|
|
摘要：
Jen-Hsun Huang
Today we reported revenue of $1.15 billion for our first quarter. Compared to last year, revenue grew 37% and GAAP net income increased 34%.
Year over year, our GPU revenue grew 55%, MCP revenue grew 31%, PSB revenue grew 44%.
Overall our graphics market share grew from 30% to 33% year over year. However, if we accounted for double attach, whiappens when a PC has both integrated graphics and GPU, our market share is approximately 42%.
Marvin D. Burkett
GPU：
which includes desktop, notebook, and memory, grew 45% year to year but declined by 8% from the fourth quarter. The components of GPU had significant growth year to year.
Desktop GPUs： grew 44% year to year.
Notebook grew 99% and memory declined 36%.
MCP：
grew 31% year to year and declined 4% quarter to quarter.
The star for the first quarter was the professional solutions business, which includes workstation products. It grew 44% year to year and 21% quarter to quarter.
The consumer business declined by 24% quarter to quarter and 37% year to year.
毛利方面：
Gross margin was disappointing. We didn’t manage our product transitions as well as we would have liked and this resulted in a decline in gross margin to 44.6%.
原因：The issues in gross margin were focused primarily on product transition issues in the GPU business. Therefore, margin for the GPU business segment declined by almost two full percentage points quarter to quarter. Again, it was not an ASP issues, as ASPs held relatively flat. We had hoped and expected to make progress in improving the yields for the new products. This did not happen.
专业解决方案方面：Margin in the professional solutions business remained very good.
MCP方面：Margin in MCP were down because of the movement to the integrated chipsets, but that was expected.
汇率造成的影响：Other income dropped by $7 million quarter to quarter. This was a combination of several factors. The impact of the continuing weakness of the dollar on our foreign currency denominated liabilities resulted in foreign exchange losses of approximately $4 million in the quarter. Also, the lower cash balance and the lower interest rates drove down the interest income.
Jen-Hsun Huang对G80 -> G92迁移以及未来55nm的评论：
The transition from G80 to G92 was challenging and dragged our margins. Every few generations, because of how our new products are aligned with process nodes, a new GPU is both faster and cheaper. As a result, it is difficult to move the stack down to layer the new GPU on top of the stack. This was the case in the G80 to G92 transition. G92 was both lower cost and higher performance than G80, so therefore repositioning G80 while ramping G92 aggressively forced us to sell out G80s at lower margins.
Going forward, although we still have some G80 inventory to work through, we believe we are in an excellent competitive position. We are also launching new exciting products in Q2 and the move across the board to 55-nanometer is expected to help margins
CUDA方面的进展：
In additional to visual computing driving our GPU growth, we have a new growth driver with the invention of CUDA. It is important to understand that CUDA is not in just any GPU. CUDA is a C programmable general purpose computing mode available only in our GPUs, from GeForce 8 and beyond.
In the CUDA mode, our GPUs transform and reconfigure its shader processors into a many core parallel computing processor. For example the GeForce 9800GT has 128 processor cores. CUDA enables a new parallel computing model that the industry calls heterogeneous multi-core computing, where the vastly different architectures of the CPU and GPU are used in a collaborative way to achieve huge speed-ups in computing. CUDA has raised the attention of programmers and researchers all over the world. Since its launch, over 60,000 downloads of the compiler, over 10,000 per month, have been downloaded by software developers, scientists, students, game developers and researchers around the world that have started programming with CUDA.
We recently released a beta version of CUDA 2.0 and for the Mac. We are constantly seeing reports of astounding application speed-ups. We are routinely seeing 20 to 200 times speed-up over multi-core CPU alone. Chevron has shown excellent performance using GPUs for computing in oil and gas exploration. Jack Collins, from the National Cancer Instituted, recently reported excellent results on AutoDock, a key application in cancer drug discovery.
[Gouda] and [Navcentric] announced EDA tools using GPU for computing optical proximity correction, or OPC, and [inaudible] for design.
CEA, the French Atomic Energy Authority, and Bull released news of a design win in super computing using NVIDIA GPUs.
Last week, we announced our sponsorship of Stanford University's new Pervasive Parallelism Lab, PPL. The PPL will develop new techniques, tools, and training materials to allow software engineers to harness the parallelism of the heterogeneous multi-core computing. Stanford now joins the growing list of prestigious research programs, including the universal parallel computing research center, UPCRC, both at Berkeley and the University of Illinois Urbana-Champaign, each designed to accelerate developments in mainstream parallel computing.
