Mantle Benchmark Thread

[ICDP]

Just an FYI, SimBY posted it, but the Starswarm stress test has been updated, i just done 4 runs on it for comparison and posted them in the Starswarm thread

http://forums.overclockers.co.uk/showthread.php?t=18576886&page=8

Interestingly OC'ing the core on the GPU made some difference, OC'ing the mem on the GPU had little impact at first glance, however i think the card processed more batches and stuff so infact was probably working harder for a negligable fps gain, im impressed.

I did a complete uninstall of 13.12 using DDU (Display Driver Uninstaller) and reinstalled 14.1 to test the new update of Star Swarm.

Scenario RTS (the most consistent for me as it has a static camera)
DX = 11.4

Mantle pre update
29.27

Mantle post update
36.14

A very nice update from oxide there on Mantle. It is 224% faster than DX. Off course Star Swarm is taking extreme examples but none the less this is impressive for Mantle. I'm far more impressed by this demo than the farce that is BF4 with the fogging, spiking and crashing.

 

[GoogalyMoogaly]

If this is a Mantle benchmark thread, why so many DX11 benchmarks?
Shouldn't Mantle be judged on its own merits rather than compared to DirectX all the time?

I still stick with my idea that until we get a proper benchmark that can be re-run identically 100 times then a lot of these benchmarks are subject to randomness that's not connected with DX or Mantle. Too many variables to make anything mean too much at the minute.

Gonna get cramped in this thread once we have a decent number of proper benchmark tools for Mantle though.

 

[MouseMat2004]

Gets me a good increase too:

Mantle:
Minimum FPS: 4.78767
Maximum FPS: 263.852
Average FPS: 166.057

DX11
Minimum FPS: 25.2589
Maximum FPS: 262.467
Average FPS: 156.411

Same map/game, just played half on each.

FX8320 @ 4Ghz & 7850 2GB. Mostly on low @ 1080p except a few on high like Mesh etc. I prefer a smooth frame rate, although BF4 runs so well, I'm sure I could increase things a little!!

 

[LtMatt]

Gets me a good increase too:

Mantle:
Minimum FPS: 4.78767
Maximum FPS: 263.852
Average FPS: 166.057

DX11
Minimum FPS: 25.2589
Maximum FPS: 262.467
Average FPS: 156.411

Same map/game, just played half on each.

FX8320 @ 4Ghz & 7850 2GB. Mostly on low @ 1080p except a few on high like Mesh etc. I prefer a smooth frame rate, although BF4 runs so well, I'm sure I could increase things a little!!

Scores added. What clocks were you using on your gpu please?

 

[EL_JOCK]

Yeah, i'm 100% sure I did. But now you've made me doubt myself.

Ha ><. I'll retest if you think I really should.

edit. Just done another quick go, trying to follow as closely as I did with the last run and frames were almost identical. Who knows eh.

Fair play, just thought I would ask as the results seemed at odds a little with what i had been getting over the last few days, not trying to make you think you are going crazy or nothing, had enough of that for myself since doing these benches.

 

[MouseMat2004]

Scores added. What clocks were you using on your gpu please?

A modest 1040/1250

 

[EL_JOCK]

Think i maybe bringing this back from the dead, however now we have proper crossfire support with mantle in BF4, I thought why not re-run the bench test LtMatt had set up to see how mantle now does vs DX...

This run should be about as apples for apples as we can now make it, you might remember when we did this before, the DX runs had working crossfire & mantle could only have a single card running..

So here is the Mantle (14.2) with working crossfire score....

i5 3570k @4.3 BF4 ultra preset @1080p / crossfire sapphire tri 290p's 1000/1300

now compare that with my DX run. which did have crossfire working and we can now see mantle doing an 82% improvement over DX apples for apples on my system...

Also with the news today that DX & openGL could be going low level the future looks great for PC

 

[LtMatt]

Great news there Jock and impressive results. 82% is not be sniffed at and dx is likely a bit bottleneck for your xfire and i5 setup.

Have you noticed any bugs with Mantle and 14.2 so far? Any random big cpu spikes? What OS are you on?

 

[EL_JOCK]

Great news there Jock and impressive results. 82% is not be sniffed at and dx is likely a bit bottleneck for your xfire and i5 setup.

Have you noticed any bugs with Mantle and 14.2 so far? Any random big cpu spikes? What OS are you on?

running windows 7 at the mo, not sold on 8 just yet!

I did notice a very small spike which to be honest was only now and then & could just have been a load spike to be honest as there is much more time between them vs the big spikes on 14.1,

To compare the old spikes on (14.1) which where big and often, the new smaller spikes on (14.2) are much less a thing and give it another few drivers and think we could be gold....

On a side note have noticed that crossfire on DX9 is much better (as per the slides from the other thread), been playing mass effect the last week and since 14.2 is so much better its almost night and day better

 

[chief barker]

Awesome results

I just wish that my 7970s would play ball with Mantle.

 

[Warsam71]

Awesome results

I just wish that my 7970s would play ball with Mantle.

I deleted my previous post since I realized I asked you an obvious questions ("have you tried the latest 14.2 Beta driver) - sorry for any confusion I may have caused.

 

[BuzzLloydyear]

From the previous tests I have run, I made 2 changes. I have lowered my cpu clock from 4.7ghz to 4.4ghz to see if that caused a random crash I get from time to time. I also just switched from default bios, back to Matt's bios for 7990. I am surprised by the results.



The top one is Mantle, the 2nd one is Directx 11.1. This is running at 2560x1440 @ Ultra setting.

This is a higher score on Mantle than with previous bios, despite lowering my clock .3ghz. The directx 11.1 is a fair bit slower, I will have to do some more testing when I get chance to see if this directx score was an anomaly or not. I am happier with getting extra mantle frames out by using matts bios (which has never actually given me any real tangible benefits in the past).

 

[BuzzLloydyear]

I upped the CPU to 4.6ghz and ran it again and these are the results (Mantle is the top result again).

 

[LtMatt]

Not bad considering your card has not had performance optimizations yet.

 

[BuzzLloydyear]

Yeah, the issue with it maxing VRAM within minutes has definately improved too, I don't know if it happens now, but the numbers were ok in these runs. I was maxing my VRAM on low yesterday for some reason, so I don't know what's going on, but I won't complain.

Also I don't know if the loss of FPS in Directx 11.1 (around 7fps average) is driver related, or whether it's the bios, as I never seemed to benefit with your bios like others did in the past, but it's clear to see it's giving me a boost in Mantle.