The word on the street today is that server maker Sun Microsystems - which is in the midst of being eaten by software giant Oracle for $5.6bn - has formally killed off the 16-core "Rock" UltraSparc-RK processor that has been in development for more than five years. The official response from Michelle Parkinson, the spokesperson …
end of an era
that's a shame; rock sounded quite exciting and i understand it was getting fairly close to production. certainly not off in vaporware land like ultrasparc V was. i think this announcement marks nothing less than the end of sun microsystems as an independent hardware designer. soon enough sun will be nothing more than a name plate on a generic intel box, but i suppose it's been going that way for a while.
i'll just keep on running my old ultrasparcs as long as i can get away with it.
incidentally, the ultrasparc II wasn't really a paragon of a successful hardware launch. yes sun was flush with cash and riding high back in the .com boom but certainly the early ultrasparc II processors had some hardware issues e.g. with caches that really drove a lot of customers crazy back then. the story is not all roses but it is true that the ultrasparc II did eventually turn into a very solid performer albeit somewhat uncompetitive in terms of raw performance versus its contemporaries.
really, in retrospect, i wouldnt say that sparc really ever had the edge in terms of pure performance. even the early sparc risc chips were not vastly more powerful than cisc contenders like the 68040 and early mips processors, and the ultrasparc was often bested in straight benchmarks by the alpha and power2. what always did it for me with sun is that they had a good all around package with excellent hardware reliability (imho; stop laughing!) and performance that while not eye popping was always acceptable.
Talk about waiting until the last minute
Next week Sun was to present a paper at the International Symposium on Computer Architecture:
"Simultaneous Speculative Threading: A Novel Pipeline Architecture Implemented in Sun's ROCK Processor"
Sorry Sunshiners Sparc Shocker!
I can just imagine all the Sunshiners out there wincing with each passing sentence of that article and dreading the inevitable Bryant beasting that is to follow. Bring it on Matt. I'm in my comfy chair and the pop corn is warm, crisp and ready to eat.
It's not too late
Very interesting article. Sun's Rock failed because it was not innovative enough, in my opinion. Anybody who thinks that the multithreading CPU technologies of the last century are viable in the age of massive parallelism should lay off the dope, in my opinion. This is a lesson for all the big players in the business and it will not be the last big chip failure either. Get ready to witness Intel's Larrabee and AMD's Fusion projects come crashing down like so many Hindenburgs.
As I wrote on Ashlee Vance's NYT blog (http://bits.blogs.nytimes.com/2009/06/15/sun-is-said-to-cancel-big-chip-project/?hp), after the industry has suffered enough (it’s all about money), it will suddenly dawn on everybody that it is time to force the baby boomers (the Turing Machine worshippers) to finally retire and boldly break away from 20th century’s failed computing models.
Sun Microsystems blew it but it’s not too late. Oracle should let bygones be bygones and immediately fund another big chip project, one designed to truly rock the industry this time around and ruffle as many feathers as possible. That is, if they know what’s good for them.
How to Solve the Parallel Programming Crisis:
Unfortunately, it makes sense...
Right now nothing software-wise can take advantage of the parallelism that Sun can bring to the table with Rock. It would mean a major investment in rewriting software for a single chip that maybe might have some gains on Intel and other architecture.
Right concept, bad timing.
very interesting articles, thank you for the links
I really think there is space for these processors in data warehousing environments run by oracle software, i wish you could add 4000 cheap nodes to Oracle RAC .... I wish
Expect SPARC64 to be the next casualty
Fujitsu's SPARC64 is going to quickly fade away also because they are simply just adding more low performance cores to a chip.
The latest SPARC64 VII is a perfect example of a poor design with lower performance cores.
Sun had the same problem with UltraSPARCIII to UltraSPARCIV.
Sun/Fujitsu M-Class systems suffer from low performance cores which provide an inefficient software value platform.
The SPARC64 VII quad core processors provide 26% less performance per core than the prior dual core versions.
