the refunds probably made little difference (if you were talking about that) or if you were talking about how well it sold then yes it probably made a bundle of money for them
returns were less than 5%, (citation needed) and not all of those 5% were due to the ROP and Memory Allocation. Some just took advantage of a 3 month return window, and some just flat out returned and didn't replace with anything.
Who the hell would return their 970 because it turned out to have 1/8th less addressable VRAM? Benchmarks are still benchmarks, and the card is a steal for the price, even if by some unlikely scenario games within the next few years adequately utilize 4GB of VRAM.
I don't really consider the GTX970 a 4K-ready card, anyway. Not because of the 3.5GB VRAM situation, but because the card can't even run CURRENT games at high detail at 4K let alone future games. I run BF4 at 2560x1440 and can't have everything maxxed without losing 60FPS, and Frostbite 3 is almost a 2 year old engine.
But the card will run 1080P in any game for many years.
"Who the hell would return their 970 because it turned out to have 1/8th less addressable VRAM?"
Not because of the vRAM, but the reduced ROP count. But I'm running a very non-typical task, pushing 8k and above (sometimes far above) output sizes for testing supersampling methods for VR. That 7/8 ROP count is nearly a 7/8 drop in pixel output rate (not particularly complex scenes). Hard to test that directly with the move to a 980 as the extra SMs make it an only indirect comparison, but I am seeing that range of performance improvement.
From what I understand the ROPs can feed the shaders all the data they need, it's 14/16th the ROPs and 13/16th the shaders so it actually has more ROPs per shader than the 980. It's only addressing the last 0.5GB that's borked.
I think it's not even an issue of the technical deficiencies - the benchmarks already take these very minor issues into account. What people are annoyed about is NV apparently lying. I never saw it as more than a marketing mishap though, and was happy recommending a 970 with full knowledge of the controversy.
Sounds like you may have benefitted from the full spec 980, but I am sure the 970 was attractive to you because it offered roughly 80% of the performance at 60% of the price of the 980? Interestingly, this paradigm has not shifted at all, despite Nvidia restating some other specifications on the 970.
Exactly, and we have been over that again, and again, but of course the idiocy in the gamers arenas, even what everyone calls the high end is enormous. If the gaming gooober doesn't "FEEEEEEEEEEEEEEL" good "about his rig", well then no amount of smooth FPS gaming can soothe the mental midget. Have no doubt many of them will a few years from now fire up a game at 4k, wait 15 minutes, then get the 3.27fps fraps report and say "see, I told you so ! " - no amount of explaining that 3fps is not playable will suffice. The insane nutter butter peanut butter hamwhich kookie needs his ego massage more than life itself.
Not yet. And it will be likely very small. Reports state that only 5% or so are returned. (Also might include other causes)
Some e-shops (like Czech CZC) report defect rates if available. (An ASUS 290X DCII has for example ~24% rate, while MSI with 290X got 8%. GTX 970 ASUS DCII got ~%, didn't check any further)
@jackstar7, the 970 undoubtedly played a huge role in driving Nvidia to record profits and margins, and the price cuts it forced upon AMD's flagship 290/X parts clearly had the opposite effect on AMD's quarter (greatly reduced revenue, profits, margin on their GPU business).
For most buyers, the relevance of Nvidia correcting and restating the 970's specs will never materialize beyond a few bits in an article or on a PDF, because the paper changes do not change the performance level they paid for when they bought the card. I guess if Nvidia's pricing of the 970 at launch seemed too good to be true, it probably was and one of the warts compared to the 980 was exposed.
Is that enough to make people return their cards or pay the premium for the 980? Looks like the answer is no.
What has been years in coming has finally happened (albeit not under the circumstances either company expected). Nvidia has overtaken AMD in revenue. AMD is now officially second fiddle in both technology and revenue to all their main competitors sans consoles. Even as Tegra continues to flounder, Nvidia continues to prosper. I don't see AMD ever coming back to be a significant force. Their next GPU release might be their last big high-end competitive launch unless Nvidia massively stumbles in future GPU iterations.
Tegra revenue for FY15 was $579 million up 45% from FY14's $398 million.
The data from the article is only for ONE QUARTER and the reason: Tegra Processor sales declined 15 percent from a year ago, driven by the product life cycle of several smartphone and tablet designs.
You should not use one quarter's data as a measurement as there is too much variability from one quarter to the next you should use Y-to-Y.
Listen to their post results call at min 35 and 20 sec http://investor.nvidia.com/eventdetail.cfm?eventid... " between automotive and Shield (our gaming platform) that represents the vast majority of Tegra now " So they admited SoC sales to others are almost nothing
In other call related news while talking about Tegra they said they have gaming news soon so the March 3 event is Tegra related for sure.
No they didn't made 579 million,they had revenue of that much and if you want to look at historic data look at previous years too. I'll get you started by letting you know that in 2012 Tegra revenue was 591 mil, in 2012 Tegra was 764.447 millions revenue.Then look at their SEC fillings and you'll find how much money Tegra is (mostly) losing. In 2011 Tegra op income was 43.736 mil , in 2012 Tegra op income was 40.508 millions and in 2013 Tegra op loses were 268.068 million while in Q1 2014 the op loss was 61.440 million. Tegra as a mobile SoC is dead ,it it gets revived or not it remains to be seen.
ROFL. Car revenue from SOCS/infotaintment systems is expected to be 15B by 2017. NV is shaping up to be a leader there and only needs 1/3 of that to more than double the size of their whole company revenue. They only need to crack ~$1B revenue for tegra to be profitable. They are not that far off and we have all been watching the fakes up to x1 which is the FIRST one will a TRUE mobile gpu on it (kepler wasn't built for this, maxwell was but we lost denver due to tsmc for this rev).
