Mips And Gflops


TylerAW said:
Tom` said:
On the other hand, memory size, past a certain point, matters much less than speed and latency - IIRC the Pandora suffers a bit in this area (although maybe this is mitigated by the L2 cache, that isn't present, at least, on older ARM platforms).
So the Pandora has latency issues, how bad are they?
Is it the cache, ram speeds, design etc...?

PS More RAM in my book means extra performance than less RAM :p .

supposing that you already have 4Gb and could add more than 4Gb, and you had win XP installed,
then NO.

there is a minimal point for smooth operation, and beyond that it would be an overkill.
 
Last edited by a moderator:
The PSP has 32MB of RAM if you add an extra 100 lets say than you can get better performance in certain tasks :p
 
but it would be still behind the PS2,
even after the 100+

see!!
you just tripped by yourself.
 
DroneB Dev said:
but it would be still behind the PS2,
even after the 100+

see!!
you just tripped by yourself.
WHICH IS WHY I SAID MAYBE THE Ps2 is better than Pandora and maybe its not.
 
Last edited by a moderator:
so?, im not arguing about that, but about the use of RAM in that comparation.
 
TylerAW said:
Tom` said:
On the other hand, memory size, past a certain point, matters much less than speed and latency - IIRC the Pandora suffers a bit in this area (although maybe this is mitigated by the L2 cache, that isn't present, at least, on older ARM platforms).
So the Pandora has latency issues, how bad are they?
Is it the cache, ram speeds, design etc...?

PS More RAM in my book means extra performance than less RAM :p .
Short answer: it's an OMAP issue you'd have to ask Exophase :p
Not too bad, I don't think, but it might be a bottleneck for some things.

And more RAM is not necessarily better - look at the Wii, for example - it's only got 88 MB of total RAM, but it's quite fast and overall probably more powerful than the Pandora by a decent amount.
 
Last edited by a moderator:
Tom` said:
TylerAW said:
Tom` said:
On the other hand, memory size, past a certain point, matters much less than speed and latency - IIRC the Pandora suffers a bit in this area (although maybe this is mitigated by the L2 cache, that isn't present, at least, on older ARM platforms).
So the Pandora has latency issues, how bad are they?
Is it the cache, ram speeds, design etc...?

PS More RAM in my book means extra performance than less RAM :p .
Short answer: it's an OMAP issue you'd have to ask Exophase :p
Not too bad, I don't think, but it might be a bottleneck for some things.

And more RAM is not necessarily better - look at the Wii, for example - it's only got 88 MB of total RAM, but it's quite fast and overall probably more powerful than the Pandora by a decent amount.
Good comparison but than again the Wii is not running a Multitasking OS. And if you mean better in 3D terms the GPU is faster and backed by expensive tools we don't have for the Pandora. Also Wii is a Gamecube with DDR3 memory. I mean I think it's safe to say the Gamecube is faster than Pandora. I never said more ram means superior I was just saying RAM helps, since FLOPS and MIPS and MHz don't seem to?
 
Last edited by a moderator:
There are many factors to how fast an application runs on a hardware. Here, let me simplify why what you are asking is so difficult to answer:

Consider two systems. Because of their differences, they perform differently at different tasks, like integer math, floating point operations, branching and so on. This gives us a profile of the system's performance.

Let's assume that the profiles for the two systems are (in vector form):
System 1: [4 5 6 1 2 3]
System 2: [1 1 3 8 2 6]

Now, each application uses the system differently. This is why their performance depends more on some features and less on some.

Let's assume three pieces of software, with feature usage profiles like this:
Application 1: [8 7 6 5 4 3]
Application 2: [3 2 1 9 8 7]
Application 3: [5 5 5 5 5 5]

To get the sum performance, we multiply the corresponding numbers in a system profile and an application profile, then get the sum of those products (dot product):
Application 1 @ System 1: 32 + 35 + 36 + 5 + 8 + 9 = 125
Application 1 @ System 2: 8 + 7 + 18 + 40 + 8 + 18 = 99
Application 2 @ System 1: 12 + 10 + 6 + 9 + 16 + 21 = 74
Application 2 @ System 2: 3 + 2 + 3 + 72 + 16 + 42 = 138
Application 3 @ System 1: 20 + 25 + 30 + 5 + 10 + 15 = 105
Application 3 @ System 2: 5 + 5 + 15 + 40 + 10 + 30 = 105

See? System 1 performs better in application 1, system 2 in turn in application 2. Both perform equally in application 3. Which system is more powerful? Of course if one system is better across the field, the comparison is much easier (like comparing the i7 setup with an omap device). The thing is, in many cases the comparison depends very much on the application.

