Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

Serverworks today officially announced that it will be working on an Opteron
chipset.

http://www.eweek.com/article2/0,1759,1605020,00.asp

It's going to do a derivative chipset of its Xeon chipsets sharing some
functionality between the two chipsets.

Interestingly, an official at Broadcom said back in 2003 that it wasn't able
to do an Opteron chipset because of "IP firewalls" in place due to its Intel
agreements.

These days, I wonder if it wouldn't be AMD that would be concerned with IP
firewalls against Intel.

Yousuf Khan

--
Humans: contact me at ykhan at rogers dot com
Spambots: just reply to this email address ;-)

Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

>>>>> "Yousuf" == Yousuf Khan <bbbl67@ezrs.com> writes:


Yousuf> These days, I wonder if it wouldn't be AMD that would be
Yousuf> concerned with IP firewalls against Intel.

Nope no reason for AMD to worry because intel has take care of that.

Later

Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

Yousuf Khan wrote:

> Serverworks today officially announced that it will be working on an Opteron
> chipset.
>
> http://www.eweek.com/article2/0,1759,1605020,00.asp
>
> It's going to do a derivative chipset of its Xeon chipsets sharing some
> functionality between the two chipsets.
>
> Interestingly, an official at Broadcom said back in 2003 that it wasn't able
> to do an Opteron chipset because of "IP firewalls" in place due to its Intel
> agreements.
>
> These days, I wonder if it wouldn't be AMD that would be concerned with IP
> firewalls against Intel.
>

My usually very active imagination is failing me. Can you sketch out a
plausible scenario of what you see happening, even if it's not
necessarily what you think will happen?

My private oracle says that the incentives for Intel to pursue its
Intel-proprietary motherboard strategy are stronger than ever.

RM

Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

Robert Myers <rmyers1400@comcast.net> wrote:
> Yousuf Khan wrote:
>> Interestingly, an official at Broadcom said back in 2003 that it
>> wasn't able to do an Opteron chipset because of "IP firewalls" in
>> place due to its Intel agreements.
>>
>> These days, I wonder if it wouldn't be AMD that would be concerned
>> with IP firewalls against Intel.
>>
>
> My usually very active imagination is failing me. Can you sketch out
> a plausible scenario of what you see happening, even if it's not
> necessarily what you think will happen?

I really can't see how there could be any IP conflicts if Broadcom decided
to do Opteron chipsets and share some technology with its Xeon chipsets.
Xeon has got its own bus, Opteron has its own. The only things that would be
shared would be things like i/o southbridges (ATA controllers, Ethernet,
etc.). Can't see how Intel could block on things like that, since anybody
has the capability to make those things and they're not owned by Intel. If
anything, it's AMD that might have something to worry about, because chances
are high that Broadcom might start using Hypertransport to connect its
southbridges even in Intel motherboards. I'm not sure AMD would want its
Hypertransport being used to help Intel -- but then again AMD tends to be
less pedantic about its turf than Intel in every respect so maybe it won't
put up a fuss. Broadcom would obviously have to become (if it isn't already)
a member of the Hypertransport consortium so I doubt AMD could put up any
roadblocks to Broadcom to use Hypertransport as it sees fit. Of course, the
only thing AMD could control is the useage of Coherent Hypertransport.

> My private oracle says that the incentives for Intel to pursue its
> Intel-proprietary motherboard strategy are stronger than ever.

Intel is going to pursue this strategy more, and it is going to lock
Broadcom out more and more. There was a time when Intel needed Serverworks'
expertise in designing properly working chipsets, but now that it's got its
own set, there is no incentive for it help Serverworks anymore.

Broadcom's initial Opteron chipset will be a 4-way chipset. That means it's
really just another hanger-on for AMD. It's not going to start adding value
to AMD until it starts reaching for 16-way (or possibly 8-way), since the
Opteron does 4-way on its own right out of the box. Probably a good starting
point for Broadcom to gets its feet wet with Opteron, but nothing else.

Yousuf Khan

Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

On Wed, 02 Jun 2004 00:39:29 -0500, Post Replies Here Please <spamme@edge.net>
wrote:

>>>>>> "Yousuf" == Yousuf Khan <bbbl67@ezrs.com> writes:
>
>
> Yousuf> These days, I wonder if it wouldn't be AMD that would be
> Yousuf> concerned with IP firewalls against Intel.
>
>Nope no reason for AMD to worry because intel has take care of that.