Tegra is just starting, not ending. I don't believe we'll see the full affect until the chip AFTER X1 though when they put back Denver on 14nm samsung with maxwell in too. Both of those together should be very good. You have to remember Denver was NOT made for 28nm, but TSMC screwed up the schedule forcing them to go with quick IP from ARM instead of remaking changes again to 20nm (after having to already do that to put it out at 28nm). So the real deal got delayed a bit, but we'll see it at 14nm and then I think we'll see it be able to be used in many more phones/tablets paired with samsung modems most likely (since they just hit cat10 while qcom is now seeming behind at cat9). We will also know more about the NV suit vs. samsung in another quarter or two and how that affects pricing on NV getting stuff fabbed over there, or if there's a huge payout over 6yrs like Intel etc, or if it was to get free modems from samsung included etc. Who knows how that works out.
I can't remember the stat group that said auto socs/systems would be $15B, but this year semiconductor automotive crap as a whole is 34B. That's a lot of dough, and with cars doing FAR more in all this soon and across the board, we'll see them use 16-20 socs per car (which is why this market is so huge) with all kinds of driver assistance crap, etc.
http://www.statista.com/statistics/277926/projecte... The whole ball of wax. I'd have to dig further to find the part that singled out the soc part of that for 2017 but it was 15B estimated (it was an article about NV infotainment systems/ADAS for auto IIRC and the data point was mentioned). They make more from the car "solution" than from a soc sale to a mobile device and that stuff is just starting. They can wait a few revs for their desktop gpu drivers (now used in mobile) to take over gaming on android etc. Cars will more than make up for it shortly.
Intel is losing 4B+ a year on mobile. They aren't going to quit...LOL. They are hoping to shrink their way in, same way NV is hoping to shrink in a modem at some point while keeping great gpu power for gaming and utilizing 20yrs of gaming drivers to help take over anything gaming in mobile (already gaming has taken over all the app stores). The first 5-6 years of cuda were to get them where they are now also (owning 80% of the workstation market and some server stuff). You don't seem to get the LONG game. Intel and NV do. Having said that I think Intel either buys NV or loses that war (and maybe at some point a good portion of notebooks/desktops as sheer numbers drive apps/games up in quality from devs on 64bit arm). To beat the enemy here I think they have to join them (buy NV and make a better ARM than all other arm's...LOL).
NV can lose 2.5B over 5-6yrs of tegra if it makes them 1-3B a year for years to come in the next few years. Profit over revenue I mean, as again, they are chasing an estimated 15B revenue here alone, assuming NO gain in mobile devices. So if they took say 1/3 of that at $5B they're chasing (and they could end up leading it all taking 70-80% like discrete/workstation stuff), they could easily make 1-3B profit on it as they sell a whole solution here. That would only be a 60% margin at 3B profit on 5B or so revenue, more than doubling the revenue of the company and blowing up profits big time. If you don't get the point of tegra, I guess you just don't understand where we're headed. Once gaming takes over mobile even more (with xbox360 and up type gaming quality), you will see NV start to take some real numbers from mobile. Intel is basically paying for any tegra losses...LOL. If samsung gets bit for just as much or more (they make far more than Intel), and of course Intel has to re-up again in 2016 (or go back to court and lose again) they could take it in the shorts for a few more years until auto is at 15B and laugh. I expect a few people to lic NV gpu IP soon (samsung? Maybe apple if they start losing big to 14nm maxwell and beyond). It's a LONG game, not 5yr war. Can you imagine gpu revenue from PC's etc being the smallest portion while cars pulled in even 1B+ PROFIT? ROFL. Quit tegra? You must be kidding. As 64bit apps and far better games amp up they could fight for a piece of the PC desktop/notebook pie for real (way past chromebook crap). Yearly x86 pc numbers vs. ARM numbers are pathetic and getting worse. Wintel is in trouble soon.
Nvidia never made more than ~825mil in a year (2007/2008 depending on how you count quarters), but we're gaining on that now and this isn't including cars exploding, share buybacks that are now getting them to 2007 levels (when stock was ~$38 at ~800mil income, far weaker company), the 1000 companies testing grid, or at some point making something from mobile. Grid and shield each cost 10mil to develop, but may end up yielding huge returns in the end game. I think Jen just laughed at your post as he ponders Tegra 5yrs from now ;) Cuda took 8yrs to get to basically domination. Google's car solution is 150K of high tech gear for self driving crap, while NV plans to do it with 16-20 socs and a dozen cheap cameras (with grid helping) making it affordable for everyone at some point. What is that $400-600 or so in socs, and a few hundered on cameras vs. 150K per car?...LOL - add a few hundred more in software or something etc, the point is far cheaper. Google will have to go NV (someone else? Maybe apple, we'll see how they go) or give it up.
https://www.abiresearch.com/press/global-advanced-... Driver assistance crap blowing up huge soon. As stated from 11.1B 2014, $90B 2020, $200B 2024 as it goes mass market shortly. No longer just in a tesla etc. That's a lot of dough to fight for (not all a soc, but it is a huge part of the equation here, predicting everything, dash crap, infotaintment etc). You have ARM, Apple, Google, Nvidia, Qcom, probably Intel at some level all fighting over portions of it all but no denying NV has a decent shot at a good bit of it. NV's tech is a good fit for auto which is why it is no surprise they left mobile to some degree (where they were not a market share leader for now) and went to where it's a wide open HUGE market with no real leader.
