Messages in this thread | | | Date | Fri, 23 Apr 1999 11:44:13 -0400 | From | Aaron Tiensivu <> | Subject | Re: Re: bogomips ??? == Bogus MIPS |
| |
On Fri, Apr 23, 1999 at 06:03:00AM -0800, ddavis@carolina.rr.com wrote: > This is a question I was wondering about. Why do the AMD chips > usually return a double MHZ value? Are they really twice as > fast as an Intel chip when checking the loop you mentioned? > > Daniel
It's a false penis-size CPU measurement. The K6 has always done null-loops faster, and this is also why the K6 exposed a bug in Windows 95 which took a result of such a loop and divided a number by the result. Well once the K6 was hitting 350mhz and above, the result ended up being 0. Bam. Divide by zero in kernel-space == BSOD on every other boot (in general). Get into the 400mhz range and it would do it everytime. M$ had the balls to charge people for the fix, even though the bug was in their code and not the CPU.
I might have botched the explaination up above but that is the general case of what was going on.
My old AMD 5x86-133 has better bogomips (by far) than any of the original Pendulms but its performance is in the P90 range (unless you run it at 160/40 and it is closer to a P100)
The general rule is that K6-based cores will get 2x the bogomips of the actual Mhz rating and .25 micron P2's will get 1x the actual Mhz rating. My DEC Alpha gets almost 1-to-1 corrospondance too.
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |