Slashdot Mirror
Student and Professor Build Budget Supercomputer
Luke writes "This past winter Calvin College professor Joel Adams and then Calvin senior Tim Brom built Microwulf, a portable supercomputer with 26.25 gigaflops peak performance, that cost less than $2,500 to construct, becoming the most cost-efficient supercomputer anywhere that Adams knows of. "It's small enough to check on an airplane or fit next to a desk," said Brom. Instead of a bunch of researchers having to share a single Beowulf cluster supercomputer, now each researcher can have their own."
A beowulf cluster full of these!
(Okay, now back to responsible mature posting)
Hopefully I didn't put any [] around my words.
It's just four motherboards sitting in a single frame. connected by an ethernet switch.
True supercomputing machines (sun, ibm) have a little bit better interconnectivity between the components than a mere 1Gb/s line. This can serve its purpose though, VASP will run wonderfully on it. GAMESS probably as well.
B.
Every experiment which ends in a big bang is a good experiment.
It's small enough to check on an airplane
With security concerns nowadays, it's the amount of cables coming out of it that worries an airline, not the size or weight of this machine.
Virtual Betting on Facebook for non-geeks.
It looks rather fragile, quite like the iRack (http://www.youtube.com/watch?v=xcjLEwZqcQI), and I dont think it would survive checking in on an airplane given how some suitcases looks like at baggage claim.
Cool achievement nevertheless.
Probable impossibilities are to be preferred to improbable possibilities.
Aristotele
They just linked 4 motherboards together. My cat could do that.
..can you play Minesweeper on it?
And it looks like they'll be running into heat buildup issues. An enclosure ventilated by one or two desktop fans would have provided sufficient cooling. Mere convection (outside of the tiny on-board fans) is often not enough. The Sun E450's were well ventilated machines, with a clear air path going from the front to the back. The temperature monitors (ambient, cpu (x4), PSU (x3)) were useful as well. One was used for a long time at Stack (www.stack.nl) as a room temperature monitor.
B.
Every experiment which ends in a big bang is a good experiment.
Now Microsoft have their next development target for Office.
Deleted
As seen in other comments, this isn't quite so extraordinary. If you can't build one too (given the $2500), you probably have to turn your geek badge in. So on to the interesting question - what are we gonna do with them once we build 'em?
...
I'm thinking along the lines of hosting my very own MMOG in the basement. Or maybe decyphering cell calls in real time. Or
Well?
Yes, I am a biological organism. All rumors to the contrary are just that, rumors.
is it powerful enough to run Windows Vista?
I am impressed with how amazingly lame this story is. It should have been entitled, "College Senior and Professor discover Ethernet, MicroATX, and PXE boot. Funding dried up before paying for cases. News at 3 am because we can't find anything else to report."
Honestly, our whole research lab is filled with PXE booting MicroATX computers connected via ethernet. And I guarantee that four "nodes", aka Linux PCs, are cheaper than $2500. Whoop-de-freaking-do.
So this Microwulf is about as powerful as my laptop. Well done. Glad it looks so good.
This seems pretty similar to the way google builds their racks, with just mb's and no cabinets. What would have been really cool was if someone made som e kind of network driver for a pci express slot, with them being able to use external cables, is it possible to use a dedicated pci express slot as a interface to another computer, skipping the network bottleneck ?
Doolittle :
Bomb no.20 : To explode of course.
Out of curiosity, what actually constitutes a supercomputer? Is there a certain threshold of OPS that must be achieved above and beyond what's considered a desktop or entertainment solution (five times, ten times)? Is it a matter of hardware structure? Degree of specialization?
I can't find anything that clearly outlines the criteria.
What's going to be used for? Well, Prime number hunting, pi crunching, computing obscure mathematical constants 99.995% of the US hasn't heard of before....creating my own fractal Deep Field images...trying my hand at cracking RSA numbers with GGNFS.
to name a few
Not all the comps are old, old. Several are Ghz boxes.
...this is *hardly* a supercomputer. This is 152.57 times slower than entry number 500 on the Top 500 List. There isn't a nice neat definition of what a supercomputer is anymore, but "capable of running Beowulf" isn't it. Leaving aside the more custom machines that the company I work for (and a few others) build, there are plenty of Linux clusters that *do* qualify. The fastest one seems to be number 8 on the current Top 500 list (a Dell Infiniband cluster at NCSA).
Go Badgers! -- #include "std/disclaimer.h"
One of the problems with supercomputers is that there aren't really very many of them, because of the size and cost. It means that the tools you use to run your supercomputing applications are similarly unusual. The skills to use and develop on parallel systems are then equally scarce. Access to a supercomputer isn't exactly common.
Microwulf could make all of the above common. For the price of a high spec PC. The commodity nature of it could bring super computing and super computing applications to the masses.
Then you can scale your application from microwulf to miniwulf to superwulf with little more effort than installing it on the bigger machine.
Course, they'd have to produce a commodity pre-built system.
Deleted
I see a few people making the expected "It's just four motherboards wired together with Gig E"-comments. While I won't object to that, I'd say this is not about a groundbreaking evolution in hardware, more a case of demonstrating what's possible today with COTS parts. Adding to that the compact packaging, and the ability to run off of a single power cord, it's a nice setup IMHO.
While it does not have the interconnect of "true HPC" hardware (a bit of a fleeting distinction, but bear with me) it'll surely be suitable for a lot of the simpler, yet still compute-intensive tasks out there ("simple" here meaning not needing a lot of intra-node communication).
On the flip side, it might fuel the "hell, I'll just build my own cluster"-mentality going around these days. I work in the HPC group at a university, running linux clusters, IBM "big iron" and a couple of small, old SGI installation, and we certainly see a bit of that going around. Problem is, sure, the hardware is cheap and affordable, but getting it to run in a stable and sensible manner without spending large amounts of time just keeping the thing together is a challenge, mainly due to the immature state of clustering software. As many researchers are not exactly keen on spending time solving problems outside their specific field, they're usually better off letting somebody else administer things, so they can just log on and run their stuff.
But for individuals and small groups of people who are computer savvy enough to handle it, things like these are definately a "good thing" (TM).
