Oracle unveiled an overview of its Sparc plans on Tuesday, and in doing so proved to anyone not yet convinced that things are different in the Sun Microsystem universe now that Larry Ellison & Co have taken charge. If the past decade has taught us anything, it is that Sun Microsystems was a little too cavalier and ambitious with …
It's SPARC not Sparc!
It's long about time that you got it right: SPARC(tm) not "Sparc".
Its not SPARC its SPANC!
The SPARC architecture has been getting SPANCed since year 2000.
We also call Itanium the Itanic.
Power we call Power.
x86 we call Rabbits.
The UltraSPARC T3 looks like it is arriving!
I am really looking forward to the specifications and systems involved in these new UltraSPARC T3 platforms!
"Bring out your dead..." *bong*
A good article with the inferences correctly interpreted, for the most part. Obviously a roadmap is entirely different from actual delivery, let alone success, but hopefully this underwrites my assertions in other article comments that Oracle is not planning to "kill off" SPARC or Solaris - despite fellow commentators crowing to the contrary.
With a nod to M. Python, its like the "Holy Grail" collector of the dead segment...
The Market: 'Ere, he says he's not dead.
Commentators: Yes he is.
Oracle SPARC/Solaris: I'm not!
The Market: He isn't.
Commentators: Well, he will be soon, he's very ill.
Oracle SPARC/Solaris: I'm getting better.
Commentators: No you're not, you'll be stone dead in a moment.
If they do die out, it'll be down to strategic mismanagement rather than intent. Of course, one could argue that mismanagement prior to the acquisition may have done fatal damage already. Maybe we'll need that ninepence to drop the corpse on the cart after all - time will tell!
With any luck, they'll actually do it.
Now that Sun's not in charge anymore, maybe the list of abandoned UltraSPARC projects, ordered chronologically, can start with UltraSPARC V and end with RK. Or at least I damn well hope so.
Time to update this article...
The presentation is available in pdf format on Oracle's website:
Sure, I believe 5 year roadmaps
Who in their right mind believes a 5 year roadmap from any vendor, let alone Oracle trying to convince people to stay faithful to SPARC ? Oracle is hemorrhaging customers to (linux) x86.
Spraklen and Trembley? left, it will all be derivative work and core additions.... little innovation.
And PS: the other article saying Oracle believes flash and tape are the future and the rest of the industry is wrong. Classic arrogance, and hoping you bite. Guess what they sell ? Flash and Tape...
Thanks John for the powerpoint...I believe you less than larry
I am not sure John has actually admitted publicly that ROCK is dead.
I like how he said in an interview "I cannot confirm the 16 core T3 chip for this year" well John that means Rainbow Falls is sometime next year.
Interesting they did not use the chip roadmap charts from the past January presentation.
It looks like Casdade Falls the 8 socket / 16 core chip will be "By 2015" if you look at prior roadmaps it was 2H 2012, so he put a 3 year buffer in the chip. It looks like Yellowstone falls is dead as it was only a 4 core chip and was even further out. As we all know Oracle like lots of cores on chips and low performance cores.
I was surprised to see an M-class in 2014 as we all know SPARC64 VIII is dead, but then again John Fowler said ROCK was coming out any day now and has yet to say it is dead. The Oracle/Fujitsu negotiations should be interesting as everyone knows larry does not like OEM and especially systems that sell instead of the CMT software license insanity.
Our Oracle rep has been telling us Oracle is going to drop all the SPARC core factors to .5 retroactive so we should stop moving things to Power6, but SPARC is dead to us and we never went to T or M class. I told her I don't think Larry is that stupid give the maintenance stream is over 59% of their revenue, but only time will tell.
See you are Oracle Open World,
Economics Will Kill SPARC
Since the mass production of the 80486 processor, SPARC has been largely pointless. SPARC never had a performance edge over x86 since then (it was actually slower for long periods of time). The only remaining argument could be "enterprise" features in SPARC which x86 does not have.
I seriously doubt these features could not be delivered with a custom chipset "around" the CPU.
