AMD Explains Asynchronous Shaders on DirectX 12

[NeverBackDown]

I will apply the same level of skepticism regardless of who it is, even if I feel more 'loyalty' to one company over the other. I mean FFS I'm currently waiting on the furyX2 to come out so I should be happy about this turn of events. So excuse me while I hold judgement until I see a load more games with async, not just this one example.

While nVidia are now supposedly being shady as f*ck when it comes to Oxide now they've got the talent to make async computing work for them. Heck it actually does work, albeit in a buggy manner from what I've read thus far. nVidia are not amd, they'll either make it work or pay people off so that it works in the way that favors nVidia cards.

Early days innit. My next card is almost certainly going to be AMD for a number of reasons but this is not a factor for me yet.

So much attitude in the post for no reason. Must have woke up on the wrong side of the bed? Especially that "loyalty" bit. Interesting...

I never said it didn't work. I said it is limited and not as good as AMDs solution. They can improve it all they want in drivers, it's limited on a hardware level.. they would have to sacrifice something else to improve that specific performance like driver latency.

The Elephant in the room is despite all NVidia's whining it is their cards that are the fastest on the bench in DX12.

Your benches I found werent accurate. From what I have seen elsewhere all the scores you got were lower than everyone else's including mine. Heck my scores before the improved driver for the game were higher than yours at 1080p(no 4k for me)

 

[Kaapstad]

Your benches I found werent accurate. From what I have seen elsewhere all the scores you got were lower than everyone else's including mine. Heck my scores before the improved driver for the game were higher than yours at 1080p(no 4k for me)

We are using the highest settings possible hence the lower scores we are getting.

There are enough AMD users in that thread who find the results accurate as well.

Why don't you use the settings in that thread and post what you get ?

 

[Dicehunter]

For proper AMD results we'll have to wait until we can have unlocked voltage as comparing a 980 Ti with unlocked voltage at 1500MHZ to a Fury X at 1100MHZ is a little unfair IMO.

 

[Kaapstad]

For proper AMD results we'll have to wait until we can have unlocked voltage as comparing a 980 Ti with unlocked voltage at 1500MHZ to a Fury X at 1100MHZ is a little unfair IMO.

It is very unfair but even with voltage control the Fury X won't be able to make up the difference on this bench.

I also consider this bench flawed along with a lot of other people so this is another reason people should take the results with a pinch of salt.

We really need more DX12 games and benches to test on before forming conclusions.

 

[Kushiro]

Yeah, don't see the Furys getting clocked much higher than they already are, perhaps around 1300 without going nuts with the volts (that is of course assuming the VRM cooling can handle it)

 

[NeverBackDown]

It is very unfair but even with voltage control the Fury X won't be able to make up the difference on this bench.

I also consider this bench flawed along with a lot of other people so this is another reason people should take the results with a pinch of salt.

We really need more DX12 games and benches to test on before forming conclusions.

How is it flawed?

 

[Kaapstad]

How is it flawed?

DX11 performance on AMD cards is very poor and I don't think it is down to the drivers or AMD.

The bench has trouble holding it's settings.

DX11 performance is better on NV cards than DX12.

It does not use multi core (5960X) CPUs that well on DX12.

Visually there is hardly any difference between DX11 and DX12 so what is the point.

I can probably think of a few more if I put my mind to it.

 

[NeverBackDown]

DX11 performance on AMD cards is very poor and I don't think it is down to the drivers or AMD.

The bench has trouble holding it's settings.

DX11 performance is better on NV cards than DX12.

It does not use multi core (5960X) CPUs that well on DX12.

Visually there is hardly any difference between DX11 and DX12 so what is the point.

I can probably think of a few more if I put my mind to it.

DX11 Drivers for AMD aren't nearly as good from a pure performance benefit compared to Nvidia. They don't have as good as CPU optimization as Nvidia do, so in a CPU based scenario of which the entire bench is, it makes sense.

Trouble? Could just be bad performance on cards
It's still in Alpha too btw. Can't expect perfect everything yet. Devs even said nothing is close to finished yet too so we'll have to see later on if it's performance limited or a bad optimzation on a setting or two currently

DX12 drivers for Nvidia aren't really working well. Sometimes they get a small boost, sometimes no boost, or lose some fps. Makes sense and is why Nvidia has come out and whined about it. They are limited in Asnyc Comp atm on a hardware level. Drivers will need to be another "wonder driver" to get it working well.

8 Core CPUs don't automatically make it faster than a 4 core. Skylake has better IPC anyways so if testing on that you will see a bigger benefit. And again, Alpha, may take some time to change.

Dx11 vs 12 doesn't mean any visual changes It's an API not a texture It's designed for performance boosts, of which it does give(once Nvidia get's the crap together about the drivers) while maintaining a massive amount more draw calls per scene. And as being an RTS at the core, it's not about the graphics anyways.