The more computing power is available in the world, the less it will be used to its potential. If everyone had an Earth Simulator in their basement, how much of that power would be wasted?
Not saying that proliferation of computers is bad, just food for thought.
-:sigma.SB
P.S. SETI@home, Folding@home, etc. are cheating. :P
WARN
THERE IS ANOTHER SYSTEM
Look at this design: http://www.mini-itx.com/projects/cluster/. It uses DC-DC converters on each motherboards (mini-itx, so low power), a single 12V PSU and a UPS for regulation:
bundaegi is good for you
Is 26 GigaFlops significant anymore? I hear that the PS3 can do 20-25 from Folding@Home people. And it is only about a 5th the price. But I hear so many different numbers that I can no longer make sense of them. Why do they bother comparing with DeepBlue, an over 10 yr old super computer? Can anyone with a PS3 can report what their PS3 with Yellow Dog Linux is doing? And what are the numbers for the latest desktop processors? Any recommendations on software to benchmark in flops for my own computers?
"a portable supercomputer with 26.25 gigaflops peak performance, that cost less than $2,500 to construct, becoming the most cost-efficient supercomputer anywhere that Adams knows of."
Dear "Dr." Adams,
One PS3, converted to Linux, has been demonstrated to have a performance of around 100 GFlops. That is significantly cheaper, more cost efficient, and higher performance than your $2,500 pile of crap. Plus, it comes with a shiny case. For $2,500, you could even buy 4 of them and wire them together into a small cluster. That's around 16x the cost-efficiency of your rats nest. Your claim has been directly refuted. Maybe you should read more modern research conference proceedings.
Sincerely,
The Research Community
CPU: AMD Athlon 64 X2 3800+ AM2 CPU $67.50 * 4 = $270
Main Memory: Kingston DDR2-667 1GByte RAM $48.49 * 8 + $4.99sh = $392.91
Power Supply: (can't beat price): $76.00
Network adapter (node to switch): (cant beat their price) $164.00
Network adapter (switch to node): (cant beat their price) $15
Switch: Trendware TEG-S80TXE 8-port Gigabit Ethernet Switch $46.99+$7.04sh = $54.03
Hard drive: Seagate 7200 250GB SATA hard drive $69.99
DVD/CD drive: (can't beat their price): $19
Cooling: (can't beat their price): $32
Fan protective grills: (can't beat their price): $10
KVM: (can't beat their price): $50 Grand total (incl. 15 in hardware): 1416.89 $1000 saved by using Newegg!
along the same lines... i have a calculator that is more powerful than the fastest supercomputer 50 years ago. I should send my findings.
I'm to lazy to run the numbers tonight to compare actual speeds but our dual CPU four-core Xeon (8 cores total) servers cost around $2500 each to build. Looking at their specs I doubt they could be doing much better and they require special clusterish programming.
At what price learning? At what cost wisdom? The price is a man's peace of mind, and the cost is his life.
Move along. Come on, even Hollywood studios can put out more flops than this.
Calvin College most recent headline was Bush as the commencement speaker. Maybe Jesus helped build this supercomputer (sic)?
www.itjerk.com
gigaflops, schmigaglops.
/.
this is
i thought performance was measured in fps?
Revolutionary? Everything old is new again...
http://www.mini-itx.com/projects/cluster/
http://news.taborcommunications.com/msgget.jsp?mi
http://www-03.ibm.com/systems/bladecenter/ -- a 7U chassis that holds 14 blades, and is a bit spendy, but not completely unreasonable for some situations
http://www.linuxjournal.com/article/8177 -- My personal favorite, this page talks about several small portable miniclusters that have been made over the last six or seven years...
Yes, 8 cores of Athlon64 is faster than 8 cores of low power VIA CPU's from several years ago, but the concept isn't revolutionary, and there isn't a lot of headline worthy engineering that goes into a project like this... I'm sure it's a very handy tool, and I'm not suggested it shouldn't have been built, or that it was entirely trivial to build, but in the end, it's just four ordinary motherboards and ethernet.
Anyone know if there has been a top500-compatible measurement of a PS3? If PS3 costs about $500, one could build a "ps3wulf" with four nodes and some network equipment for $2500. Anyone have any idea how it could compare with the Microwulf?
You can get single chips that outperform this. Specifically Intel's quad-core Xeon (Clovertown), which has a peak performance of 4 flops per cycle per core. Clocked at 3 GHz, that's 4 cycles times 4 cores times 3 GHz, or 48 gigaflops. This "supercomputer" is very unimpressive.
Did anyone else notice the poster of Monty Python and the Holy Grail behind him?
now what can we run on it?
Seriously, this looks within the capability of every slash-dotter to make. I know I could really use one of these to do CFD computations, rendering in CAD programs and home computer graphics. But that means running standard software, Linux or (pause to spit) M$. Can the Micorwulf do this?
I was going to post the same thing. It was the first thing that popped into my head after reading the headline.
n i-itx-cluster
Another group is producing much the same thing commercially, in a nice case and all. A 4 node Core 2 1.8Ghz with 1 gig ram per node and 2x 250Gb storage is about $7000 (USD)
(Wonder how that stacks up to what he built speed/cost wise, though I'd bet the Via cluster beats all in power use (140W max load))
See the link at Mini-ITX
http://www.mini-itx.com/2007/02/26/the-octimod-mi
Company site
http://ainkaboot.co.uk/octimod.php
"Cheeze it!" - Bender
Discovering that you can build an even more cost effective supercomputer than these guys: priceless
which is totally what she said
I thought one part of the definition of a supercomputer was that the cost exceeds a million (used to be that cost exceeds ten million). Dollars or Pounds, doesn't really matter even with the current exchange rate :-)
If you can't build a teraflop capable supercomputer for under $1000, you just aren't being creative enough.
This is so old-hat it has fungus on it!
m s.htm
Seti farms have been doing this since they started - here are a few pictures:
http://bhs.broo.k12.wv.us/homepage/staff/seti/far
Tell us something new!
Just wondering?
Just cat 5! (OK, I'll get my coat)
Wouldn't that be a Beowulf Cub?
