A hardware question

[Ponti]

What would be an engine`s strengh running:

- an Atom netbook
- core i7
- core i5

I`m thinking of switching back my #1 engine for analysis to... Shredder, Hiarcs, Zappa or Deep Fritz. I`m tired of all that discussion about... you know...

[bob]

What would be an engine`s strengh running:

- an Atom netbook
- core i7
- core i5

I`m thinking of switching back my #1 engine for analysis to... Shredder, Hiarcs, Zappa or Deep Fritz. I`m tired of all that discussion about... you know...

Strength will depend on engine and clock speed. The core i7 is extremely fast. I have not yet run on a core i5 so can't offer any data. Atom is nice with regard to power usage, but it is way slow compared to normal processors. Clock speed is significantly lower, as is FSB, etc. But there are some very nice form-factor mini-notebooks built around this that offer spectacular battery life.

[alpha123]

The i5 can overclock very well, but the i7 is still a bit faster. The Atom netbook.... isn't too good for chess. It's a bit slow. The i5 is probably best from a speed/price standpoint, the i7 is the fastest, and the Atom is the cheapest by a lot.

alpha123

[Ponti]

Well, I´m rated only 2029 FIDE.

What is the expected rating of, say, Fritz 12 running in a netbook with Atom processor?

[CRoberson]

Well, I´m rated only 2029 FIDE.

What is the expected rating of, say, Fritz 12 running in a netbook with Atom processor?

CCRL rates Fritz 12 single proc at 2978 on an AMD 4600+. Some claim that CCRL over rates the programs by 200 points. So, we get a lower
bound of 2778 and a upper bound of 2978.

Now, we have to adjust for the speed issues. I don't know how much slower the Atom is. So, lets make some bounded guesses. At the worst
it would be 16x slower. That would equate to 4 plys and about 400 rating points. Thus, putting it in the range of 2378 and 2578 at the worst.
I doubt it is that much slower. For an upper bound, lets try an estimate of 4x slower. That equates to about 2 ply and about 200
Elo. Thus, the upper bound estimate is in the range of 2578 and 2778. My guess is the Atom is only 2x slower, so add another 100 points to the upper bound.

So, at absolute worst, it is 2378 and at best it is 2878. If we average the two, it comes to a rough estimate of 2628.

IMHO, the guys that claim CCRL is 200 points over rated are wrong. The best version of my program Telepath scores between
2700 and 2900 against GMs. That is the 64 bit version on an i7 using only 1 processor. My program Ares (32 bit version) is about 200 points
weaker than Telepath and add another 200 pts for my faster hardware. That suggests Ares should be in the 2300s. CCRL rates it at 2340 on 40/40 TC.

If we assume the CCRL team is correct, then the average becomes 2728 instead of 2628.

[BubbaTough]

IMHO, the guys that claim CCRL is 200 points over rated are wrong. The best version of my program Telepath scores between
2700 and 2900 against GMs.

Perhaps in online blitz this is true, but I suspect this would not be true in a normal FIDE time control. I don't think Telepath is up to Anand strength yet for example. Or even Nakamura.

-Sam

[Eelco de Groot]

There is a Fritzmark posted by Ingo here, that must be Ingo Bauer! The Atom is very similar to the old Pentium III but with lower electricity consumption. I think this Atom 330 is two cores or else I don't see how with hyperthreading there are four cores seen by the benchmark. To be sure you would have to look that up, there are also single core Atoms I think, which are cheaper.

Hypertheading pushes up the Fritzmark, but it means probably nothing in real nodes per second, that is probably lower with hypertheading than without hyperthreading, if you run the chessprogram under full load. If you use just one core for chess and you have two then I'm not sure what happens exactly with hyperthreading. There are some discussions on the Rybka forum but that is mostly about hyperthreading on the i7, hyperthreading can be disabled probably in the bios. Anyway, just concentrate on the two cores figure, 2.02 that is very close to two Pentium III processors, so from that you can compare with what you would get with a Q6600, every doubling of the processor speed is very roughly 50 elo.

On Ingo's screen you also see two cores, maximum of 4 threads but that is a bit misleading.

Eelco

[Alex Lobanov]

http://www.sedatchess.com/craftybenchmarks.html

[CRoberson]

IMHO, the guys that claim CCRL is 200 points over rated are wrong. The best version of my program Telepath scores between
2700 and 2900 against GMs.

Perhaps in online blitz this is true, but I suspect this would not be true in a normal FIDE time control. I don't think Telepath is up to Anand strength yet for example. Or even Nakamura.

-Sam

Telepath is much improved since you last tried it. Its on ICC now. Give it a go.

[DomLeste]

Is the Telepath engine a private engine i dont seem to recall?