Criticism of Nvidia's TWIMTBP Program - HardOCP's Just Cause 2 Review

[Benetanegia]

I don't agree. DX is the first thing that should die and open the door to OpenGL again. DX is not open, it's not a standard and most times than not is not what game developers want. That is why they use things like CUDA.

Not to mention that with CUDA or OpenCL you can do much much more things than you can with DirectCompute.

And if DX is what AMD choses, of course they should promote it because it's their job. M$ and DX is just the bridge between the real players on the gaming making bussiness. M$ should not be the one making decisions about which fetures are in and forcing their implementation. That's the job of game developers and hardware providers. M$ should just act as the bridge that it really is between them and offer those features that devs and GPU makers demand.

Then there's the Xbox. As long as M$ has a console, they should have nothing to do with PC gaming. Period.

Regarding GPU physics. I know very well why I want GPU physics and yes it is performance related. Rather than performance it's the amount of detail. I laugh everytime I see someone say how great physics on Bad Company 2 are. They are garbage compared to what it could be done (and has been done on PhysX demos) with just a 8800GT. JUst because the GPU power is not being used in games, doesn't mean it couldn't b used for real fully destructible environments. In the sledge demo Nvidia showed a single Fermi card simulating 1 million interactive particles at a time smoothly. Bad Company 2's physics are a joke technologically.

 

[cadaveca]

Not to mention that with CUDA or OpenCL you can do much much more things than you can with DirectCompute.

Actually, that's not true. DXCompute can be used for basically anything...it's just that the solutions you mention (maily CUDA)have ready-baked code. That's lazy...understandable, but still lazy.

This is what "support" for these api brings...ready baked code. That is it.

Why the hell should AMD, a hardware company, be publishing software?

nVidia is NOT a hardware company...they are a softwarecompany that also sells hardware. I can provide video of Jen Hsun saying this exact thing.

If that's what you expect of AMD...well...nV doesn't make cpus. They wish they could, but will NEVER get an x86 liscence. That's why they are doing what they do...they are creating a software platform.

AMD is NOT in the business of selling software, or a software platform..they provide HARDWARE platforms.

 

[Benetanegia]

Actually, that's not true. DXCompute can be used for basically anything...it's just that the solutions you mention (maily CUDA)have ready-baked code. That's lazy...understandable, but still lazy.

This is what "support" for these api brings...ready baked code. That is it.

Why the hell should AMD, a hardware company, be publishing software?

nVidia is NOT a hardware company...they are a softwarecompany that also sells hardware. I can provide video of Jen Hsun saying this exact thing.

If that's what you expect of AMD...well...nV doesn't make cpus. They wish they could, but will NEVER get an x86 liscence. That's why they are doing what they do...they are creating a software platform.

AMD is NOT in the business of selling software, or a software platform..they provide HARDWARE platforms.

Hardware is nothing without software. Like I said AMD expects the other do all the work for them. If you want to sell your hardware you need to give a reason why consumers should pay for it. And I'm not talking about publishing, I'm talking about sending engineers and helping develop and include features in games.

 

[cadaveca]

I understand what you are saying.

What I am saying, is based on those points i jsut mentioned, it's unrealistic to expect AMD to do that.

It's RIGHT to expect it...heck..Intel does the same(although not to the extent that nV does), but that doesn't mean it will happen.

The market is SCREWED UP, period, and nV started it all, pushing forward PS3.0 in DX9. DX is only broken, thanks to nV.

BATMAN AA is the perfect example..running MSAA, using DX10.1 code in DX9. Why the heck didn't they just use DX10.1? Business.

And nV's business...isn't selling hardware. It's selling software.

You are asking AMD to become a software company...and THAT would only make this situation worse!

 

[MadClown]

Wait, aren't game developers supposed to make games? And aren't GPU companies supposed to make GPUs and not co-develop games? If both types of companies stuck to their own thing, we as human's could have made greater progress.

 

[newtekie1]

Wait, aren't game developers supposed to make games? And aren't GPU companies supposed to make GPUs and not co-develop games? If both types of companies stuck to their own thing, we as human's could have made greater progress.

