Jonathan Schwartz Shows 32-Way UltraSPARC Chip

Megaslow writes "The latest entry in Jonathan Schwartz's blog has pictures of Sun's Project Niagra chip, with 8 cores * 4 threads per core for a 32-way computer on single chip. He also shows what looks to be a test rig reportedly already up and running Solaris 10."

Comments

Jonathan Schwartz

[Anonymous Coward]

Wow.

Can someone say haircuit?

Re:Jonathan Schwartz

[michael path]

Would that be like, the merging of 'haircut' and 'circuit'?

I'll try.

'HARE-kit'.

How'd I do?

Re:Jonathan Schwartz

[Mr.Dippy]

Way to keep the stero-type alive that all hardware developers/programmers/general geeks look like social freaks that never heard of hair products. Seriously, how much is this guy making as CEO? I hear you can get a haircut at the Hair cuttery for like 13 bucks.

/off to go groom myself

Re:Jonathan Schwartz

[Hieronymus Howard]

Isn't that the standard issue hairstyle (and I use the world 'style' very loosely) for geeks the world over?

Re:Jonathan Schwartz

[REBloomfield]

well, that's what my hair 'cut' looks like anyway...

Re:Jonathan Schwartz

[spektr]

Can someone say haircuit?

This man clearly has no time to support his hair, because it is not part of his core business. Haircuts designed by an underpaid committee simply don't cut it. We suggest that he should open source his haircut using a GPL style license and let the community do all further development. ESR is already preparing a groundbreaking new paper named "The wig museum and the hairdresser's shop".

Re:Jonathan Schwartz

[TheRealMindChild]

So he doesn't have a buzz cut... what era do you live in? 1950's?

Re:Jonathan Schwartz

[LizardKing]

He wouldn't last five minutes in a school playground.

Re:Jonathan Schwartz

[Master of Transhuman]

So since he is in the "core" business, he should allow at least eight people to cut four "threads" of hair per second.

Should make for a fast, if unusual, cut.

I foresee a new method of cutting hair in those hair shops that have multiple stylists. All of them work on one customer, but it's done in seconds.

Wow

[lewp]

You know what I'd do if I had one of these?

Two chicks, man.

solaris fan

[BoldAC]

Okay, I am a solaris/sun fan boy. But this sounds like it was crafted by a professional commerical writer...

Ahhh... to be 38 and be this guy. President of Sun at 38 years old... what a life.


This is the silicon for our Project Niagara chip: 8 cores * 4 threads per core = a 32-way computer. On a chip.
And did I mention we have silicon, and not just a JPEG file?
And I saved the best for last. Are you ready?
It's already running Solaris. A volume OS that eats threads for lunch, on the world's most advanced massively parallelized silicon.
That's not just a box.
That's what we call a system. A system built for internet workloads. Not for the expedience of a press release. And a system that gives customers yet more choice, rather than taking choice away.
(And before you ask, yes, we are planning a nicer box when we ship :)


These guys deserve to Microsoft level of success...

Several of sun gurus have given us suggestions and hints at solaris section of our site. Without their early input and links from within the sun website, we would have never been as successful.

These guys are trying to do things big and correctly.

Re:solaris fan

[AlexTheBeast]

It does read like a commerical... ... a commerical against IBM.


I'm watching with amusement as IBM prepares to stub its toe with their new, curiously named "OpenPower" low-end boxes.

Now, I will freely admit I am entirely confused by what they're doing. Why on earth would you ship a proprietary computer that doesn't run your own operating system (AIX)? If I were trying to freak out my installed base, that's exactly what I'd do.


These guys are attacking IBM (and linux?) directly. The first part of his blog is a calculated attack against IBM--step by step he breaks down IBM's strategy. Just when you are wondering why you ever thought about using IBM, he introduces his new baby. He must have spent hours crafting this blog post.

Yes, it's a commerical.

I just can't believe that blog posts are this important now. I remember when we would just finger the inside guys we knew to see the plan. Now, it's been turned around into a commerical like everything else.

Re:solaris fan

[Dogers]

A system built for internet workloads.

Looks like blogs.sun.com isnt running one of these puppies..

And could this be the first anti slashdot device, which works decently??

Re:solaris fan

[jeif1k]

These guys are trying to do things big and correctly.

I think "big" and "correctly" are themselves contradictions in terms these days. Solaris is bloated, Java is bloated, and now they are producing a bloated CPU chip.

None of this will stop commodity hardware and open source from kicking Sun's butt and driving them out of business, because ultimately, people don't want "big", they want manageable and cost-effective. And that Sun isn't delivering anymore.

Ahhh... to be 38 and be this guy. President of Sun at 38 years old... what a life.

I'm sure he is well paid, but so are lots of other jobs. Other than that, I would imagine his job is causing him ulcers because deep down, he must know that his company is in deep trouble.

Re:solaris fan

[Quixote]

A system built for internet workloads

... and slashdotted already ;-)

Back to the drawing board again, eh, Jonathan?

Re:solaris fan

[BoldAC]

I think "big" and "correctly" are themselves contradictions in terms these days. Solaris is bloated, Java is bloated, and now they are producing a bloated CPU chip.

As much as I love sun, it's hard to argue with your points. Java and Sun both have some bloat... agreed. Even most main stream linux distros are bloated compared to a few years ago. However, it's hard for me to understand how a chip can have bloat? The bigger and faster and more the chip does, the better! Right?

People a few years ago were complaining because programs were too big. People don't really complain about this anymore because hard drive space has gotten so cheap. Now we usually complain about bloat for other reasons... mainly speed, usability and security.

Will cheaper, faster, bigger, and better chips help decrease this problem? Maybe, maybe not. But I don't know how it could hurt.

Re:solaris fan

[jeif1k]

However, it's hard for me to understand how a chip can have bloat? The bigger and faster and more the chip does, the better! Right?

One of the biggest bottlenecks in computers is the interface between the CPU and memory. It makes little sense to make the CPU so fast that it requires data faster than the memory system can provide it. So, 32-way CPU may, in fact, not be much faster than a single or dual core CPU when you build a real computer running real applications around it. Even if it turns out to be significantly faster, there may be other, simpler, cheaper ways of achieving the same speedup.

Based on the presentation in the blog, Niagara sounds just like more of Sun's "bigger is better" attitude, as opposed to sound cost/benefit tradeoffs. Keep in mind that the blog brags about the fact that Niagara is "32-way", not that it were faster than the best PowerPC chip.

Whether Sun's chip actually delivers good bang for the buck remains to be seen. I don't think any of Sun's recent machines have delivered good bang for the buck, so I won't be holding my breath (we used to be a big Sun customer but haven't bought any Sun hardware for several years now).

Even most main stream linux distros are bloated compared to a few years ago.

Linux distros, yes. The Linux kernel is still being kept fairly lean feature-wise relative to Solaris or NT (it just has lots of drivers and loadable add-ons).

Re:solaris fan

[elmegil]

Yes, it's a commerical.

Duh. Just like "Blog For America" was political advertising, and anyone who thought it was simply whatever Dean felt like writing was fooling themselves. Look at the author's agenda. Just because it's called a "blog" doesn't mean it's in any way authentic.

Re:solaris fan

[pegr]

I accidently modded the parent post off-topic. I'm replying to it to cancel the mod. Would some other mod mod this off-topic? (Love the irony!) Thanks...

Re:solaris fan

[Kehvarl]

I would, but having already opted out of modding, I decided to post this absolutely pointless bit of text instead.

Re:solaris fan

[Cajal]

The Solaris kernel is modular as well. Do you have any basis for claiming that the Solaris kernel is bloated? I doubt you've seen the source code.

Re:solaris fan

[jargoone]

I just modded this post off-topic for you. You're welcome!

Re:solaris fan

[Anonymous Coward]

I just can't believe that blog posts are this important now

Actually, not all blogs are important, nor are many/most CxOs that interested in Blogs. But it just so happens that Jonathan is VERY enthusiastic about blogs... that's why his blogs are like marketing press releases (read his blogs to see how fanatic he is about blogs... bit of recursion, but it's good read in case you care about blogs), and he apparently spends lots of energy on writing them.

Now, as to attacking IBM... hello? earth to major Tom; Sun has been fighting primarily with IBM for past 3 years, if not longer. Nothing new there. And as a result, they are trying to belittle IBM's linux strategy as well.

Re:solaris fan

[fsbilly]

Not to mention getting to appear opposite Bill Murray and Mickey Rourke!

Re:solaris fan

[jeif1k]

The Solaris kernel is modular as well.

I made no claims about modularity. What does modularity have to do with being bloated?

Do you have any basis for claiming that the Solaris kernel is bloated?

Sun proudly proclaims so themselves; just look at feature lists.

Re:solaris fan

[Kehvarl]

what the hell? why did I get anything for that, funny or not?! bloody insane mods I tell you.. bloody insane.

Re:solaris fan

[kaffiene]

Um... you do realise IBM and Sun are competitors, don't you?

Do you expect press releases from Intel claiming that the new Opterons are "really bitchin"?

Grow up.

Re:solaris fan

[Anonymous Coward]

After your joke, jargoone, gotta love the mod who completed it all by giving you the "off-topic" ;-)

No, I'm not that mod, obviously ;-)

Re:solaris fan

[Cajal]

Well, in a modular kernel, you can avoid loading those modules you don't need....

Further, you've provided absolutely no evidence supporting your claim that the Linux kernel is "leaner" than the Solaris and WinNT kernels. You only linked to a feature list for an entire operating system, which has nothing to do with the size of a kernel (and even if Solaris' kernel is bigger than Linux, what does that have to do with the suitability of Linux or Solaris to any given task?).

But let's look at this feature list you've offered up as proof:

Predictive Self-Healing is mostly in user-space. It's a framework in which various daemons report error codes to a user-space daemon, which diagnoses the problem and takes appropriate action (restarting services, increasing filesystem size, taking hardware offline, etc). There's likely little new kernel code here.

N1 Grid Containers does require new kernel code. But so does User Mode Linux, which accomplishes a similar goal. Conviently, you neglect to mention that.

DTrace is kernel space, but it's implemented in a modular fashion - the various DTrace probe providers don't have to be loaded. And since the probe points are dynamically generated, there's little bloat.

Dynamic File System (ZFS) is a new filesystem driver. Solaris already has a lot of these. So does Linux. In fact, Linux has more (how's that for bloat).

Process Rights Management is a new permission system. This will add some kernel code, but not reams. This really is just integration of existing Trusted Solaris code.

IP Filter - Sun removed SunScreen and replaced it with IPF. Big woo.

So, your claims don't add up.

Re:solaris fan

[jeif1k]

you've provided absolutely no evidence supporting your claim that the Linux kernel is "leaner" than the Solaris and WinNT kernels.

Nor have you provided evidence to the contrary.

In any case, Sun itself likes to brag about all the enterprise features Solaris has relative to Linux and that's why high-end users should all be using Solaris. Are you saying Sun is not telling us the truth?

Re:solaris fan

[JonAnderson]

@jeif1k You do realize that whole fucking point of Niagara is to address the problems you talk about when discussing CPU and memory? Niagara is a single die, 8 cores each capable of running 4 threads simultaneously. Sun have realized that core performance is far outstripping memory subystem performance. Just look at itanium. Niagara hides memory latency (stalls) by scheduling different threads (1 cycle latency to context switch). Individual thread performance will suck, particularly in Niagara1. However, throughput of massively multithreaded workloads (webservers etc.) will be impressive. Particularly as these will be in 1U and 2U rack systems. Sun are really starting to turn the corner now.

Re:solaris fan

[JonAnderson]

Nor have you provided evidence to the contrary.

WTF!!! Are you saying that you can post whatever unsupported bullshit you like but it's other peoples responsibility to point that out? And then call them out for NOT doing so. Man you are too funny.

Re:solaris fan

[Cajal]

Do you understand how operating systems work? No, I'm serious.

Just because Sun claims that lots of "enterprise features" doesn't mean they're all in the kernel. You do understand the concept of userspace, yes?

Re:solaris fan

[Kehvarl]

Much better :] was that so hard?

Re:solaris fan

[jeif1k]

WTF!!! Are you saying that you can post whatever unsupported bullshit you like but it's other peoples responsibility to point that out?

First of all, you are missing the point. We are discussing design, not quantifiable parameters. It is Sun marketing, not me, that keeps harping on how wonderful all the new functionality is they put into the Solaris kernel. That's the wrong direction in kernel design. (And, incidentally, Sun is spending millions of dollars to publish that sort of "unsubstantiated bullshit" in the form of marketing materials.)

Second, yes, if you want to change the discussion into one about quantifiable criteria (rather than design attitudes), the burden is on you to define the criteria you want to discuss first.

