DDR on the Powermac in July: two possibilities

2

Comments

  • Reply 21 of 44
    nandonando Posts: 1member
    I think we are getting a NEW BOX!



    just check out the new Xserve.
  • Reply 22 of 44
    Of course I can only offer idle speculation, but I think we can probably look forward to a minor upgrade in clock-speed - Around 1 to 1.3 Ghz, and the same DDR solution as the Xserve, probably in a new enclosure. I wouldn't be surprised if they also threw in a new 2 button mouse and updated SuperDrive for that extra "Wow" effect.

    I'm not trying to be pessimistic, but as others have pointed out the Apollo G4 series is fairly new, and still has some legs to it. A process shrink and some minor revisions to the bus and cache controllers and it can easily scale up to about 1.5 Ghz. Apple has already come out and said that there will be no G5 in the near future.

    If the Powermacs move to DDR 266 while the iMac bus is still at 100 Mhz SDR, and the Xserve is at 133 Mhz SDR, then it makes the other two (new) product lines look outdated already.

    The new iMac is still too new for a major update to the bus design and the Xserve will only have been shipping for about a month. If Apple could release a full DDR system this summer, I think they would have put it in the Xserve. Whether it would be truly useful in a server is beside the point, if the desktops have a percieved performance advantage over the servers, the server sales are going to suffer.

    Frankly, I don't care one way or the other. I'm actually going to buy a 800 Mhz QuickSilver now. The price is right, and the performance will be close to twice that of my 400 Mhz iMac DV. I guess that's all I need.

    For those of you who need more, though, I hope Steve and Co. have some real surprises in store this summer... <img src="graemlins/smokin.gif" border="0" alt="[Chilling]" />
  • Reply 23 of 44
    amorphamorph Posts: 7,112member
    [quote]Originally posted by Neofelis9:

    <strong>If Apple could release a full DDR system this summer, I think they would have put it in the Xserve. Whether it would be truly useful in a server is beside the point, if the desktops have a percieved performance advantage over the servers, the server sales are going to suffer.</strong><hr></blockquote>



    Uh, no. Servers and desktops have different requirements.



    If anything, the XServe is unusual for being ahead of Apple's towers. Usually it's the other way 'round for rack servers, and particularly 1U servers, which are the rack equivalent of notebooks.



    Even if the towers get a new chip that supports a DDR frontside bus, the XServe will have a proven processor, all that monitoring hardware/software, space efficiency and rackmountability (which is a make-or-break deal for a lot of people - that's why Apple built XServe instead of telling people to rackmount their towers), ridiculous ease of maintenance, and massive disk throughput (which is far more important to a server than CPU to memory bandwidth).



    A radically redesigned PowerMac will not affect XServe sales significantly. It should affect PowerMac sales significantly, though.
  • Reply 24 of 44
    jonasvdljonasvdl Posts: 12member
    the reason why we are all waiting for an apple g5 is because we want to beat the pc's

    because most apple users are fanatics ,and i am one to.

    but why are we waiting for those chips.

    is it realy gonna effect the way we computer?

    apple is faster anyway but pc users just dont admit it, well we know better.

    i even bet that almost none of us have the fastest g4 so it is not gonna change the way you computer ,when the time is there apple will introduce an 2 ghz machine you wil be computering with an 933 so lets wait for a real anwser to this question.



    grtz
  • Reply 25 of 44
    hornethornet Posts: 76member
    The way I see it:



    We used to have higher than x86 performance, and paid the higher for it

    great



    We used to have equal x86 performance, and paid higher for it

    ok



    We now have lower than x86 performance, and pay higher for it

    not on at all



    I'm willing to pay the apple tax (the $500 that just shouldn't be there on the low model... sigh) OR pay for a slightly slower than x86 machine, but for the SAME price. Meaning if the mac line went to 1, 1.2, DP1.2 (a "lower than x86" outcome), I would expect the towers at $999, $1500, $1999.



    Take a pick apple!
  • Reply 26 of 44
    spookyspooky Posts: 504member
    [quote]Originally posted by Junkyard Dawg:

    <strong>I'll sum it up for everyone:



    Apple's testicles are in Motorola's vice, and ever since the clone wars Motorola has been winding it tighter and tighter and tighter...



    Until Motorola offers Apple a G4 that can support a DDR front side bus, there is NOTHING that Apple can do other than divert attention from the guts of their hardware with bitchin' software and the almighty OS X. It is so tragic that Apple is leaps and bounds ahead of Wintels in every area under their control, yet just as Apple is poised to capture new market share, Motorola rips open Apple's scrotum and gives 'em a bucket to keep the blood off the floor.