(Disclaimer: the above is a rough simplification to raise a point, not an accurate model)
 
SO couldn't we compare consoles like Pandora and Dreamcast by technical specs one by one?
 
TylerAW said:
SO couldn't we compare consoles like Pandora and Dreamcast by technical specs one by one?

Why would we want to, What would be the reason and what would actually be achieved ?, Willy waving ?.

Trooper
 
Last edited by a moderator:
The 'numbers' give a good idea of what hardware is best at, and how fast it is compared to similar hardware. It wont tell you how good it is compared to dissimilar hardware. A mini cooper can get you across a busy city centre faster than a ferrari. So does that make it *faster*? I think the FLOPS highlights this the best, if the program uses fixed point arithmetic, it doesn't matter how fast the FPU is, if the program doesn't use it. The other 'numbers' have similar properties (though its more about curves than off/on).

It sound like you want a generally overall 'x' is better than 'y'. For this you really want is a set of benchmarking figures, (done by someone who knows their onions). I'm sure there are benchmarking figures, but I don't know if anyone has done a comprehensive benchmarking across the systems you are interested in. Of course no benchmarking can cover everything, but its the best you will get.
 
Wow what a bloodbath ... (this is a dangerous place for a Medic to tread)

PS More RAM in my book means extra performance than less RAM

True, but past a certain point it just means its no longer the bottleneck.
Remember you are only as fast as your slowest component (or was it "as strong as your weakest link"?)


On the other hand, memory size, past a certain point, matters much less than speed and latency

Just confirms what I just said before.


this is like counting cars horsepower comparing with the size of the fuel tank

Haha lol. The analogy is actually true (backed by an example: Wii)


DroneB Dev said:
did you understand what i am saying?

I actually do. It is impossible to compare the "power" of various different CPU's ACCURATELY without knowing what to define and how to define it.


you all keep saying comparing is impossible

And they're right, but so are you. I got the feeling you were seeking the power of various systems to accomplish 3D graphics, am I right?
(no you idiot not like Avatar 3D)
They're right because you can't use Mips & Gflops to compare what you perceive as a powerful system.
But you're right as some systems are more powerful than others at certain tasks
(and sometimes at all tasks ie; my laptop can emulate GBA games 300% fullspeed something the GBA cannot)


Voila!

*snip...Highest being better at graphics than the lower ones.
1. i7 Gaming Computer
2. PS3
3. Xbox 360
4. Xbox
5. Wii
6. Gamecube
7. PS2
8. iPod Touch/3GS
9. Pandora
10. PSP
11. Dreamcast (maybe it stays about here or maybe its better than Pandora)
12. Nintendo 64
13. PSOne
-------------------------
14. DSi (afaik)
15. SNES (afaik)
16. Sega Genesis (afaik)
17. Commodore 64 (afaik)
18. Atari (afaik)


CCCUUUTT!!!
This is off-topic.
I think what OP (original poster and openpandora) wanted to know was:
How can we determine the "power" aspects of different CPU systems?
SO couldn't we compare consoles like Pandora and Dreamcast by technical specs one by one?
Not by technical specs (too difficult and mostly guess work) but by actual testing more accurate "point system" may be better alternative.
SO What does that mean = benchmarking. (3D benchmarking tools and WinSAT! anyone?)
Now to device a benchmarking system that can be applied to a wide range of devices ... (near impossible)
So if we don't have a benchmark, how can we even compare it?
We use technical details of the systems, the type of task/app we want tested, observe them in action... then add in common sense!


It is not just about power but also about who has the better software that takes advantage of this.
 
Last edited by a moderator:
trooper said:
Why would we want to, What would be the reason and what would actually be achieved ?, Willy waving ?.


I think that is a dumb question/reply (not implying anything about your iQ).
The more information we have the better, in all geeky circumstances.
And the more accurate the info the better.

I still understand your point: What the hell do we do with these figures?
Well if it is about consoles, it would make it more clear to coders on the emulator.
"Hypothetical:
I've figured the PS2 has a Kangal rating of 36,000.
If I can write a very good emulator, the minimum hardware necessary to emulate this would need to be 2.8x this to run fullspeed.
There is no Atom line that has a Kangal rating of 100,800.
There are many C2D (P-series), i5 and i7 that reach this figure.
The next Pandora has the new Dual-2GHz-core ARM Cortex A9 on 30nm with SGX543MP8 and 1GB DDR RAM.
This achieves 88,000 Kangal rating, so it would only pump-out 52fps max.
Should I continue coding...?"
 