Re:solaris fan

[JonAnderson]

These guys are trying to do things big and correctly. I think "big" and "correctly" are themselves contradictions in terms these days. Solaris is bloated, Java is bloated, and now they are producing a bloated CPU chip. None of this will stop commodity hardware and open source from kicking Sun's butt and driving them out of business, because ultimately, people don't want "big", they want manageable and cost-effective. And that Sun isn't delivering anymore. Ahhh... to be 38 and be this guy. President of Sun at 38 years old... what a life. I'm sure he is well paid, but so are lots of other jobs. Other than that, I would imagine his job is causing him ulcers because deep down, he must know that his company is in deep trouble.

You didn't post this complete pile of FUD? And don't you feel the need to justify any of the staements contained therein? And who are you to say what is and isn't the right direction in kernel design You do realize that Niagara is for the LOW END don't you? These are not going to be expensive machines, in fact, they are more like alternatives for racks of commodity (x86) hardware.

Re:solaris fan

[jeif1k]

You didn't post this complete pile of FUD?

What you call a "pile of FUD" is an opinion. People have them, and they talk about them. Sorry if you don't understand the difference between, say, an opinion and some other kind of statement.

As for the term "FUD", give me a break. The term "FUD" refers to a sales technique by which big, established companies try to cast doubt on the viability of a new competitor. I couldn't "FUD" if I wanted to because I'm not a company and I don't have a product to push. Sun sales force has learned well from IBM and Microsoft, however, and Sun employees are masters of FUD: they are trying to cast fear, uncertainty, and doubt on everything that competes with their proprietary offerings: open source projects, on open source desktops, on IBM's Java efforts, on C#, on Linux, on IBM's Power processor, etc. Schwartz's uncalled-for rant against IBM and Power is just another example of Sun's FUDding.

You do realize that Niagara is for the LOW END don't you? These are not going to be expensive machines, in fact, they are more like alternatives for racks of commodity (x86) hardware.

We used to buy Suns, but Sun hasn't been able to produce a machine with a price/performance ratio that would interest us in years. Furthermore, the fact that they are not x86 compatible and that they will (apparently) be using Solaris as their OS are two more big strikes against them because what "LOW END" customer is going to have the expertise to manage those machines or the software to run on them?

Re:solaris fan

[JonAnderson]

Well, FUD is FUD - It's not restricted to marketeering. You make several 'opinions' which when asked to substantiate cannot or will not. Opinions should be based on facts. Can you please post the facts which you base your opinions on please so we can all be enlightened? Whats your beef with Sun?

Furthermore, the fact that they are not x86 compatible and that they will (apparently) be using Solaris as their OS are two more big strikes against them because what "LOW END" customer is going to have the expertise to manage those machines or the software to run on them?

Well, who said they were going to be Solaris only? I am pretty sure they will run Linux. And the argument regarding expertise to manage these machines is vacuous. Apparently people have resources to manage Linux but not Solaris - you GIVE ME a break!

Re:Wow

[Anonymous Coward]

Eh?!

Its says 32-way Think bigger!

Blog?

[Roofus]

By blog, you mean a press release from an officer instead of the Marketing/PR dept right?

I don't care if this guy is a miracle child from the gay marriage of President Bush and the Pope, this blog reads like it was written by a pompus asshole. Apparently everyone but Sun sucks.

Re:Blog?

[Bender_]

It may help to use the right spelling, it is: P O M P O U S

Re:Blog?

[His name cannot be s]

Fuck.

I thought it was, but then I made the mistake of looking at the parent post.

Ah, well. Shit happens, life goes on.

Pity I got no comeback...

here: just for fun.

His name cannot be s : ...bla bla bla.. Pompus Asshole ...

Witty Retort Man: What no U?

His name cannot be s : No, just You!

ha ha ha...

Feh.. I need some coffee..

Re:Blog?

[His name cannot be s]

Fuck. I *REALLY* need coffee.

I just realized I missed the O not the U.

That could have been funny, instead it wasn't.

Re:Blog?

[Anonymous Coward]

I'm starting to think a lack of coffee isn't the problem here.

Niagra?

[gutman]

Is that like a cross between Niagara and Viagra?

Re:Niagra?

[Anonymous Coward]

Almost. Both helps your critical "systems" stay up longer.

Re:Niagra?

[JohnboyHolmes]

> Is that like a cross between Niagara and Viagra? Yes, the designer was standing on his head watch the water flow straight UP when he came up with the name :-)

Re:Niagra?

[Anonymous Coward]

Almost. Both helps your critical "systems" stay up longer.

Only if you don't play with it too much.

Re:Niagra?

[eigerface]

Actually, you can copyright Niagra.

Niagara is in the public domain.

Apparently Not Such a Powerful Chip After All

[The Other White Meat]

So far, there have been like 8 posts on this article, and the article itself seems to have been slashdotted. If they have Four Processors per Poster, you'd think they could keep the page up...

Re:Apparently Not Such a Powerful Chip After All

[Anonymous Coward]

It's not exactly confidence inspiring if Sun's own Solaris servers can't handle a Slashdot size load...so that begs the question, does Niagra run Linux?

Re:Apparently Not Such a Powerful Chip After All

[Anonymous Coward]

Listen, a threeway keeps you busy enough, but a 32-way?

Yikes, so wonder Schwartz can't keep it up!

Why doesn't he stick to a smaller, more manageable, harem size?

Re:Apparently Not Such a Powerful Chip After All

[Anonymous Coward]

FYI

blergh

[gustgr]

BLOGS.sun.com

enough said.

I can say 'haircuit'...

[johannesg]

...but I have absolutely no idea what it means.

Mirror

[AlexTheBeast]

If you are going to run an ad... keep the ad web server up, shmucks.

Quote:

The difference between humans and white mice.

I'm watching with amusement as IBM prepares to stub its toe with their new, curiously named "OpenPower" low-end boxes.

Now, I will freely admit I am entirely confused by what they're doing. Why on earth would you ship a proprietary computer that doesn't run your own operating system (AIX)? If I were trying to freak out my installed base, that's exactly what I'd do.

Surely they should read my earlier entry here, regarding the history of OS blunders, and the difference between humans and white mice. (White mice learn from history, while humans have a harder time in far broader fields of endeavor.) Chips don't matter if they don't have software (see Dec ALPHA for the ideal example), and software doesn't matter if it doesn't run in volume (see HP/UX on Itanium).

Second, saying "it's ok, we run linux" is like saying you "run the internet." Sure feels like IBM is trying to avoid specifying the distro. Why? Because they'd be doing demand creation for Red Hat. And why buy WebSphere when you can just use what comes in Red Hat? - "Jonas (Red Hat's app server) is just a toy, it's just for the low end" said IBM's exec at the Smith Barney Tech Conference I just attended in NYC. Notwithstanding the familiarity of that refrain to how linux itself was mistakenly positioned a few years ago, the irony is that IBM is positioning these new boxes as low end boxes. Presumably ideal for running a low end app server, and just using what's in Red Hat.

Finally, the 'P' in Power5 stands for Proprietary. You can't claim your chip is open if you're the exclusive supplier, guys - at least you can dual source SPARC from Sun or Fujitsu. Perhaps we should rename SPARC OpenSPARC. Nah, I like what AMD is doing with "industry standard" better. And while SPARC is outshipping Power 3:1 (so sayeth IDC), sure sounds like we're the industry standard.

IBM saying they're using this to come after Sun really suggests they've gone a few degrees shy of plumb - the single biggest threat to low-end SPARC isn't a funny low volume Power5 box without an operating system. The big alternative to SPARC arose years ago from volume in the x86 market. That's why we've built out the most complete family of Solaris/Opteron systems the industry has to offer, and we're starting to drive into the $20B+ x86 market. Volume has spoken.

That's also why we changed tack with SPARC, to move away from the single thread approach, to truly parallelized multi-core computing. And not just a tepid two core approach - the internet is one massive, multi-threaded application environment. Every user is, for all intents and purposes, his own thread - whether they're shopping for chandeliers on eBay, or managing wealth at Lehman Brothers. So if you want to see what multi-core computing looks like, allow me to help. It looks like this:

This is the silicon for our Project Niagara chip: 8 cores * 4 threads per core = a 32-way computer. On a chip.

And did I mention we have silicon, and not just a JPEG file?

And I saved the best for last. Are you ready?

It's already running Solaris. A volume OS that eats threads for lunch, on the world's most advanced massively parallelized silicon.

That's not just a box.

That's what we call a system. A system built for internet workloads. Not for the expedience of a press release. And a system that gives customers yet more choice, rather than taking choice away.

(And before you ask, yes, we are planning a nicer box when we ship :)

(2004-09-10 21:59:34.0) Permalink

Sunday September 05, 2004
Volume wins.

As I've said, I'm a big believer in the idea that volume wins. And we invest (much to the occasional befuddlement of our friends on Wall Street) to support that thesis - most notably in the propagation of our programming platform, Java.

And in the J2ME mobile handset

Sounds Pretty Bitter

[Apple Acolyte]

Schwartz's blog may not be representative of the general corporate attitude at Sun, but he comes across as bitter and even hostile. Perhaps he is just a passionate believer in his company's work, but his whiney tone smacks of unprofessionalism. I'm not particularly well versed on the continuing saga that is Sun, but should not product performance be speaking for itself? In any case, if they have achieved something noteworthy with this "32-way" chip, I hope they figure out a way to make it useful. This MPR Paper on the processor may be of interest to some.

Re:Sounds Pretty Bitter

[Jeff DeMaagd]

but should not product performance be speaking for itself?

It is not necessarily the case. Yes, it seems like the product is pretty good but the days were heavy iron were typically faster than mainstream processors are gone. Instead, heavy iron is even more of a niche, which may be eroded. Sun is kind of working from reputation in order to sell their low end stuff, and their high end would seem the be eroded by the increasing prevalance of x86-64 systems.

Re:Sounds Pretty Bitter

[rs79]

Schwartz's blog may not be representative of the general corporate attitude at Sun, but he comes across as bitter and even hostile.

He's a programmer. You're lucky he cleaned up that well for a photo.

Beware of smiling programmers.

Re:Sounds Pretty Bitter

[Animixer]

If a company could get by on "product perfomance speaking for itself", DEC would still be around and we'd all be running VMS on Alpha.

Re:Sounds Pretty Bitter

[Kuad]

Why isn't there a "+1 - The bitter truth" option?

Re:Sounds Pretty Bitter

[Anonymous Coward]

It's called "insightful".

Why isn't there a "-1 - Dork" option? Oh yeah, it's called "troll".

Re:Sounds Pretty Bitter

[Mateito]

he comes across as bitter and even hostile

He's 38, he's worked his way up to COO of a decent sized multinational IT company, he's got dud hair but at least he looks like a 90s geek instead of a 70s geek (Mr Gates).

Because of all this time dedicated to work, he's probably never had a pussy he didn't rent.

Of course he's bitter and hostile.

Re:Sounds Pretty Bitter

[jamiethehutt]

but should not product performance be speaking for itself?

It should, but as DEC and the Alpha show it doesn't.

Re: Sounds Pretty Bitter

[Leadmagnet]

Opps your link has an error in it. It should be - http://en.wikipedia.org/wiki/Slashdot

Re:Sounds Pretty Bitter

[JonAnderson]

Where does it say Niagara is for heavy iron? Where? Niagara is for network facing loads and is meant to go, intially, into 1U and 2U boxes. In fact, re-reading your post it's 100% crap. Where did you come up with:

heir high end would seem the be eroded by the increasing prevalance of x86-64 systems.

Hold on...

[StevenHenderson]

Sun's Project Niagra chip, with 8 cores * 4 threads per core for a 32-way computer on single chip

Doesn't mean a damn thing unless software is written to take advantage of it. Damn PC developers can't write software to take advantage of HT (with some exceptions, I know), but hopefully this chip's power can be realized fully.

Re:Hold on...

[ban]

Doesn't mean a damn thing unless software is written to take advantage of it. Damn PC developers can't write software to take advantage of HT (with some exceptions, I know), but hopefully this chip's power can be realized fully.

Why is it that people keep stating that you have to write software that targets HT specifically? This is not true. Any multithreaded application will benefit from it. It is up to the OS to present you with the CPUs, real or virtual.

Yes, there are specific issues with handover during tight spinloops et.c., but only people writing locking or timing code should have to deal with those issues. Not your average application programmer.

Re:Hold on...

[Anonymous Coward]

Damn PC developers can't write software to take advantage of HT

Why should "PC developers" target a one-manufacturer buzzword of a soon to be dead product line?

Re:Hold on...

[StevenHenderson]

Hyperthreading is a marketing name, yes. A buzzword only, though? No. Saturation of execution units = better performance under most cases. Whether it is living up to its potential is a different story, but it is a great idea in terms of engineering, not just marketing.

Re:Hold on...

[Anonymous Coward]

Software is already written to take advantage of it. When the Apache receives a HTTP request, it will start a new thread to handle it and continues to listen for more requests. 30 simulatenous requests => 31 threads for Apache alone.

