NVIDIA Tegra Architecture
- Thread starter french toast
- Start date
-
- Tags
- nvidia
well NV is a profitable company with more than 4 billion of cash flow in the bank and without any debt. A financial situation that lot of semi conductor company will die for.
seriously, you really think NV will not be there in 2015 ?
no idea of the volume, but during last conf call, NV said they have 2 billion of multi-year sales contract in hands for Tegra in automotive market
A $310 million projected drop in Tegra consumer business revenue relative to last fiscal year sounds "optimistic"? I don't think so. And Tegra doesn't drag down corporate margins as much as you think (it is only a little bit lower than corporate average). Even at 45% gross margin the difference in COGS is only $25 million. The ironic thing is that, even though Tegra 4 was at least 3-6 months behind schedule and not contributing to revenues while Tegra 3 production was ramping down, the projected loss for Tegra in Fiscal Year 2014 is far less than the actual loss for Tegra in Fiscal Year 2013, which goes to show the advantage of heavily leveraged investments moving forward. Along that theme, considering good timing for Logan (and presumably Parker), and considering the year-over-year doubling in Tegra automotive over the next few years, Tegra should be nicely profitable in Fiscal Year 2015 and Fiscal Year 2016. Last but not least, any GPU licensing deals that come to fruition will actually raise Tegra's gross margins even closer to corporate average. In other words, Tegra will contribute to profitability without significantly dragging down corporate gross margins, which is a very good thing.
Last edited by a moderator:
Being hopeful for Tegra 5 is one thing, but claiming Tegra 4 is one of the highest performing SoCs for graphics in the market is a bit off.
Comparable products don't use cooling fans, so Shield can't be counted. Tegra 4 tablets perform behind last year's iPad in Gfxbench's T-Rex test and well behind Qualcomm's smartphone/phablet SoC which launched around the same time as nVidia's newest application processor. Depending upon when products with the Exynos 5420 and its new Mali configuration hit the market, direct comparisons to it may be warranted as well and have the potential to further diminish Tegra 4's competitive standing.
Comparable products don't use cooling fans, so Shield can't be counted. Tegra 4 tablets perform behind last year's iPad in Gfxbench's T-Rex test and well behind Qualcomm's smartphone/phablet SoC which launched around the same time as nVidia's newest application processor. Depending upon when products with the Exynos 5420 and its new Mali configuration hit the market, direct comparisons to it may be warranted as well and have the potential to further diminish Tegra 4's competitive standing.
So this means what, over the next few years there will be like 40+ million cars with Tegra in them? Seriously, how is this possible? Especially when the named list only includes Volkswagen, Audi, and Tesla. Volkswagen has done very well selling ~8m vehicles, but I doubt that many of them are the luxury sort with high end media systems. Audi sells barely 1m a year and Tesla is not even worth mentioning. But maybe there's something I don't know about how much money nVidia makes per vehicle and it's somehow about more than selling an SoC, maybe they're selling entire boards or something.
Think logically about what you are saying. In shipping tablets/phablets, Tegra 4 is within ~ 10% of the GPU performance of A6X and within ~ 20% of the GPU performance of S800 (at least on any graphics benchmark that is even remotely smooth with playable framerates on these ULP GPU's). These are relatively small differences (and small enough where some or even most of the difference can be made up with newer software drivers; Shield actually gained ~ 15% GLBench 2.5 performance between May and August, and that was likely due to software updates). Obviously T4's GPU is not the fastest ULP GPU, but to suggest that it is not one of the fastest ULP GPU's at this time is disingenuous at best (and if that is the case, then you might as well say the same thing about the A6X GPU because it is behind the Adreno 330 GPU too). And obviously T4-powered products still have a very fast quad-core CPU with super smooth UI performance, extremely fast web browsing, and very fast application load times too (and I should know too, because I actually own a T4-powered product).
Last edited by a moderator:
Automotive design cycles and contracts last for many years. NVIDIA has $2+ billion in Tegra automotive revenue lined up over the next ~ 6-7 years. NVIDIA's partners include not just Volkswagon, Audi, and Tesla but also BMW, Mini, Porsche, Peugot, Lamborghini, Rolls Royce, Citroen, Fiat, Maserati, Bentley, Skoda, Lancia, Seat, etc. and the list is growing too. These automakers are not looking for anything that is merely modest and competent when it comes to an SoC or anything else that they place in their vehicles.
Don't expect to get a reasonable answer from the gentlemen here; all you get is the same marketing/PR nonsense repeated in every possible fashion.
At worst they'll just tell you that the infotainment system costs as much as half the engine to justify the figures; if that doesn't work then they'll include the next couple of decades as if firms make such long terms plans. Just as much as the supposed "exclusivity" for the VW group while for some strange reason Jacinto SoCs should then also be developed by NV.
You are going too far. It's not "marketing/PR nonsense", its real information from investor conference call. As much you hate NV, JHH cannot lie to investors.
As long as I'm not insulting you as a member or am misbehaving I am entitled to my own opinion as much as you'd think it's going too far and no you don't have to like it either. And yes my "negative sentiments" for NV are well known around here. Repeat it a couple of times here on the B3D boards and it'll become the joke of the day.
you can say whatever you want, but your freedom of speech doesn't allow you to spread FUD, especially regarding official financial information of a company listed on stock exchange. it's called defamation and can be punished by law.
