Sun Microsystems has launched a new business unit to sell its Sparc processors, a return to an idea it had dropped years ago. David Yen, currently executive vice president of storage but previously in charge of Sun’s Sparc work, will lead the new group and retain his executive VP status, the company said Thursday.
…but can I put one in my Dell Linux laptop? ๐
go and suggest it to Dell
“””
go and suggest it to Dell
“””
They wouldn’t go for it. The Sun chip community is too fragmented. If they used Sparc IV, the Niagara users would be offended. If they used Niagara, the Spark IV folks would complain.
So the best thing to do is not to offer anything at all.
Sun would need to consolidate their chip offerings like Intel has before Dell would consider preinstalling them.
ROTFL. Nice one ๐
Sure i’m not going to be picking up a sparc at my local pc hardware store but it should be good to see what other vendors can do with them.
nice to see that sun is living into the whole innovation happens everywhere thing.
Time to move on.
The economies of scale have made it impossible for Sun to be able to compete with Intel in the processor business. Intel has the entire market range from embedded to massive covered, and they’ve got better competitors than Sun against them in every market.
If AMD can give Intel a run for their money, why not Sun?
Edited 2007-03-28 01:42
Because AMD is competing head-to-head in a clone market. Customers can leverage all of the x86 based expertise towards an AMD system.
Sun is competing with a product without the same level of third party support.
Besides, it’s in Intel’s (anti-trust) interest to leave AMD alive. No reason to do that with Sun.
“””
Besides, it’s in Intel’s (anti-trust) interest to leave AMD alive. No reason to do that with Sun.
“””
Indeed. They might ramp up their marketing of Itanic II and crush Sun. ๐
Edited 2007-03-28 01:58
SPARC’s dead
Please enlighten me as to why Sparc is dead?
It costs somewhere between 300 and 500 million dollars to do the R&D for a new processor, if you want to stay competitive with Intel.
Intel can afford this because they tend to sell 100s of millions of each new chip so the cost the R&D adds to the part is 3 or 4 dollars.
Sun can’t afford to do it because even if they’re doing 1/10 of Intel’s market share, they have to add 30 or 40 dollars to the price of a chip.
The exact numbers were different 10 years ago when HP got out of the processor business and sold the wide word architecture to Intel as a basis for Itanic, but at that time HP was the #2 computer maker in the world and couldn’t make money competing against Intel.
Sun’s a much smaller company and has an even worse position with respect to competing with Intel. They’re about where Intergraph was in ’89, except that unlike Intergraph, they don’t have the option of dropping their own processor and going to using Intel, er, wait, they’ve already started doing that. . .
Sun’s dead Jim.
They’re doing awfully well for something dead. It isn’t nice to spread FUD.
typo. it was supposed to say “SPARC’s dead” again.
Anyway, Sun is dying. It takes a company that got as large as they did a long time, but anyone whose old enough to remember CDC will recognize the signs.
“””
Anyway, Sun is dying. It takes a company that got as large as they did a long time, but anyone whose old enough to remember CDC will recognize the signs.
“””
It depends on the leadership and the employees. After years of progressive atherosclerosis in the management department, Sun is showing signs of a reversal, and is beginning to think and act like a much younger, more vibrant company. They are willing to take risks like OSSing Java and Solaris. Those are the kinds of risks that can either kill your company or reinvent it. They’ve shown some courage and a willingness to “think young”.
Sun may be dying. But if they do, I suspect that it will not be without a valiant and admirable fight. Should be a good show, either way.
Edited 2007-03-28 13:58
They’re doing awfully well for something dead. It isn’t nice to spread FUD.
If it’s FUD then you’re going to have to read the comment above and say why to each bit.
Edited 2007-03-28 09:25
Sun can’t afford to do it because even if they’re doing 1/10 of Intel’s market share, they have to add 30 or 40 dollars to the price of a chip.
A V890 on the Sun store with 16 cores and 128 GB is over $200K. Do you really think that adding even $1000 to each chip makes a significant difference here?
The thing you do not seem to get is that Sun is not competing with Intel. They are competing with HP and IBM , whose non-x86 prices are similar and quite likely worse.
A HP Integrity 16 core / 128 GB box with top-end Itaniums is easily over $200K, but the dual core Itaniums themselves list (from Intel) at maybe $4K each. HP, of course, charges a lot more than that for each CPU (so much for the benefits of an “industry standard CPU” – last I checked, the price vs PA-RISC was about the same) , but the point is, what does Intel get here, 10, maybe 15 percent of the sale?
