Playing DOSBox D-Fend against Dedicated Computers

This forum is for general discussions and questions, including Collectors Corner and anything to do with Computer chess.

Moderators: Harvey Williamson, Steve B, Watchman

Forum rules
This textbox is used to restore diagrams posted with the fen tag before the upgrade.
Post Reply
User avatar
spacious_mind
Senior Member
Posts: 3999
Joined: Wed Aug 01, 2007 10:20 pm
Location: Alabama
Contact:

Playing DOSBox D-Fend against Dedicated Computers

Post by spacious_mind »

With the new strong Millennium Exclusive dedicated computer, it becomes more interesting now to find strong opponents for it to play against. Some interesting ones are obviously Tasc R30's, Chessmachines, Mephisto TM Machines and Ruud's Resurrection, Revelations and Reflection. But even if you were fortunate enough to have all these, you would still really not have that many opponents to play so it becomes interesting to also look at the old historic classic DOS programs and DOSBox with D-Fend as this would potentially add another 30 - 50 programs to play and have fun against your top dedicated computers without having to invest in anything special like an old DOS computer.

I have a 4 GHz AMD computer and a couple of Dell Latitude Pentium I5 2.67 GHz computers, therefore the challenge was to see how I could accurately use all these computers in tournaments to allow me to play one DOS program on one computer and another DOS program on another computer. Therefore I needed to first calibrate these computers under DOSBox D-Fend and find a setting that would also be very competitive for matches against dedicated computers.

Well, I have tried various programs and tests for calibrating different computers and most of them are time consuming and less than satisfactory in their accuracy at the end of the day.

But fortunately there is one DOS chess program that does allow you to accurately calibrate different PC's !!

Ed Schroeder's Rebel 10 is ideal for this. It has a Bench Test that works very well for calibration.

Below are my Bench Test Results using Rebel 10's built-in Bench Test on my AMD 4.0 GHz computer:

Bench Test - AMD 4 GHz FX-8350 8 Core
Dynamic + Max = 2551 ELO
Auto + Max = 2554
63000 CPU Cycles = 2400 ELO
75000 CPU Cycles = 2432
84000 CPU Cycles = 2450
85000 CPU Cycles = 2452

And here are my results for the Dell Latitude Pentium I5 2.67 GHz:

Bench Test - Dell Lattitude I5 M560 2.67 GHz
Dynamic + Max = 2518 ELO
Auto + Max = 2514
63000 CPU Cycles = 2400 ELO
75000 CPU Cycles = 2432
84000 CPU Cycles = 2450

Obviously as you can see under the DOSBox D-Fend Dynamic + Max and Auto + Max settings the computers as a result of their different speeds perform differently and therefore would be useless to play against each other or against dedicated computers. Their scores of over 2500 ELO's would also be too strong and therefore hardly interesting when playing against a dedicated computer.

However by using the CPU-Cycles Value option it is possible to set two totally different computers to play identically against each other. As you can see from above by setting the cycles to 63,000 both computers are rated by Rebel 10 with a strength of 2400 ELO and 84,000 cycles were rated with 2450 ELO.

Since these cycle values fall below these computers max strength rating it becomes easy now to find an acceptable setting that works on your different systems accurately, so long as you remain below the max settings.

So now as first test I played Millennium Exclusive against Rebel 10 with the 63,000 = 2400 ELO Rebel 10 Bench setting:

[Event "Test Match"]
[Site "Alabama"]
[Date "2017.05.26"]
[Round "1"]
[White "DB63F Rebel 10, 40/20."]
[Black "Millennium Exclusive, 30S."]
[Result "1/2-1/2"]
[ECO "E11"]
[WhiteElo "2400"]
[BlackElo "2300"]
[Annotator "Spacious_Mind"]
[PlyCount "116"]
[EventDate "2017.05.26"]
[EventRounds "2"]
[EventCountry "USA"]

