Carmack to use shadow map in the next game?

[Scali]

So pointing out that you neglect that adding the code you would like to see in DOOM3 would require additional reprogramming of the game, thus adding a significant amount of work, qualifies as "harassment"?

No, but wrong assumptions such as this one do.

Why post here in the first place if you are not prepared to consider the input and position of others? Did you simply want an audience for your whining?

Again you wrongly assume I am not prepared to consider the input and position of others.
I am not prepared to hear the same stuff I've been hearing for weeks, and repeating myself over and over again.

Forget it, I'm just going to email Carmack himself and see if he will answer me. At least I can have a technical discussion with him. I doubt that he will just tell me that I need to upgrade my CPU, or give some lousy fake excuses.

 

[Cat]

Your code that you wrote is homebrewn, whether it was for commercial or personal gain. Once again, you've told us all that we're not qualified to comment. Why are you? I also write software for a living, and some of it *gasp* even uses OpenGL...and VBOs, and fragment programs, and pbuffers, and ... I'm not an expert on the subject, but neither am I unable to grasp why people might do things the way I do, or do not.

http://steampowered.com/status/survey.html Many people that play games have fast CPUs. Make what you will of the PR ratings of processors and how they fit into this graph.

For the third time, it's a discussion forum, you asked for discussion, and you're getting it.

We all gave you technical reasons for not using a fully GPU-skinning extruding system, but you've just decided to dismiss them. If you really want to beat your chest like a gorilla, why don't you go over to gd-algorithms and take your monologue--- I mean discussion, there.

 

[Scali]

Your code that your wrote is homebrewn, whether it was for commercial or personal gain.

If that is really the case, why did you bother to mention it at all?
I think you wanted to imply that it was just amateur/hobby code.
I get the feeling you are trying to provoke.

Once again, you've told us all that we're not qualified to comment.

I didn't say you weren't qualified to comment. I said you weren't qualified to answer why John Carmack designed his code the way he did. You are just twisting my words yet again.
Again I get the feeling you are trying to provoke.

http://steampowered.com/status/survey.html Many people that play games have fast CPUs. Make what you will of the PR ratings of processors and how they fit into this graph

As said before, that data is useless, since it doesn't show any relation between CPU and GPU. Since the claim is that my PC with 1800+ and Radeon 9600Pro is a unique combination, you will need data with such a relation to statistically show that this is the case.
If we look at the groups that are closest to my CPU, the 1.5-1.7 and 1.7-2.0 GHz, we can see that these groups are among the most popular.
Then we look at Radeon 9600-based PCs... We see that again this is one of the more popular cards.
So it is entirely possible that there are combinations of such popular CPUs and GPUs. There could even be quite a lot of them!
But we have no idea how representative this is for Doom3, since these are HalfLife statistics. HalfLife requires entirely different hardware. It doesn't require a fast videocard at all. And we do not know how representative HalfLife players are for Doom3 players.
So these figures are useless to prove anything.

We all gave you technical reasons for not using a fully GPU-skinning extruding system, but you've just decided to dismiss them.

Yes, since none of them convinced me.

If you really want to beat your chest like a gorilla, why don't you go over to gd-algorithms and take your monologue--- I mean discussion, there.

Again, wrong assumptions, arrogant remarks... completely uncalled for.
I wanted answers. I have mailed John Carmack, perhaps he will give them. If all you are going to do is provoke and insult me, I prefer you just don't post at all.

 

[Cat]

I don't think the CPU measurements are in PR ratings. I think they're actual GHz measurements. About half of those are AMD processors. If we're still dealing with GHz, then the majority of those processors are rated at least at 2500+. ( ~1.8GHz )

So, almost every AMD owner has a 'fast' CPU, and at least half of the Intel owners have a 'fast' CPU.

If most people have fast CPUs, then you are most definitely in a small minority, and indeed, a special case.

 

[Scali]

I don't think the CPU measurements are in PR ratings. I think they're actual GHz measurements.

You THINK? Well that's rather useless.
What if I THINK they're not?

About half of those are AMD processors.

Since when does AMD have a 50% marketshare?
Or do you just THINK that again?

If most people have fast CPUs, then you are most definitely in a small minority, and indeed, a special case.

Nonsense.
"most people" could be as small as 51%.
"small minority" could be as large as 49% in that case.
Special case? Not at all.
As I said before, these numbers prove absolutely nothing.
They prove even less, since we don't have an actual figure that decides when a group is large enough to support.
Just give it up, I've seen the Steam numbers before, I don't want to repeat myself yet again. You can't prove anything from those figures.
But you don't want to prove anything.... You just want to provoke, right?

 

[Cat]

I 'think' it's measured in GHz, because it says it is.

There is a clearly labeled category that tells you the Processor Vendor on that page. Half of the reported CPUs are made by AMD.

This is the largest survey that's available to us. It's not perfect.

 

[Scali]

I 'think' it's measured in GHz, because it says it is.

Yes, but you don't know for sure if it uses PR to correct for AMDs, since that is not explained.

