Mips Vs Arm

[DijiTao]

While it is pretty much impossible to know till the actual hardware is out there, I think a lot of us had our hopes up the 200Mhz ARM processor of the GP2X would be very close performance wise to the 333Mhz (let’s not argue about weather or not the psp’s processor is 333Mhz and just accept it is) since “ARM is a more efficient architecture then MIPS”. We’ll I’ve looked and looked and have found precisely nothing to prove that. In fact what I have found is a benchmark run on PDAs using ARM and MIPS processors and the results look to me like Mhz per Mhz ARM and MIPS are about the same.

http://homepage2.nifty.com/qta/bmq/index.htm

So can anybody disprove me here, or can we just expect the GP2X to be about 30% slower then the PSP in terms of raw computing power?

 

[Jarska333]

As my GF uses to say, it ain't about the size, it's about how you use it. B)

 

[Digital Awakening]

The ARM have hardware 2D acceleration and GP2X have 2x200MHz, one of them with less cache.

 

[skeezix]

Its more complex than that of course; (and the PSP, when set to 333, really is 333.. otherwise it'd be fraud, right?)

Some (not all!) ARM-spec processors are quite aggressive.. others, less so. Factor in the speed of the bus they're on, the chip latencies etc etc and you can get a better picture. Now, an older MIPS generally isn't up to par with the latest ARM specs, but its no slouch either.. and can very by operation, heat, etc. I figured the PSP main processor would be more sluggish than it is, but its pretty darned snappy.. especially for floating point for instance. (I don't recall if MPIS has an FPU always or some of the time or what) -- CaSTaway with the exact same CRC32 code as on the GP32 just _flies_, even when clocked down rediculously slow, for CRC32 calculation. Now, the PSP CaSTaway is much faster than on the GP32 (duh), but at the same time.. its not as fast as it could be on another 333MHz processor. I didn't pound-for-pound compare against the GP32 but I could, as I made the PSP version able to underclock too.. so perhaps clocking the same could be interesting. Of course, that throws other thing sinto the mix.. the GP32 being an older ARM spec, and having various oddities due to the GP design.

But of course, the largest factor by far is the fanboism factor

A 200MHz ARM9 is very likely about the same as (say) 250MIPS; sayings its the same as 333 or 400MIPS.. well, who knows?

Go look at comparisons to DEC Alpha to Intel until your head spins.

In short, don't sweat it. It'll feel fastre than a 200MHz, but its still 200MHz, capiche?

jeff

 

[Radek]

While it is pretty much impossible to know till the actual hardware is out there, I think a lot of us had our hopes up the 200Mhz ARM processor of the GP2X would be very close performance wise to the 333Mhz (let’s not argue about weather or not the psp’s processor is 333Mhz and just accept it is) since “ARM is a more efficient architecture then MIPS”. We’ll I’ve looked and looked and have found precisely nothing to prove that. In fact what I have found is a benchmark run on PDAs using ARM and MIPS processors and the results look to me like Mhz per Mhz ARM and MIPS are about the same.

http://homepage2.nifty.com/qta/bmq/index.htm

So can anybody disprove me here, or can we just expect the GP2X to be about 30% slower then the PSP in terms of raw computing power?

​

In a raw CPU power? Potentially the GP2x will be significantly faster if some of the doc's infos are to believed. But until actual hardware shows then it's just a speculation.

As whole machines it's still not that clear as the GP2x's MMSP2 chipset has some hardware coprocessors going far beyond than just the 2d accelerator.

For an example video decoding in the GP2x is actually done by its Video Processor and cpus are just used for controlling and simple data parsing. And then VP of GP2x isn't dump either - it has many functional blocks for maths such as "VLD/VLC, DCT/IDCT, Q/IQ, Prediction, ME, MC" (as stated in docs).

So if you have to do a DCT (Discrete Cosinus Transformation) then GP2x might terrorize the PSP in such case.

 

[DijiTao]

I'm pretty well versed in the Mhz don't mean much gospel, but what I’m really getting at is that elusive “pound for pound” comparison between ARM & MIPS. It seems your saying that a 200MHz ARM chip performances comparably to a 250Mhz MIPS chip and if that’s the case I think some of us are a little disappointed as we were hoping for 200Mhz ARM performing like 333Mhz MIPS.

 

[Digital Awakening]

To the question of why they used 2x200MHz instead of 1x400MHz GPH said that the 200MHz thanks to it's 2D accelleration it would actually outpreform a 400MHz when dealing with 2D stuff. So for what GP2X is meant for, 2D that is, it should outpreform PSP which is designed for 3D and floating point calculations.

 

[carlgeorge]

I never thought that the gp2x had similar performance to the PSP
I don't think I'll be disappointed with the gp2x either.

