The AMD Mantle Thread

[krumme]

Johan told me to ignore websites that bench on an empty server.

Yes and who wouldnt agree with him. Thats why the above test is with a minimum of 60 players.

 

[sontin]

Johan told me to ignore websites that bench on an empty server.

Maybe good old Johan had provided the press with a "real" benchmark. Testing on a server with active players is useless.

 

[repi]

Johan told me to ignore websites that bench on an empty server.

At least if you are primarily playing on servers that actually have people on them

-- Johan

 

[desprado]

http://www.golem.de/news/amds-mantle-api-im-test-der-prozessor-katalysator-1402-104261-3.html

Direct3D 11.1 Mantle
Siege Of Shanghai (720p) 65,4 avg / 52 min 116,1 avg / 88 min
Siege Of Shanghai (1080p) 63,8 avg / 49 min 112,2 avg / 81 min
Siege Of Shanghai (1080p, 4x MSAA) 57,7 avg / 44 min 78,2 avg / 63 min
Siege Of Shanghai (2160p) 48,9 avg / 38 min 52,2 avg / 41 min

64 Spieler, Ultra Details (Core i7-3770K, 16 GB DDR3-1333, Radeon R9 290X @ 1.000/2.500 MHz, Windows 8.1 x64, Catalyst 14.1 Beta)

-------

Well i am sorry but the 50% is an understatement. You will easily get in a situation with 80% difference.

Arent you happy?

In which setup and resolution u are talking about?
in 1440p or 1600p there hardly 8 to 10% difference and Cf differnce is not even 7%.

Where John and AMD said that CF will have huge improvement which is false.

Coming to 4k there is barley a difference there.

 

[krumme]

Isnt that PR?

I usually play Multiplayer with players on. It tends to improve gameplay.

 

[Grooveriding]

Uh, these are multiplayer tests apparently:

So now we need to find the 2 people in the world out of 6 billion who paired a 7850k with a GTX780 or 290X.

That is a 4670K....

 

[rtsurfer]

Maybe good old Johan had provided the press with a "real" benchmark. Testing on a server with active players is useless.

I think you meant without active players here.

 

[ASM-coder]

What is this? You can put a GTX card on a Kavari APU?

Edit: Never mind. I suppose it would work, but it would defeat the entire premise this APU(HSA etc).
At least if you xfire with an AMD GPU, then the benefit would be incremental.

 

[OCGuy]

Either we are being trolled or this thread just became an official PR vehicle..

 

[Crap Daddy]

Multiplayer results are the wow factor:
(because of more CPU stress in these situations)


From: http://www.golem.de/news/amds-mantle-api-im-test-der-prozessor-katalysator-1402-104261-3.html

Core i7-3770K + 290x (tiny OC):

1920x1080 (no aa)
63,8 fps (dx11) vs 112,2 fps (mantle) (~75% perf. gain)




On smaller CPUs:

A10-7850k + 290:

DX11 = 31,9 fps
Mantle = 58,9 fps (~85% perf gain)

i3-4330 + 290:

dx11 = 43,3 fps
mantle = 79,7 fps (~84% perf gain)

Huge differnces in performance in multiplayer mode.

There goes out the window Kaveri APU. The i3 destroys it. Not that anybody sane woould pair these with a 290x.

 

[rtsurfer]

Either we are being trolled or this thread just became an official PR vehicle..

Or your mind has lost its ability comprehend facts.

Infraction issued for personal attack.
-- stahlhart

 

[sontin]

I think you meant without active players here.

No, because here you can reproduce the scene.
With active players is more a gambit than anything else.

A benchmark should always be reproducible. Otherwise you end with something like the Star Swarm test which is quite useless for benchmarks.

 

[blackened23]

Either we are being trolled or this thread just became an official PR vehicle..

It's been that way for a while as far as I can tell. But yeah. Independent verification of that account would be needed IMO.

 

[krumme]

In which setup and resolution u are talking about?

At 1080 where most people play and at mp situation that all people play. Sounds familiar?
With a fairly typical cpu power in a taxing situation.

Siege Of Shanghai (1080p) 63,8 avg / 49 min 112,2 avg / 81 min

Its at least a 50% improvement.

