Will OpenGL ES 3.0 prove to be a short-lived standard?

[Ailuros]

Required voltages are depend on the architecture, ALU pipelining, VLSI shematic, fab's tech process characteristics and etc, etc. I don't think it's clever to extrapolate A7 characteristics on other SoCs. For example http://images.anandtech.com/doci/7622/Screen%20Shot%202014-01-06%20at%206.19.21%20AM.png here A7 consumes more than 550 Mhz Adreno 330 with 2x of ALU horsepower - http://gfxbench.com/result.jsp?benc...e&arch-unknown=true&arch-x86=true&base=device

1. One thing that isn't clever for sure is to provoke and then jump to conclusions, especially since I didn't specify any other platform. There are ULP SoC GPUs with much higher frequencies than 450MHz shipping in devices.
2. Most if not all IHV's will continue to optimize for synthetics like that and scores will continue to improve.
3. You might "think" the 330 at 550MHz has more ALU power, but it's not nessecarily the case. If you really want to interpret the ALU test results as you do, you might want to think why the 330 hasn't 2.6x (141 vs. 55fps medium ALU scores) the arithmetic throughput of an Adreno 320. From the GFXbench description for the ALU test:

Measures the pure shader compute performance of your device by using a complex fragment shader and rendering a single full-screen quad.

But since we're at it counting beans where what excels at most you might also want to check the "driver overhead offsreen" results. It basically measures the CPU driver overhead of the OGL driver with very frequent state changes.

 

[Ailuros]

Do we stil have anyone from Kishonti here reading in the fora?

http://www.glbenchmark.com/device.j...ndroid&api=gl&D=Lenovo+P780&testgroup=battery

While I'd love the possibility for a MT6589 to last roughly 13 hours under 3D I don't think it's even remotely close to being realistic.

 

[tangey]

... (continued) Funny also, that in all this time, no one has benchmarked a new iPad mini with its A7 (on either GfxBench 2.7 or 3.0); I'm curious to see just how it ranks against the Air and 5S to see the effects of the different power profile for its form factor vs clock speed vs slight variations in the tablet "version" of iOS7 versus the phone "version".

Its up there now with the latest drivers, pretty much identical to the other "i" devices, although the phone seems to have scored higher than both the tablets, but by such a small amount I guess it could be a statistical error.

 

[kishonti]

Do we stil have anyone from Kishonti here reading in the fora?

http://www.glbenchmark.com/device.j...ndroid&api=gl&D=Lenovo+P780&testgroup=battery

While I'd love the possibility for a MT6589 to last roughly 13 hours under 3D I don't think it's even remotely close to being realistic.

The Lenovo P780 is removed.

 

[Mariner]

Well, the P780 does have a huge battery (4000mAh) so a good battery life should certainly be achievable, but 13 hours running 3D benchmarks does seem a little bit too much to hope for!

 

[Ailuros]

The Lenovo P780 is removed.

It's always good to know you're around. Thank you

 

[Ailuros]

Well, the P780 does have a huge battery (4000mAh) so a good battery life should certainly be achievable, but 13 hours running 3D benchmarks does seem a little bit too much to hope for!

I haven't bothered running the test on mine since even a single TRex run at 2+fps is utterly painful (imagine that 30x times), but with half the battery size I get around 10 hours browsing time on it? In 3D and such a demanding 3D benchmark no chance in hell even with a 4000mAh battery IMHO

 

[Kaotik]

4.9 Offscreen in my HTC One.

Which test?
1080p Manhattan Offscreen 314 (5.1 FPS)
1080p T-Rex Offscreen 837 (15.0 FPS)

HTC One with Android 4.4 KitKat (Official, not rooted, no custom roms)

 

[Ailuros]

The Lenovo P780 is removed.

Two further cases where something is definitely wrong:

http://www.glbenchmark.com/device.j...timus+L9+(P760,+P765,+P768)&testgroup=overall

http://www.glbenchmark.com/device.jsp?benchmark=gfx30&os=Android&api=gl&D=Asus ME371MG

There's no chance in hell a SGX540 get's more than 2fps or so in GLB2.7.

 

[kishonti]

Two further cases where something is definitely wrong:

http://www.glbenchmark.com/device.j...timus+L9+(P760,+P765,+P768)&testgroup=overall

http://www.glbenchmark.com/device.jsp?benchmark=gfx30&os=Android&api=gl&D=Asus ME371MG

There's no chance in hell a SGX540 get's more than 2fps or so in GLB2.7.

The fake results are removed. We did not recognise, because we have a bug on the site: the median for the GPU results is not the median of the medians, but the top of the medians. We're working on it.

 

[Ailuros]

The fake results are removed. We did not recognise, because we have a bug on the site: the median for the GPU results is not the median of the medians, but the top of the medians. We're working on it.

Don't worry it could happen in the best family

 

[Lazy8s]

As always, Samsung delivers the most performance-tuned power profile of the common high-end SoC platform of the time, the S801 in this case, for their flagship:

http://gfxbench.com/result.jsp

Not enough to catch the G6430 but higher than other MSM8974AB and MSM8974AC 550 and 578 MHz Adreno 330s.

 

[Ailuros]

The fake results are removed. We did not recognise, because we have a bug on the site: the median for the GPU results is not the median of the medians, but the top of the medians. We're working on it.

Aren't those Adreno320 results a bit too high? http://www.glbenchmark.com/device.jsp?benchmark=gfx27&os=Android&api=gl&D=Pantech IM-A870 Vega Iron

 

[jedibeeftrix]

OpenGL ES 4.0 (with google extensions) too?

http://www.phoronix.com/scan.php?page=news_item&px=MTc0MjA