Hey Guys,
Our community
really put NVIDIA to work with a monster-sized question list! It was so huge that we had to split the interview into two parts, and we're pretty sure that you guys will agree: we saved the best for last. There’s a LOT
of meat in this interview! Nvidia had a
handful of their specialized guys answering your questions to assure you’d get
the most comprehensive answers to each Q.
That said, keep in mind that the guys aren’t able to talk about
undisclosed products due to company policy. Still, they did an awesome job of getting the best possible answers for everything they could. 
Ok guys, here's Part II of the NVIDIA interview. We hope you enjoy! 
1. Xpert: ATi has released
Quad-Crossfire, and you can get it rather cheap (4 HD3870's or 4 HD3850's
aren't that much money), is there going to be an answer to this from the
NVIDIA front (Like Triple or Quad SLi compatible 9800GT's or 9800GTS's)?
Andrew Fear, SLI
product manager: NVIDIA is always investigating new solutions for our
end users that offer maximum flexibility when building an SLI PC. One of the reasons we have not enabled 3-way
or Quad SLI on all GPUs is that typically the target resolutions and
performance for those products is less than 3-way and Quad SLI, which is
2560x1600 with 4x anti-aliasing and higher.
In addition, the price/performance benefit on having
this feature on GPUs that cost $249 MSRP doesn’t always make sense. Many users have told us that rather than
spending $750 on three GPUs they would prefer to have two GPUs that cost the
same, enable the same performance, and allow you to have an additional PCI-E
slot, like two GeForce 9800 GTX. We’d love to get your feedback: is having
3-way or Quad SLI on performance class products such as the GeForce 8800 GT
something you would want? Please post your comments in the forum. Thanks!
2. Xpert: Why are you still making gpus
with freaking 64bit memory interface? When is NVIDIA going to switch to 512-Bit
interfaces for their GPUs?
Ujesh Desai, general manager of desktop
GeForce products: NVIDIA is a worldwide company and there is still demand for 64-bit
SKUs in parts of the world.
3. Xpert: When are you going to switch
to GDDR4, or are you waiting for GDDR5?
Ujesh Desai: We constantly
evaluate memory technology and right now GDDR3 delivers the performance we need
for our products.
4. Xpert: When will NVIDIA try catching
up to ATi when it comes to video and movie rendering and quality? Are
you interested in designing a GPU dedicated to video encoding/rendering?
Patrick
Beaulieu, video technologies product manager:
NVIDIA PureVideo HD technology is the
best solution on the market for enjoying high quality video playback on your PC.
A number of independent sites scored PureVideo HD higher than ATI’s competing
technology in their tests with HQV and HQV HD. PureVideo HD also provides
unique features like Dynamic Contrast and Tone Enhancements, dual-stream decode
acceleration and Blu-ray playback in AeroGlass mode.
For more
information, go to http://www.nvidia.com/page/purevideo_hd.html
As for video encoding, we’re working on some
exciting technology that will dramatically speed up encoding and transcoding.
Stay tuned!
5. Xpert: When do you plan to fix the dreaded nvlddmkm has
stop responding and has closed error in Windows Vista?
Chris Daniel, software product manager: Let me provide the background first.
Windows Vista has a new feature called Timeout Detection and Recovery (TDR).
TDR attempts to detect problematic situations and recover to a functional
desktop dynamically. In prior operating systems, these situations would have
resulted in a system freeze and forced customers to reboot their PC. More
information about this Vista feature can be
found here http://www.microsoft.com/whdc/device/display/wddm_timeout.mspx.
Therefore, when you see the TDR error message "Display driver stopped
responding and has recovered," you know that the display driver is being
reset without requiring a reboot.
Now to answer
your question: TDRs can occur for a variety of reasons, many of which
are unrelated to the graphics card or graphics driver. Since Vista launched, NVIDIA has been working hard to address
TDR issues that are specific to the graphics driver. Last year, we set
up the NVIDIA Vista Quality Assurance Site to record and address
user issues: http://www.nvidia.com/object/vistaqualityassurance.html.
