logo Welcome, Guest. Please Login or Register.
April 25, 2019, 07:41AM
Home Help Search Login Register
News: Visit a member's site: http://perso.wanadoo.fr/pthfndr (mychess)

  Chess Computer Community
  General Category
  Discussions about Chess Computers
(Moderators: Endspielgott, Overtom, bobosse)
  emul of occ
previous next
Pages: [1] Reply Notify of replies Send the topic Print 
   Author  Topic: emul of occ  (Read 1053 times)
emul4ever
Newbie
*

Offline

Posts: 2

I love Boardnation!

View Profile
Re:emul of occ
« on: January 5, 2010, 02:21PM »
Reply with quote

Hi,

I come back with mach 3 :

rsteller                           Fidelity                                             
Gerät                               mach 3                                             
                                    1024 K Hash                                         
BT 2630                             M68000                                               
                                   80 mHz       
                                                                          
BT 1   Sxg7                        100                                                 
BT 2   Lxb6                        155                                                 
BT 3   Te6                        20                                                   
BT 4   Df7                        224                                                 
BT 5   Ka6                        25                                                   
BT 6   ...e3                      900                                                 
BT 7   ...Td6                      16                                                   
BT 8   Txc6+                      2                                                   
BT 9   ...g5                      32                                                   
BT 10   Txg7+                      7                                                   
BT 11   ...Dxh2                    6                                                   
BT 12   ...De4                      110                                                 
BT 13   ...Le6                      900                                                 
BT 14   Txh7                        18                                                   
BT 15   e5                          175                                                 
BT 16   ...Sxg2                    900                                                 
BT 17   Dxf4                        35                                                   
BT 18   d6                          12                                                   
BT 19   ...f3                      13                                                   
BT 20   Ta2                        32                                                   
BT 21   Te1                        900                                                 
BT 22   a3                          540                                                 
BT 23   g4                          900                                                 
BT 24   g6                          112                                                 
BT 25   ...Sd3                      900                                                 
BT 26   f5                          1                                                   
BT 27   e6                          176                                                 
BT 28   e5                          900                                                 
BT 29   0-0-0                      900                                                 
BT 30   f4                          900                                                 
Summe:                            9911                                                 
BT - E L O :                       2307                                                 



Next time : DIABLO from novag

EMULATED regards (yes, it is an emulated 68000.
« Last Edit: January 7, 2010, 04:45AM by emul4ever » Report to moderator Logged
mychess
Full Member
***

Offline

Posts: 108





View Profile WWW E-Mail
Re:emul of occ
« Reply #1 on: November 9, 2009, 01:39PM »
Reply with quote

Kostea,

Well, it's always risky to do math during night after 18 hours without sleep.
Indeed, my calculation is wrong, not least because the Excel Club 68000 runs at 12 MHz, not 16.
But your calculation is equally so, because we must use the exponential function of 2 to find the number of time that  the increment of 50 points Elo applies: 2 ** n = 80/12, so n = 2.737, and there is 50 * 2.737 = 137 Elo points to add.
The excel at 80 MHz is therefore 1857 + 137 = 1994 Elo, error = 3.16%

If we used your too simple rule of three (proportion), a computer 16 times faster give 16 / 2 * 50 = 400 points Elo
while it's 4 * 50 = 200 Elo points (2 ** 4 = 16)

For the Elite V11, we  have 2 ** 2.17 = 72/16, so the Elite V11 should have:
1993 + 2.17 * 50 = 2101 ... At below 200 points of its actual valuation (error = 9%).
I can see 2 main reasons that increase his rating:

- The size of the Hash table (32 times of the Mach III)
- Cache memory CPU 68060, not avalaible on the 68000.

My math teacher would disagree with your math teacher from elementary school for your correction. ;-)

Best regards
Report to moderator Logged
Kostea
Newbie
*

Offline

Posts: 38



View Profile
Re:emul of occ
« Reply #2 on: November 9, 2009, 03:17AM »
Reply with quote

I feel the need to add another post. I was at 13 posts before this one, and don't want to stay at 13 for too long for the same reason you don't walk under a ladder, or whenever a black cat crosses your path you take the first 3 steps in backwards.
The other reason is that I have to make a correction. In my previous post I said that you can't overclock a Fidelity Excel Club with a 68000 CPU from 12 to 80 MHz. I still think that way.
But I think that emul4ever actually said that he saw an emulation (software) of a Fidelity Excel running at 80 MHz, and not the physical chess computer. I don't have any problem with that. Actually the elo rating of 2057 is around what is supposed to be.

