[NTLUG:Discuss] Why do people have to change the subject on a thread reply? -- WAS: Sun ships single Opteron with AGPx8 and 4 PCI-X busses! -- WAS: HDTV card for Linux

brad angelcyk bradangelcyk at gmail.com
Wed Sep 8 12:52:55 CDT 2004


"WAS: HDTV card for Linux"

Thank you for breaking the message thread display in Gmail... I was
very easily ignoring a thread that I had no interest in until you felt
the need to change the subject.




On Wed, 08 Sep 2004 13:38:14 -0400, Bryan J. Smith <b.j.smith at ieee.org> wrote:
> On Wed, 2004-09-08 at 04:23, Bryan J. Smith wrote:
> > Until PCI-Express comes out, I agree.  It's not worth the cost to get
> > 0.5GBps+ PCI/PCI-X.
> 
> Sorry to introduce a tangent, but a vendor has _finally_ shipped a
> single Opteron platform with _true_ multiple PCI busses starting under
> $2K for Socket-940 starting with the Opteron 144 (1.8GHz, 1MB L2) upto
> the latest 150 (2.4GHz, 1MB L2).  In all offerings prior, the _minimal_
> setup I've seen with the AMD8131 (dual-PCI-X) chip was a dual-Opteron
> mainboard starting at $450+ -- ouch!
> 
> Ironically, it's Sun Microsystems:
>   http://www.sun.com/desktop/workstation/w1100z/datasheet.pdf
> 
> - (4) PCI busses ... how?
> 
> The Sun Java Workstation W1100z has 4 PCI busses, which leaves me
> perplexed on what HyperTransport tunnel/bridge chips they are using.
> Normally the AMD8151+AMD8131/AMD8132+AMD8111 combination will only give
> you (1) AGP 3.0 (8x), (2) PCI-X 1.0/2.0 (8131/8132) and (1) PCI/Legacy
> PC.
> 
> Especially given the fact that there is (1) PCI-X 133MHz slot, and (2x2)
> PCI-X 100MHz slots -- that's (3) PCI-X channels right there, plus all
> the on-board peripherals (another PCI-X/Legacy PCI -- bridged?).
> Normally that wouldn't be possible even with the "maxed out"
> AMD8000-series configuration.  Furthermore, it's shipping with (5) USB
> 2.0 ports, and the AMD8111 only offers USB 1.1, so it's likely it's
> another chipset.
> 
> - Proof is in the architectural document:  dual-AMD8131 chips!
> 
> But then I checked the architectural document:
>   http://www.sun.com/desktop/workstation/w1100z/arch-wp.pdf
> 
> Looking at pages 9+ in the architecture guide -- sure enough --- it's
> the AMD8000 series.  How do they do it?  They put (2) AMD8131 chips
> on-board!  Wow!  That's _true_ (4) PCI-X channels -- _plus_ the (1)
> PCI/Legacy PC in the AMD8111 _and_ (1) AGP 3.0 (x8) slot.  Then they add
> their own IEEE1394, SerialATA, USB 2.0 and Ultra320 peripherial chips
> too.  Not bad for a $1,995 base system!  Wow!
> 
> Man, Sun's single processor Opteron _blows_away_ most _quad_ Opteron
> designs!  And it _equals_ the HP Proliant DL585 quad-Opteron which
> starts at $10,000+ (for just 2 CPUs).  And it's got all the "goodies" of
> the "latest'n greatest" systems.
> 
> - Modular HyperTransport connector = future upgrading!
> 
> It gets even better!  Looking through the pages, you'll note this
> "E-connector."  Looking closer, you'll note the mainboard is actually
> _two_ mainboards, in a single ATX form-factor.  What what's all that?
> 
> The "top" portion has (1) AMD8131 (dual-PCI-X) and (1) AMD8111
> (PCI+LegacyPC) with most of the peripherials (IEEE1394, USB 2.0,
> SerialATA) and (2) PCI-X slots (1 @ 133MHz, 1 @ 100MHz w/2nd master
> being used by the Broadcom GbE).
> 
> The "bottom" has the (1) AMD8151 (AGP 3.0) and (1) AMD8131 (dual-PCI-X)
> with (3) PCI-X slots (2 @ 100MHz, 1 @ 100MHz w/2nd master being used by
> the Ultra320).
> 
> The bottom can be _replaced_!  Yes, that's a full 32+32 at DDR800 (6.4GBps)
> HyperTransport channel that connects the two.  The idea is that when Sun
> wants to offer PCI-Express, it can!  Very cool!
> 
> This is the ultimate extension of the flexibility HyperTransport offers
> engineers.  Glad to see Sun take it as far as it can go!
> 
> --
> Bryan J. Smith                                  b.j.smith at ieee.org
> ------------------------------------------------------------------
> "Communities don't have rights. Only individuals in the community
>  have rights. ... That idea of community rights is firmly rooted
>  in the 'Communist Manifesto.'" -- Michael Badnarik
> 
> _______________________________________________
> https://ntlug.org/mailman/listinfo/discuss
>



More information about the Discuss mailing list