Intel to pay AMD $1.25bn

[Owenb]

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=3675

It seems AMD will have enough money in the bank to last until Fusion...

edit: Changed the link to a better article.

 

[VoG]

sensible move on Intels part, i'm surprised they didn't do it sooner tbh. they're getting so tied up in litigation with various trade/governmental enteties that if they can make there problems with AMD go away at a cost of $1.25bn, probably seems like a bargain to them.

 

[mdjmcnally]

Wonder if this means that they will drop that increase on the 58x0 cards now that had a cash injection.

 

[DragonQ]

Probably a good thing for both companies tbh.

 

[feeddagoat]

TBH this doesnt surprise me following their anti competition fine. cant really complain tho, thats great news for AMD. Just makes me wonder why intel went ahead with the suit in the first place. There obvously was a chance of successfully pulling the licence from AMD but the market is set up so that doesn't happen so one company doesn't get a monolopy. Guess you play withfire you get burnt.

 

[jaket]

Wow, might see some things to come from amd then.

 

[something_totally_different]

BUGGER. i searched for this, as i was organising my thread, youd got yours up with five replys *sigh*

 

[Trunks9486]

Makes you wonder how confident Intel feels about their own technologies...

I really wish that AMD will be more competitive in the future. Bulldozer needs to be a killer micro architecture and so does Fusion.

 

[Corasik]

The interesting part is that Intel/AMD are extending their cross licensing for another 5 years, its their existing cross licensing agreements that allowed intel to use AMD's 64 bit extension of the X86 processor, and its the same agreements that allow AMD to use intels SSE extensions.

Im sure AMD will enjoy the extra cash, but the benifits of the cross licensing benifit both companies, and end users, and Intel will no doubt be happy to know that AMD wont be throwing out more and more anti-competitive litigation etc. Im guessing that it could open the doors to Hyperthreading on AMD's processor line at some point too.

 

[AndrewN]

Nice analysis at AnandTech, seems it is good news for everybody apart from the lawyers but especially good news for AMD, they can keep going for a few more years now, and possibly (shock) make a profit

Cheaper processors ***

Edit: That's odd, I didn't know F(or) T(he) W(in) was swearing, learn something new every day.

 

[Hotwired]

WTF isn't censored and that certainly contains more naughtiness than reversing the letters.

Anyway MEGA win for AMD.

Unties them and hands them a pile of dosh at the same time, go forth and develop

 

[DrBombcrater]

Im guessing that it could open the doors to Hyperthreading on AMD's processor line at some point too.

Hyperthreading (aka SMT) is an old technology that existed long before Intel chose to use it, so AMD is free to use it if they want. But they've explicitly ruled out HT, their new Bulldozer design will use a very different system known as Clustered Multi-threading (CMT). CMT is almost the exact reverse of HT, it can execute a single thread on multiple cores while HT handles multiple threads on a single core.

 

[drunkenmaster]

Doesn't really untie them, frankly the licensing issue was just that Intel were worried the licences AMD were allowed to use, would now be intergrateable by anyone that uses GloFO as they could "help" out companies. THe new agreement just states that only AMD products have access to the licences that use stuff, and GloFo really won't be helping other companies design their products, basically the licencing situation now, is identical to before, just properly defined which with the split it wasn't quite as defined anymore.

The payment is big, and probably has an agreement in place that prevents AMD from sueing Intel pending the results of the New York states prosecution, which was a case being formed by the government(with a lotta help from AMD). But if Intel loses that case AMD have a huge case to them sue them for damages based on Intel being found guilty of anti-competitive practices.

However if the government wins that case, which looked likely, and AMD were able to sue for damages after, they'd probably get quite a bit more than the 1.25billion, saying that civil cases like that can be appealed until everyone involved has retired(almost ). So even if they got awarded 5 billion, 12 appeals and 10 years later they might finally get paid, maybe a reduced amount, maybe the get 2 billion, but 10 years later thats not as much money and they've spent 500million on lawyers over that time. This is easier as they'll get the cash within 30 days.

OTher than confidence in the company it doesn't make much difference to AMD right now, cash isn't a problem, they aren't in danger and they have enough backing to go till 2150, they have vastly more resources than INtel these days. They just need time to build fabs and build the business now, those are the only limits.

IN the future though, with a 3rd fab, a new architechture(which by all accounts as been annouced in the last couple days and appears, at first glance ridiculously good) the lack of anti competitive action, IE Dell and the likes not being paid to ignore AMD, and they can really start to push forwards and fight to gain market share. This second they don't have as strong a CPU to really push forwards, nor the manufacturing capability to even supply say 30% of the market.

But AMD have a very strong platform even now, gpu, cpu, mobo AMD are cheap, fast enough for basically any Dell base unit buyer, and it will help there. Though again, AMD are so competitive as a platform and gpu maker that Apple, Dell and lots of others are already taking them on now anyway.

