NVIDIA Game Works, good or bad?

[NThibieroz]

Which, as I've mentioned before, AMD still will not allow other IHVs to post optimized versions of as Mr. Huddy knows first-hand AMD folks please correct me if you've changed your minds on this issue... as you've said yourselves let's all collaborate on making the graphics industry better, right?

Here we go again. And then you wonder why I'm questioning your agenda...
There is no ambiguity about AMD SDK source code access and use by game developers. Such game developers are free to download, integrate and ship AMD effects from the SDK into their game. And we do not preclude them from optimizing or modifying code prior to shipping. That's all there needs to be said on this.

I'm sure TressFX is a neat effect, but in terms of scope, it's in the 'hey, cool research project', league. Not exactly the kind of stuff GW seems to have.

Sorry for the off-topic but if there is one thing NVIDIA is good at it is marketing, and that includes their own effects . The Hairworks authoring pipeline they have shown indeed seems fairly complete, but my (obviously biased although informed) opinion is that the effect in itself lacks in quality and performance. It really requires MSAA enabled to start looking OK, which I think is an unreasonable requirement to ask of (usually deferred) game engines. The use of isoline tessellation is also a performance hog (especially on AMD hardware where the effect can be up to 7x slower compared to equivalent NV parts). Finally the lack of a proper OIT solution can make it look like "spaguetti rendering" since without MSAA hair strands thickness becomes an issue (you may get away with it with very dark hair/fur though). I would encourage anyone to try out the fur in COD Ghosts and observe the quality of the effect when MSAA is disabled.

 

[silent_guy]

The Hairworks authoring pipeline they have shown indeed seems fairly complete, ...

I'm not arguing about which hair implementation is better. I've seen neither, but have no reason to doubt you on that.

My point is that AMD is only (publicly?) championing hair and nothing else. If Nvidia only had Hairworks, I don't think we'd be having this discussion. Because GW simply wouldn't exist.

You're saying Nvidia should make source available to AMD for everything. I think that's unreasonable for complex pieces of middleware like Optix etc. AMD doesn't have anything of that scale, which makes it very easy to scream about how it should be open.

 

[DavidGraham]

The Hairworks authoring pipeline they have shown indeed seems fairly complete, but my (obviously biased although informed) opinion is that the effect in itself lacks in quality and performance. It really requires MSAA enabled to start looking OK,

These effects are typically offered for high-end users with high end GPUs, which can handle these effects with combination of MSAA. And in fact these games usually require MSAA or TXAA to look good, even on simple geometry. FXAA alone is not enough.

The use of isoline tessellation is also a performance hog (especially on AMD hardware where the effect can be up to 7x slower compared to equivalent NV parts).

So now you are blaming NVIDIA for AMD's lack of Tessellation performance?

I would encourage anyone to try out the fur in COD Ghosts and observe the quality of the effect when MSAA is disabled.

In the meantime you also forgot to mention how TressFX implementation of hair in TombRaider was as buggy and unconvincing as it can get, hair was flickering, glitching and swaying radically, it was also a performance hog, even without MSAA, which the game actually don't have.

 

[Dave Baumann]

In the meantime you also forgot to mention how TressFX implementation of hair in TombRaider was as buggy and unconvincing as it can get, hair was flickering, glitching and swaying radically, it was also a performance hog, even without MSAA, which the game actually don't have.

All of those points improved rapidly with the updates, to the point that they were able to include the effect in the "Definitive Edition" on XBOX One and PS4.

 

[DavidGraham]

All of those points improved rapidly with the updates, to the point that they were able to include the effect in the "Definitive Edition" on XBOX One and PS4.

Nope, the only thing the updates did to the PC version is make the hair more stiff (especially on Lara's horse tail) to avoid radical behaviors. Also concerning the TressFX 2 implementation in the DE version of the game, I can't take your word for it since it was never released or tested on PC, which is the only relevant platform in this discussion.

 

[pMax]

Nope [...] I can't take your word for it since it was never released or tested on PC, which is the only relevant platform in this discussion.

That is very very likely a SV choice, not AMD one. The Software Vendor choose what to do, and why they have to spend money for integration&testing.

 

[Dave Baumann]

That is very very likely a SV choice, not AMD one. The Software Vendor choose what to do, and why they have to spend money for integration&testing.

And its perfectly illustrative of the points being made here - the developer has taken the implementation and further enhanced it for use on other platforms that not even related to AMD's PC business; its an open concept with developer control and up to the developers where they utilise it...

 

[SuperCow]

So now you are blaming NVIDIA for AMD's lack of Tessellation performance?

No. I'm arguing isoline tessellation is not efficient for hair rendering on any HW.

In the meantime you also forgot to mention how TressFX implementation of hair in TombRaider was as buggy and unconvincing as it can get, hair was flickering, glitching and swaying radically, it was also a performance hog, even without MSAA, which the game actually don't have.

Sorry to hear this was your experience. I disagree about "as buggy as unconvincing as it can get" though. While there were definitely a few issues the overall experience was great on most configurations able to handle the workload. We've made a huge number of improvements since the Tomb Raider implementation.

 

[Andrew Lauritzen]

There is no ambiguity about AMD SDK source code access and use by game developers. Such game developers are free to download, integrate and ship AMD effects from the SDK into their game. And we do not preclude them from optimizing or modifying code prior to shipping. That's all there needs to be said on this.

If you don't see how this policy is in direct contradiction to the rhetoric you've been spouting then you're in too deep... And yeah on this point I do have a clear agenda: to point out that you're not backing up your claims about being "open" and such with actual actions. People deserve to know the incongruity with your PR.

