New Server Chip Niagara

* * Beatles-Beatles writes "Sun recently announced their latest release in server technology. The UltraSparc T1 processor, code-named Niagara, has eight computing engines on a single chip, with each core capable of handling up to four tasks at once." With this new processor Sun hopes to get a leg up on the competition. The Niagra chip is being billed as an "eco-friendly" chip because of its low power requirements. From the article: " [...] removing the world's Web servers and replacing them with half the number of UltraSparc T1-based systems would have the same effect on carbon dioxide emissions as planting 1 million trees."

Comments

Sun has the fix for global warming!

[sgtboost]

Just buy their new processor...it's equivalent to planting a megajillion trees!

Re:Sun has the fix for global warming!

[beh]

Unfortunately, it won't solve the global warming problem - because if everyone switched their web servers to the Sun machines, that would only be used as an excuse to log another million trees for some short-sighted profit...

Re:Sun has the fix for global warming!

[kpwoodr]

Unfortunately, they neglect to tell you that the making of the new processors consumes a megajillion trees. So after you plant the megajillion equivalent, you're back to where we are now. Only with better servers...

Re:Sun has the fix for global warming!

[level_headed_midwest]

Many people are against logging because they have no experience with it firsthand and also hear a lot of the really irresponsible cases (i.e. the slash-and-burn in the poor soil of Brazil.)

Logging is a very good tool to harvest a renewable resource for forever if done correctly. You log mature trees on a parcel and leave the immature ones to grow. The immature trees now have sunlight and will grow big and tall and feed the squirrels, rabiits, and deer with their mast crop. This not only produces lumber but rids the forest of a lot of old, dead, useless wood and provides lots of food for the wildlife. Old trees produce little for nuts and other mast, they just block out everything else underneath them. They are also perfect fuel for a big forest fire.

Logging is almost always done as I said above because if a crew cuts down everything on their plot and does not re-plant anything, they will cut out all of their future income as no new trees will grow, only scrubby thickets.

Re:Sun has the fix for global warming!

[Anonymous Coward]

A million trees, alright I will buy that. But how many trees die, and how much damage is done to make those 8 cores per chip. The manufacturing is a dirty chemical and energy intensive operation.

Re:Sun has the fix for global warming!

[Anonymous Coward]

And wouldn't it be a lot cheaper to just plan a million trees?

Re:Sun has the fix for global warming!

[Jozer99]

We could save another billiontrillion trees or so by replacing all cars with magical unicorn drawn buggys' driven by charismatic gnomes!

Re:Sun has the fix for global warming!

[CapnGrunge]

Yeah, but they don't do jack about pirates!

Re:Sun has the fix for global warming!

[LordoftheLemmings]

Are people still going on about loggin? Shortsited logging companies huh? And what happens after a logging company clears out part of a forest? The replant the trees? Yes they do which is why we have more trees today in america then we did over 100 years ago.

In future news

[Stephan Seidt]

The Fall of Niagara

nasty stuff

[Anonymous Coward]

What about all the real nasty chemicals that go into the manufacturing process of chips .. eg arsenic and acid !!!

Re:nasty stuff

[youngerpants]

Arsenic = Naturally occuring substance (you know when potatoes grow a green scum... thats arsenic)

Acid = Oh no, not acid, which type, I really hope it isnt ascorbic acid, nasty that one, I try to avoid it at all costs... hey, my teeth have all fallen out

Re:nasty stuff

[wytcld]

Chips will be manufactured anyway. The question is what the useful lifetime of the chips that are manufactured is, and what the power consumption of those chips over their lifetime will be, in ratio to the work they perform.

Sparc's strength in the early Internet days was always throughput - even under load - rather than speed. Sun also built more reliable hardware. I switched from Sun to AMD/Linux for Webservers early on, but with energy costs rising quickly, I'll be taking another look at Sun. Where these probably can matter most is for large Web farms, which currently tend to be commodity Linux boxen. But those are throw-away machines - chips headed onward to the landfill after just a couple years.

Re:nasty stuff

[Anonymous Coward]

Your comparison of the types and amounts of arsenic and the acids used in photolithography is disingenuous at best. Go into any chip-fab plant and look at the cyanide emergency kits on the wall to get some idea of what they deal with in reality. I worked with people who did chip-fab in another part of our lab and one of them died from a phosgene leak. Acids they use include hydrocyanic acid (cyanide poisoning) and a variety of heavy metals for doping.

By the why, the green discoloration on potatoes that grow exposed to the sun is chlorophyll and solanine with small amounts of arsenic. You'd get a tummy ache from eating a green potato as opposed to the amounts used in chip-fab.

Re:nasty stuff

[moro_666]

indeed, you have acid's all over the place.

alcohol is acid, quite a weak one, but still having the effects and reactions as the rest of acids have.
most of the batteries that you throw into your ipod contain acid, so does the battery in your car.
if you let CO2 and water together, they create a nice (and a bit unstable) carbonic acid, and that is basically all over the place.
even your body contains so many acids that i would get a ban on slashdot for even naming all of these.
compared to the acids created by gasoline engines and powerstations in the atmosphere, the electionics production isn't even worth mentioning.

sun is making the right move, more computing power for less watt-per-hour, and if they can spare the energy used during producing too, it's even better (and more profitable for them, since they pay for that). having a 200W P4 screaming under your table just to play solitaire is really wicked from the energetic point of view. so is driving an engine overbloated suv just to get one fat butt from one place to another. regular swedish buses that carry 30 people have the same size of engines as hummers or corvettes, sniffing the word 'wasted' anywhere ?

while this cpu will be nogood for playing doom3, it will be a very good chip for handling many many many threads'n'processes at once and therefor be ideal for running webservers and mailservers and other type of multiple client handling services. way to go sun, i hope amd will do an amd athlon 64 X32 some time soon too :)

too bad i can't afford this stuff anytime soon :(

sadly when you are interested in the price of the latest server, you're not rich enough to buy it

Re:nasty stuff

[cbreaker]

I like Sun hardware, always have. But what happens when in "just a couple of years", these Sun chips aren't all that fast anymore? Do you keep them around just because you paid a lot for them?

That's the *benefit* of the so-called throw away machines. They're cheap. They are pretty fast. Lots and lots of bang for the buck. If you go with a really expensive Sun box, you can't do that.

Re:nasty stuff

[absinthminded64]

"compared to the acids created by gasoline engines and powerstations in the atmosphere, the electionics production isn't even worth mentioning."

Yes, please. let's not mention Acid's involvment in the electionics..

acid + counting = funny numbers. Ask Florida.

Re:nasty stuff

[MECC]

A friend of mine was showing me a chip fabricating lab, and pointed out an alarm that went off if there was an accidental release of a certain chemical. He also pointed out that if that alarm were to go off, only people right next to the exit would survive....

Re:nasty stuff

[Anonymous Coward]

Great Stuff, but how much is it? As my work depends on so many X86 applications, is this chip support them?

Re:nasty stuff

[AKAImBatman]

I like Sun hardware, always have. But what happens when in "just a couple of years", these Sun chips aren't all that fast anymore? Do you keep them around just because you paid a lot for them?

I don't know about you, but I can always find a use for a Sun machine. They're built to last, and can often still be useful for a decade after their manufacture. The worst case is that you can resell your old machines to a refurbisher like AnySystem so that it can gain new life in someone else's possession. I know of plenty of companies where the AnySystem servers are powerhouses for the work they need to do. I also know of a lot of developers and sysadmins who would like a Sun Workstation, but can't afford new. Again, AnySystem (or Ebay, take your pick) can provide them with a system that meets or exceeds their needs. :-)

Re:nasty stuff

[Yokaze]

> Naturally occuring substance

It always puzzles me, how people think, that natural is some indicator for being harmless. You know, like: "Here my love, take that medecine, it is made from plants, perfectly harmless, all natural", which triggers my reposte, "So, is a cup of hemlock ".

What puzzles me even more, is that you take such a perfect counter-example as an argument.

Re:nasty stuff

[jafmuse]

If you are suggesting that Sun etches their chips with Vitamin C, you are WAYYYY off. Flintstone Chewable UltraSPARC anyone? I thought not.

Re:nasty stuff

[Anonymous Coward]

while this cpu will be nogood for playing doom3

What do you mean? That CPU would make a damn fine Doom3 server... ;-)

Re:nasty stuff

[InvalidError]

"Sparc's strength in the early Internet days was always throughput - even under load"

Are you implying that you can have useful throughput under no load? How do you measure this idle throughput advantage?

The Intel/AMD architectures are historically single-threaded desktop-centric where the most important thing usually is to run one thing really fast. Sun, however, was always in the HPC/workstation game where overall throughput matters most, latencies and single-thread performance be damned. These two groups were playing pretty different games up to recently.

But now, Intel/AMD have hit a GHz and complexity brick wall. They are forced to promote multi-threading multi-core at the desktop-level and optimize their future desktop chip designs for multi-threaded application throughput rather than single-threaded performance. Imagine what would happen if AMD and Intel could afford to quit competing on single-threaded performance overnight: goodbye complex deep out-of-order execution, goodbye branch-prediction and speculative execution - those transistors would be much better spent on implementing quad-threading cores to keep every pipeline filled with useful instructions that will retire cleanly on every clock.

Sacrificing single-thread performance for simultaneous multi-threaded throughput in the above-described way has been the name of Sun's game for the last few years.

Obsession with single-threaded performance is what costs current x86 CPUs the most power. Of course, in the P4/HT case, there is the added power and transistor costs of trying to be a jack-of-all-trades who predictably turned out as a master-of-none. (The P4's uOP replay engine is a neat idea... but re-executing the same stupid uOPs until they meet retirement conditions is woefully wasteful, whoever designed and bothered to patent this should be fired.)

Re:nasty stuff

[moro_666]

playing doom versus serving multiple clients, as a dedicated server, the latter one is clearly the other case that i pointed out.

i guess doom3 server is a mixture of handling multiple clients and 1 heavy weight calculation algorithm with vectors ... if this is the case, and amd athlon x2 will outperform this slick sun cpu , since 1 core in this sun miracle doesnt have the power of 1 core in amd. ofcourse if they would have a thread with less killing algorithms it would be good for the sun cpu, but it would be a terrible miss on a regular cpu available today (and that's why i think they have the 1 heavy weight thread and some lightweight network handling threads). it would be nogood to have 31 networking threads waiting behind 1 lazily lagging realm calculation thread.

while playing doom3 is totally different weight having most of it's cpu power teared apart by the engine that feeds your superfast 3d card with correct information and doing little else like AI behaviour or networking tasks.

besides, do you really think that sun has a fpu for each of the processor handling units ? i don't think so, cause it would be very expensive and not used in any web-, file- or mailservers. they may have more than 1, but not 8x4. and calculating bullet trajectories, lines of sight, current angles here and there, is pretty clearly fpu work mixed with heavyweight memory access.

when regular pc-s should ever have that sort of multitasking build-ups, idsoftware would ofcourse kick the multithreaded stuff out of the door and you'd be right.

Re:nasty stuff

[TeknoHog]

> Naturally occuring substance

It always puzzles me, how people think, that natural is some indicator for being harmless.

Moreover, it's nigh impossible to define substances as 'natural' and 'unnatural'. IMHO, nature == the universe, and everything that exists is natural -- whether or not manufactured by individuals of a certain species. Things that don't exist are supernatural :)

Re:nasty stuff

[Anonymous Coward]

"unnatural" usualy means that every little piece you could find on this planet was created/assembled by humans. Of course it's build from "natural" stuff, but it would not have happened [here] without us helping.
At least I think that's a sensible definition.

Re:nasty stuff

[Weh]

Arsenic, though occuring naturally, is poisonous in large enough concentration. Pollution of groundwater with arsenic is an increasing problem in many countries.

Check out the link if you like...

http://www.epa.gov/safewater/arsenic.html

Re:nasty stuff

[lowe0]

"most of the batteries that you throw into your ipod"

You don't throw batteries into an iPod - it has a built-in battery.

Re:nasty stuff

[aminorex]

Another religious maniac.

Re:nasty stuff

[Anonymous Coward]

+1 funny

Re:nasty stuff

[OrangeTide]

Arsenic is a poisonous metalloid. And therefor inorganic. It (Lead arsenate) has been used in pesticides, especially for fruit trees. Sodium arsenite has been used on potatos, but you're not supposed to be finding arsenic on potatos. And it doesn't "grow" there.

