I'm buying new card and I was thinking to go Nvidia. 1)I was wondering does Ati have any similar software like Physx and what is equvalent to CUDA cores in Ati cards ? 2)Will be Physx be abandoned or replaced in future years ? If it will, is getting Nvidia still good option ? 3)How much Nvidia card gains performance using Physx ? I think theese que. will make my choice. THx
is this new card for your current system listed in your profile?? If so then you without sounding like an ass, but you may want to think about a system overhaul..... If i'm correct your motherboard will not support any modern graphics card due to the lack of a PCIE slot.....
well I love Nvidia so i'll always promote them..... BUT, PhysX isn't used in a huge amount of games, designers tend to opt for their own physics engines that can be used on any system..... That said, the few games that do use it look stunning, just look at any of the batman games etc..... It will be around forever, as its always being developed, and yes games will still come out that utilize it...... but tbh, you can always add a cheap nvidia card and run it as physX card alongside an ATI card. So its all about personal preference.... I prefer Nvida so I say go get an nvidia card!!
Well being an ATI/AMD guy I won't bash Nvidia. But the number of games that really utilize PhysX could pretty much be counted on one hand. I would never factor that into my decision on purchasing a GFX card. I would look at cards in price vs performance not a "novelty" that PhysX really is.
1. AMD can compute physics (not PhysX) using DirectCompute - component of DirectX 10/11 or OpenCL, available for all modern GPU's. I don't know if there is any game that compute physics through DirectCompute or OpenCL right now (3D Mark 2011 benchmark uses DC to calculate physics). PhysX run on CUDA and is being utilized in only a handful of games. AMD has APP as a direct counterpart for CUDA. Just not to mix up technologies: CUDA - NVIDIA's standard for using GPGPU, recently went opensource APP - AMD's standard for using GPGPU OpenCL - open standard for harvesting power of GPU (can be mixed with CPU). DirectCompute - part of DirectX 10/11, can be used to calculate physics PhysX - NVIDIA's implementation of physics calculation engine running on CUDA 2. I think NVIDIA's PhysX will be abandoned or ported to open-source or general platform like OpenCL or DirectCompute. I hope AMD and NVIDIA will come up with global physics acceleration platform that can use any modern GPU. Or maybe Microsoft will develop one itself based on DirectCompute? Time will tell. 3. You don't gain performance using PhysX. On the contrary, when using PhysX, your GPU has to divide resources between graphics and physics calculation, so your FPS could take a hit. Sometimes with less complicated PhysX games, it can be run on CPU, but it's usually X-XX times slower than using GPU. General thoughts: judging by PhysX alone, I would not consider this as an argument to go for NVIDIA. There are too few games that actually use it, with 1-2 premieres/year. CUDA on the other hand is being used in a number of applications (video editing, graphic software etc.) and it can speed up calculations significantly. So if you work with application that could utilize CUDA, then it could be an argument to go for NVIDIA. I'm still hoping for general standard for physics acceleration, so it could be implemented in more games without the need to own a specific GPU. If you are buying new setup in a couple of months then maybe wait what Keppler will bring to the table. Also prices for AMD 7000 series should drop. I would recommend buying the best performing card for your budget, no matter if it is NVIDIA or AMD. I've owned both and both were great. I have always picked up the best performer for my budget from either AMD/ATI or NVIDIA.
^^ great answer Just to expand on a couple points: NVIDIA uses the term CUDA to refer to both the software API that programs likes PhysX run on as well as the the actual hardware cores on the GPU. AMD's equivalent to this are known as Stream Processors (SP), though you can't directly compare the two types of cores. In general, ATI GPU's have far more numerous, smaller cores than their NVIDIA counterparts. For reference, the GTX 580 has 512 CUDA cores, the 7970 has 2048 SPs. The other interesting thing about all of this is that NVIDIA just made CUDA semi-Open Source, so it's entirely possible we could see CUDA applications running on AMD hardware in the not too distant future. It's worth noting that the Open Source did not necessarily extend to PhysX itself, though.
You are being a bit dismissive http://uk.geforce.com/hardware/technology/physx/pc-games You get a more immersive experience with PhysX. If the game has it, its worth turning on. Note there is hardware (GPU) PhysX and software (CPU) PhysX. The above list linked are those games using hardware.
Yeah but many of the games on that list use CPU-only PhysX that don't use the GPU at all. And of the ones that do, how many are AAA titles? Personally the only games I've played that I felt like PhysX made a big difference were the two Batman games.
The list of PhysX games is pretty big. Those games I linked above are hardware PhysX, using the GPU. There are more PhysX games linked here (including consoles). http://developer.nvidia.com/physx-games I've seen a much longer list, not sure where it is. Mafia II and Metro 2033 make good use of PhysX and the Batmans as you said. I heard the new Alice game is very good.
True, there's 300+ cpu based physx games. Would be nice to see more hardware accelerated ones but oh well, maybe physx 3.x will bring more out with its performance increase.