There is a clearly labeled category that tells you the Processor Vendor on that page. Half of the reported CPUs are made by AMD.

Yes, but again there is no relation between processor speed and brand. So you cannot make any kinds of conclusion about the marketshare in any clockspeed category. It is highly unlikely that there are that many AMD CPUs in the higher regions if it were measured by actual clockspeed, since AMD never produced a CPU faster than ~2.5 GHz.
Just give up, these figures are useless. They don't apply to Doom3, and they don't contain the relationships between CPUs and GPUs required for this discussion.
And why are you even bothering to discuss the brand of the CPU? That is completely irrelevant anyway, in this discussion.

 

[Cat]

It makes more sense to me that they're gathering actual processor speed, and not PR ratings, as that's what written, and that's what's easiest to measure remotely. But yes, you're right, I don't know this for sure.

The processor brand comes into play when you're talking about clockspeeds and varying IPC, and have only clockspeed numbers. Obviously an AMD processor at 1.8GHz is not going to be as slow as a P4 1.8A.

If enough people have CPUs that are fast enough for your skinning/shadow cutoff, then I don't think it matters what GPU they have.

You've decided this survey is meaningless. I don't think it is, but I obviously won't be able to convince you.

 

[Scali]

The processor brand comes into play when you're talking about clockspeeds and varying IPC, and have only clockspeed numbers. Obviously an AMD processor at 1.8GHz is not going to be as slow as a P4 1.8A.

A 1.8 GHz AMD processor is also more expensive than an Intel one, and it also has not been available as long as the Intel one. So it is quite unlikely that it is as widespread as the Intel one.
That's the whole problem with AMD and Intel today. Clockspeed is no longer a measure for performance or price. Which is why PR was introduced in the first place.

If enough people have CPUs that are fast enough for your skinning/shadow cutoff, then I don't think it matters what GPU they have.

I do, since you can only get ~100 fps average on the fastest Athlon64s with a 6800U. 3dmark03 gets ~100 fps average on any CPU, as long as it has an 6800U.
So there's still a difference between "fast enough" and "getting top performance from the system", even for CPUs that make the game playable. But then the issue is less critical, I suppose.
It is even quite possible that the ~100 fps figure is still very much CPU-limited, so even on today's fastest CPUs, Doom3 may run faster with GPU-acceleration on the fastest GPUs.

 

[Blastman]

I’m gaming on a Celeron-T 1.1 -->1.54 with a 9600XT. My 2 brother in laws have a P4-1.7 & 1.8A. It’s pretty hard to gauge the overall gaming community because a lot of online surveys catch more of the gaming “enthusiasts”, who are much quicker to upgrade than the causal gamer like myself or brother inlaws.

The recent game demos I’ve tried -- Farcry, Painkiller, COD, 1942 all played well on my system. (Farcry was a little taxing but generally above 30fps.) I could upgrade tomorrow if I wanted, but I will use this system for as long as possible because it runs everything else that I do outside gaming well. I have a HL2 coupon in my drawer … …and expect it to play well on my system. Judging by what scali has said, Doom3 would kill my system

 

[Reverend]

I said you weren't qualified to answer why John Carmack designed his code the way he did.

I'm not the person you addressed this to but I hope to be "qualified" enough to ask you if you have tested skinning on a GF4MX and compared that to a sub-2GHz (non-AMD rating) CPU? Take into account the speed of CPUs that GF4MX owners likely own (of which I just gave you the answer anyway). And then take into account other GPU & CPU rendering components of Doom3. Put them all together.

I don't know the contents of the email you sent to John. If it was "technical", it would've been a mistake. Better to just ask him "What is your base target market -- GPU, CPU -- when it comes to skinning?".

I'm not questioning the fact that skinning on the video card would be a better option given the current industry state. I am questioning if you know certain rationales behind some of the decisions (like this skinning issue) taken by not only John but by the entire id team.

 

[Scali]

I'm not the person you addressed this to but I hope to be "qualified" enough to ask you if you have tested skinning on a GF4MX and compared that to a sub-2GHz (non-AMD rating) CPU? Take into account the speed of CPUs that GF4MX owners likely own (of which I just gave you the answer anyway). And then take into account other GPU & CPU rendering components of Doom3. Put them all together.

Not sure what you mean, since GF4MX obviously cannot perform skinning at all (not in a way that is useable for Doom3 anyway), so you MUST use a CPU path there.
For the rest: as I said many times before, I have not seen any kind of reliable data for what CPUs are commonly paired with what GPUs.

Better to just ask him "What is your base target market -- GPU, CPU -- when it comes to skinning?".

That's a different question. We already know the answer to that: high-end CPUs, paired with NVIDIA GPUs of any kind, or ATi 8500 and up.
That's not what I want to know.
I don't even want to know why this is the base market, since we already know that aswell: these are popular combinations in prebuilt PCs.

I'm not questioning the fact that skinning on the video card would be a better option given the current industry state. I am questioning if you know certain rationales behind some of the decisions (like this skinning issue) taken by not only John but by the entire id team.

