Review Index:
Feedback

NVIDIA Tegra K1 - Kepler meets ARM, Project Denver Surfaces

Author: Ryan Shrout
Subject: Mobile
Manufacturer: NVIDIA

CPU Design and Kepler in Your Pocket

The processor design of the Tegra K1 is very similar to that of the Tegra 4 in its primary form. There are four A15 cores running at a maximum clock speed of 2.3 GHz with a fifth A15 core at a maximum of 1.0 GHz (it is typically in the 500 MHz range I’m told) used during idle and low performance scenarios to save on power. 

The engineers have built on three key areas to improve the SoC.  First, and most importantly, the chip benefits from the experience that NVIDIA built from using the A15 cores with the Tegra 4. As with any microprocessor design, you improve performance significantly the second time around. 

Tegra K1 also benefits from the updated 28HPM process node from TSMC that combines the benefits of higher performance and lower power silicon.  Finally, the latest revision of the Cortex-A15 “r3” was used with the K1 which adds ARM-developed and integrated architectural power reductions. 

View Full Size

All three of these components combine to allow the Tegra K1 to run about 30% “better” than Tegra 4.  As with all mobile processors you can target two different, but equally important metrics when analyzing performance.  K1 will see 1.4x the performance at the same power draw of Tegra 4 or utilize about half the power to get to the same performance levels of the Tegra 4. 

View Full Size

NVIDIA did offer up a comparison of Octane performance and power draw on Tegra K1 against Qualcomm’s S800 SoC with Krait 400 graphics and the results look impressive.  The single white point on the graph represents the Apple A7 SoC with the Cyclone graphics core nearly matching the performance of Tegra K1 at 2 watts power draw – the maximum level for each part more than likely.  Without a complete power/performance curve, the comparison to Apple’s product will have to wait until we have K1 product in our hands.

View Full Size

All of this data was presented on the 32-bit, A15-based verions of the Tegra K1.  The 64-bit version that includes the NVIDIA Denver cores and will obviously change the game pretty dramatically.  From the information provided at NVIDIA's press conference the Denver cores run 200 MHz faster at the maximum clock rate and the CPU has a 7-way superscalar design which should make it very efficient.  We don't yet have architectural details or even performance estimates of Denver; that analysis will have to wait until later in the year.

Kepler in your Pocket

By far the most impressive part of Tegra K1 is the implementation of a full Kepler SMX onto a chip that will be running well under 2 watts.  While it has been the plan from NVIDIA to merge the primary GPU architectures between mobile and discrete, this choice did not come without some risk.  When the company was building the first Tegra part it basically had to make a hedge on where the world of mobile technology would be in 2015.  NVIDIA might have continued to evolve and change the initial GPU IP that was used in Tegra 1, adding feature support and increasing the required die area to improve overall GPU performance, but instead they opted to position a “merge point” with Kepler in 2014.  The team at NVIDIA saw that they were within reach of the discontinuity point we are seeing today with Tegra K1, but in truth they had to suffer through the first iterations of Tegra GPU designs that they knew were inferior to the design coming with Kepler.

View Full Size

Going forward, all future GPU architectures will be built and designed with mobile integration in mind.  NVIDIA’s Jonah Alben, SVP of GPU Engineering, described it as being part of the “bones” of the design team.  Very early in the design of Kepler, NVIDIA’s key minds decided that this was the direction for the company – and this isn’t without its own risks.  Maxwell, NVIDIA’s upcoming architecture due in desktop systems this year, will be the first design that was truly and completely built with Tegra as one of the targets.

View Full Size

If they can make it work, NVIDIA’s graphics IP will be scaling from milliwatts to megawatts and from phones to HPC server racks.  Alben seems confident that this does not require a sacrifice to the high end in favor of the low end – “it can be built perfectly for everyone” was a common thread amongst discussions.  If you consider the benefits that the GeForce line has had since the power efficiency improvements of Kepler were introduced, it is easy to see how this might actually work out in favor of each market segment.

January 6, 2014 | 07:10 AM - Posted by Anonymous (not verified)

Well with Nvidia GPU joining their previously separate GPU technology between mobile ad the desktop, and entering the exclusive Top Tier ARM ISA custom design club, with Apple and others, it should not be to difficult to estimate what Maxwell will bring to the table. This merging the desktop GPU with some on die CPU cores, and maybe a large on die RAM, should begin the move towards less reliance on the moatherboard CPU. Gaming engines and other latency/bandwith constrained code will now run, and hopefully reside in a large on die RAM, to reduce these latency/bandwith issues between gaming engine code and the GPU. This puts the relevance on the motherbard CPU into question, with repect to descrete GPUs possessing their own complete gaming system ability.

January 6, 2014 | 08:06 AM - Posted by Angry

I don't think you'll be seeing the x86 CPU going the way to the dodo bird anytime soon....at least not for awhile.

January 6, 2014 | 09:27 AM - Posted by Anonymous (not verified)

For sure x86 will never completely go away, AMD will be doing the very same thing with its own ARM based APUs as Nvidia's K1, but AMD will also be taking the x86 ISA on board with the descrete GPUs for some CPU/GPU accelerated complete gaming platform capable descrete GPUs, via AMD's already deveoped for the gaming consoles x86 based technology! Both Nvidia's descrete Maxwell GPUs and AMD's future descrete GPUs will merge the CPU with the GPU, and by themselves, become complete gaming platforms on a PCI card.

