Questions: Integrated Physics (PhysX or Havok) Acceleration; Differential Impact of Higher Resolution on CPU vs. GPU

Hi guys,

I searched the forums regarding these topic but the information I found was either outdated or unsatisfactorily answered my questions. I am going to be purchasing either the M860TU or M570TU in the coming weeks, and when they become fully configurable on 7/14 or 7/15, I would like to be able to make sound decisions regarding the video card. I will almost exclusively be playing Starcraft II and the upcoming Diablo III (not much time for gaming as a super busy graduate student ), which both employ Havok physics. Please help me better understand two issues:

1a) ATi has teamed up with Havok, and nVIDIA uses PhysX. How advantageous would it be to use an ATi card solely (over an nVIDIA) because it natively handles Havok physics? In the case of nVIDIA cards, would the physics calculations be performed by the CPU instead? In a game like SC2 where potentially hundreds of units are rendered simultaneously, I would assume (in my naiveness) that a GPU performing the bulk of those physics calculations could, potentially, substantially reduce the stress on the CPU, leaving it open to handle other calculations and thus increasing performance. Of course, nobody knows specifically about SC2, but perhaps experience with other games could be extrapolated to predict the usefulness of matching an integrated physics engine to a game supporting that engine.

1b) As for hardware options related to (1a), options that might be available at (or near) release of the two Clevo models include the nVIDIA 8800M GTX, ATi HD 3780, and nVIDIA 9800M GTX. Though I have found that the desktop variant of the ATi card has "physics support", I have not been able to find a source that definitively states Havok integration. Since the Havok-ATi news was released in 2006, I am hopeful that it is integrated at this point, but I am not certain. Does anyone know if either the desktop or mobile version of this card support Havok, or is this something that's expected in the 4000 series? If the nVIDIA cards perform equally well or better than an ATi option (and I know this answer hinges on knowledge of my (1a) question), in your opinion/prediction, would choosing ATi not really impact performance (for the games I've listed) enough to justify the higher general performance of nVIDIA cards?

2a) If I chose the M860TU, I would get the WSXGA+ screen, but if I went with the M570TU and its larger screen, it would be a toss-up between WSXGA+ and WUXGA+. I have read that with most games, higher resolutions generally impact the GPU to a much greater degree than the CPU. WUXGA would be excellent for my academic work -- that screen real estate would be pretty boss. I'm sure that WSXGA+ would be fine, though, and I don't want to choose a resolution that would significantly impact my ability to play a game smoothly. With Montevina and these high end graphics solutions for RTS games, I'm not too concerned with the graphics intensity since I've read that RTS games generally stress the CPU more than the GPU. In that case, the WUXGA option should not significantly impact the performance of the game as compared to the WSXGA option given the same CPU. Do you think this is sound reasoning, or am I missing something?

Though both of these Clevo models are user-upgradable and there are better things coming (like always), I would prefer to not have to upgrade anything at least for the first 2-3 years and plan to keep the laptop for 4-5 years. There is a rumor that a powerful ATi card is coming out Q3/Q4 this year, but I really need (to have received) a new computer by the end of August at the latest, so I don't want to wait too long.

Thanks in advance for any help -- I really appreciate your opinions and expertise!

 

1a) As I understand, the Havok "engine" is simply employed by developers to add to physical interactions with the environment above and beyond the standard game engine itself. Two notable games, both of which I have played on Nvidia machines are Company of Heroes and World in Conflict. The first I play on a 9500m gs and the second on my desktop 8800gt. Both play beautifully. The difference between ATI/Nvidia I would have to believe to be negligible.
Game deveopers and manufacturers have been "teaming" together or offering "sole support" for years and to tell you the truth I'm not sure it makes much difference. I could be wrong on this subject but the above mentioned games run great on Nvidia hardware. I'm sure the CPU has a good amount of calculations to do depending on the direction of the game. I.E. FSX can pretty much kill most systems out there not because of graphical complexity but simply the mass amount of objects to be rendered and/or moved about and interacted with. I'm sure the system you will get will offer a strong processor. RTS are indeed known for an increased amount of load on the CPU. And to be honest I don't know how it splits itself between physics and graphical rendereing or if I even care. As long as it runs. High end parts as you speak of should do just fine.

1b) I would say go with the 9800mGTX because it should, by all accounts, run cooler than the 8800 and possibly use less power. Either the ATI or Nvidia would run those games.

As for screen type/size I fear I cannot be much help. All I can say is that notebook cards, even this next gen, will not touch the high end desktop parts (X2, GX2 variations) that are meant to be played at crazy resolutions. I wouldn't shoot for a super high end resolution because you may be disappointed when the game doesn't perform to spec. I play at 1440x900 and 1280x800 and I'll never know what I'm missing As far as i'm concerned though it still looks damn good.

 

From what we know the 9800M GTX is still a 65nm G92, and has 112SPs VS 96 on the 8800MGTX, thus it will run hotted and consumes more power.

I don't hink Havok has been used very much and the GF8/9/+ will be having physX soon so it's a big point imo.

At higher resolutions the GPU becomes the bottleneck because resolution doesn't affect the load on the CPU but on the GPU.

WSXGA+ is great imo, I'm using it on a 15.4'' and it's wonderfull. I'd say go for it if you want to mix smoother frame rates + decent view.

 

Both cards will still run the games though, regardless of engine. It's all in how the game is coded, and developers would'nt run the risk of losing market audience to use a certain engine that only works with one type of card. Havok is used in both games he mentioned. The point still is that any of the cards mentioned will run both games in the little amount of time he has to game.

 

Thanks for your responses guys

@JCMS -- you said "I don't hink Havok has been used very much and the GF8/9/+ will be having physX soon so it's a big point imo."

Since the 8000 series is already finalized and released, do you mean that it does not have integrated PhysX support, but that the upcoming 9000 series does? I believe this is what I have heard, too. I'm sorry I don't understand your reasoning -- why do you think it's a big point? Maybe Havok isn't as widely implemented for physics as PhysX was/is/will be, but the games I'm interested in playing will be using it nonetheless. In that case, would it be a big point in your opinion to obtain a GPU integrating Havok support?

Thanks for your advice on the screen resolution and notes on the potential heat generation of the 9800M GTX. I trust Clevo to have developed adequate cooling systems on their notebooks, though -- the good people over in the Sager/Clevo forum really put their faith in the ODM and have had good experiences on the whole with those notebooks.


@ bigepilot -- you said "I'm sure the CPU has a good amount of calculations to do depending on the direction of the game. I.E. FSX can pretty much kill most systems out there not because of graphical complexity but simply the mass amount of objects to be rendered and/or moved about and interacted with. I'm sure the system you will get will offer a strong processor. RTS are indeed known for an increased amount of load on the CPU. And to be honest I don't know how it splits itself between physics and graphical rendereing or if I even care. As long as it runs. High end parts as you speak of should do just fine."

I agree that the CPU I get should be able to handle the physics calculations fine, and thank you for that reassurance. Since, as you mention, physics can really tax systems due to the massive number of objects to be rendered and/or moved about, etc., which is what I was thinking too, I was hoping that a GPU (in this case an ATi with integrated Havok support) could take care of (or at least significantly partake in) that processing.

I am sure that whatever system I get will perform well, but as someone who loves to research, this is more curiosity on the intricacies of these physics solutions than worry about overall performance.

Thanks again and I look forward to any additional opinions/help

 

PhysX has been made to run on CUDA, and it happens that CUDA is nvidia's USP architecture. They recently added it to the 9800GTX, GTX 280 and GTX 260. Once it iwll be added on all the cards, even the 7150M will be physX enabled.