We see computer scientists all over the world jumping into heterogeneous multi-core computing research.
As the world’s leading supplier of GPUs, we are in truly exciting times. The new sensibility of OEMs optimizing PCs for visually rich applications is driving GPU adoption and spend. And with CUDA, we have transformed our GPU into a general purpose parallel processor at precisely the time when the CPU performance improvement is slowing and the industry is moving full-force to parallel computing.
We believe the GPU will be the most important process of this new era and will be a major driving force of visual computing and parallel computing revolution.
问答环节：
关于Telsa/应用处理器产品线何时能带来显著的收入影响：
Arnab Chanda - Deutsche Bank Securities：Thank you. A question for either Jen-Hsun or Marv; when do you expect significant revenues from your TESLA product and the application processor product?
Jen-Hsun Huang：I would say significant I guess later this year but those design wins tend to be a lot longer. TESLA because they go into servers and super-computers and scientific computing applications, or [design] workstations. So they tend to have a longer design cycle. We are seeing a lot of enthusiasm out there but it just takes a little longer.
And as you know, application processors definitely take longer. Their design cycles are consistent with the design of the device itself and so we are really enthusiastic about both of them and they’ve been wonderful investments for us, and I think they are going to pay off wonderfully. So we just have to wait until the end of the year.
关于heterogeneous（异质架构）方面的看法：
Arnab Chanda - Deutsche Bank Securities：Thanks, Jen-Hsun. One other question about the CUDA platform -- could you help us understand what portion of the computing market, if you include both PC as well as high-performance computing, is open to the heterogeneous computing model that you are recommending and what -- maybe it’s difficult to give an exact number but how should we think about how big that market can be?
Jen-Hsun Huang：I think that’s a good question. I happen to think that within the next five to 10 years, no longer than 10 years and probably around five years, the heterogeneous multi-core computing will be the way of computing for all computers, whether it is video processing or image processing or whether it’s in the area of games obviously for graphics, but more importantly in the future for simulation, physics simulation, artificial intelligence simulation. You’ve been hearing about how the expense of creating content continues to go up and the reason for that is because these worlds are more and more expansive and it’s hard for artists to create the entire world. And so just like movies, they use -- movies uses a simulator for cloth simulation, otherwise they would have to animate all of the cloth, all of the clothes that the actors wear, or hair simulation or in large crowds, they use crowd simulators, which is basically a form of artificial intelligence.
And so we are going to see more and more simulations as a way of creating content in online games, in virtual worlds, in videogames, and of course in the area of mainstream users -- you know, it’s -- you know that videogames is really a mainstream application anymore but in the area of photos and videos, you are going to see a lot of computational techniques starting later this year from some very important applications. And so I ought not pre-announce some of their exciting products but consumer applications, lifestyle applications are going to start showing, using GPUs in some really, really exciting ways.
So I think heterogeneous computing is the future. I mean, it is as obvious as using the right tool to do the right job and there is no reason why heterogeneous computing can’t be available to everybody. You know, we’ve shipped 60 million CUDA GPUs already and we are on track to ship 100 million, 150 million GPUs a year and in a couple, two, three years, that’s several hundred million PCs will have heterogeneous computing in it. So I think it’s a foregone conclusion that heterogeneous computing is here. The question now is what are the exciting applications that are going to come.
And just one more comment -- in the last 20 years that I’ve been in the computer industry, the CPU or the computer has improved in performance a staggering 1,000 times. It’s a staggering 1,000 times. And that’s clearly one of the most aggressive improvements in performance of any technology or anything that humanity has ever seen.
And yet CUDA, in the hands of all these 60,000 software engineers around the world, and it’s easy for you to find applications and reports of CUDA applications written, people are seeing 20, 50, 100, 200, 300 times speed-up. In just one year, we’ve helped the computer improve its performance by 100 times. That kind of a step-up and that kind of a discontinuity in computing performance has just never happened. And so that’s the reason why Stanford and Berkeley and UIUC and just about every supercomputing center on the planet and scientists all over the world and software engineers in all the major software companies that you and I both know are all looking into heterogeneous multi-core computing.
