IT WAS WORTH listening to the Nvidia conference call yesterday, as the company gave some facts and figures on the market and on its relationships with its partners, including Microsoft (Xbox) and TSMC, which will fab (make) its up-and-coming NV30.
Yesterday the firm reported a net loss of $48.6 million on its Q3, but took a charge. The firm said in the conference call that its biggest cost in Q3 was the legal costs it incurred.
But peace has broken out between Nvidia and TSMC, even if its relationship with Microsoft is still a tad rocky. Jen-Hsun Huang, the CEO and president of the firm said: "IBM has a worldclass semiconductor process but we've always known that. Our focus is very much TSMC and we're delighted with their performance on .13µ technology. The copper technology is very stable. That will give a dramatically increased performance for NV30".
TSMC had ramped .13µ since the beginning of the year. Earlier in the year it had "struggled some" but Nvidia is now seeing very good yields across the board. The combination of copper technology and .13µ technology was compelling. There were already wafers using this technology.
It had started testing TSMC .13µ in February and March. The challenges other people saw made Nvidia test it carefully, and now it's being mapped into the next family of GPUs, and there's a lot of risk been taken out of that by its testing.
The next generation would be a bigger challenge, particularly for cinematic rendering, and that day wasn't here yet, Nvidia said.
He also said that the total engineering costs for NV30, which will be launched at Comdex in Vegas later this month, will amount to a staggering $400 million. That kind of cost was an obstacle to any new start ups entering the market, he said.
He wouldn't be drawn on wafer pricing discounts Nvidia might get from TSMC, but said: "We're getting a fair price from TSMC and if anyone from TSMC is listening, I'm sure they're giving me a very competitive price, one that they'd be proud of and I'd be proud of". So Nvidia is getting a good deal from TSMC.
TSMC is the best foundry in the world, he said.
Despite its dispute with Microsoft over the Xbox, he said that "the relationship with Microsoft is terrific, we just have a dispute with them over money."
The GeForce 4 accounts for under 50% of Nvidia's business in units, and while units for desktop graphics processors (GPUs) were up, ASPs are down and the two combined created a flattish desktop graphics business, for its financial Q3.
The overall PC market was slack or a little bit up sequentially in Q3 2003, but Nvidia's unit sales were up 19% while Intel's integrated graphics chipsets were up sequentially.
Microsoft sales on the Xbox were just slightly less than 20% in the quarter, down from 29% in Q2.
Twenty seven per cent of its (Nvidia) sales were in the US, the rest were international. But products are sold to add in cards and other manufacturers. Many are made in Asia and come back to the USA market after manufacturing.
The TI 4200, said Jen-Hsuan, is the right blend of price, performance and quality. But Nvidia said supply had been tight, and Nvidia will attempt to provide more of these units during Q4.
Products like Doom III will give the overall market a boost during 2003.
Nvidia had designed two types of its Nforce 2 chipset as a result of lessons it had learned with the first Nforce.
The company said it was very successful with OEMs on Nforce 1 but not with the system builders. Nforce was targeted at the Athlon builder, and system builders were the majority of the share. For Nforce 2, Nvidia decided to to build a much more balanced product, which is why it has two configurations.
The primary configuration without integrated graphics is aimed at the system integrator market, which are built by Taiwanese motherboards and then sold into the US and European markets, Nvidia said. The other configuration is aimed at the OEMs.
He said that faster CPUs did not seem to be drawing people to buy in the commercial space. There's some evidence the commercial desktop market is looking for something completely new, Nvidia said. µ
INQUIRER graphics coverage in full
Sign up for INQbot – a weekly roundup of the best from the INQ