Nvidia/ATI Feature philosophy
- Thread starter trinibwoy
- Start date
Joe DeFuria said:
I am a developer myself (boring stuff not games
) and I would think that any lead time that a developer has to get familiar with a particular technology or feature the better the final product will be when the hardware is available to fully utilize it. I wasn't referring to those games that just hack on little effects here and there - if those were the poor products you were referring to. I'm more focused on the availability of hardware with new features at a reasonable time before products utilizing those features extensively will be available. In this way I think the API/IHV combination truly drives advances in game development so I can only see early adoption by an IHV in a positive light.
Joe DeFuria
Legend
trinibwoy said:
No, I was referring to the hardware.
In other words, a piece of hardware that's good for developers, does not make that hardware the better option for consumers.Agreed....from a developer perspective.
Bouncing Zabaglione Bros.
Legend
trinibwoy said:
The issue for consumers is the time for game development verses the lifespan of new hardware. So although it's great for developers to have SM3.0 to play with in the form of NV40, by the time developers release significant SM3.0 for any commercial game projects, the NV40 will have been superseded - maybe more than once if we are talking about the extended multi-year development of a triple-A title. It's also often the case that developers actually use the first generation of these products for programming work, knowing full well that adequate performance can only be gained for consumers by using the *next* generation product (for example, Unreal Engine 3)
In this scenario, consumers pay for (and are heavily marketed) towards new features they might not see any use of during the life of their expensive new card. Consumers might only see benefit in the form of encouraging developers (by providing a potential market) to support those features for the games and hardware arriving two years later.
In short, people tend to be dismissive of new features because for the most part, you often don't see any significant use of them for the consumers' benefit, as opposed to the developers' benefit, until the second or third generation products arrive.
Bouncing Zabaglione Bros. said:
EXACTLY!!! I can see why people don't appreciate paying for features that they will never use on their card but they also need to realize that without this feature creep better games will take longer to come out. So although they may be dismissive of it based on current software, without it the quality of the gaming experience won't improve nearly as fast. It's not like developers get some slightly advanced version of each consumer card to play with - that would be ideal.
One thing I haven't seen brought up yet here (forgive me if I'm wrong) is that we should keep in mind that these companies are currently in the design phases for their next generation parts after NV40 and R420. Perhaps in ATI's case, it wasn't worth the time to have their designers working on a PS3.0 part given the needs and goals of R500.
As an example, we've seen nVidia launch three highend cards in roughly a year time framewith some pretty drastic changes (memory controller, pipeline configurations, etc). Given how quickly they are going to replace the NV3X line (it's only been a year!), I tend to wonder how many resources nVidia has had to dedicate to get NV40 out so soon, compared to how many they have spent on NV45/NV50. In ATI's case, the R300/R350/R360 are pretty similar to each other, and the architecture has been around for a good 6-8 months longer than nVidias. If the R420 isn't that dramatic of a departure from R3XX (as some rumors have been claiming), and so long as the ATI engineers haven't been sitting on their asses for the last 18 months, they've had a lot of time to work on the next gen architectures.
Nite_Hawk
As an example, we've seen nVidia launch three highend cards in roughly a year time framewith some pretty drastic changes (memory controller, pipeline configurations, etc). Given how quickly they are going to replace the NV3X line (it's only been a year!), I tend to wonder how many resources nVidia has had to dedicate to get NV40 out so soon, compared to how many they have spent on NV45/NV50. In ATI's case, the R300/R350/R360 are pretty similar to each other, and the architecture has been around for a good 6-8 months longer than nVidias. If the R420 isn't that dramatic of a departure from R3XX (as some rumors have been claiming), and so long as the ATI engineers haven't been sitting on their asses for the last 18 months, they've had a lot of time to work on the next gen architectures.
Nite_Hawk
Bouncing Zabaglione Bros.
Legend
trinibwoy said:
In effect, early adopter consumers subsidise the use of these first generation features for developers, rarely getting to enjoy these new features until the next generation arrives.
trinibwoy said:
Using PS 3.0 as an example of this - For the developer it may make their life easyer (eg when coding the dev writes a single PS3.0 Shader with branching, rather than several PS 2.0 shaders). BUT if the output and performance is the same then PS 3.0 means nothing (from the consumer's perspective).
I think that this thread touches on a really important issue: a missmatch between new hardware and new software where the hardware often adds features that are not (really) used for a long time because the games developers have to catch up. We are now just starting to see the first few PS2.0 games coming out more than a year after the release of PS2.0 hardware (titles that have PS2.0 support hacked in late in the development cycle and where that support does little do not count as true PS2.0 games). Developers do tend to get the hardware as soon as it is available but that is basically the same time that it is commercially available. I think it has helped that the DX9 specs where put out by Microsoft prior to hardware being released and the hardware in large part built around those specs since this gives devs a good idea of what is coming. Early and precise specifications combined with decent emulation might help things out a bit but the whole field is just moving too fast to predict anything even a year ahead (imagine if full DX9 games came out at the launch of the NV30 with its horrible PS2.0 performance, it would have been a fiasco for Nvidia, well more than it already was anyway).
I think we will just have to deal with the reality that games will lag behind the hardware by about a generation for some time to come. A shame really because otherwise the IHVs might be able to sell their top range cards much cheaper (but in greater volume) right away and maybe just make two different models (high end and low end) rather than 3 or 4.
