NVIDIA, MSI, and Lucid speak about Hydra delay

[Biscuit]

going a bit off topic here guys...

[sammorris]

Due to the memory, the 4870X2 actually runs Warhead better at 2560 than the GTX295, substantially so. This is why 2GB per GPU on the next gen would have been nice, but it's unlikely to happen.
Agreed on the extra features, people are often saying we don't need more performance unless we use 'silly' resolutions. Well, those of us who buy top end cards ought to have 'silly' resolutions to match, and need the graphics grunt to take on today's demanding (but still not necessarily good looking) games engines.
The idle power consumption of the 5800s is also a big deal for me.

Anyway, agreed with Biscuit, that's enough of a Derail for now.

[ajones]

Maybe it's time that the powers that be stop trying to justify their actions (and I'm not pointing fingers here at any one manufacturer). Rather than driving the technology towards what the manufacturer believes Joe Public needs, maybe they should sit back and accept the criticism. We want good performance, low power and reasonably priced graphics card solutions.

Personally, I've not been burnt by Hydra/Batman AA/PhysX debacle, so can rather comfortably sit back and throw my oar in every once and a while. However I am fed up with fanbois.

Back to the topic;
nVidia; IMHO, either enable PhysX with an ATi powered graphics sub system, or accept that some of your customers are not gong to be happy about it.
All of this "back and forth" is becoming rather tiring and it's not doing anyone any favours by trying to defend yourselves to people who, through no fault of their own, have wasted their hard-earned

[Syllopsium]

@Biscuit - I think we can all agree that NVIDIA can choose to not spend our investment dollars to QA ATI + PhysX correct?

No, we can't agree and you're not correct.

I am not by any means an ATI fanboy, but I find Nvidia's actions completely unacceptable. This is a wider principle than graphics cards, too - the opinion of one vendor that they can disable functionality based on the presence of another card is almost without precedent and something that needs stamping on, hard.

It's very simple : if an ATI card (or indeed, a Matrox or anything else) for graphics and an NVidia card for PhysX doesn't work, the person whose fault it is fixes the issue. Since PhysX is device independent this would usually be an Nvidia issue. If by some miracle (unlikely) it's ATI's problem, then they fix it. If it's uncertain, both ATI and Nvidia test the configuration until it works.

That is a consequence of writing under Windows. You don't control all the software and hardware and have a responsibility to make things work together. Same with soundcards and Nvidia, network cards and Nvidia, SCSI cards and Nvidia, TV tuners and Nvidia etc.

Even given that I don't accept you can disclaim your responsibility to write good drivers, if this seeming incompatibility is such an issue : please document a case where an ATI card conflicts with an Nvidia card using drivers prior to Nvidia disabling the functionality.

Anyone tried running a game with a non ATI/Nvidia card as the primary device and Nvidia as PhysX?

The sad thing is, I am very happy with Nvidia hardware. I ran two passive 7600GTs (not in SLI - because of your - and AMD's - stupid multi card motherboard lock in) for years, and now run a 7600GT and a (second hand) 8800GTX. The 8800GTX is an incredible piece of engineering - quiet, powerful and stable. I've got a Zalman 3D monitor - using the Nvidia drivers and who knows - I may even buy a PhysX game at some point.

Nevertheless, whilst Nvidia's current tactics and PhysX lockout continue, I will not consider another Nvidia card - even if you release a DX11 card, and even if it's faster than ATI's unless ATI fails to fix their driver issues before the 8800GTX is so grindingly slow I can't run anything (come on ATI, get your finger out, comment on your forums, accept bugs have been replicated and actually fix them).

PK

[badass]

To be honest they could even just make a really rubbish registry hack that they "don't know about" and get really angry every time it's mentioned in public - i.e. physxdisable 1, and then in-the-know users can switch it to 0 and make it work.

Exactily. That IMHO is the best solution.

[NVIDIA_tap]

You mean, just like if a box appeared in windows stating "This is an unsupported configuration"?

Just admit it's a cop-out to sell more gfx cards. The line about being unsupported is marketing BS.

Funny how you have done nothing to compensate the people like myself who bought a nVidia card for Physx solely......it worked for months then you decided to disable my second card and now I have to resort to 3rd party hacks to use the NVIDIA card I purchased, for the purpose that I pirchased it.

Can you tell me where on the box of your gfx cards it states "Not to be used as a dedicated Physx card when an ATI card is present"?

