NVIDIA's Project Kal-El

[AtenRa]

From : NVIDIA's CEO Discusses Q4 2011 Results - Earnings Call Transcript

http://seekingalpha.com/article/253...sses-q4-2011-results-earnings-call-transcript

Uche Orji - UBS Investment Bank

At Mobile World Congress, it seems like everyone is racing -- and this is for Mike, there's a race to launch quad core. And first of all, congratulations on the success you've had with Tegra 2. But the question I have on quad core is, first of all, is there enough application now to take advantage of this level of power that you are bringing in at the end of this year? And then secondly, do you need 28 nanometers to make this profitably? Can you talk about the economics of that?

David White

Uche, I'll let Jen-Hsun answer that question. How is that?

Uche Orji - UBS Investment Bank

Sure.

Jen-Hsun Huang

Uche, so Quad core. It is the case that several of our competitors announced quad core products that will be sampled next year and potentially go into production the year after. At Mobile World Congress, we demonstrated Kal-El, and there were quite a few systems on-hand for demonstrating all aspects of Kal-El, whether it was extreme high-definition video, it has the ability to support very high definition displays. The overall horsepower and computational capability of Kal-El is about 5x the performance of Tegra 2, and has some really exciting new technologies for extremely low-powered operations even though we have four CPUs inside our Kal-El processor. These are all the breakthroughs that we have demonstrated at Mobile World Congress. Whereas some of the competitors seemed panicked to announce something on their PowerPoint slides, we tend to announce the product when we're actually sampling to customers. And in the case of Kal-El, we're now sampling to our strategic partners and our intention is to go to production this year. With respect to the importance of quad core, obviously, quad core is incredibly important otherwise it wouldn't be on all of the suppliers' road maps. We were the first in the world to announce and ship dual core. During that time, a lot of people asked us about that, and now it's quad core. If you think about the tablet or mobile device applications, you'll realize that very quickly the multitasking. You're multitasking whether you're streaming music, while you're reading a book or you're playing game and you're being connected to other players in a multiplayer environment. Or you do physics processing while you're playing the game, one of the most delightful parts of the game that people like these days, whether it's Angry Bird or others, it's just the realistic physics that it does. And the way that the bricks and the buildings fall apart are physically real, and you're going to see more and more of that type of capability over time. And so whether it's multitasking, whether it's image processing for very high-quality camera effects, browsing, multitasking, those type of applications are all very, very intensive users of multi GPU cores. And the last thing is it's actually logical and completely intuitive if you think about it, but four CPU cores working less hard consumes less energy than one CPU core or two CPU cores working their butt off. And that's the reason why if you look at the PC industry, what people have said about the PC hitting the brick wall and hitting the power wall. Remember when CPUs were cranking at extreme high frequencies, those are just bad ideas. Brute forcing CPU design and over-clocking it and causing it to run extremely fast is just not a very good idea. It makes a lot more sense to run at the natural frequency of the semiconductor process and utilize all kinds of parallels and ideas. In the case of CPUs, we have four cores. In the case of GPU, we have 12 cores. Using parallelism is the most efficient and also the highest performance approach to computing that we know.

Uche Orji - UBS Investment Bank

In terms of the economics, will you get it by 28 nanometers?

Jen-Hsun Huang

Well, 28 nanometer is not available yet, so it's not an option this year. On the other hand, 40 nanometers is actually more economical than 28 nanometers this year and we’ll likely expect it to be so until about first half, maybe even the midpoint of next year. So 40 nanometers, absolutely the right approach. It is the most mature, and we can go into very, very high production, very, very quickly because the yields are so great.

 

[Arkadrel]

The www.coremark.org

benchmark they run, and shows the Kal el beating a Core2duo T7200... thats a processor from around 2005 right? still its impressive.. I belive the T7200 has a TPD of 34watts (65nm processor).

i dont think nvidia stock will continue to rise. tegra 2 is first which got them some wins. but its not even clear its the fastest dual core out there.

you have the samsung chip that is already faster, t he qualcomm chips and marvell ones. its not like there is no competition. they just have a very slight lead.

if anything tegra 1 was mostly a failure, and no one cared about nvidia until they got tegra2 out. and if anything nvidia stock has always had wild swings. its been going from 6 to 40s to single digits for years. not t omention we are in the super fed liquidity QE2 "all boats rising" stock market of 2010/2011.

Your right that market moves fast, Tegra2 was only the fastest for like a week or two, before someone else took its spot. Its a tough market.... even at the speeds nvidia is going, its bound to get good competition.

 

[BD231]

I didn't actually pay attention to what was inside a computer until around 2005, so Nvidia already had IGP and chipsets. lol

I also knew very little of ARM, or how many devices they're actually inside when first hearing about Nvidias venture into the ARM. There was also the rumor of Win 8 running on ARM and Nvidia wanted to create APU's to compete with AMD/Intel and X86. I would still be laughing if their main target was desktops and Win 8.

