Samsung Orion SoC - dual-core A9 + "5 times the 3D graphics performance"

Yeah, I noticed that 'on store shelves within 2012' part. They're like two separate scalings of the same family, not just architecture, of cores.

 

Don't you think that this could be a little bit unfair?
If the new exFalanx GPU-cores (T604 and T658) are far smaller than a SGX then you can maybe have 4 or even 8 T658 cores with the same Die-area as a single SGX 554 (at the same nm-node of course). So without knowing the die-area of both exFalanx offerings IMHO no valid comparison can be made.

 

someone needs to ask which one is right the blog or the ARM road map (see previous page in this thread). The road map shows that T658 doesn't appear under the tail end of 2013.

 

The roadmap doesn't show 8-core T658 until 4 years from now. We also have ZERO idea of core size. given that they are now incorporating DX11 and full OpenCl compliance, one assumes that core size has increased significantly, ignoring improvements in manufacturing process.

 

New Exynos soc featuring 2 x A15 due in Q2 2012:

http://sammyhub.com/2011/11/30/samsung-unveils-cortex-a15-2ghz-exynos-5250-soc/

Interesting to note the 4x Mali 400 gpu performance. Could the GPU be related to the recently announced SGX MP licence?

http://www.anandtech.com/show/5083/samsung-licenses-powervr-sgx-mp-from-imagination-technologies

We already know the quad core A9 based Exynos 4412 will most likely feature a t604 MP4 judging by ARMs t658 roadmap:

http://images.anandtech.com/doci/5077/Screen Shot 2011-11-09 at 2.06.20 PM.png

 

First look at Samsungs quad-core A9. I assume this is the 4412 thats been showing up in some drivers.


http://www.eetimes.com/electronics-news/4236682/Samsung-gives-peak-at-quad-core-mobile-CPU

"Samsung is using the latest version of its own graphics unit for the chip. It includes four pixel processors and one geometry engine with a dedicated 128 KByte L2 cache. The graphics support the OpenGL ES 2.0 API and can generate up to 57 Mpolygons/s."

"its own graphics unit", is surely a mistake (maybe meant "Arm's own graphics unit"), given the description of 4 pixel processors and 1 vertex processor, it sounds like its still using Mali400MP4 ?

 

Yeh, that sounds like Mali 400 mp4 which is disappointing. The A5 gpu will be the the fastest for a while yet. It also means exynos 5250 most likely contains the t604.

 

I see samsung have finally made public a tech datasheet on the Exynos4210.

The most amazing thing is that not only does the 730+ page document not go into details regarding the Mali400MP4 GPU, it does not even mention that the Soc HAS a 3D GPU at all !

http://www.samsung.com/global/busin...t/Pulbicmanual_Exynos_4_Dual_45nm_Ver00-2.pdf

Its dated August '11, so not a preliminary document as far as I can tell.

 

It mentions the GPU for clock domains section 5. But that is the only reference to a 3d IP

EDIT: also look for G3D in the document, there are other references to it (interrupts, etc.)

 

A Mali-400 configuration could fit that claim of 57M tri/sec for the 32nm Samsung quad A9 SoC assuming the clock speed of the GPU was high enough. 400 MHz would have to be a given just as a start, and it would really need to be closer to 500+ MHz for even theoretical accuracy for a Mali-400 config, not that polygon performance is marketed accurately or usefully most of the time anyway.

 

Once I didn't see it mentioned either in the overview, nor at all in the index, I did bother to look for little traces of evidence. Your clock domain find is interesting as it refers to 400Mhz or 330Mhz options. As I understand it, in production they only used 267Mhz. Might imply some power/heat issues when they tried to run it at original spec.

 

Press releases for the 4212 (32nm dual core) talked about a 50% improvement in graphics performance over the 4210, which would aline exactly with a GPU clock increase from 267Mhz to 400Mhz

 

4210 is able to spend some time at 400 MHz, but its thermal/power limited. 4212 should spend more time at that clock of course.

 

http://www.samsung.com/global/busin...ct/application/detail?productId=7643&iaId=844

Samsung seems to have just made available the 4212 brochure. Besides a huge picture of wind mills to attest to it's green prowess, they are claiming up to 2x better 3D performance than the 4210.

The brochure mentions power consumption being "30% lower than 45/40nm process" so that should give some indication of the difference.

EDIT:
http://www.samsung.com/global/busin...ct/application/detail?productId=7644&iaId=844

The 4210 brochure is also now available. This one emphasizes a plant sprouting from a cup in a green field. On actual tech matters, memory bandwidth is also reported at 6.4GB/s so is unchanged between the 4210 and 4212. 3D performance is reported as 3.2Gpixels/sec in the 4210 so that's presumably their baseline for the up to 2x measure for the 4212.

 

The original press released for the 4212 in Sept said "50% Higher 3d graphics performance"
http://www.samsung.com/global/business/semiconductor/minisite/Exynos/news_10.html#mainClick

One wonders has anything actually changed, or did the marketing dept consider 50% not a big enough figure and so changed it to "up to 2x"

 

http://www.glbenchmark.com/phonedetails.jsp?D=MagicLAB+MagicLEGO4x12&benchmark=glpro21

Looks like this might be the first outing for a 4212 (Mali400mp4@400MHz).

Although it gets twice the performance of the MagicLego 4210......

http://www.glbenchmark.com/compare....icLAB MagicLEGO4210&D2=MagicLAB MagicLEGO4x12

....the magicLego4210 had very poor numbers to begin with.

compared to a recent test of the samsung GT-P6800 (4210), there is about 30% improvment in egypt offscreen and 25% improvement in pro offscreen, so no 50% overall improvement that you might expect from a 50% clock increase, and definitely nowhere near "x2" in the marketing.


fill rate shows about 75% improvement to 775K tex/s. Still below ipad2 raw fill rate of 1G tex/s @250Mhz, nevermind ipad3, and excluding savings due to overdraw removal on ipad.

 

Interesting...Your right its only 30% over the latest Sammy 4210....but over twice that of the magic lego 4210?? as you say the original Magic lego 4210 (wtf? lol) was a poor result, could it be that the new one is also a poor result?? maybe Sammy has got better drivers available for its own chips? it does actually do well with such optimisations...

Well in anycase it beats out HTC one x's Tegra 3....which i thought it would....so will be the most powerfull smartphone gpu of its class..

 

Hardly, the iphone4s gpu beats it, and it's been in production since Sep/Oct. Mali400MP4 drivers should be well optimised by now.

 

ipad2 isn't "Android Smarthpone Class".
It's in a tablet and it's not Android.

 

Sorry i did mean Android only. if you also consider the duel channel memory and likely high cpu speeds...the 4412 is going to rock..aspecially as its on the best process for ARM currently.