Somewhat surprised that Yousuf didn't remember that part ;-)

Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

daytripper <day_trippr@REMOVEyahoo.com> wrote:
> On Wed, 02 Jun 2004 00:39:29 -0500, Post Replies Here Please
> <spamme@edge.net> wrote:
>
>>>>>>> "Yousuf" == Yousuf Khan <bbbl67@ezrs.com> writes:
>>
>>
>> Yousuf> These days, I wonder if it wouldn't be AMD that would be
>> Yousuf> concerned with IP firewalls against Intel.
>>
>> Nope no reason for AMD to worry because intel has take care of that.
>
> Somewhat surprised that Yousuf didn't remember that part ;-)

Actually, I didn't understand what he meant. Care to refresh my memory, if
as you say, I should already remember it? :-)

Yousuf Khan

Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

On Thu, 03 Jun 2004 02:40:58 GMT, "Yousuf Khan" <bbbl67@ezrs.com> wrote:

>daytripper <day_trippr@REMOVEyahoo.com> wrote:
>> On Wed, 02 Jun 2004 00:39:29 -0500, Post Replies Here Please
>> <spamme@edge.net> wrote:
>>
>>>>>>>> "Yousuf" == Yousuf Khan <bbbl67@ezrs.com> writes:
>>>
>>>
>>> Yousuf> These days, I wonder if it wouldn't be AMD that would be
>>> Yousuf> concerned with IP firewalls against Intel.
>>>
>>> Nope no reason for AMD to worry because intel has take care of that.
>>
>> Somewhat surprised that Yousuf didn't remember that part ;-)
>
>Actually, I didn't understand what he meant. Care to refresh my memory, if
>as you say, I should already remember it? :-)

Intel continues to decline to issue Broadcom/Serverworks any front side bus
licenses for everything faster/newer than the 533mhz fsb family. That pretty
much brought the whole dueling NDA scenario to a crashing halt long ago...

/daytripper

Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

Yousuf Khan wrote:

> Broadcom's initial Opteron chipset will be a 4-way chipset. That means it's
> really just another hanger-on for AMD. It's not going to start adding value
> to AMD until it starts reaching for 16-way (or possibly 8-way), since the
> Opteron does 4-way on its own right out of the box. Probably a good starting
> point for Broadcom to gets its feet wet with Opteron, but nothing else.
>

I don't really expect to see anybody make an 8-way board.
Dual core Opties are only a year away and I vaguely recall
seeing a road map for quad-core in 2007. Manufacturers
may be better off developing their 2P and 4P board and letting
the multi-core processors take care of giving 8P and 16P
equivalency.

And there is still lots of room for 4P board manufacturers to
differentiate themselves from their competitors by such
features as RAM capacity, networking, integrated SCSI,
integrated RAID, remote management/diagnostics, etc.

I think the next big problem to solve for 4P Opty board
manufacturers is to squeeze more than 4 DIMM sockets per
processor onto the motherboard - either that or pray for
huge price drops and speed increases for 4 GB DIMMs. HP's
4 way system has 8 DIMMs per processor, but they accomplished
that by putting each cpu and its DIMMs on a riser card - and
the DIMMs are limited to PC2100 even if there are only two per
processor.

Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

daytripper <day_trippr@REMOVEyahoo.com> wrote:
>>> Somewhat surprised that Yousuf didn't remember that part ;-)
>>
>> Actually, I didn't understand what he meant. Care to refresh my
>> memory, if as you say, I should already remember it? :-)
>
> Intel continues to decline to issue Broadcom/Serverworks any front
> side bus licenses for everything faster/newer than the 533mhz fsb
> family. That pretty much brought the whole dueling NDA scenario to a
> crashing halt long ago...

Ah yes, now I remember. Yet, even though Intel continues to decline licenses
to Serverworks for newer processors, perhaps Intel can continue to exercise
IP rights against Serverworks through their older processors? Anyways, I'm
still trying to figure out how there could've ever been any IP clashes with
AMD's older Alpha bus on the K7 Athlons, let alone what could possibly even
common at all between the newer K8 Hypertransport processors?

Yousuf Khan

Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

Rob Stow wrote:

> Yousuf Khan wrote:
>
>> Broadcom's initial Opteron chipset will be a 4-way chipset. That means
>> it's
>> really just another hanger-on for AMD. It's not going to start adding
>> value
>> to AMD until it starts reaching for 16-way (or possibly 8-way), since the
>> Opteron does 4-way on its own right out of the box. Probably a good
>> starting
>> point for Broadcom to gets its feet wet with Opteron, but nothing else.
>>
>
> I don't really expect to see anybody make an 8-way board.

And no sooner do I say that than all of the tech news sites
start spouting stories about 8-way systems being available
before the end of the year. Sure evidence that they are
reading everything I write and just waiting for opportunities
to show me up.

