Comparing a desktop cpu to a tablet cpu

[fluidistic]

I'm wondering how to compare my desktop cpu to my tablet cpu. They are an intel core 2 duo E6300 (1.86 Ghz, you can see it at http://ark.intel.com/products/27248/) and the Samsung galaxy note 10.1, basically a quad core Exynos at 1.4 Ghz.
I would have expected that when running a chess program like stockfish (a top chess engine free to download, open source) I would get more playouts with the Samsung than with my desktop PC. However I get roughly half of the playouts number with the Samsung than with my desktop PC (456 k vs 950 k).
Considering that the RAM has little to no effect on the number of playouts, I'm wondering why basically my desktop PC can run programs faster than on my tablet.
How do they compare in general? Is the dual core PC always faster than my quad core tablet?

 

[Borek]

You just compared both in a real life application, and as far as I know that's the only way of doing it. Plenty of benchmarks that try to be universal, but usually there is no single and simple answer to the question "which of these two computers is faster" (unless it is obvious - say, Z80 vs Pentium ).

 

[fluidistic]

I see Borek. That's surprising to me :)
Thanks.

 

[256bits]

I would have expected that when running a chess program like stockfish (a top chess engine free to download, open source) I would get more playouts with the Samsung than with my desktop PC.

It would be interesting for you to explain why you would think so.

At 1.86 GHz vs 1.4 GHz the Intel chip is already the front runner in terms of instructions that can be performed per unit time.
A single core processor is just as fast as a multi core if the software is not written and taking advantage of the extras cores, and even so written, extra communication overhead between say two cores limits the processing to less than twice as a single.

Your Intel chip is designed to be a number cruncher.
The Exynos is designed to be peripheral friendly and thus limit the amount of support chips, since a lot of the kernal I suspect utilizes that aspect of having that is built into the processor chip itself. That is where the 4 cores are useful.

Very few applications such as the chess program you mentioned take advantage of extra cores of a processor. Programming for them is not the easiest.

In essence, the reply by Borek is the one to follow - test the application in real life.

 

[fluidistic]

Thanks for the reply!

It would be interesting for you to explain why you would think so.

At 1.86 GHz vs 1.4 GHz the Intel chip is already the front runner in terms of instructions that can be performed per unit time.
A single core processor is just as fast as a multi core if the software is not written and taking advantage of the extras cores, and even so written, extra communication overhead between say two cores limits the processing to less than twice as a single.

Here is my explanation: As you say, a dual core should be less than twice faster than a single core; at least for that chess program. If remember well I had read that it should be 1.6 times faster on a quad core than on a dual core (I don't remember exactly if it was for rybka or stockfish though).
Considering the clock latency of my quad core is more than 1/1.6 for my dual core, I would have expected it to get slightly more playouts per second, or at least close to what I get with the dual core. Certainly not half of them!
I just made the math, I should have expected close to 20% more playouts on my quad core than on my dual core.

 

[AlephZero]

You are comparing two chips with completely dfferent instruction sets (ARM and Intel). Just looking at the number of cores and the clock speed is much too simplistic.

There is also the issue of how mature the software development tools (e.g. compilers) are - if open source developers have strong views about using open source compilers, it's possible that they haven't yet got a compiler optimized for a relatively new chipset like the ARM, compared with the level of understanding of the Intel - or that they are more interested in spending time developing chess algorithms rather than understanding a new hardware system.

 

[Dmobb Jr.]

I don't know anything about stockfish so this might be irrellivant. Most android apps are written in java, which means the slowdown of running in a VM.