Penryn versus K8L (until Yorkfield and Bloomfield appear).
The last processor using the Core microarchitecture will be Penryn. It will also be the first processor to use a 45nm process.
We all know what that means when Intel shrinks its core:
Higher clocks are possible
Penryn will appear on the market when AMD releases its K8L microarchitecture based line of processors.
Remember, the transition from K8 to K8L is much like Intel's move from Netburst to Core.
Let's say that K8L proves to be more efficient (more work done per clock cycle) than Core.
Do you think that the shrink to 45nm will be enough for Intel to keep the performance crown (by aiming for higher clocks [since 4GHz is easily reached with a Penryn]) till it releases its Nehalem microarchitecture in 2008?
Tough question... on most hardware, yes. Intell will keep it (untill AMD drops to 45nm, which I think I read will occure before Nehalem). I know K8L is supposed to have 128 wide Float unit (I can't find what Core runs at, and I don't remember ever seeing it), and that L3 sounds like they may have finally got it right, but other than that it seems they're playing catch up on almost everything. I don't think they can squeeze the advantage out like they did on the Athlons vs. P4's at this point.
Personaly, I think AMD has admitted defeat for this round and the next 2 or so and is biding they're time. They used to be all about beating Intel at they're own game. Now looking at the end of 2008 thru 2010 for them, it looks like they want to get everyone else to help them beat intel. A good chunk of their plans revolve around working differently with other pieces of the Hardware. It's almost like they're taking a bunch of ideas from cell or something. I'll admit, I'm confused by a lot of what they are trying to do. Not all of it makes sense, but that's how I've read the very few articals on the matter.
It's almost like they're saying fuck clock cycles, and arn't even trying for it anymore.
thanks to Silverwmoon!
Well Intel's last attempt before admitting their defeat to K8 was the Gallatin (Pentium Extreme Edition for the mass) with L3 cache which was a real disaster.
About 15 months ago they released the 65nm processors (Cedar Mill and Presler) to replace the Prescott and Smithfield but it it wasn't enough to challenge AMD's K8. However, they were the ultimate overclocking processors. Heck, my Presler was happily running at 4.8GHz till it died a few months ago.
Any predictions? How good will K8L be? Will Penryn cause enough of a stir among enthusiasts (potentially great overclockers)?
Well now your asking a different question. Your asking what will sell better as opposed to what will perform better.
That question I give hands down to Intel. Like you said, at 45nm, it's an overclocker's wet dream. Besides that, the Intel's look far better on a spec sheet. I doubt many people are going to see which one's can crank out the performance better or look at the flops.
As far as K8L and the L3 goes I don't think it will be that great. Honestly, I see it more as a stepping stone. AMD litteraly HAS to get L3 to work. If they can't even make a connection between between two or 4 cores work, then how will they ever get their ideas of multiprocessor heaven going. Their xPU plans aside, they plan on making extremely efficient laptop platforms that specialize in using hybrid hard disks, and a combination of dedicated and integrated GPUs. Another artical I read said they wanted to make dedicated processors to video decoding as well. We know such tasks can be done, but I can't think of an instance where it's been done well yet on the PC platform.
To me, K8L is a proof of concept. I realize in processor design and bus design for multi-processors are very different animals, but this still looks like an important step to me.
(as a side note, I'll admit I'm quite the AMD fan. However I think they are fucking up. Ever since they bought ATI, it just seems like they are slowing right down. Although some of their ideas are pretty good for the future, such as xPU, I'm not sure I really like them. I'd rather they do one thing, make good processors, and do it right. Don't worry so much about audio, video, or storage. Just do what your good at. There ideas are... very un-unix. )
thanks to Silverwmoon!