Articles include, among many others, ..
http://www.itworld.com/Comp/1361/040604opteronservers/

> Dual core Opties are only a year away and I vaguely recall
> seeing a road map for quad-core in 2007. Manufacturers
> may be better off developing their 2P and 4P board and letting
> the multi-core processors take care of giving 8P and 16P
> equivalency.
>
> And there is still lots of room for 4P board manufacturers to
> differentiate themselves from their competitors by such
> features as RAM capacity, networking, integrated SCSI,
> integrated RAID, remote management/diagnostics, etc.
>
> I think the next big problem to solve for 4P Opty board
> manufacturers is to squeeze more than 4 DIMM sockets per
> processor onto the motherboard - either that or pray for
> huge price drops and speed increases for 4 GB DIMMs. HP's
> 4 way system has 8 DIMMs per processor, but they accomplished
> that by putting each cpu and its DIMMs on a riser card - and
> the DIMMs are limited to PC2100 even if there are only two per
> processor.
>

Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

On Fri, 04 Jun 2004 10:28:03 -0600, Rob Stow <rob.stow@sasktel.net>
wrote:
>Rob Stow wrote:
>>
>> I don't really expect to see anybody make an 8-way board.
>
>And no sooner do I say that than all of the tech news sites
>start spouting stories about 8-way systems being available
>before the end of the year. Sure evidence that they are
>reading everything I write and just waiting for opportunities
>to show me up.
>
>Articles include, among many others, ..
>http://www.itworld.com/Comp/1361/040604opteronservers/

Hmm.. IWill is throwing out some interesting things for AMD Opteron
systems. The 8-way Opteron server isn't up yet, but I did just come
across this little interesting setup:

http://www.iwill.net/whats_new/SeeMore.asp?vID=83

What's interesting is that they are using a combination of the AMD
8131 PCI-X tunnel with nVidia's nForce3 250Gb chip. They've also
announced an updated version of this boards supporting PCI Express
along with PCI-X

http://www.iwill.net/whats_new/SeeMore.asp?vID=84


In any case, for that 8P Opteron server, I'll file that one in the
"I'll believe it when I see it" category. Not that I don't think such
a design is possible, and in fact it could be quite worthwhile, I'm
just not sure that there's enough of a market for a company to pursue
it heavily.

-------------
Tony Hill
hilla <underscore> 20 <at> yahoo <dot> ca

Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

Tony Hill <hilla_nospam_20@yahoo.ca> wrote:
> In any case, for that 8P Opteron server, I'll file that one in the
> "I'll believe it when I see it" category. Not that I don't think such
> a design is possible, and in fact it could be quite worthwhile, I'm
> just not sure that there's enough of a market for a company to pursue
> it heavily.

It's a 5U rack unit, consisting of two motherboards. So it's really two
4-way motherboards. Not sure what the interconnect is between the two
motherboards though.

Yousuf Khan

Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

Yousuf Khan wrote:

> Tony Hill <hilla_nospam_20@yahoo.ca> wrote:
>
>>In any case, for that 8P Opteron server, I'll file that one in the
>>"I'll believe it when I see it" category. Not that I don't think such
>>a design is possible, and in fact it could be quite worthwhile, I'm
>>just not sure that there's enough of a market for a company to pursue
>>it heavily.
>
>
> It's a 5U rack unit, consisting of two motherboards. So it's really two
> 4-way motherboards. Not sure what the interconnect is between the two
> motherboards though.
>

The interconnect is supposed to be HT.

And that is not the only product IWill is promising.
One of the many review/PR/announcement sites I read
suggested that they have 5 different 8 way systems
under developement.

One sounds like it is supposed to be an 8-way
version of the way HP made there 4 way systems:
each processor and its RAM on a card/blade, and then
said cards/blades inserted onto the motherboard
eight in a row at the front of the case. *That* is
the one I'd like to see.

Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (More info?)

Tony Hill wrote:

<snip>

>
> In any case, for that 8P Opteron server, I'll file that one in the
> "I'll believe it when I see it" category. Not that I don't think such
> a design is possible, and in fact it could be quite worthwhile, I'm
> just not sure that there's enough of a market for a company to pursue
> it heavily.
>

8P Opteron _server_ maybe, but I suspect that a hefty chunk of Opterons
being sold aren't being used as servers, anyway. Eight way glueless
cluster right out of the box? Sounds pretty good to me.

I haven't been following the Opteron drama closely, so it's possible I'm
missing something, but interconnect is _expensive_ and _clumsy_. Unless
the glueless interconnect is a dog, I'd suspect there'd be lots of buyers.

Twice as much glueless means half as many interface cards or none at
all. The pressure to fit problems into the eight processor space would
be nearly overwhelming, assuming the price is right, and (again)
assuming that "SUMA" really _is_ sufficiently uniform at 8-way.

RM