Why do developers continue to use havok?

[~El~Jefe~]

I think the whole problem is that people cant program efficiently to save the F'ing life.

half-life2 was good/unusual. STALKER is an awesome game but kinda sucks how much it takes to do it. (the float32 homebrew-made-by-a-kid fixes what 3 years and 25 programmers couldnt)

maybe Aegia would be take games into the year 2020 but I dont think anyone knows how to program.

 

[svet-am]

O and also make sure the physx has good software cpu support.

i think is the most important aspect. while PhysX already runs in software mode, there aren't a lot of side-by-side examples of the same PhysX code with and without the card. if we could get more, I think that PhysX could pretty much crush Havok. In addition, some kind of Havok wrapper API for the PhysX card would really help turn the tide.

 

[Manny Calavera]

Cheap quad cores like the 32xx's and the g0's,and if AMDati get thier shit together,Phenom X4 will land/are landing the killing blow that will take Ageia
out of the picture,and put them down for the count,for good.

 

[SuperGee]

Well Quadcore are somehow in the same problem as PPU is. Dev need to support them. It will come but it will take some time. Well done SMP support isn't a easy task. Quake4 did it with assistance off iNtel and that was fixed dual threaded.
Valve has done some homework and put there aim at using any CPU for the 100%. with the more advanced hybrid threading. Not every engine builder licence or inhouse have this time to do research. Most will do SMP with split all task over avaible cores wich isn't efficent and doesn't scale well if there are more cores then task.

Unreal3 say a Dualcore will sufice a quadcore doesn't yield much. So I expect this nextgen engine doesn't use quadcore to there full potentional.
I wonder what crytech engine does with it. Valve and epic went public allready how it stands.
Because the PPU is dedicated and CPU not it still offer more Physics then Phenom. Instead of over focusing on Physics they must give AI also a boost that AI isn't frustrating gamers. That can be often a show stopper. More or less Physics can change a game in feel or emersion AI/AL is often a difference about fun or frustation.

Valve speaks about Physics and AI in a equal manner. They see what Quadcore can mean for Ai.

 

[Atech]

Yeah I count 9 different games, most of which I've never heard of.

People use Havok because there isn't a need for hardware for it to work well.

So what is HavokFX?

 

[Bleedblue12345]

honesly i much like a physx card better than havox fx, but i think havox fx was more of what the can do, not what they will do.

i mean imagine sticking 3 8800's in a system for physx, ure case would have to have insane cooling and a great psu to take that, but what a physx card as it stands it is fairy low heat and power

 

[SuperGee]

Yep a HD2400xT or 8600GT compete with PhysX PPU in price. For a HavokFX game I reserve the CF/SLI slots for.
But I think it the numbers of Havok FX games might rise in 2008.

 

[jmackay]

So what is HavokFX?

Its your last gen card not being obsolete.

 

[ex0du5]

But, is the bottleneck on certain games even coming from the CPU due to a lack of processing power to run the physics engines?

It seems to me that the graphics card will be the main bottleneck in games for a while yet. Developers would be wiser to use the multicore processors to process the physics requirements. 4 core processor? Just dedicate one thread to physics.

But seriously, I just don't think we're ready for high level physics. Developers have been pushing the graphics hardware since FarCry came out, and it doesn't seem like it's going to stop quite yet. After that, developers will be focusing on optomizing their engines for multi-core processors. I think only when that's done, will developers really focus on the physics engines in games. That really will probably only happen in another 5+ years.

edit: I just checked the Havok FX information site....they've got it spot on!!!

Do PC Gamers Need To Purchase Proprietary Physics Hardware To Accelerate Game-Play Physics?

Not for game-play physics. Havok believes that fast game-play physics starts with excellent software technology, and gets faster with increased general CPU computing power – specifically from multi-core and multi-cell computing architectures. The very clear trend from AMD, Intel, and even IBM going forward is to provide scalable core architectures that can be leveraged by truly multi-threaded software systems. Havok has introduced HydraCore™ technology in its physics system to take advantage of these architectures now and in the very near future. So, for acceleration of game-play physics, along with the core game-logic, special purpose physics hardware is not the answer. Multi-core CPU architectures are the path to speed.