1. c4 Nf6 2. d4 e6 3. Nf3 Bb4+ 4. Bd2 Be7 {DB63F Rebel 10 out of book} 5. Nc3 d5 {Millennium Exclusive out of book} 6. e3 O-O 7. Bd3 dxc4 8. Bxc4 b6 9. O-O Bb7 10. Rc1 c5 11. dxc5 Bxc5 12. Qe2 Nbd7 13. Rfd1 a6 14. e4 b5 15. Bb3 Rc8 16. Bf4 b4 17. Na4 Be7 18. Ne5 Rxc1 19. Rxc1 Nxe5 20. Bxe5 Nxe4 21. Rc7 Bd5 22. Ra7 Bxb3 23. axb3 Bc5 24. Nxc5 Nxc5 25. h3 Qd5 26. Qe3 h6 27. Qg3 g5 28. Qe3 Rd8 29. f4 Rd7 30. Rxd7 Nxd7 31. Bc7 g4 32. Kh2 gxh3 33. Qxh3 Kh7 34. Bd8 Ne5 35. Bf6 Ng6 36. Bg5 Kg8 37. Bxh6 a5 38. Bg5 Qd4 39. Qf3 Qh8+ 40. Kg1 Qxb2 41. Qd1 Qc3 42. Kh2 f6 43. Qh5 Qc2 44. Bxf6 Qxg2+ 45. Kxg2 Nxf4+ 46. Kf3 Nxh5 47. Bd4 Kf7 48. Bb6 a4 49. bxa4 e5 50. Ba5 b3 51. Bc3 Ke6 52. a5 Nf6 53. a6 e4+ 54. Kg3
Nd5 55. a7 Nc7 56. Kf4 Kd5 57. Ke3 Kc6 58. Kxe4 Kb7 1/2-1/2

And the game was a draw! And Rebel 10 was a little lucky to get the draw.

Well since I am not convinced that Millennium Exclusive is the strongest dedicated (excluding engines). I felt that Bench Test setting of 2400 might be a little too low for DOS. So next I played a game at the 75,000 cycle setting that scored ELO 2432 in the Bench Test.

[Event "Test Match"]
[Site "Alabama"]
[Date "2017.05.26"]
[Round "2"]
[White "Millennium Exclusive, 30S."]
[Black "DB75F Rebel 10, 40/20."]
[Result "0-1"]
[ECO "B10"]
[WhiteElo "2300"]
[BlackElo "2400"]
[Annotator "Spacious_Mind"]
[PlyCount "162"]
[EventDate "2017.05.26"]
[EventRounds "2"]
[EventCountry "USA"]

1. c4 c6 2. e4 d5 3. cxd5 cxd5 4. exd5 Nf6 5. Nc3 Nxd5 6. Nf3 Nxc3 7. bxc3 g6 8. d4 Bg7 9. Bd3 O-O 10. O-O Nc6 11. Re1 {DB75F Rebel 10 out of book} Bg4 12. Be4 Qc7 {Millennium Exclusive out of book} 13. h3 Be6 14. Ng5 Bc4 15. Bd3 Bxd3 16. Qxd3 h6 17. Ne4 Rfd8 18. Rb1 Rac8 19. Qb5 b6 20. Bd2 f5 21. Qb3+ Kh7 22. Ng3 e5 23. Qc4 exd4 24. cxd4 Rxd4 25. Qc2 Qd7 26. Be3 f4 27. Bxd4 Nxd4 28. Qe4 fxg3 29. fxg3 Rc2 30. a3 Qc7 31. Re3 Nf5 32. Rf3 Rc1+ 33. Rxc1 Qxc1+ 34. Kf2 Bd4+ 35. Ke2 Qb2+ 36. Kd1 Qxg2 37. g4 Bc3 38. Qb7+ Ng7 39. Qd5 Qh1+ 40. Kc2 Bd4 41. a4 Qg2+ 42. Kd3 Bf6 43. Qb7 Bg5 44. Qc6 Bc1 45. Qe4 Qd2+ 46. Kc4 a6 47. Qd3 Qa2+ 48. Qb3 Qe2+ 49. Qd3 Qh2 50. Kd5 Qc7 51. Qc4 Qd7+ 52. Ke5 Qe7+ 53. Kd5 Bb2 54. Qc6 Qe5+ 55. Kc4 Qe2+ 56. Kb3 Ba1 57. Kb4 Qb2+ 58. Kc4 Qc2+ 59. Kd5 Qe2 60. Kd6 Qe5+ 61. Kd7 Qe8+ 62. Kd6 Qb8+ 63. Kd5 Qe5+ 64. Kc4 Qd4+ 65. Kb3 Qd1+ 66. Ka3 Bd4 67. Kb4 Qd2+ 68. Kc4 Qc2+ 69. Kd5 Bc5 70. Qf6 Qxa4 71. Qf7 Qb5 72. Ke5 Bg1+ 73. Ke4 Qe2+ 74. Kd5 Qc2 75. Qf8 b5 76. Rf4 Qg2+ 77. Ke5 Qe2+ 78. Re4 Qb2+ 79. Kd5 Qd2+ 80. Kc6 Qc2+ 81. Kb7 Qxe4+ 0-1

This time Rebel 10 won. But still since part of the exercise is to challenge all the top dedicated computers I still feel this setting is a little too low for DOS and therefore assuming that I will have to play some more games confirm this, my inclination is to calibrate the DOS Programs to the ELO 2450 Rebel 10 Bench Setting by using 84,000 cycles.

