The first of the chipsets is the long awaited "Placer" (7505) chipset. In a note on the Intel website, the firm describes two errata for the chipset - the first is that AGP signals do not meet the AGP 3.0 specification, while the second is that the AGP Prefetch Cache has to be disabled.
Intel says in the errata notification PDF, which can be found here, that no fixes for either of these problems are planned, but the first can be fixed by a board workaround, while the second can be fixed by a BIOS workaround.
These problems also appear to affect the "Granite Bay" E7205 workstation chipset. That, as we've reported here in the last two weeks, will be displaced by the "Canterwood" chipset as early as Q2 of next year.
Intel has also made a change in the voltage specification for the E7505 chipset, details of which can be found in this PDF.
The voltage is changing from 1.2 volts to 1.3 volts, and Intel is advising a fresh revision of the chipset will be available in early January next year.
One system integrator said: "Promoted for their AGP 8X interface, how big of a performance hit will there be when the AGP Prefetch Cache is disabled? How much of the hype of the projected or preliminary performance numbers for systems built around these chipsets was calculated or gathered (under ideal circumstances) before it was determined that the AGP Prefetch Cache would have to be permanently disabled?"
And another, end user, wrote to the INQUIRER saying: "I postponed a much-needed purchase of a new workstation, waiting for the Placer chipset to be released. Now, I'm confused as to whether a system built around the 7505 chipset would even be a wise purchase."
A number of third party companies including Supermicro, Iwill and others have built third party boards based on the workstation chipsets. We'd be interested to hear what effect the changes might have on system functionality using these parts. µ