Its the system sale that is important.
Besides, if CPU speed was what kept a server company afloat, Alpha would still be here, and Sun would have died 10 years ago.
No. But we’re talking about Sun reentering the chip business in this thread.
Do not kid yourself. If they go back into the chip business, they are competing with Intel.
Not in the chip business.
“No. But we’re talking about Sun reentering the chip business in this thread”
It seems there are two topics you are covering in this thread. When you say “Sun is dying” that is dependant upon all of the products and services that the company offers, far more than just Sun reselling SPARC chips. I think the V890 example is valid counter-argument to the “Sun is dying” statement.
Your arguments about why Sun will fail in this new venture is more compelling. But going from “Sun is trying to enter a very competitive market and will most likely fail” to “Sun is going to disappear” is absurd without more information such as “their revenue is dropping” or “their market share could dip below some critical point” or “the leaking sewer gas in the building will addle the executives’ brains.” Maybe a “this venture will be an expensive failure that will divert resources from ventures that could save Sun” is a better statement to make?
I like Solaris and don’t want Sun to fail in general, but will they succeed in the chip seller business? I have no idea.
Agreed. But it’s tough to sum up 22 years of Sun watching into a single paragraph. Sun is pretty much the last of the “Unix + our processor” houses, other than IBM, and IBM is a special case. Sun is in an industry where very few companies make the transition from a long established business to a very different business model. Sun is, despite Java, mainly a systems house to this day. Sun going back into the chip business tells me that they’re staying true to their roots which at this late date in the industry is suicidal.
A small company in a cutthroat business needs to focus on one thing and do it well, preferably in a niche that’s not interesting to the competitors. Sun on the other hand, is defocusing and going in many directions at once.
No single one of these factors are enough to kill a company, and a miracles do happen (CF Jobs returning to Apple) but the horse to bet on in this rate is “CDC-style death of a thousand missteps.”
And without a chip business its hard to offer differentiated products any more and yopu margins get whacked. How much is a 16 core opteron box once barcelona is out of the door? how much faster than that sparc box?
os choice – check
ras – check
cost – check
vendor lockin – removed
performance – check
IO – check
-xxmf
The thing you do not seem to get is that Sun is not competing with Intel. They are competing with HP and IBM , whose non-x86 prices are similar and quite likely worse.
That’s what they all say, just as many companies (like Sun) say they don’t compete with Microsoft. HP got rid of their own specialist chips long ago, and IBM has the resources, R and D budget and specific markets (mainframes primarily) to keep Power relevant. Sun has nothing like that to keep SPARC relevant to anyone.
Its the system sale that is important.
Which makes SPARC even less relevant. In that context, it is the chip that costs the least and performs the best that wins out. However, we’re talking about the chip business here, and SPARC is just running out of ways to keep itself relevant.
Yep, HP killed of PA-RISC and Alpha in favor of Itanium, interestignly enough the sales of VMS and HP-UX kit started to slide about the same time and haven’t stopped…
SGI killed off MIPS/Irix in favor of Itanium/Linux and have been deliseted from whichever stock exchange they were on and gone into bankrupcy protection from which they have since emerged and are now adding Xeons to boost their low end boxes where the Itanium sucks.
IBM and Dell have abandoned the Itanium. That just leaves a couple of offerings from Unisys, Bull and Fujitsu, all of which have other server lines, usually based on x86-64 from at least one manufacturer.
The only companies showing any decent growth in the high end unix/mainframe market are IBM and Sun, both of which have their own proriatory chips (POWER and SPARC) to run their own Unix on (among other things).
IBM is pushing POWER and PowerPC as well as Cell into the high end embeded space through Power.org having sold off their low end PowerPC 400 and 440 chips to AMCC. This is exactly where Sun is now going with the UltraSPARC T1 and UltraSPARC IV. The Niagara II will be even more suitable for this market with more threads per core, a dedicated FPU per core (instead of a single shared FPU), onboard 10Gbps Ehternet and chriptographic acceleration which will allow it to replace embedded systems where you have previously needed several costly ASICs.
Yep, HP killed of PA-RISC and Alpha in favor of Itanium, interestignly enough the sales of VMS and HP-UX kit started to slide about the same time and haven’t stopped…SGI killed off MIPS/Irix in favor of Itanium/Linux and have been deliseted from whichever stock exchange they were on and gone into bankrupcy protection from which they have since emerged and are now adding Xeons to boost their low end boxes where the Itanium sucks.
