SixtyWattMan
Supreme [H]ardness
- Joined
- Aug 4, 2008
- Messages
- 4,466
Keep trollin trollin trollin trollin trollin
![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
NVIDIA works with devs to add CUDA ocean sim and graphics effects to Just Cause 2
- Thread starter BlueStorm
- Start date
Keep trollin trollin trollin trollin trollin
YES YOU CAN....ATI is free to support this if they want to.
No you can't. Nvidia owns it, nvidia controls it, nvidia determines who gets to implement it. ATI can't just make a CUDA driver without licensing it from Nvidia.
In other words I can just completely ignore you, got it.
Are you going somewhere with that statement?
orangeblue
Limp Gawd
- Joined
- Oct 16, 2009
- Messages
- 325
So can anyone confirm whether or not these features are absent if you're running ATI hardware?
You really are deep in denial, aren't you?
DirectX is a proprietary API locked to a vendor specific product.
Stop being a hypocrite.
1.1.2.3.5...
Gawd
- Joined
- Jul 13, 2006
- Messages
- 527
Yes they can!!!!!
Are you being willfully ignorant or something?
In case you are too lazy....
http://www.google.com/search?source=ig&hl=en&rlz=&=&q=cuda+on+ati&aq=f&aqi=g2g-m2&oq=
In particular
http://www.maximumpc.com/article/news/cuda_running_a_radeon
http://www.extremetech.com/article2/0,2845,2324555,00.asp
http://www.techradar.com/news/compu...d-to-support-nvidia-s-cuda-technology--612041
It is AMD/ATI that do not want cuda to run on their cards.
Last edited:
It's funny that people can be told the absolute bottom line truth and completely choose to ignore it. There's nothing more I can do but give you that. It's your choice whether you will accept it or let cognitive dissonance get in the way.
Yes,let's change points since you last straw man fell down. You stated that anyone who doesn't think what nV is doing with CUDA just wants crappy console ports. A fallacy on so many levels it's painful, from insinuating we can't get good PC games if nV doesn't add proprietray code to games to stating that without this attempt at vendor lock in we have no choice but to receive marginal console ports.
Don't try to paint CUDA/PhysX/etc... as some kind of rewards program, that is just funny stuff right there.
Do you understand the concept of "licensing"? ATI would have to pay nvidia to support CUDA.
No, you've only stated what you *claim* the truth is - and you're credibility is minimal at best. Since you can't source your claims, as far as I'm concerned you're just making shit up, *especially* since there is evidence directly contrary to what you are claiming.
1.1.2.3.5...
Gawd
- Joined
- Jul 13, 2006
- Messages
- 527
Yes I do. Nvidia is NOT charging anything for this RTFA. Do you not get that?
I have given you plenty of sources. You just give me...well nothing. You basically have been condescending towards me and others all while not supporting anything you say.
1.1.2.3.5...
Gawd
- Joined
- Jul 13, 2006
- Messages
- 527
Pot meet kettle....
Brackle
Old Timer
- Joined
- Jun 19, 2003
- Messages
- 8,568
wow 1 game that uses Cuda....and everyone is freakin out.
Either way, Cuda is not going to stick.
Remember when the FX5900 came out and they didnt want to go by MS DX standard....that sure got them far didnt it?
The water that the OP links looks just like Crysis to me....
So If crysis can do it on any video card...why do we need Cuda
Either way, Cuda is not going to stick.
Remember when the FX5900 came out and they didnt want to go by MS DX standard....that sure got them far didnt it?
The water that the OP links looks just like Crysis to me....
So If crysis can do it on any video card...why do we need Cuda
its nVidia's strategy to attract customer and make sure customer buys their GPU with higher price tag.
greenfrogman
Ad Blocker - Banned
- Joined
- Jan 5, 2009
- Messages
- 387
games should use OpenCL and ATI should get there driver team sorted out and Bundle the OpenCL as part as the Driver install like Nvidia has for Quite some time now, as ATI is holding back OpenCL support (Do not point me to the ATI OpenCL SDK website as you need to make an account that no normal user is going to do)
if i use Nvidia i get the CUDA, Open CL, Physx ( XP to win7 supported) and MS Direct compute (Vista - Win7 only so not very useful for every one as GPU programming works with less issues under XP then it does under VIsta or win7)
ATI you get Direct compute only (Vista- Win7 only) and something called CAL thats not very optimized at all that basically no one uses apart from F@H + lots of CPU use (unless settings are changed)
if i use Nvidia i get the CUDA, Open CL, Physx ( XP to win7 supported) and MS Direct compute (Vista - Win7 only so not very useful for every one as GPU programming works with less issues under XP then it does under VIsta or win7)
ATI you get Direct compute only (Vista- Win7 only) and something called CAL thats not very optimized at all that basically no one uses apart from F@H + lots of CPU use (unless settings are changed)
MrLonghair
[H]ard|Gawd
- Joined
- Jun 7, 2004
- Messages
- 1,764
Nvidia TWIMTBP Program = Reagan administration
Vermillion
Supreme [H]ardness
- Joined
- Apr 5, 2007
- Messages
- 4,417
Actually you didn't read your own article.