M5000 16 core system with dual core chips: Sun M-Value = 345,000 with M-Value/Core = 21,563
M5000 16 core system with quad core chips: Sun M-Value = 255,000 with M-Value/Core = 15,938
SPARC64 VII now provides less performance per core than Sun's discontinued UltraSPRAC IV+ systems
E4900 1.95GHz 16 core system with dual core chips: Sun M-Value = 279,500 with M-Value/Core = 17,468
M5000 2.4GHz 16 core system with quad core chips: Sun M-Value = 255,000 with M-Value/Core = 15,938 8% less performance than US-IV+ per core.
Anyone running Oracle products should never buy a quad core SPARC64 chip.
Re: end of an era
"i think this announcement marks nothing less than the end of sun microsystems as an independent hardware designer."
You may be right here, however the Niagara line is very popular and they have a new chips scheduled pretty regularly through 2012. More chips that Intel has for Itanium, that's for sure. Niagara has always been on time as well. Much better history than Power or Itanium. My biggest concern is for Sun on the high-end. The M-Series is great, but I would like to see a high-end solution from Sun, not Fujitsu. Maybe Oracle will buy Fujitsu's SPARC group...
Killed != dead
I doubt very much that Sun would have shredded all the documentation and reformatted all the servers holding the data for Rock.
The IP up to this stage still exists and can be re-awakened if it makes business sense to do so.
Thanks to Marc Tremblay, Jeff Thomas, Aharon Ostrer, Shailender Chaudhry strong leadership of Rock.
Marc Tremblay just chalked up one more success after MAJC.
Congrats to MSFT for the great catch!
Re: Unfortunately, it makes sense...
"...Right concept, bad timing..."
The right concept is a concept that has the proof, at least in the magnitude (30 times) as it was promised in 2002.
The timing is really bad, it's too late.
Regatta On a Chip Killer
What a funny name. It's like a movie studio funding a project called "box office flop". A non-functioning chip is "rock"! Doh!
On time and on budget
That was the reason for failure: how could anybody think ANY project could survive to such a huge delay without consequences?
Let's face the fact no one would wait a year for a brand new architecture to come to the market: eventually one will go with what is already present.
Sigh... I guess I'll have to look at IBM or AMD for architectural cleverness. Besides, I can't cope with a suppliers' names being longer than 3 characters.
@Ian Gumby - your conclusion is wrong. There are plenty of workloads out there that fit very nicely on multi core ships.
1) Surely you've noticed the general enthusiasm for virtualisation at the moment? Alright, there's no VMWare for SPARC, but there certainly is Solaris Containers which does much the same thing only more efficiently.
2) Sun have long had the type of compiler support you need when developing code for multi-core machines (not just multi-core chips), long before anyone else. Intel have only just got round to implementing much the same kind of support in their compilers. AMD don't do compilers, but I guess one can make do with Intel's.
3) If sun are wrong on the multi-core thing then so are AMD, IBM, and Intel twice over.
4) Sun, like IBM with Power, are/were into developing processors targeted at specific markets. Hence things like on chip crypto engines, multiple ethernet controllers, and certainly in IBM's case all sorts of wierd stuff like hardware BCD FPUs. These types of features are very specific to particular market segments. For instance, an internet banking system benefits significantly from a hardware crypto accelerator, saves having to do all that maths in software, leading to a reduction in power consumption, etc. Intel don't do that.
AMD's support for that kind of market specialisation is interesting; having open sourced Hypertransport there's companies out there building bespoke modules that fit in a processor socket to allow for bespoke hardware processing with very close coupling to the x86 cores. So in a sense this is even more sophisticated a model than that of IBM's and Sun's because the end user is at liberty to decide what sort of processing they want hardware acceleration for. In a way one could argue that AMD's model will ultimately win over Sun's and IBM's; one can be part of the x86 herd, but one can also add specialist hardware things if one wants to.
Like they said, and end of an era
If Rock is killed then, like the article says, I also think that we'll see one or two more spins of the current chipsets, and then folks will be moved to something from Intel.