We build this in the ages of Pentium I.e rcomputer/
It was really funny and interesting.
http://www.sorgonet.com/supercomputing/yourownsup
For a more hands-on approach, maybe these 200W+ DC-DC converters will do: http://www.mpegcar.com/acatalog/200w_and_above_PSU .html#aD220PSU. The 220W version is rated at 95% efficiency... can't go wrong with that!
bundaegi is good for you
Well, if you are looking for a conceptual testing ground couldn't this work be done with Vmware on a high end PC ?
In '97-98, I had a little 4-5 node Beowulf cluster on a cart with wheels. While it wasn't quite as cost-effective as this, that's the nature of pricing in the computing world.
:)
On that note... hard drives are good to have for all nodes, imo, since you may be doing things that make 'fetch/store data over the network' a bad strategy.
The entire point of the Beowulf model is it's cheap, easy, and fun. While it's great to see people building cute little clusters like this one, I wouldn't exactly call this a breakthrough moment in hobbyist cluster computing.
That having been said, I hope the machine gets a lot of use, and I like the appearance. I'm almost tempted to build one for myself, for old time's sake.
This has 4 dual core CPUs - 8 cores. That's the same as a MacPro or Dual Quad code Xeon PC who's cores are more powerful and which have much better communication between CPUs. And they have cases ;)
So a Dual Quad core Xeon a super computer too?
ccalam - acoustic versions of new songs.
Sony talk so much shit
So 1 Hz equals 1 FlOp? And a 3.2 GHz CPU can do 3.2 gigaflops, right? so how are they getting more then 3.2*4 = 12.8 gigaflops out of four CPUs? Can they execute multiple FlOps per tick then? And do we care that these will bottleneck at the rather limited bus (even forgetting about the switch).
If the bus speed is 1 Ghz x 32 bit, doesn't that mean that the whole computer is limited to 1.3 gigaflops at best (need to move at least 96 bits to perform a FlOp?), or even less if a lot of data has to travel over the 1GBit ethernet:
I know I am clueless, sorry, but that's how I learn. THanks for your help.
A striking resemblance for a box of bits. I wonder if it's got the same surly attitude.
http://upload.wikimedia.org/wikipedia/en/1/1d/B7_s eason4_blake.jpg/
A million in any currency, or just US dollars or UK pounds? how about Zimbabwean Dollars or Laos kips?
Hell, the IBM SP was a commodity pre-built supercomputer. This is much cheaper. but the concept isn't revolutionary No, the concept hasn't been revolutionary for decades, the effect might be though.
Deleted
It's still interesting to many of us, simply *because* we could probably build one ourselves. Not in spite of it.
A lot of what us humans do in life is "because we can". This doesn't appear to be any different.
(It slightly amused me that the captcha to log in to post this post was "differer".)
...Hrm, let's see. Foreign entity that we (the US) have done business with for years (let's call them "Son-eee") builds a sleek black box with multiple cores to appease the hardcore gaming m-asses in the never-ending search for the ultimate entertainment experience. Of course they equip such a device with Internet connectivity, and then secretly deploy code in the background and call it "gaming negotiation" when trying to mask the bits of encrypted traffic in the background noise. In the meantime, 80% of those processor cores are cranking out new ways to ensure Foreign entity can overtake target audience within their massive computing grid. Of course this all happens while you're sleeping at night. C'mon, did you actually think your new gaming console seriously NEEDED all those cores? Please. Think I'll go and dump some more money into Tynfoil-Hat Inc.
Am I missing something here? The Sisoft Sandra MFLOPS measurement for a top end Intel Core 2 is 47 GFlops http://www.tomshardware.co.uk/overclocking-intel,r eview-2395-28.html/. OK admitedly this is a sythetic measurement, but it's a ballpark figure right?
From Wikipedia: Supercomputer The speed of a supercomputer is generally measured in "FLOPS" (FLoating Point Operations Per Second), commonly used with an SI prefix such as tera-, combined into the shorthand "TFLOPS" (1012 FLOPS, pronounced teraflops), or peta-,combined into the shorthand "PFLOPS" (1015 FLOPS, pronounced petaflops.) It's not exactly a good quote, but looks to me like we're bumping the lower edge of the petaflop scale these days. Thats six decimal places people.
You are awash in a sea of fiercely stated opinions. Obvious exits are: 'File->Quit', 'Reply', and 'Page Down'.
The University of Kentucky (where he is coincidently going to grad school) beat his price point years ago on a "real" supercomputer. This super computer was built for about $84 per GFLOP in 2003 and it made the Top500 list when it was built. The Aggregate team at UK is one of the tops in the field when it comes to supercomputers on the cheap.
Seems like 2GB per (dual core) node is a little on the low side for practical usage. Not surprisingly though, RAM is the biggest cost of the system (992$ total) and switching to 2GB or 4GB modules will raise the system price considerably. Would still be cheap though.
Its a nice project, fun and cheap and good for the students to try out on But 8Gb is far to small memory to be of any use in the kind of simulations my customers been doing, and only 1Gb per core... but for the price its nice enough... for the tasks able to be run on it.
You could get rid of those gigabit cards and use a dual head MB @ $87.99ea. : http://www.newegg.com/Product/Product.aspx?Item=N8 2E16813138059
Every time I start to have faith in humanity, I ruin it by driving to work between 7 and 8 am.
I mean, it wouldn't even be enough to run Jita, let alone the whole EVE-Online cluster. :)
For current Intel core2, to get the Rpeak, take the number of cores total * clockspeed * 4. A single quad core 2.0 GHz gets you to 32 gflops already. You can readily build a rig with a single quad-core 2.0 ghz for less than 2,500. This is incredibly a non-event. It serves as a handy demonstration of how supercomputers are roughly architected today to people not in the industry, but the price/flop is noting special at all.
XML is like violence. If it doesn't solve the problem, use more.
They spend $180 on the Intel NIC's while the mobo had 1gbt ports. Other micro clusters pointed to here used the onboard nics so why not here. Given their credentials I am sure there was a good reason.
Wow!
/.
I repeat, wow!
How exactly does this qualify as newsworthy?
This is almost as bad as the time some goose bought a mini mac and before the sales launch was a week old he'd gone and ripped the guts out and stuck them in a frickin' PC minitower case so he could "run a cheap server". What a dingbat.