Right smack in the middle of the ExtremeTech article it states that AMD/ATI would have to license PhysX in order to hardware accelerate the CUDA code. CUDA itself is worthless to AMD/ATI without PhysX.
From what I have read, you have addressed nothing and have done nothing but complain. As someone already pointed out, the feature thru Direct Compute/OCL on anything other than DX11 hardware is gimped and more than likely does not perform up to par. CUDA, even on lowly 8800GTs runs far faster and better than DC/OCL does on even a 4870.
The probrlem you have brent tho is this. How many DX11 cards are out there? HOw many are DC/OCL(my understanding anything DX10 or newer)? Which is faster and has more support, DC/OCL or CUDA? Which has the larger install base? and lastly Which enables these features to reach a leager number of people?
Answers, 2-3M atmost right now. All. CUDA. CUDA and again CUDA.
Phoenixdancer
Weaksauce
- Joined
- Mar 3, 2010
- Messages
- 69
... My word...
You can use Cuda and PhysX on an ATI card with the specific driver hacks.
Who would want to though?
CUDA and PHYSX will be over taken by something better soon anyway.
Plus if you have an ATI card and you want PHYSX you can buy a cheapo Nvidia card from ebay for next to nowt and use that
You can use Cuda and PhysX on an ATI card with the specific driver hacks.
Who would want to though?
CUDA and PHYSX will be over taken by something better soon anyway.
Plus if you have an ATI card and you want PHYSX you can buy a cheapo Nvidia card from ebay for next to nowt and use that
1.1.2.3.5...
Gawd
- Joined
- Jul 13, 2006
- Messages
- 527
Really? can you point it out to me? I think you mean,
So ATI would have to support cuda to run Physx, not the other way around. And cuda is not worthless without Physx
Just because they have the architecture to make it possible does not mean they would be licensing-free.
I think that's incredibly obvious.
Either way, nVidia has shown in the recent past that it has no intentions on working with ATi for a common good. By having developers close off AA to ATi cards and disabling PhysX when an ATi card is present, they are already showing the kind of predatory business tactics the red team would expect if they were to go into a mutual agreement.
Disregarding that, there is no telling what nVidia would demand for PhysX licensing on future release cards. Because ATi customers would expect PhysX, nVidia could charge whatever they please for their software and leave ATi cards without the price/performance selling point that they commonly fall back on. Leaving the pricing of your cards up to your only competitor is oligopoly suicide. It just doesn't make sense.
I think that's incredibly obvious.
Either way, nVidia has shown in the recent past that it has no intentions on working with ATi for a common good. By having developers close off AA to ATi cards and disabling PhysX when an ATi card is present, they are already showing the kind of predatory business tactics the red team would expect if they were to go into a mutual agreement.
Disregarding that, there is no telling what nVidia would demand for PhysX licensing on future release cards. Because ATi customers would expect PhysX, nVidia could charge whatever they please for their software and leave ATi cards without the price/performance selling point that they commonly fall back on. Leaving the pricing of your cards up to your only competitor is oligopoly suicide. It just doesn't make sense.
Really? can you point it out to me? I think you mean,
So ATI would have to support cuda to run Physx, not the other way around. And cuda is not worthless without Physx
It was in the same damn paragraph as the quote you posted for crying out loud:
Now compare CUDA on an 8800GT to DC/OCL on an 8800GT.
Vermillion
Supreme [H]ardness
- Joined
- Apr 5, 2007
- Messages
- 4,417
Really? can you point it out to me? I think you mean,
So ATI would have to support cuda to run Physx, not the other way around. And cuda is not worthless without Physx
You make me laugh. How about you actually quote the ENTIRE statement instead of just the piece that benefits you. Here I'll do you a favor by highlighting the piece you keep leaving out.
ExtremeTech said:
EDIT: Dammit kllernohj! You beat me to it.
1.1.2.3.5...
Gawd
- Joined
- Jul 13, 2006
- Messages
- 527
Wow...just wow...so are you saying there is a contradiction in the article?
If you do not understand this, then I am done. Nothing more can be said.
The part you quoted says that ATI would have to pay for PhysX. PhysX is not CUDA.
Running cuda is free. Why would the author contradict him/her self?
Last edited:
Because it's pro-nVidia.
Says nothing about hardware, states that writing programs is free.
Driver != Hardware.
The article implies that it is free to support CUDA through software, but does not mention the cost of implementing CUDA on hardware or licensing feed associated it. It's carefully worded to state absolutely nothing. Once you have quotes stating that ATi is welcome to implement hardware CUDA, you'll have an argument. Right now all you have is nVidia telling ATi that they can write their own drivers.
What are the license fees? Is this public knowledge? Are we certain there are license fees?
I don't think nVidia would spend time and money on GPU processing and give away the results for free. 11235 was directly stating that they are welcome to use it, which I believe is doubtful.