What you quote is in less taxing environment without any gamers. Hardly interesting for anyone.

Bang. Mantle have killed dx in bf4.

 

[repi]

Can we get an independent verification of identity here? I think it would actually, be called for in this instance. If that is the "real" repi, could be a good resource for asking questions. Just hope we're not being trolled by a fake account here. Which, I suspect, there are a lot of posters with multiple accounts here for not so good reasons.

There!
https://twitter.com/repi/status/429645531045199872

The mods can also confirm that Repi is who he says he is. Welcome, Johan.
-ViRGE

 

[sontin]

BTW:
nVidia cards are much faster in the MP pf BF4:

http://pclab.pl/art55953.html

A GTX780TI should be 50% faster than a 290X. I said it in the Star Swarm thread:
Mantle is more a fix for AMD's poor DX driver quality than a fix to DX at all.

 

[cytg111]

What is this? You can put a GTX card on a Kavari APU?

Dumb question perhaps, but why not?

 

[repi]

Though I'm not here for PR, just noticed that this was the most active forum discussing Mantle and a lot of good posts here, though also some with some misconceptions or errors on them as well that I can try and help correct.

 

[blackened23]

There!
https://twitter.com/repi/status/429645531045199872

Thanks for the clarification, it wasn't my intention to be a jerk with that comment. A lot of accounts here claim to be something they're not right now , just ensuring that we weren't being trolled. Thank you.

What brought you to AT? Did someone at AMD suggest you come here? Curious. Also, any word on crossfire being workable in BF4 with Mantle? A few press outlets mentioned anomalies.

 

[sontin]

BTW: Pclab.pl wrote that they used a 64-man server.
Google-translation:

From the point of view of players in Battlefield 4 is without a doubt the most important multi player mode. After a few months of release of the game rather no one playing in the single player mission. Here we expect, in accordance with the promises of AMD, the largest increases in productivity, according to the manufacturer because it is in terms of CPU load Mantle gives the greatest advantage. As is, in reality? Tests done on a map of Shanghai Siege game mode 64-man. The first one we prepared a test on the AMD A10-7850K (Kaveri):

 

[jaju123]

Hey Repi can you possibly elaborate on the reported problems with crossfire that AMD talk about in their 'Known Issues' post for BF4? I presume this will be ironed out in the future!

Thanks, great work btw!

 

[144HzGamer]

Though I'm not here for PR, just noticed that this was the most active forum discussing Mantle and a lot of good posts here, though also some with some misconceptions or errors on them as well that I can try and help correct.

Hello Johan, we've actually exchanged a couple of tweets in the past, quick question for you:

How much performance would you say us non 290X, 290, 260X users will be missing in this initial release? (I have a HD 7970 GHz Edition), and how soon will those optimizations be implemented for the rest of the GCN ecosystem?

Thanks.

 

[desprado]

Actually i will post both 3 results using R9 290 with DX,Mantle and GTX 780 at 1080p and 1440p than let see what happens.
It will be full 64 player MP.

I cloud do CF if it dont Stutter and Lag.

 

[rtsurfer]

No, because here you can reproduce the scene.
With active players is more a gambit than anything else.

A benchmark should always be reproducible. Otherwise you end with something like the Star Swarm test which is quite useless for benchmarks.

While I agree with your concerning about reproducing the same test for all the cards.
It is useless to test on an empty servers with no players as that will not be a real world scenerio & also as it doesnt put much strain on the CPU if you are just standing in an empty map.

But repeating the same test with multiple players is near impossible..

Maybe Dice should make some kind of multilplayers benchmark for BF4.

 

[blackened23]

While I agree with your concerning about reproducing the same test for all the cards.
It is useless to test on an empty servers with no players as that will not be a real world scenerio & also as it doesnt put much strain on the CPU if you are just standing in an empty map.

But repeating the same test with multiple players is near impossible..

Maybe Dice should make some kind of multilplayers benchmark for BF4.

Where do you get the idea that the server is empty? Pure speculation on your part? The wording of the page roughly translated seems to indicate that there were in fact 64 players on the server. Nowhere was it stated that it was an empty server.