Seite 17 von 22

Re: [sub-project] Perfect cuboid

Verfasst: 19.07.2018 09:24
von Zak
Excellent, but let's move to the triple dimension. Now do the same with the other two face diagonals for Euler's brick. Then imagine that we have placed the three halves of the sweep so that the points C and A coincide for all three cases, since the length of the space diagonal is the same. As a result, I want to see from you a triple cyclic equation for the length of the CA by dint of coordinates.

Re: [sub-project] Perfect cuboid

Verfasst: 19.07.2018 12:04
von Zak
If I am not mistaken
fig02.jpg
fig02.jpg (26.83 KiB) 2240 mal betrachtet

Re: [sub-project] Perfect cuboid

Verfasst: 19.07.2018 20:16
von x3mEn
CodeCogs2.gif
CodeCogs2.gif (10.31 KiB) 2236 mal betrachtet

Re: [sub-project] Perfect cuboid

Verfasst: 20.07.2018 22:15
von Zak
O.K. Thanks. So if we have an irrational space diagonal then at least one of coordinate will be irrational. I will show you the invariance test.
I have write up a square to you:

(a^2+b^2)+c^2

(c^2+a^2)+b^2

(b^2+c^2)+a^2

It's a square. You can see the vertical rows and horizontal lines in it. And everywhere in them will be the alternation of variables: a,b,c. And we can also see diagonals, in one of which there is also an alternation of the same variables: b,a,c, and in the other diagonal there is only one variable: a,a,a.

Lets: a - odd, b - even, c - even.
Let's write our square one more time:

(odd+even)+even

(even+odd)+even

(even+even)+odd

Try to find the rows, lines and diagonals in it.
As a result, we have integer value under square root, but irrational value of square root.

Re: [sub-project] Perfect cuboid

Verfasst: 27.07.2018 20:36
von BlackObelisk
I think score system for perfect cuboid is bugged for Ubuntu and someone (deliberately or not?) abusing it. How on earth Intel(R) Atom(TM) CPU Z3735F @ 1.33GHz can outperform Intel(R) Xeon(R) CPU X5560 @ 2.80GHz ? If u and your wingman run special version of Ubuntu/BOINC u get 1200 points instead of 200. For example this host http://www.rechenkraft.net/yoyo//result ... tid=429410 I hope admin can fix it.

Re: [sub-project] Perfect cuboid

Verfasst: 28.07.2018 02:11
von Zak
x3mEn
How do you thing about this:
02.jpg
02.jpg (65.22 KiB) 2149 mal betrachtet

Re: [sub-project] Perfect cuboid

Verfasst: 28.07.2018 09:01
von gemini8
BlackObelisk hat geschrieben:I think score system for perfect cuboid is bugged for Ubuntu and someone (deliberately or not?) abusing it. How on earth Intel(R) Atom(TM) CPU Z3735F @ 1.33GHz can outperform Intel(R) Xeon(R) CPU X5560 @ 2.80GHz ? If u and your wingman run special version of Ubuntu/BOINC u get 1200 points instead of 200. For example this host http://www.rechenkraft.net/yoyo//result ... tid=429410 I hope admin can fix it.
I think you're mistaken. The Atom is the slower one.
And I think yoyo might award additional credit for fast return.

Re: [sub-project] Perfect cuboid

Verfasst: 28.07.2018 09:20
von BlackObelisk
gemini8 hat geschrieben:I think you're mistaken. The Atom is the slower one.
And I think yoyo might award additional credit for fast return.
No, if you are double-checked by normal host you will get the same score around 200, but if you double checked by ubuntu host with ridiculously high integer speed you will get 1200. Also check Top Hosts, Atom has same RAC as Xeon and some 4th generations of i7, seriously?

(note total score of atom is very low, and it was added on March, not that long ago)
top.jpg
top.jpg (133 KiB) 2135 mal betrachtet

Re: [sub-project] Perfect cuboid

Verfasst: 31.07.2018 04:53
von x3mEn
BlackObelisk hat geschrieben:I think score system for perfect cuboid is bugged for Ubuntu and someone (deliberately or not?) abusing it. How on earth Intel(R) Atom(TM) CPU Z3735F @ 1.33GHz can outperform Intel(R) Xeon(R) CPU X5560 @ 2.80GHz ? If u and your wingman run special version of Ubuntu/BOINC u get 1200 points instead of 200. For example this host http://www.rechenkraft.net/yoyo//result ... tid=429410 I hope admin can fix it.
I guess you are right, I also had found those, Ubuntu (maybe all *nix) clients are earned by more credits.
The root of the problem lies on extremely different MIPS measurement in Ubuntu versus Windows system.
Look, for example, at:
http://www.rechenkraft.net/yoyo/show_ho ... stid=25456
  • AMD Ryzen 7 1700X Eight-Core Processor [Family 23 Model 1 Stepping 1]
  • Microsoft Windows 10 Core x64 Edition, (10.00.10586.00)
  • Measured floating point speed: 4292.94 million ops/sec
  • Measured integer speed: 16230.61 million ops/sec
and compare with this one:
http://www.rechenkraft.net/yoyo/show_ho ... tid=434292
  • AMD Ryzen 7 1700X Eight-Core Processor [Family 23 Model 1 Stepping 1]
  • Ubuntu 18.04 LTS [4.15.0-24-generic|libc 2.27 (Ubuntu GLIBC 2.27-3ubuntu1)]
  • Measured floating point speed: 4745.57 million ops/sec
  • Measured integer speed: 109708.58 million ops/sec
That's why *nix BOINC clients claim significantly more credits than Windows clients.
Credits earned are calculated as 2*min(claimed_credits)
When both, you and your wingmen have *nix, you both get more credits, than any Windows-Windows pair under the same other conditions.

Re: [sub-project] Perfect cuboid

Verfasst: 01.08.2018 18:41
von BlackObelisk
Looks like it's bugged only in ubuntu 17.10 and 18.04, other Linux clients get usual amount, but that one... both claim only 600, but granted 1200.

Re: [sub-project] Perfect cuboid

Verfasst: 01.08.2018 23:23
von x3mEn
BlackObelisk hat geschrieben:both claim only 600, but granted 1200.
As I have already said above, granted credits (for each of both) = 2*min(claimed credits by both)
For instance: http://www.rechenkraft.net/yoyo//workun ... d=39448720
host=434292 claimed 596.74
host=429410 claimed 695.48
granted credits = 2*min(596.74, 695.48) = 1,193.47

Re: [sub-project] Perfect cuboid

Verfasst: 02.08.2018 06:30
von yoyo
Some days back I limited the maximum credit you can get.