On second thoughts, the mini mac destroyer's effort was *much* worse than this, at least there is some merit to what these guys did and they didn't go and wreck a nice piece of kit in the process. It's just not exactly newsworthy stuff, just some geeks having fun with their toys.
On the other hand, mebbe CowboyNeal just feels the need to see a whole load of incredibly amusing "imagine a beowolf cluster" and "but does it run linux" jokes. After all, it must be several minutes since the last one appeared on
Yes, I think that might be it.
"You can't fight in here, this is the war room!"
Modern 'supercomputers' are not fundamentally so different from any other network of more than one server. Most run at least one independent kernel instance on each server, and that kernel is generally the same kernel as anyone else would use of that platform. The brains about sending messages efficiently between nodes is generally handled by userspace software (though you might make the claim that low-latency interconnects such as Infiniband and Myrinet are mostly cluster oriented, and have drivers that provide a low level medium for the userspace tools through their kernel drivers, but the story configuration involved none of that). To make something that no one could argue against having the same basic architecture as a 'real' cluster, you just need two-three computers with ethernet connectivity, and download something like OpenMPI and off you go.
Their number sounds about right for 4 2.0 ghz Opterons. If they used core2, they are doing pretty terrible at Rmax/Rpeak ratio. Maybe they have terribly low amounts of ram per core (which precludes tho Top500 benchmark from doing well).
XML is like violence. If it doesn't solve the problem, use more.
He once predicted that computers would be so large and so expensive that only the five richest kings of Europe would own them.
-Yim
Still can't run Vanguard SOH
The cluster depends on gigE for the interconnect, which means data transfers are going to be slow, and have a high latency. He'd be better off spending a little more and using Infiniband equipment.
Clear, Dark Skies
FWIW, I was able to put together another solution around $1,322.89 using 2 motherboards and 2 quad core Xeon 2.13Ghz procs. Same amount of memory, only two of those drives, two 350W PS, the same switch, etc.
Not exactly cheaper, but it's definitely smaller.
Every time I start to have faith in humanity, I ruin it by driving to work between 7 and 8 am.
somone has to. Imagine a Beowulf cluster of these! all jesting aside, i am impressed by this.
I've decided to Diversify my Holdings. I've divided my cash between my left and right pockets, instead of all in one.
let us see, hummm...
AMD Athlon @ 600 mhz - 2.4 gigaflops (single precision), 1 gigaflop (double precision)
Pentium 4 @ 2 ghz - 8 gigaflops (single precision)
Pentium 4 @ 3 ghz - 12 gigaflops
Athlon 64 X2 4600 - 14.7 gigaflops, 17400 MIPS
G5 Dual 2.3GHz - 30 gigaflops
XBox 360 Xenon chip -115 gigaflops
XBOX 360 Xenos graphics chip - 240 gigaflops
nVIDIA 7800 GTX 512 - 200 gigaflops
ATi X1900 - 553.8 gigaflops
Even an old fashion powermac G5 goes faster for half the price !
But, did you even know your X1900 is a super-Compter by itself ?
Kwisatz Haderach
Sell the spice to CHOAM
This Mahdi took Shaddam's Throne
Nice effort, but why not just use a couple blade servers in a chassis?
If anyone wants their uni project shot down, just stick it up on slashdot! :)
In http://clustercompute.com/ you could find better design in term of compactness. Another thing is that the cluster does not need KVM (in process only at test mode) and as noted in several research papers dual 100M can beat gigabit (source http://en.wikipedia.org/wiki/Kentucky_Linux_Athlon _Testbed )
Ok, when did they bought it, and how much did the prices go down since then?
Beowulf is a good idea for a very limited number of number crunching applications, or as a student learning tool for comp sci or related studies. Yeah, we built one of those a couple years ago, the professors ended picking up intel quad core machines that were faster (no effing network latency). Beowulf is gathering dust.
Oh, and try writing your own lam-mpi code sometime...
...and running Ubuntu linux... priceless...
Plus it would be on your doorstep by noon tomorrow.
It would be like having Sony sponsor your hardware ;)
The number of GFLOPS moves up every 18-24 months. I don't think specialization is that bad of a qualifier; after all, using something that'd be considered a supercomputer a decade (or more) ago just to watch YouTube videos can hardly be called a supercomputer.
e w
http://en.wikipedia.org/wiki/Supercomputer#Overvi
Get'cher mobo humming...on the superhighway...looking for some venture...and whatever comes my way...
Do not mock my vision of impractical footwear
As someone else working at a public university tech dept. I think this tiny "supercomputer" is the worst idea I've heard in at least a few weeks. Great, you can put a machine together that does this on $2,500. Who runs it? Who keeps it patched? Who fixes it when it breaks? Grad students and researchers are there to research. Paying whatever computer services organization on your campus to maintain it is going to cost more than the machine itself!
...
All the ~100 processor (that's ~12 quad-core dual-cpu nodes these days) clusters I see going in would be much better pooled into a central academic computing cluster. Adding 10 nodes to my cluster of 500+ really isn't going to have squat all impact on the administration time required for it. You'll get more than out of your standalone cluster: people building little clusters don't tend to keep them 100% busy. Large, multi-group collaboration clusters do tend to stay 100% busy due to staggered paper deadlines, etc. So when your nodes would otherwise be idle, someone else runs them and when you need compute power, you can overrun other's idle nodes. Plus you get to share services with everything lumped into a single large machine: DNS, DHCP, Network boot/install, file services, user database,
Slashdot Patriotism: We Support our Dupes!
The Mac Pro 8 core has been clocked in excess of 80 gigaflops. See below:o re_Mac_Pro_reports.html
http://xlr8yourmac.com/systems/8-core_mac_pro/8-c
Although not as cheap as the Microwulf, Nvidia has a desktop super-computer for sale http://www.nvidia.com/object/tesla_deskside.html at 500 GigaGlops, to start.
You could get rid of those gigabit cards and use a dual head MB
The network adapters that come build into consumer motherboards are pretty simple; at $165 the ones quoted by the gp almost certainly do a lot, if not all, of the network overhead processing onboard. This is much more important when trying to squeeze flops out of your cluster than saving a card slot or a few $s.