Last edited by a moderator:
Kangal said:
trooper said:
Why would we want to, What would be the reason and what would actually be achieved ?, Willy waving ?.


I think that is a dumb question/reply (not implying anything about your iQ).
The more information we have the better, in all geeky circumstances.
And the more accurate the info the better.

I still understand your point: What the hell do we do with these figures?
Well if it is about consoles, it would make it more clear to coders on the emulator.
"Hypothetical:
I've figured the PS2 has a Kangal rating of 36,000.
If I can write a very good emulator, the minimum hardware necessary to emulate this would need to be 2.8x this to run fullspeed.
There is no Atom line that has a Kangal rating of 100,800.
There are many C2D (P-series), i5 and i7 that reach this figure.
The next Pandora has the new Dual-2GHz-core ARM Cortex A9 on 30nm with SGX543MP8 and 1GB DDR RAM.
This achieves 88,000 Kangal rating, so it would only pump-out 52fps max.
Should I continue coding...?"

If you want to compare systems like this using the technical specs, i suggest something like the following:-

  • Compile a list of all the numbers can be defined and compared for each system. (processor speed, memory size etc)
  • Adjust the figures with an appropriate factor for each architecture type, so they are comparible.
  • give each statistic a 'weighting' which is a figure on how important you think it is. Multiply all the statistics by their weightings and sum up for each architecture.
  • Compare results against any benchmarks you can find, and adjust the weighting until the figures fit.

The result will be crude of course, but will have the advantage of giving some sort of quantitative value of unreleased or untested hardware.
In practice i suspect the exercise of compiling the figures will impart much more information than the actual end result.
 
Last edited by a moderator:
:unsure:










Seriously... is this really necessary? Good software makes a good system. Who cares about hardware? Even if the Pandora is "better" than a PS2 [hardware wise], you won't find the quality titles from EA, THQ, ... etc on the Pandora (not now, probably never ;) ).
And Kangal: Have fun with your "i7 Gaming Computer" - how does the i7 contribute to graphics except for compiling shaders and forwarding graphics? Just say "Gaming Computer" next time - or just drop the entire list directly as all of these lists are useless.
We have seen enough Pandora videos - we can judge how fast the system is. "Calculating" the speed of a system might make sense for marketing reasons to show others how fast your system is approximatly. However, even if your benchmark performs super-good but your real-world application doesn't even though it should, you are fucked.
//Edit: Lulzfish: I *think* you mixed up VFP (Vector floating point) and NEON (which is the marketing name, its called "Advanced SIMD" in the docs).
//Edit: And if anyone wants to compare them for emulation reasons - don't waste your time. Spend your time writing an actual emulator to see how it turns out. Some systems can be HLE'd so much that I can't even consider it emulation anymore but more or less interpreting the rest of the binary (much more like a script). Others have add-on hardware which add so much to the system that you can't even know the real performance unless you name a specific piece of software. (Not to mention that you can interpret, make a dynarec, a static recompiler, come up with your own stuff depending on the system). Even with the best benchmarks in the world with the most comparable functions its impossible to have "2.8x" or "7.0x" rule. PERIOD.
 
JayFoxRox said:
And Kangal: Have fun with your "i7 Gaming Computer" - how does the i7 contribute to graphics except for compiling shaders and forwarding graphics? Just say "Gaming Computer" next time - or just drop the entire list directly as all of these lists are useless.
As far as i'm concerned he could have said I9, and ATi5890 In a Crossfire X4 setup. The I7 is the most capable series of desktop CPU's available today, but you cant really claim it to be the most price effective. I'm assuming he was trying to get the point across, that todays most powerful folding rigs is indeed at he top of his list. Oh, sorry, folding, gaming... Same same. You still need the CPU power every now and then you know...
Your rant on the issue seams mostly childlike, and unfit. You don't agree with the list, fine, ignore it.
B!
 
Last edited by a moderator:
JayFoxRox said:
Seriously... is this really necessary?

necessary? desireable? well not for me it isn't, but whos to say why he wants to figures? Maybe it *is* for marketing (e.g. marketing to his dad why he should buy one). Im not qualified to comment on the motivation.
 
