A simple explanation of the “MHz aren’t MHz anymore” problem. –Ed
Maybe a little math will make this MHz situation clearer. If you think someone you know ready to buy a computer will be helped reading this, please feel free to print them out a copy.
People who use MHz to judge computers think MHz is MHz. So 3=3, 4=4, 5=5. Nothing simpler than that.
Actually, there is and always has been more to it than that. As we said earlier, performance isn’t
just Speed, it’s Speed X Work.
So when people were thinking 3=3 and 4=4, it really was 3X=3 and 4X=4, whether they knew “X” was there or not.
X Equals Work Done Per CPU Cycle, or more simply, X=Work.
The last few years, both Intel and AMD CPUs did about the same amount of work, so “X” was the same. Let’s say X (or Work) equals 1.
If “X” always equals 1, saying 3x=3 and 3=3 boils down to the same thing. You can ignore the X and come up with the same answer.
What Intel did with the PIV was drop the amount of Work (or value of “X”) from 1 down to roughly .75. The PIV is designed to do less work per MHz than the Athlon or PIII, but be able to do it more often.
In CPU design, you either:
Do more work less often or
Do less work more often.
So if you do the MHz math properly on a PIV, 4x=3, not 4. It’s 4 X 0.75 for the PIV, not 4 X 1 anymore.
However, people have been so drilled into thinking 4=4, not even knowing that Work was part of the equation, that when Intel says 4, they answer “4”, never dreaming that it’s really 3.
AMD has processors that work the old-fashioned way. Work still equals 1. In their case, 3x=3. Their 3x equals Intel’s 4x, but if you don’t know Work is there and that its value can be changed, it looks like AMD is trying to claim 3=4.
Since AMD doesn’t believe it can break most people of the habit they picked up saying 4=4 a million times, what they are saying with PR is:
“OK, if you insist Intel’s 4=4, even though it really equals 3, then to be accurate, when you look at our CPUs and compare them to Intel’s, we’re now saying 3x+1=4.”
Then people start screaming, “You changed the formula!” not realizing that it was Intel that changed the formula when they changed “X”.
AMD could have just as easily said “Just multiply Intel’s PIV Mhz number by .75” or “Take 25% off Intel’s number” and it would essentially work out to the same thing.
However, that would require people to do math, since Intel is certainly not going to do it for them on the computer sticker. As Barbie once said, “Math is tough.”
So PR is an attempt to do the math and make the adjustment for them.
Intel Messed Up Your Math, Not AMD
I’ve had people say, “How come AMD didn’t complain when it had the lead?”
That’s very simple. X always equalled “1” when you compared an Athlon to a PIII. So if AMD had a 4X TBird, and Intel had a 3X PIII, 4 X 1 is more than 3 X 1. No need for any adjustment there.
AMD could have done this when the initial PIVs came out, but nobody was buying PIVs, and they cost much more than the Athlons. So while this problem has been around awhile, it wasn’t hurting AMD.
Now Intel has put together cheap systems at prices that look very good compared to Athlon systems IF you think Intel’s 4x is more than AMD’s 3x. Actually, since these cheap SDRAM-based systems make the PIV even slower, for those systems, 4x doesn’t equal 4, or even 3, but something closer to 2.5.
So the average person buying a computer will do the old math drill and think he’s getting a “4” when he’s actually getting 2.5, and ignore the AMD “3” systems.
This is the problem, and this is why it’s important people realize what’s going on.