Low-cost emerging market SoC/phone discussion

[Ailuros]

I'd think that TI will honor its OMAP5 design wins but starting from early 2013.

As for Intel I've no idea how they're counting beans. I was left with the impression that they had claimed in past marketing material that their next generation tablet GPUs will be 10x times faster than the current solution. If you'd take that with twisted GFLOP only (marketing) number crunching the 554MP4 in iPad4 is already there in theory and in real time I wouldn't be surprised if its about 4-5x times faster than a 545@533MHz in Clovertrail.

I don't know how powerful or not their first GenX tablet SoC solution will be, but I have severe doubts that it could counter whatever at least Apple has in the works. Besides that and since we're in a low cost SoC thread, both MediaTek and HiSilicon are PowerVR Rogue licensees.

 

[Exophase]

yeah, well worked out both of you. First time I've seen PowerVr stated in terms of the number of ALUs.

Still 544MP2@300MHZ is pretty decent, probably not a match for the 543MP4@250 of the ipad3, but none the less, quite competitive amd with a better compliance set.

Allwinner (?) seem to have done a decent job in Soc design time. IMG licensed to Allwinner sometime between May '11 and April '12
(was mentioned in the full year results to April 2012 as a new licencee), so looks like they went from licence to end silicon in 18 months max.

Also, Is this the first tablet boasting 4K video out ? , wonder are they implementing some PowerVR video decode ?

Allwinner already claimed it could do 4K ("Super HD 2160P") on A10 SoC (Cortex-A8, Mali-400MP1). They also claimed that its Cortex-A8, probably on 65nm or 55nm, clocked at 1.5GHz. That number was probably really the result of adding a 1.2GHz CPU with 300MHz GPU clock.

With this and their new games with calling SGX544MP2 "8 core" I have a hard time taking them very seriously.

 

[Rurouni]

I kinda hate Allwinner because they seemed try to deceive the uninformed public by the naming of their processor. A10, A13? Yes, I have seen a few ads where products using this soc claim superiority over competing products, because other products limited to A9.

Btw, there are news/rumors that Sony was/is sampling quad core Mediatek SoC for their next phone. Unless this SoC is dirt cheap, I would prefer dual core Krait instead of quad core A7 (I believe the quad core Mediatek SoC is using A7). Do you guys think this new Mediatek chip would still be cheaper compared to S4 after we include the necessary radio module? I hope that the decision on using Mediatek is because it's cheap (and pass the saving to consumer) instead of fulfilling the quad core checklist.

 

[tangey]

I kinda hate Allwinner because they seemed try to deceive the uninformed public by the naming of their processor. A10, A13? Yes, I have seen a few ads where products using this soc claim superiority over competing products, because other products limited to A9.

Btw, there are news/rumors that Sony was/is sampling quad core Mediatek SoC for their next phone. Unless this SoC is dirt cheap, I would prefer dual core Krait instead of quad core A7 (I believe the quad core Mediatek SoC is using A7). Do you guys think this new Mediatek chip would still be cheaper compared to S4 after we include the necessary radio module? I hope that the decision on using Mediatek is because it's cheap (and pass the saving to consumer) instead of fulfilling the quad core checklist.

Well, given ailuros posted a few posts ago a link to an article citing qualcomms new quad core A7 specifically for the Chinese low cost entry market, it would seem that mediatek are leading the others in spec-ing this segment.

 

[Rurouni]

Well, given ailuros posted a few posts ago a link to an article citing qualcomms new quad core A7 specifically for the Chinese low cost entry market, it would seem that mediatek are leading the others in spec-ing this segment.

How small A7 core compared to A9 and Krait at similar process? I assume quad core A7 will be smaller than dual core Krait, otherwise it wouldn't make any sense to go quad core for the sake of having quad core. Probably lower power consumption?
Qualcomm also have products using A5 core. I assume these A5 and A7 Qualcomm is a vanilla A5/A7 with little customization?

 

[Deleted member 13524]

How small A7 core compared to A9 and Krait at similar process?

