TI announces OMAP5

[roninja]

Are there any estimates out there of how much TI is going to sell of the OMAP4?

Currently Omap4430 is publically only designed into the playbook with no obvious announced smartphone wins. Motorola,HP have gone with Nvidia and Qualcomm respectively. This leaves Samsung, Nokia and Sony Ericsson as historic customers to pick up sockets.

They've recently bought forward Omap5 and it's predecessor omap4 appears nowhere like as successful as Omap36x in particular.

On various analysts calls TI themselves have said that OMAP4 had received even greater design in traction than the previous generation.

Let's see next week may obvious change my view on the matter.

 

[Lazy8s]

OMAP may have lost some initiative this generation, but I'm sure OMAP4s will still end up in their fair share of carrier affiliated devices by the end of the cycle like always.

It's definitely deserving performance/power-wise.

 

[Ailuros]

How many variants did OMAP3 have roughly 2 years before it appeared in devices? (honest question)
Today at least 9 different OMAP3 variants exist, whereby I'm not counting the Sitara, Da Vinci, Integra SoCs into them.

I don't know any specifics but it seems like OMAP3 has had an even bigger success compared to OMAP2 or not?

In any case while its definitely possible that OMAP4 might not see the same success as 3 I still think it's too early since OMAP4 seems to have at least a couple of years lifetime ahead.

Arun repeats it over and over again and I try to keep it in mind that vendors aren't buying just from one source SoCs. Just because Motorola has for example licensed SoC X up to now, doesn't mean that they might not opt for something else in the meantime and the very same goes for example for Samsung or Sony Erricsson which might license Tegra2 and/or Qualcolmm SoCs.

 

[shyam335]

Would it be viable if we were to replace the two M4's with a A5?

 

[Exophase]

Would what be?

No matter what, if you think those M4s should be replaced with something better then you're probably missing the reason why those M4s are there.

 

[Deleted member 13524]

If the M4s are doing the same as the M3s in OMAP4, they're only assisting the dedicated media processors.

There's really no point in changing them for A5s. They're never being used as general applications processors so that wouldn't enhance the system's general performance.

 

[warmi]

People are more violently opposed to fur than leather; because it's easier to harass rich ladies than motorcycle gangs.

Guess what ....

http://glossynews.com/society/20100...declaring-war-on-leather-at-motorcycle-rally/

 

[Ailuros]

Guess what ....

http://glossynews.com/society/20100...declaring-war-on-leather-at-motorcycle-rally/

LOL albeit vastly OT, after that I don't think any activists will try again *snicker*

 

[Ailuros]

I'm too bored to revive any ancient OMAP4 related thread. Anand has posted the first OMAP4430 results for GLBenchmark2.0: http://www.glbenchmark.com/phonedetails.jsp?D=LG+Optimus+3D&benchmark=glpro20

Smells like a relevant review coming up for the LG Optimus 3D.

 

[metafor]

Interesting results from Anand:

http://www.anandtech.com/show/4179/ti-omap4-and-lg-optimus-3d-tested

One interesting comment was "LG's software build also uses a newer version of Imagination Technologies' driver (1.6 instead of 1.5) which fixes some rendering issues and improves performance considerably (likely between 10 - 30% in GLBenchmark2)"

It would seem the SGX540 has a lot of life left in it.

 

[Rys]

I submitted some GLBenchmark 2.0 results for S5PC110 (SGX540 at 200 MHz) on Nexus S recently, and Kishonti verified them just before MWC started. The improvements versus the standard drivers shipped show some of the driver-level improvements possible with this generation of the code, but I didn't turn everything on just yet (and we need some Android changes for some of them anyway)

The changes will also help OMAP4 (and indeed Xmas and I developed some of the changes on OMAP4 to start with) and they're probably what's mentioned, but the Optimus 3D driver doesn't have the changes that'll help performance that much (up to 30%) just yet, so it has the potential to go even faster.

S5PC110 gets 26.8fps in Egypt with the driver I used for our MWC demo (go see Kristof there at stand 1D45 to take a look!), versus the ~21fps you can measure on S5PC110-powered phones just now.

So yeah, we'd agree with you that 540 has some life left in it

 

[Lazy8s]

After getting a good representation of the 540's GLBenchmark 1.1 performance only after multiple updates to the driver release, I anxiously anticipated a better idea of SGX's GLBenchmarck 2.0 performance. Thanks for that.

