TI announces OMAP5

Humph... other than Apple, I can't see any current handset supplier paying to get a performance bump on current models.
All Android makers seem to be keen on making your current handset outdated asap, so you'll have to buy the next-gen.
Same goes for Samsung, unfortunately.


Would you be willing to share the new driver with the guys over at xda-dev, Rys?
I'm sure they'd be all over it.
 
There's no way I can share it, but I do push the powers that be as much as I can to get better drivers out there as much as possible via the correct channels (it's my job to keep us competitive after all).
 
"If they are willing"
I am not sure, but I think there are many Netbooks out with the GMA500 and this have no real new driver since three years. The partner here is intel, I am not sure that they are willing, better I am sure, they are not willing! :(
Ok, that was OT, excuse me!
 
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 :)

You got me excited here!

There's no way I can share it, but I do push the powers that be as much as I can to get better drivers out there as much as possible via the correct channels (it's my job to keep us competitive after all).

Only for my hopes to be dashed here :(
 
Although the announcement most likely relates to Omap6, this is probably as good a thread as any.

TI expands licensing agreementS with imagination technology to include Series6.

http://www.imgtec.com/corporate/newsdetail.asp?NewsID=630

"....has signed multi-use license agreements with Texas Instruments Incorporated (TI) for new Imagination Graphics Processing Unit (GPU) IP cores from both the latest POWERVR Series5XT family and the next generation POWERVR Series6 family, codenamed ‘Rogue’"

Remi El-Ouazzane , Vice President and General Manager, OMAP Platform Business Unit, TI says:

"....TI is pleased to once again deploy Imagination’s market-leading POWERVR cores into the next generation of OMAP platforms which will be a key driving force in shaping such future applications like augmented reality, OpenCL and, of course, gaming in mobile phone and mobile computing spaces.”

So *THIS* looks to me like TI has taken 554 from the 5 series. So it calls into question my assertion that the Renesas announcement of a few days ago had to be 554 as there was no other 554 licencee, although it might very well turn out still to be the case.

So either Omap5 will get some sort of mid-life kick with variants that have 554, or Omap6 family will contain both 554 and rogue variants. Thats 3 known licencees for series 6, ST,Mediatek (which is a surprise), and TI.
 
Funny there's no Rogue/Intel announcement yet; the absence of it would be a real surprise.
 
Funny there's no Rogue/Intel announcement yet; the absence of it would be a real surprise.

I'd be surprised to see one.

There have only ever been 2 formal announcements from IMG relating to Intel.

one was 10 years ago for MBX that was initially used in the 2700G companion chip.

The other was 5 years ago and was basically a "catch-all" allowing IMG to satisfy the requirements of the stock market, and also providing a situation where cores licenced by Intel would not actually be announced.

IMG have subsequently referred to some Intel chips containing their I/P, but only after Intel have launched the chip and formally identified the I/P, or that it has come out from some other source.

For example we know that Intel have used 535, 545 and at least 2 VXD cores and 1 VXE core, but to this day the exact video cores have never been disclosed.

This is almost exactly the same scenario as with Apple, again 1 catch-all announcement and then nothing else. In fact as far as I am aware, IMG have never confirmed the cores used by Apple, and other than the graphics drivers being identifible, neither has Apple, and again the video cores being used are not acknowledged.

So if Intel does license Rogue, I'd be surprised to here about it until a physical chip is close to being manufactured.

Having said that, given we've now had 5 licence announcements in 2 weeks, including 2 that specifically include Rogue, anything is possible !


To get this thread a *bit* more back on topic, I was reminded today of the following presentation from exactly 1 year ago:-

http://armnews.wordpress.com/2010/05/19/arm-analyst-day-2010-full-report/

the following points bear reflection 1 year later:-

"ARM sees the deal between Texas Instruments and Imagination Technologies as a setback to their GPU aspirations (announced yesterday)."

This referred to TI announcing they would be mating SGX544 to the A15 cores in their new Omap5 family, and not Mali-400 or T604. Even though TI was a lead partner in the design process of the A15. I would assume ARM would have been fighting even harder to get T-604 into OMAP6, as I imagine OMAP5 might have been too tight a timeframe for them. It has to therefore be considered a bigger setback not to have won any of the multiple graphics seats that were available for Omap6.

