OMAP4 & SGX540

[Ailuros]

Any word on clocks for the SGX544 implementation in OMAP4470?

Nope. But I wouldn't be surprised if the results is damn close to the A5 GPU performance.

Are there performance-specific differences between SGX543 and 544, or is it just DX9L3?

Latter.

If they're using vec4 shaders and the same 384MHz, the OMAP4470 should be able to match the A5 in 3D performance, right?

Nope. I've no idea what the A5 GPU block is clocked at, but I'm still betting on around 250MHz frequency. In theory to reach the same performance from a single SGX544 you'd need (if the 250MHz frequency is correct) 500MHz. It's one core vs. two and that's not the entire story either, since Apple dedicated a shitload of system level cache to its MP2 and didn't scare away from dedicating die estate for it. If TI equips the GPU with equal general cache then the former estimate should be give or take correct.

Is the 4470 exclusive to tablets?

I'm not so sure if Samsung really goes for a 4460 within its GalaxyS2 family. 4470 would make far more sense especially since Samsung wants something both for tables as for smart-phones.

Tangey,

Am I the only one around here that has the gut feeling that OMAP4 will be the reference platform for Android Ice Cream Sandwich?

Really? I thought the A5 had 8-9x the graphics power of an A4, and given the A4 had a 200MHz SGX535, a 200MHz SGX543MP2 would fit within that comparison.

The smart-phone or the tablet A4 SoC had the GPU running at 200MHz?

 

[tangey]

Tangey,
Am I the only one around here that has the gut feeling that OMAP4 will be the reference platform for Android Ice Cream Sandwich?

I got the feeling the TI rep was alluding to it, but maybe that was just me.

Does the fact that Ti went for (one assumes) a higher clocked single core 544, instead of dual core at lower frequency, suggest that space on the Soc was at a premium. One assumes a dual core approach would be the better approach in terms of power granularity options.

i.e. 2x544 wil not fit into the space of 1x540 (this also assumes that the 4470 is on the same process as the other Omap4 members).

 

[Rys]

Area is always a concern, but there is another good reason to stick with single core versus MP for 4470 in the context of TI's roadmap.

 

[tangey]

http://www.prnewswire.com/news-rele...-all-about-the-user-experience-123006938.html
x2.5 graphics power of 4430.

Sounds like the 544 in the 4470 is clocking around 400Mhz

 

[Ailuros]

http://www.prnewswire.com/news-rele...-all-about-the-user-experience-123006938.html
x2.5 graphics power of 4430.

Sounds like the 544 in the 4470 is clocking around 400Mhz

I checked that one and TI's own announcement; unless I missed something I don't see 4430 being mentioned anywhere. What tells me that the comparison isn't against 4460?

 

[tangey]

I checked that one and TI's own announcement; unless I missed something I don't see 4430 being mentioned anywhere. What tells me that the comparison isn't against 4460?

The TI one says:-

"Comparative data relative to the OMAP4430 processor"

http://newscenter.ti.com/Blogs/news...t-s-all-about-the-user-experience-716621.aspx

Its immediately below the feature table.

 

[Ailuros]

The TI one says:-

"Comparative data relative to the OMAP4430 processor"

http://newscenter.ti.com/Blogs/news...t-s-all-about-the-user-experience-716621.aspx

Its immediately below the feature table.

And Anand figures the 544 is clocked at 384MHz. However if you do the math purely based on maximum FLOPs per core the 544 should be clocked less than 384MHz in order to be at 2.5x times the 307MHz/SGX540. It sounds more like either TI or IMG have set a reasonable average than anything else.

 

[tangey]

And Anand figures the 544 is clocked at 384MHz. However if you do the math purely based on maximum FLOPs per core the 544 should be clocked less than 384MHz in order to be at 2.5x times the 307MHz/SGX540. It sounds more like either TI or IMG have set a reasonable average than anything else.

Unless Anandtech has been specifically told the clock, it hasn't been released in any of the PR or video. They may be just assuming its the same clock as 4460, which *was* confirmed yesterday at 385Mhz.

However:-
385/307 x 2= 2.5 (anand table indicates x2 Gflops performance improvement from 540->544.)

 

[Ailuros]

Unless Anandtech has been specifically told the clock, it hasn't been released in any of the PR or video. They may be just assuming its the same clock as 4460, which *was* confirmed yesterday at 385Mhz.

