Big-Iron to Open Up for AMD
vincecate writes "Traditionally the key chips that have allowed companies to
scale multiprocessors to large numbers have been proprietary.
Some examples are the
Cray SeaStar,
SGI NUMAlink,
HP sx1000,
and the
IBM X3/Hurricane.
This proprietary paradigm is about to change to a more open one.
Two companies have developed key chips for
building large Opteron multiprocessors,
and they will be
commercial off-the-shelf parts.
PathScale has
released
InfiniPath
which can be used with an
Infiniband
switch to make
a high-bandwidth low-latency interconnect for a
supercomputer cluster.
The other company is
Newisys,
which
will soon release
the
Horus chip.
This chip will make it possible to build 32 socket
(64-core) shared memory Opteron systems."
It's about time! That z990 under my desk just isn't fast enough :-)
Never ask for directions from a two-headed tourist! -Big Bird
.... an Alienware game system with this chipset by the end of the week.
This is my opinion. To make sure you don't steal it, it's covered by the DMCA.
Give it another few months and I'm sure Sun will have some server with an obscene number of opterons in it, if thier current direction is any indication ....
-GenTimJS
... k, maybe not. Can't afford one anyway :-(
Real Daleks don't climb stairs - they level the building.
Why roll your own when you can get the pricing of mass produced processors and focus your value add on better software, interconnect and optimization? Scale it up and down. Bring on a practical desktop cluster for video rendering ang Foling @ Home!
I think somebody doesn't know what 'Big Iron' is if they think AMD chips have anything to do with it.
SeaStar and InfiniPath (and don't forget the XD1) are great for building non-cache-coherent clusters, but those are mostly useful for running specially-written scientific applications.
Horus is used for building Opteron ccNUMA machines with one OS instance that can run any Linux or Windows apps. It's a very different solution for a different market.
Or...you could go to eBay and get yourself a 64 processor Sun E10K for the price of a handfull of PCs. This technology's only been around for, what, 10 years now?
Opterons == "non-proprietary" Ok, cool. So are Sparc.
Just wanted to point out that the link to Newisys is just a blurb stating that AMD is releasing the Horus chip, and doesn't really have anything to do with Newisys, other than the fact that a couple of the people behind the AMD Horus release used to work there.
Oh, and the Horus link is a PDF whitepaper... please warn when a link points to a PDF.
"Trolls they were, but filled with the evil will of their master: a fell race..." -- J.R.R. Tolkien on Olog-hai
You kids, with your ultrasparc risc processing synchronous hypermultithreading vax/vms redbox pbx mumbo jumbo and your Ska music. For Christ's sake, cut the cotton-pickin' bullshit and tell me which stocks to buy and which to short. Oh and that AMD "capturing" the retail market tip the other day? Thanks for costing me six thousand dollars, my wallet was too thick and giving me a bad back. Christ.
For a minute I was excited I might get to pop an Opteron into my PDP.. that thing really lags when you try and load up Wolfenstein. Or "Adventure" for that matter.
Anyone know what the memory model for Horus is? Hopefully it isn't processor consistency type 2 which is all of the disadvantages of the TSO memory model with none of the advantages.
really a 'large number'?
A pizza of radius z and thickness a has a volume of pi z z a
My experience suggests that Suns compilers beat out GCC on single processor machines and more so on 8 and 16 processor systems.
The benifit of Sun's mature sparc compilers might let you squeeze more performance out of a sparc box. Although of course the opterons have a higher clock speed in the first place.
China has had that system running since June 2004: http://www.top500.org/sublist/System.php?id=7036
Which makes my wonder, how much of the post is really news?
PathScale has released InfiniPath which can be used with an Infiniband switch to make a high-bandwidth low-latency interconnect for a supercomputer cluster.
This is news? We've been using an Infiniband-connected Opteron cluster for almost a year now. I got bids from half a dozen companies willing to sell us one. This is old, established tech.
Operton is still proprietary, isn't it?
A 32-way SMP dual-core opteron box is a serious threat to Sun Enterprise boxes with 64 to 128 UltraSparc, even the hardware partitioning doesn't mean as much when you can just use two or more x86-64 boxes at probably less than half the price. For that matter, it also attacks HP's "superdome" Itanium2 servers and some of IBM's Power5 and Power6. The closed architectures and the proprietary Unix(tm) they run are in deep doo-doo
Proprietary is proprietary. AMD chips are no more "open" than any other vendor's chips.
The swiftness of the processor is still talked about today...
Well, it's about time people took the opteron seriously. But, it's not really designed for this. For best results it's a cpu designed for 4way (8 core) systems. They will need to do some finagling and commit some sins to get it to work, which almost always leads to an architecture that's less than optimal. This will lead people to believe that AMD isn't up to the job, since "oh yea, those AMD mainframes are garbage" when they don't get it to perform as well as the people building these things obviously think they should. I'd lay money that Intel is designing a 15 pipe interconnect for their 64 bit chips so they can swoop in and save the day when this idea fails.
