Half-Life 2 will not feature MRM Technology

[$niper]

Valve is not going to be using the technology they licensed from INTEL for Team Fortress 2 in Half-Life 2.

Instead of using INTEL's dynamic LOD based architecture, Valve is going to be going with a traditional STATIC LOD system.

Problems which arise from STATIC LOD are:

- Visual "popping" when switching between LOD models
- Increased workload. Instead of just needing to create one model with INTEL's tech, modellers will need to create - according to Valve - 3 to 4 LOD models.
- Your disk space and memory is wasted on multiple static LODs.



For information on INTEL's MRM Technology that was planned for Team Fortress 2: http://www.cdmag.com/articles/018/102/tf2_mmt.html

For Team Fortress 2's Website MRM information: http://games.sierra.com/games/teamfortress/models.html

For INTEL's official MRM Technology website: http://www.intel.com/labs/media/3dsoftware/mrm.htm

This also explains the visual "popping" the rocks have on the Half-Life 2 E3 Dune-Buggy video.


Discuss.

 

[LoneDeranger]

As long at it doesn't look anything like the LOD transitions in Vice City, I'm gona be happy.

 

[adio2001]

Has Valve said anything as to why they went to static LOD instead of MRM? I wonder if a MOD could use MRM instead of static LOD. Would it be possible or would you have to license the engine to change something like that?

 

[$niper]

Perhaps not as bad as Vice City, but there will be visual popping in-game.

 

[Chazz]

Now, *which* visual popping are we talking about? Like when something goes from a distant "low-res" model to a nearer "detailed" one? Or something else taht I'm not noticing or something I am but just ain't matching to the term?

 

[adio2001]

Well if you watch the e3 vid during the Dune-buggy part, when he drives past the rocks, you can see them sort of shift. It's not that big of a deal in my opinion.

 

[LoneDeranger]

Did Valve actually say that they would be using MRM before?

Because I remember a topic a while back where people were discussing how many models would have to be created for the static LOD.

EDIT: Also, isnt there some easy way to reduce the number of poly's in 3D modelling software? (I'm sorry but I'm mostly clueless about modelling)

 

[Ahnteis]

I never noticed the LOD popping in serious sam. (well, at least not the character LOD. The terrain LOD was a separate issue but not bothersome either.) I'm not worried about it--there' no reason they can't do it in a way that will be unnoticeable unless you're specifically looking for it.

 

[Sparky the Fox]

what are the advantages to static LoD? i can certainly see all of the disadvantages, and that makes me wonder why they would change for what seems to be no good reason.

 

[Chris D]

They licenced the MRM technology about 5 years ago when they started on TF2. Their licence probably ran out - and the MRM technology is obviously at least 5 years old so I doubt it's even that good anymore.

 

[Mr.Reak]

Good? It is better than static LOD.

 

[koopa]

I wonder if for the SP game they want to keep the CPU load as low/efficiency as high as possible, so static LOD is probably better for them. Maybe they leave the MRM tech for mods.

 

[$niper]

Originally posted by LoneDeranger
Did Valve actually say that they would be using MRM before?

Read the post I made at the top of the page. Valve licensed it for use with Team Fortress 2.

 

[$niper]

Originally posted by koopa
I wonder if for the SP game they want to keep the CPU load as low/efficiency as high as possible, so static LOD is probably better for them. Maybe they leave the MRM tech for mods.

If they wanted to do that, then they would be using MRM technology, since it's purpose is to increase performance.

 

[koopa]

Originally posted by $niper
If they wanted to do that, then they would be using MRM technology, since it's purpose is to increase performance.

I would have thought that a good static LOD is faster than MRM (unless MRM is amazing). MRM probably decreases the time to create the models, or might save diskspace - but Valve have been working on this for years, and what's a few Kb these days anyway?

 

[Mr.Reak]

Indeed, so what's the point?
Sniper, e-mail to Valve this question, maybe they will explain better.

 

[LoneDeranger]

Which games out there do use MRM?

 

[Lifthz]

Originally posted by LoneDeranger
As long at it doesn't look anything like the LOD transitions in Vice City, I'm gona be happy.

 

[$niper]

Pandemic Studios uses MRM in all of their games.

