970MP heat info from ibm

Posted:
in Future Apple Hardware edited January 2014
Click here for info from IBM



it says higher frequency MP's will run hotter than any ibm chips before.
«13

Comments

  • Reply 1 of 46
    pbpb Posts: 4,255member
    A little late, I think . Link (check out today's posts on page 2).
  • Reply 2 of 46
    looks like Apple was definitely thinking ahead when it made the PM so frikin big
  • Reply 3 of 46
    webmailwebmail Posts: 639member
    DOCUMENT HAS BEEN DELETED. Perfect way to confirm Apple is into this ;-)
  • Reply 4 of 46
    onlookeronlooker Posts: 5,252member
    Quote:

    Originally posted by The General

    Click here for info from IBM



    it says higher frequency MP's will run hotter than any ibm chips before.




    Actually thats not what it said.



    It said:



    The higher frequency grade versions of the 970MP consume higher amounts of power than earlier IBM microprocessors do, and that can cause temperature issues.
  • Reply 5 of 46
    Yet the e600 based 8641D from Freescale is Dual Core, uses DDR2, has an integrated memory controller (and North Bridge for that matter), uses an external FSB of 667MHz and an internal speed of 3-4x that (meaning CPU frequency of up to 2.6MHz), is PCI-express and even has 4 1Gb network lines, only consumes 15W of power. Who wants a 970MP when you can get a pimped Freescale which will perform better, and is cooler.



    I now know what I want in an updated PowerBook. But only if it has the new 2 button mouse/trackpad which the other rumor sites are talking about.
  • Reply 6 of 46
    onlookeronlooker Posts: 5,252member
    What gives you the idea it will perform better? Have you ever seen either, or one of them?
  • Reply 7 of 46
    drboardrboar Posts: 477member
    Comparing Freescale CPUs and IBMs is pointless as long as shipping dates and numbers are unknown.



    Sure the e600 look to be e fantastic chip to ship now and use from the big tower down to the ibook., better than current G5 and IBM 970



    However, if it ships in the future it has to compete with future versions of various CPUs.



    The IBM 970MP will consume more power that is generate more heat then previous version. If that actually makes them run hotter depends on 3 factors. 1. The way they are cooled. 2. surface heat density, that is has the surface area of the chip increased as much as the heat output. 3. Local hotspots on the chip, hopefully IBM has used the revision work making the 970 a dual core to also work on hotspots and other stuff.8)
  • Reply 8 of 46
    thttht Posts: 5,437member
    Quote:

    Originally posted by smallstepforman

    uses an external FSB of 667MHz and an internal speed of 3-4x that (meaning CPU frequency of up to 2.6MHz)



    The 8641D fact sheet doesn't state this. What it does state is that it has an internal, integrated or on-chip MPX bus that operates up to 667 MHz real clock. That's just a little bit more bandwidth than the 7 month old 1.8 GHz iMac G5's external FSB. Wonder why they even bothered having dual DDR-II memory controllers.



    For external I/O, the 8641D will have a PCIe x8 channel, a choice between another PCIe x8 channel or a RapidIO channel, the aforementioned ethernet, and a slow 32 bit local bus.



    Quote:

    Who wants a 970MP when you can get a pimped Freescale which will perform better, and is cooler.



    I still don't have warm fuzzies that the 8641D will go into an Apple machine. It just has too much extraneous I/O, and what I/O it does have, isn't that appropriate for an Apple machine. So you may not have the choice. It's going to be the 7448 or the 970GX.



    There could be an interesting choice between a 1.6 GHz 970MP or a 2 GHz 970GX though.
  • Reply 9 of 46
    onlookeronlooker Posts: 5,252member
    THT isn't the only one that is knowledgeable of processors that think the freescale processor isn't a good choice for a Mac. People have been saying that the processor is pretty much unacceptable for a while, and given lengthily reasoning as to why. I'm surprised none of you read any of that previously. Probably the least inportant, but most interesting thing about it is why the are limited to 8X PCI-E? That is no different from AGP. It makes me wonder why they bothered.
  • Reply 10 of 46
    smalmsmalm Posts: 677member
    Quote:

    Originally posted by onlooker

    Probably the least inportant, but most interesting thing about it is why the are limited to 8X PCI-E? That is no different from AGP. It makes me wonder why they bothered.



