Nvidia Tegra X1

[jdubs03]

I figured I would start this off since it was inevitable. The first leak for a Tegra X1 system has been leaked on Antutu. Take it with a grain of salt, but it looks good.

http://blog.gsmarena.com/nvidia-tegra-x1-flies-past-70000-antutu/

~75000, not too bad. ~23% faster than the Exynos 7420 in that score. For a 20nm implementation, whew, let's roll out some more info.

 

[Roland00Address]

Its antutu.

 

[jdubs03]

Yea I know. But its something. At least the score can be compared to the K1.

 

[NTMBK]

Not that impressive for a 10W part.

 

[Qwertilot]

Presume they're devoting a lot of their power budget to the GPU?

 

[dawheat]

Hate to say it, but are there enough high end Android tablet sales that are non-Samsung to make this a commercial success? The time seems ripe for a high end smartphone SOC to take advantage of the S810 stumbling as other OEMs probably would be very open to a 7420 competitor. Nvidia never seems to have the right timing here.

 

[oobydoobydoo]

My ipad scores ~65,000 in antutu, so I'm not sure why 75,000 would be impressive. If anything it's underwhelming considering the next ipad will probably beat the last one by at least 40%, which already would put it way higher than X1.



I think nvidia should concentrate on the GTX 970 fiasco and all the other controversies surrounding almost every single product they make. Why people are so enamored with them... I will never understand. And I've owned nvidia video cards, I recall them being basically comparable to AMD, but more expensive. Not seeing the appeal.

 

[jdubs03]

Not that impressive for a 10W part.

This got me thinking. I think the 10W TDP is mainly for GPU load, at least that is what I inferred from reports.

But, even if we use that 10W number, we still can compare that to the score that the 15W i5-4200U produced when it was tested in the ATIV Q, which was 54,861. So, at least here ~37% higher.

http://www.phonearena.com/news/Samsung-ATIV-Q-AnTuTu-score-shatters-all-records-Core-i5-benchmarked_id44433

So, there is another comparison point, even though it isn't the greatest of benchmarks. It does however, make it seem a little bit better.

 

[xpea]

Not that impressive for a 10W part.

and what other 10W part produces better performance ?

 

[gdansk]

Eh, I'd like to see sunspider and linpack. Wonder what clock rate the CPU is at. The GPU score is just about the same as the Exynos 7420, not impressive. But it'll put out higher frame rates in games, Nvidia usually has reasonable drivers unless most SoC vendors.

 

[jdubs03]

How about this for competition:

http://vr-zone.com/articles/imagination-technologies-reveals-powervr-series-7xt-gpus-claimed-1-6-tflops-fp16-mode/87677.html
http://blog.imgtec.com/powervr/powervr-gt7900-redefining-performance-efficiency

Seems like they've stepped up their game. Makes the 6XT series seem tame in comparison. It's aiming at 50% higher performance than the X1 GPU, it makes me wonder why not try a higher wattage level GPU? We won't see benches for awhile but we'll see eventually each GPU vendor on the same process, who will come out on top?

 

[xpea]

How about this for competition:

http://vr-zone.com/articles/imagination-technologies-reveals-powervr-series-7xt-gpus-claimed-1-6-tflops-fp16-mode/87677.html
http://blog.imgtec.com/powervr/powervr-gt7900-redefining-performance-efficiency

Seems like they've stepped up their game. Makes the 6XT series seem tame in comparison. It's aiming at 50% higher performance than the X1 GPU, it makes me wonder why not try a higher wattage level GPU? We won't see benches for awhile but we'll see eventually each GPU vendor on the same process, who will come out on top?

looks like a good improvement. 2 remarks:
- This part is intended to gaming console. Event at 14/16nm FinFet, now way it will fit on a phone (This GT7900 will be find in ~10W SoCs)
- next year, at same node, the competition will be Parker with Pascal generation GPU. Pascal, as a totally new uarch and new manufacturing process, will be a much biggest leap in performance than what Maxwell/Kepler bring to us. I bet for >2TFLOPs at FP16 with higher sustained throughput than Maxwell.

interesting times...

