Efforts by Sun Microsystems Inc to meet the performance goals it set for the upcoming UltraSparc and SuperSparc II chips are likely to fall short of the mark, sources report. Sun’s first 64-bit chip, the UltraSparc, is to be taped out in the next few weeks. The company should have first silicon by September or October, Sparc Technology Business president Chet Silvestri said – and that would be a couple months late. He estimates the part will be able to be clocked at up to 167MHz, which, sources claim, was Sparc Technology Business’s target, although performance will be 250 SPECint at best, not the 275 SPECint hoped for. The SuperSparc II, which is already in silicon – and is said to have booted immediately – will hit a clock rate of only 90MHz (125 to 130 SPECint), 10% less than the 100MHz (140 to 145 SPECint) hoped for, and will be a quarter off, a delay that by current thinking should produce a 15% performance improvement to claw back competitive position. Silvestri contends that a second iteration of the SuperSparc II chip in September could see 100MHz parts, but claims to be unworried if it fails to do so.
High-volume merchant chips
Sparc Technology’s business model, he says, has changed and he is now more interested in producing a steady flow of high-volume merchant chips than sorting through the batches produced to separating 100MHz parts from 90MHzs and losing sales of the lower-clock parts to the dumpster. Although, according to Silvestri, the SuperSparc II will be the highest volume Sparc chip in the Sun line for a year after it arrives, sources suggest its failure to clear the 100MHz hurdle could finally open Sun to the entreaties of Ross Technology Corp on behalf of its 100MHz HyperSparc chip. SuperSparc II and its genre, however, are essentially a passing fancy compared with where Sun’s more serious intentions lie – provided it is not eventually forced into the enemy iAPX-86 or PowerPC camp. SuperSparc and MicroSparc are set to fade away, to be replaced by a single pipeline of bi-endian UltraSparc chips, differentiated only by packaging, that start at 150MHz, Silvestri said. Sun’s transition to all UltraSparcs, which includes a new bus said to be dubbed the UPA, should be effected by 1996, he said. Although this seems to suggest a new roadmap that leaves in question the previously proposed 100MHz to 125MHz MicroSparc III, which was supposed to be in silicon next year, and sources question the competitiveness of a future line starting at 150MHz. Sun’s ambitions to make SuperSparc more of a merchant chip seem to fly in the face of policies enacted over the past few months that put Sun technologies, ostensibly available to any Sparcsystem builders who want to buy them, beyond their reach. Stated policies to the contrary, Sun has declared certain input-output and graphics elements of the SparcStation 20 chip set off-limits to all but a favoured few. Silvestri maintains that for the sake of volume and marketability, he does not want to turn all the iterations of the technology Sun uses into a merchant market product, and that what is at issue here is neither more cost-effective nor functionally different nor more powerful than the SparcStation 10 chip set that third parties must currently be satisfied with – a position that raises the question why Sun would then use the Sparcstation 20 technology itself. Sources claim that Sun’s covert purpose in the action, intelligence of which it has attempted to suppress, is to kill off the clone-makers and force those that continue to follow in Sun’s footsteps into the product differentiation that has so far eluded them.
