, and a dual configuration for the Fastest Powermac. But still, a dual 970 would be considerably faster than a Pentium 4. Maybe 150% faster!
Would gamers begin flocking to Apple?
( <a href="http://slashdot.org/comments.pl?sid=43568&threshold=0&commentsort=0&ti d=136&mode=thread&cid=4552400" target="_blank">reposted from /. </a> )
What I really want to know is how much this chip is going to cost. If its cheap for Apple to put 2 or 4 of these in a machine, then how much will it matter that an expensive P4 (P5) out performs it? Hmmm.... The current Wind-Tunnel G4s raised a few eyebrows when it first came out do to the new case design. These things were designed to disapate heat! A HUGE (7 lbs) heat sink w/ matching fan, a small case fan, 2 fans on the power supply, and a ton of ventalation in the back. WAY more cooling that those 2 little G4s require. I think Apple is trying to avoid the fiasco it had with the Sawtooth (1st gen) G4s where they just slapped a G4 onto a G3 mobo. This time around, I believe they're releasing a new mobo first and then put a new proc in it down the road. I've also read stuff in forums suggesting that the power supply for the Wind-Tunnel had way more juice than the system currently demands. Can anyone out there do the math on this? We know how much power the PPC 970 eats. Can we figure out how much heat the Wind-Tunnel case is designed to disapate? What about how much power the power supply is putting? With these numbers, can we figure out how many PPC 970 the Wind-Tunnel case could power and cool? I've been suffering with a 266MHz G3 iMac, and I refuse to upgrade until Apple comes out with a system that really is worth that premium they charge, and a G4 is not it.
Got that bit right.
As Diana Ross once sang, 'Still waiting...'
Lemon Bon Bon
<strong>
According to these diagrams:
<a href="http://arstechnica.com/cpu/02q2/ppc970/screenshot-1.html"; target="_blank">http://arstechnica.com/cpu/02q2/ppc970/screenshot-1.html</a>;
<a href="http://arstechnica.com/cpu/01q2/p4andg4e/figure3.jpg"; target="_blank">http://arstechnica.com/cpu/01q2/p4andg4e/figure3.jpg</a>;
The G4 and the PPC970 have similar Altivec units, each with one vector permute unit, two integer units, and one floating point unit. I don't see how the G4e has 4 units and the PPC970 has only 2.
</strong><hr></blockquote>
Look again. The G4 diagram has 4 separate units inside the vector unit - one dedicated unit each for permute, simple int, complex int and float. The PPC970 on the other hand seems to have one permute unit, and one big vector math unit that handles simple & complex ints and floats (note that the boxes for those are not "separated", unlike all the other execution units).
Bye,
RazzFazz
[ 10-29-2002: Message edited by: RazzFazz ]</p>
Double precision (64 bit) only gives you two items per 128bit vector, which doesn't have any advantages over using scalar code/units (for VMX/Altivec anyway). 256 bit vectors in VMX/Altivec may happen eventually (four years?) but it would be a major undertaking: doubling the size of the vector registers and more operations (operations on arrays of 16 bytes? :cool
Does Altivec's design make such a move possible?
<strong>Speculation:
( <a href="http://slashdot.org/comments.pl?sid=43568&threshold=0&commentsort=0&ti d=136&mode=thread&cid=4552400" target="_blank">reposted from /. </a> )
A HUGE (7 lbs) heat sink
</strong><hr></blockquote>
This is garbage. I've personally weighed one of these heatsinks, and it was a few grams over 1500. And that is about 3.3 lbs.