Posts by Allison Park 281 publicly visible posts • joined 20 Jan 2010
Power7+ and Power8 with full details.
I saw it on the website called the register http://www.theregister.co.uk/2011/08/31/ibm_power_chip_roadmap_update/
All I have seen of Poulson is 32nm and more cores.
Kittson is only a name with promised socket compatibility which now I hear will not be the case.
sorry yes Tukwila is quad core while everyone else is octo-core ..... my bad still not something to be proud of.
and you refer to benchmarks that oracle wont release as "proof" that Itanium is competitive? Sorry HP gave us the SAP numbers for tukwila and they were embarrasing compared to the DL980 let alone compared to the Power 780 we ended up buying.
Nice try....thanks for not addressing the two generation old fab tech, the incredibly slow chip speed, or the fact that kittson is the last chip for Itanium. Will be interesting to see if the initial Kittson will be crippled to give the perception that there is a kittson+ later like HP was planning.
kisses
"But you still have to use the crippled versions of the Power7 CPUs, not the full-power (4.25GHz) ones."
Are you the HP Bigot who likes Tukwila which is sadly still dual core, sadly 65nm, 1.75GHz?
How's that glass house you are throwing rocks out of? And I say glass house cause Oracle outed HP's planned demise of Itanium so now everyone knows the real roadmap.
i believe they call it a distributed switch. They took the power5 chipset and put it on every power6 chip. Interestingly that allows them to not only avoid the latency of a chipset but the switch runs at half the speed of the speedy chip. More reason fast chips are good.
I think power7 was a continuation of the switch technology
First there was scale up...buy big expensive boxes
Then there was scale out....buy lots of inexpensive boxes and puts lots of expensive cluster software on them
Finally there is scale-in...buy big boxes again but take all of those little apps and create hundreds of virtual machines.
I first heard the expression when I was talking to Brad Day in 2005. Formally of Giga then Forrester now doing his own thing.
I thought it was a cool way to describe virtualization on big boxes.
tudalu
These sound similar to what Cisco is trying to peddle, but are open. CISCO is all about their expensive switch and pushing EMC's expensive disk and vmware. CISCO does not have advanced x86 systems like IBM"s x5 technology.
I can certainly see moving our websphere farm to these systems. We are and never will be a weblogic shop so Oracle can dream about selling their exalogic.
HP is going to be in a tough spot. They bet the farm on blades and their odyssey announce is just putting future x86 chips on the superdome blades.
Personally I think the hay day of blades are over. Vituralization technology is more important than squeezing nodes sideways in a chassis.
What will be interesting is to see how long before Oracle supports this. I see no reason why they wouldn't on day one. After all they only care about software.
As far as people thinking ibm only cares about services they need to relook at the numbers. IBM is focused on profit driven by software and that is why I think they are pushing these.
What would be even more interesting would be the breakdown of exadata sales based on even discounts of hw and sw. Oracle has been artificially inflating hw sales by doing a zero % discount on hw and 75% discount on software. This is why hardware has not totally cratered and why software is missing their projections.
What does "- I'll that that bet! " mean?
Glueless is only a good thing if you have enough interconnects. the x4800 only has 4 interconnects so in order for chip 2 to talk to chip 6 it needs to interrupt and talk thru chip 4.
Thats why IBM has had a glue chip for 5 generations and why HP finally got back into the 8 socket race with a glue chip in the DL980.
The future m-class is a modified T chip and everyone knows it, the question is not how many sockets Oracle might try to patch together but how the interconnect will work.
After all its not just about the brain but the body that its in.
tata for now.......my ride is here
pure BS.
Do you cut and paste this crap you keep spewing?
Power7 is a chip that can be 4,6 or 8 cores depending on the performance customers want.
We just bought some 780's with 6 core chips so we would have the right balance of high performance cores and density per footprint.
Power7 has the ability to run 1,2 or 4 threads per core on the fly depending on the applications thread vs. multi-thread requirements. Last time I checked the T chips still only ran one thread at a time and keeps on relying on thread swapping in supporting 8 "simultaneous" threads down from 16 which did not work well.
Oracle has announced their future will not have any SPARC64 chips.
Oracle has a roadmap with M-class systems which are really T-class and Sun engineers have not been able to build a >4 socket system since 2004. I am willing to bet their M system will have the same terrible interconnect as the X4800. Oracle is not only glueless they are clueless.
I would say Larry would get out of hardware all together but he is a stubborn little man.
Tata for now........off to the spa.
Cisco's expertise is not servers. I do like their clever marketing ""potential thermal event". I guess they didn't want to say a focking fire might burn down your whole datacenter and destroy your company. We looked at UCS a year ago but where not satisfied with the costs that yet another HA/DR environment would cost us. And now it looks like not only did we make a great decision but it might have killed us.
Looking forward to seeing the flex stuff TPM disclosed. Still waiting for our IBM rep to give us a preview.
Dear Troll
Here you go. Oracle was finally allowed to release the information they got from HP. If you compare the blacked out version to the clean version that is where you get all the facts I stated.