Thanks for that.....makes me really want to purchase nVidia again........

For some time now we have been trying to communicate clearly our policy on running our GPU PhysX with a non-NVIDIA render. As you can imagine just getting a message out on something like that can be difficult. Our partners mostly have printed boxes for thier GPU months if not years ago so that won't work.

We do have our website here that try's to be explicit.
http://www.nvidia.com/object/physx_faq.html

Also rememer this was not possible under vista which prevented multiple GPU drivers from runnning at the same time. The only time this was enabled was on old XP drivers and an early win7 beta driver from us.

Nevertheless, I know is not great to buy something for a purpose that it turns out is not supported. I mentioned earlier in this thread that we will continue to improve our security on this so if you are not ourtside of your return window I recommend you bring the PhysX GPU back. Hurts nme to tell you that but honestly I don't expect the community hack to work indefinitely.

[NVIDIA_tap]

Nice work on the graphs, shame they're all Crysis, which we know ATI have caught up with now, and suffers less than perfect dual GPU scaling
The red bar in the tom's hardware benchmark though perfectly illustrates why I loathe the GTX295 as a product. So promising, but just when it gets good, it's useless...

That looks like a bug to me - and I have seen some data that shows we have a bug on some crysis levels with 8xAA which we are working on. BTW, we are not alone having issues with that config - I have seen similar issues with 5770 and some Xfire configs.

We do try to make thes things work across all games and configs but we do occasionally have some failures. i think that is also an illustration of my point about how frickin hard this stuff is and why we don't add support for configs that are not critical. Each new config costs...big time.

[NVIDIA_tap]

Nvidia also have had a few driver releases out haven't they?? Of course they also had a better chance to optimise things for their GPU too before release of the game. The same goes for Crysis Warhead which came out late last year.

Also the GTX295 cannot run Crysis and Crysis Warhead very well at 2560X1600 despite both being twimtbp games. What company puts less than 1GB of RAM per GPU on their highest end card for instance??

This GTX295 kicks ATI's butt nonsense is pointless when neither companies single card solutions are capable of running a two year game at 2560X1600 comfortably.

Perhaps GTX300 or whatever it is called will be able to do so but over two years seems a tad silly and by that time probably Crysis 2 will be released and we will be all back to square one again.

With the resolution of even cheap sub £300 monitors increasing, 2560X1600 or intermediate resolutions between 1920X1080 and 2560X1600 are not going to become that big a deal TBH. I wish ATI and Nvidia would stop adding all this value added BS to their cards and put maximum effort into the amount of DirectX compliant processing power in their GPUs especially the ones under £200 within an acceptable TDP.The value added BS only applies to a tiny percentage of all gamers and is more of a smoke screen IMHO.

I see your point about even our highend card still struggles with Crysis. It is a very heavy title. On the other hand though it is one of the few titiles that is that punishing. As a matter of fact I look at a lot of this stuff and I can say that I only know of 4 titles that are similarly intense.
Crysis, Dark Athena, Hawx, WIC.

These are the only games I see that pull 295 down below 45fps @25x16 with 4xaa. the other zillions do not, so of course it just becomes a trade-off. And IMHO while Crysis is a truly great game (would have been better if they just stayed in jungle) it is on outlier in terms of perf requirements.

[sammorris]

It's no bug, the GTX295 is physically unable to run games properly at 2560x1600 with high AA, because it simply doesn't have enough memory.
Sadly, I'm starting to disagree more and more about Crysis being an outlier for performance, it has plenty of companions now.

[NVIDIA_tap]

Maybe it's time that the powers that be stop trying to justify their actions (and I'm not pointing fingers here at any one manufacturer). Rather than driving the technology towards what the manufacturer believes Joe Public needs, maybe they should sit back and accept the criticism. We want good performance, low power and reasonably priced graphics card solutions.

Personally, I've not been burnt by Hydra/Batman AA/PhysX debacle, so can rather comfortably sit back and throw my oar in every once and a while. However I am fed up with fanbois.