    The tragedy is that regardless of the inspiration, creativity, and all out brute effort of Apple employees, it all is rendered secondary by Motorola's lack of effort. And lack of effort it is; motorola simply refuses to commit the resources needed to develop a competitive PPC chip. Since Apple killed the clones, Motorola has been exacting vengeance by simple neglect. Desktop PPC development means nothing to Moto, their profits are elsewhere, but the killer part is that even if Apple's market share were nudged up to 8 or 10%, Moto's profits would be enormous! The problem is that Moto doesn't have any faith in Apple, hell Moto hasn't even used Macs since the clones were killed!



    I can only imagine how pissed of Jobs must feel, bustin' his guts to make Apple a superior alternative to the Wintel juggernaut, only to have his dreams crushed by Motorola.



    [ 05-15-2002: Message edited by: Junkyard Dawg ]</strong><hr></blockquote>



    I still don't get this. For a man (Steve) who throws cameras at techies in front of a live audience, fires people in lifts (Allegedly), dumps ATI over a business card etc How is it that if Steve recognises the problems and agononises over them he hasn't sought to do anything about it and fast? Is the contract between apple and MOTO [I]THAT[/] binding?





    hell even George Michael got out of his contract . . .



    I reckon that apple sees no problem with what it offers and herin lies the problem . . .
  • Reply 27 of 44
    programmerprogrammer Posts: 3,457member
    [quote]Originally posted by spooky:

    <strong>I still don't get this. For a man (Steve) who throws cameras at techies in front of a live audience, fires people in lifts (Allegedly), dumps ATI over a business card etc How is it that if Steve recognises the problems and agononises over them he hasn't sought to do anything about it and fast? Is the contract between apple and MOTO THAT binding?





    hell even George Michael got out of his contract . . .



    I reckon that apple sees no problem with what it offers and herin lies the problem . . .</strong><hr></blockquote>





    2 points:



    1) This is an engineering problem and no amount of temper tantrums will make it go away. It takes time and money.



    2) We don't know what their release or product schedule is... given the security that Apple has had since Jobs returned our advanced warning of releases has been on the order of a week or two tops and often the details are incorrect.





    I haven't seen anything that says "Apple says no G5 in near future". Moto said something like that 6 months ago, and they also said that the "G4" has legs. I put G4 in quotes because they didn't say what changes they would make to the G4 and still call it a G4. People keep crying for the G5, but a G4 could arrive which satisfies all of their requirements (except 64-bit which isn't widely needed and seems to be earmarked for G5 by the roadmap).



    There isn't some magic new technology that is going into the G5 that will make it blazingly fast, and yet couldn't be applied to the G4 series. If they add in a couple more functional units (FPU, IPU), allow more instructions to complete per cycle, lengthen the pipelines to 10 so the clock rate jumps to 1.8 GHz, build in a DDR (or QDR ) memory controller, increase the cache sizes, and add a RapidIO interface... then you've got a kick butt chip. They could even add hyper-threading or multi-core to the G4.



    What Motorola and Apple decide to call these things is quite arbitrary and doesn't really matter because they would be very fast.



    [ 05-19-2002: Message edited by: Programmer ]</p>
  • Reply 28 of 44
    lemon bon bonlemon bon bon Posts: 2,383member
    Good posts Programmer and Hornet.



    The original G5 as I heard of it was nothing more than a 10 stage pipeline G4.



    Single core. That's it. I think (correct me if I'm wrong...) Motoman said as much when he was in town.



    So, in terms of what the G5 brought to the table? Not much...except mhz. A stretched G4 pipeline with linear performance increases as the mhz gets ramped.



    The only special thing about the Motorola G5 was that its mhz was higher, yes, but that it would be plopped into a new kickass architecture called Rapid Io. (500mhz bus speeds?)



    Well, plop a ramped mhz .13 Apollo into a 'Rapid Io' mobo and you've got an impressive performer. Dual Apollos at 1.5 gig on Rapid Io with 4 meg of level 3 cache and a Nvidia 5 will be an impressive performer and surely equal to what Moto' had planned with the original G5 single core 10 stager (By the way, 4meg of level 3 cache not necessary on Rapid Io because processor is communicating at full speed?)



    What changed my mind and got me waiting (sadly) for the G5 was the leaked specs on the net which showed a 'G5' trouncing the fastest Intels and AMDs by a whopping margin.



