Question for developers - Havok hiccups

  • Two Factor Authentication is now available on BeyondUnreal Forums. To configure it, visit your Profile and look for the "Two Step Verification" option on the left side. We can send codes via email (may be slower) or you can set up any TOTP Authenticator app on your phone (Authy, Google Authenticator, etc) to deliver codes. It is highly recommended that you configure this to keep your account safe.

kafros

F1 manta tryouts
Jan 21, 2005
331
0
0
49
Under Articstronghold's bridge
I case any developers are reading these forums I have this question on the havok physics engine:

I watched the videos of UE3 and there is a small freeze (very small but noticable) when some "physics" are applied. It is very evident on the video that shows the scorpion like car jumping onto a pile of barrels that explode.

Video page:
http://www.1up.com/do/download?cId=3138759

Why is that and can you get rid of it?

I have noticed it on Painkiller (very evident) when I shot exploding barrels or send a rocket into the crowd but I thought it was my crappy CPU responsible. I have also noticed it on HL2 when hitting the woden bridges with the airboat or when splitting wooden pillars.

I expect that the demos were produced by kick-ass machines (Athlon FX or Pentium extreme) so CPU power shouldn't have been a problem.

UT2004 doesn't seem to have freeze-ups like that.

Has anyone else noticed the similarity in the freeze-ups in the mentioned games and the UE3 videos or am I just going koo-koo?
 

Dark Pulse

Dolla, Dolla. Holla, Holla.
Sep 12, 2004
6,186
0
0
38
Buffalo, NY, USA
darkpulse.project2612.org
Uhh, keep in mind two things here:

1) This is Next-Gen tech, and

2) The engine isn't even CLOSE to optimized, much less final.

It'll be all smooth and squeeky-clean smooth for those with the appropriate hardware come crunch time, I'm sure.
 

carmatic

New Member
Jan 31, 2004
746
0
0
they will fix the hiccups when they offload the physics calculations onto a dedicated card, like the physx chip... that way all the processor has to do is keep on taking input from you, run the engine, and pump as much data as it likes to the physics chip instead of going through the data by itself...just like a long time ago, it just dumped data to the video card instead of rendering every frame by itself...
 

carmatic

New Member
Jan 31, 2004
746
0
0
no one knows yet... like, back in the first days of 3d accelerators, few people had these cards and the games supported them optionally... today, we need a 3d card to play ut2k4(umm no, we can use the software renderer, but i still havent seen a computer running ut2k4 using the software renderer before).... one day, we will still be paying the same amount of money to buy a computer, but the money will be distributed to more companies because if they sell something for a price that nobody wants to pay, they make no money so they are forced to sell their cards and stuff cheaper (hopefully) so graphics card makers compete with physics card makers etc etc etc... there will be those tempting high-end cards but oh well , as long as developers realize that most of their customers live without them then it wouldnt be too painful for us...
 
Last edited:

Bullet10k

New Member
Apr 9, 2005
639
0
0
I also doubt that they were using Intel DualCore-HT or AMD DualCore prototypes which could prolly have helped in loading ahead of time. Its also possible that they might not even have used the PhysX PPU. (Correct me if im wrong).

I remember this one time when UE3 managed to run at about 2-5 fps on a 5900. Then they managed to get about 25-30 fps using the new 6800. Then finally they got it running easily on a 6600GT, they even said it would work good on that. You see its actually being optimized as they develop it and the engine isnt done YET. I'm pretty sure theyve seen that problem already and fixed it by now.

Btw, there was this other bug where in one of the videos the pipe went THRU the wall (clipping problem). Anyone else noticed? They musta noticed and fixed that too by now.
 
Last edited:

Bot_40

Go in drains
Nov 3, 2001
2,914
0
36
York, UK
Yeah, I thought the clipping error on the pipe was rather poor, passes right though the wall. Really stood out :hmm:
 

carmatic

New Member
Jan 31, 2004
746
0
0
i think that the walls are just not solid... those flexible hanging things are surely alot more complicated than just a simple solid wall, maybe the people in epic simply didnt realize that theyd get it to swing that far back or something...

i think its the same thing with the scorpion scene, maybe they drove it so smoothly through the level without bumping into anything was more than just for cosmetic reasons, it might have driven right through , or worse still the physics engine was incomplete and the game might have crashed... well, envy is still a year away...
 

kafros

F1 manta tryouts
Jan 21, 2005
331
0
0
49
Under Articstronghold's bridge
I really cant see how they can introduce the need for a PhysicsProcessingUnit for a multiplayer game. What happens to the people that dont buy one?

Just as an example: one of the UC2 developers was commenting on the blur visual effects present on the xbox that are were not implemented for ut2004. The reason was that people with lower spec PCs whould have an advantage (no blur) as opposed to the ones that the blur could be shown.

I believe the same will apply for physics since this will be a multiplayer game.

Returning to my original post: Have you noticed these freeze-ups in HL2 and Painkiller?
If yes does it spoil your gaming experience?

Has anyone noticed freeze-ups in UT2004 games that seem to relate to physics calculations?
 

carmatic

New Member
Jan 31, 2004
746
0
0
eh the physics in 2004 is waay to simple to be worthy of any hiccups... its when you have a screenfull of physically accurate things that the computer would start to struggle...

anyway, i have this belief that a graphical effect shouldnt interfere with gameplay, instead it should assist gameplay if it can... like, when they mentioned the blur effect, it should be as a kind of visual cue like when you zoom in from a sniper rifle or something, or an effect when you died ala far cry ...

the same can go with physics, like the physx chip can simulate bending metal and stuff, so people with the card installed will never see the 'destroyed' vehicle meshes anymore, just like we see ragdolls nowadays and not death animations anymore... and , imagine if you can see a damaged vehicle, its still usable but it bears marks of what types of weapons have hit it , maybe a texture or some bent bits that is calculated by the physx chip , and people who have it installed will maybe have a tactical advantage by having that extra information available to them, while people who dont have it installed will just see ut2k4-style vehicles and meshes...
 

kafros

F1 manta tryouts
Jan 21, 2005
331
0
0
49
Under Articstronghold's bridge
carmatic said:
eh the physics in 2004 is waay to simple to be worthy of any hiccups...

In a typical Torlan 8vs8 match you can see (simultaneously):
1. Players ejecting from their mantas, and the mantas sliding down the small hills
2. The turret firing a raptor which spins out of control
3. A scorpion coming either bumping onto the hellbender or bouncing off the deployed spider mines
4. Near miss Shock combos sending mantas high up in the air

Damn dude; If this is THE toooo simple I can't wait to see the complex (or even the plain simple)
 

Bot_40

Go in drains
Nov 3, 2001
2,914
0
36
York, UK
These are actually all rather simple physics calculations, apart from number 3 all you are doing is applying a force which is basically just adding a bit onto the object's velocity.
Ragdolls are much more complex with all the bone constraints etc.
 

edhe

..dadhe..
Jun 12, 2000
3,284
0
0
43
Scotland
www.clanci.net
BmB23 said:
Does this mean I will have to buy a card with such a chip in it in order to play this thing?
No. Would be a bad marketing decision.

Plus it's likely that by the time this comes out your Physics will be another thread handled on another core in your cpu, will reduce the damage done to interaction by an effect processing.