Now you can have cynide produced in plants, as it is an organic compound. But the kinds of poisons you find in potatos, tomatos and other plants related to nightshade are just alkaloids. (nicotine is an alkaloid and immediately fatal in a large dose).

When you think of Arsenic think of Lead. They are both elements, they are both toxic and they both move through the environment in roughly the same way. You can't "grow" lead, plants can't produce it, it has to be present in the environment.

Electronics manufacturing produces a fair amount of Lead and Arsenic waste. Which is bad of course. But cynide is a naturally occuring compound and is really bad for you (I'd say it is on par with breathing carbon monoxide all the time). And the metal and mining industry uses many tons of cynide and you can be certain they don't reclaim 100% of it. it's cheap and not that strictly regulated, mostly it's the local city wastewater treatment plants that hand out citations.

Re:nasty stuff

[mikefe]

I disagree. There is an active recycling community in the USA.

If you throw away a system, you have missed out on selling it to a recycling company.

My link to the abc news business section?

[brejc8]

There must be 100 better sources out there to link to.

Easy

[Laz10]

Let's go plant some trees then.

It is a pretty safe bet that 1 million trees are way cheaper than Sun technology.

Re:Easy

[Anonymous Coward]

I wonder how they calculated that figure... can someone convert to LoC please?

Here's some info on Carbon Neutral Calculations: http://www.menofthetrees.com.au/calculations.html

Re:Easy

[Moby Cock]

I doubt they considered the enegy expended in creating this new design and manufacturing this new design. I'd like to see how they came up with such nonsense.

Re:Easy

[AvitarX]

First, they bent over. Then they reached in their ass. Lastly they published it as fact.

Re:Easy

[malaprohibita]

My vote is for the trees.

Re:Easy

[jacksonj04]

Just a thought here, but since when has planting trees reduced carbon emissions?

Re:Easy

[timeOday]

I wish we could just ignore their eco-friendly marketing spin and get down to whether the chip is any good. 8 quad-hyperthread cores sounds pretty intriguiging to me. I'd like to try out make -j32 on this puppy.

Re:Easy

[DJStealth]

I'm not an ecologist or anything, but planting trees cannot reduce emissions.

The definition of emissions, means the product being emitted. Planting a tree does not reduce CO2 being emitted. Although it probably does 'eat up' the CO2, reducing the amount of the gas in the air.

Re:Easy

[Chmarr]

They didn't say it would reduce carbon emissions. They said it would "Have the same effect on carbon emissions" as planting 1 million trees. Because, ya know, trees absorb carbon dioxide from the atmosphere.

The USA is actually a net carbon SINK, rather than a carbon SOURCE. This is one of the reasons the USA didn't sign the Kyoto treaty; the treaty doesn't take into account consumers of atmospheric carbon, only producers.

Re:Easy

[fbjon]

The USA is actually a net carbon SINK, rather than a carbon SOURCE. This is one of the reasons the USA didn't sign the Kyoto treaty; the treaty doesn't take into account consumers of atmospheric carbon, only producers.

That doesn't compute. It's global warming, remember.

If the US produces carbon sources, it places a global strain on the system. If the USA produces carbon sinks, it reduces the strain on the system. Is there any industry that traps more carbon than it produces? In other words, just because the USA is a net carbon sink, doesn't mean that it should be allowed to run around willy-nilly filling their local quota of carbon absorption, because that absorption does not belong only to them, it belongs to the world (as it always has).

On the other hand, if someone plants 211 900 km^2 of carbon-vacuum-cleaning forest and vegetation, then sure, use it to an advantage.

Re:Easy

[Syberghost]

I'm thinking they forgot to take into account how much driving you'd have to do to plant one million trees.

Re:Easy

[spongman]

did a quick search, Weyerhauser grows 230 million new trees anually

Re:Easy

[ear1grey]

True, but then I never wanted to be a computer scientist anyway,
I wanted to be;
a lumberjack!
leaping from tree to tree...

etc.

You'll be singing it all day now, sorry.

Re:Easy

[jacksonj04]

Therefore planting trees doesn't affect carbon emissions at all, it only affects the amount of carbon re-absorbed.

Re:Easy

[dkf]

Can I plant B-Trees to get that saving? Or does it have to be R-B or AVL trees? Would a DAG do instead?

Damn!

[ViaNRG]

My computing engine needs an oil change!

Better link

[Anonymous Coward]

Better link here.

Right.

[Skye16]

" [...] removing the world's Web servers and replacing them with half the number of UltraSparc T1-based systems would have the same effect on carbon dioxide emissions as planting 1 million trees."

I'm sure it would. But then, after replacing all these servers with the UltraSparc T1-based systems, we'd have to cut down 1 million trees just to print the money Sun would be making back in profit. Weeee!

Re:Right.

[Sockatume]

Curses, beaten to it. You ommited the golden phrase "gigantic sacks of cash" though.

Re:Right.

[FooAtWFU]

Yes, because when I order a server over the mail or the internet, I pay only in $1 bills, because I want to keep myself safe from identity theft. Paper money... I love it! 3

Re:Right.

[tezbobobo]

Well hotdamn! I'll go out and replace my home server,a 486 which serves about 30 pages a day of Redhat and provides access to my home network. I'll just go get my wallet.

Re:Right.

[mwvdlee]

"[...] removing the world's Web servers and replacing them with half the number of UltraSparc T1-based systems would have the same effect on carbon dioxide emissions as planting 1 million trees."

Removing the world's Web servers and not replacing them at all would have an even better effect on carbon dioxide emissions.

I'm curious as to how they calculated it, though. Are they talking 1 year running time or 100? Are they taking into account the energy required to build those new systems? Do they supply the new hardware's manuals on paper? How does it compare to a similar scenario using other replacement hardware?

Re:Right.

[Skye16]

Also, what kind of trees are we talking? What size/age? Are we talking lil' 3' high sprouts? Or 100 year old oaks?

Re:Right.

[Chilles]

No no no, the meant to say that they care so much about the environment that they will give you one of these UltraSparc T1 based systems if you turn in two of your devices capable of serving web pages.

Re:Right.

[Hoi Polloi]

Sadly, all of the new Sun CPUs are made out of wood.

Re:Right.

[Qzukk]

Are they talking 1 year running time or 100

Presumably the tree's contribution to the atmosphere does not consist of one giant puff of oxygen before they die, so it would be safe to assume that their statement means that they believe that the CO2 absorption rate provided by the trees is equal to the rate of CO2 generation by whatever technology they used for the calculation (probably coal) to make up the electricity difference between the two technologies.

Re:Right.

[Anarke_Incarnate]

At least in the US, money is printed on cotton pulp and not wood pulp.

Re:Right.

[Anonymous Coward]

> Sadly, all of the new Sun CPUs are made out of wood.

...and use a magnifying glass as a heatsink!

Re:Right.

[hackstraw]

At least in the US, money is printed on cotton pulp and not wood pulp.

Actually, its 25% linen and 75% cotton. "Paper" money is a misnomer. Paper cannot withstand the folding, washing, and other abuses that "paper" money goes through.

Re:Right.

[Anarke_Incarnate]

And actually they also have red and blue synthetic fibers, as well as strips embedded for identification. However, they are, in fact, paper. Paper is not solely produced from wood. The fact that it is made from an aggregate of pulp and not woven makes this a paper. Many high quality papers are not made from wood pulp at all.

Except...

[Anonymous Coward]

removing the world's Web servers and replacing them with half the number of UltraSparc
T1-based systems would have the same effect on carbon dioxide emissions as planting 1
million trees.

... exactly where would all these other systems go to? Landfill? Recycling stations/stock piles?

Sales droids. Ugh.

Re:Except...

[Pollardito]

reminds me of this Newsweek article from a few months back :

That boom comes from buyers like Roberta Gray, who threw away most of her old clothes to make room for a new wardrobe from Greenloop, a boutique in Portland, Ore. A newly converted vegetarian and yogi, Gray, 43, wanted clothes that matched her healthier lifestyle. "When I buy things there, they last," she says. "They're of good quality, and I feel good about that."

it seems like her old wardrobe was lasting just fine until she threw them out!

raw power

[emptybody]

Niagara systems take the concept of dual core processors (with which most of you are familiar), and goes to an absolute extreme - building 8 cores, each capable of running 4 jobs simultaneously (4 threads), onto a single chip. Doing the math, we'll be delivering a 32-way chip, running 9.6GHz, which sips power (about 70 watts). , JonathanSchwartz BLOG.

This is why I got into Sysadmin 15 years ago.
To play with big honkin fast machines and new technology that makes your head spin.
Just musing about the name. Think of your kitchen sink faucet.
Now think of all the faucets in your house turned on at once.
Now think of all the faucets on your street turned on too.
Add all the faucets in your community.
Keep on thinking of how many faucets in how many communities it would take to equal the raw power behind something so large as Niagra falls.

Am I hooked?
You bet.

Re:raw power

[Anonymous Coward]

I see.. They save trees and you waste water.

Re:raw power

[rkhalloran]

For a lot of uses today, a swarm of processors like this makes more sense than driving one CPU hellishly fast so it can task-switch quick enough to get around to everything.

Sun traditionally has been very good at engineering the interconnects so I expect the actual throughput on this is pretty good.

Will be interesting to see how well this does.

Re:raw power

[ickleberry]

uhh.. wtf is a faucet?

Re:raw power

[mcrbids]

Just musing about the name. Think of your kitchen sink faucet.
Now think of all the faucets in your house turned on at once.
Now think of all the faucets on your street turned on too.
Add all the faucets in your community.
Keep on thinking of how many faucets in how many communities it would take to equal the raw power behind something so large as Niagra falls.

Dude. The water in my house all comes thru a 3/4" pipe. Turning all the faucets on would mean that they all dribble. Turn on the whole neighborhood, and I doubt a sprinkler would even sprinkle.

What kind of neighborhood are YOU from?

Re:raw power

[OldManAndTheC]

Think of your kitchen sink faucet.
Now think of all the faucets in your house turned on at once.
Now think of all the faucets on your street turned on too.
Add all the faucets in your community.

Thanks a lot, now I have to get up and take a piss.

Re:raw power

[Anonymous Coward]

Think about your breathing.

Re:raw power

[engineerofsorts]

How is he doing the math here? He's implying a 32-way system running at 9.6 GHZ. What he really has is an 8-way (thinking of # of cores) or 32-way (# of cores times # of threads/core) running at 1.2 GHz. He is implying with his math the equivalent of a single processor running 32 x 89.6 = a 307 Ghz uni. All this is entertaining statisculation. What will be more entertaining are when he can post real performance numbers for a real system.

Re:raw power

[anOminousCow]

Good luck to Sun and their new Viagra chip. I'm sure it will help them get their third leg up on the competition.

Re:raw power

[Anonymous Coward]

fucking jews

Re:raw power

[Rasta_the_far_Ian]

This is why I got into Sysadmin 15 years ago. ... Keep on thinking of how many faucets in how many communities it would take to equal the raw power behind something so large as Niagra falls.

You screwed up dude! Should have gone into Chemical Engineering! We really do it with large pipes! :)

What about I/O?

[Max Romantschuk]

Eight cores at four threads is 32 simultaneous threads. Nice, but what about memory bandwith? Each thread needs proper I/O if this is actually going to do any good... Anyone have any real info on this marchitecture?

Re:What about I/O?

[SQL Error]

Specs here. Four 144-bit DDR2-533 interfaces. That's more memory bandwidth than a quad-Opteron system.

Re:What about I/O?

[HaydnH]

Sun always builds their systems to be balanced and avoid bottlenecks, it's the first thing you learn about on the internal training courses so needless to say the I/O is fast enough.

Haydn.

Re:What about I/O?

[RzUpAnmsCwrds]

"That's more memory bandwidth than a quad-Opteron system."

Yes, but with considerably more latency. DDR2 is great until you realize that the latency effectively kills any performance advantage you had with higher clocks. DDR2-533 is about as fast as DDR-400.

Re:What about I/O?

[the eric conspiracy]

The 4 threads per core is designed exactly for that issue. If a thread is waiting for memory, execution can proceed on a different thread.

Re:What about I/O?

[photon317]

The new Opterons on the new socket architecture actually will have DDR2, it was just announced publicly a week ago or so.

Re:What about I/O?

[Paul Jakma]

The whole point of Niagara is to get higher throughput *despite* memory latencies.

Re:What about I/O?

