Quote:
Originally posted by drdingo21
not all thats is true. becnhmarks are the only way to truly see how a cpu forms. The benchmark programs used are not biased one way or another. sisoft sandra for instances measure what the cpu can do MIPS, and MFLOPS. Most of the time the cpu performance won't matter in the real word but a lot of time it will, for instance compliing a kernel or using bryce to make a picture the cpu will shine. The xp1600 runs at 1400 mhz which is faster that a pentium 1.6 and just barly slower than a 1.7. As a matter of face amd did that the rating system to prove that mhz don't mean anything its how well the cpu performs and to kinda mock pentium.
|
Benchmarks are the best current way to compare processors to one another, but they're not the only way to truly see how a cpu performs. There are many other ways to test the performance of a processor than a benchmark. The only reason why I say benchmarks are the best current way is because they're considered industry standard, thus they give a somewhat level playing field.
However there are ways to optimize hardware a little for benchmarks, optimizations that may not translate at all to real-world performance. There are biased benchmarks, as well. Please don't try to claim otherwise. Sisoft sandra is a decent benchmark, as are many others. But there are plenty of bad benchmarks.
Another note on benchmarks....don't trust the numbers from companies themselves, only trust them from truly independent 3rd parties.....and make sure the results are replicated and tests are conducted fairly. Sometimes people will benchmark one processor using, say, SDR RAM while another processor will use DDR. Clearly they are not comparable systems even though the clock rates may be comparable......the numbers are easily manipulated to fool people who don't understand anything about computing performance.
Compiling a kernel or using Bryce depend greatly upon code optimization for the processor. For instance whether Bryce uses MMX or any other extensions that a processor may use can vary your results greatly. Also bear in mind that it's all FP, some processors are designed to have better FX performance than FP. Compilation will use FX rather than FP, but again, a compiler can be optimized for a processor to greatly speed things up. A processor will only shine brightest when the code is optimized for it. Whether processor X from AMD outruns processor Y from Intel depends on the optimization and the processors themselves. While one may seem faster than the other, it may not be a level playing field.
Cyrix used the P ratings back in the day before they went *poof*. One of their biggest problems was that their P rating was really based upon fixed point performance, not floating poing. Cyrix had some of the worst FPUs in industry, but they had a pretty good FXU. So a P200+ (supposedly faster than a P200) was only in integer math, doing things such as Office type work (excluding Excel). Toss in some MP3's, or some games, and performance was far below a P133. Of course the Cyrix ran at 150. AMD has played the megahertz war game....it's funny how both companies are trying to get away from the whole megahertz thing these days. AMD because megahertz is meaningless, and Intel because of their P4.
Of course in all this I am merely arguing the fact that there is much to consider in the performance department of a processor or system. Whether the P4 is
capable of being faster than an XP of the same rating or frequency, I don't know. But just because you don't experience the speed, or a benchmark tells you it's not as fast, doesn't mean it's not capable of blowing the socks off of the AMD.
One thing's for certain, the P4 will be capable of achieving much higher frequencies at a faster pace than the AMD processors will. The chip was really designed to scale high, I believe Intel claimed 10 GHz by the end of its lifetime. We'll see if they reach it since it'll depend greatly on their ability to make the technology work for it, but if they can get their technology working then the processor has a long lifetime with high clock rates. Having a high clock rate may not be the biggest deal to everyone, but it translates to more bucks for Intel since they won't pay nearly as much on development than AMD will. Time will tell, though.
The days of the desktop are limited, though. Until AMD starts releasing good mobile and server solutions they won't be able to stay in the game for too long. I'm curious to see how the new Hammer stuff goes for them. The future is in mobile and low-end servers. Why bother having a super powerful desktop that can give you 1000 fps in quake (or whatever games the kiddies are playing these days) when you could have a slightly less powerful mobile solution where you can play your games anywhere you want? Woo the things the future will bring us.....if we don't nuke ourselves by then.
I write too much.
b