NVIDIA Tegra Architecture
- Thread starter french toast
- Start date
-
- Tags
- nvidia
Besides that if I say it again it's going to be deemed as "irrelevant". Without any serious design wins and/or sales volumes it's all but a theoretical excersize on paper. From the timeframe T4 will end up in those few actual devices at normal shelf availability until Apple's next tablet it's going to take how long?
Last but not least much to my disgress Kishonti has removed 4xMSAA tests. Tegra4 supports this time 4xMSAA but I doubt it comes with an as small performance penatly as with the entire tile based competition. In a 720p resolution an iPad4 or any recent Adreno will give 4xMSAA for free, while on T4 you'll be probably looking at a quite sizeable performance drop.
Shield/T4 is again the magic wand for all mobile 3D problems that might occur; good enough, sufficient and ultra-efficient. Just about everything that might justify a 180 degree turn in at least half a year from device availability to next generation SoC announcement.
http://www.fudzilla.com/home/item/31341-nvidia-confirms-high-end-tablet-market-is-slowing
Kudos Jensen; in hindsight such acknowledgments are completely useless. Allwinner alone sold more tablets for the past year than Intel and Qualcomm combined. Could you leave the wheel of the Tegra department to someone that has the right sniffing nose for future market trends for a change?
http://www.fudzilla.com/home/item/31342-nvidia-expects-to-gain-share-on-haswell
Now that I can believe if highest end Logan tries to bite in where lowest end Haswell is supposed to target.
Right, 720p resolution on a 5 inch diagonal screen is "substantially" lower resolution for the human eye to resolve.
Like it or not, NVIDIA made a great tradeoff here. iOS games that are not "confined" to native resolution would be using a resolution even lower than 720p.Right, 720p resolution on a 5 inch diagonal screen is "substantially" lower resolution for the human eye to resolve.
A lower native resolution is only a very minor performance advantage. Plenty of iOS games use non-integer scaling ratios, just like a wide variety of console games. There's a slight loss in image quality and a slight impact in performance in having to go through another scaling step but it's not the disadvantage you make it sound like.
In fact, several top tier console games use sub-720p resolutions (and several PSVita games even use less than native resolution), so if Shield were really following suit in delivering a console like experience it'd probably do so too. But since it'll probably get little more than a few enhanced Android games I doubt this will apply.
Last edited by a moderator:
http://www.fudzilla.com/home/item/31361-high-end-phones-belong-to-samsung-and-apple
Another reality check from NVIDIA; it's reached a point that I'm starting to actually worry
Another reality check from NVIDIA; it's reached a point that I'm starting to actually worry
What NVIDIA is saying is that, outside of Apple's ipad, most consumers will be unwilling to pay $500 for a tablet, and outside of Apple's iphone or Samsung's Galaxy S phone, most consumers will be unwilling to pay $600 [unsubsidized] for a smartphone. So what NVIDIA is trying to do is to provide similar (or better) features and performance compared to Apple and Samsung high end tablets and smartphones, but at a fraction of the cost to the end user. Makes sense to me.
That's what you think they were saying? It sounds more to me that they're saying there's no point in them having a high-end phone SoC altogether.
Given that nVidia doesn't make smartphones I don't see how they alone would enable the market to have competitive high end phones at much lower prices. Even if they totally took a dive on margins the SoC cost is still too small a fraction of the entire device. Cheaper devices have to cut costs somewhere else, and/or accept lower margins.
Besides, it's not like nVidia hasn't won Samsung phones before, including some Galaxy S2s, you'd think they'd be aiming for that if they had a good SoC for it.
Given that nVidia doesn't make smartphones I don't see how they alone would enable the market to have competitive high end phones at much lower prices. Even if they totally took a dive on margins the SoC cost is still too small a fraction of the entire device. Cheaper devices have to cut costs somewhere else, and/or accept lower margins.
Besides, it's not like nVidia hasn't won Samsung phones before, including some Galaxy S2s, you'd think they'd be aiming for that if they had a good SoC for it.
If Apple and Samsung will the be only ones left standing for high end SOCs, doesn't Qualcomm have just as much to lose, if not more? Nvidia doesn't exactly have a large share to begin with...
Qualcomm is in plenty of Samsung phones today.
ST-E actually got a lot of wins for lower end Samsung phones last generation or two.
It remains to be seen how much Samsung is going to more aggressively push their own SoCs into their phones. I expect they'll have to do something to relieve the big fab volume hit they'll take from losing Apple.
Tegra 4 + separate Icera i500 modem is arguably a "high end" SoC for tablets and high end smartphones. But the market opportunity for $500 tablets for anyone outside of Apple is limited, and the market opportunity for $600 [unsubsidized] smartphones for anyone outside of Apple and Samsung is limited too, irrespective of what type of application processor is used.
NVIDIA's "Kai" reference platform was a mainstream tablet platform that translated directly into the Google Nexus 7. NVIDIA's "Phoenix" reference platform is a mainstream smartphone platform that manufacturers can use as a blueprint to help create mainstream smartphone devices.