The Tegra segment is almost entirely auto, Shield/ Shield Tab, Note and old consoles. You can count their wins on one hand and all are low volume- it's not at all hyperbolic, it's just what it is. If you look at the press release they intentionally avoid mentioning the Tegra brand , they mention it once when talking auto highlights and given the SF event on March 3 , i wonder if they even plan to attend MWC.
And Xiaomi Mipad and some Chromebooks, all low volume. Again i am not assuming or speculating , it's a fact Nvidia itself acknowledges, the Tegra segment was mostly other things not SoC sales in mobile.
805 was a stopgap chip not worth using in anything with the 810 supposed to show up sooner than it did. The Exynos 5420 had a lot more volume, so did the SD800 (and it's overclocked rebrand 801). Tegra K1 has 3 wins in tabs with 2 chips, the Mipad, the Nexus 9 and the Shield Tab, all of them low volume, the Mipad on lack of availability outside China ,the others mostly on high to very high prices- the ipad mini gen 2 has better value at 300$ than the Shield and when you can't even compete with Apple something is very wrong. The Android market is what it is because the big brands can't offer any decent buys,there is none. A couple of weeks ago a friend that has been waiting for about a year for a decent tab, just bought a 50$ used one, that's how bad the current tablet offering is (outside China), you can't find anything worth buying even when you really really need one. They make the market by failing to offer compelling products.
The return rates for the 970 following the memory allocation incident have been around 4%. Significant, but far from the embarrassing disaster some people were predicting, most of whom don't even own 970's. Honestly if I owned a 970 I would probably be pretty pissed too, but it wouldn't be enough to make me return the card (it's still a great card), much less consider legal action.
Seriously? People actually returned their cards because of a minor architectural quirk that resulted in a segmented memory address space? There's nothing wrong with the 970, people were sold on it based on real-world performance, thermals and power consumption benchmarks and the retail units that made it into consumers' hands lived up to those benchmarks. The price too was amazing, especially compared to the 980.
Or do people seriously base their hardware buying decisions on esoteric technical specifications like the number of cores, ROPs, theoretical memory bandwidth and the cache hierarchy? Because that's a terrible way to make decisions on performance - one should always benchmark and profile their software to see what makes a difference and what doesn't.
in addition to very few people with legitimate reasons, many people who returned them probably did it to upgrade after trying it out for a month. And others were maybe just scaremonged into it and don't know anything about performance.
So you don't think A SINGLE Nvidia engineer bothered to read any of the reviews? Do you not think a single BMW engineer that has worked on designing a BMW ever reads a car mag?
Wouldn't you after spending countless hours on a project would even be a little bit interested in how others are reviewing it?!
And after reading it and notice the error, don't bother to contact the press and get them to correct their "mistake?"
It would take a sort of fetish interest to be reading the for-the-public technical specifications of a card which you helped design. It's sorta like a professional athlete reading his stats in a video game, only without the ego involved that might make some of them actually care. One can look at how people are reacting to the product without reading the boring parts. Just skip to the benchmarks and conclusions. Of course we don't know if it was done on purpose or not, but things do get missed. The biggest reason in my mind to doubt that it was done on purpose is that it seems pretty stupid to do. How many extra 970 sales at the expense of AMD cards do you think there actually were because people didn't know that half an ROP unit was disabled? The benchmarks are the real selling point.
Look what Mazda did with the RX-8 when they were caught inflating the Hp by around 10. Yes, the car mags already benchmarked the cars and you knew what performance you're going to get, yet they still offered free maintenance or was willing to take back the cars to please their customers. These are $26,000 cars not $330 video cards folks! Yet, Mazda while willing to do this during the years they were losing billions of Yen. Look at Nvidia, flush with cash and they're not willing to take these cards back. Most of us weren't asking for a free upgrade to a 980, just an option for a refund as most retailers/aibs won't take the cards back
Ah, there we have it, a free 980 upgrade exchange...that's why the sour lemonage is there. So let's see what reality is - .5Gb of 4.0GB is one eighth, and we have the GPU, the driver/capabilities, and the RAM as one third each of the package, so one third of one eighth is one 24th, and I'd round to one 25th since there is .5GB just a slow .5, and 1/25th of $300 is Twelve Dollars.
I bet the engineers did after the first reviews went out, and I bet they complained loudly and the marketing department told them to shut up because it was too late. I bet they then told marketing someone will find out, you can't hide this, and marketing said we've hid plenty of stuff in the past. Now there'll be a lot of "I told you so's" going around and marketing people trying to pass on the blame.
And I'd bet no one dared say a word, because they keep their heads down and do their jobs, and I'd bet when it came to any engineer that even caught wind or noticed, they would chuckle and think the idiots will never know, and then they watched all the reviewers and concluded: "Yep they are all so stupid after bragging about their massive expertise, they just have not clue one".
And thus it went on, even though the prior 2 rounds of core releases had fits over 4GB "needed" for "future proofing", and to drive the 970 up beyond 2-3 GB used required ridiculously slow frame rates that "enthusiasts" won't tolerate for even a few minutes, and so SLI udeage or bored users eventually whined about far out scenarios, and finally, after many months... still outlying cases speculating on half broken new games needing patches....
So really, nVidia was right, the morons of the masses of self proclaimed experts are so thick and so clueless and have such a hard time understanding what a fact is (they are "journalistic marketeers" and thus liars and spin artists), they never really knew anything. nVidia had to tell them all.
Revenue isn't profit. But I've yet to hear from someone denied, and you assume they lied. Anandtech doesn't think so. But I guess you're right...LOL. Let me know when anandtech (or anyone) shows performance degrades as sold. This doesn't change the benchmarks, and doubtful many care with that being the case (as noted ~5% and some are due to bad cards etc).
