[The Inq/THW] Nvidia Tegra K1 smashes Apple and Qualcomm in early benchmarks

[OCGuy]

EARLY BENCHMARKS are emerging that show that the recently unveiled Nvidia Tegra K1 system on chip (SoC) processor seems be the fastest thing since greased lightning

Established tinkering website Tom's Hardware was given access to a Lenovo Thinkvision 28 all in one (AIO) system that had a Tegra K1 prototype in it, and found that it outperformed everything, not only in its class, but any portable machines put up against it.

http://www.theinquirer.net/inquirer...mashes-apple-and-qualcomm-in-early-benchmarks


CES 2014 was definitely a positive thing for NVidia and possibly the whole mobile graphics market in general. Will be interesting to see it play out!

Graphics Benchmarks from Tom's:

http://www.tomshardware.com/news/le...33.html?&_suid=138964069136109157009874582127

 

[Grooveriding]

So an unreleased next-gen prototype SoC is faster than current-gen SoCs from the big players in benchmarks ? Typical nvidia hype, marketing next-gen against competitor's current-gen. :thumbsdown: They did this same stunt with Tegra 3 and Tegra 4, doesn't bode well for K1 if it winds up like those two did.

Given nvidia's abysmal track record on Tegra getting to market according to promised timelines, we could very well not see this in devices until the end of the year. Even then who knows what devices as Tegra has consistently failed to score many design wins.

Wake us up when we see K1 benched against Apple & Qualcomm's next-gen. For a hint on results just look at those benchmarks ignoring K1 and paying attention to Tegra 4 vs SD800 vs A7.

/empty hype

 

[OCGuy]

Personally I have to agree with the people who used Lenovo's machine. Impressive that it was a 4K monitor, with the potential to be extremely impressive all around.

From an Anandtech article:

For the first time I’m really excited about an NVIDIA SoC. It took the company five generations to get here, but we finally have an example of NVIDIA doing what it’s really good at (making high performance GPUs) in mobile. NVIDIA will surely update its Tegra Note 7 to a Tegra K1 version (most of its demos were run in a Tegra Note 7 chassis), but even if that and Shield are the best we get the impact on the rest of the market will be huge. With Tegra K1, NVIDIA really raised the bar on the GPU performance.

http://www.anandtech.com/show/7622/nvidia-tegra-k1/5

 

[Imouto]

I remember the very same story last year with Tegra 4.

Overpromise.
Underdeliver.
Be late.
Go missing.

Could perfectly be the Tegra brand slogan.

 

[SolMiester]

wow, tough crowd.......

 

[Arkadrel]

Looks like its around ~30% faster than the "Snapdragon 800" in
3Dmark - Icestorm unlimited.

Would be intresting to see how it matches up with power consumption.

 

[AnandThenMan]

NV has quite a history of hyping Tegra then failing to deliver on promises. Will take a wait and see approach again.

 

[Skurge]

Hey. Let's give nvidia the benefit of the doubt. Its not like they never over promise or anything.

 

[cmdrdredd]

As if no other company in history gave a slideshow or presentation and failed to deliver that product.

 

[videogames101]

As if no other company in history gave a slideshow or presentation and failed to deliver that product.

Implying the same cynicism isn't present in every. other. thread. in. this. forum.

 

[Venomous]

You know I waited and waited and waited for tegra 3 to drop thinking is was going to smash snapdragon... When it finally arrived, it was late and a failure.

Until they can prove otherwise from other marketing stunts in the past, I'll pass on the hype. I'm sure there are many others who feel the same way as I do.

 

[blackened23]

Uh, tegra 3 was a very successful product. Maybe you're thinking T4. T4's woes were due to the lack of integrated LTE, and qualcomm beat everyone in that respect. Not just nvidia. And yeah T4 was late. I think nvidia would have learned from those mistakes - we'll see with the K1 I suppose.

 

[Silverforce11]

Why are they comparing a massive tablet AIO thing versus tiny smartphones that are obviously operating under TDP/Heat contraints?

 

[Enigmoid]

Personally I have to agree with the people who used Lenovo's machine. Impressive that it was a 4K monitor, with the potential to be extremely impressive all around.

From an Anandtech article:

http://www.anandtech.com/show/7622/nvidia-tegra-k1/5

If the model being showcased is going to be priced at $1000 it is extremely impressive; 4K for your desktop on a monitor that is perfectly usable on its own.

Looks like its around ~30% faster than the "Snapdragon 800" in
3Dmark - Icestorm unlimited.

Would be intresting to see how it matches up with power consumption.

And more than 100% in T-Rex HD. 3dmark is most likely bandwidth limited.

Why are they comparing a massive tablet AIO thing versus tiny smartphones that are obviously operating under TDP/Heat contraints?

