IBM opens Power8 kimono (a little bit more)

Last May El Reg gave you the little information we were able to gather on the impending Power7+ and future Power8 processor designs and their possible announcement dates in Power Systems machines. Data was a little thin, and intentionally so on the part of Big Blue. But with Oracle kicking up a big fuss over Intel's Itanium …

[Mikel]

Thanks for the update

I haven't been interested in the Power line for a decade, but it's good to keep up with what the other kids are doing.

The reminder about the Xeon E5 launch was particularly helpful. That answers the question about "where do we go when one server can virtualize the SMB?" The answer is "Cheap."

The reminder about the impending launch of the 20-22nm node was also nice.

[Kebabbert]

Interesting

to see what POWER8 is up to. It will fight Ivy Bridge which is 40% faster than Sandy Bridge (according to Intel). POWER7 is a good cpu. POWER6 was not that good.

[Allison Park]

faster intel chips

Intel is focused on adding more cores to become "faster". They do not care about core factor tables or PVU pricing and the true cost of ownership. Their focus on adding cores vs. more powerful cores will keep them out of the mission critical space.

[Anonymous Coward]

Nice to see a roadmap of the past, with something late, and possibly 1 step into the future

Comparing the past roadmap of IBM POWER (with a late POWER7+ and possible almost 3 year away POWER8) with SPARC roadmap (delineating a 5 year vision showing multiple steps with successful checkmarks along the way) shows quite a different philosphy.

http://netmgt.blogspot.com/2011/08/ibm-power-roadmap-now-late-and-almost-3.html

(in most of Register articles, I am surprised that the current Oracle published roadmap is seldom mentioned, but always a snapshot which never shows the progress made...)

[tom 99]

Power7 in 2008?? Power7+ in 2009???

> A case in point is one of the earlier schedules for the Power processor lineup, which had Power6 coming out in 2006, Power6+ in 2007, Power7 in 2008, and Power7+ in 2009.

That's interesting. I was following such roadmaps and have never seen the one with Power7 coming in 2008 or Power7+ in 2009. Perhaps it's available anywhere?

[Matt Bryant]

RE: Nice to see a roadmap of the past, with something late, and possibly 1 step into the future

I suppose the obvious comparison is that, whilst it may be late, Pee8 will probably reach market, whilst anything off the Snoreacle roadmap is more than a little doubtful.

[Allison Park]

Early, on time, late, dead

I thought Power7 was early by a quarter and if you look at the roadmap from year 2000 on IBM has hit the 3 year new generation on time with p4>p5>p6>p7

Not sure where they get accused of being late. The Snorcle fan above is throwing rocks at the wrong house. Or should I say they dont have a rock to throw. SPANCV was cancelled which was the beginning of the end. SPARC64 will not have any future chips.

So the big question is can Oracle keep the T chip going and put it in a scalable system beyond 4 sockets. Their large SMP's before were made by Cray and they had a hard time getting the switch from 66mhz>100mhz>150mhz and finally gave up for Fujitsu technology. I don't think they have the skill to develop a competitive switch like IBM has with Power or even their x5 architecture in their x86 line. Look at how long it took HP to finally get back into the 8 socket space with xeon.

Cheers Allison

[Kebabbert]

@Allison Park

"...SPARC64 will not have any future chips...."

Allison, once again:

I would appreciate if you would stop attacking Oracle by spreading false rumours about the demise of SPARC cpus. This is not the first time you have been doing that. I have numerous times asked you to show links that supports your claims, but you do not. That makes one wonder if you are making up all your claims? Because you dont have any support for your claims, are your claims made up by you?

http://blogs.oracle.com/hwpartner/entry/nyt_om_n%C3%A6ste_generations_sparc

" ...Oracle expects to work with Fujitsu to have updates to the Fujitsu-designed SPARC64 VII+ processors for the M-Series servers with announcements in 2012 and 2014, as indicated in the roadmaps shown at Oracle OpenWorld 2010, which have been published to the Internet. Both the Oracle T processors and Fujitsu SPARC64 processors are designed to the same "V9 architecture" so that they run the same applications...."

.

.

On the other hand, when I write something, I post links which proves that I never make things up, nor am I spreading false rumours. I am only spreading true rumours, I am merely quoting media. Hence, I can not be accused of spreading false rumours.

I must remark that mere speculation or opinions are not spreading false rumours. If I say that I like hamburgers, no one can object against that. Or, if I say "I suspect that it will rain tomorrow" no one can object - that is my personal belief.

