• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

Whoa! Much higher VRAM usage in Mantle!?

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

ninjacore

Member
Joined
Aug 13, 2006
Location
OH
I was doing a little comparison tonight between the DX11 and Mantle APIs with my Powercolor PCS+ 290 in Battlefield 4.

Multiplayer
map: Siege of Shanghai
resolution: 2560 x 1440
video settings: Ultra (custom), 2xAA, 80 FOV
other relevant hardware: 4770k, 8Gb 2400Mhz

2014-09-13 21:41:36 - bf4 (shanghai - DX11)
Frames: 53887 - Time: 815968ms - Avg: 66.041 - Min: 29 - Max: 113

2014-09-13 21:58:00 - bf4 (shanghai (2nd half of same map/server) - Mantle)
# of frames: 43701 - sum of frames: 598488.53ms - (total # / (sum/1000)): 73.02

You can't use Fraps with Mantle, so I used the in-game console commands (perfoverlay.framefilelogenable 1) to record frame times, then calculated the average FPS as above.

The average is higher with Mantle, but this is just half of a round on each API and since it's multiplayer, there's no way to make it even remotely consistent. I'm convinced that Mantle is better, but I have no idea by how much.

I was actually far more interested in what Mantle did to my GPU temp-wise and that's when I noticed a HUGE value difference in the GPU-z log output...

API|Avg FPS|Core Speed (Mhz)|Mem Speed (Mhz)|Temp Max (C)|Fan Speed Max (%)|Dynamic Mem Use (MB)|VRM1 Temp Max (C)|VRM2 Temp Max (C)
DX11|66|1040|1350|77|46|2411|93|64
Mantle|73|1040|1350|77|44| 3905 |96|65

Just a little less than 4GB of mem usage?! Is that accurate? I've never seen a value that high there with DX11.
 
Last edited:
Anyone else still using Mantle or did we all stop a week after it launched? ;)
 
More Mantle vs. DX11 results. This time with a Gigabyte Windforce 280X OC rev2.

Multiplayer
map: Flood Zone
resolution: 1920 x 1080
video settings: Ultra (custom), 2xAA, 80 FOV
other relevant hardware: 4790k, 8Gb 2133mhz

DX11
2014-09-15 22:54:53 - bf4
Frames: 18203 - Time: 248719ms - Avg: 73.187 - Min: 52 - Max: 111

Mantle
2014-09-15 22:54:53 - bf4
# of frames: 25731 - sum of frames: 281947.23ms - (total # / (sum/1000)): 91

Mantle
2014-09-15 22:54:53 - bf4
# of frames: 14456- sum of frames: 177830.55ms - (total # / (sum/1000)): 81


API|Avg FPS|Core Speed (Mhz)|Mem Speed (Mhz)|Temp Max (C)|Fan Speed Max (%)|Dynamic Mem Use (MB)
DX11|73|1100|1500|72|79|2181
Mantle|91|1100|1500|72|79| 3048
Mantle|81|1100|1500|72|79| 3048

Once again, I'm seeing significantly more vram usage with Mantle. The avg FPS bump is also pretty impressive here as well, though, same caveat as above: multiplayer, not strictly the same.
 
Last edited:
I think I heard about that somewhere ;)

I'm gonna keep playing with Mantle enabled (better FPS anyway) and see if the leak still exists. The 280X results show the vram basically maxed out, but the 290 results had a little room (and I don't think the usage was getting progressively larger after it hit 3905).
 
I can't give a good explanation but Anandtech backs up your findings. There was a snippet in their review of the R9 285 that Mantle was a lot more RAM hungry than DX11. I dug out the reference for you:

http://anandtech.com/show/8460/amd-radeon-r9-285-review/6

They didn't go into details but they said the "glaring exception" in their tests was Mantle, which was a lot more RAM hungry. So for whatever reasons, at least your findings are confirmed elsewhere. Apparently Mantle really wanted 3GB+ and was somewhat constrained on the 2GB 285. They didn't attribute it to a leak. They seemed to think it was just the nature of it.
 
But is it really fixed? You can't play BF4 and not update, right?

That said, I haven't had the running out of memory errors last I played it... and subsequent BSODs...but it has been a few weeks since I was on an AMD card and BF4 too...
 
Back