NVIDIA Publishes DirectX 12 Tips for Developers

[andybird123]

So you agree that in game it will be culled, so it won't affect performance

 

[Vertigo]

(and don't ask me what a 0.5 card looks like )

R9 Nano?

 

[Harlequin]

https://developer.nvidia.com/dx12-dos-and-donts#dx12

Do's
•Use hardware conservative raster for full-speed conservative rasterization
◦No need to use a GS to implement a ‘slow’ software base conservative rasterization
◦See https://developer.nvidia.com/content/dont-be-conservative-conservative-rasterization

•Make use of NvAPI (when available) to access other Maxwell features
◦Advanced Rasterization features
Bounding box rasterization mode for quad based geometry
New MSAA features like post depth coverage mask and overriding the coverage mask for routing of data to sub-samples
Programmable MSAA sample locations
◦Fast Geometry Shader features
Render to cube maps in one geometry pass without geometry amplifications
Render to multiple viewports without geometry amplifications
Use the fast pass-through geometry shader for techniques that need per-triangle data in the pixel shader
◦New interlocked operations
◦Enhanced blending ops
◦New texture filtering ops

Don’ts
•Don’t use Raster Order View (ROV) techniques pervasively
◦Guaranteeing order doesn’t come for free
◦Always compare with alternative approaches like advanced blending ops and atomics

AND

Check carefully if the use of a separate compute command queues really is advantageous

Even for compute tasks that can in theory run in parallel with graphics tasks, the actual scheduling details of the parallel work on the GPU may not generate the results you hope for
Be conscious of which asynchronous compute and graphics workloads can be scheduled together

so yes that's `DO use what we can do well and Don't use other stuff our competition can use well

 

[GoogalyMoogaly]

https://developer.nvidia.com/dx12-dos-and-donts#dx12



AND



so yes that's `DO use what we can do well and Don't use other stuff our competition can use well

And what did you expect them to write?

"Nvidia write guide to help developers get the best out of their hardware" hardly seems like a shocking headline.
I'd imagine some of the stuff in the guide probably makes sense for both sides. It's not even like they said don't use Async Compute, they just basically said don't assume it will be faster.

 

[Rroff]

wasn't the ocean under the map thing debunked by Crytek themselves as Cryengine uses occlusion culling, stick it in dev mode and you can see all the polys as it stops culling, but in the game most of that doesn't get rendered?

As I mentioned as well back when it was a big thing its a fairly common thing with game engines as well that a water plane that spans the map can be cheaper just to render always anyway than to pass it through an occlusion test (and as per humbug's post above its often not efficient to carve the water plane up to exactly match geometry that isn't part of the water).

The only real criticism is that they didn't use some kind of optimisation of the tessellation based on distance but that is more down to the developers lack of experience with an at the time fairly new technology in terms of actual usage than something that can really be levelled at nVidia.

 

[humbug]

So you agree that in game it will be culled, so it won't affect performance

I don't always disagree with you

But read Roff above, yet Crytek have managed to get it to work pretty well.

 

[Meladath]

I love how almost every single thread that bashes nvidia in any way has the very typical very notable nvidia fanboys (you know who you are) coming in and defending them and attacking AMD. Because somehow everything one company does must be related to its competition

NVIDIA has a long history of working with devs more than AMD. OF COURSE they will tell developers to work and do things in certain ways that it utilises their GPU's more than AMD's. Especially now considering the huge differences in architechture unlike the past (async shaders, 3.5 + .5GB, no double precision etc.) Why the hell wouldn't they? It would only be a good thing for them as a business.

Also, anyone actually reading this list will laugh at how pointless it is. This is purely publicity, the make the fanboys I mentioned earlier love NVIDIA and attack AMD even more. All of these "tips" are so incredibly obvious to anyone but the worst graphics programmer, its almost useless to read.

 

[Silent_Scone]

https://developer.nvidia.com/dx12-dos-and-donts#dx12



AND



so yes that's `DO use what we can do well and Don't use other stuff our competition can use well

This is correct, these workloads aren't going to be tax free. The results certainly aren't what certain users will be expecting given the amount of attention that has been drawn to it already. Just because it isn't as transparent as to how a particular game is handling the technique, does not mean it can't be detrimental to performance.

If the workload can be run conventionally without the need for simultaneous command queuing with low frame latency then there is no need to run this technique. In fact given the current hardware spectrum and capabilities of both vendors latest hardware compared to the lowest denominators, I doubt you'll see many cases where this is relevant in the near future. NVIDIA tend to put their money where their mouth is, AMD tend to put their mouth where they think they can generate money.

