Ten, you have very funny sense of numbers. Look:
"I already said that 840 is going to be overkill for "real business productivity applications." I already said that running standard business apps on 840 was like trying to put out a candle with a firehose."
Ok, if you mean here to use a fire hydrant on
a small object like a candle, it will fly out
much faster than that miserable 2.5% speedup.
And what exactly does your "overkill" mean? How it
comes that a super-duper 3.2GB/s memory suddenly
stops working on ordinary applications?
"Did you forget that the SYSMark98 is a combination of the scores from all the apps tested? Some apps in that benchmark are going to benefit more from memory bandwidth than others. For example, it's obvious that MS Word, MS Powerpoint, and Netscape isn't going to benefit at all from more memory bandwidth. I'm sure there are several other apps which aren't going to care as well. Yet in the final SYSMark score, these apps are going to HIDE the real improvements seen by the apps which do care about memory bandwidth."
Oh, thank you for lecturing on how averages work!
However what has escaped you attention is that
we are talking here about FOUR TIMES INCREASE
in bandwidth yet your average makes only 2.5%
gain!
Hello-o: 400% of investments returns only 2.5%!
Very wise! If your "graduate theory" is correct,
there must be applications that are severely
DEGRADED in performance due to those improvements
to memory subsystem... which is highly unlikely.
And how it is obvious to you that MS Word, PPoint
and Netscape do not benefit from better memory?
Do they load their codes out of air? Don't they
paint windows walking through enormous
linked lists of numerous Windows data structures
(or whatever)?
"..then why is AMD rumored to be working on a chipset with dual DDR channels?"
Maybe because it is rumored? :)
"So since you wrote such a scathing response to my "undergraduate ramblings," perhaps you'd also like to write a similar scathing response to your "heroes" at AMD working on that chipset, no? "
Even "experts" do mistakes, don't they? No need to
go far for an example - rambust or no bust?
"After all, it's obvious that a screwdriver like yourself knows more than the experts working both at AMD and at Intel, right?"
Did you say one time that you are working on
chipsets, and therefore are somewhat an expert?
In this case my answer is obvious :)
And again, you understand everything too literally.
I already told once a story to Penang Paul, and I
will do you a favor and repeat it for you:
On a barn at my back yard somebody wrote
graffiti like "prick". And what would you think
the barn holds? Firewood!
Got the idea?
In conclusion, the Intel RAMBUS platform, even
if it sometimes works, is a performance
joke. |
