Does anyone have a very old AMD machine ?

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, Dann Corbit, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Modern Times
Posts: 2635
Joined: Thu Jun 07, 2012 9:02 pm

Re: Does anyone have a very old AMD machine ?

Post by Modern Times » Fri May 19, 2017 7:59 pm

MDRedwood wrote:Probably of no help to you, but I'm running Win 7 64 on an old AMD laptop:
AMD Turion X2 Ultra ZM-82 Griffin 65nm Technology, 3 Gb RAM. Let me know if I can help.
That is a similar K8 architecture to the X2 4600+. Let's wait to see if Wolfgang can run the benches on his actual 4600+ machine first.
.

Opinions expressed here are my own, and not necessarily those of the CCRL Group.

User avatar
MikeB
Posts: 4551
Joined: Thu Mar 09, 2006 5:34 am
Location: Pen Argyl, Pennsylvania

Re: Does anyone have a very old AMD machine ?

Post by MikeB » Fri May 19, 2017 8:53 pm

Modern Times wrote:
mbabigian wrote:Instead I'd beg you to stop using this ancient machine as your reference. Mike
What I was wanting to do was to make sure that a modern engine like SF8 (on 1 core) gained much the same from faster hardware as the ancient Crafty that we use does. Then to possibly change the reference engine in future, not the hardware. But I can't propose anything unless I know how fast SF8 might have been on our reference machine, to compare to the Crafty bench.

I did play round with my X6 1100T, and I had to slow it down to 2.1GHz to match the 48 secs Crafty result. That was even after slowing down the RAM by half, the HT link and the Northbridge. So I guess just more modern architecture than the old X2.

This exercise may prove to be impossible now given the passage of time.
Your intentions are good and noble, but I think you will drive yourself nuts.

My suggestion is to do a reset, new bench baseline with SF, recalibrate to the new bench with - you have legacy testing with the legacy programs on the legacy time control - first reset the time control to something more modern - I think 40 min with 40 sec inc is very close to 40/40 - but also you have to adjust tc to be based on the benchmark set by stockfish - and reset tc to 40/40 on new hardware or 40 +40inc. Just my $.02 - you're the one that has to deal with it and live with your decision - I can definitely see your dilemma. I would check with your fellow testers - making a decision with other stakeholders requires consensus building.

Modern Times
Posts: 2635
Joined: Thu Jun 07, 2012 9:02 pm

Re: Does anyone have a very old AMD machine ?

Post by Modern Times » Fri May 19, 2017 9:56 pm

MikeB wrote: Your intentions are good and noble, but I think you will drive yourself nuts.

My suggestion is to do a reset, new bench baseline with SF, recalibrate to the new bench with - you have legacy testing with the legacy programs on the legacy time control - first reset the time control to something more modern - I think 40 min with 40 sec inc is very close to 40/40 - but also you have to adjust tc to be based on the benchmark set by stockfish - and reset tc to 40/40 on new hardware or 40 +40inc. Just my $.02 - you're the one that has to deal with it and live with your decision - I can definitely see your dilemma. I would check with your fellow testers - making a decision with other stakeholders requires consensus building.

None of that is going to happen and that was never the intention. There is no dilemma. This exercise is incredibly simple - to test if the old version of Crafty we use as a benchmark behaves similarly to a modern engine with improvements in hardware. That is all. And possibly to switch to SF8 as the benchmark engine in future. No more than that.
.

Opinions expressed here are my own, and not necessarily those of the CCRL Group.

User avatar
MikeB
Posts: 4551
Joined: Thu Mar 09, 2006 5:34 am
Location: Pen Argyl, Pennsylvania

Re: Does anyone have a very old AMD machine ?

Post by MikeB » Fri May 19, 2017 10:34 pm

Modern Times wrote:
MikeB wrote: Your intentions are good and noble, but I think you will drive yourself nuts.

My suggestion is to do a reset, new bench baseline with SF, recalibrate to the new bench with - you have legacy testing with the legacy programs on the legacy time control - first reset the time control to something more modern - I think 40 min with 40 sec inc is very close to 40/40 - but also you have to adjust tc to be based on the benchmark set by stockfish - and reset tc to 40/40 on new hardware or 40 +40inc. Just my $.02 - you're the one that has to deal with it and live with your decision - I can definitely see your dilemma. I would check with your fellow testers - making a decision with other stakeholders requires consensus building.

None of that is going to happen and that was never the intention. There is no dilemma. This exercise is incredibly simple - to test if the old version of Crafty we use as a benchmark behaves similarly to a modern engine with improvements in hardware. That is all. And possibly to switch to SF8 as the benchmark engine in future. No more than that.
You do have a dilemma, you just haven't recognize it yet. It is along the lines of "you don't know what you don't know". But we'll save that for another day - keep up the good work!