[fitten]

I haven't looked at the actual timings, but the vast majority of memory operations are on cache lines. So, you have to use some math to determine how fast a cache line is filled using each technology to see what the difference is. It isn't all about the time to get the first byte in, it's about getting the cache line in (which is 256 bits - 2x128-bit reads on our favorite x86 machines). DDR will still be a little faster, I'm guessing, but the initial read latency is amortized over the two reads in a dual-channel memory configuration.

Besides, I think Sun is saying that they will be using faster than DDR2-533 memory *and* the system is more tollerant of latency issues than the AMD and Intel parts that we are more familiar with. Of course, I'm taking their word with a grain of salt because even four threads on a single core can easily all be stalled given that at 2GHz, a single cache load may be the equivalent of over 200 instructions. So, it's quite imaginable that all four contexts can be stalled for main memory accses simultaneously given the rather small caches it has. Also, IIRC, even the L1 caches have a 3-cycle penalty which seems to me like that alone will guarantee many short stalls (like every instruction/bundle fetch).

Re:What about I/O?

[loose_cannon_gamer]

Ahhhhh, I find that parent very informative indeed. So this is really like 8 cores and quad level hyperthreading, where the first thread will be running 70%+ of the time, and the second thread will get scheduled when the first thread blocks, and run for 25% of the time, and the third thread will get scheduled when the first two are both blocked, and run for 4.8% of the time, and that fourth thread, well, 0.2% is all it gets. (And yes, I just pulled these numbers out of the air).

Just out of curiousity, shouldn't there be some kind of 'quality of service' notion when claiming how many threads you run simultaneously? I just seem to recall that hyperthreading on the P4 gives about a 15% throughput increase with a second thread, and I can't imagine that you get anything other than similarly diminishing returns if all four threads are fighting for the same memory bandwidth and execution units.

Just to explain, it is my understanding that hyperthreading is little more than opportunistic hardware level thread scheduling -- while one is blocked on I/O or something, two can use the same execution units *carefully* to accomplish work. If my analysis is right, and I could be out in space, you might as well build a chip that can run 1024 or 1048576 threads simultaneously (sure, you need a little extra hardware per thread), but you're still only going to see work get done by the first 2, maybe 3, although I doubt even that many.

Re:What about I/O?

[htd2]

You seem to have missed the point of the T1. On a conventional system a cache miss generates a stall and a wait while main memory fetches the page(s) while this happens the whole of your 100+ watt CPU is doing absolutely nothing for however many cycles it takes to get the page(s).

Not so the T1 which has another 31 threads which arn't stalled hence the T1 needs a wide but not particularly low latency memory subsystem.

Re:What about I/O?

[ChrisGilliard]

This is the purpose of the instruction cache and data cache. I'm almost positive each core has one, so it can access both instructions and data that's used frequently without going to the bus.

Re:What about I/O?

[Anonymous Coward]

Hyperthreading on the P4 is just a Cheap Marketing Trick(TM) designed to get some extra use out of the ridiculously-long pipeline. Not all execution units are available to each "thread".

The Niagara chip holds 4 full thread contexts per execution unit (in this case a processor core equivalent) which it can switch to with 0 clock-cylcle latency. They all have equal "rights" so to speak.

Re:What about I/O?

[raftpeople]

Intel Hyperthreading is a response to their pipeline problems. It is not simultaneous multi-threading, it is merely caching registers and switching quickly. 15% is about right for that approach.

IBM's SMT (and possibly Sun's, not sure) is an actual duplication of much of the processing logic and allows for better performance gains.

Not all vendor's solution is the same.

Re:What about I/O?

[DeputySpade]

Bingo! This is exactly right. Where's the mod points when I need them? It's not about making memory fast, it's about finding other interesting things to do while you wait so that there's always work getting done. These things do have a fatal shortcoming though...

Re:What about I/O?

[SQL Error]

There are two main types of hardware multithreading, referred to as "vertical" and "horizontal".

Vertical multithreading works much like you described. Only one thread can be running in a given cycle, and when one thread is waiting on a memory access another thread can jump in. But the load balancing works out better than your example. :)

Horizontal multithreading picks out instructions from all the threads and then schedules them into the available issue slots to get the maximum throughput. IBM's Power5 does this.

Niagara is designed for servers running moderately complex interactive tasks. Dynamic, database-driven websites would be ideally served by it. Although the individual cores would be much slower than a Xeon or Opteron, there are 8 of them on one chip, and multithreading on top of that. On suitable workloads, this single chip could outrun a quad-processor Xeon. On unsuitable workloads, it will suck, because it's still only 1.2GHz.

Re:What about I/O?

[htd2]

You need workloads which exhibit a high degree of thread level parallelism. These are generally workloads such as Web and File servers, J2EE app servers, mail authentication and directory servers, DBMS's and of course servers where there are largish number of independant processes.

When Sun aquired Affara (the initiators of the T1 design) they had a breakdown of the kind of apps that would benefit from multi-threaded CPU's.

Other people in this thread have commented on the suitability of the T1 as a notebook CPU. Apart from the fact the the T1 is a bit to hot for a notebook it will also suffer because desktop apps tend to be mostly single threaded with the focus of the operator on the performance of one app. They may have a large number running but at any one point in time they are only interested in the performance of one of them. Since the T1's single threaded performance is nothing to write home about it would not be a good choice for a desktop/notebook.

Rock which is the Sun's next major processor release after the T1 and its second revision will have much better single threaded performance. It is also unlikely to turn up as a desktop because it is targeted at the large server market.

Re:What about I/O?

[Anonymous Coward]

The idea is balance. You could crank up the number of strands that share a core and render memory latency less important, (decreasing sigle thread performance and requiring more threads on the box to make good use of it), or you could reduce the number of HW strands, but would need faster memory (or larger caches as is the typical solution).

The 4 strands and given memory latency is a balanced point on this continuum.

Memory bandwidth is still important, since multiple strands per core will hide latency but increase demands for the overall amount of memory per ns that needs to be moved onto a core.

Re:What about I/O?

[JonAnderson]

On Niagara, you effectively have 4 hardware thread contexts (strands) per core. If all these strands can run then context is switched every cycle. If a strand is stalled then it is parked and gives up it's cycles to other strands which can run until it's stall condition is alleviated. Note, this is different to a scheduler context switch.

Dumb

[pubjames]

Dear Sun,

RE: your statement:

Removing the world's Web servers and replacing them with half the number of UltraSparc T1-based systems would have the same effect on carbon dioxide emissions as planting 1 million trees.

Please engage brain before opening mouth.

Thanks.

Re:Dumb

[Harrakis]

You are right on the money here, unless SUN has found a way to power a computer from CO2 in the air...

Re:Dumb

[Anonymous Coward]

I wonder, if they just plain old halved the number of webservers would it still save 1 million trees?

Re:Dumb

[frostfreek]

building 8 cores, each capable of running 4 jobs simultaneously (4 threads), onto a single chip. Doing the math, we'll be delivering a 32-way chip, running 9.6GHz, which sips power (about 70 watts). [sun.com], JonathanSchwartz BLOG.

Apparently, their brain was already fully engaged in multiplying 4 and 8 together.

Re:Dumb

[wannabgeek]

Why is it so dumb? They may say that because the power required is reduced soo much. I don't know how much of power generation today results in CO2 emissions though.

Comparisons to the cell?

[Anonymous Coward]

These processors are a step in a different direction. Like the cell processor, they lack features like branch prediction, have small, very simple pipelines, etc. However, that isn't really all that bad, esp. on some tasks where your CPU is mostly just idling waiting for IO to finish anyway. I wonder if these "simple but gets the job done" CPUs will see an even wider market in the future. As the article said, they are cheaper and consume less power than their competitors

Re:Comparisons to the cell?

[Quicksilver]

Not really the same thing at all. The cell uses one master to control several specialized units. Niagara is just what is sounds like. 8 cores on 1 piece of silicon. They all are the same and the all can run any Sparc code.... unlike the Cell which isn't compatible with anything and each unit can only work on what it is specialized in.

Re:Comparisons to the cell?

[t35t0r]

The Cell is an asynchronous SMP and the SunT1 is a synchronous SMP machine. The Cell processor has one master CPU with multiple "DSP" chips or cores. The SunT1 has multiple CPU cores.

Re:Comparisons to the cell?

[DurendalMac]

The Cell uses a POWER-esque ISA. It is compatible with some things, it just runs them like crap if they're not optimized.

Re:Comparisons to the cell?

[Eric Seppanen]

While they ditched the really massive heavyweight branch prediction, it's not accurate to say it has none. Simple branch prediction is really cheap, and pretty much required for pipelined CPUs to have good performance.

Plus, they have something that appears to make up for what they lost-- the hardware scout. It executes ahead in the instruction stream, preloading necessary data and instructions into the cache. There's lots of articles accessible via google if you want to learn more: Sun's Marc Tremblay has been giving a lot of talks about how it works.

Threatening activists

[eebra82]

So now I can extract more money from tree-hugging activists by saying that I will get a first generation Itanium, UNLESS they give me whatever's equivalent to power savings and planted trees.

GIVE ME $5,000,000 OR I'LL GET ITANIUM INSTEAD OF NIAGARA!!

And...

[Moby Cock]

removing the world's Web servers and replacing them with half the number of UltraSparc T1-based systems would have the same effect on carbon dioxide emissions as planting 1 million trees

"And it has the added benefit of lining my pocket."

Re:And...

[popeyethesailor]

And probably bankrupt atleast a million small-medium web hosts ;)

Re:Apple need this

[Anonymous Coward]

low power requirements != low enough for a laptop.

Niagra = 70 watts
G4 = 19 watts

Impractical

[Anonymous Coward]

I don't know many people who have a server room large enough to hold a million trees.

(twiddles thumbs for the remaining 17 seconds. Lahdy dahdy hum dum dum)

"The Niagra chip"

[temcat]

Gosh, those pesky spammers manage to squeeze their cheap Viagra ads everywhere.

Re:"The Niagra chip"

[ettlz]

That's because "with this new processor Sun hopes to get a leg over on the competition" (e.g., fuck Intel?).

What we need from Sun...

[Anonymous Coward]

The new Niagra line... courtesy of the little blue pill.

I can't wait to see the SPAM over this one...

(Insert obJoke about "staying up" and "hardware" etc.)

Re:What we need from Sun...

[zigziggityzoo]

Haha....Uptime....Let's just hope it's not more than 4 hours, or else you'll have to consult a doctor.

Sony's Cell twin?

[digital.prion]

I'm not sure but it seems that this is another scalar/vector chip or more of a hybrid super scalar but I'm really liking the concept of a cpu with scalar and vector units. On that note is this part of Suns Majc proccessor line? cheers!

Massively multi-core x86s

[rbanffy]

AMD or VIA would build a cheap multi-core x86 based on VIA's or Geode cores... Sun could sell systems with them as developer boxes running Solaris 10.

BTW, what would happen to performance if you started with a Geode core and spent the rest of your wafer-area budget with Itanic-size caches?

For now, I have no hope to have one of these on my desktop anytime soon.

Re:Massively multi-core x86s

[rbanffy]

AMD or VIA could... OK. I should have used "Preview"...

Re:Massively multi-core x86s

[tomstdenis]

Dude, a PPC 440 is 6mm^2 and consumes 700mW of power at 667Mhz. You could easily fit a dozen on a die the size of a P4 and still take FAR less than this Niagra core. At a rated 1200MIPS per core a die with eight of them would net you close to 9600MIPS max. Of course you'd need some form of L2 cache and high speed internal bus.

Similar the new ARM cores Cortex it takes roughly the same power at 1Ghz which gives it apparently 2000MIPS. The area is about the same as PPC 440. So in theory you could hook 4-8 of these up as well and get a killer chip too..

Point is Suns quotes of being "2 possibly 3 generations ahead" is totally bullshit. They're at most one generation ahead. It takes one multi-core ARM or PPC to totally destroy this.

Tom

Re:Massively multi-core x86s

[mritunjai]

" At a rated 1200MIPS per core a die with eight of them would net you close to 9600MIPS max. Of course you'd need some form of L2 cache and high speed internal bus."

Cache is not as small issue as you think. In most modern processors, 50%-75% of die is taken by cache alone. So do you math again taking half to quarter of die size.

Re:Massively multi-core x86s

[speculatrix]

the problem with the x86 architecture is that the whole instruction set is out of date:
http://www.geek.com/procspec/features/revx86/

Re:Massively multi-core x86s

[tomstdenis]

Law of diminishing returns?