It seems they perform similarly but the A7 has a larger set of instructions and more power-saving tweaks.. OTOH it's usually coupled with less L2 cache (512KB/core in A9 vs. 256KB/core in A7) and it's in-order whereas the A9 is partially (as people have said here) out-of-order.

If I had to guess, at the same 28nm node the A7 should be a bit smaller than the A9, while performing a bit lower but with a much better power efficiency.

I assume quad core A7 will be smaller than dual core Krait, otherwise it wouldn't make any sense to go quad core for the sake of having quad core. Probably lower power consumption?

The marketing departments all over the world seem to disagree.
Look at how Tegra 3 could perform much better if, instead of the second pair of A9s it had a dual-channel memory controller and a larger GPU.
Yet it still got almost all design wins for the most popular Android tablets throughout 2012.

Qualcomm also have products using A5 core. I assume these A5 and A7 Qualcomm is a vanilla A5/A7 with little customization?

AFAIK, Qualcomm implemented a custom FPU unit in the ARM11 core for their first MSM722x chips, but since both the A5 and the A7 can already have NEON and FPU, I don't know if that would be worth it.

 

[Ailuros]

It seems they perform similarly but the A7 has a larger set of instructions and more power-saving tweaks.. OTOH it's usually coupled with less L2 cache (512KB/core in A9 vs. 256KB/core in A7) and it's in-order whereas the A9 is partially (as people have said here) out-of-order.

If I had to guess, at the same 28nm node the A7 should be a bit smaller than the A9, while performing a bit lower but with a much better power efficiency.

I haven't followed things as close to be sure, but the above sounds give or take accurate from where I stand. I'd only add that A7s from Qualcomm and Mediatek are going to be clocked at 1.2GHz, which if one would compare it f.e. to the iPad2/3 CPUs I don't see why the former two solutions would come at any serious disadvantage against the latter.

The marketing departments all over the world seem to disagree.
Look at how Tegra 3 could perform much better if, instead of the second pair of A9s it had a dual-channel memory controller and a larger GPU.
Yet it still got almost all design wins for the most popular Android tablets throughout 2012.

In NV's defense for the timeframe Tegra3 launched I don't see what else they could had done on the CPU side of things at least, in order to increase performance compared to Tegra2. One could say that they could have stayed with dual A9s and increase much further frequencies but power consumption is never really all that kind if you go too far with frequencies. On hindsight didn't Texas Instruments quote as just one example up to 1.8GHz for the OMAP4470? I haven't seen any yet beyond 1.5GHz.

I disagree though with almost all design wins especially when popularity appears in the same sentence. If you look at Samsung Exynos and Amazon Kindle whatever sales volumes, what T3 achieved in terms of sales volumes within this year is rather peanuts.

Don't get me wrong Tegra3 is IMHO a quite well positioned SoC (well if it would come to criticism there's a lot that could be said about competing solutions too...) and NV's sales for Tegra are in an upswing and most likely to continue to scale; but no they definitely didn't take the Android tablet market share by storm when it comes to sales volumes.

 

[Deleted member 13524]

I disagree though with almost all design wins especially when popularity appears in the same sentence. If you look at Samsung Exynos and Amazon Kindle whatever sales volumes, what T3 achieved in terms of sales volumes within this year is rather peanuts.

Don't get me wrong Tegra3 is IMHO a quite well positioned SoC (well if it would come to criticism there's a lot that could be said about competing solutions too...) and NV's sales for Tegra are in an upswing and most likely to continue to scale; but no they definitely didn't take the Android tablet market share by storm when it comes to sales volumes.

AFAIK the Nexus 7 surpassed the Kindle Fire in sales this year. Furthermore, at least in Europe the Asus Transformer (Prime, Infinity and most of all the TF300) line seems to be stomping everything else.
The Exynos isn't present in many tablets. There's the Note 10.1, the Tab 7 Plus and the SAMOLED+ Tab 7.7, and you can tell these last ones weren't all that successful given the lack of O.S. updates.
Also, there's the fact that Samsung lied about their tablet sales when factual numbers came up in the lawsuit vs. apple.

And nVidia also has the 2012 lines from Acer, Toshiba, Lenovo, Sony and of course the Surface RT.