So I don't see any valid flaws yet dude. If anything the performance one is the only technical thing you could make a case for. Even then, it's Alpha, Nvidia drivers are crap atm for it, and AMD get's massive improvements which would indicate it's actually working correctly. From what I have seen the Devs say, Nvidia's current performance issues are not on them or the MSAA thing they whined about. It's just the drivers aren't really working well in DX12.

 

[Kaapstad]

DX11 Drivers for AMD aren't nearly as good from a pure performance benefit compared to Nvidia. They don't have as good as CPU optimization as Nvidia do, so in a CPU based scenario of which the entire bench is, it makes sense.

Trouble? Could just be bad performance on cards
It's still in Alpha too btw. Can't expect perfect everything yet. Devs even said nothing is close to finished yet too so we'll have to see later on if it's performance limited or a bad optimzation on a setting or two currently

DX12 drivers for Nvidia aren't really working well. Sometimes they get a small boost, sometimes no boost, or lose some fps. Makes sense and is why Nvidia has come out and whined about it. They are limited in Asnyc Comp atm on a hardware level. Drivers will need to be another "wonder driver" to get it working well.

8 Core CPUs don't automatically make it faster than a 4 core. Skylake has better IPC anyways so if testing on that you will see a bigger benefit. And again, Alpha, may take some time to change.

Dx11 vs 12 doesn't mean any visual changes It's an API not a texture It's designed for performance boosts, of which it does give(once Nvidia get's the crap together about the drivers) while maintaining a massive amount more draw calls per scene. And as being an RTS at the core, it's not about the graphics anyways.

So I don't see any valid flaws yet dude. If anything the performance one is the only technical thing you could make a case for. Even then, it's Alpha, Nvidia drivers are crap atm for it, and AMD get's massive improvements which would indicate it's actually working correctly. From what I have seen the Devs say, Nvidia's current performance issues are not on them or the MSAA thing they whined about. It's just the drivers aren't really working well in DX12.

You have just described some above lol.

As for AMDs DX11 drivers not being up to the job, I have heard that one a lot and don't believe it for a minute as AMDs DX11 drivers are fine in every other game.

And if the bench is not flawed why are NVidia's inferior DX12 cards walking away with it lol.

 

[NeverBackDown]

You have just described some above lol.

As for AMDs DX11 drivers not being up to the job, I have heard that one a lot and don't believe it for a minute as AMDs DX11 drivers are fine in every other game.

And if the bench is not flawed why are NVidia's inferior DX12 cards walking away with it lol.

I didn't describe anything. This as of now is the baseline. It's only going to get better. Probably more so for AMD, they seem to be working really close with each other as they should be. It's the first DX12 game, getting the hype and performance advantage now would be crucial for them financially.
Didn't say they weren't up to the job, but their cpu optimization is not as great as Nvidias, so in CPU intensive scenes it'll take a bigger performance hit. I've never had an issue with them, even with this slight performance disadvantage they still are great competitors

They aren't walking away. That thread is literally the only thing I know of so far saying Nvidia is faster. Heck even PCper did the same testing and got results differing from that thread. My results and many others I know of got scores similar to there's. don't know what happened but something must have been done differently by a massive amount. Game hasn't updated either except a small hotfix while back. I know they plan a large update this month. Might address some issues we have with our different results maybe

 

[barnsley]

Especially that "loyalty" bit. Interesting...

I feel a bit more 'loyal' towards an partially incompetent company over a mostly evil one with a PR team gone wild . Doesn't mean I'll furiously defend them in every post though lol.

As Kaap said, the fastest cards in that bench are still nVidia (and weirdly enough I trust a large group of individuals over the media when it comes to benches) so I doubt this will cause issues and it is only one game. If it is 100% actually a massive problem all that it'll mean is that in certain games AMD will be a slightly better choice. So nothing will have changed. As we all know AMD cards aren't bad at all. The general game playing public have a different perception. That is arguably the biggest issue for AMD.

 

[Kaapstad]

I feel a bit more 'loyal' towards an partially incompetent company over a mostly evil one with a PR team gone wild . Doesn't mean I'll furiously defend them in every post though lol.

As Kaap said, the fastest cards in that bench are still nVidia (and weirdly enough I trust a large group of individuals over the media when it comes to benches) so I doubt this will cause issues and it is only one game. If it is 100% actually a massive problem all that it'll mean is that in certain games AMD will be a slightly better choice. So nothing will have changed. As we all know AMD cards aren't bad at all. The general game playing public have a different perception. That is arguably the biggest issue for AMD.

If people look closely at my bench thread the real star of the show is the R9 290P.

It is not the fastest card but it is quite a bit faster than it's direct competitor the GTX 970. The 290P is also probably the best bang for buck on there and compares well to the Fury X.

 

[AlienALX]

I will wait for some proper DX12 stuff before I start making judgement calls. One thing is for sure, the Fury X is very, very close to the Titan X in that benchmark. So that's one positive thing at least.

