The AMD Mantle Thread

[Abwx]

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.

Quite strange results when looking elsewhere..

http://www.hardwareluxx.de/index.ph...nisse-im-kampf-mantle-vs-directx.html?start=1

edit : i saw you posted said bench...

 

[SPBHM]

the A10, Pentium and i3 results from PClabs with the 290 are pretty amazing for mantle,
but I would like to see something more realistic, since the 260x seems to be the other card with it working well and low priced enough, 7850k/G3220/i3 + 260x with high settings and no AA should be good.

 

[Skurge]

Yeah, I was also about to point that out. The 290X was faster in SP while the GTX 780 was quite a bit faster in MP. I'd attribute that to the heavily multi threaded DX11 drivers that NV has, and AMD doesn't.

But it is funny, that in the XFX HardOCP 290X review people were called out for questioning the results which were quite simply, ridiculous - these were cards not overclocked and for some reason mysteriously outperformed reference by a mile. Yet here we are. Full circle. With some people questioning results from PCLab, a reputable website, just because they don't favor the side they prefer. Sounds pretty hypocritical to me, pot kettle black and such, but whatever. Just shows that there is no moral high ground here when it comes to benchmark results.

Again, I think the Mantle results at that website are in line with everything else we've seen. Big benefits for APUs, and the GTX 780 takes the lead in MP testing due to nvidia's DX11 multi threaded drivers. I'd point out that AMD does not have MT DX11 drivers, unless something changed very recently.

Yes, lets believe PC.labs and not HardOCP. Either both are relevent or niether are. They were in crossfire and they weren't throttling. No mystery there. We all know Hawaii is scales better than GK110 and is MUCH faster at higher resolutions.

As for PC.LABS, they don't even say if their server had people in it or how they tested. Golem.De does, HardOCP does. We shall see what the other more reputable sites have to say. I wouldn't dismiss any sites.

But if you are saying PC.LABS are more believable and HardOCP isn't cause their results don't match other sites, then your BIAS is showing.

 

[Carfax83]

You seemed to have missed this multiplayer review which completely contradicts your statement:

LOL, you post a review from LAST YEAR, shortly after the game launched?!! D:

Driver updates and patches have changed a lot of things since last year dude! 🙄

 

[cytg111]

Right...so DX11's multi-threading is supposed to be better than Mantle's now?

I have no knowledge of this MT dx nvda driver ... but it should be pretty simple, before and after numbers for the MT driver. Where can we dig em up? Anyway, if they'd given the massive increase in min fps's as reported for mantle, this forum section would have lightened up as a beacon seen from space.. So I doubt it.... I didnt see it.

 

[Carfax83]

As for PC.LABS, they don't even say if their server had people in it or how they tested.

Lies, they did post how many people were on the server.. Someone translated the statement on the previous page..

 

[NTMBK]

Don't want to go offtopic here, but nvidia have better solution from business point of view. They will stick some arm cores to the gpu that will offload driver overhead.

If that were the only reason that they were putting ARM cores on die, then it's a terrible reason. Instead of making the software layer more efficient, they waste power and die space budget putting a CPU on the die instead of more GPU shaders/texture units/cache/whatever. But that's not why the top end Maxwell is getting ARM cores; it's all about the HPC market. NVidia want to sell racks full of Tesla, with no Intel processor in sight. The ARM core is there to run a Linux stack for HPC.

 

[OCGuy]

Oh, and here is another benchmark from Hardwareluxx on a 64-man server - the same map Golem.de used:

http://www.hardwareluxx.de/index.ph...nisse-im-kampf-mantle-vs-directx.html?start=1

That's the problem with MP.

Wow these results are all over the place.....which is why using data that you cannot reproduce would get you flunked out of 5th grade Science. How it can be presented as legitimate by people who know better is beyond me....

64 players in the server?

64 constantly playing, or people coming and going?

Are they all spawned?

Are they causing the same environmental events to happen each time? (Building falling)

Is the connection to the server constant?



It is impossible to control the variables, which means it is impossible to verify anyone's results, which means we get these insanely wide data discrepancies.



The easy thing to do would be to release the driver and let people see for themselves.

 

[Sisyphean]

I find those 780Ti scores really hard to believe. We all know that the 780Ti beats a 290X, but it doesn't beat it with a ridiculous margin like that. So many reviews have shown that 780Ti can beat 290X with a decent margin in a few games but overall the cards have near parity in most games (within ~5%) and there are also a few games where the 290X beats even the 780Ti.