January 6, 2014 | 08:42 AM - Posted by Willmore (not verified)

In the paragraph at the end of the page on GPU-Specifications, you misuse V where you mean W.

January 6, 2014 | 10:01 AM - Posted by Anonymous (not verified)

Shield 2 or 3

January 6, 2014 | 10:28 AM - Posted by aparsh335i (not verified)

I feel like this could make for a perfect Steam OS box.

January 6, 2014 | 02:13 PM - Posted by Scott Michaud

SteamOS is developed for x86(-64) and its games will be, too. It could be a good FirefoxOS console (or whatever).

January 6, 2014 | 11:04 AM - Posted by Anonymous (not verified)

Can`t wait !

January 6, 2014 | 11:05 AM - Posted by Anonymous (not verified)

Tegra 4 repeat ?

Samsung & Qualcomm already announced their 64bit chips will be coming out for the new Phone/Tablet season last month. Both have already been leaked to be in phones already by Spring.

Unless Nvidia sells K1 32bit cheap to make it attractive I don't see how it can gain traction much like Tegra 4 was overpriced and its modem wasn't certified so it was a no go for phones or tablets that used cell service.

January 6, 2014 | 01:23 PM - Posted by aparsh335i (not verified)

Did you read the article at all? Any of it? These new SoCs will have powerful graphics embedded, and it seems that the graphics are powerful enough to be on par with PS3/Xbox 360. Potentially this could lead to being on par with the Xbox 1 & PS4 within a few years. Now that is exciting.

January 6, 2014 | 01:56 PM - Posted by Anonymous (not verified)

Cant believe you never mentioned once Tegra K1 will lack native on-chip support for LTE.

January 6, 2014 | 07:16 PM - Posted by Anonymous (not verified)

Its also doesn't support CDMA.

So it won't work on Verizon nor Sprint networks in the U.S.

January 6, 2014 | 02:29 PM - Posted by fury (not verified)

This is pretty compelling stuff. As you say, it all depends on if they get any design wins. But for my 2 cents, I'd probably buy a phone or tablet with the Tegra K1, assuming it comes out before it gets leapfrogged by the next Adreno or Apple A8.

PCI Express capability is interesting. Does that mean this chip could potentially run Thunderbolt? Might do interesting things for accessory connectivity.

I like the comparison in raw compute power with last gen consoles. At the rate things are going, we're going to catch up with current gen consoles before next gen consoles come out.

One of the other big things stopping developers from coming out with real, true-to-life console quality games for mobile chips is the lack of a standard controller. Bluetooth HID controllers, you can have a very different set of buttons on each one, so there is a barrier to entry--both to the developer who would have to try and make their game configurable enough that a wide variety of controllers is usable, and to the user who has to go and do that setup and may end up failing to get a good, workable configuration. Consoles have a single defined set of buttons, a single set of hardware, and that means the developer knows exactly what to design for.

So, even if Tegra K1 takes off, we may still have yet to see a lot of heavy hitting games put onto mobile platforms, unless someone comes along and makes a big push for a single controller definition. Apple, as a matter of fact, did this for iOS, so maybe that style of controller will become the controller for iOS, and spill over into the rest of the world, so that maybe there's one big one that all the game developers design for, and the rest of the controllers can either follow suit or fall behind.

January 6, 2014 | 07:43 PM - Posted by Anonymous (not verified)

The rumor mill is already placing the Apple A8 in some sort of sub-macbook air form factor with a full keyboard and running OS X and iOS. I'm looking for a poor man's version of those expensive professional graphics tablets, running the Nvidia K1 Denver cores, with at least a 10-12 inch HD screen, and running Linux Mint, for my Gimp graphics and Light Blender 3d mesh modeling. I wish Nvidia could have done some HI polygon mesh modeling demos on the K1, that they had with with the A15 Cortex cores! Full OpenGL should work with Blender and Gimp, as well as other OpenSource software.

January 8, 2014 | 02:36 PM - Posted by Anonymous (not verified)

Can`t wait...AWESOME...thanks PCPER for the info.

January 12, 2014 | 08:33 AM - Posted by Kingofkats

This is such a remarkable advance that I can't wait to see K1-powered hardware hit the market. There was a rumor last week that Microsoft's next Windows RT tablet (presumably the Surface III) will be built around the Tegra K1. Has there been any confirmation of this?

January 25, 2014 | 01:08 AM - Posted by Cooling tower systems (not verified)

Its a nice post. At first I want to say thank you for sharing with us this kind of tremendous posting. Its my personal opinion that, Buying is much more effective than rent because then we have the freedom to do anything for next. invites you to read
Cooling tower systems

Post new comment

The content of this field is kept private and will not be shown publicly.
  • Lines and paragraphs break automatically.
  • Allowed HTML tags: <a> <em> <strong> <cite> <code> <ul> <ol> <li> <dl> <dt> <dd> <blockquote><p><br>
  • Web page addresses and e-mail addresses turn into links automatically.

More information about formatting options

By submitting this form, you accept the Mollom privacy policy.