I think it might be better for AMD to come up with a new chip that has more interconnects, and design the chip for the job, instead of designing the job for the chip.
Things almost always don't work right when you try to make them do stuff they aren't designed for.
It's almost as if Intel put these companies up to this to generate FUD about Opterons... Maybe the Microsoft/Linux fiasco with the FUD has me suspicious of something innocent.
There is no better marketing tool than to put the competition to a task it can't possibly do well and point saying "see I told you AMD is no good!".
If I were Hector Ruiz, I'd publicly say that "..these people are doing this on their own and the opteron is not designed to do this, so it will never work well."
l8,
AC
Umm, I know there's this odd phenomenon where many people tend to label any processor that's made by either Intel or AMD "non-proprietary" and any processor made by another company "proprietary", but even still this article is a little silly. SPARC processors have been in use since the late '80s, most people consider SPARC-based machines "Big Iron", and the SPARC processor architecture is fully open -- anyone who wants to can make SPARC processors. SPARCproductDIRectory lists a bunch of companies who currently do. In fact, there are probably just as many (if not more) SPARC manufacturers as there are X86 manufacturers.
This just in: Mac users say 64 core Opteron server will be almost as fast as the new Mac G5.
AMD and Intel have introduced multicore chips to the consumer market. It won't be too long now before there are millions of home computers with SMP class processing power. I suspect we'll start seeing games that take advantage of multiple processors by the end of next year. Whether they scale past 2 cores is an interesting question.
Go slashdot!!!!!!!
More to the point, supercomputers are not mainframes, which were (and still are, at least in most circles) the only systems that were traditionally associated with that term, and most of the folks who assert that mainframes are dead are not working for (and have never worked for) the companies who have traditionally used Big Iron in the first place.
Those of us who still work in the airline, banking, or insurance industries know that most of the current mainframe systems aren't going anywhere. They run on Big Iron for a reason.
Eventually they will be replaced, as everything is, but IMO their eventual replacements will be mainframes in everything but name -- ultra-secure, reliable, recoverable systems acting as large-scale corporate data servers.
Mainframe/UNIX Bit Twiddler and long time Windows/Linux Hobbyist.
The Theorem Theorem: If If, Then Then.
How is Opteron not proprietary? Only Intel and AMD can manufacture AMD64/EM64T chips. These chips (as with their x86 brethren) are among the most proprietary of chips, up there with Power 5. The only chip that's anywhere close to not being proprietary is the SPARC, which just about anyone can make (though only Sun and Fujitsu seem to).
So where I can buy the AMD server with near full redundancy?
Or the server which can run highly debugged application written in mainframe assembler in 60's or 70's ?
Or atleast AMD computer with SINGLE memoryspace atleast 1TB in size?
And also how many decades of uptime is for the operating system which is used with the new AMD computer?
The horus is more or less getting close to midrange server in number of processor while it won't bring it to the reliability requirements of midrange server, to get that it would have to run its own memory controllers instead of cheap ass opteron controllers which lack for example hotswappable memory.
Sure you get speed, but after taking the speed there is eventually a crash.
The big iron is all about gettin continuing to function no matter what comes.
Only problems outside of box, like earthquake or something similar could bring it down.
Yeah. AMD is doing just fine...
Its eating the cheap ass market, not the big iron.
The price is cheap and its bought where the crash proof means better than windows which is like saying saying its unsinkable since it does better in open seas than normal rowboat used in lakes.
Lets put it this way. x86 is just used in low end boxes and in clusters of lowend boxes. And those things are not for everything. They can do much but not everything. They are cost effective when you compare only the purchase price. But not so cost effective when downtime costs a lot.
There is probably order of magnitude or TWO orders of magnitude of what joe slashdotter thinks big iron and what businesses have in big iron as in price range.
Emacs is good operating system, but it has one flaw: Its text editor could be better.
If I were a betting man, I'd give you my trousers.
Stick Men
Hey, you lost your kanji .signature?
Research Non-Uniform Memory Access (NUMA)
Ma-ya-hee, ma-ya-hoo...
yeah, one day I logged on and my signature had been reduced to a BR tag. it makes me sad, but it also frees me up to change my user info -- in order to preserve the Kanji sig, I had to leave my old user info untouched.
have fun at +2 you whore,
--sa
No why -- just how the instruction mix turns out. One example is PostgreSQL. Where Opterons might be 15% in MySQL, MS-SQL, DB2 versus a Xeon at the same price range, it's 100%+ faster than a Xeon in PostgreSQL. One of the things noted was large context switching was a big penalty on Xeons so there's been an effort to reduce it.
Real World Tech has a great interview between David Kanter and two of the engineers working on Horus. If you're interested in actual information about how it works and what it does, its good reading.
BTW, Linus Torvalds posts to RWT :-D
Causation can cause correlation
sorry...
Oh fuck yeah! A paradigm change! I fucking love paradigm changes, and shifts too! This is fucking awesome, we should have a paradigm shift party! Fucking roxxor!