Big Mac Officially Ranks 3rd

An anonymous reader noted that according to Wired, it will be announced officially on Monday the Big Mac supercomputer is the third-fastest super-computer. The article also talks about some of the amazing supercomputers in the planning stages. The sort of stuff that will make Big Mac look like that old TI-85 collecting dust in your drawer.

Comments

I'm still using my TI85....

[Anonymous Coward]

...you insensitive clod!

Re:I'm still using my TI85....

[feed_me_cereal]

no joke! I still use mine all the time! It still does exactly what I need it to do, and it's one of the few (relatively) small items I've never lost.

Re:I'm still using my TI85....

[pvt_medic]

well you can always make fun of the apollo space mission. What were there computers like?

Re:I'm still using my TI85....

[archen]

I thought the TI85 was like a badge of going to an actual college. You always know who really went to a college when their home calculator is this ungodly huge black THING which they couldn't even tell you what 90% of the functions do - I only dust mine off once a year when I have to do taxes. I remember thinking when I bought it, "wow I'll never have to buy another calculator ever again". Wish I could say that about my degree.

Re:I'm still using my TI85....

[DoctorPepper]

I bought mine back in 1993, when I started college. I've kept it the whole time. My wife recently decided to go back to college (after many, many years off), and I just stuck a new set of batteries in the old TI-85, dusted off the book, and that's what she's been using. Amazing what you can still do with an old Z80 processor!

Re:I'm still using my TI85....

[uncleFester]

Actually, I'm a little more concerned Senor Taco knows my Ti85 is in my drawer. WTF are you doing breaking into my house and inventorying my old dead hardware?!? Is this the geek form of TIPS?

Egad,

-'fester

Re:I'm still using my TI85....

[Spyky]

Ahh, but he got it wrong, mine is a TI-86!

-Spyky

Re:I'm still using my TI85....

[Anonymous Coward]

I have a TI-89, so *phhhpt* to you.

Seriously, though, a TI-89 (and the TI-92) has a 10 Mhz 68000 CPU with 144 KB of memory. The original Macs had 8 Mhz 68000 CPUs and 128 KB of memory. I won't be the least bit surprised if calculators twenty years from now have the equivalent of 2 Ghz PowerPC 970s and 1 GB of RAM.

Re:I'm still using my TI85....

[Alan Partridge]

their

Re:I'm still using my TI85....

[dnahelix]

Anyone still using their TI-99/4A?

Re:I'm still using my TI85....

[Ohreally_factor]

If you call your penis your "old dead hardware", you've got worse problems than Taco pawing through your drawers.

Re:I'm still using my TI85....

[Gabrill]

My poor TI-81 . . . I can hardly see through the faded screen. Anyone ever get Linux to run natively? :))

Re:I'm still using my TI85....

[Gerzel]

Here here! TI 85 4 3v3r!

See my evil 1337!

Re:I'm still using my TI85....

[ObiWanKenblowme]

I believe it's "What were them computers like?"

Kudos to the Mac

[Space cowboy]

It has to be said that Mac's haven't been famous for their speed, always pushing the "it does more", or "there are 2 procs" arguments, but this gives them some serious ammunition. Perhaps they'll even get their advert on the air in the UK now :-)

Simon

Re:Kudos to the Mac

[black mariah]

I'd give some minor internal organs for a dual 2Ghz G5. I wouldn't give the sticker price for a dual 3Ghz Xeon system. I'm not a hacker or a techie. The only numbers that impress me are the smaller ones produced by REAL-WORLD application tests on Macs (Photoshop, etc.).

Re:Kudos to the Mac

[You're All Wrong]

I'd dearly love to see G5s as commonplace as x86s (tranlation - I want one, and I want it cheap), but being able to plonk a machine in the top-10 list really doesn't lead to that at all. At various points in time DEC (Compaq/HPaq) had more than _half_ of the top-10, and look how popular the Alpha became.

YAW.

Re:Kudos to the Mac

[You're All Wrong]

If your life consists of running Gaussian blurs the whole time, then you've got a point.

However for those of us who are software engineers, or website admins, or electronic engineers, then the SPEC marks corresponding to gcc, perl or twolf are REAL-WORLD tests.

Would you like to show me the SPEC results that put an Apple at the top of the heap, as I sure can't find them?

YAW.

Re:Kudos to the Mac

[Daniel Dvorkin]

The difference is that Compaq's marketing of the Alpha was nonexistent. (And even DEC didn't market it nearly as much, or as well, as they should have.) Apple, OTOH, will market the hell out of this achievement.

Re:Kudos to the Mac

[black mariah]

Hey, that's nice. As I already made clear, I am not a hacker and I don't care about number crunching benchmarks. LINPACK tests mean dick to me. I am a USER, not ADMINISTRATOR. If my pants went crazy for the latest release of Apache and gcc, I might want the biggest number cruncher out there. But my pants stay firmly in place unless there are some shiny numbers for tests based on software I use, such as Photoshop and After Effects.

Re:Kudos to the Mac

[Anonymous Coward]

If my pants went crazy for the latest release of Apache and gcc, I might want the biggest number cruncher out there.

I'm no graphics stud, but doesn't a Gaussian blur consist of convolving a source image with a Gaussian matrix? That's pretty much the definition of number crunching.

Re:Kudos to the Mac

[be-fan]

Damn this is annoying. The price of this machine doesn't bear any relation to the real world.

First, yes, its cheaper than the competition. It also doesn't do as much. It doesn't use ECC memory, for example, which saves nearly $1M right off the bat. They make up for this by using clever software that checks for data integrity.

Also, this machine doesn't come supported by a company (like the LinuxWorx cluster was) and uses new interconnect technology from Mellanox that the other machines did not havae access to when they were built.