Which only serves to highlight the problems Sun has in front of it.
No, it shows that HP, SGI et al are having a hard time transitioning to Itanium, either because their stratagys for transitiong are flawed (but then why is it affecting all companies that have made Itanium based products?) or because Intel’s execution of the Itanium has been weak to say the least.
If Sun were facing the same problems as these companies with their high end SPARC kit then they wouldn’t be showing huge growth in the SPARC business such as the T1 powered T1000 and T2000 servers that are clawing back some of the low-mid end from the x86[-64] machines that they lost it to and the high end UltraSPARC IV machines which even Sun have said that they are suprised by the voluem of orders and can’t meet their demand.
That’s what they all say, just as many companies (like Sun) say they don’t compete with Microsoft.
Question, who is “they all”, IBM and Sun? Last I checked, Sun (and IBM) most definitely do compete with Microsoft and that is no secret.
HP got rid of their own specialist chips long ago
Sort of. But the funny thing is, between the announcement and the actual discontinuation of PA-RISC, there is something like 10 years. The last PA-RISC design was released relatively recently and is still buyable. I think the EOL is 2009. If in system sales it is the chip that costs the least and performs the best that wins, why is this?
IBM has the resources, R and D budget and specific markets (mainframes primarily) to keep Power relevant
IBM mainframe is definitely not POWER based. It is a completely separate CISC line using (interestingly, relatively slow and very expensive) custom chips. There may be some POWER coprocessors (for IO, invisible to userland code) in a zSeries, but IBM doesn’t really say. POWER5 is only used in pSeries (AIX) and iSeries (AS/400). I would be willing to bet that they ship less cores than Sun in both lines combined. There is essentially no difference in the market for POWER5 and SPARC.
In that context, it is the chip that costs the least and performs the best that wins out.
Regarding system sales, IBM mainframe is perhaps the best example that this has little to do with it.
However, we’re talking about the chip business here, and SPARC is just running out of ways to keep itself relevant.
To the extent that all non-x86, non-embedded chips are, maybe. But I see no competitor that both doesnt have a chip line in the same boat (Itanium, PA-RISC, POWER) and is capable of building 8-socket and up boxes. There is still money there (or SPARC would already be dead).
Question, who is “they all”, IBM and Sun? Last I checked, Sun (and IBM) most definitely do compete with Microsoft and that is no secret.
Sun, IBM and Red Hat all try and pretend that they don’t compete with Microsoft in the areas they’re in. However, the niche markets are becoming ever more congested as they do that.
There is essentially no difference in the market for POWER5 and SPARC.
Well, it certainly doesn’t look like it. IBM have managed to diversify Power into many meaningful things, such as the Cell, but Sun simply haven’t don anything with the SPARC that recoups any of their investment.
There is still money there (or SPARC would already be dead).
It’s not dead simply because there are still some very old SPARC systems running in the world. However, there’s certainly no new business in it and what is left is gradually dwindling.
Sun’s a much smaller company and has an even worse position with respect to competing with Intel. They’re about where Intergraph was in ’89, except that unlike Intergraph, they don’t have the option of dropping their own processor and going to using Intel, er, wait, they’ve already started doing that. . .
I don’t see how the two compare. Intergraph core business was in the workstation market, not servers and they never designed their own processor line completely in-house. The Clipper processors they used at the time were developed by Fairchild and then the line was bought by Intergraph. The only 100% in-house design they had (the C5) was ultimately axed. On top of that I don’t see why Sun couldn’t transition to Intel processors. Their newly reborn x86 business is doing very well and suggests the contrary. Their success there was also quite surprising as they entered that very crowded market very late.
The C5 wasn’t the only completely in-house design. The C300 was designed in-house after the C5 cancelled and was Intergraph’s last Clipper based product.
The similarity was that both companies were in the processor design business with far too small budgets and far too small of a chip run to justify the R&D. The C300 was done on 10% of the budget of a typical Intel turn and the corner cutting that required showed up in the product being not competitive.
This is why it would be a mistake to claim that Sun is “dead”. They’re dying, but something like that might save them. On the other hand, it is a very crowded market and it requires that sun focus on being an integrator rather than a full system developer.
Having watched CDC try to reinvent itself and fail I have significant doubts about Sun.