The amount of times I've been called out on things when posting here, not least of all with Mantle 2 years ago, you'd think people would realise that the realist is always right

 

[humbug]

NVIDIA tend to put their money where their mouth is, AMD tend to put their mouth where they think they can generate money.

I think the opposite is true, Nvidia are very inward while AMD are very outward. Nvidia don't do anything unless there it benefits them and them only.

 

[Harlequin]

This is correct, these workloads aren't going to be tax free. The results certainly aren't what certain users will be expecting given the amount of attention that has been drawn to it already. Just because it isn't as transparent as to how a particular game is handling the technique, does not mean it can't be detrimental to performance.

If the workload can be run conventionally without the need for simultaneous command queuing with low frame latency then there is no need to run this technique. In fact given the current hardware spectrum and capabilities of both vendors latest hardware compared to the lowest denominators, I doubt you'll see many cases where this is relevant in the near future. NVIDIA tend to put their money where their mouth is, AMD tend to put their mouth where they think they can generate money.

The amount of times I've been called out on things when posting here, not least of all with Mantle 2 years ago, you'd think people would realise that the realist is always right

the biggest money makers in the market are all AMD based - the consoles , therefore devs will code for them as they generate the biggest profits. so by default games will favour AMD hardware and add on gameworks , unless gameworks break the game

 

[Silent_Scone]

the biggest money makers in the market are all AMD based - the consoles , therefore devs will code for them as they generate the biggest profits. so by default games will favour AMD hardware and add on gameworks , unless gameworks break the game

I dont feel like repeating myself so i will just say this is incorrect and shows a lack of insight into any development process.

 

[~>Dg<~]

yes devs will just make games as they see best.that suits their product.

 

[Ayahuasca]

Harlequin why are you going on a rant about fanboys when the only one coming across as one in this thread on the last page or so is you? It seems this has made you hate nVidia more and has you defending AMD. Everyone else is having a rational discussion.

If you want to bash fanboys maybe take a look at yourself first.

 

[layte]

Even a developer who has traditionally been very close with AMD thinks the list is valid. http://www.overclock.net/t/1575062/...rectx-12-tips-for-developers/40#post_24453908

This Do and Don't list is pretty sound advice, particularly the DXGI swapchain information. I wish we had this info 6 months ago, as we had to figure out a lot of this information the hard way. The Do's list could almost be a description of how Nitrous is implemented.

 

[mmj_uk]

Even AMD's pet developer thinks the list is valid. http://www.overclock.net/t/1575062/...rectx-12-tips-for-developers/40#post_24453908

Impossible, we all know this is some nefarious plot for NVidia to shaft AMD.

 

[Insanties_birth]

Impossible, we all know this is some nefarious plot for NVidia to shaft AMD.

But also it's arguably impossible that AMD would weigh in neutrally and give clarification to the devs that this is sound advice. Should the Nvidia fanboys be heard then AMD are too lazy to lift a finger for developers and we must all chastise them for existing.

 

[LambChop]

Article: http://www.pcper.com/news/Graphics-Cards/NVIDIA-Publishes-DirectX-12-Tips-Developers

Direct link to Nvidia's do's and dont's: https://developer.nvidia.com/dx12-dos-and-donts

It's the standout thing for me. Nvidia are always assisting developers, building relationships and trust. Again, going that extra mile by producing this type of information. I can see in the future Nvidia again going that extra mile by offering their services to developers on site.

 

[humbug]

Even AMD's pet developer thinks the list is valid. http://www.overclock.net/t/1575062/...rectx-12-tips-for-developers/40#post_24453908

Impossible, we all know this is some nefarious plot for NVidia to shaft AMD.

I think its a problem with certain people on the outside looking in rather than anything else, one developer or anyone with anything positive about AMD is instantly marked as an "AMD Fanboi"

It seems that for a some people you can't be a "PC Hardware Fanboi" you have to be in one camp and hate the other, its pathetic and a pretty sad state of affairs. oh and just in case its not clear, no i'm stating an opinion with you

 

[LambChop]

I think its a problem with certain people on the outside looking in rather than anything else, one developer or anyone with anything positive about AMD is instantly marked as an "AMD Fanboi"

It seems that for a some people you can't be a "PC Hardware Fanboi" you have to be in one camp and hate the other, its pathetic and a pretty sad state of affairs.

I think they are called rivalries, and happens in every walk of life. From football, to business, to manufacturing. It's natural for me, healthy, good for banter and no getting away from it.

 

[humbug]

I think they are called rivalries, and happens in every walk of life. From football, to business, to manufacturing. It's natural for me, healthy, good for banter and no getting away from it.

That would be fine if it was just banter, but it really isn't. what it is is bizarre.