 

[Radek]

To the question of why they used 2x200MHz instead of 1x400MHz GPH said that the 200MHz thanks to it's 2D accelleration it would actually outpreform a 400MHz when dealing with 2D stuff. So for what GP2X is meant for, 2D that is, it should outpreform PSP which is designed for 3D and floating point calculations.

​

GPH worded this very badly and reason for 2x200 instead of 1x400 is a very simple. The GP2x uses a common hardware (the MagicEyes' MMSP2) not a custom one. So it has exactly the same specs as the MMSP2.

 

[Digital Awakening]

From GPH's FAQ:
"Compared to other cpu like Samsung 2440, itel, AMD, the one that we adopted has 2D accelerator that enables efficiency to be enhanced.
Through the hardware test It is confirmed that the only high clock speed doesn’t make movie or game run fast.
We inform you that 400Mhz without 2d accelerator would rather be lower than 200Mhz."

 

[xzen54321]

To the question of why they used 2x200MHz instead of 1x400MHz

​

Because two heads are better than one? No, realy, dont 2 cpus each at half the clock speed on one faster cpu have better throughput?

 

[junker]

yeah, the comparisons on that site doesn't show much of a difference between the power of pda's with Arm or Mips.
since i've been hearing a lot of good about the Arm architecture, i'd like to see more direct comparisons(unbiased)

 

[skeezix]

The one true answer is.. if you want a PSP, get one; if you want a GP2x, get one; they are not the same, so comparign is basicly useless. ie: If they had the same games, it might be a useful comparison. As for emus.. the gp32 does emus pretty well, so you already know the gp2x will do them very well.

xzen - no, not at all; it 'depends' on the nature of the traffic. Imagine two roads and one car, versus two roads and two cars, versus one car and one road, and two cars and one road. IT depends on what they're doing.

jeff

 

[Radek]

From GPH's FAQ:
"Compared to other cpu like Samsung 2440, itel, AMD, the one that we adopted has 2D accelerator that enables efficiency to be enhanced.
Through the hardware test It is confirmed that the only high clock speed doesn’t make movie or game run fast.
We inform you that 400Mhz without 2d accelerator would rather be lower than 200Mhz."

​

And this is this bad wording I was talking about.
For a tasks the GP2x has to perform it's better to have 2x200 cpus + various hardware coprocessors than a single 400MHz CPU without coprocessors. Is beyond me why the GPH did only talk about this "2d accelerator" when there is far more in the GP2x.

To the question of why they used 2x200MHz instead of 1x400MHz

​

Because two heads are better than one? No, realy, dont 2 cpus each at half the clock speed on one faster cpu have better throughput?

​

Because the chipset of the GP2x is one chip solution. Yes, it's a system on chip (SoC) and clocking the cpu much higher could affect the rest of chip probably. Remember also the main uses of MMSP2 are ones like digital cameras, osd displays of TVs, display panels and such and here fast general purpose processor isn't really needed.

Why then the MMSP2 ended in the GP2x? Cost as it's doubtful the GPH could afford a custom hardware.

 

[Cyclops]

LOL. I hate to be one so say this, but what a load of rubbish.

The PSP is clocked at 222Mhz, saying it isn't just noncence, its like saying the GP32 is 200Mhz. I'm not arguing with anyone. I don't know any sony engineers.

Its a pretty graph, what is it ment to prove?? It compares the speed of PDA's ish on one benchmark. Which might be useful is I was buying a PDA. Although anyone who would buy one based on CPU speed alone is a pratt. As only one small part of buying one is the processer(how cool it looks/does what I need?.Come higher). Where is GP2X and PSP on that list of even this benchmark. What MIPS chip/ARM chips do they have?? Or anything else?

What the graph proves. Is that MHz means next to nothing. Devices on this benchmark clocked at 400Mhz run at the same speed as 200Mhz. Devices run up to twice as fast for the same Mhz on ARM alone, and yet you have a prediction.

This is all without the issues of dual core. I thought this may contain an insight rather than a random link to some noncence benchmark. That has nothing to do with anything. Its behaviour I'd have only expected of DaveC.

Please Please stop pulling things out your ass.

 

[junker]

no one was pulling anything out anyone's ass. true, those graphs don't mean much and aren't a direct comparison, but they're the only thing that was found when searching for Mips/Arm comparison.

and yes, the psp is really clocked at 333mhz. it is usually downclocked to 222mhz for battery conservation. 333mhz is NOT an overclock.

 

[Cyclops]

no one was pulling anything out anyone's ass. true, those graphs don't mean much and aren't a direct comparison, but they're the only thing that was found when searching for Mips/Arm comparison.

and yes, the psp is really clocked at 333mhz. it is usually downclocked to 222mhz for battery conservation. 333mhz is NOT an overclock.

​

Really I must have missread the first post that says, of a benchmark of PDA's in a different language. Someone who formed the conclusion of 30% slower.

