Direct3D feature levels discussion

Yes this would be a concern for me. What would happen for games like Avatar and Forsaken for example that only include FSR frame generation which is tied to FSR upscaling. Would that mean no frame generation for Nvidia users?

That won’t be an issue because games won’t be “implementing FSR”. They will just invoke DirectSR apis and leave the implementation up to the IHV. This is a great thing and will avoid the current situation of developers choosing inferior implementations due to marketing partnerships.

The only potential downside is IHVs limiting which SKUs support upscaling e.g. no DLSS on Pascal.
 
The only potential downside is IHVs limiting which SKUs support upscaling e.g. no DLSS on Pascal.
It's hardly the only potential downside. AMD users will be blocked from using XeSS. Nvidia and Intel users will be blocked from using FSR FG.

What I struggle to think of are the advantages really.
 
You can make it work on any platform.
Who is that 'you', exactly? Streamline is currently bound to DirectX and Windows. NVidia didn't announce any plans for a multiplatform implementation and no-one else is capable of doing it. That's the big problem with open-source - in theory, anyone can do anything with the code, but in practice, very few have the resources to maintain a team of paid full-time developers, unless they raise public donations and private grants. Only code maintainers are capable of major reengineering efforts, and big corporations have their own priorities and urgent projects for the developers to spend their time on.
 
Last edited:
It's hardly the only potential downside. AMD users will be blocked from using XeSS. Nvidia and Intel users will be blocked from using FSR FG.

What I struggle to think of are the advantages really.

That’s literally what I said.

The advantages are pretty clear. No marketing drama for developers. Consistent features for consumers and greater adoption in games. No need for IHVs to twist developers arms. Upscaling will be just like “texture mapping”.
 
The advantages are pretty clear. No marketing drama for developers. Consistent features for consumers and greater adoption in games. No need for IHVs to twist developers arms. Upscaling will be just like “texture mapping”.
That's not advantages, nobody cares about what "dramas" developers go through to build a good product and the way this could work won't solve the "dramas" at all since if you'll block users from a different solution you will instead get these just like you do now - and yes, there is a different way which doesn't lead to that, the one which is being used right now.
 
That's not advantages, nobody cares about what "dramas" developers go through to build a good product and the way this could work won't solve the "dramas" at all since if you'll block users from a different solution you will instead get these just like you do now - and yes, there is a different way which doesn't lead to that, the one which is being used right now.

Having developers directly implement each IHVs proprietary upscaler is a terrible and unsustainable solution for obvious reasons. I don’t understand why you’re advocating for that. Yes less drama willl lead to higher adoption which is good for everyone. Who does DirectSR hurt exactly? Pascal users who want FSR? If Nvidia cares enough can provide their own non-tensor based upscaler for those folks.
 
That's not advantages, nobody cares about what "dramas" developers go through to build a good product and the way this could work won't solve the "dramas" at all since if you'll block users from a different solution you will instead get these just like you do now - and yes, there is a different way which doesn't lead to that, the one which is being used right now.

Less drama for developers means greater adoption of the tech means more end users will get to experience it ... if they want to. I'd say that's a clear win for the end consumer versus the tech not always being implemented by developers due to "drama" or difficulty or changing interfaces, etc.

A consistent interface that can then be used by developers for any IHV's hardware is a clear win both for developers (ease of implementation) and end users (greater adoption of the tech by developers means more games using it).

Unlike something like Streamline (where NV aren't invested in attempting to get other vendors to adopt it), DirectSR will have Microsoft incentivized to get all vendors to adopt it. Additionally, IHVs will be more incentivized to support a solution that isn't tied to one IHV. Microsoft will be actively working with IHVs to get it adopted. Are NV actively working with other IHVs to get Streamline implemented and adopted?

Regards,
SB
 
Having developers directly implement each IHVs proprietary upscaler is a terrible and unsustainable solution for obvious reasons.
Which are? All of them need the same from the engine and implementing one is about as complex as all at the same time. The differences which you may have there are required for each solution to work best, and losing that on a "common API" is a disadvantage.

I don’t understand why you’re advocating for that.
Because removing user options is not an advantage for anyone.

Yes less drama willl lead to higher adoption which is good for everyone.
As I've already said there will be more drama, not less.

Who does DirectSR hurt exactly? Pascal users who want FSR?
How about AMD users who want to use XeSS?

If Nvidia cares enough can provide their own non-tensor based upscaler for those folks.
That's a completely backwards way of viewing the whole issue. "Hey guys here's a solution which REMOVES what you have now, if you don't like it blame Nvidia!" Guess who they will blame instead.

Tbf we are discussing the idea of providing an upscaling solution within the driver which may not in fact be how the API will function (and I hope that it won't be). It may be a different implementation, a common interface doesn't mean that the solution sitting behind it needs to be in the driver.
 
Last edited:
Which are? All of them need the same from the engine and implementing one is about as complex as all at the same time. The differences which you may have there are required for each solution to work best, and losing that on a "common API" is a disadvantage.

What differences? As you said the implementations all take the same inputs. Hence the common API. It’s not rocket science.

"Hey guys here's a solution which REMOVES what you have now, if you don't like it blame Nvidia!"

What they have now is a mess of inconsistent feature support depending on which IHV’s marketing team does the best pitch. DirectSR opens the door for far greater adoption and a streamlined user experience.

Does Starfield supporting FSR but not DLSS at launch make any sense? DirectSR eliminates that kinda silliness.
 
