Sneak Peek At Sun's SPARC Server Roadmap 113
The folks at The Register have gotten their hands on Sun's confidential roadmap from June, which outlines the company's plans for SPARC product lines. The chart has some basic technical details for the UltraSPARC T-series and the SPARC64 line. The long-anticipated "Rock" line is not mentioned. "We can expect a goosed SPARC64-VII+ chip any day now, which will run at 2.88 GHz and which will be a four-core, eight-threaded chip like its 'Jupiter' predecessor. This Jupiter+ chip is implemented in the same 65 nanometer process as the Jupiter chip was, and it is made by Fujitsu, a company that is in the process of outsourcing its chip manufacturing to Taiwan Semiconductor Manufacturing Corp. ... not only has Sun cut back on the threads with [the 2010 UltraSPARC model, codenamed Rainbow Falls], it has also cut back on the socket count, keeping it at the same four sockets used by the T5440 server. And instead of hitting something close to 2 GHz as it should be able to do as it shifts from a 65 nanometer to a 45 nanometer process in the middle of 2010, Sun is only telling customers that it can boost clock speeds to 1.67 GHz with Rainbow Falls."
Overcome by events (Score:2, Insightful)
The only things on Sun's roadmap now are signs to the effect of "Road Closed 1000 feet".
Re: (Score:2)
Re: (Score:3, Insightful)
it has throughput and power efficiency (Score:5, Informative)
But why? 10 years ago I thought sharing an 8 CPU Sun with a big devel team was a privilege. Now any decent Dell workstation has that. What does SPARC have over Intel? (No vague claims of superior "throughput", please!)
It has throughput. Back in 2006, when the first T2000 was released, a dual Xeon could handle 980 req/s from Apache and the T2000 could handle 15,000 req/s:
http://www.stdlib.net/~colmmacc/2006/03/23/niagara-vs-ftpheanetie-showdown/
http://www.stdlib.net/~colmmacc/2006/03/27/niagara-benchmarks-update/
At the same time the Xeon used a peak of 2.2 Amps, while the T2000 peaked at 1.2 A. Things have only gotten faster.
Throw-in on-board crypto, and you can do AES-128 at 38.9 Gb/s with a single socket (eight core) T5220:
http://blogs.sun.com/bmseer/entry/ultra_fast_cryptography_on_the
A T5440 can do 22,932 MB/s (183,456 Mb/s = 179 Gb/s):
http://blogs.sun.com/yenduri/entry/t5440_crypto_performance_numbers
If you're a site that cares about SSL/TLS, how many x86 machines would need to buy, maintain, and cool to handle that load? How many F5 load balancers/SSL accelarators would you purchase? According to F5's own data sheet, the 8900 (with dual 850W P/S) can handle 9.6 Gb/s--and you still have to buy web servers on top of that (more power)
So the T5120 can do roughly four times the raw encryption rate, uses dual 720 W P/S, and also do work as web servers. You're also using less rack space.
Let's also compare to AMD-based systems (which Sun also sells):
http://blogs.sun.com/bmseer/entry/web2_0_consolidation_sun_sparc
Now the Niagara (UltraSPARC-Tx) CPU isn't good for every work load out there, but if it's highly parallel then it's something that you should be looking.
Re: (Score:2)
Re: (Score:1)
Re: (Score:2, Informative)
That was only true for the first chip. The T2 series has 1 FPU per core.
Re: (Score:2)
Oracle + Niagara = expensive? (Score:3, Interesting)
If Oracle still charges per core, the Niagara approach of many core CPUs could be more expensive.
Looking at the roadmap they seem to be going fewer cores, or at least sticking with 8.
As for power consumption, I wouldn't bet on the Intel x86 always consuming more power than a SPARC for the same performance. They are a scary competitor. They keep introdu
Re: (Score:3, Informative)
Can Sun/SPARC keep ahead of them? They might only be ahead in SSL/TLS. And if that becomes a big enough demand, some taiwanese/chinese company start producing cheap pcie cards to do that
Crypto accelerator cards have been available for a long time. Don't know about the price though.