Nonetheless, I think the Nexus 7 alone has made the Tegra 3 the most sold SoC for tablets in 2012, apple thingies aside. Asus said it was selling over a million a month back in October, and I bet it's selling a lot more now that there's a 32GB version and people are getting stuff for Christmas.

 

[Ailuros]

With all due respect the above isn't based on any statistics or tablet market share. I suggest we leave it as is until official statistics appear and then we can revisit the topic again. In the meantime I'd like to stand corrected but you'll have a damn hard time convincing me that Tegra3 will surpass in sales volumes Samsung and Amazon combined.

 

[tangey]

The following block diagram of the all winner A31 would indeed suggest they are being rather wide in their description of the gpu as 8 core.

It does also spec out the video decode in more detail, and if you click on the A10 info, will see that its a major upgrade in that department.

http://www.allwinnertech.com/en/product/A31.html

 

[Exophase]

How much do these high end decode capabilities, that is, beyond H.264 1080p30, really matter? Is there really any "4k" content? There's 3D but I'm not sure a lot of people care about that either.

I remember nVidia was hyping better than 1080p decode capabilities on Tegra 3 a long time ago.. I wonder if anyone has ever put that to the test. If we're not already hitting diminishing returns it's going to start happening soon.

 

[Lazy8s]

Imgtec described the scale of individual SGX cores in terms of their number of shader pipelines, so seeing Allwinner rebadge those as cores in accordance with the current marketing fad didn't make for too tricky of a reference nor is it so unexpected.

 

[Ailuros]

http://www.glbenchmark.com/phonedetails.jsp?benchmark=glpro25&D=Onda+V812+Core4

Well that Allwinner SoC is doing quite well and yes it sounds like a 544MP2@300MHz.

 

[liolio]

I will end up paying royalties to Anandtech...

More detail about the AllWinner A31 SoC.

 

[Exophase]

I want to see some real CPU performance numbers. Quick, someone come up with a mobile benchmark that isn't garbage.

 

[Mariner]

The initial reviews I've seen of the Onda V812 mentioned in the Anandtech piece seem to indicate that battery life isn't particularly great at present. In the order of 4 or 5 hours which isn't really good enough, if you ask me.

The performance is reported as very good with a 'buttery' smooth interface, however.

 

[liolio]

The initial reviews I've seen of the Onda V812 mentioned in the Anandtech piece seem to indicate that battery life isn't particularly great at present. In the order of 4 or 5 hours which isn't really good enough, if you ask me.

The performance is reported as very good with a 'buttery' smooth interface, however.

Is the battery replaceable? I mean I could definitely see such a product going with indeed a pretty cheap part but "standard".

 

[Mariner]

No - I don't know of any tablets which have easily-replaceable batteries.

Some sellers claim that the Onda V812 has a 6000mAh battery but others state 4500mAh. I think it likely that the lower figure is correct.

These budget Chinese tablets can have decent battery life but all too often, the manufacturers are so keen to cut their costs that the first thing they pare back is battery size. The sellers can be very slippery in their sales tactics and dishonesty about battery sizes is rife.

 

[Ailuros]

http://www.fudzilla.com/home/item/30472-archos-rolls-out-platinum-line-tablets

The SoC in that new Archos Platinum line of product sounds suspiciously like the Allwinner A31; that idiotic "8 core SGX544" sounds familiar.

 

[Mariner]

Pretty good specification for the price. The two main issues with these budget tablets seem to be poor (or at the very least average) battery life and poor wifi reception, something which is a particular issue for tabs with aluminium shells.

Will Archos have considered these issues or is it more likely to be a bog-standard model from one of the Chinese OEM manufacturers?

Actually, a quick google around answers my own question:

http://www.slatedroid.com/topic/529...-are-most-likely-rebadged-onda-v972-and-v812/

Looks as though the Archos Platinum line are probably rebadged Onda devices. They have a reasonable reputation but there appear to be issues with the V812 - a ridiculously small battery so less than 4 hours battery life is achievable (!), frequent crashes when the A31 is running at 1.2GHz (1.0GHz is apparently OK) and excessive heat production.

The build quality and screens of these devices are pretty good but it appears that the guts of the device aren't up to standard.