Server buyers who have been waiting for IBM to ship the Power6+ chip will be surprised to discover that it already has. It just didn't tell anyone. Crazy, right? Remember the Power Systems revamp Big Blue did back in October 2008? Well, guess what? Several of those machines were based on the Power6+ processor, the kicker to the …
All rather odd
It is all rather odd, but if it's working commercially for IBM one can't really question their marketing strategy.
One thing I don't really understand though is why machines with such enourmous single thread performance don't make a bigger splash than they do. I mean, there's plenty of techies, engineers and scientists out there who could really do with a core running at 5GHz! That's way quicker than anything from Intel or AMD, especially if one contemplates an alitvec unit running at that speed too. I know that IBM make these processors and machines with an aim to put them in to business applications, but there's far wider applications out there beyond merely adding up people's bank account balance. Wasted I tell you, wasted.
So is anyone out there using one of these machines for a scientific / technical purpose?
Slow Power6 debut explained at last?
I always wondered why it took IBM so long to get the Power6 across the whole server range, and it looks like there was an issue with the design that meant some systems had to wait for the "Power6+" CPU. And then the Power6+ cores had to be wound down for some reason - don't tell me IBM wouldn't have loved to release them at full speed. All very interesting. It'll be interesting to see if IBM see fit to answer TPM's queries.
@All rather odd
It's because the people who control the budgets and a good many of the people who use the computers don't understand how the work, or the limitations of the architecture. And the fact that IT is a very "Me Too", industry, so if it has worked for one application it must work for another, otherwise why would we have Windows.
Sadly I'm in the school of choose the right tool for the right job, rather than most of industry these days that seems to think that making all problems fit the tool is the right answer.
Makes perfect sense to me
IBM have done this in the past and it's all part of their "OnDemand" model - I think. We had an iSeries at a company I worked for and were sold on the fact that, if we wanted more power; simply call Big Blue and they'd send a wee update down the line to the machine to crank it up. Fantastic in the retail world where the bulk of your sales, and therefore processing, is at Christmas. The rest of the time, it all depended on whether Tricky Dicky in Merchandising was creating a desktop database or not... You know who you are!
Anyway, the reason you don't have one of these babies on your desktop running Crysis is cost and design. Sure you could run WinXP on a virtual machine on an i-, p- or probably even z-Series, however the machines are designed for out and out processing performance not 3D computation or multimedia. OLAP, OLTP, big database crunches are what they are good at: not replicating real-life environments through physics models.
The reason they aren't used more widely in the science and engineering world is that the software is not available for those sorts of applications. You can write C or even ForTran for an iSeries, so theoretically you can write applications for these sectors; it's just that nobody has really done it yet and to my knowledge.
WTF is a kicker?
A design revision?
Ask the Oracle about hardware futures.
The pressure on Big Blue to do something will come through the quad-core "Tukwila" Itanium arriving from Hewlett-Packard and Intel, which was pushed out to mid-2009 as this year was getting rolling. Tukwila is a year or more behind schedule,
====> Tukwila will not ship in systems until November, after HP's fiscal year end.
====> The HP reps have been telling customers this already.
====> Not only will the chips have twice the cores expect Oracle to also increase the multiplier to .75 so you will need three licenses per chip instead of 1.
Sun Microsystems' 16-core "Rock" UltraSparc-RK chip is at least a year behind, if not more.
====> Rock is shelved, ask Mark Tremblay who was fired and now works for Microsoft. Seriously look at all the systems Sun currently sells and you can see why they cannot make money. Expect Oracle to eliminate over half of the models.
RE: Ask the Oracle about hardware futures.
"....Tukwila will not ship in systems until November, after HP's fiscal year end. The HP reps have been telling customers this already...." Strange, but hp wouldn't give us a ship date without an NDA, but all I can say is the date wasn't November. Are you sure your "customers" got the NDA briefing or just the FUD?
"....Expect Oracle to eliminate over half of the models...." Nah, just the ones with any Sun SPARC chip, so all the CMT line. Not sure what Larry will do about the SPARC64 boxen, I suppose that's down to renegotiation with Fujitsu. After all, SPARC64 is a Fujitsu concern so Larry doesn't have to shoulder the development costs, that's if they can convince Larry there is a market for future SPARC64s. Galaxy might survive in some form for Oracle appliances. The rest? Scrapped or sold off. By my estimation that would leave less than half the models.
By the way, when is Power7 actually due now? The IBM roadmap used to have Power7 as 2009, then 2010, now the latest version just says "future".....
Isn't it out there in the public domain that Power 7 = H2 2010 or there abouts for GA?
The prices on those new SSDs for the kit is.... interesting.
HP needs to be careful
I think HP needs to plan the Tukwila launch carefully bearing in mind when their year end is (October I think) this could have a big impact on their BCS/Itanium numbers if they don't get it right.
Funny how people here like to dis-inform...
Everybody calls the Rock CPU from sun, dead, shelved or whatever .... meanwhile opensolaris got support for this dead processor in snv_100, gets bug fixes for it (http://dlc.sun.com/osol/on/downloads/b111a/on-changelog-b111a.html). Pretty good for a dead processor...
the facts are: as of last week Rock is NOT dead.
@Matt Bryant: get your head out your a**, you might notice that there is other options than HP ...
RE: RE: Ask the Oracle about hardware futures.
Never confuz Intel's chip announce and when HP can finally ship products to customers vs. early ship programe for testing.
Itanic was delayed because it looks so bad vs. Nehalem. Sorry HP-UX customers we wont port it to Xeon it would make too much sense for everyone.
Oracle still refuses to even put its ebusiness suite on Itanium.....amazing after all these years HP still does not have support for Oracle ERP
A few things happend on the way to the ROCK concert
Since "Solaris got ROCK support"
1) Sun revenue has been in free fall
2) Sun started positioning ROCK between the Tclass and Mclass
so much for replacing Fujitsu
3) Mark Tremblay was fired or something like that and now works for Microsoft....go figure
4) Sun waved the white flag and begged every company in the world to buy them
5) Oracle bought sun for 7.5B - 2B in cash = 5.5B cash cow
6) Oracle will fire 15K Sun employees and halve the number of systems sold
7) if Oracle gets fed up with Solaris/SPARC they will sell it to the highest bidder Fujitsu or HP
According to Jonathan Rock arrived over a year ago and the best he can say publicly is
"I THINK we said it would be the end of 2009" I would have thought he would know.
Just a few points.
You need a 2 drawer power 560 to house 16 cores and 384 GB of RAM, the power 560 is basically a neutered power 570-32.
@Matt Bryant on POWER6 power usage.
Yes it is irritating that power usage from one processor to another isn't using the same numbers.
But try to go to one of the IBM system energy estimator http://www-912.ibm.com/see/EnergyEstimator
And try to punch in a firstname.lastname@example.orgGHz way power 570 (2-16) and then a email@example.comGHz way power 570 (4-32).
Now the difference in estimated power is 120 Watt that is for 2 additional POWER6 chips and 2 additional L3 Modules. That's not that bad.