Intel *and* AMD?

I was reading through some HyperTransport specs recently, and the following struck me:

nVidia used HTT to link the North and South bridge chips in the nForce 4 Intel edition.

Gigabyte exploited this to create their 'Royal' mobo, having an Intel P4 connected to an nForce 4 Intel edition Northbridge, then instead of the normal southbridge placing an nForce 4 AMD edition chip below it, using the PCI-E lanes of each.

If you were to use a 200 series Opteron (whch therefore has 2 HTT busses), could you not create the same board, but with an Opteron 200 between the two nVidia chips, thereby creating a 'hybrid' AMD/Intel Asymmetric multiprocessing system?

This might require special drivers, or an OS rewrite, but how cool would that be, being able to use the advantages of both archetectures??

I know there is one intel board on the market that can have an AMD CPU added via a daughterboard, but if I remember correctly it didnt allow both to run at once...
9 answers Last reply
More about intel
  1. I understand what your getting at but I think your basic assumptions need to be revisited.

    HTT (Hyper Transport Technology) is exclusive to AMD. There's no information on the nVidia site to suggest that they lisenced HTT from AMD to use with their chipsets.

    While nVidia mobo chipsets are used for both Intel and AMD solutions the implementations are very different due to the different core architectures; with the most obvious difference being that AMD has an on die memory controller and Intel uses a Northbridge memory controller. The Gigabyte Royal motherboard use the CK804 as a Southrbidge. The CK804 has also been implemented on AMD mobos ;ile the Tyan S2895. But, while both mobos use the same chip, their purpose and implementation are very different. By viewing the mobo block diagrams of the Gigabyte Royal and the Tyan S2895 you can see the same chip, the CK804, is implemented but used to achieve a different result.

    The Opteron 2xx series has 3 HTT links that can be used depending on the chipset, implementatin, and mobo manufacturer. Some mobos already do what you suggest with an Opteron 2xx being between the two nVidai chips.And, an implementation as you suggest would not result in a hybrid because AMD has an on die memory controller and does not need a northbridge.

    If there is an Intel mobo that can in fact have an AMD cpu connect vie a daughterboard, please povide a link, name, or model of the motherboard. I'd really like to learn about it.

    good luck...
  2. The board
    Too bad it's by ECS. Check out the live demo button though. The list of supported chips is impresive.
  3. The chips would have to recognize each other as a DMA capable controller.
    In DMA mode, the other chip would need to be at rest (max 20% over idle), so it wouldn't be much change at all. Memory access for the opteron would also be a problem. Adding in the north bridge would slow things down for it.
  4. Asymmetric multiprocessing is not a good idea.

    Even Symmetric Multiprocessing with two CPUs that are different revisions / stepping can lead to very minor problems, let alone running two totally different CPUs as you are suggesting.

    The main gain of the Quad-Royal was 4 x SLI slots, and that would not scale well in games per dollar invested. It would also waste much electricity and likely lead to premature PSU failure. - The HyperTransport Consortium are not AMD exclusive by any means, and would welcome Intel with open arms.

    The only decent Video cards with multiple GPUs interconnected (on the same card mind you) are from: ; and none of us can get them. (They are also dated, but their design scales well for its purpose).

    AMD have had True SLI 2 x PCIe x16 for ages btw:
    The same chipset pair (nForce Pro 2200 + 2050) could be used to make 4 x PCIe x8 slots for quad-SLI... but quad-SLI is an ugly beast at best. (as of March 2006, but that may change in 2-3 years.... however I suspect quad-SLI will only use 2 slots when it becomes popular and use 2 GPUs per card for power reasons).
  5. Yeah, but have a quick look at the site, notice all the logos down the right side of the page. They want the Intel one there aswell.

    Sure it may have been invented by AMD, but is is certainly not exclusive to AMD.

    Just like the MMX extentions and current FSB / Northbridge system where 'invented' by Intel, they are not exclusive to Intel.

    One could argue Intel didn't 'invent' the FSB, and AMD didn't 'invent' HyperTransport though... they both 'borrowed' the technologies from others.

    HyperTransport is very similar to Sun, SGI, MIPs, Alpha (and others) designs that predated AMD64 Direct Connect Architecture by 10 years.
  6. Quote: - The HyperTransport Consortium are not AMD exclusive by any means, and would welcome Intel with open arms.

    Whewwww, I thought I was crazy, here is a blurb from AMD's site:

    HyperTransport technology was invented at AMD with contributions from industry partners and is managed and licensed by the HyperTransport Technology Consortium, a Texas non-profit corporation.

    Note, not invented BY :)

    It says.. at :roll:

    How do you know if there wasn't an Intel guy or former Intel guy there helping?

    :oops: . o O <(J/K!! J/K!!)
  7. Quoted from a THG article Here:

    Gigabyte's latest stroke of genius is the GA-8N SLI Quad Royal, and is a very high-end motherboard that powers an Intel socket 775 Pentium 4, Pentium D or Pentium Extreme Edition processor via an NVIDIA nForce4 SLI Intel Edition chipset. NVIDIA chose HyperTransport to link the north bridge and south bridge components of the Intel Edition chipset. Gigabyte saw an opportunity to drop the south bridge (called the MCP or Media Communications Processor) in favor of an nForce4 SLI chipset - the AMD version!

    I just think it would be an interesting concept, it looks to me like it would be at least possible with these components :)
  8. The fact that an Intel edition nForce 4 can sit above an AMD edition nForce 4 also inducates that the intel nForce 4's implimentation of HTT is such that it can act in place of an AMD CPU. The AMD nForce 4 in this mobo basically sees the P4+northbridge as an Athlon 64.

    So what I'm suggesting is effectively a Dual Opteron mobo with an nForce4 intel edition and a P4 cpu in place of one of the opterons. The Intel Northbridge controls the P4s memory etc and handles 'translation' between the P4 and Opteron....

    The Netburst archetecture has advantages over AMD64 in some areas, and it would be nice to gain these advantages by simply setting a thread's affinity to a certain core :)
  9. Wow! A lot of good info in this thread. Picked up a few tidbits along the way myself...just proves that an old dog is never too old.

    So what I'm suggesting is effectively a Dual Opteron mobo with an nForce4 intel edition and a P4 cpu in place of one of the opterons. The Intel Northbridge controls the P4s memory etc and handles 'translation' between the P4 and Opteron....

    Give some of the replies and the statement above, there is a clearer understanding of what your original post was about. All things being equal, what you suggest could be done, but I'm curious as to what the end result of such a mobo would be and what would be the purpose of making such a mobo, i.e.; gaming, server, workstation, etc. It seems like an interesting idea, almost like the P4 would be acting as a co-processor. Interesting indeed...

    Any hypothesis or prognosis of what the end result and purpose of such a mobo would be? What would the block diagram of such a mobo look like? Sounds like an idea worth exploring...
Ask a new question

Read More

CPUs AMD Intel Nvidia