What good would a refund do you for perf? Not much. You're going to buy 980 or be happy you got a 970 for the price you paid. Anandtech still working on a corner case (surely NV had a budget to search for them too if desired). hmm...Whatever.
I want to see the corner case so I can LOL and watch the grunting idiots demand a 980 replacement. I also want the grunters to produce the "case" two or three years from now since they are "future proofed" but aren't now...even capable of playing top games/slam the mem/ as the gpu core pukes out.
What happened to their LTE development? Tegra 4i was launched almost two years ago, had only few design wins and nothing has followed. Are they still committed to phones or is Tegra going to be only higher TDP parts for tablets and cars?
Yeah I wonder what they will do with Icera. It's just sitting there, it seems. No new updates to the modem. Well either they laid all those people off or they left or they have them doing something. As far as the actual modem itself, is it dead?
256mbit memory bus anyone? Lack of thermal/voltage headroom? Less than stellar performance at 4K/DSR/high levels of AA because of bandwidth limitations? TBH I love my Strix 980, and it overlocks decently (not as great as some others) +170 core and +400 memory, but I dunno. I feel you could have got a lot more out of this chip with half decent memory and they are taking their sweet time with the Ti version.
4K=2+ cards (or two chips on one card). Either way, a single chip card is not made for 4K. Call me back when we hit 14nm gpu. Then we'll talk about what you're saying here. fps drops to below playable so why spend more on bandwidth that is wasted when you can't play there? IF going 256bit hurt perf it would not beat 780ti. What you feel and reality are clearly different ;) I think they're selling everything they make, and have no time to cherry up a TI yet. Again though, bloated memory for under 30fps anyway is dumb and complicating things for no reason which would NOT have allowed $100 cut from 780ti. If they gave you that extra lane, you'd be complaining about the extra $100 right? 980 was a pretty good deal over 780ti. Maybe not as an upgrade from it, but I mean if you passed on 780ti, 980 looked pretty awesome.
Guess you were wrong again with your comments a few months ago, huh? Guess price:performance aren't the most important factor when it comes to GPU sales and buying criteria, as once again, the overwhelming majority of the market has rejected AMD products and chosen to go with Nvidia instead.
AMD is so far behind now on features and extras it's an embarrassment to think about it. I hate Physx I hate variable vsync I hate 3D I hate gsync I hate automatic driver setting per game I hate game day drivers I hate recording my gaming real time live streaming and on and on and on.....
I hate EVERYTHING except my 3fps over 121fps on xxxgame at xxresolution with AA off, and the driver crashing after a while is ok, I don't hate that. (AMD gamers manifest manifestoe'd)
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
61 Comments
Back to Article
jackstar7 - Wednesday, February 11, 2015 - link
I wonder how the 970 factored in there.f0d - Wednesday, February 11, 2015 - link
probably made so lttle of a difference it wouldnt even be measurablef0d - Wednesday, February 11, 2015 - link
the refunds probably made little difference (if you were talking about that)or if you were talking about how well it sold then yes it probably made a bundle of money for them
Wreckage - Wednesday, February 11, 2015 - link
Considering the 970 helped give them something like 80% market share in the 4th quarter. I'm sure it was a pretty big factor.Morawka - Wednesday, February 11, 2015 - link
returns were less than 5%, (citation needed) and not all of those 5% were due to the ROP and Memory Allocation. Some just took advantage of a 3 month return window, and some just flat out returned and didn't replace with anything.Klimax - Thursday, February 12, 2015 - link
Here is one of reports on it:http://www.kitguru.net/components/graphic-cards/an...
Samus - Thursday, February 12, 2015 - link
Who the hell would return their 970 because it turned out to have 1/8th less addressable VRAM? Benchmarks are still benchmarks, and the card is a steal for the price, even if by some unlikely scenario games within the next few years adequately utilize 4GB of VRAM.I don't really consider the GTX970 a 4K-ready card, anyway. Not because of the 3.5GB VRAM situation, but because the card can't even run CURRENT games at high detail at 4K let alone future games. I run BF4 at 2560x1440 and can't have everything maxxed without losing 60FPS, and Frostbite 3 is almost a 2 year old engine.
But the card will run 1080P in any game for many years.
edzieba - Thursday, February 12, 2015 - link
"Who the hell would return their 970 because it turned out to have 1/8th less addressable VRAM?"Not because of the vRAM, but the reduced ROP count. But I'm running a very non-typical task, pushing 8k and above (sometimes far above) output sizes for testing supersampling methods for VR. That 7/8 ROP count is nearly a 7/8 drop in pixel output rate (not particularly complex scenes). Hard to test that directly with the move to a 980 as the extra SMs make it an only indirect comparison, but I am seeing that range of performance improvement.
Kjella - Thursday, February 12, 2015 - link
From what I understand the ROPs can feed the shaders all the data they need, it's 14/16th the ROPs and 13/16th the shaders so it actually has more ROPs per shader than the 980. It's only addressing the last 0.5GB that's borked.dananski - Thursday, February 12, 2015 - link
I think it's not even an issue of the technical deficiencies - the benchmarks already take these very minor issues into account. What people are annoyed about is NV apparently lying. I never saw it as more than a marketing mishap though, and was happy recommending a 970 with full knowledge of the controversy.Flunk - Thursday, February 12, 2015 - link
I wonder what percentage of purchasers know what ROPs are?chizow - Thursday, February 12, 2015 - link
Sounds like you may have benefitted from the full spec 980, but I am sure the 970 was attractive to you because it offered roughly 80% of the performance at 60% of the price of the 980? Interestingly, this paradigm has not shifted at all, despite Nvidia restating some other specifications on the 970.FlushedBubblyJock - Sunday, February 15, 2015 - link
Exactly, and we have been over that again, and again, but of course the idiocy in the gamers arenas, even what everyone calls the high end is enormous.If the gaming gooober doesn't "FEEEEEEEEEEEEEEL" good "about his rig", well then no amount of smooth FPS gaming can soothe the mental midget.