    Either Moto upped the anti or somebody else is making this 'G5' chip. If it was a dual core G3 with Simd unit running at over a gig. That would surely over power what Apple has now in its top end?



    This 'G5' spec rumour showed a bold projection of Apple not only catching up...but leapfrogging ahead.



    And, as anyone in 3D will tell you, ya can use that kind of power. Especially if you can slot in another G5 to double the rumoured spec marks.



    True? I don't know. Nobody does. I think it came up on the Arsehole technica boards. Can't quite recall.



    But then, I'm not technical guy, have a poor memory for detail and will gladly leave it to the experts on these boards to convince me of what the hell is going on with cpus on 'Power'pc over the next year.



    Lemon Bon Bon
  • Reply 29 of 44
    spartspart Posts: 2,060member
    [quote]Originally posted by Lemon Bon Bon:

    <strong>Good posts Programmer and Hornet.



    The original G5 as I heard of it was nothing more than a 10 stage pipeline G4.



    Single core. That's it. I think (correct me if I'm wrong...) Motoman said as much when he was in town.



    So, in terms of what the G5 brought to the table? Not much...except mhz. A stretched G4 pipeline with linear performance increases as the mhz gets ramped.



    The only special thing about the Motorola G5 was that its mhz was higher, yes, but that it would be plopped into a new kickass architecture called Rapid Io. (500mhz bus speeds?)



    Well, plop a ramped mhz .13 Apollo into a 'Rapid Io' mobo and you've got an impressive performer. Dual Apollos at 1.5 gig on Rapid Io with 4 meg of level 3 cache and a Nvidia 5 will be an impressive performer and surely equal to what Moto' had planned with the original G5 single core 10 stager (By the way, 4meg of level 3 cache not necessary on Rapid Io because processor is communicating at full speed?)



    What changed my mind and got me waiting (sadly) for the G5 was the leaked specs on the net which showed a 'G5' trouncing the fastest Intels and AMDs by a whopping margin.



    Either Moto upped the anti or somebody else is making this 'G5' chip. If it was a dual core G3 with Simd unit running at over a gig. That would surely over power what Apple has now in its top end?



    This 'G5' spec rumour showed a bold projection of Apple not only catching up...but leapfrogging ahead.



    And, as anyone in 3D will tell you, ya can use that kind of power. Especially if you can slot in another G5 to double the rumoured spec marks.



    True? I don't know. Nobody does. I think it came up on the Arsehole technica boards. Can't quite recall.



    But then, I'm not technical guy, have a poor memory for detail and will gladly leave it to the experts on these boards to convince me of what the hell is going on with cpus on 'Power'pc over the next year.



    Lemon Bon Bon </strong><hr></blockquote>



    So you are saying that basically they are going to do the same thing Intel did PIII-P4? Jack up the pipelines, put in some new tech, and get a chip that scales like a mofo and is still, considering pretty damn fast. Seems like the way to go...and it will look good on the spec sheets.



    As for the L3, it will still be pretty useful depending on what type of RAM they use. Connected to the proc more directly, less latency, probably faster DDR-RAM, etc. It just wont make such a huge difference. But also, it makes the specs look that much better.
  • Reply 30 of 44
    mmicistmmicist Posts: 214member
    [quote]Originally posted by Lemon Bon Bon:

    <strong>Good posts Programmer and Hornet.



    The original G5 as I heard of it was nothing more than a 10 stage pipeline G4.



    Single core. That's it. I think (correct me if I'm wrong...) Motoman said as much when he was in town.



    So, in terms of what the G5 brought to the table? Not much...except mhz. A stretched G4 pipeline with linear performance increases as the mhz gets ramped.



    The only special thing about the Motorola G5 was that its mhz was higher, yes, but that it would be plopped into a new kickass architecture called Rapid Io. (500mhz bus speeds?)

    </strong><hr></blockquote>



    RapidIO is not a kick-ass new architecture, for primary memory acess it does'nt really cut it. Although it may run at 500MHz, it is only 16 bits wide, as opposed to the usual 64, so it's bandwidth is equivalent to only a 125MHz standard bus. It is, however, bidirectional so it can send and receive data at the same time with no loss of bandwidth.

    RapidIO is a good bus for connecting to I/O systems (PCI, audio, video etc.), and the RapidIO chip must carry it's own memory controller, and that is the big advance because of considerably reduced latency (with maybe some increase in bandwidth), which would also probably make the L3 cache superfluous if not actually damaging.