Which makes the Unix server market interesting - in one corner you have Intel, backed by HP and soon (?) Sun, and in the other IBM. Heck, HP support Solaris on their excellent ProLiant boxes (unlike the HP/sUX trash) so maybe they'll work even closer together. On the other side, doesn't this more or less gift the high end (32-way and above) market to Big Blue for Power6/6+/7?
@Macka: "Bring it on Matt" - hey, give him time to get in and get his orders from Mark Hurd! ;)
RE: Sorry Sunshiners Sparc Shocker!
Patience, grasshopper. You may have to put the popcorn on hold as it will take the Sunshiners a while to get over their shock, they're probably still in denial right now. But that popping sound isn't your Rice Crispies, it's Sunshiners' heads exploding.
Good morning, Sunshiners. And how are you all this morning? Please try and concentrate, now - the Grand Nova Master Ashlee Vance has said The Rock is dead. Now, which Sunshiner wants to put a starter in by claiming that Ash is just an hp salesman in disguise, trolling his own blog, and nothing but a Sun hater? Form an orderly queue - no pushing at the back! Wait your turn, Novatose.
Seriously, though, was there anyone out there that didn't realise Sun had lost the plot by 2001? They've been on the backfoot and going backwards ever since, and Rock was largely doomed by Sun's own actions before it even hit the drawing board. If Rock had arrived in 2003, when Sun still had some gravitas with the software houses, then it might have stood a fighting chance. Amusingly, it was the acceptance of Linux by those big software houses that put the first nail in Sun's coffin, and one of the most telling was Oracle's drive to make Linux a key OS. Larry is just getting Ponytail to finish what he started.
/Smug? Me? Never!
Oracle never intended to carry on the Sparc line ... there is nothing Oracle cannot achieve on a, intel or amd.
Plus it's better if Sun does the slaughtering rather than Oracle. If it were Oracle to do it, they may have some issues with the SEC & other anti-trust thingies.
"time to force the baby boomers (the Turing Machine worshippers) to finally retire and boldly break away from 20th century’s failed computing models."
In shorthand Internetese: "Learn2TheoreticalComputerScience". Nobody worships the TM and the TM is not implemented _anywhere_ (except maybe in Lego, as seen on Youtube).
Sadly, the referenced blogpost does not exactly propose a valid solution; 12-year olds have such ideas between supper and bedtime all the time.
As for Rock; well, these are sad times when soon-bought companies can say "NO" to promising chip projects. I was really looking forward to a transactional memory implementation. But what will Oracle do with all the research & development and, I suppose, the attendant patents? Will they go to Intel or to IBM?
We were the first group to vomit in the bar
Rock is dead. Long live Rock!
For Sale Rock T-shirt
For sale 1 'Solaris Rocks' T-shirt.
Orange in colour worn once (whilst decorating) a few paint marks otherwise in good condition.
Given to me as a freebie (remember those days when Sun gave out freebie's)
Paris - because she would look good wearing one.
Not all doom & gloom
Sun customers need to know that Sun has other CPU designs in its processor roadmap beyond tha Rock that are still actively being developed. Skipping Rock is not such a big deal, the successors are way cooler (in features, dunno about die temp), as you would expect...
Dont tune out yet, Sun is still in the scene...
sun just scheduled rock demos in hamburg
this is weird - sun just scheduled a presentation of rock in hamburg, supposedly live hardware will be demoed:
also there is the presentation at isca next week:
"Simultaneous Speculative Threading: A Novel Pipeline Architecture Implemented in Sun's ROCK Processor"
I doubt that the last word is spoken on this.
RE: sun just scheduled rock demos in hamburg
This is just the news not having yet been internally announced to all and Sundry, and what's left of the Sun marketing team getting on with what they have been told to do. When they canned UltraSPARC V, we had a Sun Veep (not naming names) come round only five days before the announcement to tell us US V was taped out, the silicon was ready, etc, etc.