Only retards would attempt to build a super computer with the X2 3800 AM2 processors, most "sensible" computer users would use an Intel Core 2 Duo, as they are the leading processor at this current time and are by FAR cheaper performace:money wise. Not to mention they can overclock like mad!
This link would be more appropriate for slashdotters:
http://www.clustermonkey.net//content/view/211/1/
If you mod this up, your slashdot background will turn into a beautiful sunset!
The cool part of microwulf is that you could also scale this baby easily. All you'd need is a better gigabit switch.
... or, for $4.99 more per cpu, you could use the 3GHz Athlon 64 X2 6000+. That ought to have a positive effect on the performance numbers :-)
I bet this project started out as an attempt to build a machine that could run the latest version of a popular mainstream o/s, play media w/ DRM AND access the network at the same time.
Not that it costs $1000, but you forgot the components to build the cart.
The Spoon
Updated 6/28/2011
Cheap, commonly supported, so I guess good enough for a low budget attempt.... ....but seems that choice for connectivity hinders them right off the bat.
I am very small, utmostly microscopic.
Didn't Apple in fact market that silly little cube computer a few years back as A GENUINE SUPER COMPUTER?
you can save even more by cutting out the network cards and getting a nforce 570 or 590 board with dual gig-e and a low end video card. or cut out the pci one and use a pci-e and the on board gig-e.
And move to 2x512 or 2x1gb of ram. Why pay $48.49 for 1gb of ram when you can get 2x1gb for $20-$40 more.
The nforce 570 and 590 have tpc/ip off load and teaming.
why not just a 1 amd 4x4 system with ddr2 or dual opteron board with ddr2 ecc or a quad dual board with ddr2 ecc and HT bus is faster then a gig-e link.
Lower cost ram then dual xeon with FB-DIMMS.
Maybe it is buried below the level that I read, but I am surprised someone did not point out the available intel quad core chips for less than $300 as another basis for a microwulf. I'm glad these folks did this, but I have some pretty serious reservations about hauling around a no case Frankenstein like this to the schools. This is dangerous. MIGHT be ok at the College but again if someone sticks their pinkies in for a look-see... So I'd suggest thinking about suitable cases for dual quad core setup. My guess is that newegging could lead to a system that would be even cheaper and perhaps have better performance. Some information about what flavor of 'wulf was being used would also have been interesting. Of course very shortly Octa Core will be available and if the memory interconnect is done better than in the current intel quad core these things will then be the sweet spot. With respect to interconnect GigE is fine for a lot of important problems. If you want cheap then in most cases the highspeed interconnect system is going to cost more than the computer. Just some random thoughts. Nice article.
The summary doesn't state that this baby runs Ubuntu Linux.
Actually, I've done some simple benchmarking and been impressed that the Core2Duo chips are not all that much better at floating point calculations (structural biology programs) than the AMD chips. Now, the C2D cpus are much better at integer work, just not the double precision FP. So, I think the choice of the inexpensive AMD cpus is reasonable.
duke out
All I know is I want one! The price has been well proven LOW!
You have to admit. Its simplistic and good.
Kudo's to them
List of #500 on the TOP500 by year
Year .
06/1998 - 15.0 GPLOPS | Southwestern Bell, USA. . . | HPC 6000, Sun
11/1998 - 20.5 GFLOPS | Koeln Universitaet, Germany | HPC 10000 Sun
06/1999 - 34.2 GFLOPS | CIEMAT, Spain . . . . . . . | T3E900 Cray
11/1999 - 38.4 GFLOPS | Bank, United States . . . . | HPC 10000 400 MHz, Sun
06/2000 - 51.2 GFLOPS | EDS, United States. . . . . | HPC 10000 400 MHz, Sun
11/2000 - 78.0 GFLOPS | Zurich American, USA. . . . | SP Power3 375MHz, IBM
Really, calling this a supercomputer is lame. It has only one 250GB disk; it will have utter crap IO performance. Most compute heavy jobs are also disk heavy because you want to checkpoint your intermediate results in case of a crash. Since there is only one disk, one machine must be serving it up to the others (NFS, ISCSI, whatever). It is clustered through gigabit ethernet, which will act as a limit on performance. They even skimped on the connection to the outside world and got a 100MBit card. "Real" clusters use Infiniband or Myrinet, both of which are optimized for high throughput with low latency and low contention. Gigabit is not. Linpack is rather kind to clusters; more finely grained parallel tasks will pay more for the poor linkup.
Also, with only 4 processors one could also build a 4-way SMP machine which would then not have to deal with any sort of message passing at all. You instead get one shared memory interface. It may be slightly NUMA, but the extra latency cost of hypertransport is amazingly low. Instead, by putting only one dual core die per motherboard, you have to jump through hoops to move work from one die to another, and pay really bad latency costs. You could also do better with a 2 quad-core processors on the same mobo (although you'd have to go Intel for now...). It's easier to program, supports finer grained parallelism, and allows potential savings on other parts.
I can get 2 quad-core Xeons at 2.4 GHz each and a 2 socket motherboard for $820 at newegg. They spent $980 on 4 dual-core 2GHz processors and 4 single socket motherboards. They also spent $240 on gigabit cards and the switch. So, for $400 less, I can have an SMP machine; one which probably has higher floating point performance as well. Rather than 4 cheap power supplies I can get one nice one (which is probably more efficient too). Further, I don't have to run 3 of my nodes diskless. Really, at this small scale a cluster is not the way to go.
The Cell in a PS3 can do north of 150 GFlops, which at a price of $600, gives it about $4/GFlop. And that's not even taking advantage of the GPU on it. That is definitely the cheapest computing available right now.
My email address contains no numerals.
I just thought of something.. is it more powerful than a PS3? If not, then that would probably be the way to go for cheap supercomputing needs..
which is totally what she said
People now place boundary of a super mid-teraflops. Your game machine or desktop now reach tens of gigaflops.
Umm no i don't think so.
A true super computer has an architecture behind it. Just throwing some ethernet connections between some cheap PC's isn't nearly the same.