Back to the topic;
nVidia; IMHO, either enable PhysX with an ATi powered graphics sub system, or accept that some of your customers are not gong to be happy about it.
All of this "back and forth" is becoming rather tiring and it's not doing anyone any favours by trying to defend yourselves to people who, through no fault of their own, have wasted their hard-earned

I thought I was clear on this point. We have no plans to enable PhysX running on an NV GPU with rendering on an AMD GPU. I accept that some users were burned by this decision and some are probably going to resent NV for some long period of time if not forever. And I do think we could have done a better job communicating this restriction when Win7 beta's became available. We should have been able to predict that folks would want to take advantage of a dedicated GPU for PhysX with and ATI card. We duffed that one.

[NVIDIA_tap]

It's no bug, the GTX295 is physically unable to run games properly at 2560x1600 with high AA, because it simply doesn't have enough memory.
Sadly, I'm starting to disagree more and more about Crysis being an outlier for performance, it has plenty of companions now.

Not so sure about that. Ill check with the boys here about the CW perf at 25x16. BTW, I love games with high GPU loads that look good. Big monitors, cool effects....that is why I come into work in the morning! And I hope you are right that this type of game become the norm. We push the horsepower of our GPUs up the curve every year and always have to hope that the content shows up to take advantage of it.

BTW, that is why we have a TWIMTBP program. It is not some strange obscure marketing only deal. It is the way we can systematically engage with new content designers to enable studios to quickly leverage new GPU technologies. We just build GPUs and hope content shows up. We help make it happen.

[NVIDIA_tap]

No, we can't agree and you're not correct.

I am not by any means an ATI fanboy, but I find Nvidia's actions completely unacceptable. This is a wider principle than graphics cards, too - the opinion of one vendor that they can disable functionality based on the presence of another card is almost without precedent and something that needs stamping on, hard.

It's very simple : if an ATI card (or indeed, a Matrox or anything else) for graphics and an NVidia card for PhysX doesn't work, the person whose fault it is fixes the issue. Since PhysX is device independent this would usually be an Nvidia issue. If by some miracle (unlikely) it's ATI's problem, then they fix it. If it's uncertain, both ATI and Nvidia test the configuration until it works.

That is a consequence of writing under Windows. You don't control all the software and hardware and have a responsibility to make things work together. Same with soundcards and Nvidia, network cards and Nvidia, SCSI cards and Nvidia, TV tuners and Nvidia etc.

Even given that I don't accept you can disclaim your responsibility to write good drivers, if this seeming incompatibility is such an issue : please document a case where an ATI card conflicts with an Nvidia card using drivers prior to Nvidia disabling the functionality.

Anyone tried running a game with a non ATI/Nvidia card as the primary device and Nvidia as PhysX?

The sad thing is, I am very happy with Nvidia hardware. I ran two passive 7600GTs (not in SLI - because of your - and AMD's - stupid multi card motherboard lock in) for years, and now run a 7600GT and a (second hand) 8800GTX. The 8800GTX is an incredible piece of engineering - quiet, powerful and stable. I've got a Zalman 3D monitor - using the Nvidia drivers and who knows - I may even buy a PhysX game at some point.

Nevertheless, whilst Nvidia's current tactics and PhysX lockout continue, I will not consider another Nvidia card - even if you release a DX11 card, and even if it's faster than ATI's unless ATI fails to fix their driver issues before the 8800GTX is so grindingly slow I can't run anything (come on ATI, get your finger out, comment on your forums, accept bugs have been replicated and actually fix them).

PK

I loved 8800 GTX as well. When it came out it was just so substantially better than anything that came before it. And 8800 GTX's success enabled us to reinvest in the next cycle of innovation. Today we see the product of that virtuous cycle:
- Build something compelling
- Bring it to market
- Make profit
- Reinvest to make the next big thing.

The reason we got to build PhysX, Stereo and even our next GPU is because of our prior successes and our focus on our core market - gaming, while simultaneously controlling and running an efficient business.

The decision to enable or disable PhysX is only revelevant because we have again created something our users value. NVIDIA could have done a better job communicating our policy on this on - no doubt. But accept that we need to run a business that generates profit to continue our cycle of innovation. We need to both control expenses and generate margin, and sometimes we need to make an unpopular decision to help the cycle and this is just one of those examples.
BTW, with a little work on your part I suspect you can find many examples of this. Have you ever noticed that many toyota parts don't fit on lexus? No good reason for that is there? How about the fact that Intel sells thier Atom CPU with a chipset for less than the price of Atom alone? What about MS just turning off the hacked xboxes? All of these are examples of companies making decision to protect thier businesses or IP.

