will AMD introduce HT into their CPUs?

[OS]

or atleast some version of it?

 

[zmzhang]

Intel has the patient on HT. If AMD wants to use it, they would have to pay intel to put it into their chips. I hope AMD gets it soon, cause it would suck if intel has no competition and starts to raise prices.

 

[Electrode]

I wouldn't be suprised if we saw it, or some approximation of it, in a future AMD processor. Let's not forget that AMD did pay whatever fees Intel wanted for SSE instructions. Why wouldn't they do the same with HT? 🙂

 

[Adul]

they could go one step further and produce a dual proc core 😉

 

[imgod2u]

Simultaneous multi-threading is not an Intel patented technology. The way Intel implemented it is patented. So AMD or any other microprocessor manufacturer is free to implement their own version of SMT. Also, AMD and Intel have a cross-liscensing agreement that allows each to freely use the other's x86 extensions. So no, AMD didn't need to pay a fee for SSE, just like Intel wouldn't need to pay a fee for x86-64 if they chose to implement it.

 

[tart666]

funny you should ask. There's an article at digit-life.com today with some K9 speculations. Here's the babelfished rumours (just search for K9 in the page):
link

Edit: looks like K9 will work pretty much like a dual core system, executing two branches in parallel. Also, compression on cache and Hypertransport 2 are looking pretty promising.

 

[Superman9534]

AMD DID pay a fee for SSE, because they own the rights to SSE. Just like with Mac, motorola owns altivec and the only reason Apple hasn't switched to IBM is because they rely on Altivec on their PPCs. IBM can license it from Moto but the fee would be high for not much gain. But PPC and 64bit PPC are not owned by anyone.

 

[ChampionAtTufshop]

dont sun processors already use a form of "hyperthreading"?
(or maybe its not sun...another company that i cant remember at the moment 😉 )

 

[BD231]

Originally posted by: imgod2u
Simultaneous multi-threading is not an Intel patented technology. The way Intel implemented it is patented. So AMD or any other microprocessor manufacturer is free to implement their own version of SMT. Also, AMD and Intel have a cross-liscensing agreement that allows each to freely use the other's x86 extensions. So no, AMD didn't need to pay a fee for SSE, just like Intel wouldn't need to pay a fee for x86-64 if they chose to implement it.

Yeah 2u has it right, having extentions limited to an individual company could cause software support problems. Cross-Liscensing dodges any possible issues and is also the best way to create a standard for software developer's that would have use for the extentions IMO.

 

[Wingznut]

Originally posted by: Superman9534
AMD DID pay a fee for SSE, because they own the rights to SSE.

Well... Yes and no. From what I understand (and this is not from any "inside" information), AMD and Intel have a long standing cross-licensing agreement (which they just renewed last spring) which allows AMD to use Intel's microprocessor patents and vice-versa. I'm fairly certain that this agreement would involve some sort of royalties to Intel based on AMD sales.

So, I suppose you could say that AMD did pay Intel to use SSE, but not directly.

(As far as I know, anyway.)

 

[Rand]

I'm sure AMD will implement it at some point in time but I doubt it will be anytime within the next 2yrs or so.
If they had put much engineering resources into it I'm sure we would have heard by now, so it's a safe bet their not going to be able to implement it anytime soon. At the very least we almost certainly would have heard were it intended for inital ClawHammer/SledgeHammer processors, or the Barton.

Personally I think we will see single chip multiprocessing from AMD long before we see HyperThreading.
They have numerous patents covering it, and the very first rumours for the K8 initally seemed to indicate that it would be designed for it. While those rumours have long since been proven false it certainly does seem like an architecture very well suited to single chip multiprocessing, and the old K8-2 patents seemed geared towards it.

I wouldnt be at all suprised if we saw a dual core SledgeHammer chip for high-end servers when AMD switches to the .09u process.