    The SLI solution from Nvisia also uses 8 PCI-E lanes per card. Did anybody see a speedlimit therefore?



    BTW the internal MPX bus of the 8641D has 5300 MB/s peak bandwidth (in or out), the 970 in the 1.8 GHz iMac has 2100 MB/s (in) plus 2100 MB/s (out).
  • Reply 11 of 46
    onlookeronlooker Posts: 5,252member
    Quote:

    Originally posted by smalM

    The SLI solution from Nvisia also uses 8 PCI-E lanes per card. Did anybody see a speedlimit therefore?



    BTW the internal MPX bus of the 8641D has 5300 MB/s peak bandwidth (in or out), the 970 in the 1.8 GHz iMac has 2100 MB/s (in) plus 2100 MB/s (out).




    No. The SLI solution from intel uses 8. The SLI solution from Nvidia uses 16, The one for AMD provides 16 (both ways) for each card as Nvidia had it in their documentation.
  • Reply 12 of 46
    benroethigbenroethig Posts: 2,782member
    The 8641D really isn't meant for computers.
  • Reply 13 of 46
    thomazthomaz Posts: 42member
    SLi running on a amd runs each video card on a PCI E x8 or 1 video card on PCIE x16.THere no Sli on intel side other than 1 iwill mobo thats for the xeon.



    PCiEx16 is not slower than agp but not much faster at the moment but it allows the option of sli,muilt gpu's on 1 card,alot more bandwidth up to 4gigs per sec bi directial,There highier power than can be drawed from the bus and when using muilt gpu's it runs the bus at full speed unlike agp when muilt gpu's are on 1 card the agp bus becomes a 66mhz 32bit PCI bus with 266mb per sec bandwidth.



    Now with heat of the new 970mp it's goin to be alot hotter only if they don't use something like SOI with the 90nm die srink like Amd and if they go the root like intel with pressottt because the die srink causes alot more power leakage.
  • Reply 14 of 46
    wizard69wizard69 Posts: 13,377member
    It is pretty much true that the 8641D would not be ideal for some Apple hardware but this doesn't mean anything in and of itself. If it gets out the door ideal or not it might be compelling from a power standpoint.



    The other issue is that the e600 is a family of processors and Freescale has yet to announce a version to fill out its product line. Here I'm talking about their integrated system processors. Freescale could very well have something up its sleeve more suitable for PC use, that is e600 based. Evne at that I don't see the 8641, "D" or not being that bad of a processor for a number of Apple machines.



    Of course the is dependant on what IBM has up its sleeves with respect to PPE's and high integration devices. Apple has a huge need right now for a procssor that ups performance to a significant extent and lowers power demand. I think business will go to whomever can deliver that processor.





    Dave





    Quote:

    Originally posted by onlooker

    THT isn't the only one that is knowledgeable of processors that think the freescale processor isn't a good choice for a Mac. People have been saying that the processor is pretty much unacceptable for a while, and given lengthily reasoning as to why. I'm surprised none of you read any of that previously. Probably the least inportant, but most interesting thing about it is why the are limited to 8X PCI-E? That is no different from AGP. It makes me wonder why they bothered.



  • Reply 15 of 46
    eric_zeric_z Posts: 175member
    Quote:

    Originally posted by THT

    The 8641D fact sheet doesn't state this. What it does state is that it has an internal, integrated or on-chip MPX bus that operates up to 667 MHz real clock. That's just a little bit more bandwidth than the 7 month old 1.8 GHz iMac G5's external FSB. Wonder why they even bothered having dual DDR-II memory controllers.



    The latency will be way, way better though. The latency of a 167Mhz MPX bus is, iirc, better then that of a 1.25 Ghz Elastic bus for example.