What happens when the speed is increased. The cases to study is when the software is run at higher speeds either on the physical chess computer it belongs (by overclocking the chess computer hardware) or the software is taken out of the chess computer memory and ported to run on a different computer. The computer can be a PC, for example. How many elo points you get by doing that?

I have not read about any recent experiments. All I know is about stuff that happened in the past.
In the 80s, Ken Thompson did extensive experiments by playing two BELLE engines. For example a 5-ply version of BELLE would play against a 6-ply version, then the 6-ply would next play against a 7-ply and so on. He measured the elo gain per extra ply, and found out that each extra ply led to an increase of about 200-250 elo points in the lower scale below 2000, but past 2000 elo rating one extra ply would gain less and less points until it eventually plateau around grandmaster strength.
Since ply does not mean much to some people, let's convert it to something else. About 20-30 years ago it was said that a speed increase of 6 times would allow a chess program to search for moves one ply deeper. You're probably wondering why they were using a 6 times speed increase instead of 19, for example. What's so magical about 6? When the move trees are generated, the branching factor from one ply to the next ply is about 30-35 on average, but it is dramatically reduced to 6 branches by the cutoffs of alpha-beta algorithm used in the chess software. In fact the branches are not reduced from 30-35 to 6, the total number of evaluated end positions are as if there would have been 6 full branches to search instead of 30-35.
In the 70s and 80s all chess engines employed the alphabeta algorithm to search for moves, and they eventually hit a brick wall. Once they reached a certain playing strength, the progress in playing strength was minuscule year after year. The software could not be significantly improved. They realized that the only way to improve elo rating was to take advantage of the hardware advances. Just improving the software was not enough anymore. Several chess programmers teams built custom hardware with hundreds of CPUs in order to get the upper hand against the competition. For a while, whoever had the bigger and faster hardware was king of the hill. Then beginning in the 1990s some improved move searching algorithms such as NEGASCOUT, NULL MOVE, etc. came into regular use, and it was possible for even a lowly PC to beat a supercomputer just on the strength of the better software.
I still see on websites today, that a speed increase of 6 times is credited for an extra ply, which in turn gains about 200 elo points, but that would be correct only if the software uses ancient alphabeta. If it uses the most recent algorithms, it cuts down the branches maybe down to 3-4. Nowadays I think that the software with the better position evaluation function is on top. After all, all chess programmers have access to all the most up to date move search algorithms, fastest PCs, so the only area where someone can make a difference is in the position evaluation. That position evaluation also plays a big role in defining a particular style of play for which some are cherished and others are hated.

To sum it up:
- A chess computer such as one manufactured prior to 1990, they used alphabeta algorithm only, and can gain 200 elo points if the hardware is sped up about 6 times. Depending of the original CPU, that involves the redesigning of the PCB and then the software needs to be adjusted because the move clocks would be all messed up.
- Chess computers made in the 1990s only need a speed increase of 3-4 times to gain maybe 100-150 elo points.  They have some or most of improved move search algorithms.

The above applies to the top of the line chess computers.
I still had some other things to say, but this is a long post and it's time to put an end to it.

Regards

Report to moderator Logged
Kostea
Newbie
*

Offline

Posts: 38



View Profile
Re:emul of occ
« Reply #3 on: November 9, 2009, 01:15AM »
Reply with quote



mychess,

You have an interesting point of view, which gave me a very good idea.
I love the way you manipulate numbers to get to the result needed to prove the overclock rating.
Next time Uncle Sam will question why I paid so little taxes, I will use your mix and match formulas to prove that the calculations on my income tax form are correct, and they are the crazy ones.

My elementary school math teacher would have disagreed with your math calculations and would have beaten you with a stick. That is after you got slapped (read punched) in the face first and sat down on your knees on nut shells. That whole process made me the tough man that I am today. Never really got the opportunity to thank my teacher twice, once for teaching me math, and second for toughening me. He knew that life is hard and math was not going to be enough to succeed, I needed to be knocked out a few times and then brought back to life, just like the life has ups and downs.
Sh**t, I forgot that this is a chess computer forum, and not my shrink sessions. Can I delete all the above?