Re:Hold on...

[jarich]

Why is it that people keep stating that you have to write software that targets HT specifically? This is not true. Any multithreaded application will benefit from it.

This statement is true, but...

HT or a 2nd CPU will get you somewhere between 10 and 20% boost on your software. It does this letting OS operations like disk IO, video, etc, run on the second "CPU".

If you learn to write good threaded code, you can see nearly 100% speed increase per CPU. That's the difference.

And just to turn this into a more interesting thread, I do with this Java. :) And yes, it scales nicely if you use it properly.

Anyway, when people talk about how no one knows how to take advantage of HyperThreading or multiple CPU machines, this is what they are referring to. 10% boost versus 100% boost.

Re:Hold on...

[Detritus]

I can see an application for this sort of chip, adding intelligent I/O processors to the system, like the channel controllers on mainframes. Much of the I/O on modern computers is done at an embarrassingly low level.

Re:Hold on...

[TigerNut]

In general (and unless you force them, by writing the code that way) Windoze apps are not at all shy about holding on to their CPU time to the exclusion of other tasks and threads. As a (Borland) example, if you're doing a bunch of processing in a loop, then you must include the occasional call to Application->ProcessMessages(), or else your button presses etc. won't get any airtime until the loop concludes. Nothing wrong with that - it helps to keep the program state sort of known when you're handling events - but if you don't do it then your thread interaction is severely limited.

If you want to take optimal advantage of a parallel multithreading environment then you have to explicitly tell the compiler about what parts of your code are independent of each other, what events can be processed when, and what actions are dependent on what events. The compiler isn't going to make those decisions (correctly or optimally) for you.

Re:Hold on...

[davecb]

So why would a PC developer want to? They think they are a single program taking over the machine.

A Unix/Linux developer knows that there may be multiple instances of a given program running on a server, sharing executables and having better locality of reference.

And if the program has a need to do more than one thing at a time, the authors will thread it or make it run multiple instances.

Imagine a Beowulf cluster of 16 blades, each running a 32-way CPU for 512 processors per shelf. For about 1.5 times the price of 16 uniprocessor Pentiums. (512/16)/(1.5/1) = 21 times the price-performance, and a really small footprint to boot.

--dave

Re:Hold on...

[TheRaven64]

SMP (of which HT is one implementation) is not magic. You do not have any more execution units than you had without SMT. The first thread to be scheduled gets to pick what execution units it wants to use. The second thread gets to pick from the left-overs (then the third and fourth threads on each core in this chip get a go in this example, although HT only supports 2 threads). This means that for each core, you have two non-uniform CPUs. Scheduling threads optimally on uniform CPUs is relatively easy. Scheduling on non-uniform CPUs is a whole lot harder. But wait, it gets worse - the capabilities of the second virtual CPU vary depending on what you're running on the first one. In order to get a significant benefit from SMT, you need a scheduling algorithm that takes all of this into account (I don't know of any that exist on paper, let alone in implementation).

An alternative is to design your threads so that they will be using different parts of the chip (for example, run a floating point intensive thread and an integer intensive thread at the same time). This, however, will only work in very specific environments.

Re:Hold on...

[DarkMan]

You can't get a 100% speed boost with HT.

HyperThreading presents a virtual CPU - there are the same number of compuational units in the core. The speed advantage from HT is that with threaded code, there is additional silicon to save the state of each thread.

With the same number of execution units, however, you can't get better then optimal for a single non-HT CPU; once all the execution units are full, that's it, whether it's one thread or two.

HT is a great boost for a threaded app where it is not CPU bound in general.

On the other hand, a second CPU carries an aditional set of execution units; that means you can, in theory double the CPU output.

Neither helps with the memory - cpu bandwidth issues, which can limit performance. Dual CPU can mitigate that with dual memory controllers (see, e.g. Opteron), but that has it's own complications.

So, HT is a small step between dual CPU, and dual CPU.

However, even the most optimistic benchmarks from Intel that I have ever seen quote a 30ish% speed increase with HT. I have never, ever, seen anyone claim that HT can give a 100% speed boost - can you reference that claim?

I quote from
http://www.intel.com/business/bss/products/hyper th reading/server/index.htm":

While Hyper-Threading Technology will not provide the level of performance scaling achieved by adding a second processor, benchmark tests show some server applications can experience a 30 percent gain in performance.

Re:Hold on...

[Brandybuck]

Why is it that people keep stating that you have to write software that targets HT specifically?

Because otherwise you're not using the HT efficiently. It is a single CPU pretending to be two. Unless your software was optimized for HT, you won't see much advantage to it. Heck, if you're running a scheduler that assumes SMP, you might even get a performance decrease!

Re:Hold on...

[akuma(x86)]

Not true. Any multithreaded app may NOT show a performance improvement on an SMT machine.

You can have 2 or more threads that want a shared resource (like the cache) that thrash each other leading to a degradation in performance vs. a single threaded processor.

Other resources are typically shared like front-end bandwidth, execution bandwidth etc...

The threads can be NON-symmetric. It's SMT (Simultaneous multithreading).

ultrasparc

[GMail Troll]

This is fairly impressive, althrough really with chips written in high level hardware description languages it probably isnt very difficult to just add more cores. I'd certainly like to see what the power and heat requirements are for this thing (forgive me if this is mentioned in the article, it appears to be slashdotted so I cant read it). I guess one of the advantages of the SPARC architecture is that the relative simplicity of the instruction set (compared to x86) makes it possible to do things like this.

gmail invite

Re:ultrasparc

[Anonymous Coward]

The chip is supposed to run very cool and is aimed at webservers and similar applications, it's not a 32 way Ultrasparc IV that you can do lots of floating point operations on.

Re:ultrasparc

[Anonymous Coward]

The accomplishmeny is not just that the fact that each CPU has an 8-way core (which is kind of cool since just about everyone else only has either a dual or quad setup). It's also the fact that the rest of the system (I/O buses, memory, disk, network) is designed to do a decent job of keeping the CPUs fed so they do useful work.

That's one of the reasons why Sun boxes are so expensive: they're designed from the ground up to keep running. I know of people who still use Sun LXes (9+ year old system) running Linux/BSD for light-weight FTP servers. They run just as well as when they were first purchased (original hard drives). Then there are the Sparc 5s, 10s and 20s.

(Though I find the Ultra 5s and 10s "cheap".)

Re:ultrasparc

[Boone^]

Even in the layout were fully custom, you can macroize the finished core and replicate as many times as you want. All that's needed is the global net routing and you're done.

Re:ultrasparc

[Mateito]

I've got an LX that I use as a remote X-terminal for my principal RedHat box. Boots off a floppy. :)

Re:ultrasparc

[sydres]

it almost looked like it had a water block on it at first then i though maybe peltier but no heatsink
plus it had no ram so I think that it was just a bare mobo

/. article. But a 32 way processor?

[tod_miller]

I can see these lovely ladies being applied to some serious film fx... I wonder what kind of advantages these systems would give to rendering houses, or is the cost of these for farming cpu power too high, and there is more bang for using Durons?

I guess this differs based on each application and resource requirements.

Still, nice.

Re:/. article. But a 32 way processor?

[Anonymous Coward]

The entire chip shares a couple of floating point units, it's not a number cruncher.

Re:/. article. But a 32 way processor?

[Anonymous Coward]

This isn't made for rendering, or for stuff that requires floating point power.

Since the chip has 8 cores, each core is quite simple. This kind of chip is more suited for database and web servers, where there are lot of simultaneous requests, but fullfilling a single request is quite simple task.

You can find more information about Niagara here.

Re:/. article. But a 32 way processor?

[davecb]

Actually it's very close to the Ultra II core. The idea is to be able to keep running during a cache refill by duplicating decoders and register files. The ALU isn't much different from the one I'm running as I type this (;-))

--dave (Solaris and Linux on SPARC, Linux on Intel) c-b

Re:/. article. But a 32 way processor?

[davecb]

Ok, assume you have a full 19" rack of dual 1U Intels. Into the same rack you can put 3U blade chassis, with 16 blades per, and 32 logical CPUs per blade.

The benefit is (1/3) * (16*32/2) or something like 85.3 to one. And the chip that they stareted with is pretty good at graphics, bitblit and floaty-point.

--dave

Re:/. article. But a 32 way processor?

[tod_miller]

I guess that answered my question! Thanks :-)

I wonder if we will see a film using this kind of firepower any time soon...

Thanks again

A bit off topic, but what about OpenSolaris??

[Anonymous Coward]

Does anyone have any information about the status of the open source version of Solaris 10?

I am realy looking forward to that. I'll be a great OS for open source people like me incase Linux doesn't survive it's legal hurdles.

Plus it would be great to integrate it with other FOSS based OS implimentations. Once people see that FOSS has compition between IBM vs Sun vs Novell and other companies, then that will make Windows seem more old fasioned and stale (and expensive).

Monopoly breads inferior products, as MS has demonstrated with WinXP's security record.

The heart of successfull capitolist system is compitition, and this compitition breeds the best products at the cheapest prices.

Linux desktops + Solaris servers in a completely open enviroment has a great deal of appeal to me.

I'd be ALL SET FOREVER

[gunnk]

I'd be all set FOREVER if I could only get 20 of these! I mean...

640 Processors should be enough for anyone! :-)

Re:I'd be ALL SET FOREVER

[Anonymous Coward]

I'll wait for the specs of Duke Nukem Forever thank you very much.

And their blog server still gets slashdotted!!!

[notany]

Apparently blogs.sun.com is very bad marketing.

Sun is doing what they do best

[sofakingon]

Sun is back to doing what they do best, designing extremely reliable , high-performance hardware. I really do hope they can adapt their business model VERY soon. I know that the mission critical systems and workstations at my place of employment DONT use x86 hardware and/or M$ products- but then again there are how many people with Solaris experience and certifican VS the M$ crowd (myself included)? On the flip side, we don't need all of a 2 story building to house a classroom for MS training because everyone already knows Windows (in theory at least).

Maybe the open sourcing of Solaris will help them pool their resources better and re-direct their efforts?

What is Open?

[peterdaly]

Does Johnathan not get it, or is he playing the FUD game? The IBM Open chip is not a chip without an OS. It runs linux...a commodity OS. That means two major things.

1. People who run Linux on a different box may be more likley to upgrade to the Open chip since they won't have to take an OS change into account as well.

2. People not happy with big blue can migrate to another vendor without having to take an OS change into account. That means less lock in.

Sun doesn't get it. Or more likley they do, but don't want to help their customers figure it out.

-Pete

Re:What is Open?

[Mark Round]

2. People not happy with big blue can migrate to another vendor without having to take an OS change into account. That means less lock in.

Well, not exactly. If they are running IBM's POWER processor, then they can't really move their applications to another vendor, as no-one but IBM "does" POWER. They could move to another platform and still run Linux (say, x86 for example), and manage to apply _most_ of their sysadmin experience - but any proprietry, binary-only applications running on that box would have to either be bought again or re-licensed. So there would be an OS change, even if it's only from one architecture to another.

-Mark

Re:What is Open?

[jeif1k]

Does Johnathan not get it, or is he playing the FUD game?

He is playing the FUD game, and it's nothing new. For example, he has referred to Gnome as "open source crap" that supposedly requires the genius and brillians of Sun's GUI designers to put right (CRN article). He has also tried to portray Linux as a kind of low-quality entry-level UNIX that isn't enterprise ready, with Solaris being the high-performance choice.

The most annoying and dangerous thing about Schwartz is that he keeps trying to redefine the meaning of "open" and "open source" in order to get Sun's highly proprietary platforms and products (e.g., Java) to be more widely accepted.

Why is Schwartz doing this? Because he must know that Sun is basically finished as a company: their hardware is uninteresting and Solaris is uninteresting. The only piece of software they own that anybody cares about is Java, but as soon as Sun tried to use their legal control of Java to extract any significant amount of revenue from it (and, make no mistake: Sun has tighter legal ownership of Java than Microsoft has of any of their platforms), they would have a rebellion on their hands.

Re:What is Open?

[Anonymous Coward]

As another follow-up mentioned, POWER is IBM-only. SPARC actually has two suppliers: Sun and Fujitsu. (Anyone can actually license the SPARC architecture and fab their own chips. I would say the SPARC is the most "open" of architectures.)

Second, if you want to try out Solaris you can download it (for both SPARC and x86) from Sun's site. You are able to use it for both non-commercial and educational use. While it's not libre, it is gratis.

Re:What is Open?

[Decaff]

Because he must know that Sun is basically finished as a company

Nice to know you have such predictive abilities.

: their hardware is uninteresting and Solaris is uninteresting.

I would have thought that their SMP hardware would be interesting to geeks - unlike the compromised NUMA architecture that lesser Unix boxes run.

Solaris should definitely be of interest to anyone interested in UNIX or Linux - unless features such as partitioning and scalability are dull?