However, what I want to see are real gains, like the ones we saw on AMD cards using Mantle. BF4 showed at least 6-7 FPS over DX11's minimums.

This 'benchmark' is incredibly early. I'm not going to pick on the fact that the settings don't stick properly I would more hazard a guess and say it may not use the full feature set of DX12.

 

[barnsley]

It is not the fastest card but it is quite a bit faster than it's direct competitor the GTX 970. The 290P is also probably the best bang for buck on there and compares well to the Fury X.

Hasn't the 290 always been the best bang for buck card you can buy since they chopped the price? Unless you get a bit moany about power draw that is..

 

[NeverBackDown]

I feel a bit more 'loyal' towards an partially incompetent company over a mostly evil one with a PR team gone wild . Doesn't mean I'll furiously defend them in every post though lol.

As Kaap said, the fastest cards in that bench are still nVidia (and weirdly enough I trust a large group of individuals over the media when it comes to benches) so I doubt this will cause issues and it is only one game. If it is 100% actually a massive problem all that it'll mean is that in certain games AMD will be a slightly better choice. So nothing will have changed. As we all know AMD cards aren't bad at all. The general game playing public have a different perception. That is arguably the biggest issue for AMD.

I'm not defending anyone. Stating facts really. AMD perform better in my experience and others I have seen on their forums. I don't doubt kaap, but he seems to be he outlier here from what I have seen. Don't know if he's using the latest beta driver like I am but it's off for sure.

I will wait for some proper DX12 stuff before I start making judgement calls. One thing is for sure, the Fury X is very, very close to the Titan X in that benchmark. So that's one positive thing at least.

However, what I want to see are real gains, like the ones we saw on AMD cards using Mantle. BF4 showed at least 6-7 FPS over DX11's minimums.

This 'benchmark' is incredibly early. I'm not going to pick on the fact that the settings don't stick properly I would more hazard a guess and say it may not use the full feature set of DX12.

Have you even looked at the performance increases for amd going from 11 to 12? It destroys what mantle ever did for an increase. It's a massive jump. AFAIK from what he Devs said, they can be using all of DX12 but not all cards support it. So they have to adjust things. They are capable of using it, but not everything supports it. So in reality it's just not being used

 

[SpencerUk]

Naive? Not at all. They don't offer the same thing. One is hardware, one is software. The hardware solution is much better currently. Drivers for Nvidia are the problem. They need to get them to work very efficiently to avoid bottlenecks. That's all I was saying.
HBM has nothing to do with it tbh. They aren't bandwidth limited. They are CPU limited and then GPU limited.

How do they not offer the same thing?. Nvidia and AMD both deal in hardware, specifically graphics cards. It's not like one offers a visual heads up augmented display and one supplies milk cartons to Asda.

Or another example. Its not like AMD is the sole hardware vendor and Nvidia just makes the software for it. They both do hardware and they both do software. Bottom line is they are both graphics cards both offering different tech and values on how they go about it.

My perception is that you favour AMD which is fine, no problems with that, I personally have no affiliation to any brand. I'm going for whats best for money at the time I'm purchasing.

Hell I don't even know how we got to the subject of drivers. All I said is that AMD needs to make use of HBM quickly and market the monkey out of it because Nvidia's strong graphics card position will hold sway to mass market.

 

[NeverBackDown]

How do they not offer the same thing?. Nvidia and AMD both deal in hardware, specifically graphics cards. It's not like one offers a visual heads up augmented display and one supplies milk cartons to Asda.

Or another example. Its not like AMD is the sole hardware vendor and Nvidia just makes the software for it. They both do hardware and they both do software. Bottom line is they are both graphics cards both offering different tech and values on how they go about it.

My perception is that you favour AMD which is fine, no problems with that, I personally have no affiliation to any brand. I'm going for whats best for money at the time I'm purchasing.

Hell I don't even know how we got to the subject of drivers. All I said is that AMD needs to make use of HBM quickly and market the monkey out of it because Nvidia's strong graphics card position will hold sway to mass market.

They don't offer the same thing, as I said, One is using a hardware solution. One is using a software solution. Yeah they both deal in hardware.. and? Means nothing. AMD on a hardware level has support for it. Nvidia's hardware can support it through drivers.. which is software. The hardware implementation is better. That's all I am saying. It is not the same thing. If it was, both would have support on a hardware level. Nvidia is currently limited and there driver team is going to have to work wonders to get it to work correctly and not limit their performance.

I don't favor anyone. I favor the honest companies along with the best price/performance company as I want the most for my money. Currently Nvidia fit neither bill. So AMD it is for me. If the situation was different I would be on Nvidia's side. Heck I would even go Intel if they were in the market. I'm only debating with you because you think they are using the same thing when they are not. AMDs way is far superior atm. HBM again has nothing to do with the scheduling which is what this whole thread is about.