Dropped like a rock
Bad bad execution. Horrible, horrible leadership. EVPs, VPs, Directors, ... Sure fire way to sink any project.
re: On time and on budget
Tukwila anybody? Later than ROCK and twice as expensive to produce.
re: Dropped like a rock
Can't agree anymore. Good riddance marc trembly.
Impossible to avoid an Oracle takeover...
Timothy Prickett Morgan writes, "Sun has two more weeks before its fiscal year ends and another four weeks before Sun shareholders vote on the Oracle takeover"
It sounds like management in Sun are making it impossible for the shareholders to vote against the Oracle acquisition.
Remove one of the legs of the stool, which was planned to take their midrange hardware business into profitability, there is no plan except acquisition.
Re: RE: sun just scheduled rock demos in hamburg
Good on ya' Matty. You kept your comment down to a relatively small paragraph. Though, you could have cut out half of it and still got your point across, but I think this is progress. Of course, your other frothing rant was still a bit long. You're getting there little Matty. Remember, short is just frothing, while long is a frothing rant. Keep trying...
RE: re: On time and on budget
"Tukwila anybody? Later than ROCK and twice as expensive to produce." <Yaaaaaaaawwn> Look, I know you Sunshiners are gonna try and live out the denial by squealing on about Tukwila, but the truth is Rock is dead - you can't get any later! And as for the idea of Tukwila being twice as expensive to produce, how do you know? Do you have production costs for either? Well, actually do you have cost projections for production costs for Tukwila, seeing as it's not in production yet, and as the only costs now for Rock are disposing of all remainder of the Rock team.
Instead of wasting your time on more Sunshining, I suggest you get on over to https://www.redhat.com/training/ and start anew.
they had a box last year. but only single chip.
couldnt make it run properly with two. or more.
come the next tape-out, the credit crunch, more fumbling.
then the guys start talking down a little the to be expected performance compared to competitors.......
very curious to see what money-machine oracle is going to think about more spending on sparc while there's little hope for solid profit.....
btw, 3rd march they put solaris INTEL on the "obsolete platform" list.
No oracle 11g coming for that, is what it says.
Will be funny to see the sudden change of mind, say next month... ;-)
re: Drooling Matty
"And as for the idea of Tukwila being twice as expensive to produce, how do you know?"
Because it is based on the most complex chip design of all time, EPIC. Even Intel's architects comment on it (looking for the article... I'll find it as soon as I select Post). Nothing new in the Itanium camp for how long? All they do is add more cores and more cache, and slowly at that. Nothing else. Niagara and SPARC64 seem to have regular updates, and on time, unlike Power and Itanium. RISC is simple and easy to update while EPIC is complex and expensive to design. Complexity translates directly to cost. Ask anyone in the microelectronics field Matty. You seem to have a lot of friends at HP... wait, that's right HP does not have any chip designers.
@AC re: Drooling Matty
"Nothing else. Niagara and SPARC64 seem to have regular updates, and on time, unlike Power and Itanium."
Regular updates unlike POWER? Gimme a break...
POWER5, 5+, 6, 6+ what was that about? IBM released POWER4 dual core 1.1 and 1.3 GHz back in 2001 and current line is 5.0 GHz on POWER6 is that not regular update for you? You serious???
ROCK (Regatta on a chip killer) was targeted at POWER4 Regatta machines that was like 8 years ago... come on... wake up dude...
And regarding Itanium yes, they upgraded from 1.1 to 1.66 GHz in the last few years, and with Tukwila delayed again, they don't have regular updates lately
Watch the industry more carefully before posting...
Paris: 'cause she also does like the POWER
RE: re: Drooling Matty
"....All they do is add more cores and more cache, and slowly at that. Nothing else....." Lol, so you missed all that stuff about DDR3 and QPI then? And so funny that a design that you say has additions so slowly and so lightly has been trouncing any SPARC design for years!
