Power Mac 'Quadra' G5 quad processor coming soon?

Posted:
in Future Apple Hardware edited January 2014
Quote:

RUMOR: Power Mac 'Quadra' G5 quad processor coming soon?





Friday, February 27, 2004 - 09:31 AM EST





The following is an unconfirmed rumor, so take it (or leave it) for what it's worth. Our lone source on this one indicates that Apple Computer may be reviving the "Quadra" moniker for an upcoming quad processor G5 Power Mac. Yup, four G5's in one tower - now we know why the case is so roomy. Of course, Mac Quadra's were single processor models only; this is a branding name deal only. All about reviving a bit of the nostalgic past and making it new, we guess.



That is the extent of the information. Rumor only. Grains of salt and all that, of course.



We'll have more when/if possible.






Ahhhhhh, new rumors!



http://www.macdailynews.com/comments.php?id=P2240_0_1_0
«13

Comments

  • Reply 1 of 43
    yevgenyyevgeny Posts: 1,148member
    Quote:

    Originally posted by tink

    Ahhhhhh, new rumors!



    http://www.macdailynews.com/comments.php?id=P2240_0_1_0




    Well I for one would like to see such a machine, but I doubt that it is coming. Quad G5s would just suck your bus dry.



    It seems odd that Apple would revive the quadra name since they have done nothing else along these lines with the G4's or G5. More likely, Quadra would be a tongue in cheek codename (kinda obvious what it is) and they would just call such a beast the "PowerMac G5 Workstation/small office heater".
  • Reply 2 of 43
    mellomello Posts: 555member
    Wouldn't it be more plausible to see a dual processor - dual core g5

    instead of an actual 4-processor ubersystem? And with that new

    hyperthreading technology (not sure if this is the correct name), it

    would act like 8 processors.



  • Reply 3 of 43
    jcgjcg Posts: 777member
    Quote:

    Originally posted by mello

    Wouldn't it be more plausible to see a dual processor - dual core g5

    instead of an actual 4-processor ubersystem? And with that new

    hyperthreading technology (not sure if this is the correct name), it

    would act like 8 processors.







    Dual Cores won't make it into the 970 line for at least one more revision. I think that most of the 980 romour s(now 975?) stated it would still be a single core chip. However they could build Quads Apple could build quad processor systems with the current line of chips from IBM.
  • Reply 4 of 43
    Quote:

    Originally posted by Yevgeny

    Well I for one would like to see such a machine, but I doubt that it is coming. Quad G5s would just suck your bus dry.



    It seems odd that Apple would revive the quadra name since they have done nothing else along these lines with the G4's or G5. More likely, Quadra would be a tongue in cheek codename (kinda obvious what it is) and they would just call such a beast the "PowerMac G5 Workstation/small office heater".




    The thought of it is sucking my 'bus' dry right now.
  • Reply 5 of 43
    Quote:

    Originally posted by Yevgeny

    Well I for one would like to see such a machine, but I doubt that it is coming. Quad G5s would just suck your bus dry.



    No it wouldn't, actually. Each 970 has its own point-to-point FSB connection to the system controller. The tough part would be building a system controller with 4 FSB ports, and a memory subsystem that has a hope of keeping up.



    Much more likely, in my opinion, is a dual core 9xx variant replacing each of the chips in the current machine. I don't put any stake in any of the processor rumours so far, but I don't think a dual core chip is impossible this year. Whether it shows up depends mainly on where IBM and Apple's priorities are.
  • Reply 6 of 43
    jubelumjubelum Posts: 4,490member
    ...and what would be the cost of FOUR processors ? Price points are right out the window with four chips. How much does a 970 cost (90nm) just for the processor....? Dual core sounds more plausible.
  • Reply 7 of 43
    Goooo so expensive!
  • Reply 8 of 43
    3.14163.1416 Posts: 120member
    Quote:

    Originally posted by Jubelum

    ...and what would be the cost of FOUR processors ? Price points are right out the window with four chips. How much does a 970 cost (90nm) just for the processor....?



    We don't know, but there have been several rumors that they're not much more than G4s. Even if a quad G5 cost $5000, it would sell.



    I don't believe this rumor, but I don't think processor price is the obstacle.
  • Reply 9 of 43
    jubelumjubelum Posts: 4,490member
    Quote:

    Originally posted by 3.1416

    Even if a quad G5 cost $5000, it would sell.

    .