I don't believe we would be anywhere near where we are right now if everyone just stuck to what they were supposed to do and never ventured into new territory.

 

[entropy13]

I don't believe we would be anywhere near where we are right now if everyone just stuck to what they were supposed to do and never ventured into new territory.

Specialization contributed greatly to the modernization of the world. I don't know where you're getting that conclusion. It's just that recently, because of those advances themselves specialization was forgotten and multi-disciplinary "specialization" became the vogue. Hence a hardware-making software company is possible.

 

[newtekie1]

Specialization contributed greatly to the modernization of the world. I don't know where you're getting that conclusion. It's just that recently, because of those advances themselves specialization was forgotten and multi-disciplinary "specialization" became the vogue. Hence a hardware-making software company is possible.

So you really think we would be as advanced as we are if no one tried anything new?

I mean, lets just use an example of your logic.

There is a company, that started as a logic chip maker, then moved into the RAM industry, then moved into the processor business, and today is the primary competitor to Intel...yes I'm talking about AMD. And because they changed their business focus, and expanded what their business does, we have advances in CPU technology that we would have never seen without them making that move. Do you really think Intel would have advanced the industry as much if it wasn't for the need to try to stay ahead of AMD? We would all be still stuck on single core processors as powerful as PIIIs. Hell, we wouldn't even have x64.:shadedshu

 

[entropy13]

So you really think we would be as advanced as we are if no one tried anything new?

I mean, lets just use an example of your logic.

There is a company, that started as a logic chip maker, then moved into the RAM industry, then moved into the processor business, and today is the primary competitor to Intel...yes I'm talking about AMD. And because they changed their business focus, and expanded what their business does, we have advances in CPU technology that we would have never seen without them making that move. Do you really think Intel would have advanced the industry as much if it wasn't for the need to try to stay ahead of AMD? We would all be still stuck on single core processors as powerful as PIIIs. Hell, we wouldn't even have x64.:shadedshu

So AMD is currently in the RAM industry as well?

I thought you're using my logic? Since when did I even imply that once you get a specialization you can't change it? What I was saying is that instead of being focused on one thing, people (and companies) keep on "focusing" on many things, albeit all related, ALL AT THE SAME TIME. Which is for all intents and purposes hardly something you can call "specialization". But I'm not saying as well that doing that is necessarily a bad thing. It's just that sometimes, it just really doesn't work out.

Even in AMD itself there's still a "specialization" of sorts.

During this time, AMD attempted to embrace the perceived shift towards RISC with their own AMD 29K processor, and they attempted to diversify into graphics and audio devices as well as EPROM memory. It had some success in the mid-1980s with the AMD7910 and AMD7911 "World Chip" FSK modem, one of the first multistandard devices that covered both Bell and CCITT tones at up to 1200 baud half duplex or 300/300 full duplex. The AMD 29K survived as an embedded processor and AMD spinoff Spansion continues to make industry leading flash memory. AMD decided to switch gears and concentrate solely on Intel-compatible microprocessors and flash memory, placing them in direct competition with Intel for x86 compatible processors and their flash memory secondary markets.

A more common example of my reasoning in this:
http://blogs.static.mentalfloss.com/blogs/archives/35163.html

In Avon's case, he started selling books, then added some perfumes. The perfumes were more popular, so he SPECIALIZED on that (perfumes and cosmetics). Avon's not selling BOTH cosmetics AND books anymore. In Nokia's case, they started as a paper mill, but they aren't BOTH a paper mill AND a phone maker.

The ones that do well in diversified industries are quite big, hardly the size of an Nvidia.

 

[ctrain]

And if DX is what AMD choses, of course they should promote it because it's their job. M$ and DX is just the bridge between the real players on the gaming making bussiness. M$ should not be the one making decisions about which fetures are in and forcing their implementation. That's the job of game developers and hardware providers. M$ should just act as the bridge that it really is between them and offer those features that devs and GPU makers demand.

DX features are primarily implemented based on developer request, MS doesn't force anything. Actually, I think geometry shaders might have been their idea though, but in reality, basically everything is requested.