"....Complexity translates directly to cost....." Ah, so what you're really saying is you haven't got a clue how much Tukwila will cost, you're just swinging wildly in your grief. There, there, go wipe your eyes and blow your nose and try not to be such a whiner in future.
Re: RE: re: Drooling Matty
"... stuff like DDR3 and QPI..."
Ha ha ha!!! Nice Matt. You're right, those are great advances... LOL. and when will they be released? You don't know, do you? There have been regular Niagara and SPARC64 at least every 6 months. When was the last Itanium update????
Montecito came out in 2007 (2 years late) and just added dual cores... Montvale came out in 2007 and only added more cache. Two years and still no new chip! Where's Tukwila? Intel has been promising Tukwila since 2003, of course they called it Tanglewood then. Intel first said 2007, then they said 2008, then 2009, and now 2010!!!
Don't believe me...
Heck, with Intel's 3 year old chips both Sun and IBM can beat the snot out of Itanium.
Sure there are a couple of cherry picked benchmarks that HP does well on, but when
you actually run your application on the systems you see a massive performance advantage
for both IBM and Sun. Itanium is behind and if they don't start releasing chips more often than
every couple of years, then they will fall further behind. No frequency advances, no core advances, no thread advances, and their bus is sorely behind... QPI better be as good as they say, 'cuz they're hurting if not.
I was looking forward to ROCK, but SPARC64 screams at the high-end and Niagara is quickly moving up the performance/scalability ramp. You can actually see the advances of Niagara on a regular basis, while Itanium is a snail in comparison. Is Intel actually investing in Itanium anymore?
Re: @AC re: Drooling Matty
I agree. Power is updated much more often that Itanium, but Power7 is very late. They promised Power7, but give us Power6+, but where is it? Oh yeah, it came out, but no once cared.
"Watch the industry more carefully before posting..."
I know that you love your Power, but you are blinded by your love. Power6 was promised in 2006, and then moved to 2007. Power7 was promised in 2008, but then Power7 was removed from their roadmap and Power6+ showed up. Then Power7 was there, but no date (To Come). Then Power6 showed up in 2007(more than a year late). Power7 was then pushed out to 2010 (2 years late). Of course Power6 was only put on some low end systems and not pushed to the high end until 2008, more than 2 years later than promised. Power7 was removed from roadmaps again and Power6+ was promised instead.
Forgive me if I can't keep up with IBM's Roadmaps. Apparently, not even IBM can.
Of course, IBM was smart to keep their core count down and their megahertz up considering the licensing issue, but they've apparently hit the wall with this strategy. Where is their new chip?
Also, though better than HP/UX, AIX is no where near what Solaris is, let alone Linux. No wonder people want to run Linux on Power.
About IBM Power roadmap
I agree. IBM had some delays on Power roadmap.
However, I was searching for detailed roadmap information in El Reg and went through this:
A 2003 IBM roadmap, and it seems IBM did a great job on their processor launches so far.
Never saw IBM cancelling a processor project like the ones at Sun and HP (Compaq).
In terms of roadmap, I believe IBM is the only player that can really say when and what to be released.
@Bazza... not quite...
The same problem exists in Sparc VII+ chips in their m900 series (or is it m9000?)....
If you follow comp.databases.informix, there was an issue concerning how to tune the database engine in the environment because while you have 64 cores, there are two threads to a core.
The bottom line was that Informix (IBM's IDS engine) is probably the best engine to take advantage of the virtual cores, yet it can't. If you go to rock you have the same problem.
So your current off the shelf packages like Oracle, IDS, DB2, Sybase, etc ... can't take advantage of the extreme parallelism that Rock can offer.
Yes you are correct that you can have special applications developed that would take advantage, yet how many customers are there that have the budget and the staff to write these apps? (Not enough to make the chip viable)
Many articles, including those in El Reg have talked about having to upgrade the software in the OS and in applications to take advantage of parallelism offered not only by Sun, but also Intel's new chips. CRN (Computer Reseller News) had an article that talked about creating 15+ virtual microsoft servers on a 2 socket 8 core xeon 5500 box. Think about that. Instead of being able to have the software scale up, you have to instead scale down the box into multiple virtual servers to get the most out of the hardware.
