x86 mobile SoC future

I think you mean the NDK. The SDK is mostly java.

The Tegra 3 was not as bad as the tegra 2. In fact it was bloody good. I dont other chipsets being much better if any than the Tegra 4.

of course I dont think you would see it that way since if I remember correctly you seem to have a personal vendetta again Nvidia and the Tegra

I would have no issues buying a Tegra device. However, with each generation, they seem to create some hardware-based cardinal sin.

Tegra 1 - ARM11 at a time when A8 and Scorpion are hitting second generation devices. Few devices bothered with it.

Tegra 2 - Solid chipset, but they stripped NEON support. This became a larger issue than I expected, causing it to lag behind other A9-based chipsets in performance.

Tegra 3 - Same GPU for the third straight generation. More cores but still weaker than the competition.

Tegra 4 at this time looks attractive. The GPU is based on mobile Kepler, the CPU is not stripped of any key features, and they have a decent companion core for battery savings.

I suspect that Tegra 4 should handily beat S4 pro in performance, but not battery life. We'll see.
 
i dont put much faith in those sites, but i did read else wear souces sa the gpu is still legacy based on VS and PS cores.
 
i dont put much faith in those sites, but i did read else wear souces sa the gpu is still legacy based on VS and PS cores.

It's legacy in that regard, but it's no longer exclusively based on the same architecture they had used for years. While NV did make claims about Kepler coming to mobile, it seems all we got was, as I said, something based on mobile Kepler. It is finally DX11 compliant, but I'm surprised to see that NV hasn't advertised OpenCL support, or even their own CUDA. Other GPUs for mobile are supporting OpenCL, Google has expressed interest in going that way, and NV even started this discussion when they mentioned the usefulness of CUDA on mobile when they were talking up future iterations beyond Tegra 2.

Bottom line is, it's not full blown Kepler, but it's a huge improvement over what they had tried to push on us the last few versions.
 
No matter how efficient the power draw of. X86 cpu becomes in the future, you can't make up for how bloated and upoorly optimized x86 apps are.in terms of resource usage.
 
Not saying it isn't an improvement though. I think nvidia would be tickled pink if they could get get cuda logic on their gpu but I have a feeling it requires way too much power.

With cuda or opencl their could be a ton of acceleration from the gpu, more so then what ios gets from the A6.
 
No matter how efficient the power draw of. X86 cpu becomes in the future, you can't make up for how bloated and upoorly optimized x86 apps are.in terms of resource usage.

Cant look at the normal desktop app market, these small atoms will for sure get their fair share of the "applet" market which will be hacked down versions of the same programs you are talking about.
 
Not saying it isn't an improvement though. I think nvidia would be tickled pink if they could get get cuda logic on their gpu but I have a feeling it requires way too much power.

With cuda or opencl their could be a ton of acceleration from the gpu, more so then what ios gets from the A6.

Amazon seems interested in Open CL
http://www.youtube.com/watch?v=-twOwM4LP9o

Newer Adrenos support it so maybe in the future we would see it being implemented in mobile.
 
In fact it was bloody good. I dont other chipsets being much better if any than the Tegra 4.
Really? Every major chip apart from Tegra has universal shaders, Open CL, decent GPGPU support and Tegra is missing OES3.0 features llike being limited to FP20. Beasicly the Tegra 4 GPU is way behind the times. Tegra 4 GPU looks to not only be the slowest of the new 2013 chips but it looks to be lacking in basic features.
 
Completely disproving the Intel cannot compete myth:
http://m.seekingalpha.com/article/1104561

P.S. NVIDIA representative said: "Kepler isn't in Tegra 4, but Tegra 4's GPU leverages multiple GeForce architectures and elements - including Kepler and others." - 2013-01-09

I dunno why you insists on saying that it is.
 
Really? Every major chip apart from Tegra has universal shaders, Open CL, decent GPGPU support and Tegra is missing OES3.0 features llike being limited to FP20. Beasicly the Tegra 4 GPU is way behind the times. Tegra 4 GPU looks to not only be the slowest of the new 2013 chips but it looks to be lacking in basic features.

Open CL is great to have but I somehow don't see them making using of this tech within this upcoming generation of GPU's.
I hope I am proven wrong!
 
Open CL is great to have but I somehow don't see them making using of this tech within this upcoming generation of GPU's.
I hope I am proven wrong!

Android has support for GPU compute, but it is done through Renderscript as opposed to OpenCL or CUDA.
 
My guess is Tegra 4 will once again be the most underhwelming mobile SoC. Over promise and under deliver is practically Nvidia's motto.

It'll be interesting to see if Intel can make an impact. They have the fab and the R&D to basically out-engineer any performance gains on ARM by just moving to a lower process fab, and they are already matching the power draw. It'll come down to price, the license fees Intel will charge, and whether the big players like Samsung and Apple will continue to favor their own ARM designs or pick Intel once the cost/performance ration becomes favorable.
 
The lack of integrated baseband hurts Tegra compared to Qualcomm. Tegra 3 had a lot of decent tablet wins but not much in the phone market IIRC (except the One X I guess).
 
The lack of integrated baseband hurts Tegra compared to Qualcomm. Tegra 3 had a lot of decent tablet wins but not much in the phone market IIRC (except the One X I guess).

Because qualcomm owns it, hard to come out with your own radio without trading on patents.
 
Back
Top