Compare a 2MB L2 cache on a P4 to a box with 1MB of L2 or 512K ... the cache ends up contributing less and less to the overall performance.

What is also important is associativity. If you have a low-assoc cache, meaning a given address has few places in the cache it could reside you end up wasting more space. That's why [iirc] the AMD processors have high associvity L2 caches. They make good use of the 512K available.

At my previous job we built Gentoo distros on 128 and 256K semprons and the time to build wasn't really that different even though we were building 100s of packages.

So you could have a relatively high performance web server or file store [or whatever] without the 2MB of cache stuck on the back of the thing.

Tom

Re:Massively multi-core x86s

[IvyKing]

I think you're being optimistic about the cache for the multicore PPC 440.

As far as power requirements - once you fill up all 16 memory slots, the memory will be consuming more power than the CPU. That implies that you wouldn't see much of an energy savings with the multicore PPC 440. To get further energy reduction would require more attention to reducing the memory I/O - maybe better cache design, maybe better software.

The lowdown

[gtoomey]

http://www.sun.com/processors/UltraSPARC-T1/specs. xml

Since the story is devoid of content:
- up to 8 cores, 4 threads per core
- integrated RSA
- 3MB L2 cache
- 90nm process
- 1.2 GHz

Re:The lowdown

[southpolesammy]

Since the story is devoid of content...

No worries, since most of the comments here (excepting yours) have also been devoid of content.

Re:The lowdown

[should_be_linear]

1.2 GHz is interesting part. As we already know that "x threads per core" makes almost no difference, high end version with 8 cores * 1.2 Ghz = 9.6 GHz. This means that dual opteron 2GHz (with dual core chips) should have comparable performance, while complete system costs around $3000. Sun with 8 core Niagara wil be 5x more than that, at least. Am I missing something except trees and CO2?

Re:The lowdown

[Anonymous Coward]

The Opteron cores are single threaded, so a stall for memory access means that no other work is being done on the entire core. A multithreaded chip can switch to another thread instead and make some progress. Some workloads scale well with this approach.

Re:The lowdown

[imsabbel]

You are also missing that those cores are single inssue in order, IIRC.
So 4 opteron cores have the same numbers of clocks than a Niagara, but can do a LOT more per clock.
But otoh, the targets are vastly different, so both solution could have their place...

Re:The lowdown

[Mr. Storm]

The big thing you're missing is that the Opteron spends a lot of time waiting for data to come back from memory. The higher level post that claims "As we already know that "x threads per core" makes almost no difference" is only correct for Intel's hyperthreading. Not for Sun's Niagara design. The UltraSPARC T1 uses the 4 threads per core to hide the memory latency that cause instruction pipelines to sit idle and wait on other CPU designs. That makes a bit of a differnce in total throughput. Check out the memory bandwith. Maybe there's a reason for it.

Re:The lowdown

[Anonymous Coward]

The Opteron is an Out of Order CPU, so if there is a memory stall, it will pick some other instruction (with no hazards) and execute that instead.

Not appropriate for all types of workloads

[XNormal]

You need at least a dozen concurrent threads or processes before you can make good use of this CPU's power. Certainly not a good idea for desktops. An excellent match for web servers. Other server-type workloads (e.g. database, file server) may need some tuning to make the best of this architecture.

Re:Not appropriate for all types of workloads

[Knetzar]

I'm pretty sure Sun is aiming this at the J2EE space.

Re:Not appropriate for all types of workloads

[Anonymous Coward]

Databases like PostgresSQL and Oracle that utilize a multiple process model
automatically use additional threads/cores/processors.

Re:Not appropriate for all types of workloads

[root_42]

You need at least a dozen concurrent threads or processes before you can make good use of this CPU's power. Certainly not a good idea for desktops.

Not for the average Joe's desktop. But definitely a good developer workstation. Concurrent compiling... It might also be quite nice as a graphics workstation. If you have a threaded rendering application, this would also be nice.

Re:Not appropriate for all types of workloads

[zerocool^]

What I'm thinking is VMWare-type virtualization.

I recently was hired by the CS department at a major university, and they're looking into a way to replace the rows and rows of desktops in a cs lab with a fewer amount of servers, while at the same time, being able to give their students access to the entire OS. Getting one dedicated server per student wouldn't fly, but something like this running vmware, or a small cluster running usermode linux would allow for quite a few small virtual machines.

I'm intrigued.

Plus, there's nothing like the warm embrace of Sun hardware. It costs arms and legs; but there's no substitute for the stability of sunOS+sun hardware. When I worked in webhosting a couple of years ago, we had quite a few Sparc IPC's, IPX's, and Sparcstation 1's (12, 20, and 40 Mhz, respectively, i think), chugging away as static content servers, dns servers, and a variety of other functions.

~W

Re:Not appropriate for all types of workloads

[pooh666]

Unless you consider what Sun has done with Zones in Solaris 10.. Now we can divide up that 8 core chip into 8 virutal machines or more. That combined with a decent RAID 10 setup would be a pretty beautifl thing, and as easy or easier to maintain as 8-10 real machines. And guess what about price?

Re:Not appropriate for all types of workloads

[psavo]

Databases like PostgresSQL and Oracle that utilize a multiple process modelautomatically use additional threads/cores/processors.

I'd really like to see database that benefits more from processor power than from memory bandwidth. remember, it's database, not spreadsheet.

Re:Not appropriate for all types of workloads

[juancn]

Not necessarily, at the time of this writting, my desktop machine has 510 threads belonging to 48 processes (I'll grant you that is not a typical setting, it is a development workstation).

Single threaded applications are pretty rare, most modern applications have more than one path of execution (you cannot afford to freeze the screen while saving for example). Network I/O is much easier to program with threading, in oposition from asynchronous I/O (think a browser with several tabs/windows open).

I wouldn't so lightly assume that multi-threading support is only desirable for server systems.

Anyway, this has been designed mostlly for servers, consider that a typical Java server will have around 50-100 threads (maybe more than 300). This is where this technology makes more sense.

Re:Not appropriate for all types of workloads

[Anonymous Coward]

A dozen concurrent threads/processes? On this desktop I see: init, 16 kernel processes, udevd, syslogd, atd, crond, cupsd, dbus, famd, hald, dhcpcd, ivman, gettys, XDM, X server, window manager... and that's before you start counting applications.

Re:Not appropriate for all types of workloads

[Anonymous Coward]

Easy. Databases written in QBasic, running in an XP cmd.exe, in bochs, under linux.

Re:Not appropriate for all types of workloads

[Anonymous Coward]

Zones on the backend (seconds to create, no emulation slowdown like with VMWare, and you can still maintain it like a single system when applying OS patches) with Sunrays on the desktops. The Sunrays are fanless and diskless and maintenance-less and let students carry their persistent sessions around with them. Let me tell you, being in a silent lab after a lifetime of unknowingly listening to fan and disk noise is freaky!

Sun's AMD systems are nice too.

Re:Not appropriate for all types of workloads

[Anonymous Coward]

This technology is leap-frogging all the other chip designs, but they are all headed in this direction.

What is holding this technology back is application developers. Threading (fork) has been around for a very long time, but very few developers make use of it.

This processor will be ideal for a workstation if applications were threaded, imagine StarOffice where every feature was one or more threads, the spell and grammar check may utilise several threads to track your typing etc....

As MHz limitations put pressure on more threaded CPU designs the applications are going to have to catch up. Now is a good time to change.

Re:Not appropriate for all types of workloads

[Anonymous Coward]

A problem that no one seems to be discussing is that most commercial software for these other workloads (database, J2EE server, mail server, ...) is licensed on a per core basis for Unix. This makes per core performance very important for your overall price-performance. This per cpu licensing issue is causing problems for the grid vendors as well. Of course this doesn't apply to open source products.

I don't know about the rest of the world

[brokeninside]

But in the US, most schools would gladly accept the old systems as donations.

Re:I don't know about the rest of the world

[ray-auch]

In that case the old inefficient systems would still be running, using power, hence no environmental gain.

The only way to get the claimed environmental gain would be if the old systems were never used again - which then does raise the landfill etc. issues

Better yet

[tezbobobo]

no wait, imagine a PowerMac with a dual processor setup. Why, that would be... practically a super compooder.

that's power

[tehwebguy]

i bet this baby can pump 1.2 jiggahertz!

The processor after this

[Xerp]

Will obviously have the name "C1ala1s"

A new measure of CPU performance...the tree (T)

[ip_freely_2000]

We can add this performance criteria to our system selection process...how many trees are saved. Featuring the Sun chip which is the world's first Megatree (MT) performer.

Why I remember when I was a lad we had Kilotree performers, and we were glad for it!!!

Re:A new measure of CPU performance...the tree (T)

[amliebsch]

They weren't talking about saving trees (since we don't get our electricity from wood-burning power plants), they were talking about saving CO2 emissions. Since trees absorb CO2, it's a measure of how many trees it would take to absorb the CO2 emitted by the burning of fuel sufficient to power the chips. So it's really more of a "Tree-Reduction Equivalency Estimate". In this case, one MegaTREE.

Re:A new measure of CPU performance...the tree (T)

[TV_Slug]

But to truely break down the numbers, we'll have to know how large the trees being discussed are supposed to be. When chopped down, the wood from those trees would be measured as a cord which conveniently is 128 cubic feet of wood, so a MegaT, multiplied by tree size and divided by 128 will determine its MegaCords of equivalent burning power. For tree size, ask Sun what standard they were using in their million tree hyperbole... er, calculations.

Re:A new measure of CPU performance...the tree (T)

[codeButcher]

I'll rather wait until performance reaches the 3 brazillian Tree mark before I buy in...

So that's no effect at all, then

[Flying pig]

Or even positive because the emissions from the tractors used in planting all those trees. Trees don't reduce emissions, people do. They might reduce the environmental effects a bit...but they definitely don't deliberately fall on SUVs or jump out in front of trucks to cause a real reduction in emissions.

Re:So that's no effect at all, then

[Yartrebo]

Right on.

To be pedantic, planting trees (unless it's done on soil that was used for industrial agriculture, which has pretty much giving up its carbon already), will generally cause a release of CO2 from the ground. Even once the forest becomes mature, the net release of CO2 is positive in many cases (especially if the land used to be grassland).

But assuming that is ignored, a million trees:
  - Is nothing. Assuming they're Christmas trees, it's about a square kilometre. It's also about 1/100th of the annual harvest in the USA.
  - Is meaningless. Tell me in megatonnes of CO2 or gigawatts how much this will save, and if it doesn't equal a megatonne/yr or gigawatt, then it is just a drop in the bucket. Probably less of an effect than eradicating all spyware (thus causing less PCs to be replaced by lazy or ignorant or rich PC owners).

Re:So that's no effect at all, then

[cbiltcliffe]

No, I guess they don't, do they?

Re:So that's no effect at all, then

[zerocool^]

What if it conserves slightly more than a gigawatt? Say, maybe 1.21 gigawatts?

Will you travel back in time to eat your own words?!?

~W

Too bad

[CastrTroy]

Too bad you can't get windows to run on one of these things, or any microsoft software. The fact that they usually charge per processor would really be make the software a lot cheaper to buy. Just think, 1 processor that acts like 32. Anyway, it doens't matter since most of their software doesn't support that many processors to begin with. Stupid built in limits and all.

Re:Too bad

[Anonymous Coward]

I bet that they will never publish the chip specifications. Therefore Solaris will be the only OS that can run on these processors. Hey Sun, how about letting the open source community support Linux on your hardware. You'll sell more hardware if you let people know how to use it. Reverse engineering the sloppy code in Solaris won't be any fun.

Sounds like an anti-slogan

[suso]

How about: I'd rather plant a tree than run on a Sun.

Re:Sounds like an anti-slogan

[sheimers]

Do you meen an appletree? An apple a day keeps Bill Gates away ;-)

Misread Code Name

[FJ]

Am I the only one who read the code name as Viagra?

I knew sun was having troubles but not THAT kind of trouble.

Re:Misread Code Name

[ceeam]

Well, Suns are quite well-known for their decent uptimes!

Re:Misread Code Name

[frank_adrian314159]

I heard they were having problems keeping their margins up.

Re:Misread Code Name

[GrievousMistake]

I read it as 'Nigeria', but were sadly unable to come up with a decent 419 joke.

Re:Misread Code Name

[Anonymous Coward]

> Well, Suns are quite well-known for their decent uptimes!
What are you talking about? Mine goes down once a day!

Re:Misread Code Name

[anOminousCow]

That, and the post did say Sun was expecting to get a 'leg' up with this chip.