The Nexus S jumped from the 2400 range of all of the other Galaxy S1s to running right into Tegra 2 with the update. Samsung really shouldn't be shy about marketing the 1.2GHz Hummingbird platform to fill this lull in their update cycle; it'd top Tegra 2 in graphics performance and work nicely with a GPU-optimized Android UI like Samsung uses on their Galaxy S devices, helping to nullify the impact of their current lack of a shipping A9 dual core.

OMAP4430 unsurprisingly takes the top spot, topping even the 800x480 comparable Armada. Additional headroom is obvious considering the potential implied by Nexus S's movement and OMAP4's implementation advantage (clock speed, etc.); maybe nVidia needs to rewrite their Tegra 2 "technical doc" where they pull some not-so-represntative numbers for SGX540 out of the air.

 

[Lazy8s]

A day in the life of OMAP5:

http://www.youtube.com/watch?v=wo0qpQxdL3k

Unintentionally amusing, the video is still pretty neat.

 

[Lazy8s]

The Optimus 3D is insane.

It's the most powerful phone by a nice margin. It not only brings 3D, but HD 3D for recording/capturing (with a live 3D viewfinder), playback, and sharing with Youtube 3D. It comes pre-loaded with a few high-end games in 3D, and it's also a 1080p (in non-3D) camcorder and playback device with a lot of versatility for standards and output options.

 

[Erinyes]

I submitted some GLBenchmark 2.0 results for S5PC110 (SGX540 at 200 MHz) on Nexus S recently, and Kishonti verified them just before MWC started. The improvements versus the standard drivers shipped show some of the driver-level improvements possible with this generation of the code, but I didn't turn everything on just yet (and we need some Android changes for some of them anyway)

The changes will also help OMAP4 (and indeed Xmas and I developed some of the changes on OMAP4 to start with) and they're probably what's mentioned, but the Optimus 3D driver doesn't have the changes that'll help performance that much (up to 30%) just yet, so it has the potential to go even faster.

S5PC110 gets 26.8fps in Egypt with the driver I used for our MWC demo (go see Kristof there at stand 1D45 to take a look!), versus the ~21fps you can measure on S5PC110-powered phones just now.

So yeah, we'd agree with you that 540 has some life left in it

Rys, would the updated driver be available publicly only for OMAP 4 or for Hummingbird as well? It would be very disappointing if it was not available for existing Galaxy S phone users


OMAP 5 has dual A15's with a whopping 2 MB L2 cache. Nvidia has stuck to 1 MB L2 even with their quad core A9 implementation. Tegra 2 had 1 MB L2 for two cores. In comparison Qualcomm has gone with just 512 KB for their dual core Snapdragon. Would be interesting to see what effect all this has on performance, and which direction the other SoC makers go

 

[Rys]

Rys, would the updated driver be available publicly only for OMAP 4 or for Hummingbird as well? It would be very disappointing if it was not available for existing Galaxy S phone users

It's up to TI and Samsung, and so sadly I can't say for certain what new drivers they'll ship and what fixes and improvements they'll contain. That said, the point of the exercise was to show them what's still possible with 540, in the hope that they'll take the code, so I'm personally crossing my fingers that existing Galaxy-based phone owners will eventually get a driver that's as good as the one I demoed.

 

[Erinyes]

It's up to TI and Samsung, and so sadly I can't say for certain what new drivers they'll ship and what fixes and improvements they'll contain. That said, the point of the exercise was to show them what's still possible with 540, in the hope that they'll take the code, so I'm personally crossing my fingers that existing Galaxy-based phone owners will eventually get a driver that's as good as the one I demoed.

Thanks for the reply Rys. If made available is it possible for end users to install the driver on our own? Or incorporate it in a custom ROM? (I ask because i recently acquired a Samsung Captivate and a Vibrant )

 

[Rys]

The driver could be integrated into a custom ROM without any problems (as long as the device uses Samsung's kernel), and could be installed on a stock ROM device if you had root (with the same kernel caveat). It's pretty trivial to replace the driver binaries

 

[Deleted member 13524]

Would this driver be compatible with other SGX GPUs or just 540?
There's a ton of OMAP3 devices that take custom ROMs.

 

[Rys]

It'd just be for 540, but there's nothing (other than time and money) stopping more optimised drivers coming out for any SGX, if our partners are willing.