Intel vs AMD debates are kind of like preaching religion to athiests - nobody is really listening to what you say. It's a lot like politics, most people already know what political party they are going to vote for and no matter what anyone tells you you won't change your mind. That said, I'm bored and I feel like speaking my mind on this subject.
I think the community here is generally "mature" enough to have a rational discussion on this topic. (It will also give me something to read tommorow, considering I have to work on Saturdays for the next few months.
)
First, let's talk about benchmarks:
I have built and worked on a LOT of Intel systems and have built and worked on a LOT of AMD systems. As far as performance goes, reviews on hardware sites are simply not comprehensive enough and are often biased one way or the other. I can run benchmarks that run twice as fast on a P4 than on the same rating Athlon, just as easily as I can run benchmarks that are the exact opposite. By very selectively picking and choosing which applications I want to run I can make either Intel or AMD CPUs look about as attractive as a 1980's vintage Yugo.
An example would be Lightware 7.5 tests - a few of the often benchmarked filters are very heavily SSE 2 optimized and the P4 dominates the Athlon XP in those tests since the Athlon XP does not support SSE 2 instructions. By the same token, running a benchmark like Sciencemark or a few x87 FPU stressing Spec benchmarks can highlight weaknesses in the P4s FPU or purposely create extremely branch heavy code that will favor the Athlon. (For even more fun, look at Linux application benchmarks. Why? Because Linux applications are usually compiled (and thus optimized to run and take advantage of features of each CPU architecture) on the hardware in the machine. What's interesting with these tests is that generally the scores of similarly equipped P4 and Athlon systems are within +/- 1% of each other.) I feel a review that only lists around 5 CPU benchmarks are proclaims that one CPU "dominates" another is fundamentally flawed.
Why?
If you look at what most people do with their systems you will find that hardly anyone stresses their systems hard enough and rarely runs ANY software where there is a noticable difference in performance between even something as "slow" as an 1800 MHz system and a 3000 MHz system. What software packages do most consumers run on their computers? I think it's safe to say that most people who read this forum spend the majority of their time on a computer either playing games or browsing the web.
If anyone ever tries to sell you a new computer claming that upgrading your CPU will do anything to speed up your Internet browsing, alarm bells should be going off in your head. The fact of the matter is that any system built within the last 4 years or so is going to be limited mainly by your Internet connection. Even something as "old" as a P2 300 MHz system is more than fast enough for web browsing.
What about gaming? I think most of you know that gaming performance is almost completely determined by video card performance, given sufficient CPU horsepower. (Which is usually anything around 2 GHz class or faster with the current fastest consumer level video card available, the Radeon 9800 Pro.) If you ever look at gaming benchmarks you'll probably notice that most tests are run at very low resolutions with montior vertical sync disabled. If you do this and run a test like Quake 3 at 640x480 at 16 bit color you'll probably find that a top end P4 or Athlon XP system can score between 350 and 450 frames per second with the fastest graphics card available. I don't know about the rest of you, but I only really look at the tests run at 1024x768 or higher with all the goodies like anti-ailiasing and anistropic filtering all enabled. Generally you'll find that all tested systems score within 10 FPS of each other because the graphics card has become the limiting factor. Often, you'll see fractions of a FPS seperating one machine from another. Even then, you'll find that P4 based systems edge out the Athlon XP based systems in one test, and the Athlon XP based system edges out the P4 based system on another. It makes you wonder about a few things: How many serious gamers only play one game where one CPU is much faster than another? How many of you can afford to buy a Radeon 9700 or 9800 Pro that can take advantage of a newer CPU? How many people with 9800 Pros run at 640x480 16 bit color with vsync off?
What else do people do with their computers?
1. Write CDs. CPU speed has almost no influence on this. Any system from around a P2 300 MHz machine can burn a CD just as fast as a P4 3.06 GHz machine, given the same drives.
2. Office applications. In the "real world" these tests are almost totally influenced by hard disk performance. Take a P3 1 GHz system and throw a pair of Seagate 15000 RPM LVD SCSI drives in a Raid 0 configuration and test it against a P4 3 GHz machine with a single 5400 rpm IDE drive and guess which will win.
In fact, if you look at the tests run in benchmarks like Sysmark 2001 and 2002 you'll find they usually run around 12 tests spread around 4 applications or so and run those same tests over and over to come up with a score. The office tests generally sort a spreadsheet over and over again, where the content creation tests usually run a couple of photoshop filters 3 or 4 times in a row to generate a score. Nobody does that on a system; why would you run a "blur" filter on a photograph 3 times in a row or sort a huge spreadsheet, undo, and repeat 3 more times? (For fun, look at the scores of AMD and Intel systems on Sysmark 2001 and 2002 and take notice of the differences. Then look at the tests run to generate those scores. You'll soon discover just how useless and biased these tests are. The office portion of Sysmark specifically runs tests that generally run faster on Athlon systems. The content creation portion was changed between 2001 and 2002 versions to run filter tests in Photoshop that favor the P4. The combined score in 2001 shows Athlons (and P3s coincidentally) beating the P4s. The 2002 test scores are the opposite. If the benchmark runs the same applications, why would the scores change so much from one year to the next? Both of these benchmarks are so flawed the scores they generate are all but useless.)
Very few people (and even fewer home users or gamers) do things like continuous video editing, serious 3d rendering, work with large SQL databases, or run enginering/scientific applications. In reality, it is only in these types of applications that you see real noticable differences between the CPU types. I think it's safe to say that the people who run these tests already have their mind made up if they are buying an Intel or AMD based system. If you do video editing as a profession, you want a dual Xeon or one of AMDs new Opteron based systems (if you can afford to pay $750 per CPU and 2 GBs of ECC ram!). With a single CPU you want a Hyperthreaded P4. If you do 3d rendering, you need to look at what applications you are running to decide which CPU to use. SQL database applications and most engineering/scientific software is faster on AMD Athlons and Opterons. These are the only applications where one CPU type is far superior compared to the other.
What is really needed is a universal benchmark that runs 100s of tests and compares the results to a reference platform. It's only in that matter that the benchmark can truely reflect realistic performance in a way that matters to the home user. The problem with this is that such a benchmark will end up showing that neither AMD or Intel CPUs are greatly superior to that of their competitor. Neither AMD or Intel will want to support or fund such an orgainization because it would be detrimental to sales.
I have one last point to make. In general, a new CPU architecture comes out of AMD or Intel about every 3 to 5 years. What's interesting about this is that many of the engineers who design CPUs are experts in one particular area of CPU design or went to the same schools as one another. Some of these people have worked for multiple companies or even move from Intel to AMD or vise versa when a new architecture is being designed. I'm sure that most of you don't know that the Athlon and much of the new Opteron and Athlon 64 were based on a server processor known as the Alpha, which was designed and produced by Digital Equipment Co. (DEC). (The Athlon even supports some of the Alpha's memory addressing modes and uses it's EV6 bus protocol!) AMD now has many of the engineers from this company working for them and Intel, Compaq/HP have many others and the rights to much of the Alpha design itself. HP and Intel are co-developing the Itanium. It's a pretty safe bet that some engineers who worked on the Alpha and/or Athlon are now working at Intel on the P4 or Itanium.
Here's my take on the whole situation: Buy what you want. Ask for recommendations and listen to advise from those whom you trust which motherboards, video cards, etc are best suited for which ever CPU you want to run. Both platforms are mature enough at this point that either an AMD or an Intel system can be equally stable. What primarily differentiates them at this point is features of the chipsets which support them and the price of the components needed.
What do you guys think? (I need something to read tommorow!
)