 

[Benetanegia]

DX features are primarily implemented based on developer request, MS doesn't force anything. Actually, I think geometry shaders might have been their idea though, but in reality, basically everything is requested.

Acepting requests and implementing some in the way M$ wants, is not listening to developers exactly. The fact of the matter ir that OpenGL is the one API that is created and updated by a consortium formed by hundreds of companies, game developers and academic figures. And another fact is that OpenGL and DX are so different, not in the features they have, but in how they are implemented, which is what I was talking about. Is been long since M$ cares much more about their console and making multi-platform migration easier than it is about making PC gaming the best thing posible for PC gamers. And again I'm not talking about the features, but how and when they are implemented and how and when and to what extent M$ decides to promote them. When a OS launch is near they will promote it to death, but after that they go mute as if pretending to hide the dust below the carpet. The fact that they pay for Xbox exclusives that not only prevents the game from going to the PS3, but also delay the PC release by several months doesn't help create the sense that M$ is promoting PC gaming either.

 

[Mr McC]

Acepting requests and implementing some in the way M$ wants, is not listening to developers exactly. The fact of the matter ir that OpenGL is the one API that is created and updated by a consortium formed by hundreds of companies, game developers and academic figures. And another fact is that OpenGL and DX are so different, not in the features they have, but in how they are implemented, which is what I was talking about. Is been long since M$ cares much more about their console and making multi-platform migration easier than it is about making PC gaming the best thing posible for PC gamers. And again I'm not talking about the features, but how and when they are implemented and how and when and to what extent M$ decides to promote them. When a OS launch is near they will promote it to death, but after that they go mute as if pretending to hide the dust below the carpet. The fact that they pay for Xbox exclusives that not only prevents the game from going to the PS3, but also delay the PC release by several months doesn't help create the sense that M$ is promoting PC gaming either.

There is little that I can not agree with in your post; however, given that the Xbox is nearing the end of its life, we may have a small window of opportunity for advances in DirectX 11 gaming on the PC between the demise of the current console and the next generation replacement.

A number of contributors have pointed out that ATI, and more importantly Microsoft, whilst promoting open standards, have failed to deliver tangible results. I believe that there is a great deal of truth in this accusation, but I don't see Nvidia's push towards proprietary technology as the solution.

 

[cadaveca]

If nVidia did nothing but provide the HAL for Windows, then they'd be a much stronger force. Everything they do is convoluted by teh fact they sell hardware as well. This is why I think AMD and nV should merge...ATI hardware engineers, and nV software engineers...would be unstoppable.


But really, they are selling hardware so that they can pay all the employees. Note that through this "global recession", nVidia didn't lay anyone off.


As a business, few can compare to nV. But the process of making money is interfering with the industry moving forward as a conglomerated unit.

 

[newtekie1]

So AMD is currently in the RAM industry as well?

I thought you're using my logic? Since when did I even imply that once you get a specialization you can't change it? What I was saying is that instead of being focused on one thing, people (and companies) keep on "focusing" on many things, albeit all related, ALL AT THE SAME TIME. Which is for all intents and purposes hardly something you can call "specialization". But I'm not saying as well that doing that is necessarily a bad thing. It's just that sometimes, it just really doesn't work out.

Even in AMD itself there's still a "specialization" of sorts.



A more common example of my reasoning in this:
http://blogs.static.mentalfloss.com/blogs/archives/35163.html

In Avon's case, he started selling books, then added some perfumes. The perfumes were more popular, so he SPECIALIZED on that (perfumes and cosmetics). Avon's not selling BOTH cosmetics AND books anymore. In Nokia's case, they started as a paper mill, but they aren't BOTH a paper mill AND a phone maker.

The ones that do well in diversified industries are quite big, hardly the size of an Nvidia.

Yes, but the original argument, the one you are defending, is that nVidia is a hardware company, and they should stay a hardware company and not attempt to move into new areas.

AMD was in the RAM industry, they aren't currently, because they have moved into something new and more sucuessful. However, at one point they were doing both RAM and processors. They also don't just sell processors currently, they do chipsets and graphics cards now(through the ATi merger). That is hardly specialization also.

