Chess Forum / Computer / March 2010

Thanks, Tony.  I wasn't sure because the Chessbase page at
http://www.chessbase.com/shop/product.asp?pid=488 says that the following
items are new to Deep Fritz 12:

Completely redesigned ergonomic interface
Twelve months premium membership on Playchess.com
New DEEP FRITZ engine: even stronger!
New openings book by Alex Kure
Twelve hours chess tutorial videos with international trainers and world
class players
Updated database of 1.5 million games from 1625 to 2009

But I was aware that Fritz 12 already had the 12 months free, 12 hours of
tutorials and the databas eof 1.5 m games.  I think it also had the openings
book by Kure.  As those weren't 'new to Deep', I wondered if the 'completely
redesigned interface' was the same as in Fritz 12 too.

Every try to do a reasonable parallel chess searcher?

Flipping compiler switches aren't going to do it.  There are *NO*
compilers or compiler libraries (from Intel, Microsoft, who ever) than
can automatically convert a single cpu searcher into a multi-cpu
searcher.  And going beyond a few cpu's into 16, 32, etc. is even harder.

Even with the earliest parallel chess programs (like Ostrich), the
authors struggled with good ways to split the search to get reasonable
performance improvements.  Later programs have better algorithms (based
on years & years of research by others) but they still aren't easy to do
and they aren't done by flipping a preprocessor directive.

People tried just simply splitting a tree into 'X' separate threads.
Fairly easy to do but very wasteful.

After years of research, people came up with various dynamic ways to
split a search based on where in the search it was, whether other
processors were still busy, etc.

It's easy to program the search so it can use a few extra cpu's, but the
trick is they have to be helpful, and not just wasted cpu effort.

So, they may be activating the advanced searching algorithms with a
preprocessor switch, but the code its activating had to be written,
tested and tuned.  It's not something inherent in the compiler or some
helpful library they are linking to.

Is it worth the $40-$50 you ask...  (shrug)  That depends entirely on
whether you are willing to pay it.  Only the consumer can say if its
worth it to them.

Me....  Not a chance.

In general though, many programs have similar situations.  A plain
version and a more advanced version for a higher price.  It's all the
same code base but the extra features are enabled with a #define or the
inclusion of a couple extra library files to be installed onto the
user's system.

Really?  I wasn't aware most netbooks, ipods, iphones, android phones
etc. etc. were multi-cpu.

Why should they if people are willing to pay extra for extra cpu support?

Also, different people have different opinions of what 'Deep' means.  To
some, it means using more than a single CPU core.  To others, it means
using more than what a normal desktop has.  To still others, it means
using some exotic hardware with dozens or even hundreds of cores.

Still, going 'deep' (in the sense of going to 2, 4 or 8 cores) is likely
only going to add 20-30 points for each doubling up into 8 or so and
then after that probably much less.