No I don't, that's the entire point. Apparently nobody gets that.
Only John knows this, so only John can answer why he ignores the features of modern GPUs, which does two things:

1) screws the enthusiast gamers out of getting the optimum performance from their expensive GPUs.
2) screws all Doom3-engine based games out of using reasonable polycount (6-sided sodacans are ridiculous, come on).
As we all know, GPUs scale much faster than CPUs... since Doom3 barely runs with its lowpoly models today on today's high-end CPUs, we can assume that polycount cannot increase all that much over the coming years in this engine).

So, did they consider this? Did their decision backfire? Etc etc.
Only John knows (and possibly other id developers, but afaik, John is the lead programmer/architect, so he should ultimately make such decisions).

 

[Cat]

For future titles, a vertex program skinning/volume path could be added, if they wanted to. You yourself said it was fairly easy.

For the current title, rendering is capped at 60FPS, so there's no real benefit with the technique, unless a person is in your situation (slow CPU, fast GPU.) This just circles back to the murky userbase statistics issue.

I also loaded up the test box map, and spawned at least 20 imps on a P4C 3.0. My framerate did not dip below 60FPS if I noclipped out of the map, and backed far away from them. Perhaps your system's slowdown is partially related to fillrate.

At least 20 imps running at 60+FPS on a 3GHz machine vs 3 imps at 10FPS on an 1800+ seems odd. I'm assuming you have vsync disabled...

 

[KimB]

There may be LOD differences if you back too far away.

 

[Cat]

I don't think there's any LOD, be it discrete or continuous.

 

[Scali]

Perhaps your system's slowdown is partially related to fillrate.

Nonsense, since as I said before, 640x480 low detail or 1024x768 high detail makes absolutely no difference whatsoever.
You can see the same trend here: http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2149&p=7

All the CPUs below 2.6 GHz get exactly the same framerates in 800x600 as in 1280x1024, regardless of the considerable fillrate differences in these resolutions, especially with all the renderpasses for stencilshadows and lighting. Completely CPU-limited (and all getting CONSIDERABLY lower framerates than Battle of Proxycon does on a 6800U, namely ~100 fps. Now if you had a 2000+ with an 6800U, and got 100 fps in Battle of Proxycon, wouldn't you feel screwed when Doom3 only gets 46 fps on the same system? Battle of Proxycon clearly shows that the system is well capable of more).

You can also see what speed a Radeon 9500Pro or 9600XT get on a 2500+ here: http://www2.hardocp.com/article.html?art=NjQ0LDY=
My 9600Pro should be somewhere in between. They don't seem to have a lot of fillrate problems.

As I said before, one factor could be that I have an ATi card. Perhaps an NVIDIA card would require less CPU-usage. But that is most probably still related to the dynamic vertexbuffers required for the CPU-based method then. But still, that 2000+ with the 6800U only gets 46 fps on average, which I suspect is the same as in my case, namely severe drops for multiple characters, and extreme peaks when there is nobody around. The peaks are just more extreme on 6800U so the average is higher, that doesn't say anything about whether actual combat is any faster though.
For that same reason I played the game in 1024x768. It was slower in the fast parts, but eqally slow in the slow parts, so it was still equally playable.

At least 20 imps running at 60+FPS on a 3GHz machine vs 3 imps at 10FPS on an 1800+ seems odd. I'm assuming you have vsync disabled...

vsync doesn't make a lot of difference since it doesn't get anywhere near the monitor refresh, obviously.
And if you don't believe me, get an 1800+ yourself (a REAL one that is, not a modern system that is underclocked to 1800+, but a real one with a board from that era (VIA KT133/266 or nForce1)) and a Radeon 9600Pro, and see.

 

[Cat]

I believe you, I just find it odd.

 

[_xxx_]

It might have been posted already, but I'm too lazy too read all the stuff...

Here's Carmack's Quakecon speech (transcribed, so you don't have to bother with that 600+ MB vid), he also speaks about shadowing and why he chose stencil instead of maps:

http://www.gamedev.net/community/forums/topic.asp?topic_id=266373[/b]

 

[Ragemare]

That's my point. You have a CPU which is only barely above the min req while your gfx card is mainstream (high end if you consider the valve survey Wink. Your system would benefit from hardware generation of shadow volumes but how many others are in your situation?

There are many parts in doom3 where I get a 100% performance boost by switching shadow volumes off... and then there are parts where it makes no difference atall.

I have a 2800xp with 9800pro running @ 800x600 on high quality (with the H-tweak). Moveing the rez up knocks about 5-10fps off.

 

[KimB]

That's my point. You have a CPU which is only barely above the min req while your gfx card is mainstream (high end if you consider the valve survey Wink. Your system would benefit from hardware generation of shadow volumes but how many others are in your situation?

There are many parts in doom3 where I get a 100% performance boost by switching shadow volumes off... and then there are parts where it makes no difference atall.

I have a 2800xp with 9800pro running @ 800x600 on high quality (with the H-tweak). Moveing the rez up knocks about 5-10fps off.

Just because there's a performance difference doesn't mean you're CPU bound in these cases. Shadows eat up quite a bit of fillrate, too.