So therefore upon completing my Division 4 Tournament I will likely start Division 3 with the 84,000 cycle setting for all the DOSBox programs.

So what is the advantage of standardizing with a DOSBox D-Fend cycle setting? Well it becomes obvious when you look at Michael's Tournament that we are currently playing where you have a Hiarcs program competing through a slow down process and a Mysticum program against dedicated computers.

In future should there be interest in another tournament, anyone that has a fairly fast laptop I5 2.4 GHz and above and the DOSBox D-Fend programs set at say 84,000 cycles (2450 ELO) should be able to join in and play as well, using their favorite DOS program such as Hiarcs Masters, Genius 1-5, Sargon 5, MChess Pros, Fritz, Kallisto, Nimzo, Psion, Rex and so and on...

Now wouldn't that make for a great dedicated against Classic DOS programs tournament?

Best regards
Nick
User avatar
mclane
Senior Member
Posts: 1600
Joined: Sun Jul 29, 2007 9:04 am
Location: Luenen, germany, US of europe
Contact:

Post by mclane »

I am Not Sure if these elo rates are not too high for the exclusive.
It only runs on 300 MHz and even on my 500 MHz via pc I had the feeling that Exclusive is outcomputed by the dos programs.

Even on my via 533 MHz pc, rebel 10 says / calculates a benchmark rating of
2554 !!

Therefore I think I switch from pc to mobile phone.
What seems like a fairy tale today may be reality tomorrow.
Here we have a fairy tale of the day after tomorrow....
User avatar
spacious_mind
Senior Member
Posts: 3999
Joined: Wed Aug 01, 2007 10:20 pm
Location: Alabama
Contact:

Post by spacious_mind »

mclane wrote:I am Not Sure if these elo rates are not too high for the exclusive.
It only runs on 300 MHz and even on my 500 MHz via pc I had the feeling that Exclusive is outcomputed by the dos programs.

Even on my via 533 MHz pc, rebel 10 says / calculates a benchmark rating of
2554 !!

Therefore I think I switch from pc to mobile phone.
Hi Thorsten,

Yes exactly it is out computed your Pentium 533 is too fast. If you have a faster computer something like a Pentium I5 with 2.4 GHZ or higher then try DOSBox D-Fend. I know Pentium I5 2.67 GHz works perfectly and so does my AMD 4.0 GHz Try out DOSBox D-Fend from Ed Schroeder.

If you look at my Post, my AMD 4.0 GHz DOSBOX D-Fend at full speed also scores 2554 ELO and this is too fast for the dedicated computers.

But in D-Fend if you right click with your mouse on Rebel 10 a menu opens. Choose EDIT and another menu opens. Under Hardware you see CPU. Click on CPU and now you have some CPU setting options. Here now you Click on VALUE under CPU cycles you enter 63000 and do the Bench Test and you will see the result should come out the same as my two computers with Rebel 10 doing a final ELO rating of 2400 ELO. Change the Cycles to 84000 and Rebel 10 will bench test at 2450 ELO. 75000 Cycles will Bench test at 2432 ELO.

As you can see from my two test games, Millennium Exclusive is competitive at playing Rebel with both 63000 cycles (2400 ELO) and 75000 (2432 ELO) But since Rebel 10 is one of the strongest DOS programs and I also want to consider other DOS Programs for being competitive against other top dedicated computers, I am going to in my tournaments set the Cycle at 84000 (2450 ELO) in order to really push all the top dedicated chess programs and not just Millennium Exclusive.

Now that the Cycle has been established, I can now Right Click on every other DOS Program that I have playing under DOSBox D-Fend and set them all to 84000 Cycles (Rebel 10's 2450 ELO setting). By doing this every other DOS Program performs at the same Hardware speed and that is perfect for comparisons and competitive matches. Also I can now play MChess Pro at 84000 Cycles on my AMD 4.0 GHz computer against Tasc King 2.54 at 84,000 cycles on my I5 2.67 GHz computer or vice versa and feel confident in knowing that neither program has a Hardware handicap as they are both playing at the same cycles.

Best regards
Nick
User avatar
mclane
Senior Member
Posts: 1600
Joined: Sun Jul 29, 2007 9:04 am
Location: Luenen, germany, US of europe
Contact:

Post by mclane »

I don't waste a new machine on that topic, I do have an old i3 laptop. I have also installed dfend on my i3 laptop.

There is a virtual HD directory. I guess I can copy all dos programs into
This directory and start them from the dos prompt.

