vesicles
Colonel
Hmmmm... Curve fitting with 2 data points? Any function would fit...
News on China's scientific and technological development.
- Thread starter Quickie
- Start date
Hmmmm... Curve fitting with 2 data points? Any function would fit...
Jura The idiot
General
gosh no, I evaluated the three parameters
(and
)...
(it's not clean but I needed to be quick...
the nitty-gritty:
a=2.2538;b=0.75104;c=2.3973;
...
goddammit are you trolling me?
how many points is this, huh?
vesicles
Colonel
No, I was serious. I saw two existing data points (2011 and 2016) and an additional extrapolated point at 2021.
Jura The idiot
General
vesicles
Colonel
When you do curve fitting, you have to use actual data points. The extrapolated info would be the predicted values based on the fit. So that data point at 2021 should not be considered as a part of the fit.
Jura The idiot
General
Mein Gott! I have three unknowns, and the exponential function, which can be linearized, so I have three equations, and there's been no 'fitting' involved, the values
are exact (now you may tell me 'within rounding errors' hahaha)!! anyway if you don't get it even from this:
then ...
oh and in
https://www.sinodefenceforum.com/ne...ogical-development.t4270/page-296#post-426366
I checked if such a crude approximation works, and to my surprise, it's predicted 2008 point (out of 1993, 1998, and 2003)
see now?
Jura The idiot
General
^^^
gee I used perhaps the simplest, more or less realistic model so that anybody could look at a picture and see the growth, and I could still get a prediction ... something which took me like twenty minutes before even drinking an early morning coffee, and was meant to be fun on a Military Forum (since it's not a Math Forum here, is it) ... and I'm hearing the model isn't fancy enough
and you bro SamuraiBlue, Yesterday at 10:33 AM
"That’s due to the fact that Linpack is a rather simple measure of floating point performance, based on calculations using linear algebra. HPCG, on the other hand, uses a number of different computationally-intensive algorithms. It incorporates calculations in sparse matrix multiplication, global collectives, and vector updates, which are said to more closely represent the mix of operations in many supercomputing codes. Overall, it's a much tougher metric than HPL; no system reached a single petaflop running HPCG.
That's mainly due to the fact that HPCG exercises data movement to a much that greater extent than HPL, and is therefore much more challenging to the memory subsystem. And given that memory is often the bottleneck on modern supercomputers, HPCG can be much more indicative of real application performance. If you glance through the , it’s immediately apparent the large difference between HPL and HPCG performance."
K Computer Comes Out on Top in HPCG Supercomputing Benchmark
| November 22, 2016 01:40 CET
EDIT
LOL now
gee I used perhaps the simplest, more or less realistic model so that anybody could look at a picture and see the growth, and I could still get a prediction ... something which took me like twenty minutes before even drinking an early morning coffee, and was meant to be fun on a Military Forum (since it's not a Math Forum here, is it) ... and I'm hearing the model isn't fancy enough
and you bro SamuraiBlue, Yesterday at 10:33 AM
shouldn't have been talking about cars here, but should've posted stuff like
"That’s due to the fact that Linpack is a rather simple measure of floating point performance, based on calculations using linear algebra. HPCG, on the other hand, uses a number of different computationally-intensive algorithms. It incorporates calculations in sparse matrix multiplication, global collectives, and vector updates, which are said to more closely represent the mix of operations in many supercomputing codes. Overall, it's a much tougher metric than HPL; no system reached a single petaflop running HPCG.
That's mainly due to the fact that HPCG exercises data movement to a much that greater extent than HPL, and is therefore much more challenging to the memory subsystem. And given that memory is often the bottleneck on modern supercomputers, HPCG can be much more indicative of real application performance. If you glance through the , it’s immediately apparent the large difference between HPL and HPCG performance."
K Computer Comes Out on Top in HPCG Supercomputing Benchmark
| November 22, 2016 01:40 CET
EDIT
LOL now
Last edited:
vesicles
Colonel
Mein Gott! I have three unknowns, and the exponential function, which can be linearized, so I have three equations, and there's been no 'fitting' involved, the values
are exact (now you may tell me 'within rounding errors' hahaha)!! anyway if you don't get it even from this:
then ...
oh and in
https://www.sinodefenceforum.com/ne...ogical-development.t4270/page-296#post-426366
I checked if such a crude approximation works, and to my surprise, it's predicted 2008 point (out of 1993, 1998, and 2003)
see now?
What's up with the hostility? I simply questioned your methodology... You should know that it is perfectly normal for people to query about methodology in any kind of analysis. And it is the responsibility of the analyzer to convey his/her methodology clearly to the audience.
This is especially the case when you try to derive anything out of a highly limited data set. And you have a highly limited data set. Even if you include all 3 points, you only have 3 points including one based on an assumption. Your Chi2 is 0. And your R2 is 0 as well. That is by any definition a lousy fit and a poor correlation, which means your correlation means absolutely nothing.
Yes, your equation happens to be consistent with some of the existing values. That is simply luck and does not say anything about the quality of your correlation. Come back when you collect enough data and your Chi2 and R2 get close to 1.
Jura The idiot
General
Last edited:
Quickie
Colonel
LOL! including also
#1 in 2001 with 7226 GFlops
and
#1 in 2011 with 10510 TFlops
I quickly scrambled this:
(at first I plotted up to one exa Flops but then even the 2011 point became like invisible
and you may tell me
now this business sure looks like a boondoggle to me
- how far it's necessary to go to be able to predict California weather, Honshu Island weather, ... weather, in real time?
- if people will go into computing?
Moore's Law is originally used to describe the development of the Microprocessor chip. It only so happens the development of supercomputers also follow some sort of an exponential curve albeit with a very different timeline of doubling the speed. Bear in mind that, since the timeline can be very variable, it can easily go the whole nine yards to get a fit of the curve.
In the case of the Microprocessor chip, Moore's Law is already starting to hit the limit because of heat problem and the fact that circuits don't work reliably when it gets near to the size of atoms.
Last edited: