More Details On NVIDIA’s Kayla: A Dev Platform for CUDA on ARM
by Ryan Smith on March 19, 2013 8:00 PM EST
In this morning’s GTC 2013 keynote, one of the items briefly mentioned by NVIDIA CEO Jen-Hsun Huang was Kayla, an NVIDIA project combining a Tegra 3 processor and an unnamed GPU on a mini-ITX like board. While NVIDIA is still withholding some of the details of Kayla, we finally have some more details on just what Kayla is for.
The long and short of matters is that Kayla will be an early development platform for running CUDA on ARM. NVIDIA’s first CUDA-capable ARM SoC will not arrive until 2014 with Logan, but NVIDIA wants to get developers started early. By creating a separate development platform this will give interested developers a chance to take an early look at CUDA on ARM in preparation for Logan and other NVIDIA products using ARM CPUs, and start developing their wares now.
As it stands Kayla is a platform whose specifications are set by NVIDIA, with ARM PC providers building the final systems. The CPU is a Tegra 3 processor – picked for its PCI-Express bus needed to attach a dGPU – while the GPU is a Kepler family GPU that NVIDIA is declining to name at this time. Given the goals of the platform and NVIDIA’s refusal to name the GPU, we suspect this is a new ultra low end 1 SMX (192 CUDA core) Kepler GPU, but this is merely speculation on our part. There will be 2GB of RAM for the Tegra 3, while the GPU will come with a further 1GB for itself.
Update: PCGamesHardware has a picture of a slide from a GTC session listing Kayla's GPU as having 2 SMXes. It's definitely not GK107, so perhaps a GK107 refresh?
The Kayla board being displayed today is one configuration, utilizing an MXM slot to attach the dGPU to the platform. Other vendors will be going with PCIe, using mini-ITX boards. The platform on the whole is in the 10s of watts - but of course NVIDIA is quick to point out that Logan itself will be an order of magnitude less, thanks in part to the advantages conferred by being an SoC.
NVIDIA was quick to note that Kayla is a development platform for ARM on CUDA as opposed to calling it a development platform for Logan; though at the same it unquestionably serves as a sneak-peak for Logan. This is in big part due to the fact that the CPU will not match what’s on Logan – Tegra 4 already is beyond Tegra 3 with its A15 CPU cores – and it’s unlikely that the GPU is an exact match either. Hence the focus on early developers, who are going to be more interested in making it work than the specific performance the platform provides.
It’s interesting to note that NVIDIA is not only touting Kayla’s CUDA capabilities, but also the platform’s OpenGL 4.3 capabilities. Because Kayla and Logan are Kepler based, the GPU will be well ahead of OpenGL ES 3.0 with regards to functionality. Tessellation, compute shaders, and geometry shaders are present in OpenGL 4.3, among other things, reflecting the fact that OpenGL ES is a far more limited API than full OpenGL. This means that NVIDIA is shooting right past OpenGL ES 3.0, going from OpenGL ES 2.0 with Tegra 4 to OpenGL 4.3 with Logan/Kayla. This may also mean NVIDIA intends to use OpenGL 4.3 as a competitive advantage with Logan, attracting developers and users looking for a more feature-filled SoC than what current OpenGL ES 3.0 SoCs are slated to provide.
Wrapping things up, Kayla will be made available in the spring of this year. NVIDIA isn’t releasing any further details on the platform, but interested developers can go sign up to receive updates over at NVIDIA’s Developer Zone webpage.
On a lighter note, for anyone playing NVIDIA codename bingo, we’ve figured out why the platform is called Kayla. Jen-Hsun called Kayla “Logan’s girlfriend”, and it turns out he was being literal. So in keeping with their SoC naming this is another superhero-related name.
Facebook
Twitter
RSS
16 Comments
View All Comments
coder543 - Tuesday, March 19, 2013 - link
It would be rather nice if they just happened to upgrade the processor to be Tegra 4 based... This platform, with Cortex-A15s, could make a rather nice general purpose computer for those of us who use Linux.hpglow - Wednesday, March 20, 2013 - link
Reading comprehension -1. "The CPU is a Tegra 3 processor – picked for its PCI-Express bus needed to attach a dGPU "jwcalla - Tuesday, March 19, 2013 - link
So.NVIDIA has a video driver for discrete Kepler GPUs built for the ARM ISA? Interesting. :)
jjj - Tuesday, March 19, 2013 - link
Too bad they aren't going for DIY desktop, fileserver and HTPC markets ,i keep waiting for the ARM guys to establish some standards and go for this market. We need some fun.Nvidia said a while ago that they'll use Denver in Tesla on the server side so maybe Kayla is not all about Logan.
It is odd that nobody mentions that Denver is late if it comes in 2015 ,granted Nvidia wasn't listing what core they are using in Logan so could we see Denver in it? Most likely they messed up and Denver is late but who knows...
Krysto - Wednesday, March 20, 2013 - link
Have you confirmed that they are using Cortex A57 in Tegra 5, or are they still going to use Cortex A15 in 2014?Krysto - Wednesday, March 20, 2013 - link
If Parker is their "first 64 bit chip" and Logan isn't, then Tegra 5 will be based on Cortex A15, unfortunately. I just hope they at least make it at 20nm. But it still shot down my chances of getting a 64 bit device in 2014.ltcommanderdata - Wednesday, March 20, 2013 - link
"Because Kayla and Logan are Kepler based, the GPU will be well ahead of OpenGL ES 3.0 with regards to functionality. Tessellation, compute shaders, and geometry shaders are present in OpenGL 4.3, among other things, reflecting the fact that OpenGL ES is a far more limited API than full OpenGL. This means that NVIDIA is shooting right past OpenGL ES 3.0, going from OpenGL ES 2.0 with Tegra 4 to OpenGL 4.3 with Logan/Kayla. This may also mean NVIDIA intends to use OpenGL 4.3 as a competitive advantage with Logan, attracting developers and users looking for a more feature-filled SoC than what current OpenGL ES 3.0 SoCs are slated to provide.I love how with Tegra 4 nVidia went around promoting how being OES2.0 based is so much more die area and power efficient compared to OES3.0 mobile GPUs. And now a year later with Logan they will jump to the opposite end of the spectrum and bring over a full desktop API which no doubt includes features that are unnecessary or inefficient for mobile and play it off as a great design decision.