The most annoying and dangerous thing about Schwartz is that he keeps trying to redefine the meaning of "open" and "open source" in order to get Sun's highly proprietary platforms and prducts (e.g., Java) to be more widely accepted.

Java 'Highly proprietary'? Ah - I guess that is why the spec is published, and why GNU can implement Java; why Java is the most in-demand language for IT jobs, and why its so widely targetted on sourceforge.

There is nothing proprietary about Java, only the name, which you must pass compatibility tests to use. Without these compatibility tests, Java would have fragmented years ago.

(and, make no mistake: Sun has tighter legal ownership of Java than Microsoft has of any of their platforms)

Evidence?

Sun already extract a lot of revenue from Java - they use it as their language of preference when the provide software consultancy services (now a significant part of their revenue). There is also J2EE licencing, and J2ME services and partnering.

Re:What is Open?

[Daniel Ellard]

1. People who run Linux on a different box may be more likley to upgrade to the Open chip since they won't have to take an OS change into account as well.

Linux runs on Sparc, as does FreeBSD. There's no reason to think that these kernels will not be ported to take advantage of the Niagra architecture. If you don't like solaris, nobody is going to force you to use it (except maybe your customers, if that's what they like).

2. People not happy with big blue can migrate to another vendor without having to take an OS change into account. That means less lock in.

I call FUD on you. Doesn't "big blue" make the Z architecture, OS/390, System/36, etc., which are text-book examples of vendor lock-in? As in all other situations, you choose the system that works best for you. If vendor lock-in is fear for you, get Linux from IBM (or Sun, for that matter).
On the other hand, if you need something that only System/36 or Windows or MacOS or Solaris provides, then you bite the bullet and buy that.

Re:What is Open?

[AusG4]

Ever consider that maybe some people won't -want- to run Linux on it?

Live in fantasy-land all you want.. the vast majority of the people buying this chip will -want- Solaris on it, even if given a choice.

How this got modded up to is beyond me.

Re:What is Open?

[Anonymous Coward]

I would have thought that their SMP hardware would be interesting to geeks - unlike the compromised NUMA architecture that lesser Unix boxes run.

You are mistaken. Sun's SunFire systems are NUMA. They have non uniform access to memory. This is a *good* thing, it allows large machines to not have to have to distribute high frequency clocks long distances, it allows multiple memory controllers and memory subsystems. It allows high speed access to memory close to a processor.

As far as scalability goes, Linux runs on "lesser" 512 CPU systems, with plans for 1024 and 2048. Solaris is at 144 CPUs. They're the cold, hard facts.

Re:What is Open?

[LizardKing]

Linux runs on Sparc, as does FreeBSD.

They run on most UltraSparcs (not sure about the UltraSparc IV), but they don't handles large numbers of CPU's or massive loads as well as Solaris does. This is only to be expected, as few people have access to high end Sun kit to do Linux hacking on it. Despite Schwartz's use of the "open architecture" mantra (something the OpenBSD guys would find contentious seeing as they can't get UltraSparc IV documentation), Sun's vendor lock in is Solaris. That's why they are pushing Linux on their low end Opteron stuff, but limiting their high end UltraSparc stuff to running Solaris.

Re:What is Open?

[SnowZero]

I would have thought that their SMP hardware would be interesting to geeks - unlike the compromised NUMA architecture that lesser Unix boxes run.

NUMA is good, as that's the only way things scale; Shared busses are what is bad. Crays, Opterons, and Itanium servers use NUMA. The Xeon and AthlonMP use a shared bus. I wouldn't call a Cray a lesser unix box.

Solaris should definitely be of interest to anyone interested in UNIX or Linux - unless features such as partitioning and scalability are dull?

Every Unix besides SCO supports that stuff at this point, its not really a market differentiator. If I wanted large-scale scalability I'd talk to SGI, not Sun. Guess which company is working on Linux scalability now? But for most things, going beyond 8 way is incredibly rare. For example, how many network interfaces would a 32-way system need to compete with a group of 4 way servers? Niagara is interesting yes, but I'll wait until they have numbers before I'd bet anything on it.

Java 'Highly proprietary'? Ah - I guess that is why the spec is published, and why GNU can implement Java; ... There is nothing proprietary about Java, only the name, which you must pass compatibility tests to use.

Java is a proprietary language with an open implementation. One company controls the language spec, rather than a standards body or industry group. It's rather like the difference between DirectX and OpenGL. DirectX has open specs and many drivers that implement it, just like Java. But ultimately you have to trust one company with the evolution and future viability of the language.

GNU GCJ is kind of a bad example since Sun dislikes them for working on a native compiler, and generally treats them like dirt. The blog link is for one of the main GCJ developers.

Re:What is Open?

[Decaff]

You are mistaken. Sun's SunFire systems are NUMA.

You are right in this case, but many Sun systems (including earlier SunFire machines) are SMP.

As far as scalability goes, Linux runs on "lesser" 512 CPU systems, with plans for 1024 and 2048. Solaris is at 144 CPUs. They're the cold, hard facts.

Scalability has nothing to do with the number of processors. Just because you can place a Linux system on that number of CPUs does not mean that it will efficiently make use of them for general purpose use. The large CPU count Linux systems are often for very specialised numerical work where a coder puts in a lot of effort (using specialised libraries) to distribute algorithms. You may be able to use kernel services (such as I/O) from all of those processors, but that doesn't mean it will scale within the kernel (although, in these applications, it doesn't need to).

On the other hand, high-processor count Solaris/Sparc systems are usually more general purpose business machines, running database engines, app servers, web services etc. These hammer the CPUs with unpredictable and varied requests.

The 'I put my OS on more processors that you' argument is very misleading.

Re:What is Open?

[Anonymous Coward]

Scalability has nothing to do with the number of processors.

Do you take me for an idiot? Scalability is exactly that. By definition.

Just because you can place a Linux system on that number of CPUs does not mean that it will efficiently make use of them for general purpose use.

Of course not.

The large CPU count Linux systems are often for very specialised numerical work where a coder puts in a lot of effort (using specialised libraries) to distribute algorithms.

You may be able to use kernel services (such as I/O) from all of those processors, but that doesn't mean it will scale within the kernel (although, in these applications, it doesn't need to).

In many of these applications, it is actually imperitive. They can be processing and/or generating many TB of data.

On the other hand, high-processor count Solaris/Sparc systems are usually more general purpose business machines, running database engines, app servers, web services etc. These hammer the CPUs with unpredictable and varied requests.

The 'I put my OS on more processors that you' argument is very misleading.

But it is better than not making arguments at all, and just *asserting* that 'my OS is more scalable than yours', like SOME people.

Re:What is Open?

[ccp]

Does Johnathan not get it, or is he playing the FUD game?

A little of both, I guess...

Cheers,

Carlos Cesar

Re:What is Open?

[Decaff]

I wouldn't call a Cray a lesser unix box.

Having used Cray stuff, I would! Its too specialised. What Solaris/Sparc is great at is general purpose scalability, both on SMP and NUMA. SGI has always been keen on scalability for specialised purposes, such as numerical work (I have used their machines for molecular modelling). SGI etc. is fun stuff, but you would not want to try to scale your web server on it.

Every Unix besides SCO supports that stuff at this point,

Well yes, but Linux doesn't. So, an open source Solaris should be something worth looking at?

But ultimately you have to trust one company with the evolution and future viability of the language.

No you don't. You are free to implement your own VM (as many have - HPs VM is clean room). If Sun 'stopped' Java, many companies would continue with the development of the language, but it could not be called 'Java' - that's all (unless, as would be likely, IBM bought it). After all, that's more or less what Microsoft have done with C# - its nothing more or less than MS 'Java', designed because MS did not like Sun's licensing. Java is backed by tens of large companies.

Re:What is Open?

[joib]

I would have thought that their SMP hardware would be interesting to geeks - unlike the compromised NUMA architecture that lesser Unix boxes run.


As others have said, the current generation of big iron Suns (the Sun Fire) is NUMA. You can find details in

Alan E. Charlesworth: The sun fireplane system interconnect, Proceedings of SC2001 (available online if you have an ACM subscription).

While Suns combination of snooping and directory based cache coherency is neat, it nothing revolutionary or so, that other companies would somehow be unable to implement if they wanted to.

If Sun tanks, the world will go on just fine without them.


Solaris should definitely be of interest to anyone interested in UNIX or Linux - unless features such as partitioning and scalability are dull?


For the vast majority of users who have no need for partitioning nor extreme scalability, yes those features are pretty dull. And those who need those features know that Sun isn't the only game in town.

If Sun tanks, the world will go on just fine without them.

Re:What is Open?

[Decaff]

Do you take me for an idiot? Scalability is exactly that. By definition.

No offence intended! I have seen (and used) many systems where software was placed on large arrays of processors, but made poor use of them. Its like putting a go-kart on a 5-lane motorway.

Re:What is Open?

[Daniel Ellard]

That's why they are pushing Linux on their low end Opteron stuff, ...

Opteron-based N-way blade servers (I'm not sure what values of N have been announced yet, but it's not 1...) aren't exactly low-end, but yes, they're for different markets.

Re:What is Open?

[pdcruze]

I think what he means is:

Linux + IBM Power5 hardware == few commercial applications available
AIX + IBM Power5 hardware == many commercial applications available

The question then becomes, will adopters of these boxes be happy to use open source software or can IBM convince enough commercial developers to port their software to it.

Re:What is Open?

[Decaff]

As others have said, the current generation of big iron Suns (the Sun Fire) is NUMA. You can find details in

I know. The point was that Solaris is very good at scaling over SMP as well, with a highly multi-threaded kernel. Whatever the architecture, a kernel which has been tuned for years to handle very large number of processes and threads is a benefit (otherwise, only specialised software can really make use of those processors). I assumed that this kind of tuning and high performance would be of interest, and not casually dismissed with the attitude 'Linux is starting to do this so we don't care about Solaris'.

If Sun tanks, the world will go on just fine without them.

This was not the point. The point was that Sun and Solaris and Sparc etc. weren't interesting.

But, anyway, considering Sun has contributed so much to open systems and pioneered the use of Unix as both a server and workstation system for more than 20 years, I find the typical slashdot hostility to them very strange. I guess its because they aren't quite 'on message' with open source and GPL - the don't yet follow the 'true religion'.

If Sun 'tank', it will be a great loss, as one of the real innovative software and hardware companies will have gone. Sun played a major role in holding off Microsoft in the server market during the 90s.

Re:What is Open?

[Anonymous Coward]

no-one but IBM "does" POWER.

Not exactly...if you are in the HPC market. Hitachi uses POWER4 w/AIX and even Federation (IBM's flagship cluster interconnect). The difference between the two is that Hitachi's systems offer more memory bandwidth than a Regatta.

Re:What is Open?

[oldmanmtn]

The IBM Open chip is not a chip without an OS. It runs linux...a commodity OS. That means two major things.

I read that to mean that it's a chip that doesn't have an OS 'based' on it. Linux runs on it, but Linux's base hardware is x86. Most of the Linux developers do their work, testing, and tuning on x86, so that's where it will run best. It will run on many other platforms, but they aren't going to get the same level of attention.

Similarly, Darwin is native to PPC. There is an x86 port, but the PPC implementation gets most of the work. Same with Solaris and SPARC (although, it seems like Sun is in the process of making both SPARC and x86 "native platforms" starting with Solaris 10).

IBM's new processor doesn't have an OS that is native to it. AIX doesn't run on it at all. Linux will run on it, but I can't imagine it will get the same attention as x86, amd64, PPC, or even MIPS.

What I don't get: if you want a low-end Linux machine, why wouldn't you buy an x86 or amd64 box? What is the incentive to buy a proprietary hardware platform just to run an open source operating system? It's bizarre. I'm with Jonathan on this: I just don't understand what IBM is thinking.

People not happy with big blue can migrate to another vendor without having to take an OS change into account. That means less lock in.

This doesn't follow at all. Nobody is going to buy these Power machines to run Open Source software. They'll run an Open Source OS (OS^2?), but they'll be running IBM's proprietary software. If you want to run Apache and/or MySQL, you'll go with cheaper, easier-to-find, and easier-to-maintain x86/amd64 hardware. If you want to run WebSphere, maybe you'll look at these low-end Power machines. Once you've done that, you are completely locked in: hardware, OS, and application.

Re:What is Open?

[Anonymous Coward]

The point J.S. is making is that the new servers don't run any version of Linux, they run RedHat Linux. Was that not obvious when he points out about history? Remember Microsoft and their OS?

Now if IBM's new servers ran multiple IBM *supported* Linux distro's that would be different.

As for lock in, do you honestly believe there is no lock by using RedHat? Any feature that RedHat has that other distro's don't have is a potential lock-in. If you've never done a port from one UNIX to another it doesn't take much to make things non portable.

Of course if Linux is the only OS you've ever used well then believe what you want.

Re:What is Open?

