but just going from 0.8V to 1V would bump that up to ~3.9W. (For comparison, desktop Kepler goes from something like 0.875V to 1.125V I believe it is?)
But at max clocks under full load 5W is not going to cut it. That is easily seen in a lot of devices.
4x A15 @1.8 ghz + GPU at full load is over 5W.
Tegra 4 had about 3000 multi-thread on geekbench 3. Their whitepaper has K1v1 at 1.4x performance. So, thats 4200+ multi thread. Denver has 2 cores, so we're already at least 2100 points for single-thread (I said near 2,000 in my last post, and assuming better multi-thread scaling).
Thanks for providing a source - http://www.nvidia.com/content/PDF/tegra_white_papers/Tegra-K1-whitepaper-v1.0.pdf for those who don't want to search for it, first paragraph of page 10. And that definitely explains the discrepancy seeing as how I'd be surprised if an 'average of populate Android games' even gets the GPU up to a 50% load. Especially if that's run on an equivalent to the Tegra Note 7 at only 1280x800 resolution with a 60 fps cap.
Denver is a massive upgrade, I think it will easily be competitive with Airmont/Goldmont CPU wise. It could be competitive with i3-U and Broadwell-i5/i7-Y. Denver will likely have higher multi-thread peformance than K1v1 (A15r3).
Tegra 4 had about 3000 multi-thread on geekbench 3. Their whitepaper has K1v1 at 1.4x performance. So, thats 4200+ multi thread. Denver has 2 cores, so we're already at least 2100 points for single-thread (I said near 2,000 in my last post, and assuming better multi-thread scaling).
i5-4200U is ~2500, 4300U is about ~2700.
And if that one GFXbench result is accurate to true GPU performance, then we're seeing HD4400 level performance with 3x less TDP.
All this on 28nm is pretty incredible, imagine 16nm FinFET (though 20nm BEOL) will give an additional 35% perf (according to TSMC), that is 4500U perf. territory.
A lot of ifs, but Denver is a massive jump in terms of CPU and GPU compared to past iterations.
1 Cyclone core consumes about as much energy as 3 Silvermont cores. Single threaded performance is only 40% faster -> performance/watt of Cyclone is 2x worse. It would surprise me if Denver would be any different.What makes you say that A7 [Cyclone] CPU consumes "tons of energy"? A7 seems to work fine in the relatively small and thin iphone 5s (albeit with some throttling compared to the variant in ipad Air).
The Tegra Note 7 prototype chassis's used to demonstrate Tegra K1 had a 1080p screen resolution.
NVIDIA's statement that K1's GPU consumes an average of 2W playing popular Android games doesn't tell us anything about actual load power consumption.
Well, to fit inside a thin fanless tablet, the TK1 GPU peak power consumption really cannot be much more than ~ 4w (and this level of consumption from the GPU would require an extremely GPU-intensive application to be running).
It would indeed be limited to 4W in a fanless tablet. But is that with the K1 throttling? If so, how badly? And how much power would it consume if running at full speed? That's the question. It's a lot trickier to get a straight answer (from anyone!) these days, with configurable TDPs, SDPs, throttling, turbo...
950mhz on a 192 Cuda Core GPU is way overly optimistic for 28nm. Nvidias 384 core mobile GPUs can't clock that high and they have a 35w tdp to work with. In any application that people will actually buy(phone or tablet) we're never going to see these peak clocks in practice.
Sure, if you think everything needed to run two SMX together will be required to run one only.
I think Anand covered this very subject.
19W seems remarkably low. I can't find any concrete info but most search results calling the 740m a 19W part come from the past few days. Prior to that most people were citing a 33W tdp.If you take a GeForce 740M (2 SMX/384 CUDA core design), youre looking at roughly a 19W GPU. Of that 19W, around 3W is memory IO, PCIe and other non-GPU things. You can subtract another 6W for leakage, bringing you down to 10W. Now thats a 2 SMX design, so divide it in half and now youre down to 5W. Drop the clock from 1GHz down to 900MHz, and the voltage as well, and now were talking around 2 - 3W for the GPU core and thats without any re-architecting. Granted you cant just subtract out things like leakage like that, but you get the point. Kepler wasnt a bad starting point for a good mobile GPU design.
Nonsense. TDP for Tegra 4 is 5W.
19W seems remarkably low. I can't find any concrete info but most search results calling the 740m a 19W part come from the past few days. Prior to that most people were citing a 33W tdp.
I think Nvidia is pulling the wool over our eyes.
...
Look at AMD's kabini chips. 128 gcn cores at max 600 mhz on a much larger power envelope. I simply do not see 192 kepler cores at 950 mhz using 4W under load.
950mhz on a 192 Cuda Core GPU is way overly optimistic for 28nm. Nvidias 384 core mobile GPUs can't clock that high and they have a 35w tdp to work with..
I never doubted that a 650m can physically run at 1100MHz if the cooling was sufficient. But at the stock clocks and tdp(a whopping 45W for the 650m) it does not.Yes they can, I run my 650M at 1100MHz stable with okay thermals..
I'm a bit less optimistic. Denver is Nvidia's first CPU, I do not expect it at 28nm to be a threat to 22nm or 14nm Tri-Gate transistors.
It might be somewhat competitive with Silvermont, but certainly not Airmont, let alone Goldmont which will be Denver's main competitor .
Look at AMD's kabini chips. 128 gcn cores at max 600 mhz on a much larger power envelope. I simply do not see 192 kepler cores at 950 mhz using 4W under load.
unless I missed it ut wasn't the 5W figure an estimate for gles3 android games? or total chip power?