Last edited by a moderator:
Mr B said:
JayFoxRox said:
And Kangal: Have fun with your "i7 Gaming Computer" - how does the i7 contribute to graphics except for compiling shaders and forwarding graphics? Just say "Gaming Computer" next time - or just drop the entire list directly as all of these lists are useless.
As far as i'm concerned he could have said I9, and ATi5890 In a Crossfire X4 setup. The I7 is the most capable series of desktop CPU's available today, but you cant really claim it to be the most price effective. I'm assuming he was trying to get the point across, that todays most powerful folding rigs is indeed at he top of his list. Oh, sorry, folding, gaming... Same same. You still need the CPU power every now and then you know...
Your rant on the issue seams mostly childlike, and unfit. You don't agree with the list, fine, ignore it.
B!

Did you notice the "better at graphics" part before the list? It would have made a slight bit of sense if he said something like: ATI### or GTX### ... Gaming Computer. But throwing the CPU in the discussion was useless.
This lets me conclude that the list is just "made up" - making it useless. The claims are not even backed up nor does Kangal seem to understand what it takes for good graphics.
There are so many techniques to make graphics look good. Just check out what demosceners do on old Amigas now and what Nintendo did on the N64. Some of that stuff is really awesome and yet it had nothing to do with the systems raw performance but the tricks used to create great visuals.
 
Last edited by a moderator:
WombleGoneBad --> thanks for your post

Mr B said:
The I7 is the most capable series of desktop CPU's available today... I'm assuming he was trying to get the point across, that todays most powerful folding rigs is indeed at he top of his list. Oh, sorry, folding, gaming...
Your rant on the issue seams mostly childlike, and unfit. You don't agree with the list, fine, ignore it.
B!

I just put the "core i7 gaming computer" because it is not difficult to find a computer today which outpaces the graphical performance of the PS3. If I had just put PS3 at the top, perhaps you would've still replied and told me off as a sony fanboy. It was just there to make a point, yes I already know its the GPU not the CPU its the combination of CPU and GPU that determine if it can play Crysis, I was just too lazy to type a kick-ass GPU up there (so many to choose from). BTW, thanks B! for getting it and I do think that rant was childlike and unfit.

JayFoxRox said:
Did you notice the "better at graphics" part before the list? It would have made a slight bit of sense if he said something like: ATI### or GTX### ... Gaming Computer. But throwing the CPU in the discussion was useless.
This lets me conclude that the list is just "made up" - making it useless. The claims are not even backed up nor does Kangal seem to understand what it takes for good graphics.
There are so many techniques to make graphics look good. Just check out what demosceners do on old Amigas now and what Nintendo did on the N64. Some of that stuff is really awesome and yet it had nothing to do with the systems raw performance but the tricks used to create great visuals.

Again I used "i7" and "Gaming Computer" to save time. I mean a true geek would understand the meaning of the point instead of arguing its details like Dr Sheldon Cooper. That list is there as to also prove a point. We can compare the performance ... by our perception. It is a crude method, but one that relies on observation, and as a scientist I know physical observations are valid parameters to be able to make predictions and conclusions. And Fox one more thing, the claim is backed ... by my observations! So you can take it as a grain of salt if you please or reference it for a journal, it does not affect my stance. I dont know many things, infact easier to say nothing about how software in computers work. So their may be techniques to make some systems look better than others and I'm not arguing that.
To you this list may be useless (as you stated), then just ignore it or say I think this list is useless instead of a long long long Dr Cox rant. But here's a kicker: Can you honestly say that an Xbox 360 has less graphical power (or however you state it) than a PS2, as to contradict the list?
Do you have any proof that disproves the list?
I don't have the patience to find hard proof for the list at this time, but I believe many people would be satisfied by the order.

Even with the best benchmarks in the world with the most comparable functions its impossible to have "2.8x" or "7.0x" rule. PERIOD.
Agreed :lol: but it is humiliating to see how literal you took a "hypothetical" scenario. I was just trying to say it isn't completely useless to have that extra piece of information, and was trying to back that up with an example. It's a shame if you didn't understand it.

Your move Fox. (fix'd)
 
Last edited by a moderator:
Kangal said:
You're move Fox.

Your! YOUR! :angry:
Now you're saying "You are move Fox", which I don't understand. You probably wanted to imply that it is Fox's move, in which case you use the form "your".

Sorry. I've seen too many of these lately, and just snapped at this one. Nothing personal. :p
 
Last edited by a moderator:
Back
Top