The core problem is that SPARC R&D expenditure has to be amortized over a small number (less than 500 000/year) of CPUs and a limited business of a few billion dollars per year. AMD and Intel CPUs are a component of a market which generates well above 100 billion dollars revenue.
A much more rational strategy for Oracle would be to negotiate with Intel and AMD to get all required enterprise features added to the high performance x86 CPU cores of these two suppliers. AMD is currently cash-strapped and buying a 30% stake would be easily possible for Oracle. They could then virtually dictate AMD all the enterprise features they needed for a "Enterprise-x86".
Oracle just donned the millstone and now jumped into the water for a long, long swim.
Are we really to believe this time Sun will actually be able to execute on a SPARC project?
Millennium/UltraSPARC V - Canceled.
Gemini, Honeybee, and UltraSPARC IIIi+ - Canceled.
Rock - Canceled.
UltraSPARC T2 was the only project to make it out during the last five years, and it was heavily based on the original Afara design, and heavily leveraged the UltraSPARC T1 core and pipeline. It was as similar to US-T1 as US-IV+ was to the original US-III.
Sun was only able to bring to market acquired and derivative processors designs for the last decade. Perhaps Oracle has brought execution discipline to Sun microelectronics engineering. I hope so. Oracle has somehow managed to completely mismanage the one thing Sun manged well, and that is Solaris. Solaris needs a successful SPARC marketplace.
What It's For
At first, I saw no rational reason for Oracle to value the SPARC technology.
But I eventually learned that they did gain a key technology from Sun. SPARC chips have RAS features, which standard x86 chips do not, but Itanium chips do. But new server chips from Intel have now brought RAS to the x86 world, and this may have made the SPARC technology superfluous.
My perception is that Oracle purchased Sun so that Oracle would be in a position to go head-to-head with IBM. Not Microsoft, not some other company, but IBM. So they wanted a good hardware platform of their own well-adapted to running their database products.
Oracle bought Sun for MySQL and Java
Oracle's #1 threat to their busines was MySQL, now they control their destiny. Their investment in MySQL reflects the revenue to the company vs. what Sun was doing.
Java is a core technology which Oracle and their main competitor use for all their products.
SPARC....it's a huge install base for millions of Oracle licenses which if they killed would move to x86 and Power and would become ten's of thousands of licences and their maintenance stream would be cut in half
No - Sentimentality
Oracle thrived on being a "horizontal" company like Intel, MS or Seagate. Their RDBMS would run on virtually every commercial Unix and a number of mainframe operating systems in the 1980s, when DB/2 ran only on expensive IBM hardware (S/390 and AS/400).
Had IBM made DB/2 available on all major platforms, Oracle would probably not be in the IT conscience. Larry would fly a MS flight simulator.
So the "rationale" that they "need" to vertically integrate is not plausible at all. They have excellent relations with HP and many other hardware vendors. Every IBM sales rep is happy to push POWER boxes to run Oracle.
MySQL is also not a good reason, as it is GPLed as much as Postgresql is BSD licensed. What next ? Hire all PSQL developers and put them on a pointless dud project ??
Airbus does not buy Cessna because the latter are "highly successful" (in selling 100 Hp Piston-Engine planes). They might buy them because the Airbus CEO is a chummy of the Cessna CEO and does not want to see them fail. Especially if Cessna has made some key components for Airbus.
Larry has a sentimental attachment to the Stanford University Network and he thought buying them will count as a Good Deed whenever he knocks at Petrus' door in 30 years. "See, I helped my chummies who were losers and soon would be unemployed. Please reduce my time in hell to 10^15 years. 10^155555 years is clearly doing injustice to a nice American Criminal Tycoon like me !"
Fire burning in The Eternal Hell.
Actually it has...
Actually, SPARC *has* done pretty well against x86. In peak performance, x86 has beat SPARC for quite a while. But, SPARC's performance drops off much less under workloads involving a lot of processes (or threads), random memory accesses, and so on, conditions where x86 performance drops precipitiously. So under realistic loads it's remained pretty competitive with x86 for quite a while. I don't know if it's competitive the last few years, since they've not come out with a new chip (just more UltraSPARC cores on the die), but we'll see how the newer SPARCs pan out.