Or Intel could decide to use some transistors to do it - they have lots of transistors to play with on their chips, it's just a matter of priorities.
See "Sandy bridge", Intel's next 32nm chip, due Q1 2011, will have extra instruct
Re: (Score:2)
I have yet to see any numbers that indicate that UltraSPARC-Tx can be beaten by the same number of Intel- or AMD-based boxes when it comes to highly parallel workloads. (The main provisos are with regards to floating point. :)
At spec.org you can see the previous generation of Xeons caught up and started peeling away from T2 performance, ad the new Nehalem Xeons absolutely throttle everything else out there (similar to how Sun's CMTs leapt ahead at launch). The most interesting thing I learned from staring at spec.org charts is that sooner or later a vendor will pull away from the competition like this, and it never lasts. Sort the charts by available date, and you can see Sun/Intel/AMD pulling away at times. I think Sun and A
Re: (Score:1)
Just run it in a single core Virtual Machine, e.g. VirtualBox [virtualbox.org].
Re: (Score:2)
And if you run your DB single core, you'd probably do better running it on a powerful single core like i7 rather than a weak T2 core.
Re: (Score:1)
OK, well that is what I recommended. If I am not mistaken VirtualBoz and Xen are both paravirtualization tools. Maybe I am mistaken, but I said something along the lines of VirtualBox. I cannot imagine why it would matter actually. Perhaps you can elaborate?
Also, I concede it might not be a great idea. I was just theowing it out there as an option.
Re: (Score:2)
OK, well that is what I recommended. If I am not mistaken VirtualBoz and Xen are both paravirtualization tools. Maybe I am mistaken, but I said something along the lines of VirtualBox. I cannot imagine why it would matter actually. Perhaps you can elaborate?
Also, I concede it might not be a great idea. I was just theowing it out there as an option.
Yeah, Virtualbox is NOT paravirtualization.
Re: (Score:1)
Re: (Score:2)
Now the Niagara (UltraSPARC-Tx) CPU isn't good for every work load out there, but if it's highly parallel then it's something that you should be looking.
Highly parallel with *low* cpu needs.
Niagara is good at dispatch and switching, but not computation.
Re: (Score:2)
Re: (Score:2)
no, oracle said they would spend more "than sun does now". which is next to nothing for R&D since their sales have tanked.
Re: (Score:3, Insightful)
Unfortunately, that's looking more true every day. I remember running a network of Sparcs and bragging to my family members about how they (the Sparcs) were sooo much more powerful than PCs that we had in our homes. Seven years later I was replacing all our Sparcs with x86_64 Linux boxes... too bad Sun just couldn't keep up with hardware development. It would be nice if Oracle really did ramp up hardware R&D for Sun, but I can't see those announcements being anything more than reassurances to nervous
No Economies of Scale (Score:2, Interesting)
By contrast, though Sun Microsystems often boasted that it has -- actually, had -- the largest microprocessor team after Intel, the team could not design a chip that sold to hun
Re: (Score:2)
Believe it or not, but back in 1990, I worked on a Sun workstation with an Intel 80386 processor that ran SunOS and *DOS* ( http://en.wikipedia.org/wiki/Sun386i/ [wikipedia.org].
Re: (Score:1)
I have a Sun 386i too.
However, at the time, Sun didn't have the Sparc yet, and were looking for the migration path forward from the 68000.
Now, I can run DOS on my Macintosh SE/30 (bochs on NetBSD) and have.
Re:No Economies of Scale (Score:5, Interesting)
Yeah, lots of the original Samba code was written on Sun 386i's. Ah, memories :-).
Jeremy.