TheJian - Wednesday, March 20, 2013 - link
In 2015 we will actually use it. Well, maybe...LOL - I'm saying devices show up late 2014 and games using those features AFTER that, not talking chip debut here. In 2013 it's a waste of die space and power. Do you see anyone using DX11.1 this year on a mobile product? NOPE. Yet, Rogue6, Mali, Adreno etc are all bragging about this ES3.0 & compute...You don't see them producing benchmarks, just "we've got this feature, and that feature...blah blah". We're just now getting unreal 3 type games and they won't be pushing the details you get in that engine on the desktop. NV chose wisely allowing them to hit better power targets/cheaper die size and jack up perf on what we will actually USE during T4's life. I think everyone else missed the whole point this gen. As we start to get REAL game benchmarks this year we'll see the point of nvidia's move. Dx11.1 & ES3.0 won't speed up your unreal 3 game nor make it look better on mobiles (without hitting single digits anyway). To turn all that crap on will take us back into a Nexus 10 situation or worse where it's not quite fast enough to push their res.I think T4 barely cuts the mustard at 2560x1600 and exynos5 is woefully underpowered for 2560x1600 in nexus10 (for gaming). T4i with 12 less cores should be about the same on 1080p (as T4 at 2560x1600 roughly) where top phones will be for the next year...There's really no point in going higher res in either case right? I won't see the difference anyway. I think we'll just keep seeing them jack up perf at these two resolutions (tab/phone respectively) or enabling more graphic features going forward. But for this year and most of next year you should have designed for cut down unreal3 engine type graphics (think dx9.3 type). GPU compute is pointless on mobile too for quite a while. I don't see anyone doing WORK that requires that on a phone or tablet (like I'm going to dev content on a tablet or phone...yeah right), and games barely use it on a PC. So any space dedicated to this is wasted until you can show some legitimate USE for it. I can't really name much on a home desktop that compute is for (civ 5 benchmark to prove it works? - Bitcoining? Folding at home?...LOL@all of these). I'm not saying it won't be important at home one day, just that it's pretty pointless right now (which is why NV majored on gaming for kepler and NOT compute).
http://www.unrealengine.com/showcase/mobile/
I'm guessing most of these (and more) will be on android soon. Some already are but devs aimed at ios first as it looked like apple would rule forever, but android has ended that cycle this last year. More devs will head for android going forward I think as android is dominating now and pre-orders of galaxy s4 are 4x S3's at some places...WOW.
Can anyone name a released game on mobile using ES3.0 or DX11.1? If so I guess I missed it :) I predict I'll be asking the same question with no answer for over a year :) Heck any in development even? The biggest thing for me here is all gpus from here forward in mobile should be ~xbo360/ps3 and be able to get a lot of crap ported from previous dx9 stuff on these. Everyone is jumping on the phone/tablet or ouya type stuff with gamepad as a console. Samsung just demoed a gamepad with S4 out to tv in the last week, Shield, ouya, moga, greenthrottle etc. X720/Ps4 might be irrelevant before they even ship this time. They'll be lucky to sell 20mil units for Q4/Q1'14 total for all 3 consoles, vs. ~25+x that for phones/tablets in those two quarters and most will be xbox360/ps3 quality power. How do you survive that onslaught? EA/Sony/Nintendo/Activision/MS will be pretty much the only people producing stuff IMHO as no small devs can afford the failure possibility of the devices themselves. Right behind their launch we'll be hearing about T5/T678/Rogue7/A340 (whatever they call all these). Sales of consoles will continue to face more an more power in your hands that can get to a TV. This gen was doomed on the vine; they just didn't realize how powerful your phone or tablet etc would be when they started their work on them.
One more comment: I noticed stark is missing and no mention of boulder. Are these the same, or just further out or something? Boulder was supposed to be 2015, denver 2014. But now that denver is being discussed by some as potentially used in server is boulder just not on the map now? Stark on a whole other roadmap these days that's purely server/hpc related? This was pretty much the desktop/mobile talk so maybe this explains the absence?
One more thing, just got a warning for dailytech being a dangerous site...LOL.
http://safebrowsing.clients.google.com/safebrowsin...
Infecting pc's or is google just wrong?
"What happened when Google visited this site?
Of the 999 pages we tested on the site over the past 90 days, 178 page(s) resulted in malicious software being downloaded and installed without user consent. The last time Google visited this site was on 2013-03-20, and the last time suspicious content was found on this site was on 2013-03-20.
Malicious software includes 175 exploit(s), 5 trojan(s). Successful infection resulted in an average of 5 new process(es) on the target machine."
Hmmm...interesting.
lmcd - Wednesday, March 20, 2013 - link
I personally would use a Parker-based laptop (with a KDE-based distribution), particularly with a 1600x900 13"/14" screen at $200-250 (would want mSata SSD accessiblelmcd - Wednesday, March 20, 2013 - link
As such I'd want gpu computer and such, hence my rant.