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

[Lazy8s]

That Exynos is a strong enough competitor that I don't think Samsung's CE division will pass over it for Tegra 3.

Both SoCs are still some ways off, though; the consumer products division apparently felt they had to get something on the market near this transition period so Tegra 2 became the obvious processor platform. Caught me by surprise.

 

[Ailuros]

So their internal design isn't ready and they're sourcing from an external vendor? The new Galaxy tabs aren't due until June right?

Would they have some other tablet product using their own SOC later this year?

http://www.anandtech.com/show/4233/...1-89-smaller-than-ipad-2-competitively-priced

Samsung is playing the ambiguous SoC game again and only referring to these two tablets as having a dual-core applications processor. I'm guessing we'll see a mix of Tegra 2 and Samsung Exynos based versions depending on region. Samsung's 8.9 prototypes at the show floor had NVIDIA's Tegra 2 under the hood.

Isn't T3 tablets suppose to be out later this year with quad-core CPU? If that's the case, Samsung may have to use that as well, unless their SOC can at least match the T3.

If Samsung clocks the Mali400MP4 high enough, it's anything but a slouch.

Seems like every major tablet except iPad is using T2, at least for the first half to 8 months of 2011.

Being as early available with a dual core CPU and at the same time being the reference platform for Honeycomb made it the obvious choice for vendors. I don't expect much from Samsung; the Motorola Xoom could stick out of the bunch in terms of sales and SONY's upcoming T2 tablet might be another interesting option to watch for.

 

[Deleted member 13524]

If Samsung clocks the Mali400MP4 high enough, it's anything but a slouch.

Hum.. how high would you have to clock that single vertex shader to match the quad vertex shaders @ 300MHz from T2? Or the quad vec2 unified shaders in the SGX540 @300MHz from OMAP4430?


At least looking at preliminary performance benchmarks from Galaxy S2, it seems that Samsug isn't clocking it high enough in Exynos 4210. That or the drivers are a mess.

 

[Ailuros]

Hum.. how high would you have to clock that single vertex shader to match the quad vertex shaders @ 300MHz from T2? Or the quad vec2 unified shaders in the SGX540 @300MHz from OMAP4430?

It's most likely a single Vec2 unit in Mali400. The ULP GF in T2 sounds like a Vec4 VS ALU but even then you still can't make a viable estimate unless you know what each architecture's triangle setup for instance is capable of. However if it should get clocked at say 400MHz that's a shitload of fragment processing in a Mali400MP4 and 1.6 GTexels/s of raw fill-rate.

At least looking at preliminary performance benchmarks from Galaxy S2, it seems that Samsug isn't clocking it high enough in Exynos 4210. That or the drivers are a mess.

The Galaxy S2 smartphone might have the Mali400MP4 clocked somewhere around 275MHz. However its GL Benchmark2.0 scores have been removed which smells like crappy drivers and the recent appearance of the Hardkernel ODROID-A running at 1366*768 is most likely clocked significantly higher and should have a lot better drivers:

http://www.glbenchmark.com/phonedetails.jsp?D=Hardkernel+ODROID-A&benchmark=glpro20

Low level geometry related tests apart, over 4000 frames in 1366*768 isn't exactly a bad score; au contraire. Here against the T2 based Malata Zpad running at 1024*600:

http://www.glbenchmark.com/compare.jsp?benchmark=glpro20&showhide=true&certified_only=1&D1=Hardkernel%20ODROID-A&D2=Malata%20Zpad

 

[tangey]

slightly OT, but anyone know the Mali configuration in the SmartQ T10 ?, Glbenchmark scores are here:-

http://www.glbenchmark.com/phonedetails.jsp?benchmark=glpro20&D=SmartQ+T10&testgroup=gl

(scores were added a few days ago)

details only say mali-400MP, but not how many cores.

 

[Ailuros]

slightly OT, but anyone know the Mali configuration in the SmartQ T10 ?, Glbenchmark scores are here:-

http://www.glbenchmark.com/phonedetails.jsp?benchmark=glpro20&D=SmartQ+T10&testgroup=gl

(scores were added a few days ago)

details only say mali-400MP, but not how many cores.

Sounds like a single Mali400.

 

[wco81]

What is HP going to use?

Really there are only a handful of vendors which have a shot at shipping tablets in any kind of volume.

Maybe Dell will come with something and undercut everyone.

Or some Chinese brand will sell a knockoff that catches on, kind of like those 99 Euro Android phones.

 

[Ailuros]

What is HP going to use?

Really there are only a handful of vendors which have a shot at shipping tablets in any kind of volume.

Maybe Dell will come with something and undercut everyone.

http://www.dell.com/us/p/mobile-streak-7/pd

Or some Chinese brand will sell a knockoff that catches on, kind of like those 99 Euro Android phones.

As what? Serious competition to iPads? It doesn't take rocket science to see that Apple's success is by far not due the hw alone. I still can't personally find a real purpose for purchasing any sort of tablet (at least not yet), but unless some contender manages to pull a comparable sw ecosystem like Apple it doesn't look any "iPad killer" in sight.

 

[wco81]

Well the T-Mobile 4G network is about to become defunct. No if that Streak was $400 or less unsubsidized, it would have a chance of taking off. As it is, apparently Lenovo and Asus have announced tablets and so far, they're over $500 for 16 GB models.

http://blogs.wsj.com/chinarealtime/2011/03/29/lenovo-asustek-launch-tablets-to-take-on-ipad/

I don't think any single model will outsell the iPad but if a company like Huawei or ZTE brings out the tablet counterpart to the ZTE Blade for $300 or less, they should move some units.

