Comments Locked

52 Comments

Back to Article

  • Cellar Door - Thursday, November 14, 2019 - link

    What clearly is needed are RTX 3080 cards, that will cost 30% more then last RTX generation and just barely beat a 1080Ti.
  • Arkturus - Thursday, November 14, 2019 - link

    So, for the entire year Nvidia is just trying to milk the consumer part of their business (gamers) to fund their failed attempts to compete in other areas such as automotive & data center; neat... Though, hopefully this shows Nvidia that they can't pull off shitty Apple style pricing schemes and still expect everyone to buy their GPU's like there's no tomorrow. Sadly though, until there's any good financial reason for them to make better GPU's like we want, the milking part will probably continue, let's just hope prices will be somewhat reduced for 3000 series cards.
  • michael2k - Friday, November 15, 2019 - link

    You sound salty.

    NVIDIA can't both be trying to milk gamers, and milk them. They are either successful at it, or they aren't. Given the lack of competition for the most part I think they're successful.

    Likewise, you don't fund failed attempts if those attempts are growing.

    From 2014 you can see how far they've come:
    https://www.anandtech.com/show/8003/nvidia-q1-2015...

    GPU $898m -> $1,659m
    Tegra $139m -> $162m (automotive)
    Other $66m -> $143m (OEM & IP)
    Visualization $0 -> $324m
    Data Center $0 -> $726m

    Put shortly, the two new groups, Visualization and Data Center, bring in more revenue now than all of GPU + Tegra in 2014.

    You can't win bets if you don't make them, and these things don't happen overnight. Tegra was originally a mobile strategy, and they had to switch course in 2016 when they sold of Icera and refocused on Jetson, Automotive, and sales to Nintendo. Automotive is a really long bet, but it will be a much slower ramp as cars are sold at much lower rates than handheld gaming units.

    I wouldn't bet against NVIDIA.
  • eva02langley - Friday, November 15, 2019 - link

    I would not bet on them either. There is competition in town now...
  • Operandi - Friday, November 15, 2019 - link

    There is?
  • michael2k - Friday, November 15, 2019 - link

    Technically, Intel has their Xeon Phi products for deep learning and and GPGPU clusters, as well as purchasing MobilEye for their automotive division.

    So far no one else has a competitive GPU (AMD doesn't quite count, in that if AMD can't hit the high end mark then they aren't competing in the same markets), nor a competitive mobile part (Qualcomm definitely has a decent CPU in the SnapDragon, but their Adreno isn't designed to scale the same way Maxwell, Tesla, Volta do, since Volta cores from their desktop part are found in their Tegra) or embedded part (Their Xavier part can use on board Volta or a discrete GPU)
  • eva02langley - Friday, November 15, 2019 - link

    Yeah, AMD is not having a 1080 TI competitor for 400$... (sarcasm)
  • webdoctors - Friday, November 15, 2019 - link

    I think you misunderstood that table. The decrease in automotive revenue and the tiny automotive revenue is bad for consumers, not just Nvidia. They're the leader in autonomous driving platform, if they're not selling any that means no one's gonna have selfdriving cars.

    The sad fact that automotive manufacturers haven't migrated to selfdriving cars sucks for all of us, they dont have any incentive and the industry is stagnating. I thought by next year we'd already have lvl 4 selfdriving cars and it looks like we're going to be several years past that target.