    Well, I might get one for myself after begging my accountant, but my employees would just have to cope with dual 2ghz machines. Five grand is moving toward pro-only, and then only the prosperous pros. I don't think sales numbers would stay where they are if we saw PMs at more than $3499... the current price points are steep for most, but doable. I don't think there is much left to add and keep people buying top-end PowerMacs.



    Don't give Fred any ideas.
  • Reply 10 of 43
    mikemike Posts: 138member
    Quote:

    Originally posted by 3.1416

    We don't know, but there have been several rumors that they're not much more than G4s. Even if a quad G5 cost $5000, it would sell.



    Try $10,000 and up...most likely.
  • Reply 11 of 43
    onlookeronlooker Posts: 5,252member
    HIghly unlikely.
  • Reply 12 of 43
    kurtkurt Posts: 225member
    Quote:

    Originally posted by Programmer

    Whether it shows up depends mainly on where IBM and Apple's priorities are.



    My guess is portables.
  • Reply 13 of 43
    Quote:

    Originally posted by 3.1416

    We don't know, but there have been several rumors that they're not much more than G4s. Even if a quad G5 cost $5000, it would sell.



    I don't believe this rumor, but I don't think processor price is the obstacle.




    The 970FX is probably cheaper than the G4 -- it is roughly the same transistor count, but on a 90 nm process instead of Moto's poorly yielding 130-150 nm process. $250-500 x 2 added to the price of the current high end Mac. I suspect a lot of pros would jump on it. Biggest problem would likely be getting enough processors to make all those machines!
  • Reply 14 of 43
    amorphamorph Posts: 7,112member
    Quote:

    Originally posted by Programmer

    The 970FX is probably cheaper than the G4 -- it is roughly the same transistor count, but on a 90 nm process instead of Moto's poorly yielding 130-150 nm process.



    On 300mm wafers vs. 200mm.



    Quote:

    $250-500 x 2 added to the price of the current high end Mac. I suspect a lot of pros would jump on it. Biggest problem would likely be getting enough processors to make all those machines!



    I remember a thread where we hashed out how Apple could go from 2 to 4 to "8" processors without going into expensive NUMA architectures: First two CPUs. Then two dual-core CPUs (or two SMT CPUs). Then two dual-core SMT CPUs. And the basic design of the motherboard never has to change significantly. It's still two chips and two busses to the controller ASIC.



    We can get to the "4 CPU" stage this year, I think, and the "8 CPU" stage when IBM moves to 65nm in 12 months or so...
  • Reply 15 of 43
    Quote:

    Originally posted by Amorph

    On 300mm wafers vs. 200mm.







    I remember a thread where we hashed out how Apple could go from 2 to 4 to "8" processors without going into expensive NUMA architectures: First two CPUs. Then two dual-core CPUs (or two SMT CPUs). Then two dual-core SMT CPUs. And the basic design of the motherboard never has to change significantly. It's still two chips and two busses to the controller ASIC.



    We can get to the "4 CPU" stage this year, I think, and the "8 CPU" stage when IBM moves to 65nm in 12 months or so...




    sure, you could...but what would feed these beasts? the current implementation of the elastic bus is stuck at 1.1GHz (according to IBM). so how do you keep 4 logical processors, each running at 2+GHz fed on a single 1.1GHz bus? and beyond the bus, RAM, and most significantly, HD?! I seriously think the biggest advances made in "consumer" computing in the next couple of years will revolve around storage performance. the PC-DIY crowd has already seen the light of RAID. far better performance payoffs than OC'ing a P4. add to that advances in SATA command queueing, etc. i just think the industry is starting to realize the growing bottleneck created by mass storage. either mass storage needs to become significantly faster, or the systems-model we've been using for years needs to be revamped to better address the limitations of mass storage.
  • Reply 16 of 43
    wizard69wizard69 Posts: 12,695member
    I'm not so sure that feeding the beast will be all that difficult. The extra processor mean more cahce and chae that is used more effectively as you maintian a greater chance of previous code being in the cahces after a context switch. Now obviously not all programs are chewing through that much code but the same goes for data. Add a little intelligence to your scheduling algorithms and it may be possible to keep a process on a processor undestrubed by any context switches.



    In the case of logical CPU's that are the result of SMT you do realize that the secondary thread may only see 30% or 40% of the performance of the primary thread so there is less demand on band width than may be first implied. The threads are not equal. It is interesting to see what the linux world has done with their kernel to make use of SMT. When SMT first came out SMT was a fairly negative feature, now after a bit of work on the kernel we are starting to see real benefits. While not wanting to go into a great deal of details part of the problem was somewhat bandwidth related due to cache issues. Just having a scheduler that maintains processes on a specific processor greatly reduced the need for cach refills do to processes moving around the machine.