Avon, your example, proves my point perfectly. By the argument that you are trying to defend, he should have stayed selling books, and never moved to perfumes and cosmetics. And at one time, Avon WAS selling books and perfumes at the same time. He didn't just decide one day that he was going to instantly stop selling books and move to cosmetics/perfumes.

Focusing on what is profitable and you are good at is one thing, and a smart business sense, but at the same time trying new areas is important for progress.

Just look at Henry Ford as an example. By all means, if we follow the logic you are trying to defend, he should have stayed a farmer. Never trying anything new, never venturing into new areas. He was a farmer, he should have farmed. But instead he did move into new areas, first moving into steam engines, and then becoming a founder for the automobile industry as we know it including inventing the modern assembly line method, something that allowed humans to progress to the point we are at today. And all that would not have happened if we simply said a farmer is a farmer, they should't do anything else. So, do you still want to defend the logic that a GPU company is a GPU company, and they shouldn't do anything else?

 

[SNiiPE_DoGG]

where to start.... for one thing you are completely off base with this "trying something new" thing. Lets stick to the companies were actually talking about eh? and not abstract analogies to different markets and companies in different eras.

Nvidia goes to game developers and gives them money. The developer reciprocates by optimizing for nvidia hardware. Thats not "trying something new" that's paying to stop the devloper from optimizing for both brands.

IF nvidia were to give the game devs no money, is it in the developers best interest to optimize for only 1 hardware brand? hell no. They want to optimize for all hardware so people buy their game and don't complain about poor performance on either side.

Instead they are getting nvidia's money so they have incentive to only optimize for one hardware brand; this hurts progress.

 

[Mr McC]

where to start.... for one thing you are completely off base with this "trying something new" thing. Lets stick to the companies were actually talking about eh? and not abstract analogies to different markets and companies in different eras.

Nvidia goes to game developers and gives them money. The developer reciprocates by optimizing for nvidia hardware. Thats not "trying something new" that's paying to stop the devloper from optimizing for both brands.

IF nvidia were to give the game devs no money, is it in the developers best interest to optimize for only 1 hardware brand? hell no. They want to optimize for all hardware so people buy their game and don't complain about poor performance on either side.

Instead they are getting nvidia's money so they have incentive to only optimize for one hardware brand; this hurts progress.

That is more or less my interpretation of the situation and my only real possibility of protest is to refuse to buy where I feel that a given game has overstepped the mark by optimising features for Nvidia that should be universally available.

 

[newtekie1]

where to start.... for one thing you are completely off base with this "trying something new" thing. Lets stick to the companies were actually talking about eh? and not abstract analogies to different markets and companies in different eras.

Nvidia goes to game developers and gives them money. The developer reciprocates by optimizing for nvidia hardware. Thats not "trying something new" that's paying to stop the devloper from optimizing for both brands.

IF nvidia were to give the game devs no money, is it in the developers best interest to optimize for only 1 hardware brand? hell no. They want to optimize for all hardware so people buy their game and don't complain about poor performance on either side.

Instead they are getting nvidia's money so they have incentive to only optimize for one hardware brand; this hurts progress.

However, what we are talking about is not just the optimizations parts that included in TWIMTBP, but also the additional software elements that nVidia is developing to be added to games, and at this point in the discussion CUDA based software.

Now, as for nVidia giving money to the developers to optimize for their hardware, I think you're wrong there. I believe if nVidia didn't give the money for optimization on their hardware, we would see games that had little or no extra optimization for either. The games would still run more than reasonably on both ATi and nVidia hardware. It certainly isn't the case where TWIMTBP games run like shit on ATi hardware, despite you believing they have no optimization for ATi hardware. If nVidia didn't pay for the optimization for their hardware, the developers would certainly not be forced to optimize more for both, that is a false belief, they instead would not optimize more for either.

Think of it like this:

The develper designs the game and optimizes it to the point where it run at 30FPS with Max Settings on equal hardware from each company(say an ATi HD4890 and a nVidia GTX275). They are done with the amount of optimization they are going to do. Then nVidia comes in and says they will pay for the developement costs to optimize the game so it runs at 45FPS on the GTX275. If nVidia hadn't use the TWIMTBP program to improve the optimization for their hardware, would the ATi side get better optimization? No. If nVidia didn't come along would the developers be forced to do more optimization for both sides? No, they already met their optimization goals.

