Pcoket PC's Processor Question and Theory

[Andy_7676]

Hey I heard that my Casio E125 has a 64bit 150mhz processor..if thats the case would it be comparable to a 300mhz Pentium processor, since intels chips are 32bit?<br><br>I understand that cache and such plays a big role, but just curious on your thoughts on how the chips in the Pocket PC compare to say a P3 or P4?

[Dan East]

No, it would not be twice as fast as a 32bit processor. Let me give you an example I used before:<br>Say you and I are both processors that simply add two numbers together. I can add only 1 digit numbers at a time ( 4+5=9 ), but you can add 2 digit numbers at a time( 12+45=57 ). Since I can only add 1 digit numbers, if I need to add 2 digit numbers I must do it in two steps, and combine the results. However you can do it in a single step, since you can handle the larger values. Thus you would perform at least twice as fast as me when adding 2 digit numbers. But what happens if we compare adding single digit numbers? We have the same performance (assuming we're at the same mhz).<br>A 32 bit processor can handle an unsigned number from 0 to 4,294,967,295. Most programmers never need to work with values larger than that, so they never need to use 64 bit integers. Graphical algorithms can be written to make use of 64 bit integers when it comes to binary math. The problem is that a program must be written to make use of the extra 32 bits a 64 bit processor affords to see any performance improvement.<br><br>Next issue, is what is the bus and memory width of the Casio? 32 bit. So there's your bottleneck regardless.<br><br>Dan East

[Dan East]

Also, the Pentium processor has a very robust instruction set, while the Windows CE processors are have smaller instruction sets (RISC). Thus the Pentium can do many things with one single opcode that a CE processor must break down into multiple steps. This is very much true of Floating Point math. A CE processor will take hundreds of steps to perform a floating point operation that can be done by the Pentium in a one or two steps.<br><br>Dan East

[Moose or Chuck]

I can multiply floats in one step too...

[TBone]

------------------<br>Announcing MOOSEMASTER(C)-- Now available in a CISC architecture! Handles floating point arithmetic with near-infinite precision easily! Concatenates strings in a single operation! Easy to use Speech-To-Text or Text-To-Speech Algorithm! <br>------------------<br>"Yeah, but does he come in different colors? And what's this Lexus add-on I keep hearing about?"<br><grin>

[Andy_7676]

very enlightening posts...thanks guys......<br><br>so what would you say the Pocket PC's speed would be close to?<br><br>can you give an educated guess at what a 150mhz MIPS chip would be in relation to a desk top processor?

[TBone]

I don't know that you can really equate it to the desktop, due to the disparities in things like floating point support. It depends on the application for stuff like that, and it's also not helped by the fact that you can't really do coding in assembly for PPC, making it sometimes tough to get full power out of the platform.<br><br>I would say that when it comes to floating point performance like Quake, my iPAQ currently runs around the same as an old 486/DX 66MHz I used to own. This will not stay current as Dan and other developers who are farther into the engine than I am replace the floating point emulation with a C++ class, at which point we will see double the performance. <br><br>On the other hand, when it comes to browsing the net and performing most GUI tasks, like using PocketWord or doing light web-browsing, I can't tell a huge difference between the PPC and my laptop, which is a 366 Celeron running Win2K. Granted, these applications are stripped down shadows of their desktop versions, but I certainly don't go beyond their capabilities very often. I would point this out as the point where a computer becomes fast enough that it doesn't really make any difference. Once you reach a certain level, the only way you'll notice performance differences is by going to high-performance apps like PQ. The rest of the time, it's fast enough for your reaction, and that's usually all that matters. <br><br>There are some benchmarks over at pocketpcpassion that might help answer your question. What someone really needs to do is grab the source for something like Spec95 and run it on the PPC. If we had a benchmark that would test integer and floating point performance, we might be able to come up with some sort of equivalency in a limited number of situations. Other than that, I think the altered role and usage of the PocketPC means you can't really equate its speed with the desktop. <br><br>Anyone with a Casio want to give a more exact answer?

[Jaybot]

yeah here's my results from<br><br>CPU<br>Floating point: .87<br>Integer: 15.43<br><br>Graphics<br>Bitmaps<br>BitBlt 5.00<br>StretchBlt 5.0<br><br>Filled<br>Ellipse: .83<br>Rectangle 2.82<br>Rounded Rectangle .54<br><br>Memory<br>Allocation 6.40<br>Fill .20<br>Move .13<br><br>Text 3.21<br><br>.....<br><br>Now I'm not quite sure what you can compare htose to on a desktop, but i'm sure there is something.  Also there is Buseful benchmark which gives actual drysone whetstone scores but its a large download.

[Andy_7676]

thanks guys..I was just curious...you guys are a helpful bunch..unlike some other forums I hang out at....

[Moose or Chuck]

who brighthand...

[randall]

Thomas, don't forget the most important feature of the MooseMaster(c) processor: it divides by zero.