Sure, might be cool to show your friends, and isn't useless, but don't call it a 'super computer'.
---- Booth was a patriot ----
Then I wonder how it would compare to a single 8-core Mac Pro running the same software. The Mac Pro might be limited in memory bandwidth, but if you're not going out over Ethernet and handling the overhead of Beowulf, it should be faster, no? If the Mac Pro can double the performance of this cluster, then it's PPR is even lower.
That's that sticking out of the 19" monitor? It looks like a little black hand.
$2,500??? That's what I paid for a Dell P-II-233Mhz w/128 MB of RAM and 6.4GB of harddrive space just 10 years ago! Of course, I got a 17" crt monitor and 56KB modem at the time.
"It's the height of ridiculousness to say for those 9 lines you get hundreds of millions."
Depends on what you need the box for.
:-/
I've used databases with 800 small powerpc computers (each with 1gig ram and 500gig HD). It's likely 800 times faster than any single DB-box out there. I'm also currently using 16 dual-core cpu box (32 cores!)---that's for heavy computational (kind where threads need to talk to each other often) lifting. In a few weeks, I'll make a cluster of those boxes (got 3 more shipped to us). We also got a grid environment (40 dual core blades)---for heavy computational (kind where threads don't talk to each other) lifting.
It's unfortunate that financial institutions are much better funded than universities
Steve Jobs introduces the iRack: http://youtube.com/watch?v=rTOoXpeNwMI
Either way... I did end up pricing our a 2x Quad-AMD system right after that and it was a around the same price. This whole project just looks like a way to build a cheap cluster and not really a small high performance machine for a small amount of cash.
Every time I start to have faith in humanity, I ruin it by driving to work between 7 and 8 am.
O.o Why didn't they choose the new NanoITX or PicoITX? You could double or triple the CPU's per box.
So what should I call my "learning cluster" built from 1+n vm images? SoftWulf is more descriptive, but I like SquishyWulf...
It's way cheap, slow, and certainly capable of providing a learning platform. And unlike MicroWulf, I can distribute it, as it's all free software.
Oh wait, VirtuWulf! I like that one!
I looked at the photos and, not to burst their bubble, anyone trying to check something that looks like that onto an airplane is going to get a cavity search...
It must have been something you assimilated. . . .
I thought the hip thing was GPU based supercomputing. NVidia even has a dedicated GPU based, desktop sized, scalable supercomputer line called Tesla.
The basic Tesla unit c870 = 518 Giga flops for ~$1300.
Tesla s870 = 2 Terra flop for ~$12000 (still desktop size)
NVidia Tesla
I'm in ur computer
Connectin ur motherboards
Apology to Ubuntu forum.
I code for systems with 800 4 Optron nodes, with 10GB/s interconnect, and a couple hundred terabytes of SAN attached storage. That is a supercomputer :) Well sort of, a lot of people in the HPC community consider it just a cluster, as some programs need 64+ CPU's in SMP mode, so any loosely coupled memory model would be considered a serial farm :)
Also, note that high end platforms, would have redundant power, redundant high end interconnects, redundant hot swap drives etc. There also would be enough of them to need, high end switches, blowers, power conditioners, air circularators, and various other room coolers. Of course a custom built workstation without a graphics card, monitor, or even case is going to beat the pants off of HPC architecture price per flop, good work to the group, but hardly newsworthy in the HPC community.
But you aren't really a supercomputer at that point, you're a cluster. These days the line is more blurred than in the past but more or less the difference is interconnect speed. In a real supercomputer, there are very high speed interconnects, so you can run things that heavily rely on one part communicating with another, like particle simulations. That's why the US Department of Energy buys so many, rather than clusters. They do things like weather simulation and simulation of nuclear weapons, where every node as to be able to talk to every other node with essentially no penalty.
Now if you have a job that doesn't use a lot of inter-node communication, like say 3D rendering, then a cluster is a better answer. Normal hardware with Ethernet interconnects. Works great and is cheap since you can use commodity parts. But don't confuse that cluster with a real super computer, you throw one of those intense inter node problem at it, it'll fall over because the interconnects are too slow.
Unfortunately these days people really blur the distinction. You'll see systems on the top 500 list that are really questionable. It'll be commodity hardware connected with something like infiniband. Ok, great, that is faster (both more bandwidth and less latency) than Ethernet, but it still isn't necessairily up to what you'd get from a real supercomputer.
However in the case of this deal, no, not a super computer. It's a small cluster and they are just calling it a super computer as marketing, effectively.
OK, I give them credit for the following:
They needed a fast computer and they built one.
They had a limited budget and worked within the limits it imposed.
They intend to use it, for illustration purposes, to get young kids interested in computer science.
That's commendable. Great.
However, as others have said, this is nothing more than simple beowulf cluster with 8 cores on 4 commodity boards using gigabit ethernet in a home-made frame. It's been done. And with the rate of growth in processor performance being what it is; next month I'll be able to go out and build a more powerful machine, for less money.
RTFA. The $2500 figure is based on January 2007 prices. Behind one of the links in TFA is a better article, which states that with August 2007 pricing, it would cost less than $1300.
Gigaflops are a very bad way of defining performance. All that means flops is "floating point operation per second". Ok but the problem is, what kind? Well that's not specified. So depending on how well your benchmark conforms to what your hardware is good at, depends on what you'll get. For example high end graphics hardware these days really is capable of approaching around 500Gflops in ideal conditions. If you give it the right kind of problem to solve, it can deliver that kind of power. However there's only some things it is good at.
In terms of a more general test for CPUs though, no 26Gflops really isn't anything impressive. My desktop benches at 16.3Gflops in SiSoft Sandra and it's theoretical max is about 21Gflops. That's just a Core 2 Duo, and not an especially expensive one (about $200).
This is nothing more than a professor being self amazed with their new toy, and trying to put some marketing spin on it. It's not a super computer, and not even a particularly impressive cluster, other than the size.