Cause its the only thing they have and many smartphones maintain the turbo long enough for short benchmarks to take place?

 

[AnandThenMan]

We need to see perf/watt in a broad array of workloads. Perf/watt is everything (in Jensen's own words) and I agree it is.

 

[DarkKnightDude]

So a chip in a massive tablet that is plugged in outperforms smartphone chips from six months ago?

 

[NTMBK]

Uh, tegra 3 was a very successful product. Maybe you're thinking T4. T4's woes were due to the lack of integrated LTE, and qualcomm beat everyone in that respect. Not just nvidia. And yeah T4 was late. I think nvidia would have learned from those mistakes - we'll see with the K1 I suppose.

NVidia lost a lot of money on Tegra 3, same as Tegra 1, 2 and 4. They only shifted so many because they sold them to Google at a knockdown price for the Nexus 7.

 

[Arkadrel]

And more than 100% in T-Rex HD. 3dmark is most likely bandwidth limited.

Whats weird is how slow the K1 is whenever your not hooking your smart phone/tablet up to a monitor.

I wonder why that is? does it draw extra power when hooked up or something (to get more oomf)? consume more? why does it go from 16 fps (when not useing a monitor) to 48 when useing a monitor with it?


If you look at the "on-screen" only the K1 is going 16 fps, which is the lowest of the bunch there shown.

So a chip in a massive tablet that is plugged in outperforms smartphone chips from six months ago?

My guess too.
It has some insane turbo funktion when its not power constrained by the small battery of a laptop/tablet.

 

[red12355]

Not really a fair comparison... The Lenovo likely has far better cooling than the other devices in the comparison.

 

[Silverforce11]

Whats weird is how slow the K1 is whenever your not hooking your smart phone/tablet up to a monitor.

I wonder why that is? does it draw extra power when hooked up or something (to get more oomf)? consume more? why does it go from 16 fps (when not useing a monitor) to 48 when useing a monitor with it?

If you look at the "on-screen" only the K1 is going 16 fps, which is the lowest of the bunch there shown.

It has some insane turbo funktion when its not power constrained by the small battery of a laptop/tablet.

Yup, the others drop performance when they have to drive a monitor, as expected, this thing speeds up 3 times.. which means there is a very high boost function available for high end models that can cope with the heat output. Nobody wants a smartphone that burns your hands.. so it will be interesting to see the real result, perf/w.

 

[NTMBK]

Whats weird is how slow the K1 is whenever your not hooking your smart phone/tablet up to a monitor.

I wonder why that is? does it draw extra power when hooked up or something (to get more oomf)? consume more? why does it go from 16 fps (when not useing a monitor) to 48 when useing a monitor with it?


If you look at the "on-screen" only the K1 is going 16 fps, which is the lowest of the bunch there shown.

"On-screen" means that the GPU is running the benchmark at the device's native resolution- the K1 is hooked up to a massive 4K display, so its framerate plummets. "Off-screen" means that they GPU is rendering to a fixed size resolution in an off-screen buffer, probably 1080p. (The Nexus 5 has almost the same score for on- and off-screen, and has a 1080p display.) Off-screen is the fair comparison between SOCs, while "on-screen" is meant to give you an idea of how well balanced a given device is.

 

[NTMBK]

Yup, the others drop performance when they have to drive a monitor, as expected, this thing speeds up 3 times.. which means there is a very high boost function available for high end models that can cope with the heat output. Nobody wants a smartphone that burns your hands.. so it will be interesting to see the real result, perf/w.

It's driving a 4K monitor...

 

[krumme]

There actually was a time when nv didnt release bm bs to sites like thg long before the products was available on the market.

The product is very interesting but the marketing is unworthy and pathetic.

 

[Galatian]

Whats weird is how slow the K1 is whenever your not hooking your smart phone/tablet up to a monitor.

I wonder why that is? does it draw extra power when hooked up or something (to get more oomf)? consume more? why does it go from 16 fps (when not useing a monitor) to 48 when useing a monitor with it?


If you look at the "on-screen" only the K1 is going 16 fps, which is the lowest of the bunch there shown.

Off-Screen means it renders on an internal resolution of the benchmark on all devices (e.g. 1650x1050) so you can compare the respective power of the chip

On-Screen means it is using the internal screen of the tablet/smartphone to show how good the chips actually is to power a given device. Of course using a high resolution tablet screen (e.g. 2560x1600) needs more power then running a smartphone screen (e.g 1024x768).

(Numbers are just made up to explain)

Edit: ninjaed

 

[Silverforce11]

It's driving a 4K monitor...

Ok, I understood it now, thanks!

Those are impressive numbers, but we will see when there are phone vs phone comparisons.