However:-
385/307 x 2= 2.5 (anand table indicates x2 Gflops performance improvement from 540->544.)

If you'd go into hair splitting mode the table isn't correct.

 

[tangey]

Well if the tables wrong, all calculations based on it are naturally screwy. When I saw your reference anandtech article, I was assuming you were implicitly accepting its data

 

[Lazy8s]

Targeting Windows devices currently doesn't necessitate quite so conservative power consumption (especially for max performance when they already have the M3s for the lighter weight stuff), and seeing as the CPU has gone up in clock from the OMAP4460 to the 4470, the SGX544 could really be clocked way up... ~400 MHz doesn't quite hit that profile.

 

[Xmas]

If you'd go into hair splitting mode the table isn't correct.

Maybe you could point out what's incorrect?

 

[Lazy8s]

TI included the qualifier of "overall graphics performance increase" to their OMAP4470 claim, so they appear to be providing some kind of conservative performance guideline in the context of all performance areas as opposed to referencing the difference in GFLOPS. Anandtech made a reasonable observation but still a very suspect assumption.

 

[tangey]

I notice in the TI PR, and again highlighted in the Anandtech piece, special display composition I/P is in the 4470 (as distinct from the rest of the series). Anandtech said TI confirmed that this was 3rd party I/P, but would not specify who.

"The next part of what's new in OMAP4470 is inclusion of a new hardware composition system for doing display composition without taxing the SGX544. TI wouldn't disclose whose IP this is, but did acknowledge that it's from a third party and includes a dedicated 2D graphics core for compositing the entire display. Ordinarily this is done on the GPU, but TI hopes to accomplish the same composition on this hardware accelerator in a more power and bandwidth efficient manner for driving large displays while maintaining low power profile."

So hardware composition and 2D graphics core.

I know IMG have I/P in this category, so I wonder could it be their I/P and if so, why not disclose it.

If it isn't, I wonder whose I/P it is.

 

[Deleted member 13524]

Aren't the other OMAP4s using proprietary IP from TI for 2D acceleration?
Why would the 4470 use a different one?

 

[Lazy8s]

Considering TI don't currently even go for the obvious win of using PowerVR's video solutions, I haven't had especially high expectations for Imgtec to be the IP provider of that low power graphics/display-compositor core.

 

[tangey]

Aren't the other OMAP4s using proprietary IP from TI for 2D acceleration?
Why would the 4470 use a different one?

Don't know.

Anandtech's quote however was not ambiguous.

".....but did acknowledge that it's from a third party and includes a dedicated 2D graphics core for compositing the entire display"

my only thought is that for the composition hardware to work it needed to be mated to specific 2D accelerator.

 

[Erinyes]

Given the modest clock increases im guessing its still on 45nm. Given that its only sampling in October couldnt they have gone 28nm? OMAP 5530 which is on 28nm is sampling in October as well so i dont see why it wouldnt have been possible

 

[Lazy8s]

Availability of the 4470 will be better all around -- predictability, reliability -- and would be a reason for a multi-tiered roadmap for SoC performance.

 

[Lazy8s]

The answer directly from TI:

There’s no expected overlap, Granger said during our phone call. “Both the OMAP4470 and OMAP 5 chips will sample in the second half of 2011, but products using the 4470 are expected in the first half of 2012, while products using OMAP 5 are slated for the second half of next year,” he told me.

http://gigaom.com/mobile/texas-instruments-omap4470-smartphones-tablets-windows-laptops/

And the author sees the reasoning as:

That timeline makes sense for a few reasons. First, the OMAP4470 has the same pin configuration as existing OMAP 4 chips, such as those used in the BlackBerry PlayBook. That means device*manufacturers currently using an OMAP 4 can quickly upgrade their hardware with little development cost. Second, while Microsoft hasn’t yet announced when it would have a version of Windows available for ARM processors, it’s widely expected to be some time next year. Whether that happens in the first or second half of 2012, TI is covered by both the new OMAP4470 or the future OMAP 5 chips to power such computers.

It's another indicator of how TI is stepping up their game to compete on a more frequent update cycle, and my belief that their Series6 implementation wouldn't have to wait for OMAP6 was also apparently confirmed by the Digitimes article mentioning the new licensees.

TI has licensed Imagination's PowerVR Series6 for its OMAP 5 processor

http://www.digitimes.com/news/a20110608VL200.html