[KZigurs]

...and CTO's get fired when linux starts to show all of it's problems: "fuck it, this bug is unimportant", "who cares about optimisations anyway", "what do you mean - thread synchronization?".

In my eyes IBM offers chip without an os. Linux is far, oh so far, from proper unix. Just as far as averange linux fan from fucking a real girl ;D

Re:What is Open?

[LizardKing]

Sun's Opteron is high end for most Slashdot readers (who think a computers performance is measured by how well it runs Doom 3), but compared to something like an E25K they're distinctly entry level ...

Re:What is Open?

[jeif1k]

Java 'Highly proprietary'? Ah - I guess that is why the spec is published,

The Java spec isn't "published", it is available only under license. It just happens to be the case that it is dangerously easy to accept the license conditions. But the license is printed for you to read on Sun's Java specs, as well as the specs produced under the Java Community Process, and it states that the spec is owned by Sun and that you cannot implement it unless you meet specific conditions.

and why GNU can implement Java;

You can read what the GNU folks have to say about Java here. It's no accident that RMS calls Java "a trap". Besides, GNU hasn't really implemented Java, and just because they haven't been sued yet by Sun doesn't mean that they aren't in violation of lots of licenses, copyrights, and patents.

why Java is the most in-demand language for IT jobs,

So? Is Windows "open" because there are lots of job postings for it?

There is nothing proprietary about Java, only the name, which you must pass compatibility tests to use.

Many Java specs are proprietary and available only under license, and in addition Sun has lots of patents on parts of Java. That makes the Java platform proprietary.

Without these compatibility tests, Java would have fragmented years ago.

The irony is that Java is not only proprietary, it is already fragmented. One dialect of Java, namely C#, is just different enough to try to work around all of Sun's license requirements and patents. And all the attempts at open source implementations are wildly incompatible.

Basically, with people like you, I can't figure out whether you are just too lazy to read Sun's licenses, or whether you have a stake in Java and are lying through your teeth. But the truth is easy enough for anybody to see: just read the licenses at the beginning of the Java and JCP specs and search on the USPTO site for Sun patents related to Java. Or just read RMS's article.

Use Java if you like. But don't keep pretending that it's an "open" platform. Using Java is just like using Windows: you are using a single-vendor, proprietary platform.

Re:What is Open?

[jeif1k]

No you don't. You are free to implement your own VM (as many have - HPs VM is clean room). If Sun 'stopped' Java, many companies would continue with the development of the language, but it could not be called 'Java' - that's all (unless, as would be likely, IBM bought it).

People keep claiming that, but it is just not true. Sun's ownership of Java is not just ownership of the name. Sun's specs are available only under specific licenses, and, in addition to their ownership of the specs, Sun has patents covering aspects of Java implementations.

HP's clean-room implementation of a JVM (and just a JVM!) tells you nothing about whether their JVM is entangled in Sun patents or licenses.

After all, that's more or less what Microsoft have done with C# - its nothing more or less than MS 'Java',

Yes, but while C# looks and works like Java, it is not Java compatible, so it is of little help for companies or open source projects stuck with millions of LOC based on a proprietary platform.

However, in the long run, we can hope that Java source code translators or on-the-fly bytecode translators can avoid Sun's patents and licenses while still allowing people to migrate their legacy Java code to an open platform.

Re:What is Open?

[Decaff]

Many Java specs are proprietary and available only under license, and in addition Sun has lots of patents on parts of Java. That makes the Java platform proprietary.

No. Many parts of Sun's implementation of Java are patented and proprietary. This does not apply to clean room versions of Java (like HPs), or Kaffe, or many other Java implementations.

However, if (like GNU) you don't want to use Sun's code, you don't have to - that is why there is the GNU Classpath project, to develop a Java that fits the GNU definitions of 'free and open', showing that the Java platform definitely need not be proprietary.

As for C#- sure, its kind of a Java dialect, but as its not called Java, you know it is not compatible, and you don't get to run your Java code on it. This is proof of Java NOT fragmenting.

Use Java if you like. But don't keep pretending that it's an "open" platform. Using Java is just like using Windows: you are using a single-vendor, proprietary platform.

Java is not single vendor - that is its huge strength. I can get it from IBM, HP, Sun, SGI and many others.

As for quoting RMS - as much as I admire him, he has his own political axes to grind, and his own definitions. He was not the first to coin the term 'open'.

And yes, I have read the licences. I just honestly disagree with you.

Re:What is Open?

[Decaff]

People keep claiming that, but it is just not true.

Its not people who claim that you are free to implement your own VM - its Sun - James Gosling, for example.

HP's clean-room implementation of a JVM (and just a JVM!) tells you nothing about whether their JVM is entangled in Sun patents or licenses.

Actually, it does. That's why they did it. That is why there are several clean-room implementations of the VM (e.g. Kaffe) and the libraries (e.g. GNU Classpath).

Yes, but while C# looks and works like Java, it is not Java compatible

That is not the point. The point is that companies can implement Java-like languages without legal problems. You can even write a Java VM if you like.

it is of little help for companies or open source projects stuck with millions of LOC based on a proprietary platform.

I don't understand the use of 'stuck with' here. It implies some sort of problem, as if the lines of code will suddenly vanish or stop working.

However, in the long run, we can hope that Java source code translators or on-the-fly bytecode translators can avoid Sun's patents and licenses while still allowing people to migrate their legacy Java code to an open platform.

You may hope this, and - strangely - term Java 'legacy', but I - along with hundreds of thousands of fellow developers - am happy to use a free language that is guaranteed compatible and binary portable across all major OSes. If an 'open' platform turns up that has the same degree of power, range of deployment platforms, and multi-vendor support, I would be happy to use that too.

Re:What is Open?

[jeif1k]

Its not people who claim that you are free to implement your own VM - its Sun - James Gosling, for example.

What Gosling says is not legally binding on Sun. In fact, it is a huge problem that Schwartz and Gosling and people like them say one thing ("Java is open", "you can implement it", etc.) and their lawyers write something rather different into their license agreements and go off patenting one piece of Java after another.

"it is of little help for companies or open source projects stuck with millions of LOC based on a proprietary platform."

I don't understand the use of 'stuck with' here. It implies some sort of problem, as if the lines of code will suddenly vanish or stop working.

Yes, they will: when Sun stops shipping Java for my platform or starts charging an arm and a leg for it, I will have no other runtime to run my code on.

Furthermore, users have no control over what Sun actually puts in their implementations. Look at Sun's own bug parade: Sun just isn't responsive to user feedback or wishes. On more than one occasion, a new JRE release has seriously broken Java software we have developed.

That is not the point. The point is that companies can implement Java-like languages without legal problems.

Of course, people can implement "Java-like languages"; after all, it's technology that predates Java by many years. But implementing a "Java-like language" does not solve the problem that the only way you can run Java is by using a Sun proprietary Java implementation.

Actually, it does. That's why they did it. That is why there are several clean-room implementations of the VM (e.g. Kaffe) and the libraries (e.g. GNU Classpath). [...] You can even write a Java VM if you like.

You can also rob a bank, and you may even get away with it. But the fact that you can do it doesn't demonstrate that it is legal to do so.

HP has implemented a clean-room VM. They either have to have patent licenses from Sun to do so, or their implementation is not fully compatible, or they are infringing Sun's patents. Kaffe and Classpath probably violate both Sun's license agreement for the specifications and Sun's patents (Sun just has chosen not to sue them for now--Kaffe is no threat to them because it works so poorly and a lawsuit would be a PR nightmare for Sun, destroying Sun's carefully crafted myth of openness and friendliness towards open source).

If an 'open' platform turns up that has the same degree of power, range of deployment platforms, and multi-vendor support, I would be happy to use that too.

I have no problem with you using a proprietary platform if it serves your needs. What I have a problem with is when you and others (in particular, Sun representatives) falsely claim that Java is "open" or that it can be "freely implemented". In addition to Sun's ownership of the Java trademark (which is mostly harmless), Java is covered by patents, and the Java specifications can only be implemented under specific licenses (which you can find at the beginning of each specification).

Re:What is Open?

[jeif1k]

No. Many parts of Sun's implementation of Java are patented and proprietary.

The patents look essential to building compliant Java implementations. If you try to work around them, your implementation will probably not pass the certification tests. But the burden of proof here is on Sun: if Sun wants to show that Java is an open platform, they have to demonstrate that it can be implemented without violating their patents.

Java is not single vendor - that is its huge strength. I can get it from IBM, HP, Sun, SGI and many others.

Java does come from a single vendor; IBM, HP, SGI, Blackdown, and all the other compliant Java implementations all rely on code licensed from Sun. That means Java stands and falls with Sun--if Sun decides it's bad for their business to ship, say, on Linux, they can kill Java on Linux.

And yes, I have read the licences. I just honestly disagree with you.

The very fact that there are licenses on the specifications means that your claim that the specs have been "published" and "anyone can implement them freely" are false. So, you admit that I'm right, you just choose to ignore that fact because it doesn't fit into your Sun-friendly world view.

As for quoting RMS - as much as I admire him, he has his own political axes to grind, and his own definitions. He was not the first to coin the term 'open'.

And Sun has a political axe to grind, too: they want to maximize profit by any legal means available to them, and one great way of doing that is to tie users to their proprietary platforms. And Sun invests millions of dollars in crafting PR messages, which right now happen to be centered around being "open source friendly" and "the alternative to Microsoft". In fact, Sun has taken more than a page out of Microsoft's playbook.

So, we have a financially disinterested third party (RMS) plus our own eyes when looking at Sun's actual licenses, and we have a company (Sun) whose financial future depends on owning some piece of proprietary technology and getting it to be widely adopted by whatever legal means they have. I trust RMS and my own eyes a lot more than I trust Sun's self-serving PR messages.

Re:What is Open?

[JonAnderson]

Actually AIX 5.3L (required for advanced features of p50 = few commercial applications available.

Re:What is Open?

[JonAnderson]

This is one of the first sane posts I have ever seen on slashdot. 7 thumbs up!

Re:What is Open?

[Decaff]

You can also rob a bank, and you may even get away with it. But the fact that you can do it doesn't demonstrate that it is legal to do so.

So, you believe that GNU is breaking the law?

Yes, they will: when Sun stops shipping Java for my platform or starts charging an arm and a leg for it, I will have no other runtime to run my code on.

Firstly, there are plenty of non-Sun VMs, secondly, this is an extreme point of view that should stop you using just about anything.

Your PC is patented. Your car is patented. Do you use a GNUTV? Do you play computer games? Are they 'open'?

I think I see where you are coming from, but there is a big world out there away from open source - most software and products don't confirm to the GPL. However, the IT industry has been working and people have been using software and developing products for decades before GNU and the GPL. I like the ideas behind the GPL, but that doesn't mean I have a religious insistence that every product I use conforms to it, and it doesn't mean that I run scared from anything that does not.

Re:What is Open?

[Decaff]

So, you admit that I'm right, you just choose to ignore that fact because it doesn't fit into your Sun-friendly world view.

No, I don't admit you are right.

Java does come from a single vendor; IBM, HP, SGI, Blackdown, and all the other compliant Java implementations all rely on code licensed from Sun.

No. This is not true. There are many compliant implementations (such as HPs) that are clean room.

That means Java stands and falls with Sun--if Sun decides it's bad for their business to ship, say, on Linux, they can kill Java on Linux.

No. IBM have non-Sun VMs internally, and ship Java for Linux (which does not use Sun's Java libraries - no com.sun.*).

And Sun has a political axe to grind, too: they want to maximize profit by any legal means available to them, and one great way of doing that is to tie users to their proprietary platforms. And Sun invests millions of dollars in crafting PR messages, which right now happen to be centered around being "open source friendly" and "the alternative to Microsoft". In fact, Sun has taken more than a page out of Microsoft's playbook.

What a cynical view!

Look - Sun has a history of an open strategy. Its a strategy that works like this: If you are good at a technology, its good to make sure that that technology is widely used, even if that means that other companies also supply it. If you are a UNIX provider, as Sun is, it makes sense to ensure that there is a large market for UNIX out there. This is why Sun (almost alone) encouraged 'open systems' in the 1980s - encouraging the use of TCP/IP-based networking, and providing open APIs such as NFS to the community free of charge. The same applies to other technologies: Sun have ensured that their Sparc chip is multi-vendor - you can buy it from Fujitzu, for example. Java is also multi-vendor - companies can either develop their own VMs and pay for compatibility tests, or licence code from Sun. Having Java as the de-facto language for server-side development allows Sun to make a considerable amount of money selling software services using Java.

You CAN provide free technologies and APIs to the community AND make money from it - Sun pioneered this decades ago when companies like IBM and HP were trying to tie users into proprietary systems such as VM/CMS and VMS.

Comparing Sun to Microsoft is lazy and shows an ignorance of history. Its not even bad history, as no matter how bad Microsoft are, IBM (now 'good guys') was just the same decades earlier!

