Andrzej's monthly column #2

Column

Column: nVidia unveils their first CPU at CeBIT?

OK. I may mean CeBIT 2008, but whether it is real or a bluff to help leverage a deal with a major player, there seems little doubt that competition for the BOM is going to get fierce and anyone without a CPU will not be able to play the ‘platform programme' game. The BOM is the Bill Of Materials that makes up a PC. From a peak of around €2,500 in the nineties, the UK average is now probably closer to €900. Just like a water hole drying up in Africa, less resource means that the ‘hunters' need to get closer and closer together in order to take a drink. PLUS, just for fun, in a little while there will not be enough water for everyone - so someone has to die.

Since their famous line of Socket 7 chipsets, owning as much of the BOM as possible has been an important part of Intel's strategy. Multiple components within a system gives you a platform - and ‘Platform Power' came into its own with Centrino. Launched in March 2003, with promises of mobility for all, over a third of notebooks shipping by Christmas 2003 had wireless capability. Experts at the time predicted that there would be strong competition for this market. However, they did not allow for the Intel marketing muscle that would follow. As long as your advertising followed the ‘Butterfly effect', there were plenty of Intel marketing dollars available. Instead of having to find the dollars from a single component - Intel owned the CPU, chipset, wireless etc - and had plenty more marketing money to go round.

The effect of Centrino was stunning - even to people like me who have been in the business for more than 20 years. Centrino platform marketing rocked the world and you still hear people ask for a ‘Centrino processor' in their laptop, because the message/branding was so powerful. Even when they were enjoying huge growth in revenue, neither ATI nor nVidia were able to come close to the margins available to the CPU companies. While GPUs can struggle to get past 20%, the CPU giants can regularly command margins over 50%.

So how can the smaller GPU manufacturers compete for the BOM? Rumour has it that nVidia have stuffed over $500M into a warchest to create a CPU business - at a time when demand for their traditional products is at an all-time peak. Licensing issues with x86 aside, some senior people within nVidia must feel that it is crucial to their survival for them to have a CPU offer. No CPU means no platform. No platform means no ‘Centrino'. And that is not a happy place to be with FUSION etc coming.

FUSION is AMD and ATI's first step toward unification - the fruits of which will have ramped up to full volume by the time their heavily subsidised New York plant opens in 2009. When Koen first saw the internal diagrams for X1600 and X1800, he must have thought ‘R520 looks a bit like four X1600s stuck together' and he would have been (partially) right. From graphics capability on the same substrate - to ‘CrossFire within the core' is a small and inevitable step. The GPU represents a big chunk of the BOM and the CPU guys want a piece of it.

Which brings us to Intel and GPUs. People always seem surprised to learn that Intel is the world's largest supplier of graphics parts - with their integrated solutions for desktop and mobile. This is a market which they have dominated for a long time - starting with things like the i740 products and leading up to today's GMA parts. Now, it seems, they are getting ready to take their first few steps toward discrete graphic cards - as you can see from this link.

Creating ‘world class silicon' won't be Intel's biggest challenge - only part of the battle is won in the hardware. Far more difficult will be drivers and developer relations. Drivers for graphic cards are like drivers for Formula 1 teams. With identical hardware/capabilities, they are the difference between winning and losing. ATI and nVidia have massive teams of the best driver writers in the world - dedicated to creating optimizations for every benchmark/game on the market. Does it really make that much of a difference? 100% YES!

A guy called ‘Andy' in ATI's driver team came up with the idea to replace a ‘table look up' with a ‘pure maths calculation' for Quake 4. After 3 solid years of losing to nVidia in OpenGL applications - this single ‘shader replacement' allowed ATI's R520 series to ‘win' at Quake 4. One driver update, one change - and you win. Driver writers are a different kind of animal, spending all day, every day, trying to save nanoseconds in game code execution. All of the best ones already work for ATI and nVidia.

ATI's strategy for years has been to try to make great hardware, provide Huddy & Co for DevRel (Developer Relations) and hope that the developers will ‘Get In The Game' by using your ‘feature' (e.g. HDR with AA). nVidia have always shown themselves ready to buy their way into every game in town - and work from within to convince the developers that their way is the only way a game is meant to be played. Intel don't have an army of developer relations people (yet) - but they do have the deepest pockets.

Can they buy their way into AAA-titles like Crysis? Definitely.
Can they make decent silicon? Sure.
Can they create a high-end discrete solution that will beat nVidia and ATI? I have to say ‘unlikely'.
Will that make a difference ? Probably not.

Most graphic cards sold are priced at less than €150. That is a market where absolute performance is not as important as ‘feature set tick boxes', ‘famous brands' and ‘price'. Intel have the biggest brand, great channels to market, all of the features and - more importantly - no AIBs. An AIB is an Add-In-Board partner who takes the ATI/nVidia GPU and makes it into a ‘graphic card'. No AIB means that you have a 5-15% price advantage... and that could be pretty serious. Above all, having the graphic solution as part of your ‘platform programme' means - you guessed it - more marketing dollars available !

Overall, I think nVidia have a better chance of making cheap CPUs than Intel do of making powerful graphics accelerators - and ATI/AMD already have both. One thing is for sure, it will be ‘fun' for us spectators watching the big beasts tearing each other apart at the waterhole !

In stark contrast to AMD's rather nervous approach to telling us about their future plans, Intel have been very gung-ho over the past 12 months - especially around the Conroe pre-launch activity. My money is on Intel being very bullish about graphics at IDF next time around. They may not talk exact specs, but they could lay the ground to discuss how they will ‘own' the DX11 market in a couple of years time. Right now, they might be content with what they have - being number 1 overall - helped by a swollen integrated share. But, somewhere around 2008/9, I expect they will want to ‘puff their chests out' and take serious market share in the discrete arena. In terms of competition, a lot will depend on where competing products from ATI and nVidia fall. Assuming we mean (and what uber-geeks might call) G100 and R800, Intel is certainly setting their aim far higher than anyone could have imagined.

Andrzej Bania is Marketing Director for OMEGA SEKTOR which is a Cyber Theme Park with almost 600 PCs and Consoles. Previously, he ran PR & Marketing for ATI in Northern Europe and, before that, helped run the UK's number 1 award winning PC manufacturer.