 

[Wrigleyvillain]

If nVidia didn't pay for the optimization for their hardware, the developers would certainly not be forced to optimize for both, that is a false belief, they instead would not optimize for either.

But they need to sell their game. With NV cash not in the equation is it not in their inherent best interest to optimize for both? That's not to say all would in every case but as long as we are making generalizations...

 

[newtekie1]

But they need to sell their game. With NV cash not in the equation is it not in their inherent best interest to optimize for both? That's not to say all would in every case but as long as we are making generalizations...

Obviously a certain amount of optimization will be done regardless, and is still done with ATi hardware or the games wouldn't play at all on ATi hardware. I did word that poorly. However, the assumption that nVidia is paying for all the optimization to go toward their hardware is false, I believe. Instead, they are paying for optimization beyond what the developer would normally do. This is evident by the fact that TWIMTBP games still running perfectly fine on ATi hardware.

And if you want some examples of what I mean look at a game like Modern Warfare. It is part of the TWIMTBP program, with really the only thing done seems to be optimizations for nVidia hardware, and we see this optimization with a GTX275 getting about 20FPS more than an ATi HD4890. However, the HD4890 is still getting 80FPS@1920x1200. So without nVidia stepping in the GTX275 might be right at the same 80FPS, but if nVidia didn't step in, I doubt the developers would have spent any more time and money getting better framerates for both. I'm guessing 80FPS@1920x1200 would have been good enough for them.

 

[Benetanegia]

but I don't see Nvidia's push towards proprietary technology as the solution.

For me it's the lesser evil. In fact, is the natural way for new things to take traction. First a feature has to exist, has to be created has to be established as a worthwile feature and before that happens there is no interest from anyone to create a standard for something that doesn't exist yet. I come back to the fact of how AMD says they support open standards but they are not creating those standards themselves, nor helping too much on their creation either.

Everything we have today has been a propietary standard at some point. Without going any further graphics API's started as propietary tech. IMO that's the only way that things can evolve in this capitalist world, because every company has its interests and none of them is to create something for free for everyone. In the early days of computing/gaming joining forces was a rewarding venture, because they only fighted against anonymity. There was a world full of potential customers that didn't have what these companies sold nor even something similar. The challenge was making those people buy that new thing they had created and if joining forces was the best to do they would simply do it. There was enough fish in the river.

Nowadays that doesn't happen anymore, anything sold is a replacement for another thing, it's a replacement of a competing product, so collaboration is not as rewarding as it was and that's why the only way for something to happen is making it yourself and that's what Nvidia is doing, because AMD will not create their own things and will not support Nvidia in the things they are making, as they already demostrated with PhysX. It's just a matter of finding a market for them, and although they might not be for eveyone, they surely are amazing features for some. Take Badaboom for example. A lot of people say it's useless, but for me that I record TV programs to watch them in my mp4 during my relatively long travels on the underground+train it was a blessing, I'd have my mp4 movies encoded in minutes instead of hours. Without this CUDA app I wouldn't be able to do that and I've been enjoying that app for almost 3 years if I'm not mistaken. Is it better to wait 3+ years to have it made with an open standard? Maybe for some, but not for me, 3 years is a long time. And same goes with PhysX. Forget about the fact that it has not been extensively used where it's been used (we know it's just because of lack of support from AMD), even the very little that it does in Mirror's Edge was worth for me. It was worth the effort of trying to have better physics in games and the only thing I had to pay was $0, the only requirement was going with a Nvidia card.

Things would be different if there were other alternatives, but there's none. there are promises, but again, promises were made 2-3 years ago and we have nothing yet. I can only think they will follow the path of Duke Nukem Forever.

And that's a similar point I've been making regarding Batman's antialiasing and the Just cause 2 features. If there were other similar features being made on OpenCL or DX compute now or soon or if most of UE3 games were coming with built in AA, I could see the situation as you do. But there's none. Those are the only examples of something like that, which is proof enough that developers don't want to make those things on their own. And I rather have those features than waiting forever. If more people had supported PhysX instead of bashing it, AMD would have been forced to make something, propose a standard, make their own thing, support PhysX... After that, we would already have an established feature, one worth making a standard off. Anything would have been better than what we have today, which is nothing.