    My biggest fear when I leave the house is getting killed on the road. There's so many idiot drivers and they keep increasing. Ppl don't know what turn signals are for, and think alcohol makes them smarter behind the wheel.
  • bobhumplick - Friday, November 15, 2019 - link

    we dont need better cars. we need better people. we need the highway overhaul that has been put off for the last decade. we need cars that people can afford so they dont have to work 60+ hours a week and fall asleep at the wheel driving home. the tech industry cant get a game launch right and yet we want them to drive our cars? no thanks.

    i have no TV, no high end stereo, my phone cost 50 dollars new and i just use it for calls and a couple authenticator apps. my one luxury item that i make room for is my computer so ill spend a bit more for a gpu than i should. but i dont want my life staked on ANY companies drivers.
  • flgt - Friday, November 15, 2019 - link

    That's because making autonomous vehicles is super hard. And once you start making computer systems that can literally kill people the level of regulatory scrutiny goes up a 1000X. It's not a good fit for the silicon valley whiz kids who have a 5 minute attention span and expect to be millionaires in 3 months. I think we're at another decade away.
  • UltraWide - Friday, November 15, 2019 - link

    LOL
  • eva02langley - Friday, November 15, 2019 - link

    They are not, Black Berry and QCOM are more present in automotive than them. The self-driving solution from Nvidia is entirely relying on GPU, surprise, surprise... while the solution CANNOT rely only on visuals.
  • Drkrieger01 - Friday, November 15, 2019 - link

    Don't quote me on this, but I'm fairly certain their mobile (auto) platform doesn't just support camera inputs, but lidar/radar as well.
  • eva02langley - Friday, November 15, 2019 - link

    Their main focus is on selling GPUs and all their AI is either looking at Neural network, which makes sense, but all they do is analyzing video feeds.

    Basically with AI, your goal is to reproduce human cognitive senses for being able to make cognitive links which we call learning. Nvidia is focusing on visuals. Anybody arguing here is just plain old clueless.
  • Phynaz - Saturday, November 16, 2019 - link

    Lol, speaking about clueless.

    Fanboys are a pestilence. AMD fanboys are the are the most insecure of the all.
  • Kjella - Friday, November 15, 2019 - link

    "The self-driving solution from Nvidia is entirely relying on GPU, surprise, surprise... while the solution CANNOT rely only on visuals."

    Hello time traveler, it's been about 25 years since graphics cards went from using 2D to 3D engines. Any sensor that can record depth so you actually get (x,y,z) points speaks way more "native GPU" than any cameras trying to reconstruct 3D from 2D images.
  • eva02langley - Friday, November 15, 2019 - link

    They relate on video feeds.... and yes, radar, lidar, whatever... is video... feeds...

    Hello hello genius...
  • michael2k - Friday, November 15, 2019 - link

    You don't understand AI nor neural nets. AI is used to perform multidimensional pattern matching, which can be used for video, but is not only used for video. Anything you can do on a CPU you can do on a GPGPU, and certain data sets perform better on a GPGPU, given certain algorithms and design, and one of the things NVIDIA focuses on is neural networks, both training and recognition.

    And... what else is there other than video feeds if you're going to reduce everything down to video feeds? A self driving car can only operate on the input it is given, and if you remove radar, lidar, cameras, and ultrasound, what else is there? Vibrations from the road?
  • Phynaz - Saturday, November 16, 2019 - link

    It’s data dumbass
  • michael2k - Friday, November 15, 2019 - link

    It's a good thing NVIDIA's GPUs can work with lidar, radar, GPS, and ultrasound then.
  • Phynaz - Saturday, November 16, 2019 - link

    What’s AMD’s solution?
  • rahvin - Friday, November 15, 2019 - link

    This is absolutely incorrect. Most of the automakers are choosing to user their own developed IP rather than rely on and be at the mercy of nVidia.

    Anyone with any knowledge of the automotive industry knew they would go that direction. The Automotive manufacturers learned a long time ago that it's bad for business if your business is dependent on an external company for an essential part. Where would they be if there was only one brake manufacturer and they decided to raise prices 400%?

    Each company is investing billions to develop their own solutions, some like Tesla, Toyota and Daimler are even working on their own silicon. I don't know if the others are, but they will be if not now, soon. Their business model is mostly to use the nVidia product to jumpstart their development then move to their own or commodity silicon.

