Linked by Eugenia Loli on Fri 20th May 2005 20:13 UTC
SGI and IRIX Can a proprietary Unix be a desktop OS that competes with Windows, Mac OS X, and Linux desktops? Although it may lack the visual effects of OS X, and installation is tricky in parts, Irix is a stable desktop OS -- possibly because it runs only on SGI's own hardware. Read the rest at NewsForge.
Permalink for comment
To read all comments associated with this story, please click here.

"Why would Solaris/Opteron/Numaflex boxes make sense when switching to Linux from IBM is "inevitable"?"

IBM is putting its bets on Linux because IBM was the sore loser from in Unix wars and AIX simply can't stand up to Solaris. I guess for IBM switching to Linux is "inevitable" because it can't compete with Solaris having just AIX -- AIX is years behind Solaris technologically and can't compete price-wise (Solaris is a screaming bargain compared with AIX). I guess Linux is just a very comfortable FUD that IBM can use to its own advantage.

There was quite a good article a while back; I think the most glaring issue was the threads implementation of AIX, which apparently is crap-o-la and in terms of fine-grained scaling, its a gludge.

With that being said, however, personally, I think that the better step for IBM would have been to team up with Apple, and take on the heavy work of ultra-fine-graining the Darwin kernel, and start adding features that are found in high end operating systems like Solaris.

IBM would then be able to start selling POWER based servers to small, medium and large businesses, once intrenced into Windows because of the 'ease of manage-ability', but can now move easily to MacOS X because of its ease of use, and its UNIX core.

Either that, or simply port OpenSolaris accross to POWER, and make that part of the larger portfolio of operating systems to support. People would then be able to get the raw speed of POWER and the great reliability of Solaris.

"Maybe SUN should buy out SGI and start producing friggin huge Solaris boxes using Opteron using NumaFlex - now *THAT* would be a great system - the only downside? it would cannabalise their existing SPARC systems, that run slower than a dog with no legs - and no, I'm not exaggerating - just look at the TPC and SPEC2000 benchmarks for their SPARC workstations"

No, Opteron on NumaFlex wouldn't cannibalize SPARC systems. NumaFlex is fine for High Performance Computing but it wouldn't fare as well the SPARC FirePlane interconnect based systems -- two thirds of all Oracle installations out there go on Sun SPARC for a good reason. NumFlex is just not suitable for latency and data skew sensitive applications, such as transactional databases.

Sounds a bit like mainframes, and how they're more suited to large batch transactions rather than lots of piddly little ones.

And your reference to SPARC performance as "dog with no legs" is too harsh in all fairness (to both the dog and the processor ;) ). SPARC delivers reasonable performance even though it is currently lagging behind Power5 -- processor performance is leapfrog game and SPARC is likely to leapfrog Power with the release of Sun-Fujitsu APL line pretty soon. If UltraSparc performance is not good enough for you, just buy Fujitsu SPARC64 IV system which have performance comparable to Power5. Like it or not SPARC has still got by far the absolute best 64bit story -- guaranteed backward compatibility and thousands of certified application, something no other 64bit platform can claim.

Well, IMHO, its taking SUN *TOO* long to move over to SPARCIV - they should have done it at this latest conference; their *WHOLE* line should have been moved over, and by now there should have been a joint SPARC development facility where by the two companies work together on the processor, thus reducing the over all costs.

Regarding their workstation, today, they should having SPARCIV workstations, PCI-E throughout the machine, along with the latest ATI FireGL graphics card. Believe me, if they did that, bundled it with a SATA drive, DVD Writer plus a 1.9ghz SPARCIV processor, you'd see SUN workstation sales double overnight. The perceived advantgae of Lintel stations would vanish, and SUN would regain the title as the "king of the bean hill".

Why don't they do that? they'd much rather spend the $7billion they have in reserves on "new pricing models" (yes, I'm as confused as you are, as to how changing their pricing structure should cost them money; then again, its Schwartz, and alot of the stuff he does and says doesn't make a whole lot of sense at all).