Async Compute is super hard & too much works for devs

[GoogalyMoogaly]

Where did devs say that or was this pulled from where the sun don't shine?

MS stated only the big boys(likes of Unreal/Crytek) will get full use of DX12 and the rest will simply stick to DX11.*



Same as GW's, takes extra time and a bung to use it

FC4/Primal is a perfect example, if GW's was that easy to implement, imagine it would have carried over to Primal, some argue the devs want to use it, it saves them time.



+1

You can shout from the rooftops all you want, but if you don't pay devs to include it then it's a lot of noise for nothing.

Not sure where GameWorks comes into this, but I agree it'll take time. Don't think anyone ever claimed GameWorks was a free performance boost for Nvidia though...

(5 to 10%), its really not a good idea to listen to a developer who also says "its too hard"

There is no reason why this technology should not yield a great deal more than that.
Some developers are getting more.....

And surely we want developers to explore new technologies? if we do; maybe not be so childish as to put a measure on the benefits for us, what are we saying? that developers should not bother modernising unless it yields a 20% performance boost for us?

Really? are we all just elite console gamers now?

Edit. misread that ^^^^ somewhat.

Inevitably yes, your right, however, some developers will modernise while others will cross their arms and say "burrrh.. its too hard"
I know whose games i will and will not buy.

Some developers have got more than a 5-10% performance gain just through ASync compute or more than 5-10% using DX12 in general?

 

[tommybhoy]

Not sure where GameWorks comes into this, but I agree it'll take time. Don't think anyone ever claimed GameWorks was a free performance boost for Nvidia though...

Wasn't comparing performance boosts(GW's is an anti performance boost) just comparing time and cost as you said it wasn't free, I was agreeing with you, it'll take a bung like GW's does to implement(as it only works on AMD perhaps until Volta).

 

[Orangey]

When low-level APIs first reared their head, some of us said all along it wouldn't be the silver bullet you all were thinking. We said it would enable the big-money guys to push their engines to new limits but for the rest too much hassle.

Few years and umpteen threads later...

 

[ICDP]

When low-level APIs first reared their head, some of us said all along it wouldn't be the silver bullet you all were thinking. We said it would enable the big-money guys to push their engines to new limits but for the rest too much hassle.

Few years and umpteen threads later...

What a load of twaddle quite frankly. I don't recall everyone but you and an enlightened few saying low level APIs were going to be a silver bullet that would make all games awesome. DX11 was stagnating and holding game devs back due to the fact they could not utilise multi core CPUs anywhere near their full potential.

Mantle (Vulkan) then DX12 changes that and gives any devs who put in the time to learn them much better tools to get the best performance from their games. Don't just look at how a high end CPU/GPU combo works in DX12, look at how lower end CPUs fare in comparison to DX11.

The article the OP links to seems to miss the point of low level APIs entirely. 5%-10% performance increase on Async Compute at high level such as Fiji does not look that spectacular but what it does is open up a larger potential customer base as from those with lower end CPU/GPU combos. A 50%-60% performance increase on a low end Intel or AMD CPU with a lower/mid range DX12 GPU is not an insignificant increase and transforms a game from unplayable to playable. This is where the real benefit to games developers comes from. More potential buyers.

http://www.pcper.com/image/view/60354?return=node/63600

 

[bru]

When low-level APIs first reared their head, some of us said all along it wouldn't be the silver bullet you all were thinking. We said it would enable the big-money guys to push their engines to new limits but for the rest too much hassle.

Few years and umpteen threads later...

What a load of twaddle quite frankly. I don't recall everyone but you and an enlightened few saying low level APIs were going to be a silver bullet that would make all games awesome. DX11 was stagnating and holding game devs back due to the fact they could not utilise multi core CPUs anywhere near their full potential.

Mantle (Vulkan) then DX12 changes that and gives any devs who put in the time to learn them much better tools to get the best performance from their games. Don't just look at how a high end CPU/GPU combo works in DX12, look at how lower end CPUs fare in comparison to DX11.

The article the OP links to seems to miss the point of low level APIs entirely. 5%-10% performance increase on Async Compute at high level such as Fiji does not look that spectacular but what it does is open up a larger potential customer base as from those with lower end CPU/GPU combos. A 50%-60% performance increase on a low end Intel or AMD CPU with a lower/mid range DX12 GPU is not an insignificant increase and transforms a game from unplayable to playable. This is where the real benefit to games developers comes from. More potential buyers.

http://www.pcper.com/image/view/60354?return=node/63600

I've got to hand it to you there ICDP, I really love the way you have told Orangey that his post is a load of twaddle and then basically said the same thing as he did anyway.
Yes I now the part in red is reversed, but if some were saying it was going to be the silver bullet, then he is correct in saying that some must have been saying it wouldn't be.