 

[Mondozei]

Hate to say it, but are there enough high end Android tablet sales that are non-Samsung to make this a commercial success? The time seems ripe for a high end smartphone SOC to take advantage of the S810 stumbling as other OEMs probably would be very open to a 7420 competitor. Nvidia never seems to have the right timing here.

That's already the case. Mediatek has released a high-end SoC that crushes the S810.

 

[jdubs03]

looks like a good improvement. 2 remarks:
- This part is intended to gaming console. Event at 14/16nm FinFet, now way it will fit on a phone (This GT7900 will be find in ~10W SoCs)
- next year, at same node, the competition will be Parker with Pascal generation GPU. Pascal, as a totally new uarch and new manufacturing process, will be a much biggest leap in performance than what Maxwell/Kepler bring to us. I bet for >2TFLOPs at FP16 with higher sustained throughput than Maxwell.

interesting times...

If the gt7900 is 10w then its > than x1 perf which has a tdp of 10w, but the gt7800 seems to be around the same perf as the x1, so i guess we'll see about the perf for each range, it definitely helps its on 16ff+ vs the x1 at 20nm pln. i agree with the pascal uarch, it should be a leadership gpu at all tdp levels.

 

[dark zero]

Sorry for necroposting, but... there is not new info about this little beast?

I mean, they destoy all the current x86 Atoms in a normal day, but... they could get better with DX12?

 

[NTMBK]

Sorry for necroposting, but... there is not new info about this little beast?

I mean, they destoy all the current x86 Atoms in a normal day, but... they could get better with DX12?

Shield console is out soon.

 

[dark zero]

I am not refering the console itself. But the Mobo just like nVIDIA Tegra K1 (who people managed to launch Gamecube apps)

 

[tweakboy]

This is not meant as a gaming card. gb

 

[tweakboy]

This is for Studio Max, Maya, Premiere etc...

 

[monstercameron]

I am not refering the console itself. But the Mobo just like nVIDIA Tegra K1 (who people managed to launch Gamecube apps)

Shield console out soon.

 

[Systems analyst]

According to ARM, the A72 core at 16nm scaling will be available in 2016, and will have almost twice the performance of the A57 at 20nm. In larger form-factors, the clock-speed will go up to 3 GHz. This predicts rapid performance increases, in affordable SoCs, and may find its way into convertibles and notebooks, with the advent of Windows 10.

 

[NTMBK]

This predicts rapid performance increases, in affordable SoCs, and may find its way into convertibles and notebooks, with the advent of Windows 10.

Not a chance. With Windows 10, Microsoft is limiting use of ARM processors to tablets which are 7" and less, and mobile phones- and both of those platforms will not run desktop applications. >8" tablets, laptops, and desktops will all use x86.

 

[Systems analyst]

Not a chance. With Windows 10, Microsoft is limiting use of ARM processors to tablets which are 7" and less, and mobile phones- and both of those platforms will not run desktop applications. >8" tablets, laptops, and desktops will all use x86.

Satya Nadella on the subject of cross-platform applications:

"We absolutely are going to have our services and their application end points everywhere. But we absolutely believe that Windows is home for the very best of Microsoft experiences."

Source:
http://news.microsoft.com/speeches/...elfiore-and-phil-spencer-windows-10-briefing/
Posted January 21, 2015 By Microsoft News Center

Can this exclude convertibles, ultrabooks and laptops sporting an ARM chip?
If you know of a categorical denial of availablility of Windows 10 on larger form-factor devices with ARM processors, I would be interested to have a reference.

 

[dark zero]

If MS denies ARM, they will stick in everything X86, losing market and the ARM comunity who will stick on stay on Android ARM.

 

[dark zero]

http://www.zdnet.com/article/google-testing-chrome-os-with-nvidias-tegra-x1-super-chip/

Seems that Google is trying the Tegra X1... why nVIDIA is not giving more for their chip? Also they shuld leave big LITTLE in create their own custom chips