Re: (Score:2)
In the 1990s, Sun could have easily built their company on the unglamorous ARM RISC processor, but Sun management wanted to exhibit the "pride" (and arrogance) of homegrown technology
And they were right to do so. ARM focussed entirely on the embedded market and left companies like Acorn in the cold for workstation chips a few years later. They were lower power, and maybe cheaper too, but they were much slower than anything else on the market; much slower than. Sun's mistake was to choose not to compete with ARM; they had low-power SPARCv7 designs, but never pushed them into the mass market. If they'd sold a stack based on the *7 prototypes (low power SPARC+Solaris running in 1MB of
Re: (Score:2)
Sun kicked super-lower-power microsparcs out into the market; the market Did Not Want them. They cost too much (surprise!) and weren't as fast as offerings coming from basically everyone else at the time.
Re: (Score:1)
Re: (Score:2)
The problem is that's what UltraSPARC ended up competing against, despite protestations to the contrary from a lot of people. SPARC machines have been replaced by the boatload with x86 and x86_64, which is why the architecture is in trouble. When you benchmark an UltraSPARC III as I did with pystones against a 1.4GHz Athlon and f
Re: (Score:2)
Re: (Score:1)
For MIPS and Digital, they have hit the end of the road
For SUN, their end of the road is near
And I am afraid the same would be for the now fabless AMD
One day in not that long in the future we gonna wake up to the fact that only IBM, Intel and some Taiwanese companies (Nvidia, VIA, TSMC) gonna be the only one left still making power processors for the world
And if I am not wrong, IBM may end up not making chips as well
Re: (Score:2)
I want a SPARC IIIi running at 65 or 45nm with modern gig ethernet. It would be far more than I need for my apps. I would like it in a box the size of the Netra X1 and running any OS the old ones could run like Solaris 9. It would be cool if they were $1000 each like the old X1 or V100. I might buy several hundred in that case. Meanwhile I'm buying old X1 systems and putting in SSD and replacing fans and power supplies and hoping for the best.
And I have loads that are faster on the old X1 than the t100
Not a good idea to publish this (Score:4, Informative)
If it's confidential then the Reg shouldn't publish the details. Unless they want to give Sun's competitors a leg-up. I'm sure Sun's competitors marketing teams are happy to have this. [sigh]
Re: (Score:2)
Re:Not a good idea to publish this (Score:4, Interesting)
Then Sun should, in fact, keep it confidential.
I'm betting it was leaked to give some assurance to the customer base that there will actually BE a Sun in the future.
Re:Not a good idea to publish this (Score:4, Insightful)
Re: (Score:1)
Comment removed (Score:4, Insightful)
Re: (Score:2)
Which anybody who actually thought about it for more than 2 seconds would know that Oracle would be keeping SPARC and Solaris around for a LOONG time.
I don't agree. They might TRY to, but it looks very much like SPARC is out of steam. POWER is beating it like a pinata.
Sun couldn't keep SPARC on top no matter how they tried. Oracle pledges to spend more money, but it's not clear where the money will come from or that it will do any good.
Re: (Score:2)
The largest box is 4 proc. Power procs are designed to scale and they sell 32 and 64 proc systems (as does HP with Itanium). T2 doesn't go here and Sparc (as in UltraSparcs) gets beat easily.
Re: (Score:2)
What workload needs 64 POWER or SPARC procs anymore? More often than not if it needs that much CPU it is horizontally scalable anyway, in which case buying 2x as many T series boxes would be cheaper anyway. Most of the time the reason you have boxes with 64 CPUs installed is for partitioning with LPARs or domains.
And for scaling it all depends what you are doing with the box. We have an application which consumed a full 48 core E6900 (i.e the box was 100% on CPU) because it ran all its components on the
This is hardly a secret (Score:5, Informative)
Sun have been providing theses details to their Partners at the Sun Partner Advantage Summits, I got this info last month.