These benchmarks do look promising, however. If Mantle plays out as planned gamers should be able to pair high end GPU's with much cheaper CPUs. Thus, making the access point of high fidelity gaming at 1080p+ much cheaper and more available. When I read people saying that no one pairs a lower end CPU with a higher end GPU like its an argument for something makes my head hurt. Of course the price points of enjoying modern games at the highest settings as cheaply as possible is only good for the industry. A lot of gamers that don't have the cash, or aren't willing to spend, don't really experience a lot of the games at the level of detail that a small portion does. This hurts PC gaming in the eyes of the masses because they never really see the possibly significant difference between PC games and console games when it comes to fidelity, performance and replayability. If Mantle can drive down the cost of building a rig that can handle high/ultra settings in modern games by $1000 then it's a huge win for the developers.

 

[cytg111]

LOL, you post a review from LAST YEAR, shortly after the game launched?!! D:

Driver updates and patches have changed a lot of things since last year dude! 🙄

Then what is the review data as close to todays date as possible?

 

[ShintaiDK]

Then what is the review data as close to todays date as possible?

This:

 

[Skurge]

Lies, they did post how many people were on the server.. Someone translated the statement on the previous page..

Yes, someone did translate and said they played on 64 man server. Not if they had 64 players. Still their results match nobody as their are much slower than any other test on the internet.

 

[Gloomy]

Then what is the review data as close to todays date as possible?

http://hardocp.com/article/2014/01/20/gigabyte_gtx_780_ghz_edition_video_card_review/4

[H] revisits BF4 on every GPU review. These are using the Jan. 7th drivers.

 

[Slomo4shO]

LOL, you post a review from LAST YEAR, shortly after the game launched?!! D:

Driver updates and patches have changed a lot of things since last year dude! 🙄

Yes, apparently enough to swing performance in favor of the GTX 760 from being 10% slower than the R9 290X to now having a 40% performance lead :whiste:

 

[ShintaiDK]

http://hardocp.com/article/2014/01/20/gigabyte_gtx_780_ghz_edition_video_card_review/4

[H] revisits BF4 on every GPU review. These are using the Jan. 7th drivers.

Same result. Besides the slower older drivers. HardOCP looks like a mess tho.

 

[cytg111]

http://hardocp.com/article/2014/01/20/gigabyte_gtx_780_ghz_edition_video_card_review/4

[H] revisits BF4 on every GPU review. These are using the Jan. 7th drivers.

Nice, thx.

 

[sontin]

I dont understand the complains about pclab. They tested the MP after the release of BF4:

http://pclab.pl/art55318-5.html

And a GTX780 was 28% to 43% faster than the 290X.

 

[desprado]

Just to point out that they dont use the same settings.

R290 uses FXAA low. While the GTX780 uses 2xMSAA.

I swear i was about to post this.

 

[Grooveriding]

Then what is the review data as close to todays date as possible?

 

[Gloomy]

I swear i was about to post this.

Yes, there is an apples-to-apples down at the bottom...

 

[ShintaiDK]

I dont understand the complains about pclab. They tested the MP after the release of BF4:

http://pclab.pl/art55318-5.html

And a GTX780 was 28% to 43% faster than the 290X.

This is also a classic example that FX series does much better with nVidia cards. Simply due to the multi capable nVidia drievr vs AMD. I wodner if there is actually any improvement in mantle outside the multithreading part. But CPU usage should show that. The Oxide demo is rigged as already found out.

Another example of AMDs lack of multithreading:

 

[ASM-coder]

...my layman's understanding is that I should benefit a reasonable amount with my Phenom 965@3.8ghz and 280x.
Hope so, my board is AM3 only otherwise I would have popped an fx in there by now...

My understanding is yes, if fact maybe a decent amount. What are you currently seeing for FPS?

 

[Imouto]

I find those 780Ti scores really hard to believe. We all know that the 780Ti beats a 290X, but it doesn't beat it with a ridiculous margin like that. So many reviews have shown that 780Ti can beat 290X with a decent margin in a few games but overall the cards have near parity in most games (within ~5%) and there are also a few games where the 290X beats even the 780Ti.

The fun fact is that it isn't even a GTX 780 Ti, it's a GTX 780 Ghz.

 

[ShintaiDK]

I swear i was about to post this.

HardOCPs first enlarged result is almost always misleading. Its terrible.

 

[Grooveriding]

HardOCPs first enlarged result is almost always misleading. Its terrible.

Why ? Because you say so ? If so, that's not good enough.