Here is the problem. As a consumer, there needs to be a definitive measure which is applied by ALL makers. I want to be able to compare with some meaningful measure.
If Intel uses one standard and AMD another, comparison is intentionally impossible. (not unlike other industries where comparison shopping is made next to impossible by playing games.)
New Coke jokes aside. Superior archetecture means nothing. Might as well say this chip has more flux capacitors. If the chip is faster then the chip is faster, it can be measured and stated.
Same as any other speed based product.
There is a meaningful measure. Actually several. They are benchmarks from different software apps. What benchmark you pay attention to is based on what you're going to be using your pc for. However they are pretty much only available on PC hardware review sites. And when looking at them, most computer users would have their eyes roll back in their heads. Which leads us to the same problem as before.
The easiest thing for a consumer to do is just buy the fastest amd chip they can afford. That way you'll know you're getting the most power for the money.
If Intel uses one standard and AMD another, comparison is intentionally impossible. (not unlike other industries where comparison shopping is made next to impossible by playing games.) It's very difficult to produce meaningful benchmarks for processors, no matter how honest ones objectives might be, because many factors affect performance.
Different processors have different performance strengths, and most if not all have certain performance weaknesses. For example, a certain PowerPC could perform floating-point addition, subtraction, or multiplication twice as fast as a Pentium, but the Pentium was twice as fast at division. IIRC, the approximate time ratios were:
|
Add/Sub/Mul |
Div |
Pentium |
2 |
34 |
PowerPC |
1 |
66 |
If a program used a uniform mixture of add, subtract, multiply, and divide operations, the Pentium would leave the PowerPC in the dust (performing one of each operation would take 40 cycles on the Pentium or 69 on the PowerPC); if an algorithm could be written to eliminate division, however, the PowerPC would smoke the Pentium.
So which processor would one say was faster at floating-point math?
[BTW, how well do processors do floating-point division these days? I came up with a good method for fast division a few years ago, but it turned out someone else had beat me to it].