As for Gflops benchmarking, you can more or less treat it is marketing. It is mildly useful, but mostly just BS. Because there's no defined way to do it, there's no basis for common comparison. Again you consider my processor, with a max theoretical Gflops speed of 21 and my graphics card with 500Gflops theoretical performance. Why don't systems just axe the CPU, if the GPU is that powerful? Well because my CPU can do that on pretty much any floating point task. The GPU can't. It would be easy to design a problem that the GPU would be lucky to get 1 or 2 Gflops on, because it was the kind of problem it wasn't designed to deal with (one with a lot of branching for example). As such we have a CPU for general purpose calculations, and a GPU for special purpose calculations.
Comparing the Gflops numbers of the two gets you nothing useful.
The PS3 comes out of the box with a Cell uP that gets something like 20 GFLOPS on each $500 PS3. It's already networked into clustered supercomputing like this MicroWulf.
A $500 PS3 has 20 of the 26.5 GFLOPS the $2800 MicroWolf has. MicroWulf runs Ubuntu, which can also run on PS3. If people can port Linux libraries like Mesa/OpenGL/X to the PS3 SPEs, where most of the power lies, then we'd be looking at $25:GFLOPS, not the $94:GFLOPS on the MicroWulf.
And while taking a break, you can play Gran Turismo 5, and 40 more games you can afford with the money you save on HW.
--
make install -not war
Apart from the fact that it's far from a supercomputer. It's just 4 PCs stuck together.
That's what I was thinking of building. Just the idea of having all those Gumstix sticking out of a mounting board on my wall provides days of laughter, but it could be fun, educational, and even do some of the computing I need for a couple of projects.
Or, even easier, buy a Playstation 3. They can peak at between ~25 and ~256 Gigaflops depending on precision, and run far less than the $2500 price tag in the article. Additionally, with the number of Linux distros and toolkits supporting them increasing, they are a fairly cost efficinet "supercomputer" if you can even claim that ~25 Gigaflops is a supercomputer anymore.
Realistically, the one in the article is nothing more than a powerful workstation by modern standards.
Microsoft Sucks, F/OSS Rocks. I get mod points now right?
Yes...Unfortunately, they are oriented with their "poles" pointed towards the chicks effectively repelling them.
Check on an airplane?!?!?! Did you see the picture of that thing?! I have trouble getting my Treo through security!
To boldly use to and too two times and get it right too! They're not gonna believe their eyes when they see it there!
But I would NOT try to check that shit onto an airplane if I was you, not unless you want to get pulled aside, questioned, strip-searched and slapped around security for a couple of hours.
The reality is that we have this same 'novel' kind of server configuration commercially available today and it's called a fully-stocked blade enclosure.
I'd rather see a successful use of $2M in tax dollars than 1000 instances of a waste of $2000s in tax dollars. A waste of tax dollars is a waste of tax dollars. This will inspire every university in the country to go repeat the same waste.
Or, for $3500, you can get a dual quad-core Xeon machine (8 cores total) with 8GB of ECC memory in a rack enclosure from Dell (PE2950) for... let's see... $3,513.20. I'm sure that the Dell machine will significantly out-perform their stated configuration. It won't even be close. Oh, and you don't have to build it yourself. And it even has support! As a system administrator in an academic environment, just the thought of having a bunch of researchers with this frankenstein creation in their offices makes me physically ill.
...is also one of several profs from the same college that put together one of the most widely used C++ programming books, which was also followed up by him and a couple others with a Java version. (Don't know how popular the Java version is though.)
For more check out his publications page.
He is also behind his colleges primary super computer, Ohm, which was put together back in 2003/2004 (1 Ghz/1GB RAM by 16 systems); and apparently he might have a new grant to upgrade it.
Truth is like the sun. You can shut it out for a time, but it ain't goin' away. - Elvis Presley (source: imdb.com)
I for now welcome your Microwulf overlords...
(duck, removing the clown suit)
I ate dinner last night with a friend, just starting as a professor at Penn; for $5,400 (or was it $5,600?) he just ordered an 8-core, 16GB MacPro, including a 30" monitor. He considered dual 30" monitors, but then thought, "Eh, why go crazy?" Yes, that's twice the price of the system under discussion here, but it's a pretty powerful machine for the price. Can anyone comment on what the machine here could do better than a beefy MacPro?
:) Obviously, a few thousand here, a few thousand there, and pretty soon you're talking real money ...)
(I am not discounting the price difference!
timothy
jrnl: http://tinyurl.com/c2l8yr / foes: http://tinyurl.com/ckjno5
For 32 and 64 bit operations, it is probably much faster than the PS3. Last I read, the PS3 really slowed down on large floating point intensive operarions, which is where AMD has traditionally excelled.
Analyzing the stock market and the Torah to find the one number that explains both.
With the waste heat I'm baking some Pi.
Build a man a fire, he's warm for one night. Set him on fire, and he's warm for the rest of his life.
Many great things come from tinkerers who go into the garage or basement and just try out an idea they have. Most of the things that we have today are somewhat illogical extenstions that were derived from something else. The inventor(s) took off-the-shelf items and assembled them in a unique way. Examples abound in the computer-world: HP, Apple, Microsoft and many other companies got their start in exactly this way.
This computer is nothing extrodinary, it is built from off-the-shelf parts and constructed in a somewhat standard manner loaded with off-the-shelf software configured in a understood manner. I'm not cutting on him; I'm just stating something that I think is fairly obvious.
What is extrodinary is that in this high-tech age, we are still showing the desire to create. Inventions don't belong to big companies with big R&D budgets. Anyone can try to build something a bit different. Most of these inventions aren't earth shattering and most of them are not commercially viable but they are born of necessity or of a desire to make somethig better/faster/cheaper. It is part of the human spirit I think.
Good, they can use the money they saved to buy a real switch. That 8 port GigE switch has a 3.8Gbps backplane.
I have been following photonics for the last 20 years, and today I came across something that seemed it needed to be slashdotted. "This development makes evident that Intel's anticipated Laser Processor is more than just a rumor."
http://ightfeet.com So are electrons obsolete yet? Guess we will have to see if this new cool optical server ships on time. My guess is that we will soon see the light;) -- R
Ross Youngblood
All of the boards also have an additional 1g NIC installed, for additional throughput.
Just disrupt the deflector shield with a tachyon burst.
What is a supercomputer?