Plus Sun Partners just have to contact their Sun Sales managers and just ask for a Roadmap Session(Under Signed NDA)
The Register are just publishing what already is pretty common knowledge amongst most people working with Sun/SPARC hardware already, it won't give their competitors a huge advantage at all, the fact that Sun are already revealing this stuff to their wide partner network means that the development of it is well and truly in its final stages, and if their competitors are finding this out through The Register, then they really are not doing their jobs properly.
Re: (Score:1)
I had this image of Intel, Motorola, AMD, (I can't think of more) sending really hot Russian women to go and seduce the SUN engineers getting them to divulge everything. After I stopped my 007ish fantasy, I realized all they'd have to do is send in a pretty lady and have her just say "Hi" and those engineers would divulge everything - they are geeks after all.
Re: (Score:1, Funny)
Idiot. Sun has a lot of female geeks, too.
Peak??? (Score:5, Informative)
I'm just sayin'
Re: (Score:1)
Re: (Score:2)
"Peak" SPARC just like "peak oil".
And costs are astronomically solar. (Score:1, Troll)
And it will cost 50x the cost of cheap PCs?
Surely as google does it, its better to have 20 x cheap pcs each running E5400's or anything thats less than $100/cpu with $50 mb's
Everything eventually fails, its not worth spending 50x more for something that lasts 2x longer, when replacement costs are super
cheap, and the replacements are going to be even faster.
Re:And costs are astronomically solar. (Score:5, Insightful)
Google goes for the lowest watt per processing, the actual hardware cost is probably negligible compared to the cost of years of electricity for powering the systems and cooling the surroundings.
Re: (Score:1, Insightful)
And it will cost 50x the cost of cheap PCs?
Surely as google does it, its better to have 20 x cheap pcs each running E5400's or anything thats less than $100/cpu with $50 mb's
Everything eventually fails, its not worth spending 50x more for something that lasts 2x longer, when replacement costs are super
cheap, and the replacements are going to be even faster.
Google? How many tens or maybe hundreds of millions of dollars has Google spent developing the software that can run on piece-of-shit boxes?
It sure as hell is relevant to be able to buy one box that simple non-redundant apps can run on when the alternative is trying to pay massive amounts to develop fault-tolerant and redundant custom apps that can run on two or three cheap boxes.
Because unless you can run your software on lots and lots of boxes like Google does, it's cheaper to throw high-end hardware at
Re: (Score:2)
Google? How many tens or maybe hundreds of millions of dollars has Google spent developing the software that can run on piece-of-shit boxes?
Not many, but then Google has a problem that is naturally parallel with very few data dependencies. Not all of us are so fortunate.
Re: (Score:2)
My guess (just an opinion) is that it would absolutely require a large investment in custom software and manpower to create the infrastructure that google has created. And it makes sense for them and is probably worth it.
Re: (Score:2)
Clusters of machines aren't good for all problems, and Google doesn't use their huge clusters for everything... Some problems can only really be solved on a larger machine.
Hardware costs aren't the be all and end all any more either - one of the biggest costs is electricity (both for the machines and the cooling). It may well be that it is cheaper in the long run to have less of these than a huge cluster.
Anyway, to me it is quite clear why Oracle want Sparc and Solaris - have a good look at the Oracle produ
Leak to make Oracle look good? (Score:1, Interesting)
How'd they get this roadmap? More than likely from someone inside Oracle. Now when Oracle gets Sun and the SPARC chips are better than this, Oracle will get the credit for "saving Sun".
Or am I too cynical?
Re: (Score:2)
Also a SPARC/Solaris fan... SUN really needs to polish their stuff up. They have a ton of cool tech, but it's very inaccessible to the average administrator. I consider myself to be fairly smart, and willing to go great lengths to learn new things, but I'm not who they need to market to. I wish they looked at the big picture and made their technology easy to use with fewer gotchas. The problem with Solaris.. I can't believe I'm saying this.. is UNIX. It really needs to go. I'm not talking about thr
Doesn't mean they'll build them (Score:3, Insightful)
Re: (Score:3, Interesting)
On the first day I worked for Sun, an x86 implemented emulator of a SPARC ran at 2-5 % the speed of the SPARC of the day. Within 4 years, the same emulator on x86 of the day ran at about 50% the speed of SPARC of the day. By now, it may well be 1.2 times SPARC of the day. That is the nature of commodity electronics.