This is turning out to be a romance novel gone bad.
http://www.scribd.com/doc/79962880/HP-v-Oracle-Amended-Cross-Complaint
Meow, Hiss
Here you go. Oracle was finally allowed to release the information they got from HP. If you compare the blacked out version to the clean version that is where you get all the facts I stated.
This is turning out to be a romance novel gone bad.
http://www.scribd.com/doc/79962880/HP-v-Oracle-Amended-Cross-Complaint
OK Oracle cannot sue HP for fraud because of their lying. Big deal.
Oracle certainly caught HP not being truthful to its customers and that is 100 times worse.
"10 years of commited roadmap" "the longest roadmap in the industry" and they all knew the intel agreement only lasted until kittson in 2014 and Intel wont even make the chips after 2017.
The other key thing to see is Kittson will not be tukwila/poulson socket compatible like HP is still saying. It will be a xeon socket. When will HP finally admit that.
I guess they can't go forward with the handicapped Kittson idea to pretend there are two versions.
Oracle is not clean either. SPARC64 is dead and they are still "no commit" except for that one solaris vp that let the cat out of the bag.
Meow
Oracle kills off Itanium support while there are two more chips in the design process.
Oracle kills off SPARC64 (no more chips in the future) and refuse to admit it until a software VP slips up and says that they won't be using the fujitsu chips any more.
I always buy stock of companies on their way up in the patent race and short companies on their way down. Thanks to Mark Hurd I made a killing on shorting HP stock while everyone else thought it had nowhere to go but up. I am taking a few vacations to the Caribbean thanks to Mark and his cost cutting (the womanizing strangely enough was an unexpected bonus).
Not ready to short Oracle yet. He seems to be just a spokesperson for hardware vs. really setting the direction.
One thing is for sure when people cut R&D it takes years to reverse the trend vs. when people invest more its a quick ROI.
Oracle co-founder and CEO, Larry Ellison, said that the company sold over 200 Exadata and Exalogic systems in the second fiscal quarter – by systems, Oracle appears to mean "racks" – and added that the company will do 300 machines in Q3 and 400 machines in Q4
"appears to be systems" what does that mean? Full racks? Quarter Racks? or are they counting every compute node and storage node as two?
Now is the time for real numbers and the truth to be told. Larry didn't seem to know what to say when asked what happened to the 3,000 number you guys said before.
Watch that analyst not be invited to the next earnings call.
yes IBM updated their roadmap. while IBM did not want to release too much data about future chips so intel copies them with x86 they also wanted to show that poulson is not as good as p7, kittson not as good as p7+ and since there will not be an Itanium chip after Kittson it is obvious p8 continues the leadership.
Clockworks....new generations have been every 3 years and a + between 18-24 months. Itanium has been years late and will never be competitive before it is discontinued
Yes IBM saw the writing on the wall that Itanium was dead when Merced became just a development chip.
Unix actually grew 2% so far this year and if you look at the US market IBM is now 55% vs. Itaniums 15%. Revenue is what matter because that is who vendors invest. Oracle is getting out of the "shipment" business asap.
Look at IBM's recent earnings reports websphere and Power have been the stars.
What you missed is the bit that this is exactly what Intel has been planning and announcing for a very long time - socket-compatibility between Itanium and Xeon, which means Itanium gains even greater cost savings by maximising the investment in Xeon tach, whilst Xeon gains many of the high-availability goodies that have traditionally only been in RISC/EPIC servers. And it will be an industry unique tech.
Where is the IBM equivalent 32-socket Xeon servers?
HP and IBM both have an 8 socket box....and Odessey is not until 2015...first to announce a strategy because Oracle outed the dead of Itanium does not make you a leader.
Would they be able to offer hardware and software partitioning tech inside that same Xeon frame? Both IBM and HP have blade enclosures with x86 and itanium/power whats your point
Could they do it and offer the same mainframe-class resilience as Superdome2 already has?
Superdome2 does not have "mainframe-class" resilience
Getting Xeon into Superdome2 is not detracting from the Itanium-based Integrity line,....no distraction at all...everyone knows Itanium is dead and HP is quickly moving their whole "Integrity" team to x86.....why the date is so far from no is a mystery...I thought the common socket, same QPI links would make it sooner than a target date of 4 years out
Itanium would not be profitable for Intel unless HP paid X.X Billion to keep it going and that is really just base mfg cost.
Intel would rather develop mission critical for xeon than itanium then wait to put it in xeon.
April 13, 2011 - "We used to position Itanium up here: highest performance, highest reliability .. It is really now a choice of operating system, Xeon's reliability and performance is now equal and in some cases better than Itanium…. So, if you like HP-UX, OpenVMS, HP NonStop, other mainframe operating systems, we are fully going to support you on Itanium,” –
Kirk Skaugen, vice president of the Intel architecture group and general manager of Intel's data center group, during his keynote at the Intel Developer Forum
Interesting how HP was so scared of their ex-CEO working for Oracle they tried to secure support for Itanium. And the "Itanium Collaboration Agreement" obviously has an end date to Intel support which was blacked out but their obviously is one. If we still bought Itanium systems these are the questions I would be asking my HP rep.