For now the 1.25billion really just means their oil money partners(who own 45% of AMD) have 1.25billion less debt to pay off when its due, around 2011/2012. They have basically trillions between them, they've already spunked around 20billion in total on AMD, GloFO, the New York State fab and Chartered Semiconductors(another manufacturing group in Singapore that owns around 11% of the market itself). This they've done within a year, they'll probably invest another 5-6billion on new equipment for Chartered Semi con and another few billion upgrading Dresden to 32nm. 2billion or 3billion AMD debt, tbh, doesn't make the slightest jot of difference to these guys.

 

[Corasik]

Hyperthreading (aka SMT) is an old technology that existed long before Intel chose to use it, so AMD is free to use it if they want. But they've explicitly ruled out HT, their new Bulldozer design will use a very different system known as Clustered Multi-threading (CMT). CMT is almost the exact reverse of HT, it can execute a single thread on multiple cores while HT handles multiple threads on a single core.

No, Hyperthreading, SMT and CMT are all just the same thing, they can be implemented differently, but its still about getting more than 1 thread through a single core at the same time. Sun for example can push up to 8 threads through a single core on their SMT chip Niagra.

Reverse HT doesnt exist, and in all likelyhood will never exist, because single threaded apps only have a very limited ability to be executed out of order, 1 command relying on the results of the previous command.

BTW, CMT is normally considered to stand for Chip Multithreading, and intels Hyperthreading is a simple two thread implementation of it.

Intels HT for example simply allows unused execution units to be utilized by a second thread, but for a more efficient hyperthetical example, imagine if a 128 bit processor was designed, that could execute two 64bit instructions in a single execution unit at the same time, or 4x32bit commands in the same execution unit. That would allow massive levels of parallellism within a single core.

AMD have said implied they wont be simply copying intels HT implementation of SMT, fair enough, with a bit of luck they will come up with something even better , but I am willing to bet that it wont be reverse hyperthreading (Myths of RHT have been thrown about since long before Phenom).

 

[Alfie]

blimey, some of you really know your stuff!

 

[feeddagoat]

Hyperthreading (aka SMT) is an old technology that existed long before Intel chose to use it, so AMD is free to use it if they want. But they've explicitly ruled out HT, their new Bulldozer design will use a very different system known as Clustered Multi-threading (CMT). CMT is almost the exact reverse of HT, it can execute a single thread on multiple cores while HT handles multiple threads on a single core.

I read recently that microsoft was was looking into that. That would be a huge benifit to single threaded apps on quads but doesn't do much for core intensive work.

 

[drunkenmaster]

I read recently that microsoft was was looking into that. That would be a huge benifit to single threaded apps on quads but doesn't do much for core intensive work.

Bulldozer is an interger behemoth basically. Its got dual interger "cores" within each single Bulldozer core. Rather than needed HT enabled on Intel to see core cores within the os for each physical core, the OS will see two physical cores within each core. Intel basically "fake" a second core within a normal core, with very little hardware duplication so you just can't max out both threads, the only time HT offers improvements is when a thread is only half using a core so sticking another thread uses the available power better. Its a good way to improve efficiency, but if you have a thread that uses the available resources effectively, HT offers nothing.

Bulldozer has two sets of a 4 pipeline interger, with lots of duplicated hardware and I think separate L1 cache for both, so it can run two real threads, at all times, and ALWAYS gain more performance from it. its looking incredibly powerful, its saved die space by only having 2 FPU pipelines, they are 128bit so can run multiple smaller instructions and while theres normally one fpu pipeline per core, when ones not being used the other interger"core" can use the extra FPU pipeline.

Its smart because at some stage Bulldozers will have gpu's on die also, which will be FPU monsters anyway, so no need to waste excess die space with FPU pipelines.

The one downside is the first iteration will be a pretty damn huge 8 core cpu with immense interger power that likely Intel can't match(it will, without question spank Nehalem on Interger, which is odd for AMD as they normally build strong FPU cpu's), but will likely have a slightly weak FPU, because neither company will be able to fit a GPU in their 8/16 core products till another die shrink, when thats added, the fpu power will go through the roof(for both companies) but till then FPU might be a little underpowered. Though by then we could see a lot more offloading of FPU work to discrete GPU's anyway.

 

[Darkmatters]

You guys do indeed know a lot of stuff

 

[Shakeel]

good news for amd

 

[Owenb]

good news for amd

It's good news for both AMD and Intel, only time will tell if it's good for the consumer. This money gives AMD ~18-24 months to become competitive. If they haven't done it by then then I can't see them doing it- at least as an indepedant company. They have got a cash injection from the sale of their fabs, and now this $1.25bn. 24 months from now they should have release of everything that they have spent the last few years working on: the replacement to Phenom, Fusion, their new mobile chips and ultra mobile- possible for phones etc. Not all of these need to be run-away success stories, but at least 1 of them needs to a cash cow. Unless they start making serious cash they just won't be able to fund the R&D that is needed to compete with Intel/ARM clients. A weak AMD is much worse than no AMD as if AMD were forced out of the market, or bought out, then someone with the financial muscle, like the juggernaut that is Samsung, could step in and take their place.