The story isn't SPARC vs x86; it is low-run specialized electronics vs huge commodity production. There is no craftsman-like advantage to the low-run specialty; this isn't furniture, it is et
Re: (Score:3, Interesting)
Re: (Score:2)
IBM seems to do ok keeping POWER up with and ahead of Intel/AMD.
Yeah but Sun never was IBM even though they seemed to think so themselves. IBM has a much broader range of products so they can prop up other business lines for a while if they have too. What's more they can sell during a downturn (buying IBM always being the safe bet) which is something Sun never could. Sun is very much a "boom" company, every time there's an economic slump they collapse like a bad souffle.
Re: (Score:3, Interesting)
IBM seems to do ok keeping POWER up with and ahead of Intel/AMD
Depends on your metric. Raw performance? Sure. Performance per Watt? Maybe. Performance per dollar? Not for most workloads. And don't forget that IBM's latest POWER chips use the same execution engines as the SystemZ CPUs, just with different instruction decoders and a few specialised parts unique to each design; the majority of both chips is the same. When you have customers who think $1m is cheap for a machine, this helps subsidise your workstation processors.
Re: (Score:2)
Does it really matter how they do it? Their bigger problem is going to be finding commodity customers if they don't continue providing processors for all the consoles. The low end brings them into economies of scale that permit them to upgrade fabs etc.
Re: (Score:1)
The big power boxes are really nice for a set of problems. we have a couple of big ass ones for oracle servers for very large databases and they work great. They do cost big $, as somebody said in a thread long ago when your data matters you get what you pay for. IBM makes a killing on these and the 'frames because for a lot of businesses it's easier to pay some big bucks now then later when your data is fucked. It's not as hard as you think to sell these guys, plus the virtualization is top noch, and d
Re: (Score:2)
And why do you think that is? Because they have been adding the mainframe ops to Power every generation so they can eventually have 1 single proc line instead of 2 (used to be 3 for servers, 4 if you include the mac stuff).
What is your point with this statement? That having al
Re: (Score:2)
Re: (Score:2)
Sun employees are not allowed to use Powerpoint.
CPU multi-threading (Score:2)
Can anybody give real life examples where the CPU multi-threading brings anything?
And please only real life examples: no theory, no official PR - I know them well myself.
Re: (Score:3, Informative)
Multi-threading per core helps with video encoding. I saw benchmarks just today at http://www.anandtech.com/weblog/showpost.aspx?i=642 [anandtech.com] showing the results of the same processors run against the same tasks with and without HT enabled. How many thousand more examples do you need to see?
Re: (Score:2)
Thanks!
Re: (Score:2)
How about any sort of web-server type task. I do development on web-based portal software that is highly threaded. Each thread doesn't due a huge amount of work but there are a lot of them (multiple threads per web server request) so having a machine that can run 128 threads (though each is fairly slow) easily outperforms a machine with much faster CPUs but only 4 or 8 of them.
Generally webserver type loads do better on hardware/clusters that can deal with lots of threads even if they aren't all that fast.
Re: (Score:2)
SunFire vs. blade? Quite unfair comparison.
Re: (Score:1)
Just Give It Up Now (Score:1, Flamebait)
Re:Just Give It Up Now (Score:5, Interesting)
Sun, if this is the best you can do -- 4 cores, 8 threads, arriving at 45nm just as everyone else is getting to 32nm
Sun's performance as a chip vendor is far better that your performance as a Slashdot troll. According to Sun's roadmap, a 16 core times 8 threads processor (128 threads just to be clear) at 40 nanometers arrives in 2010. That would be four sockets per blade, 48 blades per chassis for a respectable 768 multithreaded processors per chassis. As Sun says, it comes down to the TPC-C numbers. I'm no Sun fanboi, far from it, but I could be convinced by the right performance/heat ratio.