    For SMP systems you would need multiple RapidIO busses to avoid considerable bandwidth problems in the resulting NUMA architecture. (see AMD's answer to this problem in their Hammer architecture)



    Michael
  • Reply 31 of 44
    spartspart Posts: 2,060member
    [quote]Originally posted by mmicist:

    <strong>



    RapidIO is not a kick-ass new architecture, for primary memory acess it does'nt really cut it. Although it may run at 500MHz, it is only 16 bits wide, as opposed to the usual 64, so it's bandwidth is equivalent to only a 125MHz standard bus. It is, however, bidirectional so it can send and receive data at the same time with no loss of bandwidth.

    RapidIO is a good bus for connecting to I/O systems (PCI, audio, video etc.), and the RapidIO chip must carry it's own memory controller, and that is the big advance because of considerably reduced latency (with maybe some increase in bandwidth), which would also probably make the L3 cache superfluous if not actually damaging.

    For SMP systems you would need multiple RapidIO busses to avoid considerable bandwidth problems in the resulting NUMA architecture. (see AMD's answer to this problem in their Hammer architecture)



    Michael</strong><hr></blockquote>



    So you are saying that it is effectively a 250MHz bus?
  • Reply 32 of 44
    mmicistmmicist Posts: 214member
    [quote]Originally posted by Spart:

    <strong>



    So you are saying that it is effectively a 250MHz bus?</strong><hr></blockquote>



    Unidirectional bandwidth like a 125MHz DDR bus. However, note that available, as opposed to theoretical, bandwidth is very architecture dependent, Motorola's 133MHz SDR MPX bus achieves benchmark results very similar to AMD's 133 DDR bus, because it is beautifully engineered.



    Michael
  • Reply 33 of 44
    brendonbrendon Posts: 642member
    [quote]Originally posted by mmicist:

    <strong>



    Unidirectional bandwidth like a 125MHz DDR bus. However, note that available, as opposed to theoretical, bandwidth is very architecture dependent, Motorola's 133MHz SDR MPX bus achieves benchmark results very similar to AMD's 133 DDR bus, because it is beautifully engineered.



    Michael</strong><hr></blockquote>



    I believe that Moto lists the bandwidth of rapidIO at 4 gigabyte. Now go to Apple.com and checkout the picture of the server architecture and behold, the Apple ASIC is what talks to main memory and the CPUs talk to main memory through the ASIC. Now replace the 1 gigabyte pipe from the ASIC with a 4 gigabyte pipe. Tweak the CPUs and bump MHz to 1.2GHz or 1.4GHz and you have 100+% increase in performance.
  • Reply 34 of 44
    mmicistmmicist Posts: 214member
    [quote]Originally posted by Brendon:

    <strong>



    I believe that Moto lists the bandwidth of rapidIO at 4 gigabyte. Now go to Apple.com and checkout the picture of the server architecture and behold, the Apple ASIC is what talks to main memory and the CPUs talk to main memory through the ASIC. Now replace the 1 gigabyte pipe from the ASIC with a 4 gigabyte pipe. Tweak the CPUs and bump MHz to 1.2GHz or 1.4GHz and you have 100+% increase in performance.</strong><hr></blockquote>



    Yes, but that's 2 GB/s in each direction (not that I claim that's not a good thing)



    Michael
  • Reply 35 of 44
    lemon bon bonlemon bon bon Posts: 2,383member
    "I'm not angry. "







    I am!







    Interesting thought, Brendon.



    We'll see if it comes true.



    I still like Hornet's post. He's about summed up Apple's lurch from 'ahead' and pricey to 'on a par' and pricey' and behind and damn well overpriced and out of bloody date.



    Lemon Bon Bon
  • Reply 36 of 44
    prestonpreston Posts: 219member
    meh



    boring
  • Reply 36 of 44
    so would you say that the core of apples real problems are b/c of motorola being lame? (sorry, i'm not too hip to all the mega-tech info)



    so it's actually motorola to be frustrated with b/c apple really truely wants to give us what we want? *shrugs*
  • Reply 38 of 44
    lemon bon bonlemon bon bon Posts: 2,383member
    "meh"



    Bahhhh (said the sheep)



    Lemon Bon Bon
  • Reply 39 of 44
    thalmrastthalmrast Posts: 12member
    The G5 is supposed to bring 64-bit computing to the table.
  • Reply 40 of 44
    [quote]Originally posted by thalmrast:

    <strong>The G5 is supposed to bring 64-bit computing to the table.</strong><hr></blockquote>



    i'm sure i'm confused somewhere. i thought the mac already did 128-bit computing??? isn't that the strength of the G4?
Sign In or Register to comment.