We Might Be Getting Doom 3, So What Else Is Possible?


Derek said:
I think some people read to much into my post. I meant a generalized comparison.
Example: ( please forgive me as I have never used glxgears and am unfamilar with it so I don't what exactly it outputs when you run it. )

Lets say you run glxgears on the minimum spec PC running linux
so P3 1Ghz, 256MB Ram, and a Geforce 3 and you got 100fps (reminder I am just throwing numbers up for my example)

then on the Pandora you ran it and the output was only 20fps.

At that point you would have to assume that Doom 3 isn't going to run very well.

That's what I was talking about when referring to benchmarking a min spec P3 vs Pandora. Just a ball park comparison.

I mean if the Panora's CPU can't do this and can't do that, but the P3 can, then it would take one hell of a port to make it run comparable. Unless it was built from the ground up based off of the source...then we wouldn't have the same game would we. ;)

No, people were right. You can't even have a ballpark comparison. It's oversimplifying the subject. A good analogy is say a 100-integer vector, which is the gpu's performance on different areas and a 100-float (ranging 0.0...1.0) vector showing a specific application's needs. The overall performance would be something like the dot product of these two. Using glxgears for testing is like having a vector with one 1.0 while the rest are 0.0. As you can see, you are then dismissing 99% of the gpu's performance.

And no this is not accurate, but it should give you a picture of what you're asking for.
 
Last edited by a moderator:
I thought they HAD released the doom 3 source code already?

CODE
ftp://ftp.idsoftware.com/idstuff/doom3/source/linux/


I remember hearing about this a while back (more than a month so no not the april fools) on some linux boards somewhere and found it in my history... I don't really know what it is tho maybe just an sdk?

Someone with a little more know how should maybe try to download that file and check it out?
 
Last edited by a moderator:
'mitosis' said:
I thought they HAD released the doom 3 source code already?

CODE
ftp://ftp.idsoftware.com/idstuff/doom3/source/linux/
I remember hearing about this a while back (more than a month so no not the april fools) on some linux boards somewhere and found it in my history... I don't really know what it is tho maybe just an sdk?

Someone with a little more know how should maybe try to download that file and check it out?


CODE
http://www.iddevnet.com/doom3/downloads.php


that is just the sdk 3.
 
Last edited by a moderator:
'B-ZaR' said:
'Derek' said:
I think some people read to much into my post. I meant a generalized comparison.
Example: ( please forgive me as I have never used glxgears and am unfamilar with it so I don't what exactly it outputs when you run it. )

Lets say you run glxgears on the minimum spec PC running linux
so P3 1Ghz, 256MB Ram, and a Geforce 3 and you got 100fps (reminder I am just throwing numbers up for my example)

then on the Pandora you ran it and the output was only 20fps.

At that point you would have to assume that Doom 3 isn't going to run very well.

That's what I was talking about when referring to benchmarking a min spec P3 vs Pandora. Just a ball park comparison.

I mean if the Panora's CPU can't do this and can't do that, but the P3 can, then it would take one hell of a port to make it run comparable. Unless it was built from the ground up based off of the source...then we wouldn't have the same game would we. ;)

No, people were right. You can't even have a ballpark comparison. It's oversimplifying the subject. A good analogy is say a 100-integer vector, which is the gpu's performance on different areas and a 100-float (ranging 0.0...1.0) vector showing a specific application's needs. The overall performance would be something like the dot product of these two. Using glxgears for testing is like having a vector with one 1.0 while the rest are 0.0. As you can see, you are then dismissing 99% of the gpu's performance.

And no this is not accurate, but it should give you a picture of what you're asking for.
I get exactly what you are saying but if glxgears has a extreme low fps compared to the low spec P3 wouldn't you be able to assume that most other benchmarks would reflect similar results? I mean some people are saying its not as fast as most netbooks. I mean spec wise that makes since. Well I tried to run Doom 3 on extreme low settins on the Lenovo S10 and it didn't run too hot. Which would lead me to believe that having Doom 3 running at anything playable would be very difficult. thats why I wanted a comparison.
 
Last edited by a moderator:
Performance can't be simplified like that, though.

Let's assume for example, that Doom 3 requires a certain minimum fill rate and a certain minimum vertex rate, and that the Pandora just barely meets both of these.

However, due to having a higher resolution or some other design decision, the desktop that can barely play Doom 3 has a ridiculously high pixel fill rate.

Then you run a benchmark that's like glxgears, where there's a bit of CPU work and a handful [100s] of vertex transformation, but the actual output is in FPS, and is mostly fill-rate-limited.

You wouldn't even be getting a ballpark estimate, you would be getting a comparision of which GPU can run that benchmark better.
So, we certainly can't do anything now, but once Doom 3 is opened, somebody is bound to at least try, and there's not really an accurate way to benchmark whether or not they will succeed.

Incidentally, I'm getting 3500 FPS on glxgears. Core 2 Duo @ 2.16 Ghz, Geforce 7900 GS. Interestingly, the CPU clocked up to the full 2.16 GHz when I ran glxgears. Maybe it's actually CPU-bound.

Anyway, whatever the parameters are, no benchmark other than Doom 3 itself will be able to properly guess how well it will run.
somebody is going to try it eventually, extraneous benchmarking will not help or hinder that.
 