Not that great..

[jjeffrey]

It only helps if you have a very parallel application that is enormously heavier on CPU cycles than on disk or anything else - there aren't many like that.

Also, I think it's quite funny that blogs.sun.com seems to have been slashdotted...

Re:Not that great..

[mikael]

Most of Sun's application software vendors (CAD/CAM/animation) were busy porting their applications over to multithreaded mode back in the mid 1990's.

Even if an application isn't multithreaded, the window system may very well be. And you can always have all those background processes (clock, TCP/IP, X-server) running on different CPU's, leaving at least one CPU free to run your application without swapping/scheduling out memory.

Re:Not that great..

[zerocool^]

*ahem*

Database applications come to mind. Granted, you'd have to have some pretty nice high end hard drive configurations, but that's par for the course - if you look at any enterprise solution, they have all the front end application servers, which are generally toned down, and then they have the database server, which is basically as powerful as possible.

Also, given fiberchannel or large disk array solutions, I can think of a lot of applications this would speed up. Anything that spawns new processes when accessed comes to mind - qmail for a university for instance, or webserving for a high volume site. Which is funny.... blog.sun.com...

~Wx

Re:Not that great..

[Daniel Ellard]

It only helps if you have a very parallel application that is enormously heavier on CPU cycles than on disk or anything else - there aren't many like that.

Too bad TFA is slashdotted... the point is not parallel applications, it's applications with lots of threads. If you're putting together a service nicely decomposes into lots of threads (i.e., lots of httpd's) then this is very nice -- lots of threads running side by side, no context switches...

Re:Not that great..

[mestlick]

Each of those 32 threads is not going to be very fast.

These chips will not be used for databases or scientific computing where you would like high single thread performance.

They will be used for web servers and the like, where you have lots of users hammering on one server.

Re:Not that great..

[timeOday]

forget CAD, this chip is very weak for floating-point.

Re:Not that great..

[kpearson]

There are actually quite a few applications like that. This kind of CPU is perfect for distributed computing applications, which use every CPU cycle and thread they can get. Clients based on the new BOINC computing platform, and the distributed.net client, are already set up to take full advantage of this kind of CPU.

Re:Not that great..

[davecb]

No,actually it's for any cloud of applications that run on a machine where the memory speed is much less than the cpu speed. This is true of everything save 390s (;-))

--dave

Re:Not that great..

[Anonymous Coward]

These chips will not be used for databases or scientific computing where you would like high single thread performance.

Actually, although I do agree that such chips won't be used for FP-intensive scientific computing, I disagree with the DB part. Most DB servers actually are heavily multi-threaded due to concurrency (dozens or hundreds of simultaneous requests) and thus it's not necessarily the case that the highest single-CPU performance is a must. It helps, but you can scale both ways.

I mean, since more and more DB servers are used solely serving data for web servers, it should be obvious that if and when web servers can greatly benefit from parallelism, so do DB servers...

something for google?

[kc_cyrus]

Google's architecture might play well with this design with lots of processors in a dense package with relatively good power efficiency per processor.

Re:something for google?

[agby]

Google's architecture might play well with this design with lots of processors in a dense package with relatively good power efficiency per processor.

It would be interesting to see if google would benefit from this, but their system is a grid computer built from commodity hardware, so is most likely optimised around that rather than any single powerful nodes (apart from the obvious benefit of more powerful individual nodes)

At the end of the day, I suspect it cheaper to plug in 50 more nodes than replace swathes of nodes with expensive, proprietary hardware.

Re:something for google?

[mikrorechner]

One of Google's advantages over their competitors is their own linux-based cluster-ish software that allows them to use cheap commodity PC hardware to run their business. I don't think they would change that and buy (relatively) expensive Sun hardware without a *very* good reason.

Re:something for google?

[Anonymous Coward]

Power, cooling, maintenance (yes, google's hardware requires maintenance too), housing, networking fabric, and "parallelizability" limits of the problem all suggest smaller, faster, and more efficient hardware. There will be a point where the money balances.

We're supposed to be impressed?

[xyote]

Sun has never explained or shown what this Throughput computing is all about. More multi-processors. Yeah, so? You need concurrency mechanisms to exploit it. Pthreads by itself isn't going to hack it. They won't scale up. Even if Sun has "parallelized" Solaris, it's in user space where most of the processing is done and where most of the problems will occur.

Re:We're supposed to be impressed?

[Anonymous Coward]

And if one of things that you do is Apache then how many different processes (or threads in the 2.x line) are spawned after you start up?

If you think of a traditional app as in mutt or Mozilla then yes, you will probably have to do some code changes. However, if what you're doing involves many daemons or instances of Java (which Suns are used a lot for), then you have parallelism built-in since you have many external requests.

Re:We're supposed to be impressed?

[haggar]

But aren't those the same problems one faces when writing apps for a multiprocessor system? And yet, you are still impressed if you get an 8-way system for the prce of a 4.

Heck, just to have an 8-way system in the SPACE of a 1-way, is very damn impressive. But of course your apps have to be written to take advantage of them.

Re:We're supposed to be impressed?

[flaming-opus]

Sun sells 144-core systems now. Quite a few of them, in fact. Just becuase doom3 doesn't scale to dozens of processors doesn't mean that real world workloads don't. Web serving, transaction processing, mail servers. These things parallelize very well on an SMP.

Tried it, deployed it, been using it for years.

Re:We're supposed to be impressed?

[guacamole]

Sun sells 144-core systems now. Quite a few of them, in fact. Just becuase doom3 doesn't scale to dozens of processors doesn't mean that real world workloads don't. Web serving, transaction processing, mail servers. These things parallelize very well on an SMP.

Maybe they do though I think it is a pretty dumb idea to buy a propietary SMP box to run "commodity" services like mail or WWW. Using clusters of cheap x86 based servers scales just as well for those applications and costs a whole lot less. Companies that buy large unix SMP machines usually do so to run their enterprise applications on them or for massive number crunching.

Re:We're supposed to be impressed?

[xyote]

I'm talking about things like databases and file servers which use lots of threads sharing large common data structures. Conventional synchonization doesn't scale very well no matter how well you parallelize the application. Sun isn't doing anything here or hasn't said what they're up to.

The only system that anyone may be doing something on is Linux. But nobody has publicly announced anything there either.

I do have some lock-free algorithms that are portable but those aren't in widespread use. In fact, I did three different versions of RCU for preemptive user threads on Solaris just for the heck of it. Under deliberately severe contention, they really cranked, even on a single processor SB100. Worst case for RCU still did better than mutexes. Worst case for mutexes were ctl-C'd to avoid waiting forever vs. almost instant for RCU, a dramatic difference. The only thing that did better was lock-free reference counting but I wouldn't expect that to scale as well as RCU with increasing numbers of cpus.

But so far I haven't heard from Sun offering me one of these 32-ways to play on. I don't think they're serious. Talks cheap, Sun.

Re:We're supposed to be impressed?

[davecb]

Yes, if you're a nerd (;-))

The current problem in compuring is that memory speeds are going up far slower than processor speeds, causing huge cache-fill delays. Sun came up with a simple architecture to keep the processors running anyway, and it is compatable with multiprocessing and multithreading:

1. Run decoder A until cache blocks on a read
2. Clear ALU and switch to decoder & register file B
3. Run B until cache blocks on a read...

.. and so on for C-F. then go back to A. Put two ALUs and two sets of 8 decoder/register sets, so as to use the whole of the current memory interconnect bandwidth.

Given this much raw compute power from the same size (and price-range) silicon, the marketplace will rapidly multi-thread or at least multi-instance their programs. They've already done the latter to run on Beowulf clusters, after all!

--dave

Re:We're supposed to be impressed?

[Anonymous Coward]

You need concurrency mechanisms to exploit it. Pthreads by itself isn't going to hack it. They won't scale up. Even if Sun has "parallelized" Solaris, it's in user space where most of the processing is done

It all depends on what you're doing with the machine. If you're trying to run a desktop and your goal is to make Mozilla and OpenOffice run fast, then yeah, you'd have to come up with some pretty kick-ass technology to auto-parallelize things. But Sun is aiming more at people who run database servers or http servers. Software like, say, Apache 2.x is poised perfectly to take full advantage of this. With a system like this, you could easily run a high traffic web server and put an SQL database server on the same box and get really great performance. You'd have a lot of CPU work (especially if you are running SSL on your web server, since that really does take actual computation), and you'd have a lot of I/O. And my guess is, these machines combined with Solaris are going to be excellent for I/O. When the Solaris kernel gets an interrupt, it handles it by saving the info and scheduling a thread to run, then it goes right back to whatever else it was doing. With so many processors, there will virtually always be a free processor to handle any interrupt immediately without impacting any other work.

Re:We're supposed to be impressed?

[fred fleenblat]

Solaris, both in the kernel, and in user space, has been fully multithreaded for years, i.e. since Solaris 2.1 or something.

What's been happening is that gradually, the kernel threading has adopted more fine-grained and performance-oriented locking. The first solaris versions 2.0 and 2.1 were indeed kernel multithreaded, but mainly for network and disk i/o. Other activities just got coarse-grained locks that basically held up the whole system for some obscure activity (heh, like fork). More recent solaris's are much better at only locking the activities that truly conflict, and locking them for the shortest time possible and amongst the fewest threads possible. Year after year, the solaris kernel has been scaling up to higher CPU counts, faster than linux has been able to and so far beyond windows it's not funny. It is entirely appropriate to leverage the multithreading capabilities of their OS by introducing hardware that is up to the task.

That's the kind of thing that should be in Jonathon's blog.

Re:We're supposed to be impressed?

[Anonymous Coward]

You need concurrency mechanisms to exploit it.

Like, maybe, those new-fangled Web server thingies? That serve many simultaneous requests? And database servers that they access to output pages for the requests?

Hello? There's plenty enough things such servers are good for; even if they wouldn't be optimal as your work station.

... and somehow there were people who thought your comment was insightful... :-p

Re:We're supposed to be impressed?

[cartman]

More multi-processors. Yeah, so? You need concurrency mechanisms to exploit it... They won't scale up. Even if Sun has "parallelized" Solaris, it's in user space where most of the processing is done and where most of the problems will occur.

I'm relatively sure Sun has thought of this already.

They're intending this to be used for commercial workloads, like Oracle. Oracle already has all kinds of fine-grained locking mechanisms and scales relatively linearly up to hundreds of CPUs, and has for years. A great many other applications in user space can already exploit many CPUs. This is a software engineering problem that was solved when SMP machines started coming out in '91. This problem is totally independent of Sun's Niagra chip.

Software licensing

[Mark Round]

What will be interesting is how the software market adjusts to these multi-core processors becoming more widespread and popular (particularly with dual-core Opteron on the way). They're going to have to rethink things a bit with regards per-processor licensing. From what I recall, Oracle (and many others) consider a dual-core processor two separate processors, and charge accordingly. Anyone running one of these chips would then get stung for a 8 (or possibly 32) processor license.

Perhaps a better solution would be to adopt the approach taken by IDC (which Sun obviously seem quite happy to back) of counting processor sockets, instead of cores.

Anyone know what other software companies are planning on doing with their per-processor licensing ?

-Mark

Re:Software licensing

[qwijibo]

Looking at processors like these, they're aimed at the server market. The people who can buy this hardware can afford the licenses. If one of these processors can compete with a 32 processor Sun Fire 12K in a functional sense, do you think Oracle is going to want less money? A count of "processors" presented to the system is as good a way of scaling licenses as any other in the server market.

Licensing probably won't change until everyone has multi core processors. Even then, licensing per "computer" without counting processors is still viable for desktop applications. Licensing is just income. Nobody cares how they do it, they just want to make sure they get as much of it as they can.

Re:Software licensing

[elodan]

So partition it.
I've played with a Sun box running dual core processors, and each core seems to be roughly equaivalent to a normal CPU. So the Oracle licensing scheme DOES make sense.

The solutoin is to use the built-in software capability to partition into processor sets, and lock applications to processor sets.

So you might have a 32-core box, but only 4 of them are running Oracle.

Re:Software licensing

[Spoing]

1. Looking at processors like these, they're aimed at the server market.

Yes, though only in a specific way. The lag on most servers isn't CPU it's disk I/O. Having so many cores on the same package would make short work of some calculations though mostly for financial, military, nuclear, other sciences.

In a typical company, more CPU girth is wasted unless you find a new application for it; server-side execution of client programs (Sun's own Java Desktop Linux distribution) and server consolidation come to mind. This may require a management change as much as a server room change.

Re:Software licensing

[Wesley Felter]

Why is it better to charge by socket? Just because it's cheaper? If your only goal is to reduce licensing cost, just charge per machine.

Re:Software licensing

[oldmanmtn]

If one of these processors can compete with a 32 processor Sun Fire 12K in a functional sense, do you think Oracle is going to want less money?

