Nvidia’s ace in the hole against AMD: Maxwell is the first GPU with full DirectX 12 support

Ayahuasca · 31 May 2015 at 14:45

When I tested Mantle on a 295x2 it felt no smoother than it did on my 970 or Titan X, certainly never seen the game stutter like it did in humbugs video. AMD CPU's must really be potatoes and Nexus your CPU is over 5 years old now, we already know it helps in your situation.

layte · 31 May 2015 at 14:51

To be fair, it is perfectly understandable that Mantle does give the claimed boost on GCN hardware, as we know just how terribly inefficient AMD's DX11 driver is in comparison to the market leader, especially for those poor individuals suckered in by the Bulldozer hype. http://www.eurogamer.net/articles/digitalfoundry-2015-why-directx-12-is-a-gamechanger

tommybhoy · 31 May 2015 at 14:54

layte said:
To be fair, it is perfectly understandable that Mantle does give the claimed boost on GCN hardware, as we know just how terribly inefficient AMD's DX11 driver is in comparison to the market leader.

Market leader DX V Mantle.

layte · 31 May 2015 at 15:03

If that's your screen shot, you'll have to excuse me if I don't believe for one moment you made a fair and unbiased comparison.

Especially when the forums very own benchmark thread tells a different story: http://forums.overclockers.co.uk/showpost.php?p=27194303&postcount=3

A lack of Mantle numbers for some strange reason...

but those that are there show a 970 setup with a slower CPU beating a Xfire setup.

tommybhoy · 31 May 2015 at 15:05

Yep I'm that biased, I sold up and moved to Nvidia.

Doesn't change the fact Mantle>current DX.

humbug · 31 May 2015 at 15:37

layte said:
To be fair, it is perfectly understandable that Mantle does give the claimed boost on GCN hardware, as we know just how terribly inefficient AMD's DX11 driver is in comparison to the market leader, especially for those poor individuals suckered in by the Bulldozer hype. http://www.eurogamer.net/articles/digitalfoundry-2015-why-directx-12-is-a-gamechanger

Only in Nvidia sponsored games, all other games AMD's DX11 efficiency is on par with Nvidia, including BF4.

Both use Driver hacks to go around the DX11 Draw Call problem, or at least improve it a bit, in games where both use it there is no difference between them.

JediFragger · 31 May 2015 at 15:41

humbug said:
Only in Nvidia sponsored games, all other games AMD's DX11 efficiency is on par with Nvidia, including BF4

I think AMD have some issues with DX11 Multithreading in their drivers? I know they're looking at fixing this (due to an advert seen hiring such a person)

layte · 31 May 2015 at 15:43

humbug said:
Only in Nvidia sponsored games, all other games AMD's DX11 efficiency is on par with Nvidia, including BF4.

Nope. The DF article directly contradicts you, and I find them a far more reliable source.

humbug · 31 May 2015 at 15:44

No it doesn't ^^^^

JediFragger said:
I think AMD have some issues with DX11 Multithreading in their drivers? I know they're looking at fixing this (due to an advert seen hiring such a person)

No, AMD Drivers thread just as well as Nvidia's

JediFragger · 31 May 2015 at 15:44

Ok fine, no boost incoming. Doesn't really affect me does it

layte · 31 May 2015 at 15:47

humbug said:
No it doesn't ^^^^

Yes it does. We have gathered already that reading comprehension is a skill you lack, but if you read the article enough times the words should sink in.

humbug · 31 May 2015 at 15:49

It just doesn't ^^^

JediFragger said:
Ok fine, no boost incoming. Doesn't really affect me does it

The problem isn't how many threads can be used to render, the problem is how many Call threads DX11 limits to, which is one and a very slow one at that, whenever the CPU and GPU need to talk to eachother they can only do it through a peep hole, causing massive bottlenecks.

layte · 31 May 2015 at 15:55

humbug said:
It just doesn't ^^^

I'm not going to play your silly game where you bury your head in the sand and ignore facts and figures because they don't fit your narrative.

The charts and text of the article have direct proof of AMD's DX11 inefficiencies, that you choose to ignore it and carry on in ignorance, most likely surprises nobody.

humbug · 31 May 2015 at 15:56

layte said:
I'm not going to play your silly game where you bury your head in the sand and ignore facts and figures because they don't fit your narrative.

The charts and text of the article have direct proof of AMD's DX11 inefficiencies, that you choose to ignore it and carry on in ignorance, most likely surprises nobody.

humbug said:
Both use Driver hacks to go around the DX11 Draw Call problem, or at least improve it a bit, in games where both use it there is no difference between them.

layte · 31 May 2015 at 16:01

Read the article.

humbug · 31 May 2015 at 16:01

layte said:
Read the article.

I have.

humbug said:
Both use Driver hacks to go around the DX11 Draw Call problem, or at least improve it a bit, in games where both use it there is no difference between them.

layte · 31 May 2015 at 16:03

humbug said:
I have.

Understand the words on the screen then.

Crying about hacks is a smokescreen, unless you are saying that NV have put in hacks for every piece of software that was tested, in which case that's no longer a hack, but a general efficiency advantage.

humbug · 31 May 2015 at 16:06

layte said:
Understand the words on the screen then.

Crying about hacks is a smokescreen, unless you are saying that NV have put in hacks for every piece of software that was tested, in which case that's no longer a hack, but a general efficiency advantage.

How do you think AMD improved the performance of Project Cars when the problem was CPU bottlenecking, by a good chunk i might add, they implemented their hacks to improve call efficiency.

layte · 31 May 2015 at 16:13

Was the terrible AMD performance in PCars drawcall related? I've not seen that reported anywhere.

I'm sure Futuremark would have some very strong words for NV if they were cheating the system with hacks as you seem to be obsessed with. If not, surely AMD would optimise the FM DX11 DC test as well to counter the huge performance advantage shown to NV hardware. I can see what you are trying to do here, as it's something you pull every time you are found out or cannot counter independent facts. The bottom line is, testing has shown a clear and measurable advantage for NV hardware in DX11, something that you are unable to counter hence the slight of hands tricks you are pulling.

humbug · 31 May 2015 at 16:15

layte said:
Was the terrible AMD performance in PCars drawcall related? I've not seen that reported anywhere.

I'm sure Futuremark would have some very strong words for NV if they were cheating the system with hacks as you seem to be obsessed with. I can see what you are trying to do here, as it's something you pull every time you are found out or cannot counter independent facts. The bottom line is, testing has shown a clear and measurable advantage for NV hardware in DX11, something that you are unable to counter hence the slight of hands tricks you are pulling.

Isn't it obvious? if your GPU can't run at full speed its bottlenecked by the CPU or the API which sits between them.

Nvidia are not cheating, what makes you say that? what Nvidia or AMD do to thier drivers is part of the package.