Let me give you the status quo on our issues with the Tiger MP and the upcoming workstation article. As mentioned we’ve been talking to support engineers at both Tyan and ATi and have sent and received many emails detailing our problems and their suggestions to solve them. Last week we’ve received a Tyan Tiger MP Revision 3 as a replacement for our older revision and that turned out to solve a number of issues right out of the box. Shutdown from within Windows worked fine now instead of before when the system simply halted with a blank screen. Tyan also sent us a custom BIOS to take care of some problems people have with the Adaptec FireConnect, unfortunately this didn’t work with our older revision board.
Furthermore we were able to verify that the Revision 3 of the Tiger MP indeed carries a new PLL clockgenerator, which has FSB settings up to 150MHz. Unfortunately there’s no BIOS support for these alternate FSBs so you’ll have to revert to 3rd party utilities that can interface with the PLL through the SMbus to make use of these. Naturally we gave it a try and we unfortunately weren’t able to get the system up and running stable at FSBs over 140MHz, 145MHz was possible but caused for many errors and even some BSODs within Windows, so we haven’t explored any further. As for our testing and evaluation of the motherboard and the 760MP platform we naturally didn’t use any alternate FSBs.
Fig 1. ATi's Radeon 8500, causing problems in our Tiger MP testbed. The jury is still out though whether it is ATi's drivers or Tyan's Tiger MP BIOS.
One problem remains though, and that’s one that has gotten the full attention of both Tyan and ATi, and that’s the inability of the Radeon 8500 to finish a run of SYSmark 2001 and the random lockups in 3D games. The system halts with a ‘page_fault_in_nonpaged_area’ in ati3dag2.dll, which would point towards a driver issue or some memory paging problem between the videocard and the motherboard. However this problem seems to only occur with the Tiger MP and the Radeon 8500 as the Radeon worked fine with other motherboards, and so did the Tiger with other videocards.
As you read this, engineers at both ATi in Canada and Germany as well as engineers at Tyan in both the USA and Germany are working at the issue. We’re hoping to see some results this week and maybe give you an update on what actually caused these BSODs. For now we’ll be working on other articles and continue testing the Tiger MP system to see how it holds up running all sorts of different workstation class applications.