Re:Misread Code Name

[SiggyTheViking]

>>Well, Suns are quite well-known for their decent uptimes!
Not around here.
Our sun doesn't seem to stay up more than about half the time.
(It seems to vary a bit seasonally.)

Re:Misread Code Name

[Exter-C]

I have to agree with you. We have several tens of sun servers ranging from the baby all the way up to 880s and beyond. We have very good uptime even with very high load. However many of the applications that we use are all web based these days so that we can be platform independant.

NIAGARA FALLS!

[Anonymous Coward]

Slooooooowly I turned. Step by step. Inch by inch...

/3 Stooges fan :)

Re:NIAGARA FALLS!

[zipfaust]

LOL! I'm glad I wasn't the only guy who had that bit come to mind. :)

Three Stooges Forever!!!!!!!!!!!!!!

carbon dioxide emissions , errr O.....K

[zenst]

UltraSparc T1-based systems would have the same effect on carbon dioxide emissions as planting 1 million trees.

Thats great but did they offset all the marketing bumf about this; Now that surely made large dent into those newly planted tree's - caching in before even planted some might say.

There is also the factor of all the extra web/page hits and respective computer power and as such extra carbon dioxide ommisions related to that.

So in the end to be truly green you would just release the product and keep stum or the papers/marketing etc will offset the savings by eating up extra paper/computer resources saying how green things are.

How green is a new CPU that uses lower power; Well initialy less so due to hype and marketing using more paper/computer resources. This is up there with SETI and there search for extra human life by chewing up loads of CPU generating way more carbon monoxide and as such killing life on our planet in the hope of finding intelligent life on another planet. Lets face it if i was an intelligent alien would i want to be sat next to a snotty kid who farted alot or would I do my best to ignore them. Begs the question about what is actualy green and what is actuly intelligent.

Ultrasparc III Cores

[the eric conspiracy]

The Niagara uses Ultrasparc III type cores which have limited single thread performance. This limits this design to certain applications that are highly concurrent in nature. More interesting is the Next Gen Rock CPU which will have highly parallel Rock CPUs.

Re:Ultrasparc III Cores

[argent]

The Niagara uses Ultrasparc III type cores which have limited single thread performance.

So by "up to four tasks at once" I assume they mean "four register contexts per core" rather than something like hyperthreading?

More interesting is the Next Gen Rock CPU which will have highly parallel Rock CPUs.

But will they actually have better single-thread performance, or will they just support more concurrent threads? It's hard to see how you could resolve the "small register file" problem that register windows lay on optimisers... intel sidesteps it by applying massive amounts of silicon, but that's hardly an option here.

Re:Ultrasparc III Cores

[the eric conspiracy]

But will they actually have better single-thread performance

Yes, I miswrote the original comment - I meant that Rock would contain Sparc IV cores.

Does the million trees ...

[Anonymous Coward]

... take the energy used in producing these chips into account? Mining the silicon, growing the perfect crystals, making all the necessary chemicals, etc.

Wow

[pureseth]

I want one of these beasts for my home PC.. O_O Anyone see a price range on this thing? Also, I heard Intil is working on these kind of low power processors as well.. Maybe there will be some competition between them and Sun..

It's obvious you see...

[FF8Jake]

"Removing the world's Web servers and replacing them with half the number of UltraSparc T1-based systems would have the same effect on carbon dioxide emissions as planting 1 million trees." Because everyone who runs a server is a treehugger, this makes a lot of sense! Oh wait...

Coming to a spam email near you

[Anonymous Coward]

Want to buy some cheap N1ag4ra?

The Spelling of NIAGARA

[CrazySpence]

It is not Niagra. Thank you

Re:The Spelling of NIAGARA

[Anonymous Coward]

You sure? Me and the wife just got tickets to this Niagra Phallus I've been hearing about.

Re:The Spelling of NIAGARA

[aminorex]

It seems someone is being niggardly with their letters "A".

Why is it named Niagra?

[Unnamed Chickenheart]

...is it good at memory leakage? ^_^

Or is it perhaps not as low-power as they clame: maybe it require a huuuge current? ^_^

One should carefully name ones product. Its fate may stand or fall on it ^_^

Re:Why is it named Niagra?

[Black Perl]

As I'm responding to this, the ad box is running a Sun ad that states, "No wonder their name rhymes with Hell...". So I would conclude that Sun tends to pay attention to what the names rhyme with.

Also, Sun is pressing the reliability aspect of their solutions. Therefore, I'm willing to bet that Sun is hoping people will connect "Niagra" with a rhyming word with the connotation "stays up a long time." I'm not quite sure what that word would be, however. :-)

p.s. actually they seem to be using Niagara. Perhaps this is to avoid potential trademark disputes.

Re:Why is it named Niagra?

[Blakey Rat]

Your final score:

+2 funny comments.
-3 rage-inducing anime smileys.

Sorry, there's not a net gain here.

What does the marketing speak mean?

[Chilles]

"We at Sun find that we are unable to produce cheap fast processors so we label this expensive slow one as eco friendly and hope for the Prius effect. Perhaps Cameron Diaz will buy some?"

It is a pretty good performer in its target area

[hattig]

A 1.2GHz CPU at 70W that is roughly equivalent to three dual-core Opterons (>200W) (source in the comments at Aces Hardware). I suspect that this processor will compete against some of the lower-end Power5 systems from IBM.

With its support for Virtualisation (hopefully a better implementation than Intel's Vanderpool from what I have been overhearing in the techy pubs where I live) you could have some interesting hosting scenarios ... 'yeah, it's a dedicated machine you'll get, Sun, J2EE, everything, yeah!'.

planting trees versus carbon dioxide emissions

[John_Sauter]

...would have the same effect on carbon dioxide emissions as planting 1 million trees.

I hope everyone realizes that planting one million trees has no effect on carbon dioxide emissions. Trees absorb carbon dioxide as part of the photosynthesis cycle, they don't emit it.
John Sauter (J_Sauter@Empire.Net)

New server chip Viagra

[ardchoille]

Whoa, talk about "uptime".

Re:New server chip Viagra

[TheRaven64]

The spam I get only promises around 24 hours of uptime. Hardly ideal for a server...

Mandatory sed quote

[fm2503]

s/n/v/

Re:Mandatory sed quote

[Surt]

Awww man that cracked me up. Suvs saving power! Heh heh!

Or even easier....

[oliverthered]

Stop buying wood products that don't come from renewable sources, that's basically all hard woods.

Re:Or even easier....

[nelsonal]

Please mod this up, recycling paper has zero impact on cutting of forests, paper pulp "trees" are grown on farms.

Re:Or even easier....

[SCHecklerX]

He said hardwoods. I don't think that includes paper tree farms. There is a market for hardwoods rescued from the bottom of lake ontario, for example, when the virgin forests were cut down. Those trees are 100's of years old. Not exactly renewable.

Re:Or even easier....

[SCHecklerX]

'rescued' was a poor choice of words. Salvaged is what I meant.

Re:Or even easier....

[Anonymous Coward]

Right, and he was pointing out the common misconception that recycling paper saves trees.

GP: Stop buying wood products that are non-renewable. That means just about any hardwoods.

P: Right. Recycling doesn't stop deforestation of old-growth because most consumer paper products come from farmed trees anyway.

You: No, he said hardwoods and that doesn't include paper farms.

Oh, and build homes out of bricks instead of lumber while you're at it.

Re:Or even easier....

[Some Pig!]

Please mod this up, recycling paper has zero impact on cutting of forests, paper pulp "trees" are grown on farms.

http://kleercut.net/en/kenogami

Re:Or even easier....

[haggar]

Agreed emphatically! And as shocking as it may sound, throwing paper and cardboard products in landfills (and NOT burning them) is also the best way to "sink" CO2 using trees!

Re:Or even easier....

[Anonymous Coward]

Hmm, Tasmanian Old Growth Forests=>Wood chips=>high quality white paper. Fairly major export for that Australian state.

Re:Or even easier....

[oliverthered]

I said hardwoods because it's an easy target and didn't require going way off topic and including all the possible ways that you can prevent 'more' trees from being taken from non-renewable resources. I suppose I failed to mention that in a few thousand years time the rain forrest may grow back and we may have another couple of gallons of oil too.

If you want my opinion on paper then don't buy acidic wood pulp paper, it has a lifetime of between 20 and 70 years, traditional cotton (cloth) and hemp fiber papers have lifetimes of several hundred years just make sure that deforestation isn't happening to grow more hemp or cotton.

Re:Or even easier....

[oliverthered]

What do you think happens to them when you put them in landfill? don't they biodegrade releasing CO2 and, 8 times worse, methane. Burning will produce less methane, so it's generally better.

Processing the wood into Tar (for roads) and charcoal for burying may be a better way of storing biomass.

Re:Or even easier....

[nelsonal]

When you check both sides of the story, the company owns 1,000,000 acres of Canadian forest land (in the business overview section), and per your site clear cut 27,000 acres last year. They further claim to running on an over 40 year [implying 40-45 year]long rotation growing cycle (in the Timberlands section). Both claims seem to fit 27,000 acres is just over 1/40th of 1,000,000. That sounds an awful lot like a farm to me, but perhaps the longer crop growing cycle makes it something else. They also mention that hardwood logs are sold to veneer manufacturers they are far too valuable to be used in pulp (except the scraps). If someone has an issue with their farming practices on their own land there are dealers making a market in the company daily.

Re:Or even easier....

[haggar]

Well, I won't dismiss your assertion, but I think (not sure, OK?) that cellulose does not decompose into methane and stuff - that's a process that would require high temperatures in te case of cellulose. Now, if we talk about the lignin in paper, that would probably, I think, behave the way you mentioned, but ther is notso much of it in paper.

You said "charcoal forburying". What do you mean by that? Burying what?

Warning! Read the fine print

[Anonymous Coward]

Trees sold separatly!

I don't think that follows

[brokeninside]

If the old systems were servers, they were probably on 24x7. If used for education, they very well might only be used 8 hours per day, or even only 8 hours per week. It depends on how they are put to use. For example, if donated to schools, the schools turn around and sell the machines to salvage companies to recycle the steel cases, strip the mother boards of precious metals, etc. Certainly there would still be some landfill impact, but not as much as you're thinking there will be.

Re:I don't think that follows

[Anonymous Coward]

A rack-mounted Sun for every child!

Will they have to share the JBOD though?

Competition

[hackstraw]

With this new processor Sun hopes to get a leg up on the competition.

Oh, instead of the previous attempt to get a leg "up" on the competition with the "if you can't beat them, join them" method like: http://www.sun.com/x64/

Before anybody gets weird on me, I am not an AMD fanboy. I am kinda a Sun fanboy, but I have been very critical of them in recent years for a reason. They have been for years watering down their name and reputation. Hopefully this new chip is in the right direction. We will see. UltraSPARC IVs have been on the roadmap for years. Lets see some good stuff here guys. Memory bandwidth would be something nice to have as well.

Engineers are amazing...

[Fortyseven]

That's a hell of a long way to go to avoid our yearly Arbor Day responsibilities. "Let's build a chip, instead." ;)

Has Sun still lost the CPU war?

[peterdaly]

I had been reading articles recently about how Sun had fallen behind when it comes to CPU's and that their "rein" in the CPU business was almost dead. Their pushing of the Opteron was an admission of this.

I am curious if those people and industry analysts still think that, or whether this chip is enough of a leading edge performer to keep them in the game.

I don't understand the geeky details enough to know if this is really all hype or if there is really something earth shattering in this chip. That being said, it does sound nice.

-Pete

Re:Has Sun still lost the CPU war?

[Anonymous Coward]

Rein? Is that what you meant to say?
(courtesy of www.m-w.com)
1 : a strap fastened to a bit by which a rider or driver controls an animal -- usually used in plural
2 a : a restraining influence : CHECK b : controlling or guiding power -- usually used in plural
3 : opportunity for unhampered activity or use

Or did you mean Reign?
1 a : royal authority : SOVEREIGNTY b : the dominion, sway, or influence of one resembling a monarch
2 : the time during which one (as a sovereign) reigns

Bit of a difference a 'g' can make

In other news

[frostfreek]

Sun announced today, that they will be chopping down one tree for every new system sold!

nasty stuff-Everythings nasty

[Anonymous Coward]

"What about all the real nasty chemicals that go into the manufacturing process of chips .. eg arsenic and acid !!!"