For effects physics, however, the GPU offers the most compelling and sustainable promise to added visual fidelity in games. Relative to proprietary hardware, GPUs also have a clear advantage as a pre-existing technology familiar and readily available to consumers and game developers, providing other benefits such as wide-spread availability, commodity pricing, and mature standards for hardware and software interfaces.

This is exactly what I was saying. They've even gon a step further and made it possible to use a second GPU in SLI for effects (which are not game-play affecting).

Game affecting physics really need to be restricted to the CPU, as you CANNOT alienate a large section of the market that don't own a PPU. Multicore processing will solve the problems in the end.

 

[Bleedblue12345]

ok well thats great and all

just if ageia would make a great physx engine that worked well in software and gave a boost of some visuals/fps that could not possibly be done with the current gen cpu, than i think people would buy it.

Also:
-make it $100
-make it very good
-make it with led's
-make it quiet, possibly fanless if possible,
-make a lot of GREAT games to take advantage of it
-make it to use pci-e x1

 

[svet-am]

It seems to me that the graphics card will be the main bottleneck in games for a while yet. Developers would be wiser to use the multicore processors to process the physics requirements. 4 core processor? Just dedicate one thread to physics.

This is the common approach of multi-core advocates and it's really starting to bug me. Do you REALLY think it's as simple as "just dedicate one thread to physics"? Do you really realize how long a single physics calculation would take if it was just a single thread, even if it was on a dedicated core? Physics is not a linearly processed science. It is massively parallel. This goes to the very heart of why a PPU, with it's massive vector processor cascade is the much better solution.

Any physics calculation that is useful for gaming would require many, many (several hundred or even into the thousands) of independent threads processing just their own small portion of the physics for a given scene.

Please, folks, before taking the "dedicate a thread/core to physics" approach, please understand what you're actually saying.

 

[[21CW]killerofall]

-make it $100
-make it very good
-make it with led's
-make it quiet, possibly fanless if possible,
-make a lot of GREAT games to take advantage of it
-make it to use pci-e x1

There is a PCI-E version coming out in the future (maybe by the end of the year?). My BFG Physx card has blue LEDs in the fan. I never hear it over my quiet system. It does its job very well. UT3 is coming out in November and should be interesting how it uses the card. I got mine for free (LAN parties FTW) so I can't complain on the price.

 

[svet-am]

There is a PCI-E version coming out in the future (maybe by the end of the year?). My BFG Physx card has blue LEDs in the fan. I never hear it over my quiet system. It does its job very well. UT3 is coming out in November and should be interesting how it uses the card. I got mine for free (LAN parties FTW) so I can't complain on the price.

Just thought I'd clarify. PCI-E versions are coming to CONSUMERS soon. System integrators (Dell, Alienware, et al) ALREADY HAVE PCI-E versions and have been using them for some time now.

 

[Atech]

Its your last gen card not being obsolete.

Really, would you care to link me to exsamples of this?
A game list perhaps?

 

[MrWizard6600]

nevermind

 

[MrWizard6600]

LMFAO! You use ONBOARD AUDIO (aka shit) instead of at least a decent sound card which would provide a VERY NOTICEABLE difference in ALL audio (music, games, movies) yet you buy USLESS physX cards (unless you play nothing but GRAW2 and cellfactor, and both of them suck) for THREE computers (all of which suck)

I can't figure it out...

Try spending some money on ONE GOOD computer instead of 3 POSes with useless 150+ dollar add on cards and displays that were obsolete years ago.

me and my friends have a term we throw around

YOU ARE CUT!

i use my onboard... 7.1 w/ full 24bit decoding (I think... im not an audio techie, theres a 24bit decoding option in the driver software) that's sh!t?

keep your flamebait.

(the float32 homebrew-made-by-a-kid fixes what 3 years and 25 programmers couldnt)

correction: "made-by-a-kid-in-5-years". STALKER originally debuted on an FX5200

when i first heard game physics, I came running, and was sorely dissapointed. If games had a concept of heat, momentum, density, vibration etc, they would all be completly different. Havoks opinion of "physics" as I saw it from the source engine is, while groundbreaking, pathetic. We are eventually going to have specialized hardware to accelerate physics, and, just like the verticies (remembering that prior the the Geforce FX polygons were calculated and rendered on the CPU), physics is originally going to be done on the CPU, and it wont be done as well.