1) How much does HP pay in the "Itanium Collaboration Agreement"
2) What is the end of support date Intel is committed to in the "Itanium Collaboration Agreement"
3) what is the minimum performance required of future chips
4) Is the only "contract" with HP to keep supporting Itanium a reaffirmation agreement which was created because HP fired and Oracle hired Hurd?
5) Does any contract talk to future products or just support
6) How do you justify tukwila pricing of 2X montvale when the license performance is not higher
7) Is the Odessey announcement related to the reality of Itanium?
8) HP Itanium revenue is in a free fall according to Gartner and IDC, can HP afford to invest?
9) Itanium Poulson looks like it got moved out to 2013 on the latest roadmap (chip end of 2012, systems 2013)
10) Does Meg care about Itanium and does Matt B feel betrayed?
You can argue who is better but there is one thing there is no argument about and that is HP is missing.
Itanium is contracted for only two more chips which Intel is not investing to make competitive.
Project Odyssey is the 2014 project to get x86 into the superdome chassis to replace Itanium and HP-ux with Linux.
And Meg is doing a firesale of touchpads on eBay and bragging about project moonshot which is not even their technology.
I didn't realize Power systems was twice as big as HP and Sun. When you sift thru all the FUD from vendors the best indicator of success is where are customers spending their hard earned money. (especially in tough economic times) I never gave a lot of value to the shipment numbers since we want less servers that do more. It would be cool to see the numbers for "virtual servers shipped".
Power +27% $1.21B
Oracle -11.6% $550M
HP -18.5% $540M
Fuj -35.8% $65.1M
Next quarter should be rough for HP given it's their fiscal 1Q and the recent announce to replace Itanium with Xeon but not port HP-UX.
Burrr it's cold to be back.
An oligopoly is a market form in which a market or industry is dominated by a small number of sellers (oligopolists). The word is derived, by analogy with "monopoly", from the Greek ὀλίγοι (oligoi) "few" + πόλειν (pólein) "to sell". Because there are few sellers, each oligopolist is likely to be aware of the actions of the others. The decisions of one firm influence, and are influenced by, the decisions of other firms. Strategic planning by oligopolists needs to take into account the likely responses of the other market participants
HP has decreased its R&D from 6% to 2% in recent years and cannot afford nor do they have they staff to keep up so they are moving their Integrity team to work on x86 and their HP-UX team to get their middleware certified on Linux. They will concede the O/S to redhat but want to keep openview and serviceguard.
The next two itanium chips will go into the current system but dont expect an upgrade to the SX3000 chipset. All those folks will be working on the x86 glue chip.
The "enhancements" needed in xeon are for the non-stop not HP-UX. That will be obviously ported asap. VMS has obviously finally found its end date just like Tru-64.
Just like the old wendy's commercial said "wheres the beef" or the clinton campaign slogan "it's the economy stupid" this time "it's the applications" and HP-UX is the dog that don't hunt and wont hunt on x86 either. Oracle has obviously said sorry HP but even if you port HP-UX to Xeon we still won't support it.
Funny how Oracle basically replaced HP-UX support with z/OS support this year. After all the mainframe migration claims its the mainframe that ended up replacing HP-UX for Oracle.
Oracle wants to be the proprietary old IBM
HP finds itself being the nearly bankrupt IBM that people wanted to break up.
and, IBM finds themselves doing everything right with the highest market cap ever.
If you think about it the only real "engineered solutions" left are Mainframes and Power systems
This HP announcement is basically the same as Intel's move in April
Intel Relocates Itanium Engineers to Xeon Projects .
Mission-Critical Server Specialists Now Work on Intel Xeon
04/17/2011 "Nearly all the Itanium engineers, save a small development team working on Poulson and then rotating over to Kittson, have been redeployed on Xeon-related projects,"
maybe you should try refuting the facts vs. pointing people to an article from 2003 which was a misquote and corrected in the same article
Bowen, the executive in charge of AIX, emphasised that IBM's Unix isn't being replaced by Linux on any product plans. "We've got people now who are building chips for 2007 systems. If we had any belief that AIX was going to fall down and stumble, we wouldn't be doing that," Bowen said. In particular, a major revamp of AIX is due in early 2004. And though Bowen wouldn't provide specifics, he said the AIX development team is somewhere between two and four times as large as the 250 people IBM employs to improve Linux at its Linux Technology Center.
"HP has secretly contracted with Intel to keep churning out Itaniums so that HP can maintain the appearance that a dead microprocessor is alive," Oracle wrote. "The whole thing is a remake of Weekend at Bernie's."
Now the real question is when will Oracle stop support for SPARC64 now that they confirmed it is already dead.
back to the beach
IBM does upgrade their enterprise systems....p570 to p770....and P7 p795 was just a book replace in the p595.....but then again while the next two chips will be socket compatible for the current tukwila that also means the number of QPI links will never get in sync with the systems. 5 QPI's when you need 8 in the glueless 8 socket blade and only 3 of the 5 are used in the superdome2
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