Take to the example of PC users desktops where they use CPU intensive software like photoshop. Even here the lack of parallelism shows that the hardware advances go under utilized.
And that was the point. Sun's architecture is brilliant, but unfortunately no one has written applications that have mass appeal to take advantage of it....
Again, good idea, bad timing.
A Rock and a Hard Place
Shame on all of you who are talking about information that is disclosed under NDA/CDA. These are legally binding on employees and customers. Any of you old enough to remember
"Loose lips sink chips"?
I wouldn't trust you with my mother's maiden name, and yes, unlike you loose lipped idiots, I know my mother.
AC on POWER roadmaps.
You should really try to get your so called facts from other sites than BMSEER.
POWER6+ was 'promised' for 2008, and you know what they just made it. If you check out the power 560, you will see that it packs a power6+ processor.
Now we can then agree that power6+ is a bit of a disappointment. But hey it's not like SUN or HP have anything that can match it.
Well there is Niagara, but hey it needs 64 slow moving threads to do it, which is a big enough problem in it's own right.
RE: Re: RE: re: Drooling Matty
"Ha ha ha!!! Nice Matt. You're right, those are great advances... LOL. and when will they be released? ...." They have been released already in Nehalem. One of Intel's big advantages has been the ability to bring advances from the Itanium line into Xeon and mass-produce them, which makes the cost of introdcuing them in Itanium much lower. When Tukizilla hits those Sun refugee accounts next year it will come with proven QPI and DDR3 tech, and run current Itanium binaries - no risk! Even if Sun had managed to get Rock out the door, it would have been "bleeding-edge" and unproven, with no prior successful implementations of scout threads or transactional memory to bolster confidences, and also with precious few applications even able to take advantage of the design even with a recompile.
".....QPI better be as good as they say, 'cuz they're hurting if not....." Aaaaaaaaannnnnd once again, time to remind you which combo is king in the enterprise high end, the most demanding but most margin-rich arena - yes, hp-ux on Integrity! In fact, Integrity's sales figures have been consistently upward for years. And the reason is because the whole package is better - better performance, better stability, better features, better management, better compatibility, better range, better integration with other hp products like storage, and better support and services. Oh, and also because us customers don't think hp is about to tank in the same way Sun already has.
But to get back to Rock, there is still hope for you Sunshiners. After all, this is still unconfirmed by Sun. Not that it makes much difference - customers will believe the rumours because it is what they expect, which does make it strange that Ponytail hasn't immediately been on the phone to deny it....
RE: Sorry Sunshiners Sparc Shocker!
"Pateince, grasshopper!!" Jeez Matt, thats an insight and a half into your psyche though I suppose it's expected you'd turn up and gloat given the ribbings you've taken. I've no problem saying I'm a bit gutted that Rock is speculated as dead but lets be clear, none of us know whats planned until it comes from Oracle directly these days.
I'm more gutted it give you space to be smug! :-)
If they actually have dropped it I'm guessing they'll morph the technology into the T2 / KT lines and produce a hybrid without all the faffing around in transactional memory. I'd be reasonably certain a chunk of the pre-fetch/scout alogorythyms they've mentioned could get carried across and so forth. Perhaps Sun engineering bit off more than can be chewed in current development costs with transactional memory stuff, at least the T2 lines use relatively standard RAM.
So IF it's dead here's hoping Oracle bolts in as much of the good stuff that they can do with a bucket load of DDR 3 and bang out the chips with a cheaper footprint.
You do set yourself up for some falls though Matt so remember the pebble will be snatched if Itanic sinks or this is just rumour, I'd place money that people will spoon feed you humble pie!
RE: RE: Sorry Sunshiners Sparc Shocker!