1) You're *assuming* that there have been no process changes because of enviromental concerns. e.g switch to less toxic chemicals, using less overall, not using at all*

2) You're *assuming* that there's no recycling going one. e.g. conversion to a less enviromentally active compound, output of one industry is useful input to another.

*A company I use to work for switched to an orange-based chemical in it's PCB wash. While "nasty" chemicals (is there such a thing as a friendly chemical that can be abused with no ill effects?) will never be completely eliminated. Both quantity, and kind can be because of key decisions (a different PCB design can use less materials), and different overall process developments.

Am I the only one...

[Cow Jones]

... who misread the topic "New Server Chip Niagara" as "New Super Cheap Viagra"?
That would at least be an honest slashvertisments for a change.

oracle licensing

[Anonymous Coward]

So how does this work with oracle licensing? If we're gonna pay 8 x 0.75 = 6 oracle CPU license per chip, then it's gonna be a much more expensive machine than one that that uses chips with fewer, but faster cores. Unless there is some sort of resolution on that, i don't see how this can catch on with any oracle database servers, which i would imagine is a big potential chunk of the market for these CPUs.

Re:It is a pretty good performer in its target are

[10Ghz]

200W for dual-core Opteron? Show me the proof of that particular claim. In reality Opterons are nowhere near 200W. the SYSTEM might consume that much, but not the CPU alone.

Re:raw power - Fuzzy Math?

[bitrot42]

Niagara systems take the concept of dual core processors (with which most of you are familiar), and goes to an absolute extreme - building 8 cores, each capable of running 4 jobs simultaneously (4 threads), onto a single chip. Doing the math, we'll be delivering a 32-way chip, running 9.6GHz, which sips power (about 70 watts). JonathanSchwartz BLOG.

Since the actual clock speed is 1.2 GHz, wouldn't that be 32-way at 1.2 GHz?

You can't multiply the number of cores by the number of threads and the clock speed. It's one or the other.

Regards,

-Bitrot.

Million tree quote not accurate

[Anonymous Coward]

I read that what Sun said was that replacing with Niagara half the entry-level servers sold in the last three years would be equivalent to removing 1 million SUVs.

no laptops

[Anonymous Coward]

No company AFAIK sells laptops, although the name persists because of consumer marketing inertia. Companies have two styles of these types of computers, lightweight notebooks, and portable workstations. In a notebook, yes, this processor would probably suck, but perhaps used in a portable high end workstation it might be useful for some folks with niche application requirements.

Years of tuning their strategy?

[Anonymous Coward]

Did I read that right? So wondering around aimlessly is the same thing as "tuning a strategy."

I don't know if it's just me, but I've heard "5 years ahead of the competition" from Sun numerous times. Solaris if 5 years ahead of Linux or AIX or just about everything. This is 5 years better than that. Java is 5 years better than .Net. etc... I think that they teach it in their marketing program.

Buy the Numbers

[Doc Ruby]

I just noticed that a Sparc "Ultra 5" goes for $60 on eBay these days. What's the bang:buck ratio of a U5:$60 compared to the most economical of these new T1s? Because the right programming can get multiprocessing from a LAN of U5s, along with lots of redundancy (power, disk, memory, buses, etc) and even geographical distribution (disaster resilience). Maybe even for cheaper.

Re:Buy the Numbers

[Anonymous Coward]

with lots of redundancy (power

exactly

Re:Buy the Numbers

[multipartmixed]

You also need to factor in the cost of replacing the disks in the U5s. Commodity PC disks will do.

Those piece of crap Conners (rebadged as Seagates) that they shipped with have a high failure rate when used heavily. Not to mention that they are dog slow.

I use U5s in my lab, they're cheap and binary-compatible with my production platforms. Their slowness is useful in finding bottlenecks early, however I still need to test on other platforms to identify multi-CPU issues.

Re:Buy the Numbers

[Doc Ruby]

BTW: you haven't gotten debian-sparc to run on your U5s, have you? My 3.1r0a netinst ISO gets to "Starting kernel...", then the VGA goes black (but stays on). Help welcomed...

Re:Buy the Numbers

[LizardKing]

You could try NetBSD instead: http://www.netbsd.org/Ports/sparc64/. I'd recommend NetBSD 3.0, which has just gone from a beta to a release candidate, and includes support for all of the graphics cards found in Ultra workstations. A prebuilt release candidate can be found on the FTP mirrors under /pub/NetBSD-daily/

Re:Buy the Numbers

[Doc Ruby]

How good (complete) is the current NetBSD Linux binary compatibility? Because the reason for switching the U5 from Solaris to Debian is to use the same app codebase for all of my servers. I can replicate Debian directory structure with symlinks if necessary, but of course the apps need to find the Debian facilities in those directories, in addition to kernel API compatibility. Otherwise, I could just stick with (Open)Solaris.

Re:Buy the Numbers

[multipartmixed]

Nope, but keep in mind that the m64 chip on those puppies is kinda broken. It doesn't work properly at anything other than 8 or 24-bit -- and it doesn't have enough ram to be useful at 24-bit.

IIRC trying to run at 16 bits produces a black screen with dark blue writing, at least in Solaris.

If you can fiddle the debian loader like you can with knoppix (closest thing to Debian I use), try 1152x900 with 8-bit colour. Oh, and select a reasonable refresh rate...

Re:Buy the Numbers

[IvyKing]

Pulling a number out of the air - I'd guess that one of T1 boxes would easily keep up with a dozen U5's - a single T1 would have at most 2x the power consumption of a single U5. The bang for buck of the bank of U5's is better than a T1 provided that your power and airconditioning costs are low, space is not at a premium, your application works nicely in a cluster and most importantly: you have adequate RAM on the U5's. The DDR-2 memory for the T1's is a lot cheaper than the EDO ram used by the U5 (I know - I have a U10, which is the same motherboard).

As far as geographic diversity - by the time that you add up the costs of setting up even a minimalist data center - those costs will far outweigh the difference between the T1 and the U5 cluster. One exception would be a distributed multiprocessor system running in home networks.

Re:Buy the Numbers

[Doc Ruby]

Interesting analysis. If it's really 12:1 performance, or even 20:1 "cost performance" including office space, power, AC, etc, then a T1 breaks even with 20 eBay U5s at $1200. Sun hasn't released pricing (AFAICT), but it's probably safe to say it will be well over $2500, probably well over $5000, probably closer to $10000. Even $5000 would buy 80 U5s. It's a specious comparison for an entire market, because there aren't even 80 U5s for sale on eBay in a month. But it does put these developments in proportion. And shows that the T1 really has to compete with regular P4s, not just Xeons.

BTW, you don't have Debian running on your U10, do you? I can't get the 3.1r0a netinst ISO to get past the "Starting Kernel" after boot: my screen goes blank (but stays on). Ideas?

Re:Buy the Numbers

[LizardKing]

In my experience the Linux binary compatiblity in NetBSD is perfect. Until I finally got around to building a native JDK, I was running the Sun JDK for Linux under NetBSD 2.0.2. It performed flawlessly, and I only switched to a native JDK because I was bored one afternoon and decided to see if it would compile. I have also run Linux binaries of Mozilla and Firefox to get support for the Flash plugin, and these also performed flawlessly under various versions of NetBSD.

I'm subscribed to the NetBSD-current mailing list, and I see regular messages about tweaks and improvements to the Linux syscall compatability in the development version of NetBSD. This gives me considerable confidence that changes to system calls in Linux are quickly picked up on by the NetBSD developers.

Re:Buy the Numbers

[Anonymous Coward]

Oh yes, getting debian to install via tftp onto an nfs-root on my U1 was a serious, serious pain in the ass. It *is* possible, but you may have to google up alternative kernels, for install and running. Allocate twice the time you thought would be the worst case scenario - and good luck!

Re: appropriate for all types of workloads

[davecb]

My workstation is running 72 threads, and top (prstat) says that seven of them are active while I'm browsing, sending email and watching a script on a remote machine run...

Sounds like I could use 8 threads minumum, way more if I'm unit-testing a three-tier app.

PID USERNAME SIZE RSS STATE PRI NICE TIME CPU PROCESS/NLWP
1139 davecb 101M 87M run 49 0 0:04:58 4.6% mozilla-bin/6
480 root 56M 44M sleep 49 0 0:03:39 2.1% Xsun/1
700 davecb 15M 10M sleep 59 0 0:00:13 0.8% gnome-terminal/1
692 davecb 9896K 7304K sleep 59 0 0:00:29 0.4% metacity/1
2605 davecb 4664K 4392K cpu0 59 0 0:00:00 0.3% prstat/1
694 davecb 15M 11M sleep 59 0 0:00:22 0.3% gnome-panel/1
724 davecb 7928K 5288K sleep 59 0 0:00:02 0.1% galf-server/1
1110 davecb 13M 8920K sleep 49 0 0:00:01 0.0% gnome-perfmeter/1
705 davecb 2680K 2008K sleep 59 0 0:00:00 0.0% bash/1
690 davecb 3800K 2328K sleep 59 0 0:00:01 0.0% gnome-smproxy/1
696 davecb 28M 24M sleep 59 0 0:00:03 0.0% nautilus/5
274 root 1784K 1168K sleep 59 0 0:00:00 0.0% smcboot/1
264 root 1104K 752K sleep 59 0 0:00:00 0.0% utmpd/1
276 root 1784K 696K sleep 59 0 0:00:00 0.0% smcboot/1
275 root 1784K 696K sleep 59 0 0:00:00 0.0% smcboot/1
253 root 1048K 712K sleep 59 0 0:00:00 0.0% syshwd/1
1034 root 3776K 2088K sleep 59 0 0:00:00 0.0% automountd/3
250 root 1488K 1112K sleep 59 0 0:00:00 0.0% powerd/3
200 root 2240K 1456K sleep 59 0 0:00:00 0.0% lockd/2
577 n1sps 71M 6512K sleep 59 0 0:00:00 0.0% postgres/1
1133 davecb 1176K 888K sleep 59 0 0:00:00 0.0% run-mozilla.sh/1
Total: 72 processes, 165 lwps, load averages: 0.33, 0.30, 0.18

--dave (seriously biased, you understand) c-b

Any bets on..

[OneSmartFellow]

...how many trees would have to be planted to consume the carbon dioxide emmited by the production, packaging, and shipping of one of those servers.

My guess is that you could power one of those suckers for several years way and use less energy than building one.

Re:Any bets on..

[repvik]

Probably about the same amount of power that goes into producing a quad Opteron system. But this system uses less power when *in use*, than an quad Opteron system.

You know the chip is a flop when...

[ZuggZugg]

They choose to hype it's tree saving power versus it's actual performance.

Before you beat me up I agree that efficiency is an important thing to measure, and most systems benchmarks do not take efficiency into account...that being said I can bet that single threaded performance is lousy (I know they never claimed it was going to be great) but more importantly my bet is that aggregate throughput won't be that great either...my only rationale is that Sun is notorious for making crazy claims and the best they can come up with is a tree saving comparison. Puhlease.

It will be interesting to see what real world benchmark they can concoct that will show one of their single socket (8 core) servers (the T1000/T2000) using this chip outperforming a cheap dual socket (quad core) opteron server...maybe the the energy TCO will sway it. I think most commercial customers care about response time more than throughput...unless you're fully batch oriented. We'll see.

Flame away sun biggots.

Re:You know the chip is a flop when...

[tomstdenis]

One phrase: 8 cores + with 16/8 K cache per core and **one** L2 cache.

I'm sorry but I'd like to think that part of the SMP boost AMDx2 delivers so smoothly is that each core has a fair chunk of cache [64/64 + 512 or 1024] with a dedicated HT bus.

So is this an 8-port L2 ? What is the latency on it when all 8 cores are busy? etc... I think we'll find this core will suffer greatly from this point.

I mean look at the Pentium D. It's only saving grace is the 1MB of L2 each core has. The interlink bus is just the FSB and it ain't none too fast.

Tom

Re:You know the chip is a flop when...

[Anonymous Coward]

Sun admits that single thread performance for the Niagara chip sucks. What they claim is the aggregate performance of multiple threads is better, the thoughput computing claim. So if you ignore thread context switching overhead, pipeline stalls, etc..., a single core cpu would have to be 32x as fast as the Niagara. A chip that fast runs pretty hot and heat is becoming the limiting factor in chip design. Even if you could get a chip to run super fast, it would melt.

So for web servers, Niagara is probably a good fit. For gaming pc's, probably not, though we're reaching the limits of what water cooling and really loud fans can cope with.