 

[erocker]

ATi needs to keep working on Stream, Open GL/CL and all that. They also need to pick up CUDA. Fact of the matter is CUDA is there and it works. I like my cake and I like eating it too. I could care less about what happens behind closed doors at these companies or why it isn't happening. Don't let pride get in the way, do it for your customers, make it happen. We all know CUDA can run on ATi cards (see NGHQ from last year).

 

[HalfAHertz]

Benetanegia raised some valid points. Funk it, would linux for the desktop be where it is today if it wasn't for windows? Would the internet be as interactive as it is today if it wasn't for flash? In both cases it all began with a closed standard, that gained wide adoption. Once you standardize a platform or a solution and set the bar(give an example, a starting point for others to follow), it is much easier to develop an open, free alternative. So kudos to Nvidia for at least trying.

Windows>Linux
Flash>HTML5
Photoshop>Gimp
etc.

 

[cadaveca]

ATi needs to keep working on Stream, Open GL/CL and all that. They also need to pick up CUDA. Fact of the matter is CUDA is there and it works. I like my cake and I like eating it too. I could care less about what happens behind closed doors at these companies or why it isn't happening. Don't let pride get in the way, do it for your customers, make it happen. We all know CUDA can run on ATi cards (see NGHQ from last year).

ATi's problem isn't a lack of skill or any of that...they've laid off too many people, and just plain and simple, lack the nessecary manpower. They've literally striped away that side of the company during the merger, and become, 100%, a hardware company.


I think nVidia would do well from the publicity of actually giving CUDA to ATi 100% royalty free, if ATi could hire the software engineers to help support it.

and +1 to not caring about the reasoning behind it all...but understanding that side of it has me thinking we are SOL on that one.

 

[newtekie1]

ATi's problem isn't a lack of skill or any of that...they've laid off too many people, and just plain and simple, lack the nessecary manpower. They've literally striped away that side of the company during the merger, and become, 100%, a hardware company.


I think nVidia would do well from the publicity of actually giving CUDA to ATi 100% royalty free, if ATi could hire the software engineers to help support it.

and +1 to not caring about the reasoning behind it all...but understanding that side of it has me thinking we are SOL on that one.

I have to agree with nVidia giving ATi CUDA, and I think nVidia knows CUDA running on ATi hardware would give a huge boost to it. Of course ATi is more likely to stop CUDA running on their hardware than nVidia is, since it would completely kill Streams.

 

[Wile E]

where to start.... for one thing you are completely off base with this "trying something new" thing. Lets stick to the companies were actually talking about eh? and not abstract analogies to different markets and companies in different eras.

Nvidia goes to game developers and gives them money. The developer reciprocates by optimizing for nvidia hardware. Thats not "trying something new" that's paying to stop the devloper from optimizing for both brands.

IF nvidia were to give the game devs no money, is it in the developers best interest to optimize for only 1 hardware brand? hell no. They want to optimize for all hardware so people buy their game and don't complain about poor performance on either side.

Instead they are getting nvidia's money so they have incentive to only optimize for one hardware brand; this hurts progress.

No, they wouldn't optimize for either at all, and just let it happen on the driver level. Plus we wouldn't get any new features at all.

Sorry, I just don't agree that nV is making the gaming industry worse. If anything, their current practices should be kicking ATI's ass into gear offering useful development tools and proper support for the newer open standards, beating nV at it's own game.

I blame the current state of gaming affairs on ATI, for not adapting to the times, and offering better developer help and dev kits. If they would help devs more, we'd have none of this mess at all.

I love these GPGPU apps, and I really enjoy the little eye candy boost Physx gives in games that support it. Where is ATI's support or version of these? With the way things are going, I think my next card will be an nVidia card, because at least I get the new features, and not nothing at all. I hate some of their business practices, like disabling Physx in systems with an ATI card, but at least they are bringing new things to the table.