But this is questionable: "I have heard from Oracle senior management that SPARC will be killed" because it is a declarative statement - and hence it is classified as a lie because Oracle would not say so.

If you dont understand the difference between a declarative statement or an opinion, please study some logic. But the difference is vast. In short: "opinions are never wrong, and no one can dispute them. Facts can be wrong and they are disputable"

.

.

Here is a true rumour: AIX will be killed, says IBM. IBM has officially said that AIX will be killed sometime in the future. To prove that I am not making this up, I show a credible link:

http://www.zdnet.co.uk/news/application-development/2003/01/29/ibm-linux-will-replace-aix-2129537/

Hence, I am not spreading false rumours. I am not a FUDer.

.

.

Here are some other true facts:

- POWER6 servers were several times faster than x86, and POWER6 servers costed 5-10x more.

- POWER7 is only ~10% faster than Intel Westmere-EX and costs only ~3x more.

We see that the performance advantage that POWER had, has more or less vanished. And we see that the price has decreased a lot. All this is true. The observations are true. We also know that IBM does high margin business and cancels low margin business.

Now, what do you think about POWER8? Will it be slower than Intel x86 high end cpus? If so, then price must be decreased even further or no one will buy POWER8. This means POWER will soon be low end business. And, IBM has said that AIX will be killed off. If IBM kills POWER, then AIX will follow. We know that AIX will be killed. So the future of POWER is quite uncertain? Or?

Note that I have lot of supporting evidence for my speculations and opinions. I am not making anything up.

Allison, if you can provide links and supporting evidence to your rumours, then I will be silent and will not say anything. The problem is, you can not find supporting evidence that SPARC will be killed off, because Oracle is betting heavily on SPARC.

As long as you attack SPARC, I will defend SPARC and also posts links that shows how slow IBM Mainframe cpus are, that AIX will be killed, that POWER7 is roughly equivalent of Intel x86 Westmere-EX, etc. If you want me to stop post such things all over the internet, making people doubt on IBMs technology, you better stop spreading false rumours. Of course, if you spread true rumours about Oracle, I will be quiet.

I dont like lies. True facts I can not dispute

[Allison Park]

FACTS

The "M-Class" servers in 2012 and 2014 will have Oracle T based chips in them. They will not be Fujitsu chips. Hence the statement that SPARC64 is on its last chip. Never said SPARC is dead. Oracle raised the maintenance costs which are generating a ton of profit so they will let this cash cow die slowly. Oracle is not stupid like HP and announce they will kill something at an earning announcement then charge $99 for something that costs $400 to make.

Ask Oracle who's chips will be in those boxes. They will be a forklift upgrade from the current M boxes and will have T chips in them. I highly doubt Oracle will be able to create a T based server that goes beyond 8 sockets but that would be speculation.

To reference a statement from over five years ago which is also refuted in the same article and does not match reality today is being a FUDster, but then again I am guessing you were not burned by Sun's hardware mistakes in the past.

Cheers

[Anonymous Coward]

Blog entry from 2004 when USV and the first T chip were cancelled

Pretty much on the money except then Rock was cancelled and Sun went belly up and was bought by Oracle to suck as much profit out of the install base as possible.

Old 04-09-2004, 03:23 PM

Ade Fewings

Guest

Posts: n/a

Default UltraSparc V & Gemini cancelled

[url]http://www.theregister.co.uk/2004/04/09/sun_kills_us5/[/url]

Does this signal the beginning of the end of the Sparc line? I think

maybe it does. I can imagine Fujitsu chips filling in the Sparc high-end

for a bit, but i suspect AMD64, IBM's POWER chips and [unfortunately]

Intel's offering[s] will be the sole competitors on the market in a few

years. It seems that Sun's chip development will be focused on bringing

together some of the stuff in Niagra and Rock which will presumably be

focused on multi-core lower-performance than the high end SMP stuff that

UltraSPARC used to be.

And if this all changes where does the future lie? Given time will

Solaris become non-sparc focused? Where does the line get drawn - when

Sun contributes stuff to Linux and ends up focusing on that? To some

extent it seems to me to be a question of where the line gets drawn and

how to draw it - how do Sun stop the gradual slide - in fact, should they

stop it? This news and the Microsoft stuff may be the best business

sense, but it threatens to take away some of what makes Sun what they are

and in my view a world without that would be a poorer world. Perhaps a

market niche or area of some sort of realy top-quality support and work

could be found, but i fear that Sun as we know it may be dissapearing.

This is particularly a shame as i have been more excited than for a long

time with things like Project Looking Glass and some of the Solaris 10

features coming soon.

What does anybody else think?

All the best

Ade