The similarity was that both companies were in the processor design business with far too small budgets and far too small of a chip run to justify the R&D. The C300 was done on 10% of the budget of a typical Intel turn and the corner cutting that required showed up in the product being not competitive.
Sun is not in the processor design business, they don’t sell processors, they sell systems, services and lately even software. Processor design is only part of their operations and it the last few years they have differentiated and are now in a much better situation than in the past where they were entirely reliant on their offerings. That said this announcement shows that they are willing to sell their in-house processors to external customers not that they are betting the company on processor design alone. On the contraray there is no mention that they will increase their focus on processor design, lately they have been working on designs which have inherently a lower development cost than more traditional processors and they probably hope to be able to sell them to third parties instead of using them exclusively.
Sun is not in the processor design business, they don’t sell processors, they sell systems, services and lately even software.
Which makes going back into the chip business right now a suicidal thing to do. They’re still an old world IT company that is still trying to find its place in the world and work out why things aren’t the way they used to be.
Which makes going back into the chip business right now a suicidal thing to do. They’re still an old world IT company that is still trying to find its place in the world and work out why things aren’t the way they used to be.
Read the announcement, they do not intend to go back in the chip business, they don’t want to spend R&D money for it, they don’t want to go head-on against Intel. They want to try selling the processors they already develop for their own purposes to third parties. They have already invested money in processor R&D for their internal needs, if they can make some extra ROI on that money then fine, if they can’t it won’t change their current situation anyway.
Edited 2007-03-29 08:29
not spending R&D is not the same as not being in the chip business.
It’s a defocusing move into a market they’ve done poorly at. You can’t be in the chip business as a one off. Even if you eat the entire R&D cost and subsidize the chip business out of a profitable internal division, you still have to spend money to be in the business. External support requires different documentation than internal support, different tools have to be developed and made available, and there’s a marketing expense.
Sun expects to sell these chips into the embedded market, and that’s a market with a high cost of marketing and low margins.
The economies of scale have made it impossible for Sun to be able to compete with Intel in the processor business.
Sun has no intention to compete in the general-purpose processor market Intel is in. Their offerings are geared mostly around the Niagara line which are well suited for network gear. That’s the high-end of the embedded market where the most prominent players are Freescale, Marvell, etc…
Intel has the entire market range from embedded to massive covered, and they’ve got better competitors than Sun against them in every market.
Not true. In the last few years Intel has done everything it could to focalize on their core business, that involved also exiting completely the embedded market:
http://www.techspot.com/news/22036-intel-sells-xscale-to-marvell-fo…
Intel never enjoyed much success there anyway mostly because the embedded market requires a degree of per-customer customization which is completely opposed to Intel philosophy of large scale production.
That said Sun has profundly changed the way they deal with chip design which is leading to huge savings, Niagara is a prime example for that. Their move to massive multi-threaded design and de-emphasis on clock speed turned made their new designs (including Rock) much simpler than the previous ones. Using replicated elements and fully synthesized logic (apart from memories) has cut significantly verification costs as well as time required to do it.
To cut mask costs (which are becoming huge) during the design process they also started verifying single components indipendently reducing the number of respins needed to get a working chip.
This doesn’t mean that Sun will take over Intel, or any other player in the market, nor that they will be successful but they are certainly in a better shape compared with years of overambitious designs getting canned one after the other.
Finally this move was probably also prompted by their need to move to other fabs in the future for their processors as TI (quietly) announced that they would stop developing logic processes after the 45nm node.
[p]Not true. In the last few years Intel has done everything it could to focalize on their core business, that involved also exiting completely the embedded market:[/p]
Do not assume that Intel sold Xscale to Marvell as a way of exiting the embedded market place. It was not that. It was a way of of exiting the ARM business so that they could concentrate their embedded strategy on their own low power chips, as they’ve recently announced.
Do not assume that Intel sold Xscale to Marvell as a way of exiting the embedded market place. It was not that. It was a way of of exiting the ARM business so that they could concentrate their embedded strategy on their own low power chips, as they’ve recently announced.
As it is now Intel went from a minor presence in the embedded market to being not present at all. I don’t think that they will manage to re-enter the market with low-power x86 processor because they lack the features needed by embedded designs (just look at how many components are integrated into modern embedded processors) and cost way too much to be competitive.