"....Jeez Matt, thats an insight and a half into your psyche...." Yes, it's called humour, I understand that Sunshiners just don't get it seeing as how they were the only poeple not laughing when UltraSPARC IV was announced. Trying to read anything else into it is just desperation on your part to find anything to divert from the topic of the article. The rest of your post is just wishful thinking.
In fact, the rather half-hearted response and the complete lack of ranting from the usually rabid Sunshiners indicates you guys have actually been secretly expecting this just as much as the rest of us, despite your continual denials.
RE: RE: Sorry Sunshiners Sparc Shocker!
Matt, in your infinite wisdom would you also remind us that while HP Market droids were laughing at Sun's move from Sparc-iii to Sparc-iV (2004) it meant a performance bump Itanium fans were dreaming of? The move from Sparc-iV to Sparc iV+ (2005) was another good bump as well (bump = leap in Itanium terms)
Sparc-iV to iV+ was a modest frequency increase (10% on first Sparc-iV+) but a fairly big engineering change (pipelining, cache, CPU features) and meant something like a 40% perf increase with power down to 90w/socket. Again, something dreamt of by Itanium fan's like yourself back then!
I'm sure the HP sales grunts were generating hollow laughs at figures like those especially when the Itanium was something like 1.6ghz, single core (or dual core if you count two seperate CPU's bolted on one socket at 260watts/socket!!!) http://en.wikipedia.org/wiki/Itanium#Processors
I think back then Sunnies weren't laughing as they were shaking their heads at Itanic slowly melting the icecaps......
Anyway, shall we both grow up a bit & just wait and see if this is a game to force a statement from Oracle? Or shall we chew & speculate more on whether there is any meat on this bone?
Re: RE: Re: RE: re: Drooling Matty
From MB: "They have been released already in Nehalem."
Regarding QPI and DDR3 in Itanium not being out yet as Itanium is (again) late.
Also, from MB: "One of Intel's big advantages has been the ability to bring advances from the Itanium line into Xeon..."
So which way is it? The GREAT advances in Itanium go down to Xeon or the more advanced Xeon advances going into Itanium. Have you gotten it yet Matt? Intel is not making money from Itanium due to extremely low volumes. Volume matters Matt. Even Sun sells twice as many SPARC CPU's as Intel sells Itanium. Intel can't make money at this rate. Intel is slowly moving Xeon features into Itainum so that they can kill Itanium all together. Will HP give Intel another 10Billion to keep Itanium afloat? I doubt it.
Solaris on Xeon anyone? I know HP loves it (obviously EDS does).
RE: RE: RE: Sorry Sunshiners Sparc Shocker!
Those of us that worked with UltraSPANKed will remember that US2 and US3 performances were generally so dire that even the "bump" of the US4 release still left Sun SPARC trailing existing Power, PA-RISC and Itanium. Just 'cos you managed to get SPARC of it's belly and crawling on it's knees didn't mean it was anywhere close enough to catch the competition. It was all the more amusing that Fujitsu's own SPARC64, developed on less money than UltraSPANKed 4, managed to be typically 20% faster for less money (I know 'cos I brought in FSC kit to replace some of our Sun Slowaris servers). You may want to recall that US4 did nothing to halt the dive in Sun sales and the continued lack of profits in the post-dotbomb period, whilst hp and IBM soldiered on with profitable businesses.
But, as you say, let's see if this gets a statement from Sun (Oracle can't comment as they don't own the bizz yet). We can also wait and see if Oracle deny the reports that they are still trying to sell on the unwanted Sun hardare bizz.....
HP and Intel are killing us all
Now I'm hearing that Intel will not RELEASE Tukzilla until Q1 2010. HP will then only get out new systems 2-5 months later!!! That's no Tukzilla kit until late 2010!!! HELP!!!
Re: RE: RE: RE: Sorry Sunshiners Sparc Shocker!
