Renesas licenses PowerVR SGX

[Ailuros]

http://www.imgtec.com/News/Release/index.asp?ID=623

 

[deviantchild]

Dang! you beat me by an hour and a half! LOL

 

[Megadrive1988]

SGX is a pretty large leap beyond the MBX family, IIRC.

 

[tangey]

Not totally on topic, but I see renesas just announced a new automotive Soc, which includes SGX531, the R-CAR E1.

http://am.renesas.com/applications/automotive/cis_highend/child_folder/rcar_e1_child.jsp

Interesting point here is that the only known licencee for SGX531 was NXP who designed various STB chips using it (and then sold the division to Trident).

I was unaware that Renesas (or indeed NEC who they acquired a while back), had SGX531 licence, and given that the core is in IMG terms quite old, its a strange time for an Soc to be announced with it. Either they got it via some acquisition, or they specifically needed the capabilites of 531, which as I recall has an 128bit bus as a unique feature.

Renesas now have access to 530,531,535,540 and 543MP2, either via direct licencing or via acquisition of NEC. They also have another member of the multi-core series 5 as well, via this recent announcement:-

http://www.imgtec.com/corporate/newsdetail.asp?NewsID=629

This makes them by far IMG's most prolific licencee in terms of cores.

Its most likely, given their deep engagement with IMG, that they are also a Rogue licencee.

 

[Ailuros]

Its most likely, given their deep engagement with IMG, that they are also a Rogue licencee.

What would keep IMG then from announcing it?

 

[Lazy8s]

IMG may have already fulfilled their investor obligation of disclosing it among the total number of licensees, and perhaps Renesas just doesn't want to be named until they're closer to announcing product.

 

[tangey]

What would keep IMG then from announcing it?

Could be a number of reasons I guess. They have publicly said they have 6 Rogue licencees, with only 3 specifically named (TI,STM,Mediatek).

I think its reasonable to assume the lack of names for the other 3 are due to the licencee not wanting it, rather than IMG. I can see that STM would want to be the first announced licencee, they need to raise their profile and be seen to be going for top performance, Mediatek could similarly do with the Qdos of being one of the first named licencees. TI probably want to show their road map has long legs for customers possibly looking to move away from them, or indeed courting the likes of google and microsoft.

Of the unannounced ones, I see them as Apple, Renesas and Intel. We know apple doesn't want anyone to know ANYTHING. Intel has likely issues within its own graphics division, and perhaps doesn't want to be seen to be undermining that division, effectively saying that they have determined it can not come out with competing designs in this space for 2013. Renesas, I agree, has less reason that any not to announce.


I think working with 6 different partners out of the blocks must be quite demanding for IMG, and likely taking up most of the resources that they can currently throw at it, so we may not see many new rogue licencees announced, until they have a capability to fully support them. A quick look at the website shows they are looking for around 50+ people in the graphics dept alone, include 4-5 DX11 driver guys in the Poland office, and another 20ish vacancies in the video dept.

 

[Ailuros]

What I'm thinking is (and yes its based on pure speculation) that the remaining 3 want to remain unnamed. Apple and Intel would be two relatively "certain" candidates, but I still don't see why Renesas wouldn't want to announce anything just yet. I know it's just wishful thinking from my side but I'd prefer the mystery Nr.3 to be a surprise this time

A quick look at the website shows they are looking for around 50+ people in the graphics dept alone, include 4-5 DX11 driver guys in the Poland office, and another 20ish vacancies in the video dept.

Why the heck do they always pick countries in Europe with crappy weather? How about pinning countries in southern Europe where it's hot and hardly ever rains?

 

[tangey]

I know it's just wishful thinking from my side but I'd prefer the mystery Nr.3 to be a surprise this time

To me Mediatek was a massive surprise, it shows they have major ambitions in Asia. It also shows presumably that Rogue will hit low end devices really quite quickly (within a year of the ST chips come out). The Mediatek Rogue chip could quite possibly be much more valuable to IMG that the ST one !

BTW, digitimes yesterday ran an article saying that Mediatek has had to ask its foundries for more capacity for its new phone Soc (with SGX in it), due to a surge in demand. They suggested that the chip will hit 2-3M a month by November.

http://www.digitimes.com/news/a20110823PD218.html

 

[Tabris]

Renesas have just announced their first SGX543MP2 SOC
http://am.renesas.com/press/news/2011/news20111018.jsp

TSMC 45nm
1GHz ARM A9 x4
800MHz SuperH SH-4A x1
SGX543MP2
DDR3

 

[Ailuros]

http://renesasmobile.com/news-events/news/news-20111025-Advanced-Multi-Core-PowerVR-GPU.html

...advanced multi-core PowerVR™ Series5XT MP Graphics Processing Unit (GPU) from Imagination Technologies, the platform represents a significant step towards delivering a games console quality experience on a mobile phone.

