AMD hex core

[frankp]

Thinking about upgrading hardware.
Has anyone got non-overclocked nps comparison of Crafty (say) running on the AMD six core cpu compared to (say) the i7/i5 quads?

[rbarreira]

I have a Phenom II X6 1055T, how to compile/run crafty for 6 cores (on Linux)? Preferably with the stuff like popcount if it uses it.

[bob]

I have a Phenom II X6 1055T, how to compile/run crafty for 6 cores (on Linux)? Preferably with the stuff like popcount if it uses it.

If you can install the intel compiler, just type "make profile". You might edit the Makefile and look at the linux-64 section. Add -DPOPCNT as a compiler option to use the inline PopCnt() that uses the popcnt opcode...

the make profile will make about the fastest executable you can find.

[rbarreira]

In my experience the Intel compiler either generates executables that don't run at all when they detect an AMD CPU (if one of the -x options is used) or generates a really crappy codepath which is selected at runtime for AMD CPUs (by default, or if one of the -ax options is used).

Even with the newest AMD CPUs an executable made by icc will run something that would probably work on a 80386 when it detects an AMD CPU. So it will probably be more efficient if I use gcc.

[Mangar]

Hi,

I have an 6 Core AMD too. I can´t tell you about crafty nps compared to i5/i7.
Single Core result compared to a core2 quad intel processor:
The AMD CPU has about the same Spike - nps/cpu clock as the intel processor. The same with using 4 cores.

The point I like most for the X6 (running at 2,8 GHz) is the ability to undervolt it. My processor is stable with 1,05V (running for weeks with 6 parallel test games). The power consumtion of the hole system is 105W with all cores on 100%.

Drawback: On every reboot you have to reconfigure the voltage. As I typically don´t reboot the system it is not a big problem for me.

Greetings Volker

[frankp]

I have a Phenom II X6 1055T, how to compile/run crafty for 6 cores (on Linux)? Preferably with the stuff like popcount if it uses it.

On Debian (Lenny) I just type
make linux-amd64 (on an intel machine)
but first have to remove the -DNUMA -DLIBNUMA
from the relevant part of the makefile, since I do not have the numa libraries installed.

Do not have icc installed, since it did not work with Debian.

[Dann Corbit]

In my experience the Intel compiler either generates executables that don't run at all when they detect an AMD CPU (if one of the -x options is used) or generates a really crappy codepath which is selected at runtime for AMD CPUs (by default, or if one of the -ax options is used).

Even with the newest AMD CPUs an executable made by icc will run something that would probably work on a 80386 when it detects an AMD CPU. So it will probably be more efficient if I use gcc.

The profiler also clams up if you have an AMD chip (it REFUSES to give advice like it does if the CPU is INTEL).

[rbarreira]

Frank: I will try that later tonight.

[rbarreira]

Here are the results on a Phenom II X6 2.8 GHz, after compiling with gcc without profiling (and without the NUMA stuff).

I ran the bench command thrice. It seems it was using a really small hash table, I don't know if that influences the results much.

White(1): smpmt 6
Warning-- xboard 'cores' option disabled
max threads set to 6.
White(1): bench
Running benchmark. . .
......
Total nodes: 237692745
Raw nodes per second: 20024662
Total elapsed time: 11.87
White(1): bench
Running benchmark. . .
......
Total nodes: 245529185
Raw nodes per second: 20108860
Total elapsed time: 12.21
White(1): bench
Running benchmark. . .
......
Total nodes: 276856934
Raw nodes per second: 19579698
Total elapsed time: 14.14

[frankp]

Ricardo, thanks. Here are the result for my intel socket 775 2.8GHz quad. The AMD hex core does not look shabby at all...


unable to open book file [./book.bin].
book is disabled
unable to open book file [./books.bin].
Warning-- xboard 'cores' option disabled
max threads set to 4.
Warning-- xboard 'memory' option disabled
hash table memory = 4096M bytes.
Warning-- xboard 'memory' option disabled
pawn hash table memory = 64M bytes.
EGTB cache memory = 5M bytes.
EGTB access enabled
using tbpath=/data/CompressedTb
5 piece tablebase files found
14043kb of RAM used for TB indices and decompression tables
pondering enabled.
Crafty version 23.3
number of threads = 4
hash table memory = 4096M
pawn hash table memory = 64M
EGTB cache memory = 5M
60 moves/30 minutes 0 seconds primary time control
30 moves/15 minutes 0 seconds secondary time control
book frequency (freq)..............1.00
book static evaluation (eval)......0.10
book learning (learn)..............1.00
resign after 5 consecutive moves with score < -6.
book learning enabled


Crafty v23.3 (4 cpus)

White(1): bench
Running benchmark. . .
......
Total nodes: 401067303
Raw nodes per second: 12846486
Total elapsed time: 31.22
White(1): bench
Running benchmark. . .
......
Total nodes: 444869480
Raw nodes per second: 12801999
Total elapsed time: 34.75
White(1): bench
Running benchmark. . .
......
Total nodes: 574349110
Raw nodes per second: 12924147
Total elapsed time: 44.44
White(1): quit