It can't compete with the 12K in _any_ sense. It has a tiny fraction of the memory capacity, the memory bandwidth, the I/O bandwidth, the cache, the TLB, the floating point performance, etc.

This might be a great CPU for first-tier web serving, and maybe second-tier app-server kinds of processing, but it's not even a remotely viable candidate for the third-tier database work that Oracle cares about.

Re:Software licensing

[fred fleenblat]

You often get a choice (numbers made up):

* site license for $50,000
* per-cpu license for $2,500
* per user license for $200

The vendor adjusts the numbers so that they scale roughly the same in typical installation scenarios, and there is a minimum order size when choose per-user so it often works out pretty close no matter which option you choose. If one option is significantly cheaper, obviously that's what the customer will choose, but then the vendor holds back on the 30-70% discounts that are common in the enterprise software industry.

So, getting back to your point, licensing is only per-processor or per-whatever as a way to get a dialog going with the customer. The sales force is free to offer huge discounts, or not, or make up new licensing schemes, or take the VP's golfing, whatever the heck it takes to make their numbers. Little things like "licensing issues" are only an obstacle in press releases.

Re:Software licensing

[mark_lybarger]

If one of these processors can compete with a 32 processor Sun Fire 12K in a functional sense, do you think Oracle is going to want less money?

why should one have to pay for software based on how much they're going to use it? that's like saying, you drive further than normal to work each day, so your car is going to cost 120% of the retail value. i only drive 20 miles per week, so mine is 60% retail.

oh yes, i remember software isn't a tangible asset, you "license" the right to use someone else's property. so, it's more like leasing a car where i have to pay extra if i go over on miles, or even pay per mile. but then again, companies don't lease oracle software do they? the buy the license outright (hundreds of thousands of dollars). what they lease is the support. annual maintenance fees. then we're back to the model of buying a car and paying for anual maintenance. but still being restricted to how many miles you can go.

this stuff is really really strange stuff these software licenses.

SPARC PLUG? or new arch?

[Anonymous Coward]

Can't get to the site to read details.

Is it /. factor or is it Gates and gang?

One nice thing I've heard, through the grapevine, about this chip is that the 'jumbo' patch only takes 4 hours.

Solaris is pretty cool too. With windoh's I can only have one user logged on at a time. Solaris lets me have 5 and even more if I want to pay extra. This is still a better deal than having to buy a sheetload of windohs boxes as it uses a lot less juice and it gets damn hot here in the summertime.

I ax u. where the mirror?

solaris 10 predictive self healing

[sl4shd0rk]

After upgrading the Sun server on our network to Solaris 10, all of a sudden the Exchange server stopped working, our Primary Domain Controller went tits up, and the W2K DHCP Server went offline. I've gotten six phone calls in the last 10 minutes from people calling to ask why their workstations say "Welcome to Looking Glass" when they log in. ;)

Re:solaris 10 predictive self healing

[sharkey]

I've gotten six phone calls in the last 10 minutes from people calling to ask why their workstations say "Welcome to Looking Glass" when they log in. ;)

Because you ate the wrong pill. Remember, keep your hands in your pockets when in Wonderland.

Blog ? Try SPARC plug

[Anonymous Coward]

He just goes on and on.

Wow... maybe SPARC will work wonders in my coffee machine?

*bump*

Re:Wow

[Xpilot]

It says "Niagra chip", not "Viagra chip".

That's just 31 more ways the machine can fail...

[march]

That's just 31 more ways the machine can fail... Have you noticed how the reliability of Sun's hardware has really fallen off in the past few years? I have an SS2 from '95 that is still runnnig (a seti@home client) and has not failed a single time from hardware (other than the initial Quantum disk going dead). At work, there is a hardware failure almost every day. Granted, we have a lot of Sun's, but still...

Benefit only up to a point

[xyote]

Normal threading techniques don't scale very well once you start getting lots of processors. That's why SGI had to do a lot of work to get Linux to run on a 100+ way. You need to develope new synchronization mechanisms like RCU to deal with the scalability issues.

Do you see Sun working on new synchronization mechanisms to deal with scalability issues? No.

Re:Benefit only up to a point

[Anonymous Coward]

The point isn't really to build boxes with a very large number of CPUs but to provide modest SMP-type capabilities on a single chip. Being on a single chip has potential advantages with regard to cache sharing and also lower power and motherboard real estate requirements than a number of single core chips on a motherboard.

These multi core sparcs seem to be aimed at application serving and the like where you may need to have a high density of CPU cores but a series of single CPU or SMP boxes will do the job. Modestly sized SMP boxes are attractive as the total number of OS installs, power supplies, and other error prone components as compared to N boxes is lower. If the cores are all on one chip and save you space over a 4 CPU SMP system then you can fit more in the same rack space. If they are lower heat output than a 4 CPU SMP system again you gain.

Re:Benefit only up to a point

[acsinc]

Do you see Sun working on new synchronization mechanisms to deal with scalability issues? No.

I dont think they have to much to worry about in that department this beast has way more than 32 processors.

Actually...

[Anonymous Coward]

Modern POWER processors conform to the PowerPC specification, they're only called POWER for marketing reasons.

Motorola make PowerPC chips too, the last I read.

Re:That's just 31 more ways the machine can fail..

[HalfFlat]

But you have to remember, the Sparcstation 2 was pretty special. One of the best workstations produced ... ever.

Sun may still have good server hardware -- it's been a while since I've had to deal with it -- but as far as workstations are concerned, I don't think they've ever matched what they achieved with the SS2, compared with the contemporary competition.

And it even looked good.

So much cores...

[jcdr]

...to make there blogs display faster!
Anyone have a mirror of the picture on a static page? Thanks in advance...

May the Schwartz be with you.

[Anonymous Coward]

Will it be available as an M-bus module?

32 threads...

[Anonymous Coward]

...all running slower than a 386 per thread. I can't wait!

Seriously though, this is cool. But I think this is taking things to the extreme, the bottom line is number of IC per mm^2 and effeciency win at the end of the day.

Sun's track record at producing complicated chips, or even chips on time is lackluster. The cost for their boutique chip and it being late will probably kill them in the long run.

It remains to be seen, but my guess is that each core will perform at a level that is far from industry leading, and the aggragate performance doing "real work" like serving web pages or running apps will be less inspiring than dual/quad based servers from Intel/AMD/IBM with their respective chips.

8 cores?

[haggar]

That sounds kinda lot. At least, compared to what Intel AMD and IBM have achieved (or are going to). IBM seems to be in a better position than the other two big chipmakers, but nowhere near 8 cores. And These cores are not some simple, transputer-ish processor implementation. They are quite complex 64-bit SPARC cores (but probably with a much shorter pipeline).

OK, I think this is impressive.

Now, who's going to fab this baby? TI?

Re:8 cores?

[Anonymous Coward]

Quite complex? Well, each core in itself is very slow. At UltraSparc 1 level in performance.

slashdot start using/posting mirrored links

[Anonymous Coward]

for example

http://blogs.sun.com.nyud.net:8090/roller/page/j on athan/20040910#the_difference_between_humans_and

http://blogs.sun.com.nyud.net:8090/roller/page/jon athan/20040910#the_difference_between_humans_and

would be good if slashdot story posters could finally use the nyud.net:8090 serive or other similar services to avoid slashdotting...

Not for PeeCees

[turgid]

Doesn't mean a damn thing unless software is written to take advantage of it.

This is a SPARC processor. It runs Solaris. The Solaris kernel is fully pre-emptively muti-threaded. Most of the large applications that you buy a big Solaris box to run are also highly mutlithreaded.

The beauty of this design is that there is already a mature, stable and high-performance industry-standard OS for it (Solaris) along with thousands of applications.

You could even probably run Linux on it if you wanted.

Re:Not for PeeCees

[Kehvarl]

yes but can it run... oh.. well how about BSD?

Re:Not for PeeCees

[Anonymous Coward]

netbsd runs on sparc and sparc64 - I'm sure they will port it to this chip too

C10K

[cmaxx]

I think they're particularly looking at things like the C10K problem (http://www.kegel.com/c10k.html).

The new Solaris 10 networking code reputedly pays a lot of attention to exploiting, and serving threads well, particularly hardware multithreading if it's available.

If they could squeeze one of these and maybe 8GB+ of RAM into a 1U box or into their blade centre, then I think it'd do quite nicely for serving web.

Re:That's just 31 more ways the machine can fail..

[haggar]

I have to disagree: we have about 200 Sun servers in our R&D lab, and dozens of Blade workstations. All this stuff, including the recently acquired dual Opteron box, is rock solid, no downtime machinery.

In any case, I think it's unwise to judge the total output quality of a company, based on just your particular experience, with possibly one single computer or application.

here is a double mirrored link

[Anonymous Coward]

http://www.stuwo.net.nyud.net:8090/temp/sdtemp/jon blog.html

should load fast.

software licensing

[Anonymous Coward]

wow....8 cores....just think: a quarter million dollars to license Oracle for a single CPU. Or any other major software package for Solaris that uses CPU based licensing where they count cores as CPUs.

That's nice but ..

[Anonymous Coward]

.. where's my Java Beans(TM) enabled Microwave? Been waiting for like a decade now.

Don't believe it!

[Anonymous Coward]

Sun has low quality hardware. They have been plagued by errors in their processor and system designs. By comparison, x86 processors from Intel and AMD have very few errors. Sun also uses non error corrected L2 caches! Come on, even my desktop P4 has that. Many of Sun's problems have been hidden by the strict NDA's they make people sign in order to get problems fixed, and their excellent policy in replacing parts. Neither of those diminishes the magnitude of the problems, though.

As for workstations, much of Sun's newer workstation hardware is built by the same Taiwanese OEM's as build x86 machines, to the exact same quality specs. The only differences between a cheap white box PC and a low end Sun workstation are the sticker, the processor, and the price. Reliability is exactly the same.

The only real advantage Sun has is Solaris, which is a stable and scalable operating system. If Sun makes Solaris available for Power and Itanium, there will be no reason left to buy Sun SPARC hardware except for momentum. The hardware reliability and performance already argue that something else would be a better choice.

It doesn't matter really because...

[bburton]

It is official; Netcraft confirms: Project Niagra is dying

One more crippling bombshell hit the already beleaguered Project Niagra community when IDC confirmed that Project Niagra market share has dropped yet again, now down to less than a fraction of 1 percent of all servers. Coming on the heels of a recent Netcraft survey which plainly states that Project Niagra has lost more market share, this news serves to reinforce what we've known all along. Project Niagra is collapsing in complete disarray, as fittingly exemplified by failing dead last in the recent Sys Admin comprehensive networking test.

You don't need to be a Genius to predict Project Niagra's future. The hand writing is on the wall: Project Niagra faces a bleak future. In fact there won't be any future at all for Project Niagra because Project Niagra is dying. Things are looking very bad for Project Niagra. As many of us are already aware, Project Niagra continues to lose market share. Red ink flows like a river of blood.

All major surveys show that Project Niagra has steadily declined in market share. Project Niagra is very sick and its long term survival prospects are very dim. If Project Niagra is to survive at all it will be among processor dilettante dabblers. Project Niagra continues to decay. Nothing short of a miracle could save it at this point in time. For all practical purposes, Project Niagra is dead.

Fact: Project Niagra is dying

Re:It doesn't matter really because...

[Rahga]

I completely agree. Niagra is a thing of the past in this shop.

Re:Wow

[Mysticalfruit]

When I first read the headline, I thought it said "Nigeria" not "Niagra" and all I could think of was what the boot messages would look like....

I am of great luck that I have found you in my booting time of need. ...
Before my father passed away he moved 32 MILLION BYTES of CACHE to a daughter board on the pci bus. I have contacted the pci controller, and explained your GRANT request. ...
I need you to send me a copy of your PID, your UID, your address and your IRQ.

President Bush and the Pope

[Udo Schmitz]

miracle child from the gay marriage of President Bush and the Pope

Now you implanted a really disturbing picture in my brain, that will ruin the whole week for me.

Re:That's just 31 more ways the machine can fail..

[LizardKing]

Sun's current line of workstations are pretty poor compared to the SparcStations of yesteryear. Back in the mid 1990's they were well worth the expense if you wanted a reliable Unix machine on the dekstop. My SS5 is still whirring away as a firewall and webserver at home. however, if I wanted a decent RISC workstation nowadays then I'd go for a Mac and dual boot either Linux or NetBSD on it.

Sun's servers are still excellent though, and they're my first choice for serious database machines. Alphas were a good alternative until DEC got melted down by Compaq, and I'm not really aware of what IBM market in this area (the Z Series?).

Whiny

[bhima]

Is it just me or does the whole thing sound over whiny, sort of like a Jr. SCO. Honestly I just read the text someone had in the thread so I didn't see the groovy picture claimed to be ITFA. If I was blogging about my companies products... well I guess I'd blog about my companies products and not about IBM's. did I miss something?