E.g. Battlezone 2, Dark Reign 2, and their new titles.

 

[Lifthz]

EIther way this MRM issue shouldn't be a decisive factor as to if HL2 will have bad LODing. There are other ways to do things. So far from what i've seen in the Steam videos the LODing seems to be quite nice.

 

[koopa]

Originally posted by Sparky the Fox
what are the advantages to static LoD? i can certainly see all of the disadvantages, and that makes me wonder why they would change for what seems to be no good reason.

Generally (not all of this might be true for Valve, someone will probably flame me): Easier to code. Usually less processor intensive than dynamic (and they're running CPU-nasty physics). Might be quicker to render on some cards.

 

[Lifthz]

Static LOD is technically inferior, bottom line.

 

[koopa]

Originally posted by $niper
Pandemic Studios uses MRM in all of their games.

E.g. Battlezone 2, Dark Reign 2, and their new titles.

Didn't know that ; seems like the Shockwave stuff does too. AFAIK most of the big engines (Unreal, Lithtech etc) have their own dynamic LOD, can't think of any others.

 

[Lifthz]

Exactly, real engine programmers can create their own LOD system. However if Valve is really using static LOD, I dunno, we'll see. So far it doesn't seem bad to me though.

 

[Mr.Reak]

So what's the point using static LOD if you can use dynamic LOD? Anyone?

 

[Chris D]

Chris_D: Hi Gabe, here's just a few questions I've been meaning to ask

1) After tonight's chat, some people are reporting that you won't be using
Intel's MRM Technology that you were going to use for TF2 and instead using
a static LOD system. What's made you come to this decision and what benefits

does it have over Intel's technology?

Gabe: Static LODs are faster and work better with the hardware.

There is a reason for all of Valve's madness.

 

[nw909]

:O

maybe it wont look as bad, we haven't seen it yet so don't freak out people.

 

[dead_pixel]

Well.. this is really OOOOOLD news dude... I always wondered how would the team be working with their LOD system, and thus I sent Gabey an email, a while ago, talking about the MRM (multi-resolution mesh) technology, patented by INTEL. This is the email:

MRM ended up having deficiencies that caused the tesselation to take longer than the time saved rendering, so we had to code our own.
-----Original Message-----
From: Diogo Sobral Magalhães Pinto [mailto:[email protected]]
Sent: Thursday, July 03, 2003 8:28 PM
To: Gabe Newell
Subject: Intel's MRM Technology


Another lil' question from me, Gabe:

The LOD scalability thing, that is being talked about so much this time around, is the same you guys planned to use in Team Fortress 2 (Intel's MRM)?? Or is it a new approach all coded by Valve Team?

Thanks Again,
Later,

Diogo

So in the contrary of your speech, Intel's MRM technology ended up being slower then expected in real play (render) time!

d_P


** EDIT **
Just a quick note: MRM technology was more of an automatic way of dealing with LOD, since the technology, built in the engine, would automatically adjust the poly count based on the depth of the player's view, I mean, the further from the player the model is, the less detail it presents.

The new way presented by Valve, is a more time consuming way of implemettng a LOD system for models, BUT it saves A WHOLE bunch of computing cycles, since you are just, seamlessly, switching models, not dynamically adjusting model's poly count (on-the-fly).

d_P

 

[LoneDeranger]

Originally posted by $niper
Pandemic Studios uses MRM in all of their games.

E.g. Battlezone 2, Dark Reign 2, and their new titles.

Ah yes, Battlezone 2 had pretty a good LOD system if I remember correctly. But HL2 doesn't look too bad at all. I guess we'll see once the high-quality steam buggy vid comes out.

 

[$niper]

Battlezone 2 wasn't slow at all for me with Intel's MRM tech.

 

[Apos]

Good? It is better than static LOD.

In computer programming, there is rarely such a simple thing as "better." And plenty of technologies sound great on paper, peppering themselves with buzzwords, but they never really pan out that way in practice. There are always multiple tradeoffs you have to decide about. Nothing makes dynamic any more advanced or succesful than dynamic: they are different ideas about how to scale things with distance, both with their strengths and weaknesses, and both with their own advancements and improvements that can be made.