The trend is for companies such as Apple and Samsung to become more and more vertically integrated over time, not less.
It's got the high end part down well enough, it's just falling a bit short of the "for phones" part. And nVidia is - like usual - turning their market weakness into claiming they didn't want it anyway.
The argument does have some merit. But it means they won't be selling high end SoCs in phones at all outside of an insignificant number of Tegra 4 phone wins. It doesn't mean that they'll get vendors to sell Tegra 4 phones at bargain prices. If the ZTE phone is any indication they'll probably have bigger screens to back bigger batteries, and will probably cost more because of that, not less.
That doesn't have anything to do with what I said, unless you think nVidia has the secret to designing the same quality hardware at a fraction of the price as others. Or you think that they'll be mass producing these reference phones themselves to sell at low or even negative margins like Intel is doing.
Perhaps, but I don't think Samsung is there yet for this generation. I'm keeping the comments in context with what nVidia wants to do this year, not really their long term plans.
Project Shield price announced, $349.
Not bad but it's a 5-inch display? Is the controller detachable?
Can't see it faring better than other dedicated portable gaming devices of recent vintage, particularly those which are priced way above Nintendo's devices.
If Google chooses something other than a Tegra 4 for its updated Nexus devices this year, will Tegra 4 ship on fewer devices than Tegra 3 did?
Not bad but it's a 5-inch display? Is the controller detachable?
Can't see it faring better than other dedicated portable gaming devices of recent vintage, particularly those which are priced way above Nintendo's devices.
If Google chooses something other than a Tegra 4 for its updated Nexus devices this year, will Tegra 4 ship on fewer devices than Tegra 3 did?
The reality is that Tegra 4 has always been geared more towards tablets than smartphones, while Tegra 4i has always been geared more towards smartphones than tablets, but both SoC's can be reused and reconfigured to work in something outside of their main area of strength. Anyway, like I said earlier, the market opportunity for "high end" and "high priced" smartphones/tablets is limited outside of Apple and Samsung. That doesn't mean that NVIDIA will "abandon" R&D spending on "high end" ultra-mobile SoC's, but rather that NVIDIA will attempt to provide high end performance and features at mainstream prices (whether that is through frugal engineering or lower-than-Samsung margins or some combination of the two).
Tablets and smartphones are built first and foremost around the SoC, so perf/mm^2 (and hence perf/$) matters with respect to product positioning (at least for anyone who plans to sell a mobile device for profit). If Tegra 4i can come even close to much larger SoC's such as S800 in terms of CPU/GPU performance, at a fraction of the SoC die size and at a fraction of the SoC cost, that would be huge in helping to enable high end smartphone performance at mainstream smartphone prices. It is much more practical to design a mainstream smartphone around a 60mm^2 SoC than a 120mm^2 SoC.
Last edited by a moderator:
The 5" display is integrated with the controller and flips open at up to a 180 degree angle, but is not removeable.
Market reception to Shield is unknown, but Shield can do essentially the same things as any other pure Android WiFi device, so it doesn't need to be used as a dedicated gaming device per se. A better way to describe Shield is as an Android device that is optimized for handheld mobile gaming.
The fact that retail pricing on Shield is set at $349 USD is further evidence of NVIDIA's goal to provide high end performance at mainstream (or close to mainstream) prices.
Last edited by a moderator:
Early signs are that the S800 will be faster than a Tegra 4. Whether we should believe that or not, I don't know. We'll have to wait for proper shipping devices to see how things compare.
BTW, I expect NVIDIA to sell roughly 100 000 Shields by the end of the year. Not much in the grand scheme of things.
Actually the expectation is that Tegra 4 will outperform S800 in most benchmarks (including SPECInt2000, Sunspider, Web Page Load, WebGL Aquarium, Google Octane, Geekbench, Kraken, Vellamo Metal, etc.), while Tegra 4i will be up to 80% as fast as Tegra 4. S800 will probably have an advantage over Tegra 4 in Coremark and Dhrystone, but not much else performance-wise. If you are referring to the leaked GFXBench 2.7 scores, note that this generation of ultra mobile devices are unplayable at 1080p Offscreen settings.
Incremental R&D expenditure for Shield is next to nothing compared to Tegra as a whole. At worst, Shield is an opportunity for NVIDIA to showcase new Tegra technology each year, and at best Shield is an opportunity for NVIDIA to provide a more optimal gaming experience on Android for end users.
Last edited by a moderator:
Interesting device for one of the first Tegra 4 releases:
http://www8.hp.com/us/en/ads/x2/slatebook-x2.html
Not available until August, however.
http://www8.hp.com/us/en/ads/x2/slatebook-x2.html
Not available until August, however.
french toast
Veteran
Ill believe it when I see it, in a tablet with comparable ram I think they will trade blows in different areas, tegra 4 winning most integer benchmarks, s800 winning on floating point, multithreading, power consumption and graphics.
You put too much stock on nvidia marketing slides.
Similar threads