Intel is clearly the most open here, and as such I think it's fair to criticize the situations where AMD is not when you guys started this whole openness PR campaign... If you truly cared about people being able to get easy access to the best and most efficient TressFX code you'd put it up on github and allow people to contribute/branch/etc. directly. We do that.

 

[SuperCow]

Thank you for admitting your agenda. And congratulations to Intel for releasing their SDK source code on GitHub. We have a different mechanism for providing SDK source access to game developers.

 

[NThibieroz]

(Just realized I've been posting under my old login. The above posts from "SuperCow" are mine)

 

[silent_guy]

SDK source access to game developers.

... And to Nvidia, right? Otherwise, what are we talking about?

 

[firstminion]

"Usual" doesn't equal to "good".

Which, as I've mentioned before, AMD still will not allow other IHVs to post optimized versions of as Mr. Huddy knows first-hand

Of what?

I'm not going to disagree that NVIDIA is the most manipulative of the IHVs... which is partially why this GW situation simply doesn't rate high on the scale But part of that simply comes from the fact that they have the resources to do it. "Power corrupts" and all, while the rest of us just get to complain and be jealous of their reach.

Your position is poisonous to the industry and slightly naive.

If you don't see how this policy is in direct contradiction to the rhetoric you've been spouting then you're in too deep... And yeah on this point I do have a clear agenda: to point out that you're not backing up your claims about being "open" and such with actual actions. People deserve to know the incongruity with your PR.

Intel is clearly the most open here, and as such I think it's fair to criticize the situations where AMD is not when you guys started this whole openness PR campaign... If you truly cared about people being able to get easy access to the best and most efficient TressFX code you'd put it up on github and allow people to contribute/branch/etc. directly. We do that.

Please, don't start a discussion about what constitutes "real" Open, we've seen this before.

 

[silent_guy]

Please, don't start a discussion about what constitutes "real" Open, we've seen this before.

When 'open' (or access) is the core tenet of AMD's PR campaign, it's hard to see how this can't be on the table.

The whole argument of AMD has been: "we need access to the source code to optimize."

Has Nvidia been given access to TressFX?

 

[Andrew Lauritzen]

Your position is poisonous to the industry and slightly naive.

People who think any for-profit company wouldn't do exactly the same thing given the opportunity are the ones being naive.

Please, don't start a discussion about what constitutes "real" Open, we've seen this before.

That's what this entire discussion is about... NVIDIA will happily claim their stuff is "open" too. Everyone says it - it's a PR fluff word. If you want to criticize NVIDIA's definition I can absolutely criticize AMD's.

 

[DavidGraham]

No. I'm arguing isoline tessellation is not efficient for hair rendering on any HW.

A point that is invalidated first by the number of different configurations running COD Ghosts well, and second by the absence of an actual rival implementation for Fur, in this case NVIDIA stands alone, no one before or after them (yet) attempted to render fur as convincingly, and the results speak for themselves.

Sorry to hear this was your experience. I disagree about "as buggy as unconvincing as it can get" though. While there were definitely a few issues the overall experience was great on most configurations able to handle the workload.

Typical PR statement, nothing wrong with that by the way, it's just the same answer you would get from NVIDIA as well. A great experience will be provided none the less, the fact remains though, you can't criticize an implementation on the grounds of performance when your own implementation suffers the same, that applies to quality too.

We've made a huge number of improvements since the Tomb Raider implementation.

Which are no where to be found because your claimed improvements didn't find it's way into the hands of customers just yet, thus not yet tested or proven, we should talk more once this happens. (I am guessing Linchodm will feature TressFX 2 right? though I have access to the Alpha version of the game, and TressFX is not working at this stage even though it's featured in the graphics options.)

 

[Andrew Lauritzen]

Thank you for admitting your agenda.

No issues, I have never made my feelings on that point a secret. Thanks for publicly acknowledging the situation; the only reason I kept bringing it back up is that no one from AMD had publicly admitted that this was the case. I would still encourage you to clear up the licensing terms on your site though. It's not that difficult to just make a license that says people are not allowed to redistribute derived works in source form or whatever.

Anyways I won't post further on that point - I just thought it fair for people to know the situation there as context to what you guys have been saying about TressFX, etc.

Back to arguing about how evil GW is I guess. We need a numeric scale and a poll I think

 

[NThibieroz]

Which are no where to be found because your claimed improvements didn't find it's way into the hands of customers just yet, thus not yet tested or proven, we should talk more once this happens. (I am guessing Linchodm will feature TressFX 2 right? though I have access to the Alpha version of the game, and TressFX is not working at this stage even though it's featured in the graphics options.)

Fair enough, I was just letting you know about upcoming improvements as I think this information was relevant in the context of this discussion. There is currently a compiler issue causing the corruption you're seeing in Lichdom. A fix is coming.

 

[NThibieroz]

No issues. Thanks for publicly acknowledging the situation; the only reason I kept bringing it back up is that no one from AMD had publicly admitted that this was the case. I would still encourage you to clear up the licensing terms on your site though. It's not that difficult to just make a license that says people are not allowed to redistribute derived works in source form or whatever.

Are you putting words into my mouth again?

Btw:
WIP TressFX Unity implementation on GitHub blog:
http://kennux.net/wordpress/?p=13

WIP TressFX UE4 implementation on GitHub thread (not posted yet):
https://forums.unrealengine.com/showthread.php?6855-Hair-Rendering-Simulation-test

 

[Andrew Lauritzen]

Are you putting words into my mouth again?

Not intentionally! This is precisely why I've just been asking you to clarify the situation here and, incidentally, why license files exist