Last edited by a moderator:
Derek said:
I get exactly what you are saying but if glxgears has a extreme low fps compared to the low spec P3 wouldn't you be able to assume that most other benchmarks would reflect similar results? I mean some people are saying its not as fast as most netbooks. I mean spec wise that makes since. Well I tried to run Doom 3 on extreme low settins on the Lenovo S10 and it didn't run too hot. Which would lead me to believe that having Doom 3 running at anything playable would be very difficult. thats why I wanted a comparison.

Granted, it's going to be challenging. I still have high hopes for the little bugger, and will save my judgement until after people have had time to get things going on it. B)
 
Last edited by a moderator:
The CPU clocked up to its full speed simply becasue it was under load.

Also, the pixel fillrate requirement for doom 3 at 800x480 60 FPS is 23 Mpixels/s on PowerVR hardware when there is no transparency visible.

The normally expensive stencil shadowing can be done virtually for free, but with source released the better method of using generalised modifier volumes could be implimented (if the SGX530 supports it) and it be done totally free of pixel fillrate requirements.

23Mpixels/s sounds like nothing, but that is literally the number of pixels being displayed per second, multiple textures may require multiple passes and increase this significantly. Still, I highly doubt it will exceed the fillrate capabilities of the graphics chip.

What I would think would present a far greater obstacle to doom3 on the pandora is the RAM requirements, especially being as the omap uses the pandoras system memory for framebuffer storage does it not?
 
Pandora's framebuffers are only 1.5MB a piece (do you need more than two for something like this?). If you're worried about that amount of memory then things are being cut too close.


I'm sure people would be happy with a variant of Doom 3 that's more similar to the XBox version (which only utilized 64MB of RAM, unified between the CPU and the GPU). I don't know how significantly that varied from the main source tree, though. Or how difficult it'd be to reproduce the design changes.
 
'Mr Poletski' said:
The normally expensive stencil shadowing can be done virtually for free, but with source released the better method of using generalised modifier volumes could be implimented (if the SGX530 supports it) and it be done totally free of pixel fillrate requirements.

There is no such thing as a free lunch. All those triangles still need to be transformed, tiled, and rasterized, and the stencil rate of SGX is not infinite. Modifier volumes are not supported.

QUOTE
23Mpixels/s sounds like nothing, but that is literally the number of pixels being displayed per second, multiple textures may require multiple passes and increase this significantly. Still, I highly doubt it will exceed the fillrate capabilities of the graphics chip.

The important part is the complexity of the shaders used.
 
Last edited by a moderator:
I didn't predict it would run well, rather it would be the limit of the Pandoras ability.

I still think it will be, however, as the Xbox shows, there is a lot of room for more optimisation in Doom3.
 
Last edited by a moderator:
@lulzfish glxgears is not a benchmark.... anything that runs faster than your vsync is a weak benchmark at best and glxgears is one of the worst

the only realy use for glxgears is to tell if acceleration is enabled since it causes a huge difference from sw rendering

The reason glxgears isn't a benchmark is it takes the graphics pipe nearly no time to render it... the only thing it measures is how fast the back buffer can be copied to the front
 
craigix said:
I didn't predict it would run well, rather it would be the limit of the Pandoras ability.

I still think it will be, however, as the Xbox shows, there is a lot of room for more optimisation in Doom3.
I didn't think I was claiming that it would run well. Did it look that way? Because if it did, I'm sorry I didn't word it better.

-God Ginrai
 
Last edited by a moderator:
Might want to change the title of the thread. Every time I see the thread in the forums, i think 'were getting doom 3'? and i can see someone thinking the same way, not looking at the thread and ending up having a misconception
 
Agreed. I really don't think that Doom 3 is guaranteed at all and the thread title is just going to confuse newbies into thinking that Doom 3 is a done deal.
 
'Exophase' said:
Pandora's framebuffers are only 1.5MB a piece (do you need more than two for something like this?). If you're worried about that amount of memory then things are being cut too close.


I'm sure people would be happy with a variant of Doom 3 that's more similar to the XBox version (which only utilized 64MB of RAM, unified between the CPU and the GPU). I don't know how significantly that varied from the main source tree, though. Or how difficult it'd be to reproduce the design changes.
'craigix' said:
I didn't predict it would run well, rather it would be the limit of the Pandoras ability.

I still think it will be, however, as the Xbox shows, there is a lot of room for more optimisation in Doom3.
You both make very vaild points. Heck I forgot about the Xbox running Doom 3. Of course this was probably ported and optimized which is what the very ported to the Pandora would need. The pandora is actually on Par with the Xbox in terms of specs, not sure about raw power though. I actually own the Collectors Edition of Doom 3 for Xbox (got it mainly for the original doom it came with). I have Doom 3 on PC also and it is pretty amazing. I would love to see it run on the Pandora in all (read most) of its glory! I believe that you guys have came up with a handheld that will hold its own in the market for years to come. Get wait to get one.
 
Last edited by a moderator:
arrrgh said:
Agreed. I really don't think that Doom 3 is guaranteed at all and the thread title is just going to confuse newbies into thinking that Doom 3 is a done deal.
Clarified.
 
Last edited by a moderator:
Back
Top