These are very stipped down cores...

[Anonymous Coward]

I don't want to rain on the parade but there are only 8 cores and they have been stripped down.

It's pure marketing BS to call an 8 core, 4 thread machine a 32-way system.

There ain't no free lunch. If anyone's expecting 32 complete cores on this chip they are going to be very, very, very disappointed.

...Sooner rather then later...

[mc2104dave]

Hmm, Poor Jonny..He better that that cpu out soon..
is it just me.. http://blogs.sun.com/jonathan/ errors out
I get a net timeout error on his blog... Perhaps that's the TRUE test, put one of those fan-dangled 32-way core-thingys, and post on /. and see if it can hold it's own......??

Why multi cpyu?

[marcosdumay]

Seriosly, all those companies're trying hard to put more cpus on a chip because there is an advantage on that or because of marketing decisions?
If there is an advantage, what it is? I can't see why jionning multiple microprocessors on a single chip may make a computer faster or cheaper than using mutiple (and much more cheap) microprocessors, one per chip.
Surely, there is some economy on cache memory, but the cache will not be divided between all the cpus? What is tha advantage on doing a big cache instead of a lot of smaller ones?

Re:Why multi cpyu?

[Anonymous Coward]

There are serveral reasons. One of the biggest reasons is the speed of light - wich sort of sets the limit for how fast two independent units can interact. Two cores that are verry close communicates faster. Another big reason is utilisation of resources (e.g. registers) that can be shared (no need of duplication -> more area usage -> units further apart -> slower). Most threaded processors share the same registers to save space and material.

Has anyone noticed....

[thewiz]

that the motherboard and CPU combo that he shows on his blog has no memory on it? Must be one of those magical motherboards to be running Solaris 10 with no memory.

it coulda been a contenda?

[Travoltus]

Then that severely limits its use in the market. Let's put it this way - the machine has GOT to be insanely expensive to begin with. Why not add a few more bucks to the cost and have a floating point unit (or more?) per core.

That makes the machine scale to any task. I imagine its use as a mere web server/etc. pigeonholes the server and puts it in danger of jumping the shark early on, or worse, doing a betamax.

Re:it coulda been a contenda?

[Anonymous Coward]

Actually, if you are really interested, I think Sun does have a roadmap that specifically explains the idea. One of the 'system flavours' is aimed at web/DB services (web servers, app servers etc, DB servers), where floating point performance is generally not THAT important. Then there's another set of systems that DOES require better FP performance, and for those systems, different kinds of cores are used. Ones that obviously have more number crunching muscle.

Thing is, it is GOOD to optimize for the use case, if you clearly have one. And thus it is possible that by reducing amount of FP-related functionality, you can squeeze in more cores, reduce power requirements and so on. Personally I think that does make some sense, when applied properly. There are obviously some use cases where more balanced "jack of all trades" systems would make more sense, but I would argue that more often they don't; especiallly when we talk about multi-CPU/core systems.

Re:it coulda been a contenda?

[mihalis]

Then that severely limits its use in the market. Let's put it this way - the machine has GOT to be insanely expensive to begin with. Why not add a few more bucks to the cost and have a floating point unit (or more?) per core.

I think the answer is likely to be that the two FP units on this chip can already max out the memory bandwidth when doing FP-intensive work anyway, so any more execution resources would be under-utilised.

But that's just my guess.

Actually they are not complex cores

[Anonymous Coward]

These cores have been simplified *a lot* in order to make them fit into this architecture.

Contrary to all previous Press Releases, there is still no free lunch.

Hmmm....

[tbcpp]

Imagine a beowulf clu... OKAY I'm joking!

Jonathan is useing this blog as a marketing spot.

[WindBourne]

CIO's are walking away from Unix, because it is a lockin on hardware and software. IOW, too high of cost. With OSS, you get a low cost or free OS, and true competition with hardware. That lowers the revenue on Hardware.

What is interesting is that Windows offered the same advantage until they got to be monopolized and started charging as much as Unix did.

Nice FUD

[hkb]

His FUD rant against IBM was quite amusing, considering that he's an executive in a company who's released a series of curious, low-end "funny little boxes". JavaStation... Sun Rays... did anyone really buy, use, and keep using all that crap? No.

I wonder if Microsoft taught Sun execs classes on speaking FUD as a part of the lawsuit settlement...

Yet more hot air from a dying company mismanaging a great, outstanding product (Solaris), that's quickly being swallowed by Linux, Apple, *BSD, and NT, and is so... so out of touch with its customer base, or what's left of it.

Re:Nice FUD

[Anonymous Coward]

"funny little boxes". JavaStation... Sun Rays

Actually, SunRays are kinda nice thin clients (but your point is valid for others... java station for example). What I do like about them is the quiet operation (they have no fans; it's amazing how nice silence is after noisy PCs), and somewhat decent design (the box is small, flat screen display). If they were cheaper I would recommend them for many companies.

FWIW, Sun itself uses SunRays extensively; and even though many companies "eat their own dog food", I think that in this case there are good reasons for using them.

Re:Nice FUD

[Anonymous Coward]

I've just scored a couple of sunrays for playing with at home - running against a linux server (sunray s/w 3.0 beta). Works really well (cards + quiet !!)

"That's what we call a system"

[jackrd]

If that's what they call a system, remind me not to call them; I prefer mine functional. There's no RAM, no expansion cards, and it doesn't look as if there's any CPU installed. If it's "already running Solaris" why don't they show a picture of it running Solaris? "Not for the expedience of a press release." Of course not...they just...didn't want to risk blowing everyone's minds with how amazing they are...yeah...that's the ticket. For as cocky as he is when he talks about IBM's advertising, he doesn't do much better.

Also, is it just me, or can that chip fit in the socket 4 different ways? As far as I can tell it's not keyed, unless that gold circle in the upper right is a pin. I guess they trust people to go by the corner that's shaved off.

Re:"That's what we call a system"

[mistshadow]

The CPU is underneath the huge heatsink with four screw-down bolts in the center of the board.

And these CPUs are factory-installed and tested, not plugged in by the random people who buy them, so being fully keyed is much less of an issue. From the picture, I believe the connectors are not pins, but metal pads, kept in connection by pressure from the above bolts. (which are tightened with torsion wrenches, you'd better believe)

Re:That's just 31 more ways the machine can fail..

[chez69]

IBM has intel and POWER based workstations. zSeries is the mainframe line.

Because sun can compete here

[grahamsz]

There's little point in Sun fighting with intel and amd to produce the highest Mhz chip. They can buy opteron's, stick them in boxes and provide a good low to mid-end system.

OTOH Solaris is already REALLY good at multitasking. The system i'm typing this on has almost 5,000 threads, it's at 80% utilization and it's still very responsive.

As you put more tasks onto a single CPU it'll have to burn more and more cycles doing context switches and suffer from register starvation.

Plus large boxes benefit from economies of scale and can have features that aren't practical in smaller ones:

When a CPU fails the system can take that motherboard out of circulation, then the admin can replace it at their convenience. Same for memory and psu's. Usually no downtime.

Plus we already know that it takes less resources to admin a unix machine than a windows box. Now consider a 144 CPU x 32 Core machine. Even IF it could only handle the workload of 500 windows servers the admin costs are slashed further.

Also consider that the cache might be shared, but then consider that all those cores will most likely be running the same application. I'm sure there's lots of code within oracle or java that gets reused frequently by all the processors. An eightcore chip with 16MB of cache will naturally be able to cache much more of the shared resources than 8 cpu's with 2MB cache.

limited memory?

[Anonymous Coward]

Hmm, I'm looking at the picture on the blog, and I'm just not sure about the memory expandability. I mean there are ONLY SIXTEEN slots for DIMMs on that board. At first I thought those were heatsinks for auxilliary chips to the processors, but no, it's really just the case that nearly 1/4 of the entire board (and it's a big board) is taken up entirely by slots for memory!

Actually, I wouldn't be surprised if Sun is doing separate memory for each core, or some combination like that. They might even be doing interleaving, so that you have 8 memory busses that are interleaved so that each bus can access two separate memory modules simultaneously. You would sort of tend to need to do something like that if you're going to have 8 cores, because you will need massive bandwidth to memory else they will all just sit idle waiting for memory accesses.

Re:That's just 31 more ways the machine can fail..

[davecb]

Actually it's 1/32 the parts count, for a roughly 32:1 improvement in MTBF.

Methinks this is one way Sun's improving their quality control (;-))

--dave

Coralized link to map

[Anonymous Coward]

It seems their server is starting to choke on that 685k png so here's a coralized link to it.

--
"Penn and Teller" vs. The War on Drugs

Two words.

[Anonymous Coward]

WEB

SERVER

Market dynamics

[mr_mischief]

The heart of successfull capitolist system is compitition, and this compitition breeds the best products at the cheapest prices.

Close, but not quite. It breeds a combination of price and quality as close to optimal as possible. Sometimes the best product dies because there's something cheaper. Sometimes the best product survives despite a higher price, because it brings that much extra value to the market. The best product at the cheapest price is a great thing to hope and wish for, but is not usually a realistic goal and is an oversimplification of the market tendencies.

The best realisitc goal one can usually hope for as a producer or consumer is to find a price point which makes a product the best value in the market, whether that means the best product available at a reasonably low premium in price or an acceptable but inferior product with a much lower price. This is where the idea of product classes comes in.

Few people would rather drive a Buick than a Cadillac. They're aimed at people with similar tastes and of a similar age and point of life, so if one was clearly better for everyone, the other would have no reason to exist (both being built by the same parent company and all). One, however, has most of the features of the other and is much less expensive. Even people who can afford the Cadillac may rather spend that extra money somewhere else. This in itself is of coruse an oversimplification, since there are many more than one car manufacturer building cars in these classes. However, I think it makes the point that for the right savings, a product in some ways inferior is quite acceptable to many people. For others, the better product duly earns the higher price. If a small manufacturer couldn't target both and couldn't decided which one to target, it'd have to try to match most of the features of the Caddy with a price closer to that of the Buick. In that case, someone somewhere would find that car to be their sweet spot.

Re:Wow

[Brandybuck]

That's too funny. Now I'm busy hacking my kernel to display that information instead of the usual hardware detection messages...

A thread is not a cpu....

[Glasswire]

8 cores with 4 threads per core does NOT equal 32 cpus. If so - WOW! - my Hyperthreaded Pentium 4 just became a dual-cpu - Which it is not.

HT, or other kinds of Symmetric Multi-Threading DO not give you the benefit of a fully cpu core, as the threads share some resources. Conceptual hint: If you put a full set of all cpu resources (ALUs, etc) for each of N threads - you don't bother to talk about multi-threading anymore, you simply have an N-core processor.

Jonathon would be really talking about a 32-core processor, if that's what Niagra was.

This would be fascinating...

[Digital Avatar]

...if there wasn't already a chip featuring multiple high-performance cores out there. It's called the 25x Forth Multicomputer. Charles Moore is the man to thank. HAND.

performance

[itzdandy]

in any single app, these chips WILL NOT perform any better than any other processor.

the performance comes when processing VERY SMALL operatings such as web serving and file serving where processor latencies cause a performance hit when threads must be qued up and then wait for their turn. smaller pipelines running in parallel will perform much much faster because threads avoid the wait in a processor que. each thread may actually be executed SLOWER but will be completed first because of the paralism.

i can forsee this chip competing very well agains similarly clocked/or priced dual core opterons, in fact they may perform much better in server related operations.

i see no details on memory interconnects and bandwidth, location of memorycontroller, type of ram, or effective FSB.

But how much will a box with it cost?

[Ralph Spoilsport]

$10,000?

If so, which wuold you rather have, one of these, or 10 or even 20 cheapie machines linked in a Beowulf cluster?

Sun's running down the same rabbit hole SGI did, and look where it got them.

Of course, if Apple keeps muscling in on its developers (FCP vs Avid/Adobe, Logic vs everybody, etc.) Apple could well go the way of SGI...so they need to be careful.

It's not a pretty site, and frankly, I think Sun is doomed. They make good stuff, but they've been very badly managed and picked the wrong enemies.

RS

I have two questions for Jonathan Schwartz

[Anonymous Coward]

Dear Jonathan,

On your blog, you say:

> Finally, the 'P' in Power5 stands for
> Proprietary. You can't claim your chip is
> open if you're the
> exclusive supplier, guys - at least you can dual
> source SPARC from Sun or Fujitsu.

Well, here's my first question. Will I be able to dual source your Project Niagra chip, or will it also be Proprietary? You can't possibly have meant that Power architecture itself is proprietary, since I can triple source devices from IBM, Motorola and Xilinx, so I assume you are referring to specific IBM chips. Which brings me to my next question.

Can I dual source any of your currently shipping microprocessors? I note that Fujitsu's currently shipping SPARC chips are somewhat different from yours. They don't work in the same systems, for example.

Thanking you in advance for your answers,

AC.