Here is what I would have seen my teacher writing on the blackboard in the last moments of lucidity, just before I was falling into a deep coma:

2 -> 40  (if a 2 times speed increase yields 40 elo points)
5 -> x    (a 5 times speed increase would yield )

Cross/diagonal multiply the above.

2x = 200
x = 100

So the Excel would have gotten an extra 100 elo points.

Now I realize we've just calculated the elo for a 'fata morgana'.
That 68000 CPU is really a tough one, it was overclocked almost 7 times over, and lived to see the end of a lenghty BT-2450 test. I bet it was sweating transistors at the end.

Let me get this right. The story is that a Fidelity Excel Club with a 68000 CPU running at 12 MHz was overclocked at 80 MHz.
I hardly doubt that you can overclock an 68000 CPU from 12 to 80 MHz, and it would run fine. PCB design (length of circuit traces; too long or not equal length), power supply juice, inadequate cooling, etc, just to name a few of the things that would try to sabotage you at every corner.
Is this a case of mistaken identity? No, I have no doubt that the Excel Club was an Excel Club. Don't get me wrong, I did not say that emul4ever needs bigger glasses. Please don't put words in my mouth.
What I'm suggesting, is that maybe the original 68000 CPU was replaced by a 68060, and then the 68060 was overclocked to 80 MHz. A 68060 was designed to run at 50-75 MHz. No, scratch that, the 68000 and 68060 are totally different in number of pins and shape of encasing. It would have needed a PCB makeover. I still can't swallow that without getting an indigestion.


Regards
Report to moderator Logged
Antonio Marques
Newbie
*

Offline

Posts: 18



I love Boardnation!

View Profile
Re:emul of occ
« Reply #4 on: November 8, 2009, 04:07PM »
Reply with quote

Hello Emul4ever

Where can I find this excel 68000 emulator?
Are there any site on the net?

Best Regards

Antonio Marques
Report to moderator Logged
mychess
Full Member
***

Offline

Posts: 108





View Profile WWW E-Mail
Re:emul of occ
« Reply #5 on: November 8, 2009, 02:04PM »
Reply with quote

Hello Emul4ever,

This information is very interesting.
It is well known that if the frequency doubles, it earns 30 to 50 points Elo (I use 40), and 80 = 5 * 16, then to an Excel @ 16MHz we add 80/16 * 40 = 200 ELO.
Club Excel 68000 is rated at 1857 Elo, we find:
1857 + 200 = 2057.
precision: (2057 - 2057) / 2057 = 0%!
The Elite V11 @ 72MHz is ranked 2291, and has the same program as the Mach III:
1993 + 72/16 * 40 = 2173.
precision: (2291-2173) / 2173 = 5.1%
Probably the large size of the hash table of the Elite V11 (32 times) explains the 118 ELO points between calculation and reality.
Is it 80 MHz the max frequency of the emulator ?
Sargon V (Mach II ) is rated 2400 Elo on a Pentium iV 2.5 GHz, the best performance for Dan and Kate Spracklen.

Mychess.
« Last Edit: November 8, 2009, 02:09PM by mychess » Report to moderator Logged
mclane
Newbie
*

Offline

Posts: 40



I love computerchess

View Profile WWW E-Mail
Re:emul of occ
« Reply #6 on: November 6, 2009, 02:13PM »
Reply with quote

? isn't that a little low for 80 mhz ?

also wouldn't it make more sense to bring a mach III to 80 mhz than an excel 68000 ?
Report to moderator Logged
Steve B
Jr. Member
**

Offline

Posts: 53





View Profile WWW E-Mail
Re:emul of occ
« Reply #7 on: November 5, 2009, 11:34AM »
Reply with quote

Modified Emulations ..
It That  Good??..

I think Not Regards
Steveb
« Last Edit: November 5, 2009, 11:35AM by Steve B » Report to moderator Logged
emul4ever
Newbie
*

Offline

Posts: 2

I love Boardnation!

View Profile
emul of occ
« Reply #8 on: November 5, 2009, 07:57AM »
Reply with quote

Hi,

I saw ysterday a fidelity club at 80 mhz,  2057 on bt-2450.
Very good, isn`t ?

Emulated regards
Report to moderator Logged
Pages: [1] Reply Notify of replies Send the topic Print 
previous next
Jump to:  

Login with username, password and session length

Chess Computer Community
Get your Ad Free Message Board at Boardnation!   © 2013 Boardnation  Web hosting by Nubis