Have no doubt many of them will a few years from now fire up a game at 4k, wait 15 minutes, then get the 3.27fps fraps report and say "see, I told you so ! " - no amount of explaining that 3fps is not playable will suffice. The insane nutter butter peanut butter hamwhich kookie needs his ego massage more than life itself.
Klimax - Thursday, February 12, 2015 - link
Not yet. And it will be likely very small. Reports state that only 5% or so are returned. (Also might include other causes)Some e-shops (like Czech CZC) report defect rates if available. (An ASUS 290X DCII has for example ~24% rate, while MSI with 290X got 8%. GTX 970 ASUS DCII got ~%, didn't check any further)
Could be fun time for data mining...
chizow - Thursday, February 12, 2015 - link
@jackstar7, the 970 undoubtedly played a huge role in driving Nvidia to record profits and margins, and the price cuts it forced upon AMD's flagship 290/X parts clearly had the opposite effect on AMD's quarter (greatly reduced revenue, profits, margin on their GPU business).For most buyers, the relevance of Nvidia correcting and restating the 970's specs will never materialize beyond a few bits in an article or on a PDF, because the paper changes do not change the performance level they paid for when they bought the card. I guess if Nvidia's pricing of the 970 at launch seemed too good to be true, it probably was and one of the warts compared to the 980 was exposed.
Is that enough to make people return their cards or pay the premium for the 980? Looks like the answer is no.
tviceman - Wednesday, February 11, 2015 - link
What has been years in coming has finally happened (albeit not under the circumstances either company expected). Nvidia has overtaken AMD in revenue. AMD is now officially second fiddle in both technology and revenue to all their main competitors sans consoles. Even as Tegra continues to flounder, Nvidia continues to prosper. I don't see AMD ever coming back to be a significant force. Their next GPU release might be their last big high-end competitive launch unless Nvidia massively stumbles in future GPU iterations.HighTech4US - Wednesday, February 11, 2015 - link
Tegra is not floundering.http://investor.nvidia.com/common/download/downloa...
Tegra revenue for FY15 was $579 million up 45% from FY14's $398 million.
The data from the article is only for ONE QUARTER and the reason: Tegra Processor sales declined 15 percent from a year ago, driven by the product life cycle of several smartphone and tablet designs.
You should not use one quarter's data as a measurement as there is too much variability from one quarter to the next you should use Y-to-Y.
jjj - Wednesday, February 11, 2015 - link
Listen to their post results call at min 35 and 20 sec http://investor.nvidia.com/eventdetail.cfm?eventid..." between automotive and Shield (our gaming platform) that represents the vast majority of Tegra now "
So they admited SoC sales to others are almost nothing
In other call related news while talking about Tegra they said they have gaming news soon so the March 3 event is Tegra related for sure.
HighTech4US - Wednesday, February 11, 2015 - link
So what they still made $579 million or 45% more than the $398 million from FY14.Expect FY16 to have the same Y-on-Y growth.
jjj - Wednesday, February 11, 2015 - link
No they didn't made 579 million,they had revenue of that much and if you want to look at historic data look at previous years too. I'll get you started by letting you know that in 2012 Tegra revenue was 591 mil, in 2012 Tegra was 764.447 millions revenue.Then look at their SEC fillings and you'll find how much money Tegra is (mostly) losing. In 2011 Tegra op income was 43.736 mil , in 2012 Tegra op income was 40.508 millions and in 2013 Tegra op loses were 268.068 million while in Q1 2014 the op loss was 61.440 million.Tegra as a mobile SoC is dead ,it it gets revived or not it remains to be seen.
TheJian - Friday, February 13, 2015 - link
ROFL. Car revenue from SOCS/infotaintment systems is expected to be 15B by 2017. NV is shaping up to be a leader there and only needs 1/3 of that to more than double the size of their whole company revenue. They only need to crack ~$1B revenue for tegra to be profitable. They are not that far off and we have all been watching the fakes up to x1 which is the FIRST one will a TRUE mobile gpu on it (kepler wasn't built for this, maxwell was but we lost denver due to tsmc for this rev).Tegra is just starting, not ending. I don't believe we'll see the full affect until the chip AFTER X1 though when they put back Denver on 14nm samsung with maxwell in too. Both of those together should be very good. You have to remember Denver was NOT made for 28nm, but TSMC screwed up the schedule forcing them to go with quick IP from ARM instead of remaking changes again to 20nm (after having to already do that to put it out at 28nm). So the real deal got delayed a bit, but we'll see it at 14nm and then I think we'll see it be able to be used in many more phones/tablets paired with samsung modems most likely (since they just hit cat10 while qcom is now seeming behind at cat9). We will also know more about the NV suit vs. samsung in another quarter or two and how that affects pricing on NV getting stuff fabbed over there, or if there's a huge payout over 6yrs like Intel etc, or if it was to get free modems from samsung included etc. Who knows how that works out.
I can't remember the stat group that said auto socs/systems would be $15B, but this year semiconductor automotive crap as a whole is 34B. That's a lot of dough, and with cars doing FAR more in all this soon and across the board, we'll see them use 16-20 socs per car (which is why this market is so huge) with all kinds of driver assistance crap, etc.
http://www.statista.com/statistics/277926/projecte...
