I thought I’d start changing things up a bit this week by writing more editorials than ever before. Topic number one - the coming graphics battles.
In case you missed it last week, Jen-Hsun Huang, CEO of NVIDIA, did a little rant (or shall we say whining?) about Intel Corporation. To no one’s real surprise, Mr. Huang was a bit defensive about the state of the graphics industry and was pretty peeved about how an Intel graphics and gaming "technologist" was calling discrete graphics cards "unnecessary." Jen-Hsun used NVIDIA’s financial analyst day last Thursday as a way to dish it out against Intel and their marketing machine. Here’s a quote from News.com cited:
"We don’t typically like to do this. It’s just that we’ve been taking it and taking it and taking it. Every single frickin’ day. Are you allowed to say that word? Every day all over the world. Enough is enough."
So NVIDIA isn’t planning to take it anymore. OK.. but is there something more to this perhaps? Why take time out from a financial analyst meeting to go complain about your partner/enemy in Intel? I mean, come on now, we’re talking about Intel’s integrated graphics solutions. Does NVIDIA really have something to worry about? Well.. in a word, yes. Why you might ask? It’s not your younger self’s graphics world anymore. It’s certainly not a world where it’s NVIDIA against ATI or any other smaller vendor. NVIDIA finds itself suddenly very much alone in the graphics market with ATI now belonging to AMD. It’s now NVIDIA vs. Intel/AMD ironically enough. Yet, NVIDIA has chosen to set it’s sights on Intel. Why? because of the path that Intel is working towards.
First a little background. I was part of the original Intel 740 team way back when Intel was still trying to figure out how to compete against NVIDIA, ATI, and 3dfx (remember them?). At that time (around ‘97/98), Intel had teamed up with Real3D and Chips and Technologies to design and create the Intel 740. The 740 was actually Intel’s third attempt at getting into the graphics market and there was much hope in the company that it would indeed succeed.
After a few delays, the Intel 740 came out to less than stellar reviews yet it did do one important thing to the graphics market - it paved the way towards cheap graphics. With a large inventory of 740 processors, Intel realized the only way to clear things out would be to sell them for low prices to third party board manufacturers which of course led to super cheap 3D graphics boards. The other effect the 740 and the follow-on projects created was the realization by Intel that it needed to bring graphics to the motherboard chipset - paving the way for Intel’s integrated graphics efforts that would eventually make them the largest supplier of graphics in the world.
Why take the integrated graphics route? It boiled down to economics really.
1. Intel makes a ton of money selling chipsets and I do mean a ton. Most of you think of Intel as a general processor company but consider the following - for every Intel processor sold, it’s very likely you’ll also find an Intel motherboard chipset inside. Selling graphics is no sure thing. Selling motherboard chipsets is a far easier task for Intel to accomplish so why not fold graphics into the chipset.
2. Intel discovered long ago that it could not compete from a design, development and manufacturing standpoint with the likes of NVIDIA and ATI. The smaller companies are more nimble and can execute on newer designs much more quickly. Yet being fast and nimble also means you might not have the best optimized processes in place. Intel executes well on established processes which ultimately results in higher profit margins. If a product doesn’t deliver roughly 40-50 percent profit margins, you likely won’t find Intel in it.
3. Finally, Intel realized that the CPU had much more room to spare in terms of processing capabilities. Let’s face it - today’s (or yesterday’s) applications really don’t stress the CPU. (That’s why Microsoft has Vista - an OS that can only run reasonably well on a modern processor!) It’s a conspiracy of course but most of you know that already. Intel, AMD, and Microsoft are constantly looking for ways to get consumers and businesses to upgrade their hardware and software. How do you convince consumers of this? Push the existing computing platforms to the limit which results in consumers saying "Gee.. I need a new computer." So what can Intel do to increase the load on a CPU? Make the processor do more graphics chores by integrating graphics to the chipset. As Intel adds more and more cores to their CPU, there becomes less of a need for a secondary graphics processing unit (GPU).
OK.. so this brings us back to the original rant from NVIDIA’s CEO. Was it justified? Perhaps.. perhaps not. Yeah he’s sick of Intel’s claims and marketing hype. We all know integrated graphics can’t keep up with discrete graphics right now. Yet people (a.k.a. regular consumers) aren’t aware of this so they’ll stick with whatever comes with their systems. So what’s a CEO to do? I suppose in NVIDIA’s case, it’s to whine but NVIDIA needs to continue pushing the envelope and in many ways, play the same games that Intel and AMD plays.
NVIDIA needs to continue executing at a rapid clip. It needs to stay ahead of Intel while at the same time convincing consumers to add in NVIDIA graphics to their systems. This translates to a much larger marketing presence for NVIDIA. It means working with many more graphics software vendors and constantly staying two to three steps ahead of Intel.
It’s going to be hard of course because as Loyd Case points out in this article, software based shaders will only get faster as CPUs get faster. Someday, a general purpose processor with multiple cores and integrated graphics will be all a consumer needs unless important graphics applications and games have a specific need for NVIDIA capabilities (hence why NVIDIA hopes people will program to their GPUs).
So NVIDIA - stop complaining and execute in a timely manner. Things will certainly get harder for NVIDIA but the company needs to stay ahead by innovating and creating useful products.
Technorati Tags: Intel, NVIDIA
The other major AMD related news release today was the introduction of the new ATI Radeon HD 4870 X2 graphics card which AMD/ATI calls the “world’s fastest graphics card.” Unlike NVIDIA which opted to create larger and faster GPUs, AMD opted to go the multi-processor route by building smaller GPUs and then combining them for graphics processing performance. According to AMD, the new Radeon HD 4870 X2 delivers 2.4 teraFLOPS of processing power on a single card thanks to 2GB of GDDR5 memory and two RV770 GPUs on the card clocked in at 750 MHz for a total of 1600 stream processors. 