I think this is clearly the future.
关于库存问题：
Gary Mobley - Piper Jaffray ：I have a couple of questions relating to balance sheet items. Are there any troubles in production lead times still out there? And your inventory turns are still above the historical norm of five to six and I’m just curious whether we might see a continued rise in the NVIDIA held inventory.
Marvin D. Burkett：Our day sales in inventory are 60 and that’s where I’d hope to get it because then it’s down at 50 or below, that’s too little inventory. Once you have it at 60, the real issue is what is in the inventory? Is it old stuff that is moving slowly or is it new stuff where you don’t have enough of it? And I would say at this stage that I am very comfortable with the inventory, both the profile and the amount of it. Are we going to build inventory? Yeah, I think we’ll build inventory as we go into Q3 and Q4 and with the expectation that revenue usually increases significantly in Q3 and Q4, so I wouldn’t have any trouble at all with building inventory in Q2, providing it’s the right inventory.
关于Q2 GPU价格走向的看法：
James Schneider - Goldman Sachs：Good afternoon. Could you maybe within the desktop GPU segment comment on your expectations for Q2 in terms of pricing, and whether or not you’ve seen any incremental pricing actions by some of your competitors there?
Jen-Hsun Huang：It’s hard to exactly predict what happens in the short-term, but I know what the general trends are. The general trends are the GPU is benefiting from an industry-wide recognition that visual computing is important. The GPU is doing more and more things because it is more and more programmable, and with CUDA, we’ve made the GPU a completely general purpose parallel processor. And so these things will cause it to be adopted more and therefore that will increase units. It will increase its TAM, you know, places that you would have never thought a GPU would go would include a GPU. And it would also increase its ASP because it is more and more useful.
And so we are seeing all of those mega-trends, and so I think that with respect to Q2, we know we are announcing new products and so that’s always helpful but I can’t control what the competition does.
9000系列是否会让NVIDIA在back-to-school的小旺季里增加份额：
Tayyib Shah - Longbow Research：Jen-Hsun, you are guiding to revenues in line with historical seasonality for the PC space. With the 9000 series ramping, it puts you in a strong competitive position in the desktop space. Do you expect to gain some share in the back-to-school season and if so, shouldn’t you be guiding above historical seasonality?
Jen-Hsun Huang：Did I mention the -- we’ve got to treat Q2 with respect. I think you are right that we have a very strong, a very good competitive position. We have fabulous products top to bottom and we are also announcing several new initiatives that you guys heard about at the analyst day.
One of the things is the modern benchmark is out. We’ve been using -- the benchmark that the industry has been using has been around since 2006 and in graphics years, that’s a long time, and so the new benchmark is out. It’s called 3D MarkVantage and it’s a modern benchmark and because its -- our GPUs, our modern GPUs are designed for modern applications. You would think that our position on 3D MarkVantage should be extraordinary and it is. So that’s one thing and that’s in the process of rolling out. The market doesn’t really know it yet and I think you could see some early websites, go see some -- and you might just be able to Google it and find some reviews and you’ll see that the reviews are really fabulous with GeForce 8 and GeForce 9, so that’s one.
The second thing is we are going to announce this quarter probably the highest volume parallel computing application or software in the world and that’s physics processing. It’s used in almost every modern game and it is one of the most computational intensive things that you can do in a game and it now runs on CUDA on GeForce GPUs.
And then we are going to add on top of that -- hopefully we’ll go into production later on in the quarter -- transcoding. It is probably the killer app today with the number of iPods that are being sold and iPhones that are being sold. People are downloading a lot of videos onto their portable media players and the transcoding of videos take forever. You have to start it up at night and give up your PC and go to bed and hope that it’s done before you wake up.
And so we are going to accelerate the transcoding of movies to what is called super real-time -- it’s much, much faster than it takes for you to watch is, and so I think transcoding is just going to -- people are going to be nuts over the work that we are doing in this area.
So not only are we -- do we have a fabulous position with our GPU lineup but we are adding three major things to it this quarter and I think that can’t but help. But having said all that, I think we’ve got to show Q2 its due respect. |
|
IP卡
狗仔卡
发表于 2008-5-9 16:25
QQ好友和群
收藏
分享
好贴
烂贴
提升卡
置顶卡
沉默卡
喧嚣卡
变色卡
显身卡
楼主