    In fact that huge drop in the automotive sector was the loss of Tesla as they have already deployed their own silicon in the Model 3 and all newly produced vehicles. It will be interesting to see if nVidia ever recovers their Tegra investments.
  • michael2k - Friday, November 15, 2019 - link

    Considering the hottest gaming system now, the Switch, runs on the Tegra, I'm fairly certain they aren't losing money in their continuing Tegra investments.

    And yes, even if companies are using Tegra to kickstart their internal development, they are still using Tegra:
    https://nvidianews.nvidia.com/news/nvidia-and-toyo...

    Smaller companies really don't have the resources to DIY:
    https://blogs.nvidia.com/blog/2018/11/20/delivery-...
    https://nvidianews.nvidia.com/news/volvo-selects-n...
    https://nvidianews.nvidia.com/news/yamaha-motor-ad...
    https://nvidianews.nvidia.com/news/leading-japanes...

    And, to be clear, these things will take 4 or 5 years to have any impact on NVIDIA's bottom line.
  • eva02langley - Friday, November 15, 2019 - link

    However, the rest of the game industry develop games on AMD hardware... surprise surprise...
  • michael2k - Friday, November 15, 2019 - link

    So what? NVIDIA makes 1.6x the revenue (3.014b vs 1.8b), 50% more efficient at generating revenue (63.6% vs 43%), and 7.5x as much net income. Both companies make GPUs and CPUs, license IP, have console HW, and cluster compute solutions, but only one of them is profitable, as AMD has reported an operating loss of $152m for the last nine months this year, and an operating loss for the same 9 month period last year

    AMD's quarterly results: http://ir.amd.com/news-releases/news-release-detai...

    I certainly wouldn't bet on AMD to win any races, they are constantly fighting just to stay alive.
  • Phynaz - Saturday, November 16, 2019 - link

    Your point? I mean besides continuing to point point out how ignorant you are.
  • michael2k - Friday, November 15, 2019 - link

    The average age of cars in the US is 11 years. That means even if everyone released only self driving cars today, it would take 11 years for the majority of cars to be self driving.

    So given that constraint, if it takes 11 years for everyone to have a single self driving car, and another 11 years for all cars to be self driving, then it would be about 30 years before all cars on the road are self driving cars.
  • TelstarTOS - Friday, November 15, 2019 - link

    Arkturus is absolutely right. Exactly last year NV moved up segments and midrange gpus became highend. This year they continued the same bullsnit but consumers got fed up. Only strong competition from AMD will make this trend stop.
  • eva02langley - Friday, November 15, 2019 - link

    I don't understand how they can have gross margins of 63% and only have 900M$ of net income.
  • madwolfa - Friday, November 15, 2019 - link

    They're reinvesting heavily in R&D.
  • michael2k - Friday, November 15, 2019 - link

    SEC explains: https://www.sec.gov/reportspubs/investor-publicati...

    The raw data from NVIDIA is here: https://nvidianews.nvidia.com/news/nvidia-announce...

    Gross Margin is net revenue, unreported here, minus cost of goods, also not reported here, divided by net revenue (NR - COGS)/NR == Gross Margin; that means cost of goods is $1,097m

    So Gross Profit is NR - COGS = $1,917m

    Operating Expenses is cost associated with research, development, marketing, etc, and is $989; subtract it from Gross Profit and you have Operating Income. There is a rounding error, but $1,917m - $989m = $928m (or as NVIDIA reports it, $927m operating income)

    So to get to their Net Income of $899m, you subtract interest and taxes paid, unreported here, but the difference in Operating Income and Net Income is $28m, so that means NVIDIA paid $28m in interest and taxes for the quarter.

    Tada! That's how they got a 63% gross margin and $899 of net income
    Given we have Operating Income and Operating Expenses, we can add those two to get Net Revenue, or $1,916m
  • eva02langley - Friday, November 15, 2019 - link

    ROFL... another genius... no, what I don't understand is how little money they make after all the cost of the company... after such high margins.