Re: (Score:2)
Sun tacitly admitted back in 2004 that conventional UltraSPARC CPUs were dead when it did the deal with Fujitsu to use SPARC64 and killed Millennium aka UltraSPARC V. Sun wanted to concentrate on more novel designs such as Niagara, which has been pretty successful, and ROCK which was a dud it would seem.
The Fujitsu SPARC64 CPUs are pretty competitive with IBM POWER and intel x86-64. Forget itanic. It's dead. Fujistu has always made some pretty impressive SPARC CPUs. They've always been enthusiastic about S
Re: (Score:2)
Re: (Score:2)
Sorry if this redundant, but the title is not very clear, we are talking about 8 threads *per core*, not 8 threads total like with the Intel i7.
Re: (Score:3, Interesting)
At this point, it looks pretty much the same as a (DEC) Alpha or Itanium roadmap.
Except for one thing: the SPARC circuitry is entirely open source [opensparc.net]. This has interesting implications, such as the fact that enthusiasts can build these things as FPGAs or even ASICs as fab costs come down to within the reach of clubs and schools. And emerging economies can fab these things by the bazillions without paying royalties. Not to mention big rich countries too. [theregister.co.uk]
Re: (Score:2)
Re: (Score:1)
Ultra 27- two PCIe but no video cards (Score:2)
The Ultra 27 was released with 2 PCIe x16 slots .... and it wasn't until we'd bought the damn things that we found out you can't put two FX-5600s in there- the case was designed to prevent it.
What's that got to do with their SPARC roadmap? Next x86 box we buy will be intel reference design. It's cheaper.
(not to mention there are bugs with the XVR-300 and the FX series of cards where you can't turn on 3 heads- it's 2 or 4 only)
Virtualisation missing (Score:2)
I'm missing indications about better virtualisation features, like I'm used to them on IBM gear. These days all high end installation I see are running tens to hundreds of virtual machines on a single server. It looks to me that virtualisation in this scale is not even on the roadmap.
Markus
Re: (Score:2)
Solaris Zones not doing it for you?
Re: (Score:2)
Solaris Zones, as I understand them, isolate applications from each other, but all are running within/on top of the same Solaris instance. As soon as you want to run different OS levels for the different apps or environments you are out of luck.
For example a new OS maintenance level is usually tested for a while in a test environment before being applied in production. Zones don't help here.
Often we have also incompatible prerequisite requirements of different apps (3rd party apps are terrible in this res
Re: (Score:2)
The T-series (sun4v platform actually) have LDoms which are very similar to LPARs, but a bit more simplistic in their implementation. You can virtualise storage, networking on a control domain (i.e like a VIO server) and create domains out of the available threads and memory on the box. So with this you can do individual OSes in each LDom. It even now has dynamic migration where you can migrate a live running LDom between two machines (akin to VMWare Vmotion or the LPAR equivalent, the name of which esca
Re: (Score:1)
Re: (Score:2)
I never used Solaris zones, never used the similar AIX 'workload partition' either. But I'm aware of what they do and and how they work.
I (my) experience, if you need isolation, going the entire way and use a separate VM/LPAR with its own OS is the better solution. This is why I am missing this tech on Sun's high-end servers and don't understand that they are not even seem to plan to catch up in the future.
Markus
Re: (Score:1)
Re: (Score:2)
Yes, good enough is the enemy of perfect everywhere. And I suppose if you know your hammer well you tend to see problems as nails.
In my area there in not much Sun high-end left (1-2 E20k) and plenty of fat IBM boxes (>50), this might explain part of the lack of customer interest...
Markus
Re: (Score:1)
By the way, my "area" computing potential are the measly computer labs at school, but thanks for thinking I'm higher up than I am.