Compaq has chosen Digital's Alpha processor over Intel's forthcoming Merced technology to power its Tandem Himalaya high-end server range, further securing Alpha's future. The company described the choice as a difficult one, admitting that there were some "political" issues that had to be considered.
The Himalaya servers, used mostly by banks and stock exchanges for continuous computing operations, currently use Mips processors from Mips Technologies, recently spun off from Silicon Graphics. After the next two iterations of that processor technology, it will be replaced by Alpha chips as a result of Compaq's recent acquisition of Digital Equipment. The decision is further evidence of Compaq's wholehearted backing for the processor - something far less evident when the Digital takeover was first announced and Merced was still on time.
"In the world of 64-bit computing Alpha gives us a two-year window of opportunity," said John Rose, head of enterprise computing at Compaq.
The version of Digital Unix being developed for Merced will be code compatible with that for Alpha, said Rose, and use the same data types. Version 5.0 of Windows NT will also run on both platforms, when it eventually emerges.
"We want to get high availability NT ready sooner than Microsoft would have done, left to its own devices," he said.
He insists that Alpha, compared to Merced, will have 1.5 to two times the performance for around half the cost of production.
Despite this, Alpha is likely to remain a price premium product. Compaq refuses to promise a cost dividend for customers, Rose noting that cost is not the same as pricing and Intel has yet to announce pricing for Merced.
Alpha servers will remain the high end offering.
Differences still remain between the outlooks of the merging companies on Alpha. Rose insists that Merced will be the key technology to customers - Jesse Lipcon, Unix systems vice president at Compaq, admits that customers will make the choice but that Alpha will still make the best systems.
A smartphone maker fiddling its benchmarking scores? That's unusual, isn't it?
'We are making good progress on 10nm,' claims Intel
Engineer calculates that Chengdu's plan to replace streetlights with artificial moonlight would cost $100bn
Research could also apply to other 'space weather' events involving hot, fast-moving plasma