    What it means is that management is having issues somewhere.
  • Alistair - Friday, November 15, 2019 - link

    Be nice. He explained it to you. Gross margins are not what you seem to think they are. nVidia isn't a grocery store.
  • michael2k - Friday, November 15, 2019 - link

    Man, wait till you see AMD's bottom line then!

    http://ir.amd.com/news-releases/news-release-detai...

    NR: $1,800m
    Gross Margin: 43%
    COG: $1,026m (so roughly the same as NVIDIA)
    Gross Profit: $774m (so 40% of NVIDIA)
    Operating Expense: $591m (59% of NVIDIA)
    Operating Income: $186m (20% of NVIDIA)
    Interest + Taxes: $66m (2.35x NVIDIA, indicative of how much AMD borrows)
    Net Income: $120m

    The report outlines paying $24m in interest, $36m in other expenses, but doesn't explain where the other $20m is spent; in the same report they list liabilities, or money they owe that they haven't paid out yet. They currently own $872m in long term debt, $201m in lease operations, $140m in other long term liabilities, and $7,265 in accumulated deficit (or, debt they have accumulated over several years as operating losses continue to accumulate faster than cash flow can pay them out)

    You can also look at NVIDIA's liabilities too:
    https://nvidianews.nvidia.com/news/nvidia-announce...

    They actually generated $45m interest income, and owned $13m in interest, for a net positive. Long term debt is $1,990m, lease is $469m, and other is $662, all at least twice AMD, but no accumulated deficit; meaning, they don't owe people money they haven't been able to pay.

    If you want to compare their 'cash', necessary when deciding to purchase a company, license IP, or create a new product, NVIDIA has $9,769m to AMD's $1,156m, and both have comparable inventory (an asset if not too large or too old) at $1,040m for AMD and $1,047m for NVIDIA, which would mean current GPU and CPU and not old/discounted/unsold stock.

    Or, TLDR; if you think NVIDIA makes 'little money', you'll like how much less money AMD makes.
  • Phynaz - Saturday, November 16, 2019 - link

    Rofl, you just owned yourself again you fucking moron.
  • willis936 - Friday, November 15, 2019 - link

    The same story for the past ten years. We need competition in the GPU space. Will ATI ever step up?
  • Holliday75 - Friday, November 15, 2019 - link

    ATI will step up when 3dfx does.
  • crimsonson - Friday, November 15, 2019 - link

    lol

    touché
  • Beaver M. - Friday, November 15, 2019 - link

    Easily seen how successful Nvidia with Turing really is if you compare the adaption of Turning with the older ones on Steam via the Steam survey for every month after their release.
    Even now, over one year later, Turing cards have laughable percentages. Pascal and Maxwell were adapted MUCH MUCH faster and MUCH MUCH more numerously.

    Their stock has also only recovered 50% after the huge dip caused by the Turing disaster. And it also doesnt look very stable.
    With all those facts I wouldnt invest in Nvidia right now. Maybe their next generation fixes it by having a massive performance gain and much lower prices.
  • michael2k - Friday, November 15, 2019 - link

    HAHA. You mean massive performance gain and much higher prices, right?

    Because the only way to see much lower prices is if AMD had an even larger performance gain.
  • Beaver M. - Friday, November 15, 2019 - link

    One can hope Nvidia sees the obvious. Their overpriced cards dont sell well, even though their propaganda machine is running hotter than ever, throwing them at Youtubers, Streamers, sponsoring "How RTX cards make you WTFPWN EVERYONE" videos, etc.
  • Yojimbo - Friday, November 15, 2019 - link

    Ironically, their "overpriced" cards do sell, proving that they are not overpriced.
  • Beaver M. - Saturday, November 16, 2019 - link

    Of course they sell, but not very well compared to other generations. I proved that with my other comment. You dont want to check for yourself, yet attack me?
    You also ignore that they are already leaking stuff about their next generation, and that it is coming relatively soon. That also indicated that the current gen isnt selling well at all.
  • Yojimbo - Friday, November 15, 2019 - link

    And.. NVIDIA's "propaganda machine"? I think you need a few weeks vacation in the real world...
  • Phynaz - Saturday, November 16, 2019 - link

    What? Did you read the report? Another fucking AMD moron.
  • Beaver M. - Saturday, November 16, 2019 - link

    I dont have a single AMD hardware piece. Not even a console. I dont plan on buying one either.
    Feeling stupid yet?