I also think Tablets are a fad, and once the fad ends, they'll be nothing more than a niche market. My Android phone is great, so is the mobility and it's great having Internet wherever I go, but outside of calls/texts I don't touch the thing when I'm home. My desktop, even for basic internet browsing, is superior for me. Than of course productivity is vastly superior due to far greater performance. A tablet to me removes the portability that a phone has, while also removing the ease of use and productivity of a desktop or even laptop.

Don't get me wrong, I'd love to see more powerful phones, but it's my honest opinion tablets will loose popularity quickly. I could very well be wrong, guess I'll have to watch the market and find out as time goes by.

Well good thing they're going into both phone and tablet segments, tablets are not for me either. I'm still on my 1g Samsung messenger phone that plays 3gp video downloaded via USB cable :awe:.

I'm not even 3g yet so I think I'll give nVidia some time to mature their device's. Just seems like an odd market to jump into without any truly compelling tech to offer. Hopefully they can keep up with the competition in features and quality.

 

[Idontcare]

I really like these charts/graphs. (particularly the second one)

Goto-san has the best graphics, ever. Even though his articles get mangled by google translator the graphics need no translation.

 

[zebrax2]

They are greatly increasing their performance by increasing core count one would think that at some point, say like 4-8 cores, performance improvements would slow down as programs get left behind from the multithreading stand point just like it is now for x86?

 

[Ajay]

They are greatly increasing their performance by increasing core count one would think that at some point, say like 4-8 cores, performance improvements would slow down as programs get left behind from the multithreading stand point just like it is now for x86?

Yes - that just theoretical banter; same thing as saying our new sports car can do 200mph (so long as we assume that the wind resistance is zero).

 

[Mopetar]

They are greatly increasing their performance by increasing core count one would think that at some point, say like 4-8 cores, performance improvements would slow down as programs get left behind from the multithreading stand point just like it is now for x86?

For phones at least, the performance gains are going to be minimal. We're going to see a repeat of what happened in the PC industry. Phones haven't had multiple cores until just recently and a large portion of software isn't written to take advantage of it. Even worse, phones aren't as capable of multitasking as traditional PC's simply due to the way they're designed and how people interact with them. Add in the fact that many background tasks such as playing music are handled by dedicated hardware, and having more than two cores doesn't provide much of a performance boost.

Phones are still gaining features at a fairly rapid pace which doesn't leave much time for development teams to clean up old code and redesign it for multiple cores or GPU acceleration. We may get there eventually as phones start to ship with more and more cores that could provide performance gains or better battery life if the OS is optimized to work with them, but right now I don't see too much benefit in these designs. That may be part of the reason HP elected to use a fast, single core for the Pre 3.

 

[cotak13]

From : NVIDIA's CEO Discusses Q4 2011 Results - Earnings Call Transcript

http://seekingalpha.com/article/253...sses-q4-2011-results-earnings-call-transcript

So basically the real use case for these quad core ARM processor is gaming smart phones and gaming tablets. I am not sure that will pan out. Not with tablet and smart phone software in the single digit price range. You'd want your software to be min effort and max gains because of the low price. We'll likelly see developers ending up designing it for the lower end mass market devices, reducing the draw of quad core SoCs.

 

[Dribble]

I see half the people in these threads complaining that all the phone/tablet soc's are woefully underpowered vs anything they use today, and now the other half saying they'll never find a use for all this power a quad core soc will give?

IMO if you provide the power people will use it.

 

[Ajay]

IMO if you provide the power people will use it.

+1. More importantly - developers will use the extra horse power. I see people playing games on their cell phones all the time. I wouldn't have guessed that a few years ago.

 

[Khato]

I see half the people in these threads complaining that all the phone/tablet soc's are woefully underpowered vs anything they use today, and now the other half saying they'll never find a use for all this power a quad core soc will give?.

Well I won't complain, but all current ARM chips are woefully underpowered on single-threaded performance when compared to even the mediocre Atom. And that's why they actually might have some valid multitasking usage scenarios for quad core on a phone - when it takes an entire core in order to complete a simple task, then it's good to have extras.

On another note, it's really quite amusing looking at Coremark "benchmark" scores. Such an unrealistic metric.

 

[Mopetar]

I honestly don't think more cores will add a lot. Right now most of the software isn't written to take advantage of it and a lot of the most common tasks are handled by dedicated hardware.

Watch a video and it goes through dedicated hardware as long as the codec is supported. Listen to a song and it goes through the audio processor as long as it's a known codec. Take a picture and the image processor handles it.

The biggest benefactors would be web browsing and intensive games. Most other applications aren't too terribly taxing in terms of CPU usage.

 

[itsmydamnation]

http://www.ilsistemista.net/index.p...a-soc-truly-faster-than-a-core2-.html?start=3

good old NV............

helps to compare apples to apples

 

[drizek]

lol, funroll-loops

 

[wuliheron]

Don't underestimate games.

