Nvidia finishes porting PhysX to CUDA

i would still want a 128 Cored CPU to handle the stuff in the background like video converting, big downloads, ect.

Comment by NikoBellic at 2:30pm 22nd April 2008

I won't buy a quad core any-more. An extra GPU like the cheap (8800GT) with CUDA+PhysX will easily outperform any quad/octa (or even 16-32 core intel processor) in performing the game mechanics. The game will run entirely on the GPUs (one doing the graphics and other the physics) . The CPU will only for run the driver code :)

Comment by abacook at 8:36am 19th April 2008

I may be wrong here, but if you are planning to use one card for physics and one for graphics then surely you don't need them to be in SLI... as I understood it SLI is a way of making both cards split the rendering process and stitching it back together again prior to output to the monitor, which you obviously aren't doing if one card is handling physics processing and only one is doing the rendering.

Comment by NewParadigm at 7:47pm 17th April 2008

will it be better to have a graphics card and a physx card or have a graphics card with cuda? or if you can a afford it sli setup with one running for games a stuff and the other on physx? sounds expensive to do it the sli way but it may perform better

Comment by Gz_shortass9118a at 2:00pm 17th April 2008

will it be better to have a graphics card and a physx card or have a graphics card with cuda? or if you can a afford it sli setup with one running for games a stuff and the other on physx? sounds expensive to do it the sli way but it may perform better

Comment by Gz_shortass9118a at 2:00pm 17th April 2008

Now that this is going on, that maybe the DX team will stop concentrating on physx, and leave it to the people who know what they are doing with it (ageia), and now maybe they'll clean the code up in DX before releasing Windows 7 with DX11.0

Comment by NikoBellic at 10:19pm 16th April 2008

I mean, this article is referring to the fact my 320mb GTS card wont need thrown away for some shiny 9000 series card, the series based on mine !...... So...... this is good.... i want CUDA yesterday :D

Comment by Lightning_Pete at 12:49am 16th April 2008

aint it out now on vista sp1... check out this video....http://youtube.com/watch?v=omq3O22EUx4

Comment by NikoBellic at 1:51pm 15th April 2008

The other issue I guess is the fact that the gaming industry doesn't like DX10, especially with the recent comments about Vista being rubbish and Microsoft already looking to replace it by the end of 2009 or early 2010. Maybe we won't see any real changes until DX10.1 or even 11.

Comment by PokerMuppet at 1:47pm 15th April 2008

I just hope this makes a big change to the way games are going to be developed from now on, with the API working on consoles too, i think that would just encourage the Devs to chuck it in even the smallest of games

Comment by NikoBellic at 12:33pm 15th April 2008

I wouldn't mind having to lower the graphics settings a bit to let me have better physics and particle dynamics, I think interaction with the environment is more important in making a game feel more realistic than super shiny graphics... to a point anyway, obviously a compromise and balance will have to be found, I'd just like to know what kind of graphics performance hit we will have to take to achieve this.

Comment by NewParadigm at 8:35am 15th April 2008

CUDA is a C/C++ API that allows you to write programs that can use the GPU to crunch numbers instead of the CPU. We are looking at using this where i work to improve the speed of some of our floating point calcs. The thing is, like Direct X the API asks the card what it can do in hardware, and any features that are not supported will be handled through software. This is abstract to the user so it just happens! As a developer, i dont need to worry about it! As below, I would imagine that first off Nvidia will only enable the Hardware acceleration on 2+ GPU setups, as there is bound to be some free memory and clock cycles to use. I think (through drivers) they will disable it for the single GPU market in a ploy to get you first to buy an NForce based motherboard (If you dont have one all ready) and another GPU for SLI. But this does make sense. At work we are looking at using SLI to get one card for calcs and the other for the rendering. I would imagine that Nvidia would use the same approach, maybe even trying to get you to buy a third card for the Physx stuff. I guess we will have to wait and see!