User avatar
lucasart
Posts: 3184
Joined: Mon May 31, 2010 11:29 am
Full name: lucasart
Contact:

Re: Does anyone have a very old AMD machine ?

Post by lucasart » Fri May 19, 2017 10:50 pm

Modern Times wrote:What I was wanting to do was to make sure that a modern engine like SF8 (on 1 core) gained much the same from faster hardware as the ancient Crafty that we use does. Then to possibly change the reference engine in future, not the hardware. But I can't propose anything unless I know how fast SF8 might have been on our reference machine, to compare to the Crafty bench.
You may end up comparing apples and pears, because some old engines are using software emulation of BSFQ and POPCNT, instead of hardware instructions. The main breakthrough in hardware are BSFQ which comes with 64-bit (hence 32-bit is so slow), and POPCNT, which came with AMD K10 Barcelona family.

As you review your time control, you may want to switch to Fischer clock, like everyone else. It's kind of silly to use 40 minutes for 40 moves, when you're at move 120, and engines are stubbornly avoiding 3-repetition and 50-move rule by making pointless delaying moves...
Theory and practice sometimes clash. And when that happens, theory loses. Every single time.

Guenther
Posts: 3694
Joined: Wed Oct 01, 2008 4:33 am
Location: Regensburg, Germany
Full name: Guenther Simon
Contact:

Re: Does anyone have a very old AMD machine ?

Post by Guenther » Sat May 20, 2017 8:43 am

lucasart wrote:
As you review your time control, you may want to switch to Fischer clock, like everyone else. It's kind of silly to use 40 minutes for 40 moves, when you're at move 120, and engines are stubbornly avoiding 3-repetition and 50-move rule by making pointless delaying moves...
You missed that point already made in the thread:
I would not want to get rid of all the oldies, I love them all, for me they are history. Also, if I were the author of an old chess program I would be deeply disappointed to see my engine disappear from the list.
And to include them in a new list would be impossible with our time and CPU resources.
Quite some older programs cannot be run correctly in Fischer mode or don't know Fischer mode at all, when >99.5 % can be run at moves per session.

Modern Times
Posts: 2635
Joined: Thu Jun 07, 2012 9:02 pm

Re: Does anyone have a very old AMD machine ?

Post by Modern Times » Sat May 20, 2017 9:02 am

lucasart wrote:
As you review your time control, you may want to switch to Fischer clock, like everyone else.
We are not reviewing our time control. How many times do I have to repeat myself. I am simply wanting to validate the continued use of ancient Crafty as a means of benchmarking current hardware vs the reference hardware. I did hesitate to post this request, because it had to potential to get way off topic, and that is exactly what happened. I wish I hadn't bothered.
.

Opinions expressed here are my own, and not necessarily those of the CCRL Group.

elpapa
Posts: 205
Joined: Sun Jan 18, 2009 10:27 pm
Location: Sweden
Full name: Patrik Karlsson

Re: Does anyone have a very old AMD machine ?

Post by elpapa » Sat May 20, 2017 11:48 pm

Modern Times wrote:Thanks Colin, very useful ! So 43 seconds for that Crafty bench at 2.5Ghz. So if that same CPU was running at 2.4Ghz it would probably be 45 secs.

And 7,624 ms for SF8, which at 2.4Ghz would be around 7,300 ms.
That can't be right. I guess you flipped the 25/24 part of the equation on the last one.

Jesse Gersenson
Posts: 584
Joined: Sat Aug 20, 2011 7:43 am
Contact:

Re: Does anyone have a very old AMD machine ?

Post by Jesse Gersenson » Sun May 21, 2017 9:18 am

Modern Times wrote:
Jesse Gersenson wrote:Ray, the following motherboard is available at a local shop, "GA-M69GM-S2H+CPU Athlon 64X2 4200/2,2GHz+2*1GB DDR2". I could run the benchmarks on it, but would the results from that board be useful for you?

Someone confirm the Crafty benchmark runs under Wine, and if so, which version:

Code: Select all

~$ wine --version
I don't want anyone to go and buy anything in order to help me out !
No one mentioned buying anything.

If you need me to run the benchark on that board let me know.

Modern Times
Posts: 2635
Joined: Thu Jun 07, 2012 9:02 pm

Re: Does anyone have a very old AMD machine ?

Post by Modern Times » Sun May 21, 2017 2:57 pm

OK then Yes that would be useful, thanks.
.

Opinions expressed here are my own, and not necessarily those of the CCRL Group.

Post Reply