falcodakrzz :
They shouldnt give up because intel would just become slouchy, there prices would become high and their technology wouldnt advance as much due to lack of competition. Although I don't see how AMD make any money in the cpu market, especialy of late, as no one would buy an AMD processor over an Intel one if they knew remotely what they were doing; Intel beats AMD in both performance and value.
Yeah this is very true. We all remember how slowly Intel moved on upgrading processor core tech prior to AMD putting out the Athlon, and then the K8 Athlon 64 cores. Actually I can think of a lot of people that would buy AMD processors that do know full well what they're doing, I am one. Everyone touts Intel processors OCing headroom on the newer lines. But, from what I remember, Intel procs have almost always been able to hit clocks up to 50% higher than the stock clocks. Even back in the old P2 days, the Deschutes based p2 300's and Celeron 300A's could easily hit 450Mhz. I don't think thats changed since.
turboflame :
There is a problem with simply passing the torch over to IBM or another large company. The x86 licensing.
AMD is our only hope (VIA doesn't count). Worst case scenario they will just fall back into the budget region or find a niche market.
Actually if AMD decided to leave the x86 market, they have the option to give the license to another company. At this point in time I think that'd most likely be IBM. If you take the time to read the cross license you'll be able to find the specifics of it. Besides, IBM already develops 2 or 3 different processor families as it is. But, both are RISC based architectures. If they obtained the x86 license from AMD they would be able to completely cross CISC(x86) and RISC based architectures to end up with some insane super processor(rampant speculation) which is one of the things Intel fears the most in the case of if IBM obtained the license from AMD.
jimmysmitty :
While it is great to see ATI doing what they used to do its not enough to keep AMDs CPU division alive.
I don't think a dual CPU setup like QuadFX will be the answer and making a dual core CPU now would be too costly. From what I heard AMD is taking K8 and adding some newer Phenom based features to it and releasing it as Phenom X2. So its technically K8.5 or maybe K9. But taking that even current Phenoms do not perform above C2Q I doubt that the K8.5 will be able to beat C2D easily and more than likely will not be able to compete with the Nehalem dual core.
Basically what Sailer said is make another K8. But that probably wont happen until K11 in reality.
I don't it would be too costly, they could basically just adapt the Turion Ultra to the desktop, they'd just need to get the clock speed up on it. It's just not cost effective for them at the moment due to only having one active fab. On average Phenom is about 8-9% slower clock for clock than Conroe. 10-15% slower than Penryn. So, if the claims of 10-15% performance gain due to tweaks in Deneb/Shanghai hold true, that could put Phenom in line with Penryn or at the least as good as Conroe. Even now if you tweak things properly while OCing you can get pretty close performance clock for clock with the b3's.
So no, I don't think they should leave the CPU market. What they need to do is start using better Cache memory technology. Perhaps ZRam that is only 1 transistor long, and very low latency, or the other new cache memory tech I forgot the name of that is 2 transistors long, and fairly low latency. That would allow them to either shrink die size, with decent amount of Cache, or have huge amounts of low latency cache.
One of the things that I read in the technology brief about Nehalem was that it uses a modular design. Basically if Intel wants to, they can chose how many core cpu can be plugged into the rest of the die, at this time 2 cores, 4 cores, and possibly 8 cores possible at release. AMD really needs to change to that kind of die design. Where the rest of the processor stays the same, but the number of cores can be changed on a wafer by wafer basis. They also need to improve their prefetch units.
Not many people stop to realize that the Nehalems cores are basically modified c2d cores. You also gotta really look back and think about the fact that AMD hasn't changed architectures majorly since the k7. K8 is just a modified k7 core with AMD64 support, and an on chip IMC, with improved clock speeds. All k10 is is a farther modified k8 core, made to work with a unified L3 cache, and new SSE instructions. The original k7 core wasn't made for high clocks speeds, which is why at the time it was excessively difficult to push them past 2.4ghz, it actually says a lot that they've managed to push that base architecture to 3.2ghz.
Something I've noticed over the past year or so, is that prior to Penryn's launch if you ever looked at the CPU charts and went through them you'd notice a few things. I took the time to go through all the bench numbers, and look at what processors consistently placed in the top 10. 6-7 of the processors in the top 10 on those charts were Intel core 2 extremes, which I'd expect. 2 of the other processors were the c2d e6850, and e6750. But, there was something else I also noticed, the Athlon X2 6400+ was almost consistently in there at number 8,9 or 10. And that was the top ten out of all processor that were listed on the charts at that time. So in the main stream arena that put the x2 6400+ into the range of #3 or 4 on the top 10, with the e6600 and x2 6000+ following close behind. But yet you never heard anyone give props for that, it was always, the intel proc to completely wipe the floor with the AMD proc, even though it was usually not by a large margin.