What strikes me as odd about the "pops" we see in HL2 vids is that they are so scattered and isolated. If things were really being switched in and out all over the place, it should be apparent everywhere rather than just in one or two spots.

I'm also a bit confused about the scope of the question. It seemed to be in response to a question about entities, and certainly those you would have to build several versions of for more distant versions. But LOD is even more important for physical geometry, and I can't really believe that they are making static LOD geometery for each and every brush in the world.

 

[Lifthz]

Originally posted by nw909
:O

maybe it wont look as bad, we haven't seen it yet so don't freak out people.

I've seen it. I dunno about you. It looks alright to me.

Originally posted by Mr.Reak
So what's the point using static LOD if you can use dynamic LOD? Anyone?

Easier, less time consuming, less hardware strenuous, but are inferior to dynamic LODing. Just like static lighting is inferior to dynamic lighting.

 

[dead_pixel]

Originally posted by Lifthz
I've seen it. I dunno about you. It looks alright to me.



Easier, less time consuming, less hardware strenuous, but are inferior to dynamic LODing. Just like static lighting is inferior to dynamic lighting.

Well.. those are two different stories.. Dynamic lighting, nowadays, is mostly accomplished by specific shading instructions of the Directx API, a LOD based system we are talking about, is just a set of models done with different polygon counting, ranging from low to higher poly count.

And btw, Valve's multi-sampling-mesh (can we call it this way? hehe )LOD system is less computing consuming than MRMs dynamic LOD

d_P

 

[fifty7var]

Another advantage of static LOD, is that it puts the artists in control. Has anyone ever used one of those poly-reducer functions in a 3d app, like 3ds max's "Multi-res." While it is able to reduce the polygon counts, the results are very often crap. One reason for this, is that the low-poly mesh that results is often poorly animated. putting the detail in the wrong places for properly bending limbs.

Also, with normal maps suplimenting the detail in the models, and with 7 or 8 LOD meshes, the "popping" should be kept to a minimum.

Adam

 

[koopa]

Question : Can the Maya/SoftImage/whatever automatically generate lower poly models, given a high poly original? I would have guessed this is automated by now, so static LOD (on models, anyway) isn't quite so unpleasant.

 

[adio2001]

I believe you have to make several versions yourself, they are not automatically generated. I could be wrong though.

 

[fifty7var]

Originally posted by koopa
Question : Can the Maya/SoftImage/whatever automatically generate lower poly models, given a high poly original? I would have guessed this is automated by now, so static LOD (on models, anyway) isn't quite so unpleasant.

Answer: Yes, many 3d applications have a feature like this incoporated.

 

[Lifthz]

Originally posted by fifty7var
Another advantage of static LOD, is that it puts the artists in control. Has anyone ever used one of those poly-reducer functions in a 3d app, like 3ds max's "Multi-res." While it is able to reduce the polygon counts, the results are very often crap. One reason for this, is that the low-poly mesh that results is often poorly animated. putting the detail in the wrong places for properly bending limbs.

Also, with normal maps suplimenting the detail in the models, and with 7 or 8 LOD meshes, the "popping" should be kept to a minimum.

Adam

Yeah. But the LOD has to be at a standard of realism. Unless you were trying to make some crazy cartoon game or something.

Originally posted by dead_pixel
Well.. those are two different stories.. Dynamic lighting, nowadays, is mostly accomplished by specific shading instructions of the Directx API, a LOD based system we are talking about, is just a set of models done with different polygon counting, ranging from low to higher poly count.

And btw, Valve's multi-sampling-mesh (can we call it this way? hehe )LOD system is less computing consuming than MRMs dynamic LOD

d_P

It is still relative. Static LOD is inferior to dynamic just as Static lighting is inferior to dynamic. I know it's two different stories.

 

[Archangel]

You can see the grass popping out during the buggy scene. Its not that serious, since you barely see enemys and stuff from far away, all the time you getting closer, until you stand in front of em, due to the level design. I haven't noticed any worse pop ups, so it should be okay.

 

[fifty7var]

Originally posted by Lifthz
Yeah. But the LOD has to be at a standard of realism.

That was my point, the automatic LOD routines, including dynamic LOD routines often produce subpar results when compared to those manually created by the artists.

Adam