AMD Mantle: A Graphics API Tested In Depth (TH)

[cheesyboy]

Tom's Hardware's in depth look at Mantle using a variety of CPUs and GPUs in Thief and BF4;
http://www.tomshardware.com/reviews/amd-mantle-performance-benchmark,3860.html

As you would expect, Thief sees a good boost for mantle, particularly at the higher GPU end where the CPU is the bottleneck. The 290x/FX-8 figures are particularly eye catching.

BF4 seems a complete fail, though. cards with <4GB vram suffering a performance hit, those with 4GB vram gain little.

 

[cheesyboy]

This'll be an interesting thread seeing at how much hate Tom's gets.

EDIT : There's some very dodgy looking results, My 290 sees gains across the board with a 4.4GHZ 4770K in Thief. Yet there's seems barely any? Yeah right.

They did look closer at that one, which does look odd;

Then, we swap the R9 290X into our Core i7-4770K-based system and observe dismal results with Mantle turned on.

The numbers are repeatable, and we don't have a concrete explanation to put forth. But we're reminded that AMD considers Mantle to be a technology in its beta stage.

It's worth noting that the i3 buries the i7 using mantle, which can't happen, really, so seems like some sort of bug.

 

[cheesyboy]

Nothing strange about that, The 780Ti is able to match or beat the 290X's Avrage FPS because its able to produce higher high FPS as long as it has a Powerful CPU, like the 4770K.
If not it falls way behind the 290X on the same lower powered CPU, the FX-8350.
What i find strange is that they did not test the 290X on the same 4770K, why not?

You're reading it wrong; the 4770k/290x/mantle is on there; it's just performed shockingly and that's what they're talking about

I am somewhat surprised at how much better the 780Ti is than the 290x under DX, though.

 

[cheesyboy]

The Mantle with AMD CPU results they came up with look right.

But its really strange that the performance drops significantly when they use Intel and Mantle, ........

It's only the i7 and only in that one chart. The i3 acts as you would expect. It's surely just a bug with that particular mantle build

 

[cheesyboy]

I am installing Thief at the moment, when done I will run a single 290X on my 4930k using 4 cores @3.5ghz the same as a stock 4770k.

In theory the 4770k should score a frame or two more than mine at the same clockspeed due to better efficiency. It will be interesting to see what happens.

I think we're all agreed that 4770k result is an anomaly. To be fair, TH said themselves it was an anomaly and double checked it.

I would be very surprised if it still occurred with the driver updates which have passed through since the tests.

 

[cheesyboy]

I'd say it's incompetence rather than them meaningfully making AMD look better than they are/Intel worse than they are.

That seems a bit harsh. They did spot it, retest it, and state in the article that it was both an anomaly and something which they couldn't come up with an explanation for, beyond the fact that Mantle is considered by AMD to be in a Beta phase (i.e. that it's probably a bug of some type).

Finally, with the High preset enabled, and a Radeon R9 290X/GeForce GTX 780 Ti installed, the results start getting strange. Nvidia's high-end gaming card averages about 80 FPS on a Core i7-4770K, and is matched by the Radeon R9 290X using Mantle and an FX-8350. Then, we swap the R9 290X into our Core i7-4770K-based system and observe dismal results with Mantle turned on.

The numbers are repeatable, and we don't have a concrete explanation to put forth. But we're reminded that AMD considers Mantle to be a technology in its beta stage. Overall, though, Mantle continues to help performance trend higher than DirectX

The 290x performance under DX looks to be below par all round, relative to the 780Ti. But these were old drivers TH was using.

 

[cheesyboy]

I am more troubled by the 290x performance in general. It actually doesn't make sense.

 

[cheesyboy]

I remember them doing an "early look" at their Thief results, and that was months ago. Can't find it now, though.

Edit:
doing a google search was more successful that the TH search bar;
http://www.tomshardware.co.uk/thief-mantle-benchmarks,review-32909.html
March 2014.