To be frank, Intel is used to the high-margin desktop/server segment where they do exceptionally well, the embedded market has razor-thin margins even in the high-end so I don’t see how they could justify developing lines of processors for it or even simply trying to focus on it with their investors.
i think Niagara 2 has:
– 8 core 64 threads .. Intel has fake 4 cores
– 2x 10 gbit ethernet embedded
– embedded Hardware encryption, random nuber generation (think of out of the box SSL acceleration)
– better power usage.
– embedded memory management. (ok AMD has it..)
this makes niagara 2 a very solid SERVER chip against whatever intel-AMD throws at least for the incoming years. the embedded stuff makes the server design less complicated although it has the draw back of pluggability problems.
And who besides Sun is likely to buy a SERVER chip from Sun?
Sun can cell chips into the high end market, where there’s no buyer, or the middle market, where they compete with Intel directly on Intel’s strengths, or the low end market, which they know nothing about, and which is full of established players.
You are mistaken. Niagara’s target market is low-middle end server market. that is direct competition with Intel. And yes, as long as they can deliver what they promise there will be buyers for that chip.
… even embedded sparc-cpus?
iirc olympus used them in some of their cams
und before i forget it:
the more architectures, the better
Especially embedded. that takes infrastructure that sun just doesn’t have.
Indeed there are, they’re called MicroSPARC and they’re still used in the Sun Ray thin clients.
I too question the future of SPARC. The RISC-like market isn’t growing these days, and IBM (and offshoots like PASemi, Freescale, AMCC) have it pretty well taken care of. IBM has basically been sleeping on R&D for the past 3 years for the simple reason that it can. IBM knows its stuff when it comes to silicon, and I’d place a bet that POWER6 and offshoots will continue that legacy.
SPARC, while novel, just doesn’t have much of a market anymore. Its previous niche was medium to high end servers, but Intel and AMD have really made their offerings quite competitive these days.
You could probably use a tool like Winhex to copy the drive in XP/Vista.
Not sure what you’d use to change the partition size. I have found that, even with this great Apple TV we still need a video converter (things like Apple TV Converter) in case to put our favourite movies onto Apple TV before playing them on TV Screen.
My apple tv converter
http://www.apple-tv-converter.com
> Sun’s a much smaller company and has an even worse position with respect to competing with Intel. They’re about where Intergraph was in ’89, except that unlike Intergraph, they don’t have the option of dropping their own processor and going to using Intel, er, wait, they’ve already started doing that. . .
Well, gee, the last time I checked Sun was doing pretty well with their SPARC processors lately — the Niagara processor beats the leaving crap out of any processor Intel has got on the throughput benchmarks (what counts the most on the servers). And Niagara II will be out this year at least doubling the delivered performance, which in effect will stomp all over the competition in the x86 space. Rock processor also sounds pretty interesting and juding by the sounds Sun is making, it should deliver the goods. SPARC is very much alive and kicking.
I have yet to see a “One size fits all” kill all other members on any market.
The Sun offerings really looks good for embedded networked appliances.
Hope they success as, sincerely, they deserves that.
thank Jeebus!
> Sun can cell chips into the high end market, where there’s no buyer, or the middle market, where they compete with Intel directly on Intel’s strengths, or the low end market, which they know nothing about, and which is full of established players.
So Sun knows nothing about low end market? Hmmm, gee, I guess the fact that Sun was the only vendor out of IBM, HP, Dell, and Sun that showed any growth in the revenue and unit shipments last quarter doesn’t tell you anything. So I guess they know a thing or two about making and selling low end machines after all. Fact of the matter is Sun presently has got the best server product line out there bar none, the value for money is simply unbeatable on all of their Galaxy and Niagara machines. Even though Sun is comparatively new player in the x86 space and so far, they’ve been doing great with their AMD powered Galaxy boxes. And SPARC reach will go even further into low end given the success of Niagara. If anything you’re likely to see more Sun machines in the datacenters around the world than fewer of them.
High end = servers, middle = desktops, low end = embedded. I should have been more clear.
Which is why, instead of going into the chip business again and defocusing, they should have remained in the server business.
Sun is also trying to push open/free Sparc implementations, like the open sourced T1 or the ESA subsidised GPL licenced Leon1/2/3 cores.
Maybe it will get some success in some niche but high profile markets : Space, Military, Aeronautics,… which could, in turn, help keep some activity on Sparc related products like workstations, servers, compilers…
Some people actually don’t care about bleeding edge CPUs but instead focus on long term availablity, deterministic behavior, ECC protected registers and memories.
[ESA=European Space Administration]