Need help on mobile phone chess playing software and hardware: comparing apples to oranges--how strong is this software in Elo?

rec.games.chess.computer,rec.games.chess.misc,comp.arch.embedded,alt.comp.h= ardware.pc- homebuilt

Can anybody help me figure out how fast (what Elo rating) modern cell phone and mobile phone hardware play chess? Go to the SSDF rating list

formatting link
and note that

38 Pocket Shredder Ipaq 114 624 MHz 2682 65 -59 138 67% 2557 39 Deep Sjeng 1.5a 256MB Athlon 1200 MHz 2671 31 -31 493 52% 2659 40 CEBoard Fruit 2.3.1 XScale 400 400 MHz 2642 59 -57 149 59% 2580

Which shows that an XScale 400Mhz machine plays like a Pentium IV class machine from around 2003. I think XScale is used in some older (few years ago) Nokia cell phones.

But what about more modern cell phone hardware, like the ones that use Qualcomm's SnapDraggon chip?

From what I can tell, the following rule of thumb is correct.

Snapdragon (single core) and Marvell's Armada 500/600, both based on ARMv7 implementations, are roughly equal to Intel's Atom. Intel's Atom is roughly equal to a 2003-2004 vintage Celeron. So using SSDF you can do the math and see what a mobile phone playing software chess program like Pocket Fritz can do in a cell phone employing these chips. Roughly the programs would have an Elo on the SSDF list of a little above 2660 Elo on the SSDF list, probably close to 2700 to 2750 (dual core Snapdragon cell phones, which come out in late 2011 to

2012, would be at the upper limit).

But does anybody have more precise figures? Perhaps based on the number of chess nodes searched per second for various chips embedded in mobile PDAs and cell phones?

RL

From the net...

Anybody

formatting link

Embedded processors based on the ARM version 7 instruction set architecture (such as TI's OMAP 3 series and Freescale's i.MX51 based on the Cortex-A8 processor, or the Qualcomm Snapdragon and Marvell Armada 500/600 based on custom ARMv7 implementations) offer similar performance to the low end Atom chipsets[dubious =96 discuss] but at roughly one quarter the power consumption, and (like most ARM systems) as a single integrated system on a chip, rather than a two chip solution like the current Atom line. Although the next-generation Atom codenamed "Pineview" should greatly increase its competitiveness in performance/watt, ARM plans to counter the threat with the multi-core capable Cortex-A9 processor as used in Nvidia's Tegra 2, T.I.'s OMAP 4 series, and Qualcomm's next-generation Snapdragon series, among others.

--
Mobile Chipsets: WTF Are Atom, Tegra and Snapdragon?

Dan Nosowitz =97 Low-power processors aren't just for netbooks: These
 Click to see the full signature
Reply to
RayLopez99
Loading thread data ...

.hardware.pc-

=A065 =A0 =A0 =A0-59 =A0 =A0 138 =A0 =A0 67% =A0 =A0 2557

=A0 =A0 =A0-31 =A0 =A0 493 =A0 =A0 52% =A0 =A0 2659

=A0 =A0 =A0-57 =A0 =A0 149 =A0 =A0 59% =A0 =A0 2580

Why not compare it to XScale (ARM)? However, memory types and clock rate would make a different.

This is probably old news, very old news. The LG VS740 is 600MHz snapdragon class cpu + dsp core. Newer versions are 1GHz.

Reply to
linnix

mp.hardware.pc-

=A065 =A0 =A0 =A0-59 =A0 =A0 138 =A0 =A0 67% =A0 =A0 2557

=A0 =A0 =A0-31 =A0 =A0 493 =A0 =A0 52% =A0 =A0 2659

=A0 =A0 =A0-57 =A0 =A0 149 =A0 =A0 59% =A0 =A0 2580

Yes, a comparison with ARM is in order, and has been done in some old testing sites for chess. I reproduce below some data from a few years ago.

I am now interested in the A4 chip by the iPhone, which is about 35% faster than first generation SnapDragon chips. I think the iPhone 4 playing chess would be about 2700, since it's about twice as fast as earlier generations, and for every doubling of speed you gain 100 Elo points, and from the below data the earlier PDAs had about 2600 Elo performance (see 2606 Elo below--note that inferior software will push this number down on the same hardware, so we are talking about the optimal, best software for any given hardware), so it would rate 2700 Elo on the SSDF scale, and indeed if you Google HIRACS, a chess software company that writes code for the iPhone, that is exactly what they claim on their website for the latest iPhone hardware. So the SnapDragon would be roughly about the same (35% is not that big a deal) or around 2675 Elo).

In conclusion, so far, the evidence points that Snapdragon (Android) is roughly the same as a A4 chip (iPhone) for chess playing (again, on optimal, best practices chess playing software) and roughly they play about twice as fast as the XScale platforms, and roughly equal in chess performance a Pentium IV (Athlon, most likely late 1990s first generation but perhaps a Thunderbird maybe, from around 2002) in performance, or perhaps as an upper bound (stretching it) a modern Intel Atom (though I'm sure Intel would dispute that). Roughly these modern chips play at the 2650 to 2700 Elo level on the SSDF scale.

The next generation A4 and Snapdragon will be multiple core, but other than multitasking, unless the chess software is rewritten to take advantage of multiple cores, I'm not sure if that will make the chips that much stronger for chess. But if the feature size is the same for the A4 (and Googling it I see that the newer A5 will be the same 45 nm size as the A4) I don't think the A5 (next generation A4) will be faster at chess--of course being bigger it will "do more" and be a better chip for multitasking for mobile phone feature purposes, but that's beside the point for this discussion.

Any other insights appreciated.

RL

Palm/Pocket PC Software PDA Estimated SSDF Elo Rating based on 288 games Palm Chess Hiarcs 9.46 Palm Tungsten T3 XScale 400Mhz 2606 Pocket Fritz 2.0 IPAQ XScale 400Mhz 2511 Palm Chess Genius 2.1 Palm Tungsten T3 XScale 400Mhz 2394 Pocket Grandmaster IPAQ XScale 400Mhz 2381 Palm Chess Tiger 15.1 Palm Tungsten T3 XScale 400Mhz 2297

Reply to
raylopez99

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.