    I am just stating the obvious, while you uncritically swallow stuff, even though history has proven how corporations, incl. Nvidia before, fix their reports and leave out facts to make them look not as bad as they really are. Even the article helps it by not showing 2018 in that graphic.
  • Yojimbo - Friday, November 15, 2019 - link

    Turing SUPER cards are counted separately from non SUPER cards. The big difference between Pascal and Turing on Steam is likely the Chinese market. A large percentage of those 1060s are in Chinese internet cafes, and they haven't been purchasing as much recently, probably because the Chinese government is pushing back against gaming culture in China and because of fears over the Chinese economy. Also, those Chinese internet cafes were overcounted in the survey for many months, because the same machines were being logged into with many different accounts, including while the Pascal market share was rising. Steam says they've fixed the problem now, or at least they have mitigated it.
  • Beaver M. - Saturday, November 16, 2019 - link

    Duh. Super cards barely show up outside of "others", they are that rare.
    Listen to yourself, your excuses. China is making Chinese not buy Turing? WTF???

    The percentage of 1060 (and thus others) didnt change much at all, not unusually much compared to previous generations. And those were before PUBG became popular in China. Turing however is extremely slow on Steam compared to them.
    Its no secret that the 1060 is the most successful Pascal card by far (its still getting bought and produced), just like the 970 was before it, and which also is still very strong on Steam. That 5 year old car WITH a huge scandal attached to it, is still almost 1% stronger than the strongest Pascal. And 1% difference on Steam is A LOT for something like that.
    And that popularity isnt even based on the Steam survey. Every hardware site reported about that.

    You remember how China suddenly was entering Steam due to PUBG? There was a huge change in everything. Games bought, games played, etc. But you didnt see that on hardware. It was very smooth, because they dont buy much different than the rest of the world. The 1060 didnt suddenly have a huge increase in share. Its was very similar to generations before that. There was no huge change either since then.
    And you can see the mentality against Turing in any hardware forum as well. People (no, not AMD fanboys only) are not recommending it nearly as much as previous generations. And thats not only due to the price and low performance gain, but also because of the low VRAM and other issues, like stuttering and unusually high hardware failures.
    Youre completely overstating your stuff to make your point, while ignoring other very important points and that Pascal is still being produced and sold. Very dishonest.
  • Beaver M. - Saturday, November 16, 2019 - link

    *is still almost 1% stronger than the strongest Turing.
  • jabbadap - Monday, November 18, 2019 - link

    A bit late reply but saying it anyway: There's so many 1060s and I don't think steam differentiates them. So all the laptops, desktops ones are just counted as 1060 at the steam stats. Which quite explains ~15% market share it has. I'm not saying Steam HW survey is perfect, quite far from it. But it's not far from the ground trough though, when considering that nvidia has sold something like 3:1 against AMD quarter after quarter since Maxwell came out 2015. And most of the RX570/RX580 went to miners rather than gamers.
  • Yojimbo - Friday, November 15, 2019 - link

    The article says NVIDIA spent $53 million more on R&D this quarter than a year ago, but NVIDIA spent $107 million more on R&D this past quarter than the quarter a year ago. This Q3 they spent $712 million, last year's Q3 they spent $605 million.
  • Yojimbo - Friday, November 15, 2019 - link

    Oh, and the total of NVIDIA's R&D investment in the first three quarters of their fiscal year 2020 is $2.091 billion, not $400 million. They've spent $362 million more on R&D through the first three quarters this year as compared to last year.

Log in

Don't have an account? Sign up now