I really wanna see a truly physics inspired game, if I have to wait until im 30 to see it, so be it.

 

[Betrayer00GT]

i mean imagine sticking 3 8800's in a system for physx, ure case would have to have insane cooling and a great psu to take that, but what a physx card as it stands it is fairy low heat and power

I believe that when Nvidia (pre)announced the physics acceleration that they had developed for the 8-series, they indicated that a part of a single 8-series card could be "dedicated" to physics, or you could dedicate a separate card to it. I would hope 3 8800 cards would not be needed for good frame rates and good physics acceleration.

It is all kind of academic until a game that will use the GPU for physics is released. It might still be, even then, if that game doesn't implement it well.

 

[silent-circuit]

correction: "made-by-a-kid-in-5-years". STALKER originally debuted on an FX5200

Yes and no. It debuted on a FX5200, sure... but Prey debuted on a Rage Pro or something. The Prey that made it to shelves was not the same game. The STALKER that made it to shelves might have been closer to its debut form, but it had been rewritten many times over the course of those five years. The guy that wrote Float32 did so since release of the final, retail version of STALKER, in so far as I know.

....(remembering that prior the the Geforce FX polygons were calculated and rendered on the CPU), physics is originally going to be done on the CPU, and it wont be done as well.

...Plenty of cards prior to the FX series had hardware T&L. The 3DFX Voodoo 3, for example. Not sure exactly what you mean by "polygons were calculated", but taking it literally that's even farther from the truth, as the original 3DFX Voodoo was capable of true GPU-based rendering of all kinds of solids, it didn't rely on the CPU to "calculate polygons" at all. Probably several other far more obscure cards before it as well.

 

[MrWizard6600]

...Plenty of cards prior to the FX series had hardware T&L. The 3DFX Voodoo 3, for example. Not sure exactly what you mean by "polygons were calculated", but taking it literally that's even farther from the truth, as the original 3DFX Voodoo was capable of true GPU-based rendering of all kinds of solids, it didn't rely on the CPU to "calculate polygons" at all. Probably several other far more obscure cards before it as well.

I wasn't into PC gaming back then, but I figured since the FX series came along with the first thing that looked like a Vertex Pipe it was the first to perform the vertex calculations, and I guessed that prior to that Any vertex load would have to be done on the CPU. But, that was a bad asumption, clearly.

 

[silent-circuit]

I wasn't into PC gaming back then, but I figured since the FX series came along with the first thing that looked like a Vertex Pipe it was the first to perform the vertex calculations, and I guessed that prior to that Any vertex load would have to be done on the CPU. But, that was a bad asumption, clearly.

Vertex shaders (which are what are referred to by vertex pipes -- the pipes being channels to simultaneously available pieces of vertex shader hardware) as cribbed direct from Nvidia here:

"A vertex shader is a graphics processing function used to add special effects to objects in a 3D environment by performing mathematical operations on the objects' vertex data. Each vertex can be defined by many different variables. For instance, a vertex is always defined by its location in a 3D environment using the x-, y-, and z- coordinates. Vertices may also be defined by
colors, coordinates. Vertices may also be defined by colors, textures, and lighting characteristics. Vertex Shaders don't actually change the type of data; they simply change the values of the data, so that a vertex emerges with a different color, different textures, or a different position in space.

Before the introduction of the GeForce3, vertex shading effects were so computationally complex that they could only be processed offline using server farms. Now, developers can use Vertex Shaders to breathe life and personality into characters and environments, such as fog that dips into a valley and curls over a hill; or true-to-life facial animation such as dimples or wrinkles that appear when a character smiles.

Examples of vertex shading effects include: matrix palette skinning, which allows programmers to create realistic character animation with up to 32 "bones" per joint, allowing them to move and flex convincingly; deformation of surfaces, which gives developers the power to create realistic surfaces such as waves and water that ripples; and vertex morphing, which is used to morph triangle meshes from one shape to another, providing smooth skeletal animation. These are just a few of the virtually infinite number of effects developers can create using Vertex Shaders. By customizing skinning and motion, developers can create life-like personalities for characters and scenes, thereby intensifying the graphics experience."