Since that time, we have resolved a majority of the issues specific to the
graphics driver and have also worked directly with Microsoft to
release five Vista hotfixes, all of which
are now part of Vista Service Pack 1 (SP1). If customers are still
experiencing issues, we recommend that they continue to file issues at the
quality assurance site. We also recommend that customers look
carefully into the wide range of other causes of TDR issues such as overclocked
components, incorrect component compatibility and settings (especially
memory configuration and timings), defective parts (memory modules,
motherboards, etc.), insufficient system cooling, and insufficient system
power. Throughout the forums, there are numerous reports of hardware solutions
to TDRs. Here is one example: "Just short of a year ago I built
a system that will crash after about an hour of intensive gaming on only two
games... those were the only two games I played that were intensive enough to
cause it. The reason was that my power supply was inadequate. Upgrading the
power supply eliminated all the problems." http://forums.anandtech.com/messageview.aspx?catid=31&threadid=2116981
6. Xpert: Do you plan on integrating SLI technology onto
motherboards w/ Intel chipsets with a brand new spankin license? What does
NVIDIA plan on doing with their SLI support in the future since they are having
problems with Intel and renewing their license with them?
Andrew Fear: We offer SLI support for Intel CPUs on our nForce chipsets and
will continue to offer this on future nForce chipsets. Our commitment is to provide consumers with platforms that offer the best
experience. We develop our products with careful consideration to price,
performance, power, and customer requirements. Our SLI certification program
helps to ensure consumers can build or buy PCs designed for the best visual
experience. We allocate our certification resources on our own NVIDIA nForce
MCPs, because they integrate special technology that helps to improve and
accelerate SLI graphics performance, something that Intel chipsets cannot
offer.
7. Xpert: Why
must we tinker with modifying .inf's to get the GeForce drivers to recognize so
many laptop or older GPU's? It's understandably too much to test every old card
and manufacturer's implementation of your chips, but a simple warning would do
during driver installation! Or at very least you could have a driver release
with unsupported backward compatibility to be used at one's own peril.
Chris Daniel: Unfortunately,
NVIDIA does not have approval from most notebook manufacturers to release
updated drivers since notebooks use custom power management and other features.
Therefore, NVIDIA provides drivers to notebook manufacturers and it is up
to those manufacturers to release driver updates to their customers. That said,
NVIDIA has started a new program that will release notebook
drivers on NVIDIA.com for manufacturers that choose to participate in
the program and approve the driver release. The first of these drivers was
released in January and can be found here: http://www.nvidia.com/object/winxp_notebook_167.51.html.
Legacy GPUs
are supported with the latest known stable WHQL driver.
The person who
asked this question proposed: “at very least you could have a driver release
with unsupported backward compatibility to be used at one's own peril.”
We do not
believe it’s a good idea to point customers to drivers that haven't been tested
and that could cause major system instability. That’s not an experience that
NVIDIA chooses to provide.
8. Xpert: How in the world does Nvidia
get so much more performance out of less than half the number of rendering
pipes ATI uses?
Ujesh Desai: Ancient Chinese secret! ;-)
9. Xpert: When do you plan on allowing multiple monitors
on an SLI set up? It seems rather silly having four monitor ports but only
being able to have one monitor.
Andrew
Fear: NVIDIA is committed
to adding SLI multi-monitor support. We understand it’s an important feature
for enthusiast gamers and will have it available as soon as we can. Please
keep checking http://blogs.nvidia.com/sli
for information on this feature and when it will be available.
10. Xpert: If the GeForce 9800 GX2 is shown as a single
card will there be driver and hardware support to utilize 3-way SLI
on 780i+ mobos?
Andrew
Fear: No. Our plans are to just support Quad
SLI on this product using two GeForce 9800 GX2 graphics cards.
11. Xpert: Are there any plans to create a product
designed for use with VM technology? More specifically, some type of
paravirtualization for real 3d acceleration inside of Virtual Machines. Sure
there are software hacks now that emulates 3d acceleration, but should we
expect nVIDIA to embrace this market soon? If so, would the technology be built
into the Quadro line, Geforce series, or both? Lastly, is this capability
already built into the G92 series GPU?
Ujesh Desai: Company policy prevents us from
commenting on unannounced products or technology currently in development.
12-13. Xpert: How
long do you think it will be before you reach the bandwidth cap of the pcie x1
and x16 slots? Any chance of going to cooler running cores than what we're
currently offered?
