gdgt

tno, you are misreading what Anand is saying in that URL.

Anand is saying that the PowerVR SGX544 in the OMAP4470 is 2x as powerful per clock as the SGX540. The 2x as powerful comes from this.

The SGX540 in the OMAP4460 is clocked at 384 MHz while in the OMAP4430, the SGX540 is clocked at 304 MHz. This means the GN GPU is 25% more powerful than the Galaxy S.

For the OMAP4470 GPU, since it is 2x as powerful per clock and is also clocked 25% higher, this means it will 2.5x as powerful as the GPU in the Galaxy S.

So, the Galaxy Nexus GPU is 25% more powerful as the Galaxy S GPU, not twice as powerful. And yeah, it won't come close to Mali, and won't even be in the neighborhood of the SGX543MP2, and it's at a pixel disadvantage since it's driving a 1280x720 screen compared to the 960x640/540 screens in other phones. 3D games probably should be designed at 640x360 on the GN and pixel doubled.

Actually, prior to OMAP 4, the SGX 540 was run at about 200 MHz, hence the "almost twice as fast" comment in reference to the Galaxy S performance.
www.anandtech.com­/show­/4179­/ti­-omap4­-and­-lg­-optimu...
www.anandtech.com­/show­/3891­/samsung­-epic­-4g­-review...

So, to make it clearer, the Galaxy Nexus will be 25% faster than the Droid Bionic in GPU performance in resolution independent tests such as the GLBenchmark 2.1 - Pro - 720p off-screen.
Handily, this will be the resolution that the GN will be operating at so we can say it'll score about 45 FPS in this test over the Droid Bionic's 36 FPS. It should be noted that in this test you can see that the Hummingbird in the Droid Charge (with it's GPU clock set at about 200 MHz) scores 1/3 less than the Droid Bionic exactly, nailing down that clock speed difference perfectly.

As to the SGX 544, Brian is pointing out that it will be twice as fast as the SGX 540 when both are clocked at 304 MHz, as in the Droid Bionic. That would make it much more than twice as powerful as the Hummingbird's 540.

The SGX 540 at 384 MHz, then, remains the third best GPU available in a smartphone, behind the SGSII and iPhone. And while the delta from the iPhone to the Galaxy Nexus remains large, we should not minimize the delta between the Galaxy Nexus and the Galaxy S.

I like your reply. I didn't know or forgot that the SGX540 was clocked at 200 MHz in the Hummingbird.

The GLBenchmark 2.1 - Pro 720p off-screen has an unusual result between the LG Optimus 3D and the Moto Droid Bionic. Both have a 1 GHz OMAP4430 SoCs, but the Optimus 3D is at 43 fps while the Droid Bionic is at 37 fps.

There's a lot of software behind the scenes that can have a huge affect on even the same hardware (just take a look at some of our charts showing how performance improved between Android 3.0 and 3.1). LG's special sauce in some of the GPU tests didn't hold up across the board, so it's really hard to say what they could have optimized. Either way, OMAP 4 is definitely no slouch.

And if you like my reply, why don't you 'Like' my answer! Cheers.

Happy now. ;)

@tno

Check Anandtech. In GL benchmark comparison in the Droid RAZR review, Anand seems to think the GN SGX540 is clocked at 307 MHz while the Droid RAZR SGX540 is at 304 MHz.

The benchmark comparison shows essentially the same: the Droid RAZR and the GN are basically neck and neck in the OpenGL benchmarks, the SGSII Mali is about 30% to 50% more framerate, while the Apple A5 SGX543MP2 is 2x to 3x more.

So, the story hasn't ended yet.

Yeah, our information per TI was that the SGX540 in the OMAP4460 could be clocked to 384MHz. Unfortunately it's never so cut and dry, because the OEM can spec the SoC to their clock flavor of choice. In this case, it seems they opted to ensure equivalent performance to their prior product, while being gentle on the battery. This is a shame, but I'm glad Brian was able to figure it out. This also cleaves whatever benefit was left in using the OMAP4460 over the OMAP 4430. TI is generally quite forthcoming with details about it's SoC's and their OMAP 4 info page (www.ti.com­/general­/docs­/wtbu­/wtbuproductcontent.ts...) clearly shows that the difference between the two is in clock speed. OMAP 4430 was originally pegged to be clocked at up to 1.0GHz, while OMAP 4460 would go up to 1.5GHz. Since then we've seen several OMAP 4430 variants clocked at 1.2GHz, and this first OMAP4460 variant is clocked the same. So if the GN uses the same clockspeeds (CPU and GPU) as the OMAP 4430, then there really is no appreciable difference between the two, and all the performance benefits come from the software.
That's both good and bad news. The bad news was a somewhat known factor. While the GN will be a great phone, it owes its performance more to the software than the hardware. The good news, though, will come as a treat to many that recently bought Motorola phones. With Motorola using OMAP4 on nearly all their recent phones, anyone that bought one can expect performance at or near that of the GN when they get their upgrade. Unfortunately, as Motorola themselves explained, that upgrade could be held up by the carriers for sometime.