I like the way that has been phrased; at least it's more accurate than others with half or less than half the performance of that claim already "games console quality experience".

 

[Lazy8s]

I looked again at Tabris's link and just realized that the 543MP2 in the R-Car H1 SoC is clocked quite high, going by the quoted theoretical GFLOPs.

And despite their move to ARM, throwing in some SuperH with that nicely appointed SH-4A as a "multimedia engine" is nice to see.

 

[tangey]

I see the first scores for the Renesas APE5R soc have hit glbenchmark.

http://www.glbenchmark.com/phonedet...21&D=Kyocera+HONEY+BEE+101K&testgroup=overall

46.6 for 720p egypt. The fact that it's a slightly lower figure than the standard egypt test (@48.7), would point to there not being a lot of ceiling to play with. Ok its early figures so perhaps not representative, but with dual core A9 @ 1.2Ghz, and with basically the same graphics cores, the fact that they are only getting somewhat over 60% of the iphone4S score would suggest:-

1) Android drivers for SGXmp are at an early stage
2) Renesas are clocking way below the 200mhz in the iphone
3) The Soc's graphics path is much poorer than the Apple A5s.

 

[french toast]

I see the first scores for the Renesas APE5R soc have hit glbenchmark.

http://www.glbenchmark.com/phonedet...21&D=Kyocera+HONEY+BEE+101K&testgroup=overall

46.6 for 720p egypt. The fact that it's a slightly lower figure than the standard egypt test (@48.7), would point to there not being a lot of ceiling to play with. Ok its early figures so perhaps not representative, but with dual core A9 @ 1.2Ghz, and with basically the same graphics cores, the fact that they are only getting somewhat over 60% of the iphone4S score would suggest:-

1) Android drivers for SGXmp are at an early stage
2) Renesas are clocking way below the 200mhz in the iphone
3) The Soc's graphics path is much poorer than the Apple A5s.

What you mean like bandwidth? what about Apple having say some dedicated Vram on board?? after all no would know outside apple...

Drivers could certainly explain it...look at Adreno for proof of that.
I dont think it could be just down to clock speeds, you would think A5 in 4S is clocked no higher than 300mhz any way, i would think 250mhz.

 

[Ailuros]

I see the first scores for the Renesas APE5R soc have hit glbenchmark.

http://www.glbenchmark.com/phonedet...21&D=Kyocera+HONEY+BEE+101K&testgroup=overall

46.6 for 720p egypt. The fact that it's a slightly lower figure than the standard egypt test (@48.7), would point to there not being a lot of ceiling to play with. Ok its early figures so perhaps not representative, but with dual core A9 @ 1.2Ghz, and with basically the same graphics cores, the fact that they are only getting somewhat over 60% of the iphone4S score would suggest:-

1) Android drivers for SGXmp are at an early stage
2) Renesas are clocking way below the 200mhz in the iphone
3) The Soc's graphics path is much poorer than the Apple A5s.

The standard score reminds a lot the initial iPad2 standard Egypt scores.

It could be one out of the 3 above or a combination of them, albeit it sounds suspiciously like (1) to me.

Other than that of course is the MP2 in iPhone4S clocked at around 200MHz IMHO.

 

[tangey]

Renesas have quad and octo core coming

Quad is 2x2A15/A7 targetting 2Ghz for the A15 has SGX544 graphics, Octo is 4x4 A15/A7.

Quad core gives "approximiately" 3x graphics & processing performance of MP5232.

5232 has dual A9 and SGX543. not sure what the 543 is clocked at. but even at a lowly 200Mhz, would imply quadcore is clocking SGX544 >500Mhz, perhaps the 533MHZ which seems to be in vogue at the moment for SGX MP. I suspect the octo core might have a rogue implementation.

http://www.eetimes.com/design/commu...sas-to-show-quad-octo-big-little-chips-at-MWC

 

[Ailuros]

Android drivers? *shrugs*

 

[tangey]

....and as if by magic, courtesy of our friend roninja, a renesas PR confirming Roguein their octo -core Soc. First sighting of a mobile targeted SOC implementing rogue.

"Renesas Mobile will showcase a ultra-high performance octo-core application processor, the APE6, to illustrate its power advantages and leading graphics performance. The big.LITTLE 4.4 compute engine is coupled with the latest Imagination Technologies PowerVR Series 6 Rogue graphics engine offering amazing performance in a mobile. Renesas Mobile will be demonstrating the advantages of big.LITTLE as well as a measure of the graphics capability"

http://pctelecoms.blogspot.co.uk/2013/02/renesas-mobile-showcases-next-gen-low.html?m=1

 

[mboeller]

oh damn....I only searched for APE6. Bad choice Renesas. They should choose a different name.

 

[Simon F]

oh damn....I only searched for APE6. Bad choice Renesas. They should choose a different name.

!!!