Originally Posted by Quinton McLeod
This is false and based off ignorance.
You and ownage honestly think pipelines are the "determing" factor in processor speed. You're both wrong!
The reason why the K8 outperforms the Prescott and the Presler series of Pentium processors is because of memory effciency, core design and thermal output. The AMD64 features a more rebust architexture than the pentium series. The Integrated controller makes it easy for the processor to process data. The core design features way less bugs than the core design in the Pentium counterparts. Not only that, but the thermal output is much lower due to the fact that Intel pushed Netburst and it sucked.
Pipelines mean nothing. Intel wanted longer pipelines to boost higher frequencies than AMD, but AMD has always done more per clock.
The Conroe has no proof that it does this. You're living off the same hype Intel has brought with the Prescott and the Presler. Intel has done this before. This time isn't any more different.
If you want to look what makes a processor faster, look at the benchmarks. You cannot possibly determine a processors speeds based on specs alone.
The AMD Opteron 165 with 1mb L2 Cache will operate at the same speed of a 3800+ with 512 megs of cache. The cache is irrelevant in AMD processors due to the Integrated memory controller. This is why AMD processors like low latency memory.
I've never said it is the determining factor in performance - so you should probably refrain from taking an ignorant yet aggressive stance. I said that it is one
thing that historically, and currently, determines the speed of your processor. If you compare two architecturally identical AMD64 processors with the same cache - one will notice that the reason one will perform better than the other is purely due to the higher core frequency. This is fact.
And I also never said the memory controller, or other core efficiencies of the AMD64 weren't
factors to take into consideration - but they were not the key points I addressed when he asked his question. I made mention of other
things, but I simply did not go into detail in them.
NetBurst, if we are to get into ignorant opinions, didn't suck. It is agreed that, although in comparison to the efficiency of the AMD64 architecture - hyper-pipelining only excelled in its desired task, which was media. But because NetBurst wasn't designed to originally compete against such an architecture (As much as it was made to scale upward in core frequency) things such as doubling of it's cache often decreased performance in many ways as well - since NetBurst was not built with such high cache in mind. The AMD64 however, was a new and efficient architecture built as a successor to the XPs - but this is hardly the argument, you seemed to have some misconception as if I was insulting AMD in some way. With the Intel 9XX Preslers, we finally began to see the direction NetBurst was intended to move in - where the XE Presler was even catching up (But still falling behind) the AMD64 - not bad for NetBurst, considering it competed with the XP. However, problems such as leakage were to blame for high temperatures, which made scaling difficult. The NetBurst architecture also needed a great deal of energy, that did not help. And the leakage was an issue that required yet another die shrink - but with Pentium M and AMD64 showing such success, it was obvious that NetBurst's days were over in favour of more efficient alternatives.
Pipelines mean nothing? So then by that logic - the longer pipeline of the P4 should have made no difference - so a Prescott should have identical performance to a Northwood at the same frequency. Oh no? Oh I see, this is not the case. You have presented nothing new with your argument here - we all know that frequency is not a very good determining factor of performance when not comparing identical architectures.
Those were actual benchmarks - you don't think Intel understands that we're all waiting for the processor when it launches in 4 months time? That we'll all be watching the benchmarks closely? But even if you look at the Conroe's architecture, it makes perfect sense that it beats and AMD64. Of course we will see - but doubting it is a fanboyish notion and ignorant at best. They were just benchmarks - if you hate Intel so much, try not to lash out at those who feel that the superior architecture won the benchmarks.
And likewise, benchmarks are not always a good determining factor with performance - as we have seen, even in actual gaming benchmarks, where an AMD64 dominates - we would see top-end Pentium 4s coming out on top. This
is not a very good way to determine performance all the time - but benchmarks can be a good indicator of overall system performance.
You mean at 512KBs of cache, not megs. Regardless - where AMD64 is concerned, I don't believe I've seen evidence (Unlike the Pentium 4 NetBurst architecture) that would cause me to believe that the increased cache is useless and gains exponential latency with each addition. Extra cache can aid a processor - AMD64 or not, memory controller or not, it's still nice to have. No one was arguing as to whether or not the extra cache would aid the AMD64 more than it would a Pentium 4 (Which is yes and no).
I never said that pipelines are the
determining factor in speed - but in certain applications it is, since an instruction that has to go through 31 stages still has to go through more than double what an AMD64's instruction must go through. With a little simple math, it makes sense that the AMD64 instruction would execute faster - the point of the NetBurst architecture, and something it did very well, was media. Hyper-pipelining with help of Intels very good branch predictor would fill the pipeline allowing excellent and sustained execution of a greater number of instructions. However, because of leakage and inefficiencies in NetBurst when compared to AMD64 and the Pentium M - it was abandoned in favour of superior processing alternatives.
Just to add to the fact that you are truly inept - you told me I thought pipeline was everything, but if you had actually read my response before jumping the gun with your fanboy attitude - you would have noticed this was not the case.
Originally Posted by SSE5
There are other things - but those are all things you would look at in a processor. Frequency can be important for processors of the same series, cache is generally beneficial for all processors, and under a typical mentality, the more the better. And smaller dies are also better, since the processor then uses less energy and generates less heat.
The AMD has a shorter pipeline - that 2.6GHz AMD will be faster than the NetBurst Intel processor simply because of the efficiency of the architecture itself - and the speed at which instructions are executed due to the shorter pipeline. The same cannot be said for Intel's new Core architecture, which performs better than an AMD64 at a lower frequency - just as you see with that Intel and AMD. (The difference is less dramatic though, but still there)