No as the PSP uses 222Mhz full stop, and is underclocked intentionally. Rumors are they will release a better battery(I can't belive this is the main reason though), and unlock its potential, but sony in its wisdom and they have engineers/marketing men that quite clearly know there stuff chose intentionally to do this. Its not ususally clocked down its *always* clocked down. Homebrews can take advantage of it but then there is a cost(too high for sony).

But then again look at the graph one chip running 400Mhz running slower(by this benchmark) than another at 200 Mhz, and yet some people(you) are still bleating about Mhz, becuase if there is a point to this pitiful thread that was it.

 

[junker]

homebrew is what we're talking about after all, and nearly every psp prog has support for the full 333mhz speed. we all know it uses much more battery life clocked higher, but that doesn't change the fact that it is 333mhz fullstop. it's not an actual overclock.
what the hell do you think we're talking about here? commercial games on psp or emulators/progs on psp's 333mhz vs same thing on gp2x on one core?

and did you understand me? we know that pda comparison doesn't mean much, and yeah, we know the results don't make much sense! given that, the benchmark info on the site doesn't show clear superiority of Arm based pdas over similarly clocked Mips based pdas.

if you can find direct comparisons or more info on Arm power vs Mips power, then go right ahead and enlighten us on this matter instead of coming in here with your "LOL" and empty insults.

 

[skeezix]

What are you on about?!

The PSP main processor can be software clocked to 333 and below.. and perhaps even higher. At 333 it is running at normal max speed.

Most _commercial_ games are set to much lower, since they're graphcis heavy (GU), and don't need the main processor as much.. so for battery life, Sony recommended running at 222 or less. However, the fact remains, it goes up to 333.. the next perosn who says otherwise gets a slap upside the head.

I'm not the only one on these boards who has coded on the device

I quote:

void cj_psp_setup_clockrate ( void ) {
scePowerSetClockFrequency ( 333, 333, 166 );
return;
}

See that 333 in there?

If I set it to 222, it runs slower. I've timed it

jeff

 

[Cyclops]

homebrew is what we're talking about after all, and nearly every psp prog has support for the full 333mhz speed. we all know it uses much more battery life clocked higher, but that doesn't change the fact that it is 333mhz fullstop. it's not an actual overclock.
what the hell do you think we're talking about here? commercial games on psp or emulators/progs on psp's 333mhz vs same thing on gp2x on one core?

and did you understand me? we know that pda comparison doesn't mean much, and yeah, we know the results don't make much sense! given that, the benchmark info on the site doesn't show clear superiority of Arm based pdas over similarly clocked Mips based pdas.

if you can find direct comparisons or more info on Arm power vs Mips power, then go right ahead and enlighten us on this matter instead of coming in here with your "LOL" and empty insults.

​

LOL. really you should read my posts. I never ever use the word overclocked. I use the word underclocked. The only time I mention otherwise is with referance to Homebrew, although I must point out I was the first to mention it. Actually we are talking about comparing PDA's with little to no relevance to PSP or the GP2X, and nothing at all to do with the chips in question.

Do I understand you, yes you have so little manners as not to even read my post and make a rant irrelevant to it. The comparison of ARMS and mips has been done to death. There are oodles of DS vs PSP posts all over the net. In other threads I have suggested keywords. They makes for interesting reading. I have also posted results of other benchmaking of the PSP.

Don't waste your life read my post. I draw conclusions from that data. Because I can read a graph.
Under this measure of processing ummph Mhz is a waste of time even if talking about the same processer ie ARM vs ARM. My point is its noncence.

Reread my post.

What are you on about?!

The PSP main processor can be software clocked to 333 and below.. and perhaps even higher. At 333 it is running at normal max speed.

Most _commercial_ games are set to much lower, since they're graphcis heavy (GU), and don't need the main processor as much.. so for battery life, Sony recommended running at 222 or less. However, the fact remains, it goes up to 333.. the next perosn who says otherwise gets a slap upside the head.

I'm not the only one on these boards who has coded on the device

I quote:

void cj_psp_setup_clockrate ( void ) {
scePowerSetClockFrequency ( 333, 333, 166 );
return;
}

See that 333 in there?

If I set it to 222, it runs slower. I've timed it

jeff

​

Read my post not the reply. You have only repeated back to me what has been said. The conclusion is as always there is a price to be made for that power. If the case is less battery life, how bad is it. Its not like I can swap the batteries. If comparions are being made and battery life is an issue and it clearly is(games having hidden features to fully uterlise the CPU). Then the two should be mentioned together as they are not mutually exclusive. Its why people choose one laptop over another, or why(one of the reasons) the original gameboy was so success.

The reality is and if the data given is anything to go for, Mhz is as useful a measure as how many mice on bicycles there are in your device. Its a guage of processing power, and increasing a bad one. People have GP2X and they have PSP why not benchmark them lets have a Dhrystone Benchmark.