Re:You know the chip is a flop when...

[ostiguy]

Not a Sun bigot - I am a MCSE holding Windows sysadmin. This morning, on the front page of the Wall Street Journal (below the fold), was an article about the trend of increasing wattage densities per square foot in server rooms, and the problems they cause. Right now, the Sun marketing guys that came up with the tree comparison are THRILLED with themselves because it is exactly in tune with an article that was in 2 million+ newspapers today.

Expect to see more marketing about power utilization and the resulting heat generation. Sun is doing it, and AMD should be much more aggressive about it.

ostiguy

Re:You know the chip is a flop when...

[ZuggZugg]

Agreed that power consumption is important...I personally think liquid cooling will come back as the solution of choice (think hermetically sealed CPU in fluorinet)...but I digress.

I would love to see future SPEC benchmarks include a power consumption measurement for CPU bound tests.

Unfortunately claiming that your CPU consumes very little power means nothing unless you can quantify that it was doing more useful work per cycle per watt than the competition. Using the power consumption argument alone is very misleading, but typical Sun marketecture.

Re:You know the chip is a flop when...

[adrianmonk]

One phrase: 8 cores + with 16/8 K cache per core and **one** L2 cache.
[ ... ]
So is this an 8-port L2 ? What is the latency on it when all 8 cores are busy? etc... I think we'll find this core will suffer greatly from this point.

I think you'll find that Sun engineers have thought of all that. In fact, here is a paper that specifically addresses all the points you've mentioned.

Basically, the answer is that it doesn't appear to be a problem for two main reasons. The first is that each core has up to four active threads at once. The cores have zero overhead for switching between one of the four hardware threads and another, as compared to continuing to execute the same thread. In fact, "thread select" is one of the stages of the pipeline, and the cores are designed to constantly switch between threads so that of the available-to-run threads, the least-recently-used one is selected on each cycle.

As a result, simply having to switch threads due to L2 / memory access will not impose any penalty. For the core to sit idle and any time to be lost due to waiting on memory, all four hardware threads would have to be unavailable to run. And there are four threads, unlike (say) Intel's chips that have Hyperthreading with only two threads. Increasing the number of threads increases the chance that at least one will be available to run. For example, with 2 threads that have a 50% chance of being available to run at any given time, the odds of having none that are runnable are 50%^2, or 25%. With 4 threads and the same 50% chance, the odds of having none that are runnable are 50%^4, or only 6.25%. The results are four times as good with 4 threads as they are with 2 threads.

Now, given that you have a limited amount of real estate on that silicon, the question then becomes: what is the best way to make use of it? If you have to choose between 4 threads per core with limited L1 cache and 2 threads per core with more L1 cache, which is a better choice for minimizing memory access problems? Maybe increasing the number of threads is a more effective strategy than increasing the size of a cache. Remember, the goal of Niagara is throughput for workloads that naturally have tons of threads. Blocking a thread to wait on L2 cache essentially doesn't matter as long as your core isn't sitting idle because all its threads are blocked. The PDF I linked to indicates they evaluated the projected workloads and found that increasing the size of the L1 cache didn't really increase the hit rate much, so it would seem that increasing threads per core might really be a better use of the real estate.

To elaborate on that for a second, the real serious hit is going to be when the system goes to memory. That is going to take one of the threads out of the game for a long, long time. Having to wait a while on L2 cache is not nearly as serious a problem, because although it does prevent the thread from being runnable, it only does so for a relatively short time. As long as the total bandwidth to the L2 cache isn't too low, access time for the L2 cache isn't going to make much difference either way.

The second main reason this isn't likely to be a problem is that the L2 cache is broken into 4 banks. Each of the 8 cores connects to the L2 cache banks through a crossbar interconnect, so each cache bank can be talking to a different core simultaneously. So, no, it's not an 8-port L2 cache, but it does support four accesses at once by interleaving, so contention for the cache probably isn't a significant problem.

Also, one other little minor detail to keep in mind: the SPARC architecture specifies register windows, and the Niagara implementation of SPARC has 8 register windows, with 16 registers per window (because when changing windows, you shift two 8-register positions). Each thread has its own set of register

Re:You know the chip is a flop when...

[Anonymous Coward]

It will be interesting to see what real world benchmark they can concoct that will show one of their single socket (8 core) servers (the T1000/T2000) using this chip outperforming a cheap dual socket (quad core) opteron server...

You just wait.

Re:It is a pretty good performer in its target are

[hattig]

"three dual-core opterons"

or...

[jbloggs]

" [...] removing the world's Web servers and replacing them with half the number of UltraSparc T1-based systems would have the same effect on carbon dioxide emissions as planting 1 million trees." Minus the energy & materials & waste costs of manufacturing the number of chips in the first place. This is in general quite a big problem for the environment. Replacing cars with hybrids is not necessarily a good thing, as creating a car will use about as much energy as it will consume over its lifetime. Using what already exists & recycling is often a better net gain.

Re: appropriate for all types of workloads

[LeninZhiv]

Um, yeah, but given that your two most strenuous processes are using 4% and 2% of your CPU respectively, with your machine posting mighty load averages of 0.33, 0.30, 0.18, I think you can safely hold off on upgrading to an 8-core processor any time soon--unless some of those threads aren't sleeping fast enough for you :-)

Where sun shines

[Anonymous Coward]

Need/want to model the known universe in real time?

Cant live without that 8+gb/sec of memory bandwidth?

Just gotta have enough CPU cache to store your kernel?

Feel the urge to replace an entire datacenter with 1 rack holding 2 servers?

Itching for that 128+ GB of ram?

Thirsty to do all this at 1/8 the electrical requirements of "mainstream" vendors?

Just need to know in the back of your mind that you have the toughest server in your city?

Orbital brain lazers and atomic supermen experiencing periodic slowdown?

Only have 120V AC current? .. but cant afford/justify cray?

Sun to the rescue....

-GenTimJS

i love ellipses

[Anonymous Coward]

more like

  [...] = "sun's spokesperson was probably exaggerating when he said that..."

if you run java...

[geoff lane]

...applications then even a desktop machine will benefit from multiple cores.

Arbor Day Foundation

[jbeaupre]

If you're willing to do the planting, the National Arbor Day Foundation will send you 10 trees for $10. http://www.arborday.org/shopping/Memberships/membe rships.cfm Get 10% of the registered users on Slashdot to sign up for this (and plant them) and you're close to a million.

Solving global warming with web servers

[PingPongBoy]

Global warming will not be solved by conservation measures. Mathematically, it's like running down a very fast up escalator. The population is growing rapidly with more people wanting to use technology in order to achieve a higher standard of living. Can't blame these poor schmucks, but web servers can come to our rescue!

Not only does the internet cut down the number of energy consuming activities such as trips to libraries, corporate espionage, commuting to desk jobs, visits to stores, etc., it has the potential to help us dig our way out of the destructive consumption spiral the whole world is executing.

Recently I read about some politicians on a visit to Beijing, China. They complained that their visibility radius was less than 100 metres because of the pollution. That country needs trees, but it's not stopping for anybody.

Global warming will not be affected by saving energy. Let's pull out the stops. Don't just replace web servers with more efficient ones. Instead, install them in bulk - the more, the merrier. We need faster computers. higher bandwidth, faster and bigger everything in the information infrastructure. It's time to fight fire with firepower. Ultimately we have to reach, through working smarter not harder, technological solutions that just don't pollute much, as well as raise a global mindset to limit the availability of useless or low value items that end up lying around ignored even though so many people are banking their livelihoods on producing junk.

greens

[tomcres]

so does this mean we should stop hugging trees and start hugging Scott McNealy?

New Slashdot Metric

[tekrat]

So, just as how many VW Beetles can fill a Hectar, or whatever that was, now we're going to have a new slashdot metric -- i.e., how many trees your multicore CPU is worth?

Or maybe it's how many trees can fit in a Beetle.. or How many CPUs in hectar?..

Have I worn out this joke yet?

Re:New Slashdot Metric

[chawly]

Yup ! But I got a giggle off it. But stop now - while you're ahead of the game.

The real question is....

[sweetnjguy29]

...will it run Linux?

Can you imagine a Beowulf cluster of these things?

[Anonymous Coward]

Oh come on, someone *had* to.

Vew Server Chip Viagara

[Anonymous Coward]

You probably means s/ni/Vi/

Re:It is a pretty good performer in its target are

[10Ghz]

well, if you could have three dual-core Opterons in a system, I would say that those Opterons would walk all over the Niagara. USPARC isn't really that great these days as far as performance is concerned, unless you compensate with lots and lots of cache. And doesn't Niagara use simplified USPARC?

The Niagara would run at around 1GHz and have 3MB of L2 cache. The Opteron-machine would run around 2GHz and have 6MB of L2. Yes Niagara has SMT but I don't think it' enough to compensate. And you can't compare clock-speeds directly, but 1GHz USPARC is not much to write home about these days.

Sun Fire T2000

[IYagami]

Acording to the register Sun Fire T2000 is going to be unveiled on 6th Dec with the first implementation of UltraSPARC T1 (see here and here)

8 cores old news.

[gwar11d2]

http://www.theinquirer.net/?article=12145 8 Cores! IBM had that years ago....then again maybe not as fast. Maybe AIX will run on Sun Hardware?

Re:8 cores old news.

[Anonymous Coward]

That is an eight chip module, not an eight core chip.

How many megatrees was that again?

[Nom du Keyboard]

same effect on carbon dioxide emissions as planting 1 million trees.

I see we've now created the megatree enviromental impact measurement unit.

Of course that still begs the question, are all megatrees created equal?

(Remember, according to Slashdot policy I own is post, and hence this word when used in this context. Sort of like how Microsoft owns "windows". As such I will soon go after the other 955 Google hits on megatree for possible infringement.)

Re:How many megatrees was that again?

[Anonymous Coward]

Of course that still begs the question, are all megatrees created equal?

Possibly, but I measure my processor power in bogotrees.

Re:How many megatrees was that again?

[adamdeprince]

And at the top of your google query: "Results 1 - 10 of about ... (0.59 picotrees)"

Viagra? Niagra?

[h4ck7h3p14n37]

Sun certainly needs something to raise it's earnings.

Re:It is a pretty good performer in its target are

[hattig]

As I said, the performance figures came from someone else.

Apparently in the tasks it is targetted at, this processor will perform on par with 3 fast dual-core Opterons.

Probably not anything floating point heavy, because apparently the chip has one FPU shared between all the cores.

However for multi-threaded primarily integer workloads (i.e., lots of server tasks) it is pretty good. 32 threads at 1.2GHz should damn well compete with 6 threads at 2.4GHz. I can see that you might be thinking 8 cores vs 6 cores, but with a relatively simple core and 4-way SMT per core you can probably keep each core pretty stuffed with work so that each core is getting overall much higher IPC than the Opteron (and indeed it would have to, by around 66%, to perform the same) which is single-threaded, and hence there will be a lot more unutilised clock cycles (despite being a wider-issue design, x86 rarely gets above 1 IPC).

The low Sparc of high heeled boys...

[argent]

The Sparc architecture has always had a problem getting good straight-line performance, because of the way the register windows constrain the compiler by exposing a small register set to the optimiser while still having the large context switch overhead of a large register set. Sun has long used multiple register contexts and microthreading-like techniques supported by the OS to get good multitasking performance despite these shortcomings, so this is a natural extension of the Sparc family.

And, yes, I'm sure straight-line single-processor performance will be nothing exceptional. But don't knock the impact of lower power... if it uses half the power of a comparable dual-core Opteron then you can fit twice as many processors in a rack, just because of the cooling requirements.

I also wonder what "handling up to four tasks at once" means. This could simply mean they have four traditional Sparc register contexts per core. TFA doesn't go into detail there.

Re:The low Sparc of high heeled boys...

[ZuggZugg]

Based on the reporting over at the register the densest you can go with these CPUs is 1 CPU per RU. So at most 42 * 70 watts or about 3KW of "straight" CPU power per rack...I haven't heard of any blade config that Sun has in the works. Unfortunately efficiency per rack/sq. ft. is dictated by the server form factor that Sun chooses...no different than other vendors except that there are many more server form factor choices for other CPU architectures than SPARC.

Again it all boils down to performance/per watt/$. Let's assume the capital cost to purchase is the same for the dual socket/quad core opteron X4100 as a 1U T1000...how many frames/web pages/compiles/queries/files served can the opteron do versus the SPARC T1? This is the precious info that Sun will probably never release and for good reason. Instead they'll say that their platform is 2.5 times more power efficient! So what, if I have to buy 5 times more servers to do the same work than I'm much further behind.