People don't want to lug around multiple devices or large tablets everywhere they go. With a gaming cellphone I can download free games online, hand it to my kid in the backseat to keep him quiet, or waste the occasional five minutes standing in line. No need to drive my kid to the Gamestop store or any such nonsense. Inevitably there is only so much you can do on such a small screen and the makers of Gameboys are already so worried they're trying to come out with their own cellphone.

 

[Mopetar]

Don't underestimate games.

People don't want to lug around multiple devices or large tablets everywhere they go. With a gaming cellphone I can download free games online, hand it to my kid in the backseat to keep him quiet, or waste the occasional five minutes standing in line. No need to drive my kid to the Gamestop store or any such nonsense. Inevitably there is only so much you can do on such a small screen and the makers of Gameboys are already so worried they're trying to come out with their own cellphone.

Considering that something like RAGE is capable of running on an iPhone and looking pretty damned good, I question whether or not any game you might give to a kid is going to seriously tax the system.

There are certainly a few that will give it a run for its money, but I don't believe that the most popular ones such as Angry Birds push the system to its limits.

 

[LiuKangBakinPie]

What happened to their GPU replacing CPU ramblings they use to make for so long.

 

[cbn]

http://www.ilsistemista.net/index.p...a-soc-truly-faster-than-a-core2-.html?start=3

good old NV............

helps to compare apples to apples

 

[Mopetar]

They probably realized that while a GPU can be amazing at some embarrassingly parallel tasks, that it's embarrassingly terrible in terms of general computing and wouldn't be replacing the CPU for that any time soon.

Anything that's highly sequential and branch heavy is going to run horribly on a GPU. Unfortunately, a lot of what a computer does falls into this category.

 

[cbn]

The biggest benefactors would be web browsing and intensive games. Most other applications aren't too terribly taxing in terms of CPU usage.

According to this article from April 9, 2010 Web kit 2 will only use two cores.

Does anyone know if Google (or other companies) have plans to have browsers scale on cpu cores beyond this?

 

[Acanthus]

You guys seem to be thinking phones and tablets only.

ARM will be Windows-8 compatible and easily fit into a low-mid range solution for full fledged PCs. Nvidia has a boatload of chipset experience under their belts.

Tegra 4 on an Nforce style motherboard coming in at a lower total system price for a similar intel offering will be an automatic win.

Atom and Tegra will collide in notebooks. the problem here is Intel will always have the manufacturing advantage due to the smaller process.

If intel gets serious about competing here, they can enter the market within a year and dominate.

 

[wuliheron]

Considering that something like RAGE is capable of running on an iPhone and looking pretty damned good, I question whether or not any game you might give to a kid is going to seriously tax the system.

There are certainly a few that will give it a run for its money, but I don't believe that the most popular ones such as Angry Birds push the system to its limits.

Rage is an exception that uses an advanced "megatexture" data compression that makes it ideal for portable devices. It's so radically new and proprietary that you won't see anything else like it come from anyone but ID for many years to come. If cellphones are to play legacy 3D games they'll require significantly more power.

 

[LiuKangBakinPie]

Two cores seem to be the sweetspot and the spot where to make money

 

[Mopetar]

Rage is an exception that uses an advanced "megatexture" data compression that makes it ideal for portable devices. It's so radically new and proprietary that you won't see anything else like it come from anyone but ID for many years to come.

I'm sure ID would be more than happy to license the engine considering they've done it with their engines in the past.

If cellphones are to play legacy 3D games they'll require significantly more power.

I so look forward to being able to run down my phone battery to the point where it becomes less useful as a phone. Eventually process improvements, improved chip design, better batteries, and optimized code will fix this problem, but until then running anything with all of those cores and a powerful GPU will chew through the battery.

I don't know how many people regularly play these types of games, but based on the top selling games for iOS and Android, it appears that they're not in the majority. There's not much of a point in optimizing for a use case that maybe 5% of your users actually fall into.

Most reviews of new dual core phones have said that there isn't a whole lot of use for the extra core right now. The most useful thing they've found you can do is get better performance while running several things in the background, which most phone users don't like to do because it erodes battery life.

The Tegra 3 will probably be great for ARM netbooks and allow for some very inexpensive devices. I think that the Tegra will do well in this market, but for phones and tablets it's more overkill than anything at this point. Until Apple, Google, et al. start optimizing their OS code to take advantage of a growing number of cores, having more than one or two won't be very worthwhile.

 

[IntelUser2000]

According to this article from April 9, 2010 Web kit 2 will only use two cores.

Does anyone know if Google (or other companies) have plans to have browsers scale on cpu cores beyond this?

I don't see a point beyond certain levels of performance for web browsing. I mean, monitor CPU utilization when browsing with your desktops systems. Even on the i5-661, the peaks were less than 15%. So beyond 15% of i5-661's performance, faster CPU wouldn't do much. Maybe faster single thread performance cores can help, but that's subject to very fast diminishing returns, because network speeds would come into play.