Originally posted by: NitOxYs
After researching and reading at least 5 to 10 reviews I tend to agree wtih justly. Now we all know that Phenom had major scaling issues and TLB issues with its l3 cache. I hope that they fix alot of their bugs especially with memory benchmarks and as well as their cpu code. Alot of the memory benchies put Phenom way behind. Why? The CPU was pushed due to community bashing. They were forced to release a product that had not finished. This comes with consequences.
I don't see how it was rushed. The scheduled release for K10 has been on the roadmap for some time. Granted, AMD roadmaps aren't all that reliable, but if anything K10 was delayed rather than moved up.
Furthermore, if it's true that K10 was moved up and rushed, why is the new Bulldozer core so close at hand? Granted I don't think AMD will meet their projections with Bulldozer, but the fact that Phenom seems to have a short shelf life is pretty telling that this chip was delayed not rushed.
Originally posted by: NitOxYsYou're going to reply that it should have been out way way earlier with Phenom. The reason why Intel never did a 4 core processor monolithic processor is because they had trouble scaling it and ran into many difficulties facing meshing it on a 65nm die. There are alot of bugs to be worked out with Phenom and hopefully with revisions down the line they can compete whole heartedly.
Source for why Intel didn't do a monolithic processor?
Besides, while it might have seemed like a good idea at the time, it is never comendable to make an architectural decision if it doesn't pay off. By all accounts there is little performance difference between an MCM core and a monolithic one, so if AMD made the decision to wait a year for monolithic knowing it would give little performance benefit then they made an obvious mistake.
Originally posted by: NitOxYsI just want to vent about one thing though. How many of these applications are "optimized" for Intel?
Proof for this wild accusation?
Besides, "optimized for Intel" is a silly term in of itself. Cheats are one thing, but if a software vendor decides to take advantage of Intel's architecture by tuning for it, that's not unfair in the slightest. If it's a canned benchmark you would have a point, but if shipping software shows performance improvements (like how the DIVX encoder uses SSE4 to huge benefit) then it is entirely relevant to the performance picture.
Furthermore, most people who accuse third parties of Intel specific tuning point to simply the use of the ICC compiler. This is obviously silly since not only is it a widely used compiler, but it provides great benefits for Intel and AMD architectures alike. Intel has put a lot of money into their ICC compiler and it shows.
Originally posted by: NitOxYsThe B3 revision going to be released hopefully in later december early January will fix some of these bug issues.
New steppings are great for tuning, but they aren't going to change the performance picture measurably. I defy you to point to one specific instance where a stepping drastically changed the performance characteristics of a processor.
Furthermore, by the time B3 is out, Penryn hits at the low end which makes matters worse for AMD.
Originally posted by: NitOxYsFor me personally I wish they had worked on their Phenom(agena/barcelona) architecture for dual core processors first to compete with mid range and then latter released their quad core.
This would have been a mistake. Intel is already slashing prices on quads and will drive them under $200 and probably under $150 in Q1. If AMD had released dualies only, they'd be competing on the low end, not the midrange.
Originally posted by: NitOxYsI'm no fanboy, and I had always liked AMD and Nvidia combined and I like to see them suceeed. In all reality, if AMD failes then the FCC will scratch their heads. And maybe just maybe they can fix their issues with the HT Bus and other architectual issues.
The reason for AMD's problems have more to do with Intel's success than AMD's failures. Intel is firing on all cylinders and hit a grand slam with the Core architecture, a fact that I think caught not only AMD but pretty much everyone by surprise.
I think AMD can fix some of the issues with K10, but they have a bigger problem looming in the form of Nehalem. Penryn will rule the performance roost next year, but Intel is already publically demoing Nehalem and by all accounts it will be and incredible chip that will build on the success of Core. AMD doesn't have anything to fight this except for some big plans with Bulldozer that are little more than paper promises right now.
I certainly hope AMD gets back on its feet, but I think this may be a sign of things to come. Certainly no one expected AMD to succeed with Athlon and later K8, but now that Intel is paying attention, it's going to be harder and harder for AMD to compete with Intel's economics of scale.
It's sad in a way, I'd really like to see a pair of juggernauts in this space. AMD, although certainly possessing some talented and driven people, simply can't compete very well with the much larger Intel unless Intel screws up majorly. Sure Intel made a pair of mistakes which let AMD run away with performance for a while (scaling of the P3 and the clock wars design of Netburst) but I have a hard time believing this will happen again. I think the lead that Intel has now is only going to get bigger and AMD is going to be permanently relegated to the mid range if they are lucky and the low end if Intel continues to succeed the way they have.