I think we will just have to deal with the reality that games will lag behind the hardware by about a generation for some time to come. A shame really because otherwise the IHVs might be able to sell their top range cards much cheaper (but in greater volume) right away and maybe just make two different models (high end and low end) rather than 3 or 4.
I think we've has this debate a million times over the years starting with the occurence of a card having features the competitor lacked. The conclusions you have arrived are so obvious that I fail to see the relevance of the post.
It would have been better to say "NV40 is a godsend for developers cause it allows them to play with and add features to their upcoming games sooner due to its advanced features".
The same could have been said for the Radeon9700(ps2.0) Original radeon(Rudimentary TCL), radeon 8500(ps1.4, truform?) and the geforce256(full dx7 T&L vs. no T&L on voodoo5), etc
On the consumer side this logic does not hold true.
EDIT: And to answer your question. ATI's or NV's feature philosophy cannot be described with a single generation of products like you intend to here. Look at the trend of innovative features each generation of architectures have brought to the table over the years and you can get a better picture than with a single generation of products.
It would have been better to say "NV40 is a godsend for developers cause it allows them to play with and add features to their upcoming games sooner due to its advanced features".
The same could have been said for the Radeon9700(ps2.0) Original radeon(Rudimentary TCL), radeon 8500(ps1.4, truform?) and the geforce256(full dx7 T&L vs. no T&L on voodoo5), etc
On the consumer side this logic does not hold true.
EDIT: And to answer your question. ATI's or NV's feature philosophy cannot be described with a single generation of products like you intend to here. Look at the trend of innovative features each generation of architectures have brought to the table over the years and you can get a better picture than with a single generation of products.
Bouncing Zabaglione Bros. said:
Usually, gamers don't just stop playing games and thus, will be able to enjoy those features faster with the next generation cards/games. So the early adopters at least get something for their money
Doesn't mean that i'm of the opinion that everybody should buy a NV4X though because in this case, both cards will have a very good featuresets and both will be very fast. And i'm guessing (doh) that the X-Box 2/next gen consoles will be the driving factor behind DX9+ support anyway. Tim Sweeney mentioned clearly that one of the targets for the UE 3 was the upcoming consoles. And all the new consoles should be more advanced then the cards available (or soon to be
) today.
Yes, but until HDTVs cost less than a car I'm not going to be giving up gaming on my monitor!Bjorn said:And all the new consoles should be more advanced then the cards available (or soon to be
Pretty much every IHV has been in the position of offering a hardware feature before it is even widely accepted as required.
Is there a general philosophy for a given IHV to try and do this?
I guess nVidia probably has consistantly offered features in a product that performance-wise are not pratical but are first to market.
This is one of the reasons that nVidia has for sometime been the gold standard for developers. However the noticable exception to this was R300. First to market with all the new features of Dx9 and at an acceptable performance level. In my 11 years of using 3D Hardware I have never seen such a combination of new features and performance.
Why would they not do the same with R420?
Cost / Benefit.
The major change with PS2.0 ->3.0 is a few more registers and a couple of static flow control instructions, 1 new arithmetic function and 1 texture function.
All the other major features (dynamic flow control etc) are introduced in ps 2.x
ATI may have decided that the new features were not important enough to warrant any potential performance issues a larger and more complex design may produce.
Is there a general philosophy for a given IHV to try and do this?
I guess nVidia probably has consistantly offered features in a product that performance-wise are not pratical but are first to market.
This is one of the reasons that nVidia has for sometime been the gold standard for developers. However the noticable exception to this was R300. First to market with all the new features of Dx9 and at an acceptable performance level. In my 11 years of using 3D Hardware I have never seen such a combination of new features and performance.
Why would they not do the same with R420?
Cost / Benefit.
The major change with PS2.0 ->3.0 is a few more registers and a couple of static flow control instructions, 1 new arithmetic function and 1 texture function.
All the other major features (dynamic flow control etc) are introduced in ps 2.x
ATI may have decided that the new features were not important enough to warrant any potential performance issues a larger and more complex design may produce.
gkar1 said:
Wrong. Developers getting access to new features early on will filter down to better games becoming availalable sooner. Plus for the consumer.
Please don't presume to know what I intended with my post. Your are quite wrong. Re-read my original post without your apparent shortsightedness. The original HardOCP articled touches on current affairs but my question was "intended" to get broad views on the relationships between game development and available hardware.
Joe DeFuria
Legend
trinibwoy said:
Agree and disagree.
New graphics features doesn't equate to better games. Devs don't have unlimited time and resources. Spending time on a feature that may only be usable to a certain percentage of the audience (because it's "new"), may actually detract from time spent on other, more useful things that would impact a wider audience. Especially if the feature they are working on doesn't actually come to mass market fruition (in terms of avilability and/or performance) by the time the game ships.
Having new features that are in practical purposes "only" useful for developers is a pro and con thing for consumers. It's not all good, and it's not all bad either. It's more like a "necessary evil".
Joe DeFuria
Legend
trinibwoy said:
No, there isn't really an alternative that's any better. That's why I said it's a "necessary evil."
I never criticize an IHV for introducing new features on hardware...even if they are not really usuable for consumers on that generation.
I do, however, tend to criticize:
1) Any heavy marketing and evangelization by the IHV of such features to consumers
2) Fans who argue that you should "buy" a card because of such features.