The whole ball of wax. I'd have to dig further to find the part that singled out the soc part of that for 2017 but it was 15B estimated (it was an article about NV infotainment systems/ADAS for auto IIRC and the data point was mentioned). They make more from the car "solution" than from a soc sale to a mobile device and that stuff is just starting. They can wait a few revs for their desktop gpu drivers (now used in mobile) to take over gaming on android etc. Cars will more than make up for it shortly.
Intel is losing 4B+ a year on mobile. They aren't going to quit...LOL. They are hoping to shrink their way in, same way NV is hoping to shrink in a modem at some point while keeping great gpu power for gaming and utilizing 20yrs of gaming drivers to help take over anything gaming in mobile (already gaming has taken over all the app stores). The first 5-6 years of cuda were to get them where they are now also (owning 80% of the workstation market and some server stuff). You don't seem to get the LONG game. Intel and NV do. Having said that I think Intel either buys NV or loses that war (and maybe at some point a good portion of notebooks/desktops as sheer numbers drive apps/games up in quality from devs on 64bit arm). To beat the enemy here I think they have to join them (buy NV and make a better ARM than all other arm's...LOL).
NV can lose 2.5B over 5-6yrs of tegra if it makes them 1-3B a year for years to come in the next few years. Profit over revenue I mean, as again, they are chasing an estimated 15B revenue here alone, assuming NO gain in mobile devices. So if they took say 1/3 of that at $5B they're chasing (and they could end up leading it all taking 70-80% like discrete/workstation stuff), they could easily make 1-3B profit on it as they sell a whole solution here. That would only be a 60% margin at 3B profit on 5B or so revenue, more than doubling the revenue of the company and blowing up profits big time. If you don't get the point of tegra, I guess you just don't understand where we're headed. Once gaming takes over mobile even more (with xbox360 and up type gaming quality), you will see NV start to take some real numbers from mobile. Intel is basically paying for any tegra losses...LOL. If samsung gets bit for just as much or more (they make far more than Intel), and of course Intel has to re-up again in 2016 (or go back to court and lose again) they could take it in the shorts for a few more years until auto is at 15B and laugh. I expect a few people to lic NV gpu IP soon (samsung? Maybe apple if they start losing big to 14nm maxwell and beyond). It's a LONG game, not 5yr war. Can you imagine gpu revenue from PC's etc being the smallest portion while cars pulled in even 1B+ PROFIT? ROFL. Quit tegra? You must be kidding. As 64bit apps and far better games amp up they could fight for a piece of the PC desktop/notebook pie for real (way past chromebook crap). Yearly x86 pc numbers vs. ARM numbers are pathetic and getting worse. Wintel is in trouble soon.
Nvidia never made more than ~825mil in a year (2007/2008 depending on how you count quarters), but we're gaining on that now and this isn't including cars exploding, share buybacks that are now getting them to 2007 levels (when stock was ~$38 at ~800mil income, far weaker company), the 1000 companies testing grid, or at some point making something from mobile. Grid and shield each cost 10mil to develop, but may end up yielding huge returns in the end game. I think Jen just laughed at your post as he ponders Tegra 5yrs from now ;) Cuda took 8yrs to get to basically domination. Google's car solution is 150K of high tech gear for self driving crap, while NV plans to do it with 16-20 socs and a dozen cheap cameras (with grid helping) making it affordable for everyone at some point. What is that $400-600 or so in socs, and a few hundered on cameras vs. 150K per car?...LOL - add a few hundred more in software or something etc, the point is far cheaper. Google will have to go NV (someone else? Maybe apple, we'll see how they go) or give it up.
http://ragingbull.com/forum/topic/1021687
You can google the crap they're talking about here if desired.
https://www.abiresearch.com/press/global-advanced-...
Driver assistance crap blowing up huge soon. As stated from 11.1B 2014, $90B 2020, $200B 2024 as it goes mass market shortly. No longer just in a tesla etc. That's a lot of dough to fight for (not all a soc, but it is a huge part of the equation here, predicting everything, dash crap, infotaintment etc). You have ARM, Apple, Google, Nvidia, Qcom, probably Intel at some level all fighting over portions of it all but no denying NV has a decent shot at a good bit of it. NV's tech is a good fit for auto which is why it is no surprise they left mobile to some degree (where they were not a market share leader for now) and went to where it's a wide open HUGE market with no real leader.
jjj - Wednesday, February 11, 2015 - link
The Tegra segment is almost entirely auto, Shield/ Shield Tab, Note and old consoles. You can count their wins on one hand and all are low volume- it's not at all hyperbolic, it's just what it is.If you look at the press release they intentionally avoid mentioning the Tegra brand , they mention it once when talking auto highlights and given the SF event on March 3 , i wonder if they even plan to attend MWC.
tviceman - Wednesday, February 11, 2015 - link
Nexus 9.lucam - Wednesday, February 11, 2015 - link
The most sold tablet of the year, isn't ?jjj - Wednesday, February 11, 2015 - link
And Xiaomi Mipad and some Chromebooks, all low volume.Again i am not assuming or speculating , it's a fact Nvidia itself acknowledges, the Tegra segment was mostly other things not SoC sales in mobile.
darkich - Thursday, February 12, 2015 - link
On terms of design wins, Tegra K1 was the most successful high end Android tablet chip.Can you name me one tablet that uses the Snapdragon 805?
Android tablet space is all about cheap processors from MTK, Allwinner, Rockckip and Intel(given away in tens of millions)
jjj - Thursday, February 12, 2015 - link
805 was a stopgap chip not worth using in anything with the 810 supposed to show up sooner than it did.The Exynos 5420 had a lot more volume, so did the SD800 (and it's overclocked rebrand 801).