Xpert: How
much of a standard PCIe x16 lane is saturated by nVidia's cards, and is there
any way of calculating/finding this out?
Since Dual-Core GPUs have double the amount of
information they can load and process, does it saturate the bandwidth
anymore then Single-Core GPUs, and it actually fully saturate (At full load) a
PCI-Express 16x (1.1v) or even a PCI-Express 8x (1.0v)? And if so, does getting
PCI-Express 2.0 give you better performance or will a PCI-Express 16x (1.1v) work?
Ujesh Desai: Today’s applications
don’t saturate PCIe x16 bandwidth, and it’s likely that the bandwidth offered
by x16 PCIe will still deliver plenty of headroom for future applications.
We are always looking at ways to improve
our perf-per-watt and improve our green story. That’s why we are also
developing technology like Hybrid Power.
14. Xpert: With
all pc components power requirements going up putting a bigger dent in
customer’s electric bills, do you think that we will see pc components like
video cards become more energy efficient? Or will we be seeing 1,000+ watt
power supplies as the norm of the future?
With the increased pressure to turn out
"green" components does NVIDIA have any stances concerning the
development of future items to support their customers being more
"green" (i.e. using less energy tying to power the system /
cool the environment, etc.)?
Ujesh Desai: NVIDIA graphics cards
in general are extremely efficient in terms of perf/watt. To enable even
greater efficiency, we created a new technology called Hybrid SLI. For
mainstream users, this technology will provide better performance without
requiring large power supplies. For high-end users, HybridPower will enable you
to completely shut down the graphics cards when you don’t need them (e.g.: checking
e-mail or instant messaging), thereby saving energy and money.
15. Xpert: With Microsoft throwing out
DirectX 10.1 and Direct X 10.0 graphic cards reportedly are not compatible with
the newer version(s), is NVIDIA aware of any manufacturers offering trade out
programs?
Ujesh Desai: I’ll assume from the context of the question that “throwing out”
means “releasing” or “introducing.” Microsoft and NVIDIA have repeatedly stated
that DX 10 graphics cards are 100 percent compatible with DX 10.1 games. In
addition, most of the PC game developers that NVIDIA works with say they have
no plans to add DX 10.1 features in the games that they are working on. So you
don’t need to “trade out” your DX 10 hardware.
16. Xpert: When can we expect to see 65,
45, or smaller GPUs with respective reduction in power demand?
Ujesh Desai: Company policy
prevents us from commenting on unannounced products or technology currently in
development.
17. Xpert: How much
longer will we have to wait before Ntune fully integrates with the i680 chipset?
Tom Petersen, director of platform technologies: Right now I am happy to say we have full support for nForce 680i,
780i, and 790i motherboards in all our software tools. The link to
our most recent package is here:
http://www.nvidia.com/object/nforce_system_tools_6.01.html
It is hard to tell from your question,
but I am betting you have tried it or an earlier release and you are finding
some features are not present or do not work. This can be caused by BIOSs
that are not designed to work with our tools. Some of our customers
boards don’t have full support for our system tools. In those cases they
typically make available similar tools which they develop internally and
provide directly.
18. Xpert: Why have the video drivers for the 8800GTX
taken a back seat, since the release of the 8800GT512 (G92) and 9600. We were seeing
regular driver updates. And that has since stopped.
Chris Daniel: GeForce
8800 GTX drivers definitely haven't stopped. A couple of weeks ago we released
v174.74 beta drivers supporting GeForce products and in another couple of weeks
we will release the WHQL version. NVIDIA has always launched new products with
new release branches, such as v174 in this case. This enables new products to
get in the hand of customers faster and does not slow down the driver release
schedule for the rest of the product family.
19. Xpert: Is it true
that nVidia cards select Frames to skip in games to enhance performance/speed?
If it is true, why? Some people love flowing beauty and have the eyes to pick
up even a single frame change.
Chris Daniel: No,
this is not true. Let me clear up the confusion. SLI technology uses a
rendering method for most games called alternate frame rendering (AFR). This
rendering method allows customers to get up to 2X the performance in games by
splitting the workload between two graphics card. The first card renders Frame
A and the second card renders Frame B and so on. Therefore the first card
renders every second frame and the second card renders the middle frames. So you’ll
see every frame without any skipping.