Last edited:
What differences?
I don't know. A new version of DLSS may need something which the common API doesn't provide for example. Are you okay with being at the mercy of MS for adoption of new features? The same MS which doesn't have a performance oriented ML API in 2024.

What they have now is a mess of inconsistent feature support depending on which IHV’s marketing team does the best pitch.
I disagree. What we have now is game developers deciding which features they want to support, and their decision is rarely based on "which IHV’s marketing team does the best pitch". Games which are negatively influenced by that are very rare in fact.

DirectSR opens the door for far greater adoption and a streamlined user experience.
Again, depends on how it will function - we don't know that yet - but if it will be limiting user choice and forcing IHVs to bundle their solutions into the driver then it's as "streamlined" as forbidding users from using options because apparently it's "too hard to implement". Which in itself is complete b.s. when we are talking about these mostly plug and play upscalers.

Does Starfield supporting FSR but not DLSS at launch make any sense?
Sure. Makes perfect sense. How many games in 2023 can you name which did that?

DirectSR eliminates that kinda silliness.
It also eliminates the ability for AMD users to opt for XeSS there or for non-AMD users to use FSR3. This is way more silly than what you're describing.
 
It also eliminates the ability for AMD users to opt for XeSS there or for non-AMD users to use FSR3. This is way more silly than what you're describing.

AMD users also don’t get to run Intel’s aniso algorithm or BVH refitting code. It’s fine.

A new version of DLSS may need something which the common API doesn't provide for example. Are you okay with being at the mercy of MS for adoption of new features? The same MS which doesn't have a performance oriented ML API in 2024.

DirectSR doesn’t block devs from implementing future IHV specific solutions if they want that. But that’s all hypothetical. Fact is today all the upscalers take the same inputs.
 
I disagree. What we have now is game developers deciding which features they want to support, and their decision is rarely based on "which IHV’s marketing team does the best pitch". Games which are negatively influenced by that are very rare in fact.
Starfield released without DLSS with a lot of hate from players with nvidia cards. Palworld does only have dlss, locking out a lot of players of upscaling tech. Counter Strike 2 has only FSR 1 support with no way to mod it in because of anti-cheat. Islands of Insight was just released with only FSR 1 and it was so blurry that I turned it off and just incurred the fps hit. Nightingale was just released in EA with only FSR. This is a sad state we live in today. There are a ton of games/gamers that have been negatively influenced by this. Not to mention the less known games that will never get mods for other upscaling techs.

No it's not. It's a s/w solution which doesn't need Intel h/w to run on non-Intel platforms. It was made specifically for that FFS.
Intel has both a hw-solution and a sw-solution. The hw-solution is the one people talk about when they say hey like XeSS.

It also eliminates the ability for AMD users to opt for XeSS there or for non-AMD users to use FSR3. This is way more silly than what you're describing.
For XeSS it does not change a lot as only 1/10 games released have support for it. A lot of Nvidia gamers prefer dlss over fsr anyway. Gamers which are negatively influenced by not having access to an upscaling solution not native to their GPU that are very rare. DirectSR would be an overall net-win compared to the state we are in now.
 
Starfield released without DLSS with a lot of hate from players with nvidia cards. Palworld does only have dlss, locking out a lot of players of upscaling tech.
Starfield is a known thing. Palworld is early access.

Counter Strike 2 has only FSR 1 support with no way to mod it in because of anti-cheat. Islands of Insight was just released with only FSR 1 and it was so blurry that I turned it off and just incurred the fps hit.
No amount of common APIs will solve the issue of games not providing the inputs needed for TAAU-like solutions due to a lack of TAA or some other engine technical reason.

Nightingale was just released in EA with only FSR. This is a sad state we live in today.
This state won't be changed by any API. Early access games will still support only some features while IHV marketed games will still use IHV's proprietary tech.

Intel has both a hw-solution and a sw-solution. The hw-solution is the one people talk about when they say hey like XeSS.
No, it's the s/w one which people are talking about when they say that they prefer that over FSR2.

For XeSS it does not change a lot as only 1/10 games released have support for it.
It changes everything since this 1/10 number will remain as such if the API will work as you propose.

A lot of Nvidia gamers prefer dlss over fsr anyway.
This isn't any reason to remove options from players.

We are going in circles. Let's wait for details on how it will actually work.
 
This isn't any reason to remove options from players.

How does DirectSR remove options? If as you say developers should be willing to implement multiple versions of the same thing they can continue to do so to their heart’s content and choose from:

1. DirectSR based upscaling
2. FSR
3. XeSS
4. New DLSS that’s incompatible with DirectSR
5. Custom TAAU

There’s only upside here. DirectSR adds options, it doesn’t remove them.
 
How does DirectSR remove options? If as you say developers should be willing to implement multiple versions of the same thing they can continue to do so to their heart’s content and choose from:

1. DirectSR based upscaling
2. FSR
3. XeSS
4. New DLSS that’s incompatible with DirectSR
5. Custom TAAU

There’s only upside here. DirectSR adds options, it doesn’t remove them.
I think in terms of choice during gaming players may be locked to use one or another method. It was not discussed in the MS DirectSR links and think we will have to wait for the GTC 2024 demo/Q&A to see how it will function. Personally I believe players will be able to chose any/all methods developers enable that is compatible with their card. (like today)
 
Last edited by a moderator:
Back
Top