We sat down with Nick Stam, Director, Technical Marketing of Nvidia to talk about some of the questions that had developed regarding the various product announcements that had been made prior to and during this year?s CES. We wanted to make sure that we cleared up any uncertainty and gotten some solid numbers and answers.
Tegra 4 Questions Answered
Our first questions primarily came from Tegra 4?s architectural design and the decisions made to get it to where it is supposed to be released. Our first question was about the lack of OpenCL/CUDA support, which we reported on, Nick stated that OpenCL and CUDA weren?t needed at the time of design and that is the primary reason why it wasn?t implemented. Their belief is that considering the refresh cycle of mobile SoCs and the current state of mobile gaming and applications, OpenCL and CUDA are not quite necessary. We also asked about the GPU and found out that the Tegra 4?s GPU shader is not unified and this decision was made in order to optimize for power.
That isn?t to say that unified shader architecture may come in later SoC GPUs, but their design philosophy for Tegra at the time of design does not reflect a need for a unified architecture. Their focus is on power, so we will have to see how the Tegra 4 compares to its competitors in terms of GPU power and consumption when it becomes available. They already have one named OEM onboard, Vizio, but we have a feeling that more will be announced at Mobile World Congress, as that appears to be Nvidia?s cadence – technology launch at CES in January, product launch at MWC in February.
We also learned Tegra 4 is expected to clock in at around 1.9 GHz, possibly up to 2GHz which should enable some impressive performance. Also, as has been stated before, this will be a five-core (4-Plus-1) design -where one core will be a low-power core and will be designed as a multi-purpose core in order to keep power consumption and idle power down.
Project Shield, What Can It Do?
Following our chat about Tegra 4, we obviously had a ton of questions about Project Shield that we wanted answered. The first question was about the clock speed of Shield and we were told that they don?t officially have figures for Shield yet, but we can expect a ballpark around the official figures above. It will have 2GB of RAM, without disclosing the type of RAM, however, it was confirmed that Tegra 4 supports dual-channel LPDDR2, DDR3L and LPDDR3. In terms of power consumption, depending on the gaming scenario users can expect between 5 to 10 hours of gaming on a 38 Watt-hour battery. Streaming directly through the Shield from a gaming PC over Wi-Fi should actually yield between 20-30 hours of gaming, which is actually fantastic in our opinion. This only goes to show the balance of power consumption between a Tegra 4 and the 720P 5" S-IPS LCD screen.
Skip to 2:54 if you want to see the actual device
The Shield will NOT have a camera, so anyone hoping to use it for in-game video chat will be disappointed. The device is going to have passive cooling as illustrated from the video that Nvidia showed during their CES 2013 press conference. It will run a stock version of Android 4.2, which means it will be very hackable and enable third party app developers to go crazy with custom ROMs and the like. We believe that this aspect of the Shield will be its saving grace if the final product is not considered successful by the mass market. Hackers will find a use for it and make it popular within the enthusiast community if it doesn?t gain widespread adoption and Nvidia might end up OEMing it to a partner. These are all hypotheticals if the Shield isn?t initially successful, though.
Moving on, the GeForce GTX 650 and above will support game streaming, as will notebooks with mobile GeForce GTX 660s and above. This will mostly be done through the graphics drivers and handled through the network to the device directly. In addition to that, the Shield will support Miracast for beaming Android games to your TV. The Shield can also stream from the cloud using Nvidia GRID once that service goes online, which should be around the same time as the Shield. Nvidia says that Shield availability should be towards the tail end of Q2 (launch at E3?) and pricing has not quite yet been established, but they?re asking the community what they?re willing to pay.
Nvidia’s CEO playing a game on a 4K, upscaling from 1080P
The Shield can do 4K video decoding, we asked about playing 4K video games from a PC to the Shield, and Nick wasn?t 100% sure on that but we think it should be fairly possible to stream a 4K game from a PC to the TV using HDMI 1.4a. Do note, however, that HDMI 1.4a is only capable of delivering 4K video at 24 FPS, so 4K game streaming is unlikely to be any good in terms of frame rate. So, as it stands the Shield will be for playing 4K video content even though it has a fairly limited amount of onboard memory, which is what the SD card slot is for.
We also spoke to Nick from Nvidia about the possibility of something like a 3G/4G modem being built into the device and he didn?t seem like it wouldn?t be an option in the future. Obviously, they haven?t said anything officially, but the idea of a 3G/4G modem does open up the ability for Nvidia to sell an Icera i500 modem along with a Tegra 4 in a single device. In addition to that, and most importantly, it enables carriers to sell the Shield on a contract and to allow Nvidia to charge significantly more for the 3G/4G version.
Icera?s i500 Modem Detailed
Talking about modems, we asked Nvidia for more details about the Icera i500. The Icera i500 is a soft modem that Nvidia claims to be very low power due to being significantly smaller than conventional modems in terms of die size. This is possible because of the fact that it is built on TSMC?s leading edge 28nm HP process, which is supposed to enable low power consumption and smaller dies. The i500 is made up of 8 individual cores which are supposed to drive the soft modem and can be individually gated to save power. This modem will initially be a category 3 LTE modem which makes it capable of a maximum of 100 Mbps downlink. Later on, Nvidia will release a software update which will upgrade it to a category 4 LTE modem, enabling downlink speeds of up to 150 Mbps. It will also support VoLTE has had been part of Icera?s capabilities in the past, in partnership with D2 Technologies. We really like the approach that Nvidia is taking with Icera and we?re interested to see how long it will be until Nvidia integrates their modems into their SoCs as their biggest competitor, Qualcomm, has done.
e="0" hspace="0" width="689" height="456" title="" />
GeForce Looking Into 2013
We also wrapped up our conversation with Nvidia with talking about the GeForce Experience, which was the first thing that Nvidia announces at CES this year. This program is designed to improve the overall gaming experience that gamers have based on configurations of hardware and available settings. First we asked about how it would be obtained, which it currently is in beta. In order to use it, you must download a separate app, and then install it. It then integrates itself into the drivers afterwards. It can be disabled at any point. Nvidia has gamers run through games on different systems and capture the overall experience to find the optimal settings for your configuration. We asked Nick if power consumption had been taken into account in GFE, but it is currently not a parameter of GFE as Kepler is a fairly efficient architecture that runs relatively cool. They said that resolution takes precedence over everything else in terms of settings weighting different settings in terms of priority and performance. Currently, GPU Boost (a Kepler feature) does have a frame rate limiter, but it is not integrated into GFE nor part of it at the current moment, yet still an additional performance option.
Oh, and we asked about desktop GPUs, and we were told that we should expect something this year based on Nvidia?s previous statements regarding GK110?s path. That means that we likely won?t see Maxwell products this year, but we probably will see GK110 based GeForce and Quadro cards.
Overall, Nvidia did a pretty good idea of answering our follow-up questions about their 2013 products and what we?re really waiting for is a price on Shield, GK-110 GeForce cards, and Tegra 4 powered smartphones and tablets. We can?t wait to get these devices in our hands to validate or challenge Nvidia?s claims they?ve made and to see how they stack up against the competition. 2013 will certainly be an interesting year for Nvidia and their competitors, old and new.