Tegra K1 has 3 wins in tabs with 2 chips, the Mipad, the Nexus 9 and the Shield Tab, all of them low volume, the Mipad on lack of availability outside China ,the others mostly on high to very high prices- the ipad mini gen 2 has better value at 300$ than the Shield and when you can't even compete with Apple something is very wrong.
The Android market is what it is because the big brands can't offer any decent buys,there is none.
A couple of weeks ago a friend that has been waiting for about a year for a decent tab, just bought a 50$ used one, that's how bad the current tablet offering is (outside China), you can't find anything worth buying even when you really really need one. They make the market by failing to offer compelling products.
lucam - Thursday, February 12, 2015 - link
By the way Shield tablet is not a win tablets since it's in a house product.Pneumothorax - Wednesday, February 11, 2015 - link
I guess when you're pulling in over $1 Billion in revenue you can't afford to refund a few disgruntled 970 owners who were essentially lied to.f0d - Wednesday, February 11, 2015 - link
as far as i know they got their refundstheres a fair few i know of that got a refund then got a 980 afterwards
dragonsqrrl - Thursday, February 12, 2015 - link
The return rates for the 970 following the memory allocation incident have been around 4%. Significant, but far from the embarrassing disaster some people were predicting, most of whom don't even own 970's. Honestly if I owned a 970 I would probably be pretty pissed too, but it wouldn't be enough to make me return the card (it's still a great card), much less consider legal action.r3loaded - Thursday, February 12, 2015 - link
Seriously? People actually returned their cards because of a minor architectural quirk that resulted in a segmented memory address space? There's nothing wrong with the 970, people were sold on it based on real-world performance, thermals and power consumption benchmarks and the retail units that made it into consumers' hands lived up to those benchmarks. The price too was amazing, especially compared to the 980.Or do people seriously base their hardware buying decisions on esoteric technical specifications like the number of cores, ROPs, theoretical memory bandwidth and the cache hierarchy? Because that's a terrible way to make decisions on performance - one should always benchmark and profile their software to see what makes a difference and what doesn't.
Murloc - Thursday, February 12, 2015 - link
in addition to very few people with legitimate reasons, many people who returned them probably did it to upgrade after trying it out for a month. And others were maybe just scaremonged into it and don't know anything about performance.FlushedBubblyJock - Sunday, February 15, 2015 - link
Their epeen deflated and dejected and no longer confident they trode the dark path and turned it inMikemk - Wednesday, February 11, 2015 - link
They weren't lied to, NVIDIA's left hand (marketing, website, ...) just didn't know what the right hand (Engineering) was doingMikemk - Wednesday, February 11, 2015 - link
And also, the 970 technically *does* have 4GB RAMPneumothorax - Thursday, February 12, 2015 - link
Sorry but that excuse doesn't fly with me. There's just a little too much denial in that.Klimax - Thursday, February 12, 2015 - link
I guess you never had to deal with two separate divisions or groups... (And still no evidence contrary which would prove anything worse then mistake)Pneumothorax - Thursday, February 12, 2015 - link
So you don't think A SINGLE Nvidia engineer bothered to read any of the reviews? Do you not think a single BMW engineer that has worked on designing a BMW ever reads a car mag?Wouldn't you after spending countless hours on a project would even be a little bit interested in how others are reviewing it?!
And after reading it and notice the error, don't bother to contact the press and get them to correct their "mistake?"
The fix is in folks...
Yojimbo - Thursday, February 12, 2015 - link
It would take a sort of fetish interest to be reading the for-the-public technical specifications of a card which you helped design. It's sorta like a professional athlete reading his stats in a video game, only without the ego involved that might make some of them actually care. One can look at how people are reacting to the product without reading the boring parts. Just skip to the benchmarks and conclusions. Of course we don't know if it was done on purpose or not, but things do get missed. The biggest reason in my mind to doubt that it was done on purpose is that it seems pretty stupid to do. How many extra 970 sales at the expense of AMD cards do you think there actually were because people didn't know that half an ROP unit was disabled? The benchmarks are the real selling point.bznotins - Thursday, February 12, 2015 - link
Hah, if the "fix" impacted benchmarks, I would be upset. But guess what -- the benchmarks are the same today as they were before the error.Who really buys cards on tech specs? I buy mine on benchmarks.
If tech specs were all that mattered, Porsche wouldn't be making 78% margins with flat-sixes. People would only be buying V8s.
Pneumothorax - Thursday, February 12, 2015 - link
Look what Mazda did with the RX-8 when they were caught inflating the Hp by around 10. Yes, the car mags already benchmarked the cars and you knew what performance you're going to get, yet they still offered free maintenance or was willing to take back the cars to please their customers. These are $26,000 cars not $330 video cards folks! Yet, Mazda while willing to do this during the years they were losing billions of Yen. Look at Nvidia, flush with cash and they're not willing to take these cards back. Most of us weren't asking for a free upgrade to a 980, just an option for a refund as most retailers/aibs won't take the cards backFlushedBubblyJock - Sunday, February 15, 2015 - link
Ah, there we have it, a free 980 upgrade exchange...that's why the sour lemonage is there.So let's see what reality is - .5Gb of 4.0GB is one eighth, and we have the GPU, the driver/capabilities, and the RAM as one third each of the package, so one third of one eighth is one 24th, and I'd round to one 25th since there is .5GB just a slow .5, and 1/25th of $300 is Twelve Dollars.
That's what you deserve at best, $12.