Comment by connectionlost at 8:04am 15th April 2008

Unless I've missed something here, what they seem to saying is that the GPU on Nvidia graphics cards can perform PhysX at high levels... right? Doesn't this mean that users will be asking their cards to do MORE calculations? If the top end cards still struggle to play DX10 games at smooth rates with the settings on high, how can we expect a performance boost by asking the cards to do more? If you have a single card, then yes, maybe you will be able to use the GPU to raise PhysX levels... so you have lots or particles flying around in explosions, etc... but do you really want them there if you are going to have to lower the graphic settings in the first place? It seems to me that you will have millions of little square blocks flying around... I can definately see the advantage when using a dual-GPU or SLi set-up as the second GPU or card doesn't really do too much, as its the extra RAM that enables the same level of graphics but within larger resolutions... So it seems to me we are back at basically buying a second card or a dual-GPU card purely for the PhysX benefits... £200 extra for some bits to fly around within a half dozen games??? As I said at the start, I may have missed something, so please inform me if you think I have.

Comment by PokerMuppet at 4:50am 15th April 2008

for those of us with a single gpu, how much (if at all) will this affect frame rates in 'real game use'? If it's too much then it'll just be left unused...

Comment by NewParadigm at 11:52pm 14th April 2008

so does this mean i wont need my PhysX's card anymore. I have an 8800GTX, hopfully it will be capable of the same performance as the 9800GTX. I remember the DX10 Waterfall demo that Nvidia made, it said that all 8 series cards had 128 shader processors and also dedicated PhysX processors maybe this is what they are utilising here.

Comment by CPC_RedDawn at 10:01pm 14th April 2008

The reason nVidia are cynical is that the 9 series cards didn't show a significant upgrade from previous 8 series. The 8800 GT and GTS 512 both have the same G92 core, so really they should either be 9 series cards, or the 9 series should have waited until the next core upgrade. The advantage is though that perfectly acceptable 8800's are now a fair bit cheaper than they were before. Not an advantage for me though as I bought my 8800 GT 512 OC only in January and already it's dropped £50 in price :/ Good on nVidia though for the PhysX stuff. Can't wait till the driver is released, even if only a handful of games I play use it atm.

Comment by TWeaKoR at 9:16pm 14th April 2008

good news considering i just bought my BFG 9800GTX for £199 :D If this means I get to see all the extras that the Physx card owners saw in games like Half Life2 and I dont have to pay anything for it then I take my hat off to Nvidia. Maybe this explains the apparent cynical move of releasing the 9800GTX? If they give us all PhysX for free how can they be accused of being cynical? maybe those revisions in the G92 chip(2nd rev.) found in all 9800 cards have something to do with the PhysX code? it would explain a whole lot when you think about it. Nvidia knows everyone will want physX if it performs so well. They maybe find the old 8800 320MB cards arent as good as the g92 ones. Then realises with a further revision or two they are getting spectacular results, hense GTS , GT(8 series) and now GTX(9 series). otherwise there hasnt been much sense to these endless lists of so similarly performing cards has there?.

Comment by hazed100 at 7:01pm 14th April 2008

i swear (reply to below) that with physx ported to the g80/92 gpus the graphics cards would perform better on the particle demo than the actual physx card, due to sheer processing power, that the physx card doesnt have.

Comment by Evil_Clown666 at 6:49pm 14th April 2008

One important question that hasnt been answered is : How fast would the same particle demo be if it was run on a hardware PhysX card?

Comment by Omegahunter at 6:46pm 14th April 2008

Im impressed, the speed and use of this. with nvidia's close work with game manufactures, this could bring some impressive results in the near future and some value to the 9800.

Comment by TBallS at 6:14pm 14th April 2008

That is good news, just wish the 9 series was worth investing in ... bring on the next series + Cuda PhysX

Comment by CSQuake at 5:37pm 14th April 2008

• BFG unveils hand built Phobos gaming PCs
• UPDATED: LiveJournal announces mass layoffs
• Top 10 PC games chart
• Seagate Barracuda 7200.12 gets 500GB on single platter
• Creative reveals new Zii processor
• Happy Christmas from Custom PC
• Intel might buy Imagination Technologies
• First Blu-Ray and DVD hybrid disc announced
• Abit denies closure rumours
• Download ‘AAA’ games in a few minutes with AWOMO

These are the most popular news articles over the past seven days.

• Phenom II overclocked to 6GHz
• GeIL unveils Cyclone RAM cooler
• Rumour control: GeForce GTX 260 GX2 in January
• Nvidia cuts price of 216 SP GeForce GTX 260
• VIDEO: PC version of Mirror’s Edge features hardware PhysX
• Get Half-Life for less than a dollar on Steam
• IBM plans to build virtual human brain
• RivaTuner 2.20 adds SLI and CrossFire support
• VIDEO: ArmA II trailer released