 

[tangey]

Sounds like a single Mali400.

I see another mali-400 device has hit glbenchmark. it uses an AMlogic chip that has a single A9 cpu with mali-400 (unknown cores, assume 1). screen 800x480.

http://www.glbenchmark.com/phonedetails.jsp?D=Ramos+W10

egypt=14, 2pro=44

as comparision the smartQ-T10 mentioned earlier in the thread also uses the same chip.
http://www.glbenchmark.com/phonedetails.jsp?D=SmartQ+T10&benchmark=glpro20

Its a tablet with 1024x768 resolution giving egypt=11.5, 2pro=29.

Interesting that there is a 50% improvement on the smaller res device in the 2pro test, but only a 22%.improvement in the egypt test. Something is limiting the egypt performance.

 

[Deleted member 13524]

Interesting that there is a 50% improvement on the smaller res device in the 2pro test, but only a 22%.improvement in the egypt test. Something is limiting the egypt performance.

Bottleneck in pixel shader performance?
Given the souped-up vertex shader (as stated around here), I don't think the Mali 400 was ever really meant to be sold with a single fragment shader, at least not in >=wvga devices...
Maybe the MP1 was only meant to be coupled with something like a ~700MHz single-core Cortex A5 and <=480*320 screens, for low-to-mid end devices.

 

[Lazy8s]

One Mali 400 core is relatively small, yet Samsung's quad set-up with the right drivers and clocks showed that it delivers higher performance than most, in a tablet implementation at least. Whether the heat and power at that level is also competitive remains to be seen.

 

[Ailuros]

I see another mali-400 device has hit glbenchmark. it uses an AMlogic chip that has a single A9 cpu with mali-400 (unknown cores, assume 1). screen 800x480.

http://www.glbenchmark.com/phonedetails.jsp?D=Ramos+W10

egypt=14, 2pro=44

as comparision the smartQ-T10 mentioned earlier in the thread also uses the same chip.
http://www.glbenchmark.com/phonedetails.jsp?D=SmartQ+T10&benchmark=glpro20

Its a tablet with 1024x768 resolution giving egypt=11.5, 2pro=29.

Interesting that there is a 50% improvement on the smaller res device in the 2pro test, but only a 22%.improvement in the egypt test. Something is limiting the egypt performance.

Egypt should have a lot more vertex shading going on than PRO. Here all three devices in different resolutions in the GL Benchmark2.0 tests. Hardkernel ODROID A should contain a 400MP4 while the others should be single Mali cores.

http://www.glbenchmark.com/compare....amos W10&D2=Hardkernel ODROID-A&D3=SmartQ T10

Most striking weakness in the low level benchmarks are the geometry related tests.

 

[Deleted member 13524]

Samsung has apparently bumped the exynos' default clock to 1.2GHz in Galaxy S II (version i9100).

I wonder if only the CPUs are affected by this 20% clock increase or it's the whole system bus (GPU, memory included).
Playing numbers game with the 3rd-gen Snapdragons @ 1.4GHz?

 

[Deleted member 13524]

Here's a Quadrant bench for the 1.2GHz (final) version of Galaxy S II:

With my OMAP3630 @ 1.2GHz and using Froyo, I get around 1600.
I heard that Gingerbread does quite a big performance bump in Quadrant, in part thanks to the updated GPU drivers, so the result doesn't sound all that impressive to me.
Nonetheless, looks like it's a bit faster than Tegra 2, but it should be slower than OMAP4430.

 

[Lazy8s]

This Exynos laps Tegra 2 in the performance department, as it should, easily; only heat/power/die are yet to be established.

 

[Ailuros]

http://semiaccurate.com/2011/04/26/lg-to-design-own-cortex-processors/

If all of it is true, it sounds quite hairy for Samsung's foundry business.

 

[Laurent06]

As far as TI goes, there was an article at EEtimes.

 

[Deleted member 13524]

Looking at the EETimes article, it sounds that Apple is pouring big money on sabotaging the supply for all other SoC vendors.
It may be bad for Samsung's foundry reputation, but I wouldn't worry about their income.

Nonetheless, it's nice to see UMC coming back to the headlines. They used to trade blows against TSMC in getting fab deals for GPUs back in the late 1990s and early 2000s, afair.

 

[Ailuros]

Looking at the EETimes article, it sounds that Apple is pouring big money on sabotaging the supply for all other SoC vendors.

Using one manufacturer saves time and resources. Apple could eventually use a 2nd tier manufacturer but going from 45nm@X to 45nm@Y isn't just a simple copy/paste job.

At least up until next year Apple doesn't sound like they'll move anything to a secondary foundry if at all after that. Still IMHO it would be best for all sides; Samsung could feed it's own needs more adequately and not piss off its traditional partners and Apple wouldn't depend on just one manufacturing source. The complications are the downside for Apple.

It may be bad for Samsung's foundry reputation, but I wouldn't worry about their income.

If you lose one partner after the other and especially big ones like Texas Instruments then it will inevitably affect their income.

Nonetheless, it's nice to see UMC coming back to the headlines. They used to trade blows against TSMC in getting fab deals for GPUs back in the late 1990s and early 2000s, afair.

Definitely.