Just got my instant coffee, time to take out the trash.
Originally posted by: instantcoffee
The advantage is a continous hardware optimizaton for the API for the years to come. With major support. Optimizations for everyone. Not everything is good to throw at the GPU, since the GPU has other tasks then to calculate the physics.
You do realize both AMD and Intel also understand the future of physics is hardware acceleration right? The only reason Intel would capitulate to a GPU accelerated, non-x86 based solution would be to keep Havok relevant by matching PhysX's hardware capabilities.
For us consumers, its better with Havok that has been constantly used as hardware agnostic middleware, in contrary to PhysX, which currently is more used to push hardware. It also takes longer time with PhysX to get GPU accelerated physics for everyone, since its using CUDA as platform. The result of PhysX's approach is that the GPU accelerated physcs exists only as an addition to the game instead of a part of the game. As seen with Mirrors edge.
PhysX is only used to push hardware? That's odd, prior to this week I thought it was used to push hardware PhysX capabilities previously impossible on existing hardware and software solutions (discounting Ageia PPU).
Also, how does it take longer with PhysX's approach just because its CUDA? You think it would've taken 8-9 months for AMD to write some working CUDA drivers for AMD parts?
The SDK isn't released to write the drivers, but to write applications for ATI cards that will access the cards resources. I don't think you understand what the SDK is for, if you think its there so people can write drivers for ATI cards.
Rofl, I don't think you understand neither are needed to write a driver for AMD's hardware, but they certainly would be necessary for API and software validation to ensure the drivers actually did what they were written to do. The point is that AMD's claim about "closed and proprietary standards" preventing them from supporting PhysX is a farce.
Misinformation and lies comes from you and according to your post history, you have done that for years. Guerrilla marketing for Nvidia at any chance you get. You're even trying to twist facts to make bad publicity for AMD and good for Nvidia. Just look at the quote from your post.
Rofl misinformation and lies coming from me? I've been spot on with my analysis of hardware physics and all the different API in play since day 1 with few exceptions. I've repeatedly corrected you in this thread with facts and references disputing your false claims. Seeing as you haven't yet mentioned Havok FX, OpenGL, or claimed Nvidia can't support OpenCL in this reply its obvious facts have prevailed.
But none if this is really any surprise given your brief posting history and propensity for posting misinformation, although I will say low post count + high concentration of FUD certainly makes it easy to remember people like you.
I guess you don't recall when you first posted, guns ablazing full of "facts":
- "Facts" right?
Nvidia getting sued by Microsoft for buggy drivers......We don't have all the material for the lawsuit, but Microsoft obviosly feels the evidence is good enough to bring to court. They probably have more evidence then submitted reports, since they are suing Nvidia for bad drivers.
Games crashes on Nvidia cards even when they have "worked close with the developers". A good example might be Assassins Creed, where the DX10.1 caused bugs on Nvidia cards, so it needed to be removed in the next patch. Or Fallout 3 running faster on ATI cards when it first came, even though it was a TWIMTBP title.
So yes, please let me know if I ever post anything as egregiously false as that, or anything you've posted in this thread (or here ever, really) and make sure to back those claims up with actual facts and references and not whatever inaccurate "facts" you fabricate next.
AMD is not segregating the GPU physics market. Thats another lie from you. First of all, Havok belongs to Intel, not AMD. Secondly, Havoks OpenCL approach gives GPU physics for everyone, not only geforce cards. I am defending AMD's choice of selecting a hardware agnostic alternative versus something Nvidia uses to push hardware. Since I am buying Nvidia and ATI GPU's, AMD and Intel CPU's, Havok is the best alternative for me as a consumer.
No, its not a lie. If they supported PhysX as it became available, and also supported Havok, you'd have a point. Instead they gave deceitful reasons about not supporting "closed and proprietary standards" and not moving forward with GPU Physics until it "makes sense". End result is that the resulting segregation of the GPU physics market has retarded the adoption of GPU physics for
at least another 8-9 months plus whatever lead time before Havok's OpenCL client is released for public consumption.
Thats the difference between us as well. You want to market Nvidia whenever you get the chance, even by lying as above. AMD's choice is not segregation, but a unification. With OpenCL, both ATI and Nvidia gets Havok support by default. With CUDA, only Nvidia cards get PhysX (on GPU). Nvidia's approach segregates the GPU market.
Uh, no I've pushed for GPU accelerated physics from Day 1 because I saw the potential for making games better. I honestly do not care about what standards or middleware is being used, as long as the
hardware I own supports it and the application results in better games. That's very different than all the AMD apologists providing reasons for why AMD shouldn't support a new innovative feature their hardware is perfectly capable of supporting and advancing the industry.