You're fucking taking it out of context entirely... nowhere during that video did they say this is only possible with Nvidia hardware, BUT it is Nvidia so it's also doubling as advertising for them just like when games same "Meant to be played with Nvidia".
All they said it's possible... and i'm pretty sure that if it was ATI was doing this you wouldn't have shit to say. You're just trying to flip this into something it isnt.
1.1.2.3.5...
Gawd
- Joined
- Jul 13, 2006
- Messages
- 527
What I said is true. The information is out there, and I have posted sources supporting my claims. Not a SINGLE counter example to anything I have said has been shown. Just a lot of fluff and misunderstanding.
A LOT of misunderstanding...
I am done with this thread...please feel free to spread misinformation.
The fault lies within the developers who are putting things into the game that only benefit a certain amount of their consumers. Considering the demo has TWIMTBP slapped all over it, I wouldn't be surprised at all if the development team was getting a little extra slice of the pie to make Fermi and nVidia look a little better after the numerous fiascoes they've had.
People are upset because they aren't getting the same game as nVidia users are. And rightly so.
What I said is true. The information is out there, and I have posted sources supporting my claims. Not a SINGLE counter example to anything I have said has been shown. Just a lot of fluff and misunderstanding.
A LOT of misunderstanding...
I am done with this thread...please feel free to spread misinformation.
Oh look, you completely avoided my response. What a surprise. There is no misunderstanding, just you posting that nVidia would give away their proprietary software so that both companies can hold hands and sing songs. I think we both know that is false, and you are just bending the facts to make it seem acceptable that the developers of this game are using gimmicks to implement features that could easily be done on the cpu.
Last edited:
A fair stance, but it'd do you well to substantiate it with something rather than try and discount it with conjecture about phantom licensing fees.
I don't have to prove my point; CUDA will never be licensed or implemented by ATi. The only think I "have" to do to "win" the argument is disprove or imply doubt that nVidia would love to give ATi CUDA.
The entire argument is pointless, and has little to do with the topic.
Someone in this thread has severe reading comprehension issues.
CUDA = The free to use
PhysX = Must license to use
It doesn't get any simpler than that, but some people in this thread can't seem to GRASP even the SIMPLEST of concepts when it comes to this stuff. But ohwell.
Reminds me of kids on the playground. Everybody is happy with one scoop of ice cream until they see the other guy with two. No different to what's happening here.
SixtyWattMan
Supreme [H]ardness
- Joined
- Aug 4, 2008
- Messages
- 4,466
If CUDA is free and better than OpenCL ATI should use it, end of story. PhysX needs to die quickly and painfully though.
Sure thing.
http://forum.beyond3d.com/showthread.php?t=56105
These are OCL results.
On stock HD5870 it takes:
2.27s for 16 queen
0.85s for 16 queen -local
Just for fun on HD5870@1130/1248:
2.07s for 16 queen
0.67s for 16 queen -local
66.0s for 18 queen -local
This is on the new one:
On stock HD5870 it takes:
3.40s for 16 queen
0.77 for 16 queen -local
Just for fun on HD5870@1130/1248:
2.80s for 16 queen
0.59s for 16 queen -local
30.50s for 18 queen -local
Here are some prelimeary results (with GeForce 8800GT) :
New -local 17: 4.41s
Old -local 17: 5.06s
New -local 18: 32.8s
Old -local 18: 38.8s
New -local 19: 270s
Old -local 19: 330s
Running on GTX 285:
New -local 17: 2.41s
Old -local 17: 2.44s
New -local 18: 15.1s
Old -local 18: 18.5s
New -local 19: 123s
Old -local 19: 155s
Running on GeForce GTX 285:
16 (gpu): 1.09s
-local 16 (gpu): 0.5s
17 (gpu): 4.01s
-local 17 (gpu): 2.55s
18 (gpu): 28.3s
-local 18 (gpu): 19.2s
GTX285 16 Queens
Cuda = .686s
OCL = .643s
Thats to noquater bringing something to my attention I over looked, I've added the newer result for the 5870 as they show mixed results. Still tho you have to question why an 8800GT can keep up with a 5870 when the later has so much more computational power compared to the latter.
Last edited:
That is one particular program, hardly conclusive. Indeed I could post the following link to a thread which shows the exact opposite.
http://www.ngohq.com/graphic-cards/16920-directcompute-and-opencl-benchmark.html
I will concede that CUDA is superior to OpenCL though.
Cool, shame its only benching the Compute shaders and not actually doing real computations. The n-queen bench being run in the link I posted is running actual computations on the CS, not benching how fast the CS can process a pixel or whatever. Its no secert ATI CS has been faster in synthitic benchs, but when put to use, it has come up slower which the n-queen bench proves.
Incase anyone is wondering, the n-queen bench can tak x number queen on a chess board and computes the random placement for the queens so as to not have any of them be able to take out another. It is also being developed by the guys behind F@H so its not like they are pulling favorites here.