• Yokemate of Keyboards
    Yokemate of Keyboards
    Posts: 12097 from 2003/5/22
    From: Germany
    > Because the G5 is a completely different architecture altogether, with a
    > different set of system controllers, etc; you can't just clutch a G5 onto the
    > current system controller of the Pegasos.

    Oh please, don't try to teach me on things *I* explained to others before in this very thread. To cite myself from there: "of course the Articia S does also not support the PPC970, nor does the Discovery II of the Pegasos II".
    In my question to your denial I refered to the "dual processors" part, not the "G5 processors" part.

    > the Marvell Discovery II in the Pegasos 2 has limitations when it
    > comes to FSB speed that would take away most of the fun even
    > when using a *single* CPU at higher speeds, like a 7448 which is
    > supposed to go up to 200MHz FSB. [...] And add a second 7448 to it,
    > and sit back watching how your two CPU's take turns in sitting on
    > their ass waiting for data to transfer through the bus, memory controller, etc.

    I never disputed that. I merely disputed your denial of the fact that "the Pegasos [II] design supported" "dual processors". This has zilch to do with things like speed or fun. And again, I already wrote about the Discovery II's FSB speed limitation myself.

    > even if you would manage to actually get two CPU's working
    > with the Marvell in the Pegasos2 (which I doubt is possible)

    I'll cite my statement I already pointed you to: " The Discovery II of the Pegasos II however explicitly supports SMP in both 60x and MPX modes".
    Marvell are with me on this: "The Marvell Discovery II MV6436x devices [...] support Symmetrical Multi-Processing (SMP) in both 60x and MPX modes".
    I hope you don't mind I'll stick with their truth rather than your's.

    And just FYI an example of a board with Discovery II and dual PPC750GX (1.0 GHz, 200 MHz bus):


    >> You just said it's not.

    > Correct, there was a reason to why Genesi went for the Tundra
    > for their dual CPU design.

    That reason was not that "it's not" "technically possible to put 2 CPU's to the Marvell Discovery 2" as you claim but rather that the Tundra chip might be better suited for this and/or be cheaper than a newer Marvell chip. That's a difference.

    > they had to wait for the TSI108 that was supposed to be
    > capable of dual CPU

    No, the Tsi108 (announced in March 2005) was never supposed to be dual CPU capable, the Tsi109 (renamed from Tsi108A) was. That's why Genesi waited for the Tsi108A/109 to be announced in June 2005 before announcing their HDB.

    > then a second revision of the TSI108 that was supposed to
    > be capable of dual CPU, and then the TSI109 that *finally did*
    > become capable of dual CPU's.

    That "second revision" would be the Tsi108A, later *just renamed* to Tsi109. And it definitely had been dual CPU capable right from the start when named Tsi108A, even if being restricted to 167 MHz FSB clock in dual CPU mode, which didn't magically change by renaming. You once knew these things 3 years ago (see link above).

    > Looking at specs in marketing materials is one thing, but
    > the reality is often something else.

    You're implying Marvell have been lying regarding the SMP capability of their Discovery II?

    >> Right answer is already given here.

    > having a look at this old picture again. What you see in that picture
    > is the ancestor to the Pegasos 1 over at b-plan, it is an ArticiaS development
    > board with dual CPU sockets.

    Yes, I well know this one with ZIF sockets and Sonnet CPU card. This is, in addition to the existing dual G4 card for Teron I already refered to before in this thread, a further indication that the Articia S *might* be dual CPU capable. But as mentioned before, I very much doubt it as well as you for reasons I already explained.

    > promises in marketing materials does not always correspond
    > to what's delivered in reality, as the ArticiaS later showed in
    > several ways; [...]

    No news to me, which you'd know if you had *read* my linked posting instead just glancing.

    > SMP? Yeah, right.

    I'll cite myself: "I'm very much in doubt considering the technical problems bplan/Thendic France/Genesi faced in employing even just a single G4..."

    > "support" and "specifications" in marketing materials doesn't
    > by any means guarantee support and delivered specs in real life,
    > as there has been many examples of

    And this is as well true for the SMP capability Marvell claim for the Discovery II?
  • »26.02.09 - 03:58