Rambus derides Intel chipset

'Inherent bandwidth limitations'

It looks as if California's leading law firm, Rambus, Rambus and Rambus, is spoiling for a fight with Intel.

The company is still smarting from Craig Barrett's remarks about the Rambus/Intel relationship being a mistake and his criticism of Rambus' litigious nature. Rambus CEO Geoff Tate spent some time listing his Intel counterpart's indiscretions during an analyst call last week and the Mountain View company has now apparently called into question Intel's ability to design a chipset.

In Rambus' Form 10K filing to the SEC, dated 30 November, lurks the following passage:

"In the PC/workstation market, Rambus memory technology is used to provide a high memory bandwidth connection to Intel Pentium III processors. The connections are provided via Rambus memory controllers, which are a portion of two chipsets developed by Intel. One of these chipsets, designed for use in the PC workstation market, has been highly successful and Rambus technology is now fully established in this market.

"The other chipset, intended for the much larger main PC market has been less successful due to a variety of factors including inherent bandwidth limitations of the chipset design and relatively high cost of RDRAMs compared to standard memory."

A Rambus spokeswoman confirmed that the 'highly successful' chipset referred to was the i840, and that the 'less successful' one was the i820 Camino. Asked what 'inherent bandwidth limitations' the SEC filing was referring to, she replied: "This is related to the mismatch with the FSB at 1.1GBps and RDRAM at 1.6GBps. The full potential of Rambus technology is not being used."

It is of course no great surprise that Rambus would cite the design limitations of an Intel chipset as having more of an impact on market share than the high price of RDRAM, but having a poke at the poor old Caminogate chipset is about as hard as getting a Reg hack to have a drink.

We asked an Intel spokesman what he thought of Rambus' criticism:

"The Intel 820 offers 'only' 1.6GBps. Regarding the bandwidth limitations, the only thing which could make sense in that context is a limitation in respect to the 3.2GBps with the Intel 840. If the writer sees that as a limitation, I would be surprised to hear that, but it is a free Europe where people can express their opinions."

Rambus' SEC filing also contains remarks which could be construed as casting doubt on the integrity of Pentium 4's Tehama i850 chipset.

"In connection with the introduction of its new processor, the Pentium 4, in late 2000, Intel has designed a new chipset which includes a Rambus memory controller. There can be no assurance that previous problems have been completely solved [our italics], that the pricing of Rambus DRAMs will be reduced to a competitive level or that the Intel chipset and Rambus technology will be successful in penetrating the market segment for PC main memory."

The Rambus spokeswoman denied that this was a suggestion that there might be unsolved problems in the 850 chipset:

"This is not related to technology at all. As a standard statement required to protect the company (similar to a forward looking statement) it can relate to such things as market penetration or supply."

The Intel representative added: "The Intel 850 doesn't have any issue with Rambus, or the processor and Rambus. So I don't know what the writer is referring to." ®

The Rambus Files

Rambus CEO surprises world+dog
We are Borg, er, Rambus
Rambus dropped Hyundai case to avoid tough judge
Intel slams Rambus toll collecting tactics
There is no 'F' in Rambus
Rambus piles pressure on Infineon
Rambus asks feds to stop Hyundai
Rambus takes aim at AMD and Transmeta
Rambus, Intel, Dramurai reach end game
Naughty Dramurai back DDR to hilt
Rambus threatens non-compliant Dramurai