Short version: vertex shaders allow extremely sophisticated realtime modification of existing geometry and texture data. Cards were capable of just rendering more 'static' vertex and texture data long before vertex shaders came along.

 

[Demon_of_The_Fall]

i use my onboard... 7.1 w/ full 24bit decoding (I think... im not an audio techie, theres a 24bit decoding option in the driver software) that's sh!t?

Your games audio is only going to 16bit 48KHz, absolute max. Your onboard audio dumps all the work onto the CPU and it can't do EAX or any proper audio effects. The quality of the audio you hear isn't determined by the "number of bits that the card can decode" or anything so ridiculous, it's the quality of the effects that it is able to process and quality of the DACs it will use to output this audio. Don't kid yourself about your motherboard audio, it blows.

 

[ex0du5]

But, is the bottleneck on certain games even coming from the CPU due to a lack of processing power to run the physics engines?

It seems to me that the graphics card will be the main bottleneck in games for a while yet. Developers would be wiser to use the multicore processors to process the physics requirements. 4 core processor? Just dedicate one thread to physics.

But seriously, I just don't think we're ready for high level physics. Developers have been pushing the graphics hardware since FarCry came out, and it doesn't seem like it's going to stop quite yet. After that, developers will be focusing on optomizing their engines for multi-core processors. I think only when that's done, will developers really focus on the physics engines in games. That really will probably only happen in another 5+ years.

edit: I just checked the Havok FX information site....they've got it spot on!!!

This is exactly what I was saying. They've even gon a step further and made it possible to use a second GPU in SLI for effects (which are not game-play affecting).

Game affecting physics really need to be restricted to the CPU, as you CANNOT alienate a large section of the market that don't own a PPU. Multicore processing will solve the problems in the end.

I realize I'm being overly optomistic. However, I think that the multi-core approach is needed to ease gameplay affecting physics into next-gen gaming, as people are not ready to dedicate money towards PPUs. If developers release games with gameplay affecting physics that require a PPU, they will be alienating the majority of the market....unless Ageia becomes a publisher and finances it's own games, these games will not come out. The only way Ageia will push their PPU is with AAA titles that demand it, which is why I say they need to publish their own games....of course, they don't have the financials to back that up, so another company would need to take the fall.

Havok has the best business model now. Slowly ease consumers into it, by taking advantage of multi-core processors. If ever it is really needed down the road, I'm sure they'll make their own dedicated hardware, or buy out Ageia to do so. Like said in previous posts, the middleware is needed first, not the hardware.

Your games audio is only going to 16bit 48KHz, absolute max. Your onboard audio dumps all the work onto the CPU and it can't do EAX or any proper audio effects. The quality of the audio you hear isn't determined by the "number of bits that the card can decode" or anything so ridiculous, it's the quality of the effects that it is able to process and quality of the DACs it will use to output this audio. Don't kid yourself about your motherboard audio, it blows.

Then again, Creative's products blow too (but not nearly as bad). I honestly think my iPod 4g (not the 5g...5g is shit) has better DACs than my Audigy 2 ZS (which I think has superior DACs to the Xi-Fi). But whatever, I'm not completely anal about the sound quality in gaming, as sound samples are rarely anything near lossless quality. For music though, you can't beat a good external DAC (though I can't bring myself to spend so much $$$ on them).

 

[svet-am]

the bottleneck is not in hardware. the bottleneck is in the game engines themselves and how developers approach the game engines. game engines (include UE3 from what I've seen) take a brute force approach of including all of the processing for everything (AI, physics, scene generation, damage and collision detection, etc) in one great big bundle. to move forward and really make effective use of the hardware we're starting to see on the market, they really need to start compartmentalizing this processing so that the physics is processed on its own, audio is processed on its own, AI is processed on its own, etc. when they get to this point, we'll see the bottleneck fall back to the hardware. Essentially, right now they need as much CPU power as they can just to keep up with all of the work they have tied up in their "code bundle" rather then spreading the processing out to the APU, GPU, PPU, CPU that would most effectively process what they're trying to do. It really is distributed computing coming to the desktop.