With rebel10 we can measure the speed of emulation, right ?!

When I run rebel10 benchmark originally on this i3 on dfend, I get
2329.

IMO This is a level Millennium Chess exclusive could handle.
What seems like a fairy tale today may be reality tomorrow.
Here we have a fairy tale of the day after tomorrow....
User avatar
spacious_mind
Senior Member
Posts: 3999
Joined: Wed Aug 01, 2007 10:20 pm
Location: Alabama
Contact:

Post by spacious_mind »

mclane wrote:I don't waste a new machine on that topic, I do have an old i3 laptop. I have also installed dfend on my i3 laptop.

There is a virtual HD directory. I guess I can copy all dos programs into
This directory and start them from the dos prompt.

With rebel10 we can measure the speed of emulation, right ?!

When I run rebel10 benchmark originally on this i3 on dfend, I get
2329.

IMO This is a level Millennium Chess exclusive could handle.
Yes, your I3 is too slow though that's why I suggested I5 2.4 GHz and higher. I was fortunate I have old computer shop here who sells old laptops by refurbishing them and loading Windows 10 on them. By old (here in USA) you are talking about I5's as well. So anyway I bought two Dell Latitude I5 2.67 GHz laptops for $50 each.

So know I can play both laptops and my Desktop 4.0 GHz AMD at the same time against each other or dedicated computers. This is why I experimented with finding a solution that works.

The key is that you have to establish your max speed on all the computers and then establish a cycle that works below the max speed. Because if your cycles become higher than the max speed the programs will no longer be efficient and will become sluggish.

From my test you can see and know that Rebel 10 Bench test 63,000 cycles = 2400 ELO and this setting is competitive with Millennium Exclusive. However my fear is that 2400 is a little too low for DOS programs as you don't want top dedicated computers to just win every time. This is why I think something like 84,000 cycles (2450) is about the right setting. But to do this you need a computer that when tested with Rebel 10 at max speed (auto+max setting or Dynamic + Max setting) scores higher than for example 2450 ELO. If it does, then you know that you can comfortably do 84,000 cycles and it will perform well as it is not exceeding its max performance.

Therefore technically everybody who has a computer that can do 84,000 cycles can now play each other in tournaments as the different hardware is no longer something that you have to worry about.

Your I3 with D-Fend unfortunately is a little too slow and I suspect it can only reach something around 40,000 cycles. Which means the top dedicated computers will find it too easy to win.

Best regards
Nick
User avatar
Tibono2
Full Member
Posts: 695
Joined: Mon Jan 16, 2017 7:55 pm
Location: France
Contact:

Post by Tibono2 »

Hi,

according to the figures bottom of this page: https://www.dosbox.com/wiki/Performance
84000 cycles should provide a DosBox performance close to a 110Mhz Pentium hardware.

I point out a very important statement Nick made:"find an acceptable setting...so long as you remain below the max settings". Your host CPU must
be powerful enough for the fixed cycles setting you aim at. Otherwise DosBox will no more respect real time.

Regards, Tibono
User avatar
spacious_mind
Senior Member
Posts: 3999
Joined: Wed Aug 01, 2007 10:20 pm
Location: Alabama
Contact:

Post by spacious_mind »

Tibono2 wrote:Hi,

according to the figures bottom of this page: https://www.dosbox.com/wiki/Performance
84000 cycles should provide a DosBox performance close to a 110Mhz Pentium hardware.

I point out a very important statement Nick made:"find an acceptable setting...so long as you remain below the max settings". Your host CPU must
be powerful enough for the fixed cycles setting you aim at. Otherwise DosBox will no more respect real time.

Regards, Tibono
Hi Tibono,

That's correct, I had estimated it at around 486-116 MHz but was not confident enough to quote it. So your 110 MHz confirms my approximate estimation.

Thanks
Nick
Nick
User avatar
spacious_mind
Senior Member
Posts: 3999
Joined: Wed Aug 01, 2007 10:20 pm
Location: Alabama
Contact:

Post by spacious_mind »

Tibono2 wrote:Hi,

according to the figures bottom of this page: https://www.dosbox.com/wiki/Performance
84000 cycles should provide a DosBox performance close to a 110Mhz Pentium hardware.

I point out a very important statement Nick made:"find an acceptable setting...so long as you remain below the max settings". Your host CPU must
be powerful enough for the fixed cycles setting you aim at. Otherwise DosBox will no more respect real time.

Regards, Tibono
Based on this chart, I just did a test with Rebel 10 with 77000 cycles = Pentium 100 approx. equivalent and it scored ELO 2437.