Also you seem to be forgetting the fact OpenCL was only ratified a few months ago and is very similar to CUDA (again, given Nvidia chaired the standard, and probably wrote the code to largely mirror CUDA). If AMD had worked on a CUDA driver that effort would've been value add toward an OpenCL driver, which is the same source of optimism Nvidia owners should have for seamless OpenCL Havok support. There is no technical reason the hardware cannot support it and the APIs are very similar, so the inverse should hold true for AMD and PhysX even if they were committed to OpenCL.
I think you should read it again (and note that its not about licensing agreement, but what they actually use):
<links>
Makes your post about licensing agreements being history a bit controdictive to your previoius post about those agreements. But, everything is allowed when guerilla marketing Nvidia, right?
Rofl, you still don't understand, I've never claimed PhysX support is greater than Havok support. I claimed PhysX support is
growing, as is clearly indicated from major publishers, platforms, and developers licensing the SDK. Surely you understand that allows for a greater probability that PhysX is used in games going forward, right? If PhysX was insignificant and not gaining support, why would these developers sign licensing agreements to use them to begin with?
Also, there's no need to engage in guerilla marketing, or even gorilla as it pertains to you seeing as I've had to bludgeon you repeatedly with facts to make any progress, as these points should be clearly obvious to any reasonably minded individual. No one has claimed overnight support. That's the whole point and why AMD's subterfuge with regard to PhysX was so deceitful. Because game development has significant lead time so that the earlier adoption of features means the higher likelihood of seeing those features in actual games.
I think its more your understanding thats a bit limited here. That CUDA provides support for OpenCL, DX11 etc. doesn't mean that OpenCL, DX11 etc. supports CUDA. You cannot access CUDA through DX11 or OpenCL. Its not a two way street.
LMAO,
my limited understanding? At least you haven't used Havok FX or OpenGL once in your replies yet, so we're clearly making progress at bringing
your limited understanding up to speed.
Also, I haven't claimed its a two way street and completely compatible, they'd obviously need whatever software translation needed whether via wrapper or custom driver, but again, given how similar the API are there's no technical reason preventing support. Dave Baumann's (AMD) and Nadeem Mohammed's (Nvidia) statements here and elsewhere have absolutely backed my points on this.
The only thing you have been advocating, is what you always advocate: Nvidia, CUDA and PhysX. Again you repeat the lie about AMD advocating segregation of the market. I'll repeat and add:
Intel owns Havok, not ATI.
Havok is currently the biggest physics middleware provider and existed before PhysX.
LMAO, its plainly obvious now that AMD's previous smoke and mirrors stance on PhysX has done nothing but segregate the GPU physics market and retarded the implementation of accelerated physics effects in games. While the end result might be PhysX and Havok both supporting AMD and Nvidia hardware, the cost will be another 8-9 months + however long it takes for OpenCL Havok to be made publicly available.
Havok's GPU acceleration is hardware agnostic through OpenCL. Both ATI and Nvidia benifits from it and will be able to get it by default. Unlike PhysX, where Nvidia won't let anyone else use it (on GPU( unless they adopt CUDA as well.
And? You didn't even know what OpenCL was before this week, much less OpenCL Havok (again, refer to previous use of OpenGL + Havok FX), and it certainly doesn't absolve AMD of all the rubbish they've fed everyone about GPU physics for the past 8-9 months. As for the bit about Nvidia lol, well again, its obvious your arguments have no merit.
As long as Nvidia is using PhysX to push hardware and CUDA, I don't think you will see a widespread use of GPU PhysX physics.
LMAO, you seem to be missing the big picture here. Nvidia is pushing GPGPU to extend the usefulness of their hardware and diversify into non-GPU markets, not a PhysX middleware they paid 100 million for. CUDA and by extension, OpenCL are just a means to that end and both are compute architectures largely designed and written by Nvidia to achieve that goal. They don't care what software exposes this hardware functionality, as long as their hardware is eating most of the pie (it is), they just want to grow that pie (it will with Havok, or PhysX, or both). The only way to ensure they'll continue to eat the majority of that growing pie is to support all standards and middleware while offering the best performance.
With Havok on OpenCL, developers will be able to use GPU accelerated physics and reach everyone. ATI supports it, Nvidia supports it and Intel will support it as well. Consumers win.
Which would've been the same outcome 8-9 months ago if AMD supported PhysX.
AMD is not advocating segregation of the market. Thats a lie you should stop spreading. If anything, they are supporting Intel's Havok which covers the whole market, instead of a part of it.
LMAO, you can ignore the last 8-9 months prior to OpenCL Havok's announcement all you like, it just makes you look like a revisionist. If they supported PhysX in August it would've been 100% of the GPU physics market, just as it will be going forward now (presumably).
Oh and I don't really drink instant coffee, it tastes like shit and leaves a bad after taste.
Facebook
Twitter