    So while I don't disagree that feeding the beasts could be a problem, it is not a problem that can not be dealt with. AS to advances for consumer computing I really believe that pervasive multi processing is the wave of the future. For some applicaitons it is the only way to grow future performance. For the very consumerish games market, SMP or similar multiprocessing arraingements are the only way for developers to extend their franchise in new directions.



    There isn't a single processor implementation from anybody (IBM, AMD, Intel or xyz), that will be available in the near future, that can have a significant impact on certian markets such as Games, medical research, science research, FEM and othe engineering applications. It will be a long time before the vast majority of the people are satisfied with the computer that sits on their desk. The reality is that software quickly out strips the capability of a machine, thus requiring faster processors. The only good thing is that the transisition to 64 bits gives us some breathing room.



    Dave





    Quote:

    Originally posted by concentricity

    sure, you could...but what would feed these beasts? the current implementation of the elastic bus is stuck at 1.1GHz (according to IBM). so how do you keep 4 logical processors, each running at 2+GHz fed on a single 1.1GHz bus? and beyond the bus, RAM, and most significantly, HD?! I seriously think the biggest advances made in "consumer" computing in the next couple of years will revolve around storage performance. the PC-DIY crowd has already seen the light of RAID. far better performance payoffs than OC'ing a P4. add to that advances in SATA command queueing, etc. i just think the industry is starting to realize the growing bottleneck created by mass storage. either mass storage needs to become significantly faster, or the systems-model we've been using for years needs to be revamped to better address the limitations of mass storage.



  • Reply 17 of 43
    Feeding the beasts will be an issue, but I don't think the FSB is the problem. In the current PowerMacs the combined FSB throughput is somewhere around 14 GB/sec, but the dual channel DDR400 memory system only achieves around 6 GB/sec (on a good day). That is a lot of improvement required in the memory subsystem before the FSB is again the bottleneck.



    IBM claims that their SMT implementation for the POWER5, on their test suite, runs 2 threads at full speed. Depending on the nature of the code, this is completely believable. Given how many threads are running on a typical MacOS X box you might find that SMT is quite a win. A few things don't SMT well (carefully coded AltiVec algorithms, for example), but IBM's SMT implementation has features to help ensure that this code gets a processor to itself.



    Apparently Darwin's kernel already does processor/thread affinity, so future SMT and shard cache dual cores will benefit from that.
  • Reply 18 of 43
    It won't be the first quad processor Mac however, DayStar came out with one years ago using four 604 processors:





    From Low End Mac, this description



    In more recent news Architosh had a Pro 3D user survey that found they "overwhelmingly want Quad Macs".



    With Apple's high end software ambitions (Shake, Logic, FCP) a quad workstation would be a logical extension of their line.
  • Reply 19 of 43
    onlookeronlooker Posts: 5,252member
    Quote:

    Originally posted by Aphelion

    It won't be the first quad processor Mac however, DayStar came out with one years ago using four 604 processors:





    From Low End Mac, this description



    In more recent news Architosh had a Pro 3D user survey that found they "overwhelmingly want Quad Macs".



    With Apple's high end software ambitions (Shake, Logic, FCP) a quad workstation would be a logical extension of their line.




    But that's daystar not Apple.



    Quote:

    In more recent news Architosh had a Pro 3D user survey that found they "overwhelmingly want Quad Macs".



    What would the point be of having a Quad processor Mac without a Pro 3D graphics card?

    I was one of the voters who voted yes to that survey, and I think it's safe to assume that the point I (and everybody else) was making when the votes were made was they (PowerMacs) still need to be faster. They are still overpowered by x86's built to be workstations. We need to be at least equal, or close in some areas, and faster in rendering, or something to have at least a semi competitive setup.

    Another thing is historically speaking the actual purpose of a quad processor computer is questionable as a workstation at best. Most Quad CPU systems are mainly for the server market, and how much performance increase can be expected from such a machine anyway? If Apple intended to build a workstation that was a superior 3D, and Rendering machine there would likely be a major motherboard revision to accomplish this, and it would still need the one thing it truly lacks the most. A pro 3D graphics card. I've already been talking my own ear off in another thread about this, but it's totally the missing link of an increase in future Macintosh sales. Where was the graphics card survey. If you want to do 3D that is what's truly needed because without it a quad processor system is pretty useless.
  • Reply 20 of 43
    Well, how about these then:



Sign In or Register to comment.