Bottom line is there is nothing unethical or unusual about or decision on PhysX. And to put an even better point to it, There is nothing ususal or unethical about our decision on Lucid. We have no plans to QA it and I don't expect many people do.

[Biscuit]

I loved 8800 GTX as well. When it came out it was just so substantially better than anything that came before it. And 8800 GTX's success enabled us to reinvest in the next cycle of innovation. Today we see the product of that virtuous cycle:
- Build something compelling
- Bring it to market
- Make profit
- Reinvest to make the next big thing.

The reason we got to build PhysX, Stereo and even our next GPU is because of our prior successes and our focus on our core market - gaming, while simultaneously controlling and running an efficient business.

The decision to enable or disable PhysX is only revelevant because we have again created something our users value. NVIDIA could have done a better job communicating our policy on this on - no doubt. But accept that we need to run a business that generates profit to continue our cycle of innovation. We need to both control expenses and generate margin, and sometimes we need to make an unpopular decision to help the cycle and this is just one of those examples.
BTW, with a little work on your part I suspect you can find many examples of this. Have you ever noticed that many toyota parts don't fit on lexus? No good reason for that is there? How about the fact that Intel sells thier Atom CPU with a chipset for less than the price of Atom alone? What about MS just turning off the hacked xboxes? All of these are examples of companies making decision to protect thier businesses or IP.

Bottom line is there is nothing unethical or unusual about or decision on PhysX. And to put an even better point to it, There is nothing ususal or unethical about our decision on Lucid. We have no plans to QA it and I don't expect many people do.

The Atom argument yeah ok i can agree on that one Intel are being a bit crap there but their reputation is being tarnished because of it (to a certain degree). The rest of your metaphors are lacking somewhat in relative substance though.

I still dont actually see how Nvidias IP is damaged by using the card in a system with an ATI card also present if, as before mentioned, the user clearly goes through motions to prove that he understands the risks with using a non-officially supported configuration.

As for the Lucid, if you are not going to QA systems with it then how can you possibly allow it to go ahead? If ATI users dont get the privileged of working in harmony with Nvidia without going through testing why is it ok for Lucid?

[ajones]

I thought I was clear on this point. We have no plans to enable PhysX running on an NV GPU with rendering on an AMD GPU. I accept that some users were burned by this decision and some are probably going to resent NV for some long period of time if not forever. And I do think we could have done a better job communicating this restriction when Win7 beta's became available. We should have been able to predict that folks would want to take advantage of a dedicated GPU for PhysX with and ATI card. We duffed that one.

It takes nothing to continually ignore the facts been presented in a number of posts in these forums over the last few days, but I for one am happy finally see an admission and someone willing to answer our questions rather than bleat on about it...

Thanks for that.

[NVIDIA_tap]

The Atom argument yeah ok i can agree on that one Intel are being a bit crap there but their reputation is being tarnished because of it (to a certain degree). The rest of your metaphors are lacking somewhat in relative substance though.

I still dont actually see how Nvidias IP is damaged by using the card in a system with an ATI card also present if, as before mentioned, the user clearly goes through motions to prove that he understands the risks with using a non-officially supported configuration.

As for the Lucid, if you are not going to QA systems with it then how can you possibly allow it to go ahead? If ATI users dont get the privileged of working in harmony with Nvidia without going through testing why is it ok for Lucid?

We believe that when a user buys a board that has the Lucid chip on it - there is no doubt that the Lucid chip or the motherboard manufacturer is responsible for the quality of that solution. You are right though that by allowing it we do expose ourselves to potential quality problems as some users won't be able to affix responsibility where is belongs. In this case we decided the technology will sink or swim on it's own merits - and we will just try to work though the issues as the come.

PhysX is not the same. Since we own the PhysX software and GPU and issues with PhysX and AMD will inevitably be a job for NVIDIA to fix.

[NVIDIA_tap]

It takes nothing to continually ignore the facts been presented in a number of posts in these forums over the last few days, but I for one am happy finally see an admission and someone willing to answer our questions rather than bleat on about it...

Thanks for that.

I don't know what bleating is ...does not sound good though, but I try to be straight forward with my comments and views. Sometimes...I speak a bit too quickly...inserting foot-in-mouth occasionally - maybe even tweaking out a few people... But usually I stay out of trouble and focus on the issues.

NVIDIA in general does not ignore issues...hence I am here, now. Can't do this all the time :> but it seemed worth doing on this issue.