Ha Ha... MB has to go back 6 years to find a point when Sun's chips were slower than Itanium... and that wasn't for very long. Nice. Itanium has been behind since the beginning and is not getting any better. That's why HP is the only user. Even without ROCK, Niagara and SPARC64 destroy Itanium now, and the roadmap for Niagara and SPARC64 leaves Itanium in the dust. Faster bus, higher speed, more cores, more threads, more RAM. These are facts, not just opinions, such as your claim of better support and maintenance tools. Enough with the marketing speak Matt. We all know that you work for a HP Partner, you don't have to keep reminding us.
"...Fujitsu's own SPARC64, developed on less money than UltraSPANKed 4..."
Oh, and you know this because - or are you just blowing Itaniums out your butt?
RE: Re: RE: Re: RE: re: Drooling Matty
".....Regarding QPI and DDR3 in Itanium not being out yet as Itanium is (again) late...." You have obviously forgotten that the article is about Rock being so late it has been cancelled. Tukzilla hasn't been cancelled. I'd try spelling it out for you but I might have to use words with more than one constinant, which seem to fox any chance of you actually grasping the facts.
"....So which way is it?...." QPI was developed as CSI in the Itanium stream and then also moved to the Xeon stream. DDR3 was originally designed into the Xeon stream, then added to the Itanium stream. Those of us with a clue have read and understood that from the Reg articles on Tukzilla and Nehalem. I assume you were waiting for the crayon edition with all the big words taken out?
"....Intel is not making money from Itanium due to extremely low volumes...." Wrong again. Intel reports it is making a profit on Itanium, and seeing as it is ramping up whilst all SPARC are ramping down, it looks like Itanium will go on making more money than SPARC. Well, more money than SPARC64 as it now looks doubtful any other server SPARC will exist in a year.
"....Even Sun sells twice as many SPARC CPU's ...." Care to post some facts to back up that claim? Even including the Mickey Mouse Niagara CPU and all of the Fujitsu sales it would not reach that "twice" claim. What you are also forgetting is that hp sell the ProLiant range against Niagara with great success, so what you should be looking at is all server sales. But you won't because it really pains you lot to have to admit hp is the number one server vendor. You also forget that hp makes more margin per server than Sun according to IDC, especially in the margin rich enterprise high end where Sun has virtually ceased to exist. In fact, hp makes more money in a quarter on printers and ink than Sun make losses from their whole server bizz in a year. The fact is, were it otherwise, it would be hp being bought up for chump change rather than Sun.
"....Intel is slowly moving Xeon features into Itainum so that they can kill Itanium all together....." Yeah, I heard that bit of Sun FUD too. I think it was before the UltraSPANKed V got canned. Nice to see you Sunshiners are staying up to date - not! Not much hope for an updated FUDset for the Sunshiners now that the Sun has set. Maybe soon they'll realise Intel (and hp) use Itanium and Xeon as a two-prong strategy to attack the RISC market - Xeon from below and Itanium from above. The success of the strategy can be gauged by a simple truth you'd no doubt like to ignore - the Sun server bizz has gone down the tubes with the cancellation of Rock being the final surrender flag. Even Sun's desperately late and failed attempts to jump on the x64 bandwagon haven't saved them due to the vanity of their SPARC pretensions. Had Sun got serious with Slowaris x86 as a real server offering back in 2000 then the story might be different, they might have held off Linux and Windows, but the SPARC-Slowaris crowd couldn't stand that, and they ran the company into the ground in their stupidity.
So let's try and keep it simple. I'll give you four simple questions, you just think long and hard about the answers and then see how silly you feel given your earlier posts:
1) Which company, Sun or hp, has made a profit in the last six years?
2) Which company has the larger range of products, Sun or hp?
3) Which company went from being a $200bn company in 2000 to having a market cap of less than $4bn this year, Sun or hp?
4) Which company has just been bought by a software company, Sun or hp?
RE: HP and Intel are killing us all
I see the Sunshiners are so desperate to avoid discussing the demise of Rock they've even wheeled out the fake Matt Bryant login. Maybe they hope they can wish away their pink slips simply by keeping their heads in the sand.