You know I might change my mind and instead of using 84000 (ELO 2450) I will try 77000 cycles (ELO 2437) so that the DOS programs can also be more accurately compared to a Pentium 100!

Since Rebel 10 already beat Millennium Exclusive at 75000 cycles, 77000 cycles might be the right compromise and allow comparison to a Pentium 100 as well :)

Best regards
Nick
User avatar
spacious_mind
Senior Member
Posts: 3999
Joined: Wed Aug 01, 2007 10:20 pm
Location: Alabama
Contact:

Post by spacious_mind »

OK I have done a few CPU Cycle ratings:

26800 CPU Cycles = 2126 ELO = 486-66 MHz?
43300 CPU Cycles = 2307
46000 CPU Cycles = 2325
46500 CPU Cycles = 2328 = Pentium 60.3 MHz?
47000 CPU Cycles = 2331
53200 CPU Cycles = 2365 = 486-132 MHz?
63000 CPU Cycles = 2400
69300 CPU Cycles = 2418 = Pentium 90 MHz?
75000 CPU Cycles = 2432
77000 CPU Cycles = 2437 = Pentium 100 MHz?
84000 CPU Cycles = 2450
85000 CPU Cycles = 2452

@Thorsten if what DOSBox shows in their CPU comparisons is correct then your I3 plays at approximately the equivalent of 60.3 MHz. 46200 Cycles would be exactly 60 MHz.

Besides Pentium 100 becoming an interesting standard, Pentium 90 would also be very interesting as when you go back in the years 1995-1996 in the German CS&S magazines you can follow the SSDF Ratings for a Pentium 90.

For example in the October/November 1996 magazine Pentium 90 programs were ranked as follows as Pentium 90:

1) Genius 3.0 = 2420 ELO
2) MChess Pro 5.0 = 2418 ELO
3) Rebel 6.0 = 2415 ELO
4) Rebel 7.0 = 2412 ELO
5) Genius 4.0 = 2409 ELO
6) Hiarcs 4.0 = 2392 ELO
7) Nimzo 3.0 = 2388 ELO
8) Hiarcs 3.0 = 2380 ELO
9) MChess Pro 4.0 = 2367 ELO
10) Fritz 3.0 = 2361 ELO
--
--
13) R30 v. 2.5 = 2356 ELO

So what remains to be checked is if DOSBox Pentium 90 setting really performs closely to a real Pentium 90.

Now that part is a bit tricky as it would require someone having some BT2630 Test results with some DOS Chess Programs running at Pentium 90.

I have not been able to find any. Does someone have any?

Best regards
Nick
Volodymyr
Member
Posts: 141
Joined: Sat Apr 08, 2017 1:03 pm
Location: Ukraine,Radyvyliv

Post by Volodymyr »

Hi Nick,
Any tests?
LCT II
https://chessprogramming.wikispaces.com/LCT+II
Results and rating - Selective Search 65,page 16.
User avatar
spacious_mind
Senior Member
Posts: 3999
Joined: Wed Aug 01, 2007 10:20 pm
Location: Alabama
Contact:

Post by spacious_mind »

Volodymyr wrote:Hi Nick,
Any tests?
LCT II
https://chessprogramming.wikispaces.com/LCT+II
Results and rating - Selective Search 65,page 16.
Hi Volodymyr,

Thanks but phew! that is lot of work to do all 35 tests :) I was hoping to find more some BT2630 position scores with time for the solution for one or two positions. That way I could check the time of the P90 or P100 against DOSBox D-Fend time for the same position.

Best regards
Nick
User avatar
spacious_mind
Senior Member
Posts: 3999
Joined: Wed Aug 01, 2007 10:20 pm
Location: Alabama
Contact:

Post by spacious_mind »

Ok I found something. In CS&S magazine December 1995, MChess Pro 5 was reviewed. In the report it showed a rating that MCP5 self generated for Pentium 100 with a rating of 2465 ELO. I just set DOSBox D-Fend with 77,000 cycles that is supposed to be equivalent to a Pentium 100 MHz and loaded MCP 5. MCP shows a rating of ELO 2495! Which I think is pretty close.

Next I reloaded MCP5 with 69,300 cycles which should equal a Pentium 90. It showed a rating of ELO 2480! It shows the same on both my AMD 4.0 GHz and my Dell Latitude 2.67 GHz computers!

So it works and that settles it.

My future matches are going to be set at 69300 cycles = Pentium 90.

ps.. I should add that all 3 of my computers are dedicated to chess and no other junk. So I would expect the score to be higher then the reporters score as who knows what other junk and sound cards etc he had installed on his Pentium 100 ;)

Dedicated to chess regards...
Nick
Post Reply