Dribble - Thursday, February 12, 2015 - link
I bet the engineers did after the first reviews went out, and I bet they complained loudly and the marketing department told them to shut up because it was too late. I bet they then told marketing someone will find out, you can't hide this, and marketing said we've hid plenty of stuff in the past. Now there'll be a lot of "I told you so's" going around and marketing people trying to pass on the blame.dragonsqrrl - Friday, February 13, 2015 - link
Yes, I bet that's exactly what happened...FlushedBubblyJock - Sunday, February 15, 2015 - link
And I'd bet no one dared say a word, because they keep their heads down and do their jobs, and I'd bet when it came to any engineer that even caught wind or noticed, they would chuckle and think the idiots will never know, and then they watched all the reviewers and concluded: "Yep they are all so stupid after bragging about their massive expertise, they just have not clue one".And thus it went on, even though the prior 2 rounds of core releases had fits over 4GB "needed" for "future proofing", and to drive the 970 up beyond 2-3 GB used required ridiculously slow frame rates that "enthusiasts" won't tolerate for even a few minutes, and so SLI udeage or bored users eventually whined about far out scenarios, and finally, after many months... still outlying cases speculating on half broken new games needing patches....
So really, nVidia was right, the morons of the masses of self proclaimed experts are so thick and so clueless and have such a hard time understanding what a fact is (they are "journalistic marketeers" and thus liars and spin artists), they never really knew anything.
nVidia had to tell them all.
TheJian - Friday, February 13, 2015 - link
Revenue isn't profit. But I've yet to hear from someone denied, and you assume they lied. Anandtech doesn't think so. But I guess you're right...LOL. Let me know when anandtech (or anyone) shows performance degrades as sold. This doesn't change the benchmarks, and doubtful many care with that being the case (as noted ~5% and some are due to bad cards etc).What good would a refund do you for perf? Not much. You're going to buy 980 or be happy you got a 970 for the price you paid. Anandtech still working on a corner case (surely NV had a budget to search for them too if desired). hmm...Whatever.
FlushedBubblyJock - Sunday, February 15, 2015 - link
I want to see the corner case so I can LOL and watch the grunting idiots demand a 980 replacement.I also want the grunters to produce the "case" two or three years from now since they are "future proofed" but aren't now...even capable of playing top games/slam the mem/ as the gpu core pukes out.
Taneli - Wednesday, February 11, 2015 - link
What happened to their LTE development? Tegra 4i was launched almost two years ago, had only few design wins and nothing has followed. Are they still committed to phones or is Tegra going to be only higher TDP parts for tablets and cars?dragonsqrrl - Wednesday, February 11, 2015 - link
They dropped smartphones around the time the K1 launched. They're focusing on automotive and ~5W TDP mobile devices, such as tablets and Chomebooks.Yojimbo - Thursday, February 12, 2015 - link
Yeah I wonder what they will do with Icera. It's just sitting there, it seems. No new updates to the modem. Well either they laid all those people off or they left or they have them doing something. As far as the actual modem itself, is it dead?darkfalz - Wednesday, February 11, 2015 - link
Explains why the GTX 980 was nerfed yet overpriced - NVIDIA don't even need to break a sweat.dragonsqrrl - Wednesday, February 11, 2015 - link
How was the GTX980 nerfed or overpriced at launch? It came in $100 cheaper than the 780Ti and outperformed it by ~12%.Klimax - Thursday, February 12, 2015 - link
And was full GM204.darkfalz - Thursday, February 12, 2015 - link
256mbit memory bus anyone? Lack of thermal/voltage headroom? Less than stellar performance at 4K/DSR/high levels of AA because of bandwidth limitations? TBH I love my Strix 980, and it overlocks decently (not as great as some others) +170 core and +400 memory, but I dunno. I feel you could have got a lot more out of this chip with half decent memory and they are taking their sweet time with the Ti version.TheJian - Friday, February 13, 2015 - link
4K=2+ cards (or two chips on one card). Either way, a single chip card is not made for 4K. Call me back when we hit 14nm gpu. Then we'll talk about what you're saying here. fps drops to below playable so why spend more on bandwidth that is wasted when you can't play there? IF going 256bit hurt perf it would not beat 780ti. What you feel and reality are clearly different ;) I think they're selling everything they make, and have no time to cherry up a TI yet. Again though, bloated memory for under 30fps anyway is dumb and complicating things for no reason which would NOT have allowed $100 cut from 780ti. If they gave you that extra lane, you'd be complaining about the extra $100 right? 980 was a pretty good deal over 780ti. Maybe not as an upgrade from it, but I mean if you passed on 780ti, 980 looked pretty awesome.silverblue - Thursday, February 12, 2015 - link
The 965M was launched on 5th January, whilst the 960 came out on 22nd January. I doubt either of those would've been a factor in NVIDIA's revenue.chizow - Thursday, February 12, 2015 - link
@CreigGuess you were wrong again with your comments a few months ago, huh? Guess price:performance aren't the most important factor when it comes to GPU sales and buying criteria, as once again, the overwhelming majority of the market has rejected AMD products and chosen to go with Nvidia instead.
In summary: maybe you are doing it wrong?
FlushedBubblyJock - Sunday, February 15, 2015 - link
AMD is so far behind now on features and extras it's an embarrassment to think about it.I hate Physx
I hate variable vsync
I hate 3D
I hate gsync
I hate automatic driver setting per game
I hate game day drivers
I hate recording my gaming real time live streaming
and on and on and on.....
I hate EVERYTHING except my 3fps over 121fps on xxxgame at xxresolution with AA off, and the driver crashing after a while is ok, I don't hate that. (AMD gamers manifest manifestoe'd)
haplo602 - Thursday, February 12, 2015 - link
I don't think Nvidia is able to report results for Q4FY2015 yet. Maybe Q4FY2014 but sure not FY2015.fivefeet8 - Monday, February 16, 2015 - link
Fiscal reporting periods often different that what logically would be thought of.