    And the reson for DDR-2 is two fold, 1) to give the IO unrestricted mem access 2) Future proofing.



    Quote:

    For external I/O, the 8641D will have a PCIe x8 channel, a choice between another PCIe x8 channel or a RapidIO channel, the aforementioned ethernet, and a slow 32 bit local bus.



    Yep x8 PCI-e to the GFX and x4 PCI-e to the southbridge, and the local bus is used for the boot ROM for example. Any superflous Ethernet channels can most likely be shut down.
  • Reply 16 of 46
    onlookeronlooker Posts: 5,252member
    Quote:

    Originally posted by Thomaz

    SLi running on a amd runs each video card on a PCI E x8 or 1 video card on PCIE x16.THere no Sli on intel side other than 1 iwill mobo thats for the xeon.





    Again that is not true with the 2200, and the 2050 They do not split 1 16x PCI express lane, they use 2 individual lanes. 16x both ways.



    Quote:

    Quote1

    In current motherboard chipsets, Intel's Tumwater chipset and the nForce 4 SLI break a 16 PCI-Express lane into two 8 PCI-Express lanes. The use of nForce Pro 2200 and 2050 combo in workstation configuration will allow for dual 16 PCI-Express lanes supporting Quadro SLI when it is officially launched by nVIDIA.



    Quote:

    _

    Quote2



    NVIDIA nForce Professional MCPs for Workstations:

    The NVIDIA nForce Professional MCPs offer an advanced PCI Express design for high performance and provide a comprehensive feature set for both single-GPU and multi-GPU workstations. NVIDIA also works hand-in-hand with the industry's leading workstation ISVs, ensuring that NVIDIA solutions maximize productivity for MCAD, DCC, and film and video applications.



    Key Workstation Features:

    Multi-GPU capability with NVIDIA SLI technology: Supports two full x16 PCI Express slots for workstation graphics cards such as NVIDIA Quadro.



    Certified PCI Express platform for high-performance, multiprocessor workstations



    ISV certifications: NVIDIA's long-standing relationship with industry-leading ISVs ensures that your professional applications are tested and certified on NVIDIA solutions.



    An innovative feature-rich single-chip architecture integrates native Gigabit Ethernet with NVIDIA ActiveArmor? secure networking, SATA 3Gb/s with Native Command Queuing and RAID.



    The NVIDIA Unified Driver Architecture (UDA) lowers total cost of ownership by providing a stable software image across multiple generations of hardware.




  • Reply 17 of 46
    smalmsmalm Posts: 677member
    Quote:

    Originally posted by onlooker

    Again that is not true with the 2200, and the 2050 They do not split 1 16x PCI express lane, they use 2 individual lanes. 16x both ways.



    They use 1 PCI-E 16x for each controller.

    So you need a dual processor with dual controller system?



    GPU---Controller---CPU---CPU---Controller---GPU
  • Reply 18 of 46
    wmfwmf Posts: 1,164member
    Quote:

    Originally posted by smalM

    They use 1 PCI-E 16x for each controller.

    So you need a dual processor with dual controller system?



    GPU---Controller---CPU---CPU---Controller---GPU




    If you're going to get SLI, you might as well break the bank.



    Meanwhile over here in reality I think PCI-E 8x would be plenty for an iBook or PowerBook. And I hear that DDR2 is lower power than DDR.
  • Reply 19 of 46
    onlookeronlooker Posts: 5,252member
    Quote:

    Originally posted by smalM

    They use 1 PCI-E 16x for each controller.

    So you need a dual processor with dual controller system?



    GPU---Controller---CPU---CPU---Controller---GPU




    These are Dual Opteron Boards.
  • Reply 20 of 46
    onlookeronlooker Posts: 5,252member
    Quote:

    Originally posted by wmf

    If you're going to get SLI, you might as well break the bank.



    Meanwhile over here in reality I think PCI-E 8x would be plenty for an iBook or PowerBook. And I hear that DDR2 is lower power than DDR.




    If your only going with 8x why not stick with AGP?
Sign In or Register to comment.