[Beowulf] "dual" Quad solution from Tyan

[Karsten Petersen]

Mark Hahn schrieb:
>> Not quite, this would make no use of two HT channels.  Its like
>> this:
>>
>> c--c--c--c
>> |  |   \/|
>> |  |   /\|
>> c--c--c--c
>>
>> (The two leftmost CPUs need their third HT to connect the two
>> chipsets.)

> are you sure about that (ie, inspected the traces, or the HT
> routing tables)?

Well, as far as the traces are visible they show above connections.
And the Tyan documentation contains a drawing:
    ftp://ftp.tyan.com/manuals/m_vx50b4881_100.pdf
(Page 17 of 56)
But be aware, there are other Tyan docs which contain a wrong drawing. 
(With a fourth vertical link between nodes 5 and 6.  This is not 
possible with the limit of three HT links of an Opteron.)

I have not looked at the HT routing tables, how is this possible?
Do you know some program to display them?

> by your logic, for instance, a plain 4x would have one 
> cross-link (and I don't think they do.)

The plain 4x boards I know have a simple ring layout without diagonal 
links.  (Well, 8 of the 12 available HT links are already in use for the 
ring and one or two must go to the chipset(s)...)

Best wishes
Karsten Petersen
-- 
HPC System Engineer

Tel.: +49 3722 528 43
Fax:  +49 3722 528 15
E-Mail: karsten.petersen at megware.com

MEGWARE Computer GmbH
Vertrieb und Service
Nordstraße 19
09247 Chemnitz/Röhrsdorf
Germany
http://www.megware.com