20. Xpert: Is something going to be done
about the latest nForce technology and how it currently designed? The consensus
from enthusiasts is the latest generation of the nForce chipset is
disappointing being a slightly built up version 600 series (I wouldn't try to
deny this either). Is nVIDIA going to put in the necessary RND (Research and
Development) work in to the nForce chipset to make it more competitive with
Intel or will Intel "eat nVIDIA's lunch" and take more of nVIDIA
market share?
Tom Petersen: NVIDIA continues to invest in nForce
technology. One of the most recent examples is the introduction of the nForce
790i SLI Ultra platform. That platform is the result of intense engineering
efforts all focused around one objective…being the best platform for Intel CPUs
in the world. The 790i platform uses all new silicon for the SPP and supports
all Intel’s latest CPUs and all next generation IO technologies. In short, we invest in technologies where we
can be world class. Our strategy is to
innovate to provide the best platform –SLI, Hybrid SLI, ESA, Teaming and
FirstPacket are all examples of our continued commitment to the platform.
21. Xpert: What
will you guys do after the 9000 series cards technology is finished? My thought
is that I seriously doubt that they will continue into 10000, but rather do
something like what Alienware did which was go from their 9700 laptop to their
new M15LX (or whatever its called). Also I remember recently that you guys made
a video card that had 1gb memory on it, which is outstanding! What is next when
it comes to memory?
Ujesh Desai: We’re open to suggestions! How about the GeForce Gigadeth Annihilator
series? ;-)
Other
1. Xpert: I would like to hear them
discuss plans on how we the consumer could have more choices / the ability to
customize a video card or mother board so that they match a common theme or
style. For example, I would like to be able to choose all black circuit
boards with copper heat pipe cooling features that match. We all get
tired of seeing the thing we want in the color we don't want or the video card
we want not match the theme we are trying to go with in our cases. There
should be a way for "US" to select the color and cooling
solution, not the maker. Give us a way to further customize our rig without all
this aftermarket stuff.
Ujesh Desai: Cool idea. Logistically this may be tough to manage but it’s something
we’ll definitely pass along to our board partners.
3. Xpert: Not that i am complaining, but speaking as one
who spent a king's ransom to purchase an 8 series SLI rig for my gamer, how is
it that the introductory prices for the 9 series is so cheap in comparison, and
why could not have the 8 series been done the same way. Or is it because of
R&D costs, which were paid by 8 series sales, and carried over to the 9
series?
Ujesh Desai: There are a number
of factors, R&D, process technology and maturity, as well as overall
improvements in the architectural efficiency that has enabled us to deliver the
performance and features on the GeForce 9 Series products. Our goal is to
constantly innovate and improve the price/performance we deliver to our
customers.
4. Xpert: Would
Nvidia please make an option to permit removal of (Active Armor Firewall) if it
is not wanted?
Tom Petersen: The Active
Armor Firewall has been discontinued and is no longer distributed as part of
our driver packages. To clean remnants of the software from an older system we
recommend that you download the latest driver package for your platform from
nvidia.com, then uninstall all of your old nforce drivers. Reboot and
reinstall the new package.
5. Xpert: Do you
feel like you've won the war in the Video card market?
Ujesh Desai: We strive hard to make excellent products that matter to people. We’re
proud that we’re the leader in visual computing and that our discrete graphics
products have the leading market share. People like to portray companies in
conflict with one another, but the thing that motivates our architects,
designers, and software teams is how we can make a difference in the world with
better technology. How can we show games in real time that look like animated feature
films? How can we speed up visually rich applications and make them easier to
use with GPUs? How can we speed up scientific research to help cure disease?
How can we help companies find energy reserves miles beneath the earth? How can
we help the medical community improve their diagnostic tools? There are an
awful lot of problems and challenges that remain in technology, and parallel
processing technology is going to be the key to solving them. That’s our focus,
and that’s what motivates us.
Special
thanks to NVIDIA & the guys who contributed to the interview:
Andrew
Fear, SLI product manager
Ujesh
Desai, general manager of desktop GeForce products
Patrick
Beaulieu, video technologies product manager
Chris
Daniel, software product manager
Tom
Petersen, director of platform technologies