glbenchmark 2.5. mali400 scores still broken

[stormkroe]

Just read the new article here at AT, and noticed once again that offscreen Mali400 scores are crazy in the game test. It scores 13 fps at native 800x480 resolution, but when cranked up to 1080 offscreen, they plummet to.... 13fps. This is in line with glbenchmark 2.1 scores increasing dramatically when going from live rendered 800x480 to offscreen 720p, yet scaling realistically in the galaxy note's similar-to-720p resolution when rendering on screen. I feel like I'm taking crazy pills!
/confused whining

 

[ArchAngel777]

I don't know a lot about these phone and tablet SoCs, but it appears to me that Tegra 3 is pretty 'meh' for the latest gen. Mybe I was expecting too much of it. But I really thought the Tegra 3 had the fastest GPU for the smart phones.

The tablet results are baffling to me. First, they only have 3 SoCs. Tegra 3 and 2, and then Omap or whatever.

Yet the phones have a lot more. Namely, the Exnys/Mali 400 and S4/Adreno 225, which both seem to perform better than Tegra 3 in the majority if the scenarios concerning the GPU

Now, I know that Tegra 3 games look better, but that is more because the are developed to be better, not because the hardware is truly superior, or maybe it is superior, just not clocked as high due to power consumption.

Thoughts?

 

[Puddle Jumper]

Now, I know that Tegra 3 games look better, but that is more because the are developed to be better, not because the hardware is truly superior, or maybe it is superior, just not clocked as high due to power consumption.

Thoughts?

I think you are exactly right about them being developed to be better, if you used Chainfire to run Tegra 2 exclusive games on Exynos they ran better than on Tegra 2 and still looked just as good.

 

[Bateluer]

The article notes that the Mali's architectural weakness causes the low triangle test scores.

 

[ITHURTSWHENIP]

lol this must be driving you crazy

I suggest you post this at ARMs own forum

http://forums.arm.com/index.php?/forum/10-mali-forums/

Maybe they have an answer

 

[shadow_k]

Tegra 3 games look better because Nividia works with the devs on some projects sonic 4 ep 2 is a good example

PowerVR SGX 543 is the most powerful graphics dont no why android OEM dont use it

 

[Puddle Jumper]

Tegra 3 games look better because Nividia works with the devs on some projects sonic 4 ep 2 is a good example

PowerVR SGX 543 is the most powerful graphics dont no why android OEM dont use it

The MP4 is would drive up SoC costs due to it's size while the MP2 isn't any faster than the 32nm Mali 400 used in the SGS3.

 

[stormkroe]

I just posted this in the comments, thought I'd copy it here to see what's bothering my chi.

Performance improvements have caused vsync to get in the way of glbenchmark 2.1 numbers on the Galaxy S2
Galaxy S2
Egypt HD 23.4fps
Egypt HD 720p offscreen 21.7fps
384k vs 921k pixels 240% workload, still 93% score

Galaxy Note
glbenchmark 2.1
Egypt High (1280x800) 48.1FPS
Egypt Offscreen (1280x720) 72.3FPS!!
1024k vs 921k pixels 90% workload, 150% score (these newer scores, while too high, are no longer double, my bad)

Galaxy Note
GLbenchmark 2.5
Egypt HD (1280x800) 16.5FPS
Egypt HD (1920x1080) 15.9FPS
1024k vs 2073.6k Pixels 202.5% workload, still 96.4% score

Galaxy S3
Egypt HD (1280x720) 23.9FPS
Egypt HD (1920x1080) 23.7FPS
921k vs 2073.6k Pixels 225% workload, still 99% score

The GS2 and GS3 continue the trend of scoring almost the same while more than doubling the workload (only difference being the doubled workload is offscreen)
I included the galaxy note because it's native screen resolution of 1280x800 is very, very close to the offscreen resolution in GLbenchmark 2.1, making apples to apples very straightforward and eliminating vertex shader architecture as the culprit.

I got these scores from glbenchmark's site to give you as much detail as I could on short notice, it confirms the trend that shows in the AT reviews.
Hopefully this helps

 

[ITHURTSWHENIP]

Tegra 3 games look better because Nividia works with the devs on some projects sonic 4 ep 2 is a good example

PowerVR SGX 543 is the most powerful graphics dont no why android OEM dont use it

Because while nerds like us gloat over benchmarks, in reality the difference between mobile GPUs is pretty small. The games are all v-synced so its not like having an SGX 543 is gonna give you 190 fps in a game. Not to mention almost all iOS games target SGX 535 because there are far more 3GS/iPhone 4 owners than there are iPhone 4S and iPad 2/3 owners.

With that said. OMAP 5 will use SGX 544 clocked at over 500 MHz wich should make it equal to iPad 3. Of course by then the rest of the gang will be bringing out new architectures

 

[ClowReed]

Just read the new article here at AT, and noticed once again that offscreen Mali400 scores are crazy in the game test. It scores 13 fps at native 800x480 resolution, but when cranked up to 1080 offscreen, they plummet to.... 13fps. This is in line with glbenchmark 2.1 scores increasing dramatically when going from live rendered 800x480 to offscreen 720p, yet scaling realistically in the galaxy note's similar-to-720p resolution when rendering on screen. I feel like I'm taking crazy pills!
/confused whining

Sorry for ressurrecting the topic, but I have the same problem with glbenchmark 2.5, but instead i'm on a GNote 2 which have a higher clocked mali 400mp (533mhz), even so I'm still caped at 13 fps on native resolution or the full hd one from the offscreen test. This doesn't make any sense.

Anyone have a concrete answer to that?

Thanks!

 

[stormkroe]

Finally a convert! All I can say for sure is that either the native score is too low, or the offscreen is too high. I choose the latter. Also, let me say, excellent first post

 

[BenSkywalker]

GLBenchmark exists to promote PowerVR's line of graphics chip. If you always keep that in mind, it will help you understand why the scores end up the way they do.

PowerVR's graphics chips aren't very good, they aren't terrible, but the only thing they are extremely strong at is GLBenchmark. There are *many* other benches that show this, but until we get people with some vague understanding of graphics benches handling reviews we aren't going to see it.

 

[lopri]

Tegra 3 is getting long in the tooth. Nexus 7 was just right with it. From this point on, it'll primarily drive budget end of Android/Windows RT tablets. (~$200)

 

[ClowReed]

I was thinking here with myself... Does anybody know if the GLBenchmark uses all cores of Exynos and Mali 400 at full potential? Cause this thing of the 13 fps obviously is a bottleneck somewhere, like happens with gpu/cpu on benchmarks for PCs. Or, like the friend above said, it's just a benchmark optimized for PowerVR gpus... Anyway, it's curious.

 

[fstime]

Why do people care so much about smartphone benchmarks?

I can understand PC benchmarking, but cell phones?

Battery life/Display quality/UI/Design >>>>>>>>>>>>>>>>>> Benchmark scores in 2012 where everything is pretty much dual and quad core.