"....no Mali deals have been lost due to lack of performance. They have been down to cost of switching GPU provider"

Well, we've had fujitsu recently take a PowerVr licence, when there was no previous 3rd party supplier (they used their own in-house stuff), so when the cost to switch was equal, they chose PowerVr.

But we've also had LG "switch" to IMG after they took a mali licence, and we've had ST drop ARM's graphics in favour of IMG's rogue for their top end A9600 when there was no previous relationship with IMG, so EVEN when there is cost involved in switching from Mali, these semis are seeing the cost as well worth while.

So for me, the above quote will have to go down as marketing rather than relecting the real state of play.
 
Last edited by a moderator:
TI's 554 license here in addition to Series6 shows they're being more aggressive, so the extra Series5XT license should be part of a more frequent upgrade schedule for new graphics cores rather than pushing out their integration of Series6.

The question was posed before: was the replacement of the OMAP4430 with the OMAP4460 little more than a name change?
 
Good point tangey I hadn't thought of. After all Intel licensed in 2005 something codenamed "Eurasia" in 2005.

As for OMAP4460 I don't think the name change is a coincidence either or signifies something boring like just a frequency increase, but time will tell if the gut feeling is in the right direction.

SGX554 might have been some sort of "special order" from maybe one partner that feels the need for very high floating point throughput? I'm not so sure TI would fit that picture, but if yes I could see it more as a part for the later OMAP5 than OMAP6 family.

All summed up it could very well be that TI might play with different GPU IP cores per OMAP family and not like in the past just relying on frequency increases.
 
There's no way I can share it, but I do push the powers that be as much as I can to get better drivers out there as much as possible via the correct channels (it's my job to keep us competitive after all).



Who are "us"?

If you can say ...
 
TI is several times referring to OMAP5 as being manufactured on "their" 28nm process, what is the significance of this? I know TI was looking to develop on UMC (and others - TSMC? They weren't happy with using Samsung again..); did they develop the process itself in partnership?
 
New performance figures:

http://e2e.ti.com/blogs_/b/mobile_m...-proves-its-graphic-intense-capabilities.aspx

In GLBenchmark 2.5 @ 1080p, offscreen (no v-sync) results are:

OMAP5430 SGX544MP2: 45 FPS
ipad 3's A5X SGX543MP4: 43 FPS
OMAP4470 SGX544MP1: 19 FPS
"Green Competitor" (Tegra 3?): 20 FPS
"Blue competitor" (Snapdragon S3 w/ Adreno 220?): 9 FPS


TBH, it doesn't look that good, considering that any device coming with one of these should still be half a year away, coming after Snapdragon S4 Pro, Exynos 5xxx and a bit earlier than Tegra Wayne.

But it should be competitive, nonetheless.
 
New performance figures:

http://e2e.ti.com/blogs_/b/mobile_m...-proves-its-graphic-intense-capabilities.aspx

In GLBenchmark 2.5 @ 1080p, offscreen (no v-sync) results are:

OMAP5430 SGX544MP2: 45 FPS
ipad 3's A5X SGX543MP4: 43 FPS
OMAP4470 SGX544MP1: 19 FPS
"Green Competitor" (Tegra 3?): 20 FPS
"Blue competitor" (Snapdragon S3 w/ Adreno 220?): 9 FPS


TBH, it doesn't look that good, considering that any device coming with one of these should still be half a year away, coming after Snapdragon S4 Pro, Exynos 5xxx and a bit earlier than Tegra Wayne.

But it should be competitive, nonetheless.

Considering OMAP5's possible GPU frequency, those results don't sound weird compared to the iPad3 (hypothetically). By the time OMAP5 starts shipping in volume devices I'm afraid it's bigger headache isn't necessarily Apple's current high end tablet, but rather Apple's next generation high end tablet and that includes all other fore mentioned solutions too.

In any case if the "green competitor" should be Tegra3, then the ULP GF in that one is doing extremely well IMHO, since it would within a margin of error break even with OMAP4470.
 
Back
Top