The fastest, most powerful machine to solve a problem today. This is generally credited to Dr. Sid Fernbach, George Michael and Jack Worlton, and others.
What if I qualify that with "cost?" ["for the cheapest"]
Then, it's not a supercomputer. Period. It might be a minisupercomputer, though. Don't let George know that I said that (he's much more hard-line).
Eugene N. Miya, Comp.Sys.Arch.Super FAQ
I like that idea... You should package it, make it easy to use and see if any Universities, banks are interested. If they are, I want first dibs on UK distribution. :)
Deleted
No that's not true at all. There are thousands of high-performance computing clusters out there. In my current job, I work on the top 100 sort, but in the past, I spent a lot of time working on the sort that don't make the top 500 list. They're generally 1-2 racks of rackmount sleds running some sort of MPI application, or bulk processing of video feeds, render frames, whatever. I remember one oil industry machine room that had over a hundred clusters, each with 40-80 processors. I suppose they could have connected them all together and called it a single, big cluster, but they didn't run any large jobs, so why kill your reliability by making the cluster bigger than it needs to be?
This microwolf thing is neat for a university project. However, clusters are a lot of work to administer, even if you come up with some really good tools. I don't know why everyone would want one next to their desks. That's a lot of work. They should build a smaller number of really good machines, and pay fools like me to administer them. Don't ask scientists to be cluster admins.
In a similar fashion to the cell processor there is the Arrix (http://www.mathstar.com/) which claims a peak of 500Gflops/s. I have heard that the price in 1K lots is 300 (somewhere around 250 if I remember right. Just put one of these on a board and put it in a laptop. That would be a portable supercomputer. Granted there is the programming thing - but the cell would need that too.
Thank you! I knew I'd seen a beowulf in a lunchbox in Linux Journal, and I haven't even been reading it for a couple of years or so.
Some pure speculation now : I also wonder whether this "Microwulf" would even outperform a dual dual-core PC. One can buy a motherboard supporting 2 dual-core Opterons for $500, dual-core opterons are about $200 each, and 8GB fast RAM should be about $500 or so (8x1GB). If we allow the same ($2500) budget, I bet you can make a 4-core PC that outperforms this cluster on EVERYTHING. Each CPU-CPU communication is either on chip or over hypertransport - each massively faster than gigabit ethernet. If you really want a cluster programming model use a virtualisation solution to get 4 copies of the OS running on the box and lock each one to its own core.
and who's paying the electricity bill if we have a farm of these. And does it do anything more than run those SETI junk ?
All the high speed hardware and broadband/no warez - make the thing "junk"
#50 and #71 on the TOP500 list are composed of Apple Xserve G5s.
Now, Apple claims that the new quad-core Xserve with Xeon processors is five times faster than the Xserve G5.
Seems like building your supercomputer from Xeon Xserves would be a viable strategy for achieving a high rank on the TOP500 list!
That that is is that that that that is not is not.
You could do encoding with it. Either encode a bunch of movies at once, or encode them with h.264's multi-encoding feature (with a slight drop in quality).
You could also do rendering on it. It would basically be your own personal renderfarm.
You could compile code on it, too. Or use some of them as test rigs -- run different OSes on them to test your code on -- instead of virtualizing it all under one Linux.
But I don't think either of these is really valid, even if these are both things you do a lot, because one dual-core machine is usually fast enough to do all of this as fast as you realistically need it done.
Don't thank God, thank a doctor!
Don't worry, they're Supermicro, they wouldn't be stable if you cooled them in a swimming pool of liquid nitrogen.
/.er has in their closet - just imagine a Beowulf cluster of those!
Still, meh indeed, scrape together the piles of computers your average
Give a man a fish and you have fed him for today. Teach a man to fish, and he'll say "WHERE'S MY FISH, YOU IDIOT?"
Dude, just buy a Quad or Octo core Mac already ! If you can't afford that on your college fund mega money build your own machine. The Quad core xeons just halved in price. A single machine bought from New Egg with dual processor, quad core xeons would wipe it's ass with this beopuppy piece of crap. Amateurs. Hope they don't kill themselves spilling tea on some exposed motherboard.
Read the article, those were January 2007 prices, they specifically mention it'd already cost about half what they spent due to plummeting AMD cpu prices.
Main Memory: Kingston DDR2-667 1GByte RAM $48.49 * 8 + $4.99sh = $392.91
...
Hah. I bought an 8 Mb stick for $500 once. That was a real bargain at the time. At that time, a real fast computer of the same dimensions would have been equivalent to a 300 MHz P3 at best
What shocks me is there is only one hard disk. I would imagine 1 drive per motherboard.
Know your pads. One time pad: good for cryptography. Two timing pad: where to take your mistress.
That actually fits the venerable /. business plan format:
;-)
1. Get'cher mobo humming -- Buy commodity hardware
2. On the superhighway -- Connect with fast network
3. Looking for some venture -- ???
4. And whatever comes my way -- Profit!
Scary what thye knew back then
Let's see - each core on a Core 2 can issue an ADDPD and MULPD every clock. Those are SIMD dual precision FP instructions, so that's 4 FLOP per clock.
Get a quad core processor - thats 16FLOP per clock. At 3Ghz that'll be 48GFLOP peak throughput - maybe around 36GFLOP for a typical heavy math workload.
Processor and system board will run you around $1200, complete system with lots of RAM and big drives and a monitor, maybe $2200.
So we're impressed with 25.6Gflop for $2500...why?
Maybe it's a time warp - did this story originate in 2001?
Perhaps he couldn't find one...
Crap. What did the new CSS do with the "Post anonymously" option??
it looks more like orac!
What about KASY0, which had $84 per GFLOP in 2003?
The question of whether a computer can think is no more interesting than the question of whether a submarine can swim.
this quote blows my mind: it provides more than twice the general-computation performance of Deep Blue
P R/
http://www.calvin.edu/~adams/research/microwulf/P
talk about a great example of how fast computers are changing the world! 10 years later, and 2 academics can hobble together a $2500 that has twice the performance of the IBM flagship, technology showcasing supercomputer.
Finally, something that will run Everquest 2 in extreme quality (with shadows turned on).
But can it play Doom?