 

[CAT-THE-FIFTH]

I've got to hand it to you there ICDP, I really love the way you have told Orangey that his post is a load of twaddle and then basically said the same thing as he did anyway.
Yes I now the part in red is reversed, but if some were saying it was going to be the silver bullet, then he is correct in saying that some must have been saying it wouldn't be.

He is not saying the same thing though - if any of you bothered to look at the benchmarks for the DX12 games on lower end CPUs the improvement is decent. That is a good thing for PC gaming as older rigs can but upgraded with a new card instead of a whole rig which makes consoles look viable in cost. The Ashes and TR results in DX12 have shown this.

All the anti-Mantle people said the same even in the face of people running older Core 2quada in MP matches in n BF4 and only focussed on SP benchmarks.

That has got to be good for PC gaming.

 

[AlamoX]

there is Devs who always enjoys discovering and learning new stuff, coming up with new techs etc.
and there is script kiddies, who learn from programming for dummies, and anything new gives them a headache

 

[ICDP]

I've got to hand it to you there ICDP, I really love the way you have told Orangey that his post is a load of twaddle and then basically said the same thing as he did anyway.
Yes I now the part in red is reversed, but if some were saying it was going to be the silver bullet, then he is correct in saying that some must have been saying it wouldn't be.

Lol, I can see where it looks like that at first glance but read Orangey's quote again please.

"When low-level APIs first reared their head, some of us said all along it wouldn't be the silver bullet you all were thinking."

He qualifies his quote as him and a few others being the voice of reason in a sea of idiocy "you were all thinking" is clearly implying EVERYONE ELSE but him a few others. He comes across as being arrogant but he is using historical revisionism to suit his narrative because EVERYONE ELSE did not say they believed low level APIs would be a magic bullet. Had he simply stated a lot of people incorrectly expected DX12 to be some some sort of silver bullet that magically gave us large performance increases across the board, then I would have agreed.

My sentence in its entirety.
I don't recall everyone but you and an enlightened few saying low level APIs were going to be a silver bullet that would make all games awesome

That sentence is calling Orangey's quote BS, not agreeing with it. Orangey and a select few were not the voice of reason in a sea of idiocy as he implies.

 

[ICDP]

He is not saying the same thing though - if any of you bothered to look at the benchmarks for the DX12 games on lower end CPUs the improvement is decent. That is a good thing for PC gaming as older rigs can but upgraded with a new card instead of a whole rig which makes consoles look viable in cost. The Ashes and TR results in DX12 have shown this.

All the anti-Mantle people said the same even in the face of people running older Core 2quada in MP matches in n BF4 and only focussed on SP benchmarks.

That has got to be good for PC gaming.

Exactly, the more people buy in to cheap PC gaming the better. If devs have a larger customer base they will be far more inclined to develop PC games.

 

[Rroff]

Exactly, the more people buy in to cheap PC gaming the better. If devs have a larger customer base they will be far more inclined to develop PC games.

Problem is in the long run devs aren't going to leave potential extra performance budget unutilised - it will either enable lazy programming or more advanced features.

 

[CAT-THE-FIFTH]

Problem is in the long run devs aren't going to leave potential extra performance budget unutilised - it will either enable lazy programming or more advanced features.

Some of the European sites have show decent performance improvements using DX12 on lower end CPUs already which is a very good sign.

Also,Watchdogs 2 looks to be AMD sponsored and DX12 too. I wonder if that will use async compute??

 

[Seanspeed]

When low-level APIs first reared their head, some of us said all along it wouldn't be the silver bullet you all were thinking. We said it would enable the big-money guys to push their engines to new limits but for the rest too much hassle.

Few years and umpteen threads later...

There will still be smaller developer teams who will be excited about taking advantage of DX12/Vulkan. There's a lot of talented programmers out there not working at giant AAA studios for various reasons who will have the knowhow and/or determination to tackle the extra work. It can also come with advantages in terms of tracking down bugs and performance issues they couldn't before because of the driver and who would get less attention from Nvidia/AMD in terms of communication to track them down.

And when engine-level support for the likes of Unity and UE4 comes in, it's going to make it even easier for the 'lesser' studios to mess with it.

It's not going to be a silver bullet, no. DX11 will stick around for a while. And there's still an issue of disparity between Nvidia and AMD performance potentially being bigger under this. But I do think it will start to change the landscape as times goes on. Hopefully for the better, overall. There's a lot of neat things that will be possible more than just 'better performance'.