Sue me. In the meantime it might be a bit more fruitful to try to answer Exophase's quite legitimate questions.
its not my job to sue people. I just gave you a friendly advice.
Regarding exophase questions, I don't know exactly. From investor conf call, JHH said Tegra automotive market will represent 200M revenue this year and will double next year. But without knowing what NV sells (only the SoC ? a complete board ? with or without custom software dvpt ?) and for what ASP, we don't have enough information to understand this business.
See that's exactly my problem with the whole story and not it's not for anyone to take personal: no one has sufficient information on topics like that, nor can form an opinion of his own unless there's something relevant listed in some public statement or conference call of any company?
Now I am gladly willing to listen if anyone has to say anything about it even if it's just a guess or mere speculation. The problem at hand is that in order for any of the claimed numbers to get reached (even over a multi-year timeframe) is that each automobile manufacturer doesn't sell all that many vehicles on a yearly basis for the numbers to add up.
Now I am gladly willing to listen if anyone has to say anything about it even if it's just a guess or mere speculation. The problem at hand is that in order for any of the claimed numbers to get reached (even over a multi-year timeframe) is that each automobile manufacturer doesn't sell all that many vehicles on a yearly basis for the numbers to add up.
french toast
Veteran
@ams whilst I admire you defense of nvidia, you seem to be missing the point.
A successful mobile soc has to be balanced across many points, not just outright performance as you keep cherry picking...even if peak cpu performance is slightly better and gpu is 'only' 20% worse...what good is tegra 4 is it has to consume 5-10w to achieve that?
That is a likely a heavy usage scenario but that is ulv haswell territory....whats stopping intel just taking a price hit and making haswell ulv available for smartphones? Aside from margins...that is far to high a power consumption and heat dissipation for a smartphone...its useless having a mobile soc that is not efficient no matter what the peak performance is, the fact that it offers weaker gpu performance and only slightly better cpu just compounds the issue.
You keep saying tegra 4 was intended to be a tablet soc only...I very much doubt that somehow, do you really think nvidia designed tegra 3s replacement with the aim to lose its smartphone market share when from the very start tegra competed in that very market? No I don think so.
Tegra 4 was designed to be a direct replacement for tegra 3..like for like, to be used in tablets and high end smartphones, im guessing it came out of the fab sucking far too much power and dissapating far to much heat....in short it was designed poorly, with an outdated gpu and cpu cores to big for the job..I mean really, 5 cortex a15s is laughable for a smartphone, they could have clocked the a15s lower yes, but im guessing that is also useless when the inefficient gpu is consuming much more power than they planned also...think about it...6x the alus, increased clock speed and a dual channel memory controller...28nm is sure as hell not going to ofset the difference, and thats not including the a15s.
Tegra 4i was not meant to be a direct tegra 3 replacement for smartphones...it was aimed at the midrange from the very start, so tegra 4 is basically a flop, its the bulldozer of the mobile world.
There I said it.
A successful mobile soc has to be balanced across many points, not just outright performance as you keep cherry picking...even if peak cpu performance is slightly better and gpu is 'only' 20% worse...what good is tegra 4 is it has to consume 5-10w to achieve that?
That is a likely a heavy usage scenario but that is ulv haswell territory....whats stopping intel just taking a price hit and making haswell ulv available for smartphones? Aside from margins...that is far to high a power consumption and heat dissipation for a smartphone...its useless having a mobile soc that is not efficient no matter what the peak performance is, the fact that it offers weaker gpu performance and only slightly better cpu just compounds the issue.
You keep saying tegra 4 was intended to be a tablet soc only...I very much doubt that somehow, do you really think nvidia designed tegra 3s replacement with the aim to lose its smartphone market share when from the very start tegra competed in that very market? No I don think so.
Tegra 4 was designed to be a direct replacement for tegra 3..like for like, to be used in tablets and high end smartphones, im guessing it came out of the fab sucking far too much power and dissapating far to much heat....in short it was designed poorly, with an outdated gpu and cpu cores to big for the job..I mean really, 5 cortex a15s is laughable for a smartphone, they could have clocked the a15s lower yes, but im guessing that is also useless when the inefficient gpu is consuming much more power than they planned also...think about it...6x the alus, increased clock speed and a dual channel memory controller...28nm is sure as hell not going to ofset the difference, and thats not including the a15s.
Tegra 4i was not meant to be a direct tegra 3 replacement for smartphones...it was aimed at the midrange from the very start, so tegra 4 is basically a flop, its the bulldozer of the mobile world.
There I said it.
Well, that's the problem: we don't know. Even though that testing procedure is much more relevant to measure useful gaming performance than a single run is.You mean after 10 looped runs of 3dmark where the CPU was throttling more heavily than the GPU?
ltcommander.data
Regular
http://www.anandtech.com/show/6472/ipad-4-late-2012-review/7
Anandtech loops GLBenchmark until the battery is exhausted for their 3D battery life benchmark and for the iPad 4 they also ran Infinity Blade II for an hour to measure temperature. Now they don't explicitly state the performance doesn't decrease over time, but you'd hope they would mention it if it were the case.
Similar threads
