Tegra 3 officially announced; in tablets by August, smartphones by Christmas

Well if you clock a GPU at 500MHz results like that aren't really a surprise. If Samsung as just one example clocks the Mali400MP4 under 32nm at 400MHz as projected, T3 won't have it easy.

 

I wonder how comparison between Adreno 225, Mali400MP4, T3 and SGX544 will look like.

 

I'm not sure but I wouldn't suggest that T30 has the same frequencies as AP30; the latter should be slightly lower (T20 GPU was at 333MHz and AP20 at 300MHz if memory serves well). Most others you mention will appear mostly (I suppose) in smart-phones. I wouldn't suggest that the differences between all those would be signficant in terms of GL benchmark performance. Exynos@32nm might lead the pack if its truly clocked at 400MHz after all.

 

Tegra in next Kindle Fire?

http://www.forbes.com/sites/ericsav...cessor-slot-in-next-kindle-fire-analyst-says/

 

Well Engadget gave the Transformer Prime a rave review. It got battery life almost as good as the iPad 2.

 

I would've picked one up but there are devices coming out with high-resolution (1920x1200 or so) screens that I'm very interested in.

 

Correct: peak 3D clock in T30 is 520 MHz, whereas peak in AP30 is 416 MHz.

 

Anand seems to like it as well, though he doesn't seem enamoured by ASUS's stupidity.

"I am comfortable calling the Eee Pad Transformer Prime the absolute best Android tablet on the market today."

http://www.anandtech.com/show/5163/asus-eee-pad-transformer-prime-nvidia-tegra-3-review/

 

Yeah but how long will that last before something better than the Tegra 3 on the Android side comes out?

 

Better than quad-core Cortex A9 @ 1.5GHz?
On the CPU side, I'd say quite a while, nearly half a year... Even OMAP4470 has a slighly higher-clocked dual core...

nVidia is on the road to get most Android tablet design wins, again.

 

Most design wins isn't necessarily a useful metric. There's really a very few key ones.

 

Krait would be one example, which should appear in early 2012.

An almost 40% higher peak frequency equals "slightly"? Have someone count how often single, dual and quad threading appears in today's embedded applications and then draw an average.

With how much volume per deal exactly? We've had that debate again but the number of design wins vs. a total of units sold is quite a difference. The first can create mostly impressions the second real revenue.

I wouldn't suggest that Tegra3 won't sell better than Tegra2 for tablets, but the problem remains that T2 didn't make any significant breakthrough in the smart-phone segment where the real high volumes are. trinibwoy posted a link above with some analyst already granting NV the next Amazon tablet deal. Considering what will be available at that timeframe (Krait, OMAP5, Exynos 5250 etc.), I'd say Amazon would be mighty stupid to pick something like Tegra3 over the others. And yes that's a real high volume deal. Kindle Fire for this year is estimated to reach 12Mio units.

I didn't see Jensen being enthusiastic about their Tegra2 overall business since they obviously didn't reach their targets. The expectations for T3 I've heard in the background is that T3 will gain in just one quarter more revenue than T2 all together. Who's daydreaming again?

 

Amazon has the luxury to define their own market segment, where pure performance comparisons with other tablets are not all that important. Consumers don't care. I don't know the difference between their current SOC and T3, but if it's simply large enough to feel a difference in day to day usage, that should be sufficient.
They don't need to have the absolute winner of the day.

 

Reviews say the Fire isn't that smooth. But if it sells well, especially after Xmas, then Amazon has no need to chase high-end SOCs, do they? They are trying to hit a $200 price target so they can stay low-end, especially if the customers who buy the Fire have low expectations and don't care.

 

nVidia posted an interesting paper that I somehow missed earlier:

http://www.nvidia.com/content/PDF/tegra_white_papers/tegra-whitepaper-0911b.pdf

(sorry if this was already linked here, I couldn't find anything)

Here nVidia makes the claim that Tegra 3 uses less power running all four of its cores at 1GHz than OMAP4 uses running its two cores at 1GHz. If this is fair and true to any extent whatsoever then it's pretty interesting: the LPG arrangement is shown as being useful for allowing low idle power, but everyone else was using LP cores already at much higher clockspeeds than nVidia. But this would mean nVidia's G cores have much higher perf/W at similar clock speeds vs the competition's LP cores, which is what you'd expect.

Makes me wonder if there really will be LPG arrangements of big.LITTLE in the future.

 

Wouldn't it be nice if there was an in-depth article currently being written on big.LITTLE and process technology trade-offs including future implications... (unfortunately getting delayed a bit by something else, but hopefully will be published by the end of this month anyway).

 

Well, there are other factors at work. Amazon also competes with itself and may want to appeal to a broader audience (look at how many different kindles there are.... 8 not including Fire). I can say I haven't purchased a tablet yet because I was waiting for the Fire, but it ultimately wasn't the device for me. If Amazon were to release a T3 based Fire XL for $399 or less, it would be pretty compelling.

 

But do we really think that this is because of the hardware? It's maddening that you can't get smooth operation with 2 fck'ing 1GHz+ cores with the only fix seemingly throwing more hardware at it when even the first iPhone was pretty good at it. ICS seems to be better at it: I hope it's not only because of better HW but also because of software efficiency fixes. It will be interesting to see how it performs on older 2 core phones/tablets.

Absolutely. Maybe there a Tegra 3 with only 1 memory controller will have an cost advantage?

 

IMHO, the smoothness of GUI should be entirely the SW's fault. iPhone showed that a lowly MBX could deliver the goods if you made the effort. There's no reason for 4430 to not deliver a smooth GUI barring poor sw investment.

 

ya playbook is incredibly smooth running basically the same internals as the fire