My understanding is that each physical core has 4 virtual cores, 3 of the 4 are in effect stopped until the one currently running stalls. So my understanding is at most you'll have 8 threads con-currently executing and 24 waiting. I imagine the OS scheduler in Solaris will be heavily tuned to make it all very smooth...likely preference will be to use the paramaters in the fair share scheduler to ensure no thread sits for too long.

Re:The low Sparc of high heeled boys...

[argent]

Based on the reporting over at the register the densest you can go with these CPUs is 1 CPU per RU.

If Sun doesn't take advantage of this option, and doesn't license or sell the processors to someone who will, it'll remain purely theoretical in this case. That doesn't mean that performance/watt isn't a valid metric.

Though I'd rather they took advantage of it by slamming more Niagara chips into a single computer than by promoting the evil of blades.

My understanding is that each physical core has 4 virtual cores, 3 of the 4 are in effect stopped until the one currently running stalls.

That's how "hyperthreading" works, as I understand it, but I suspect you could also describe multiple register contexts that way... the main difference being the role of the OS scheduler in the process.

Re:The low Sparc of high heeled boys...

[ZuggZugg]

My only bitch about this chip, is Sun hasn't released performance info.

Good Chip; Bad Angle

[MidKnight]

Sun has been talking about this puppy for a while now, and it's good to seem them deliver it. It does round out their processor strategy pretty nicely: AMD on the low end, and if you want obscene performance per-CPU at the high end you get this guy. I'll be interested to see some performance numbers.

Typical Sun though: crap-tacular marketing. What's the deal with the "eco-friendly" angle? See Sun's front page. Which CTO's actually care about that again? It's just stupid; saving the planet is a great corporate goal, but hopefully Sun is a bit more concerned with their bottom line, where they haven't consistently made a profit in 5 years.

Re: appropriate for all types of workloads

[aminorex]

> aren't sleeping fast enough

That is exactly the problem. Modern desktops need several threads in L1 icache in order
to be responsive. Is it bad design? Yes, but I can't help the politics of that. What
I can do is get lots of cores.

Believe me, once you've grown acustomed to KDE 3.4 or Windows XP on SMP, leaving aside the
sweet, sweet juiciness that is multicore SMP, you'll not go quietly into that dark night of
mono-threading.

Don't read too much into Linux load average numbers. Understanding their implications
relevant to performance, once you get past the crude first-order approximation, is a
delicate and subtle business.

Throw in some optical interconnects...

[sonofagunn]

Sun has also just signed a deal to start using optical interconnects in their supercomputers. These interconnects will be able to transfer data at up to 10 terabits/sec!

http://news.com.com/Deal+brings+optical+connection s+to+Sun+supercomputer/2100-1006_3-5947510.html?ta g=nefd.top

Environment-friendly?

[Ulrich Hobelmann]

Hm, maybe running Niagaras instead of Xeons might save some energy, but what about manufacturing all those chips in the first place?

spam

[Anonymous Coward]

Niagara - that the medication that is advertised in all the spam I get, right?

1 million trees = not that great a contribution

[loserMcloser]

Nickel mining giant INCO grows a quarter million seedlings a year just for its regreening efforts in the small city (and surrounding area) of Sudbury, Ontario, Canada.

Do The Math

[Anonymous Coward]

1M trees would occupy an area about 2 miles by 2 miles (approximately 4 square miles or 2,300 Acres). If all you're worried about is carbon dioxide emissions, then it's easier, cheaper, and quicker to plant the trees. Then you're not filling up landfills with the junked servers. Don't just take a flippant assertion like the one offered as truth without investigating if it makes sense.

Re:Apple need this

[level_headed_midwest]

That G4 is also far less powerful than the Niagra. We're talking about using these chips in servers, not laptops. Server manufacturers do want to reduce power consumption and heat output, but they need a lot more porcessing power than five-year-old laptop chips (such as the G4) can provide. 70W is quite low for a server as most server chips are at least twice as power-hungry as that.

Benchmark ?

[chevanne]

Hi Do you know whether Sun has performance numbers ? (specint, specfp, TPC...)

Acid comments in politics ??

[Space cowboy]

Is this a reference to Diebold ? [confused]...

Simon

Re:Acid comments in politics ??

[absinthminded64]

'Electionics' being mentioned in an acid oriented post made an attempt at humor on a Monday morning irresistible.

Who needs x86/PPC anymore...

[farzadb82]

With processors like this, i'm wondering if it beneficial in the longterm to just have CPUs that contain many cores and dedicate 1 or 2 cores to JIT x86/PPC/other instruction sets to native code.

Maybe Transmeta was onto something way back when...

VIAGRA!!!

[Anonymous Coward]

Call me nuts but when I first saw this I said, "They named this Viagra?"

Fuzzy Math

[Kadin2048]

In that case, wouldn't they be better from a marketing standpoint just to drop all the explanations of threads and cores and say "Sun introduces new 38.4GHz Chip! So bloody fast it'll make your head explode!"

NIAGRA FALLS?!!

[glassgnost]

Slowly I turned -- step by step -- inch by inch...

Is it just me.....

[carlmenezes]

or does it sound eerily similar to Viagra??
Heck...both of them are supposed to "make things look up" right? :)

carbon in plants all comes from atmosphere

[Anonymous Coward]

All carbon in green plants (ie, not counting mushrooms, etc) comes from the atmosphere. From the soil comes the nitrogen, phosphorus, water , etc. If there is CO2 coming from decay in the forest floor, it was once CO2 in the atmosphere first, anyway. Planting trees and plants is NEVER a bad idea, CO2 wise. That's why the destruction of the Amazonian rain forests is a crime.

Any carbon in a plant is not in the atmosphere.

Re:carbon in plants all comes from atmosphere

[Yartrebo]

It's more complicated than that. Grasslands typically store more carbon than temperate forests, but they store it in the soil. As the grass is replaced by forest, the carbon is released from the soil, more than offsetting any gains in wood mass. The above-ground carbon stored by trees and plants is fairly small. It's about an order of magnitude smaller than soil carbon (excluding peat and methane clathrates, which are found in soil, but are localized to specialized biomes).

Most reforestation has happened in the temperate zone and is entirely made up of primary forest (very poor at storing CO2). Deforestation is happening mainly in tropical and boreal forests (tropical forests by logging and reduced rainfall, boreal forests by increased winter and nightime temperatures).

But either way you slice it, a square kilometer of land isn't going to store much CO2 compared to a 1 GW coal plant, some of which can chew through ten million short tons of coal in a year, enough to cover that hypothetical square kilometer forest about 5-10 metres (about 16-32 feet) deep in coal, and that's just for one year.

You owe me a new keyboard.

[Anonymous Coward]

Mountain Dew(R) & saliva. All. Over. My. Desk.

The 409 chip

[BashDot]

Well, I misread it as "New Server Chip Nigeria." Don't ask how I got that from Niagara.

Now we have silicon scammers!

Re:The 409 chip

[BashDot]

Oops, that should have been 419 chip. Even used the preview button and STILL didn't catch it...

That should have been "one million acres of trees"

[ievans]

From the press release:
http://www.sun.com/smi/Press/sunflash/2005-11/sunf lash.20051114.2.html

"...research shows that UltraSPARC T1 processor performance could eliminate the number of Web servers in the world by half, slashing power requirements and having the same effect in reducing carbon dioxide emissions as planting one million acres of trees.(1)"
(1) Discover Magazine Vol 26 No. 08 August 2005

So, for all the Slashbots out there complaining about this bit of marketing hoo-haa, you'll have to up your rhetoric accordingly.

Any information about the TERA?

[renoX]

Sun's CPU isn't the first one to try using thread to hide latency, I remember that TERA's design was using even more thread, does anyone know if it has been successful in the marketplace?
Did-it work well?

1 million trees.... ?

[Davorama]

1 million trees... I can't get a handle on what good that would do. Can you tell me what impact that has in terms of how many libraries of congress could be printed on it?

Prices - thanks google

[anth]

You'll need to use google's cache to get to some of the pages found by the following:
http://www.google.com/search?q=+site%3Awww.sun.com +coolthreads
http://www.google.com/search?q=+site%3Awww.sun.com +t2000

A brief description of some systems here:
http://www.sun.com/products-n-solutions/government /configure/group/ch_throughput_servers_1.html

And "Pricing for FLORIDA STATE Customers":
http://66.102.7.104/search?q=cache:OvNX6_g9s84J:ww w.sun.com/products-n-solutions/government/florida/ Gov/5.html++site:www.sun.com+t2000&hl=en&client=fi refox-a

So matching the "order number" from the first with the "model" from the second tells us that the cheapest T2000 is "Sun Fire T2000 Server, 4 core 1.0GHz UltraSPARC T1 processor, 8GB DDR2 memory (16 * 512MB DIMMs), 2 * 73GB 2.5" 10K rpm SAS hard disk drives, 1 DVD-RO/CD-RW slimline drive, 2 (N+1) power supplies, 4 10/100/1000 ethernet ports, 1 serial port, 3 PCI-E slots, 2 PCI-X slots, Solaris 10 and Java Enterprise System software pre-installed (Standard Configuration)" and costs $8,295.

You can get the software too: "Solaris 10 3/05 HW2 Operating System - This special release is to install and run on Sun Fire TM T2000 servers. It should be used only on this hardware and will be superseded by the Solaris 10 1/06 Operating System once it becomes available."

bad comparison

[YesIAmAScript]

SUN describes it well, but people miss it.

They say removing two web servers and replacing them with one of these would cut the power usage way down. Yeah, sure it would, but it isn't because the CPU is incredible. It'd because you'd have half as many hard drives, north bridges and other ancilliary devices running as before.

But what happens if I were to take the world's servers and replace them with half as many Athlon X2s? Probably the same thing, especially if they were 90nm cores and the old ones were 130nm.

Good for SUN for coming up with a marketing angle, but it's just that, a marketing angle.

Re:Apple need this

[TheRaven64]

If you can turn the cores off individually, then it might be low enough. Have one or two turned on while you are on the road, and then all eight when you are on mains power. Two cores would give you a low latency in 8 threads, which might be enough for a laptop...

I agree with him, I guess

[Blitzenn]

"removing the world's Web servers and replacing them with half the number of UltraSparc T1-based systems would have the same effect on carbon dioxide emissions as planting 1 million trees.""

I have to agree. But I think that the savings is truely in the number of those new servers that would be powered down because most people wouldn't use them. Most of the code in use today wouldn't run on them. So it is true that if you replaced everything with Sun equipment, you could turn most of it off due to it's disuse and therefore save a whole lot of cash, as they suggest. Perhaps that is not what they meant, but it makes a whole lot more sense.

hummer vs corvette

[jbellis]

hummer gets about 8mpg
corvette gets about 25 /corvette driver

eco effects of creating all those chips?

[MMHere]

"...removing the world's Web servers and replacing them with half the number of UltraSparc T1-based systems would have the same effect on carbon dioxide emissions as planting 1 million trees."

but how does this compare to the resource cost of producing all those chips? how much energy is put into the fabs, and how much waste product is poured out into either the ground, rivers, or land fills [or, hopefully, cleaned up and recycled much as possible.]

the biggest gain is one they won't admit

[Bill Privatus]

Sun likely does not want the publicity they could get out of this. If you've ever been in a data center with a Sun E10000 "enterprise class" box in a rack...then you'd know that you can't get within 5 feet of one of these beasts without feeling like you're being irradiated as you stand there. The heat just takes your breath away.

It's incredible. "Sun" isn't just a name...And, "global warming" due to "the Sun" (in the computer room...) got to be running jokes.

If nothing else, perhaps these processors will run cooler - they certainly can't run much hotter!

Just my $0.02.

Re: appropriate for all types of workloads

[davecb]

I definitely want them sleeping faster (;-))

Joking aside, even with this light load Xsun is driven by Nerdscape, and lord knows Netscape could stand the ability to run a number of threads to keep up with pages I launch, instead of it's current sloth-like behavior.

When I'm doing useful work on the machine (running Teamquest Model, for one thing) it used to be unresponsive. I've since improved it a lot by enabling SRM and the fair-share scheduler, so I could run background jobs in a different project with a restricted share of the CPU.

. That's a good technique to ration out cycles, but I'd really prefer to have more to use. Then I could have responsiveness and performance.

--dave (still biased) c-b