Yes, the G5 is a good deal, especially if you do like they did and get an educational discount ($300 per machine) and don't buy your RAM from Apple. However, that's only because the Opteron is still very overpriced. Once those prices come down, Apple will either have to keep pace, or be out-priced again.

Re:Kudos to the Mac

[More Karma Than God]

Do you have a deep unwavering belief in your signature?

Re:Kudos to the Mac

[be-fan]

Oh, how often have I heard that one.

But, no, no I do not. I accept that there is a possibility, however slim, that I could be very wrong.

Re:Kudos to the Mac

[black mariah]

There's more to graphic design than gaussian blurs.

Re:Kudos to the Mac

[You're All Wrong]

You seem confused.
USER/ADMINISTRATOR have nothing to do with anything. Did you not know that users can compile things too?
And in what way are graphical effects not number-crunching? Most of them involve convolutions, after all.
And apache's not Perl.
And gcc's not number-crunching.
Very very confused.

YAW.

already official

[Anonymous Coward]

Clicky for the official November list

Trademark infringement

[Anonymous Coward]

Excuse me, but Big Mac is a registered trademark of McDonald's according to http://mcdonalds.com/legal/index.html

So what's going on here? Can they actually do that?

Re:Trademark infringement

[Anonymous Coward]

#1 - Apple isn't doing anything here except providing hardware
#2 - Virginia Tech, the school who put it all together, has a very un-catchy name for it
#3 - People, notably tech journalists, have assigned it the "Big Mac" nomeclature

If McDonalds' wants to go after the journalists, they'll run afoul of the U.S. constitution, but not as though that's ever stopped Big Business in the past.

Re:Trademark infringement

[lost_it]

Virginia Tech has not (and will not) call the computer "Big Mac". BBC used the name when it first started appearing in the news, and everyone else picked it up, IIRC.

The people in charge of the cluster don't want to call it "Big Mac" because (1) they don't want a lawsuit from McDonalds, and (2) who wants to be associated with nasty, greasy fast food?

They've worked out a solid candidate for a name (it's not official yet) that isn't quite as catchy as "Big Mac", but it also doesn't have any of the downsides.

Re:Trademark infringement

[Artifex]

They've worked out a solid candidate for a name (it's not official yet) that isn't quite as catchy as "Big Mac", but it also doesn't have any of the downsides.

They should call the building where it resides the Cider House, then they can talk about how it rules, and get Michael Caine to do the v/o for the documentary...

Re:Trademark infringement

[NeoSkandranon]

Since the super computer obviously isn't a food item how would McDonalds be able to sue?

Re:Trademark infringement

[Unregistered]

Since the super computer obviously isn't a food

What do you mean it's a big mac and it even comes with G5 chips. Can i get those G5 in Salt and Vinegar?

Re:Trademark infringement

[roskakori]

they should call it "freedom fries" to indicate that the cluster can compute stuff that will prevent terrorist attacks. that will give them a lot of funding.

Re:Trademark infringement

[Anonymous Coward]

" Excuse me, but Big Mac is a registered trademark of McDonald's" ...lest consumers get confused between a burger and a supercomputer?

Trademarks are not unique: The patents and trademarks office returns 13 results for big mac, most of which are owned by McDonalds. As usual, you get a more accurate picture by looking at the official register, rather than running to the mcdonalds.com website for legal advise, but you can have many trademarks all using the same wording for different products.

On a related note, anything with a TM after it means that the company is trying to claim something which is not a trademark, as opposed to the R which indicates that they do indeed 'own' the trademark.

Re:Trademark infringement

[JeffTL]

There is no "civil rights amendment" -- perhaps you are referring to the "equal rights amendment" which failed?

Re:Trademark infringement

[codifus]

Can I get fries with that G5 cluster? I guess you can supersize the cluster, but it's not like McDonald's and Apple are even remotely competitive with each other. Mcd's can sue, but probably won't.

Re:Trademark infringement

[afantee]

>> They've worked out a solid candidate for a name (it's not official yet) that isn't quite as catchy as "Big Mac", but it also doesn't have any of the downsides.

The official name is TCF - Terascale Computing Facility.

Re:Trademark infringement

[AaronD12]

What might be fun is if Apple and McDonald's have the rumored (and since pretty well disproven) iTunes music give away promotion. Then you can get a Big Mac (sandwich), win some music, and play that music on the Big Mac (supercomputer).

Re:Trademark infringement

[Tuross]

Since the super computer obviously isn't a food item how would McDonalds be able to sue?

Since the justice system was replaced by the "legal" or "court" system.
Like, last millennium.

To paraphrase a common saying, "don't let the facts spoil a good litigation".

Re:Trademark infringement

[hawkfish]

The people in charge of the cluster don't want to call it "Big Mac" because (1) they don't want a lawsuit from McDonalds

Back in the day (1986 IIRC) when the first 1Mb Mac was announced, Apple went to some trouble to stop people calling it the "Big Mac" (The preferred name was "Fat Mac".) Eventually it was released as the "Mac Plus".

Internet distributed computing

[ergo98]

Given the basic benchmarks used to rank supercomputers, could a cluster of loosely coupled machines compete, or is the bandwidth demands for the benchmark set too demanding? I'm just curious how projects like what is detailed at distributed.net compare: 1100 dual-processor macs would be vastly outranked by the hundreds of thousands (or millions) of PCs taking part in distributed processing for various code cracking or cancer curing purposes.

Re:Internet distributed computing

[CAlworth1]

Be that as it may, the point of the article, I think, is to show how a very powerful, yet compact supercomputer can be built cheaply - The examples from distributed.net can process huge amounts of data, true, but with only 1,100 computers, all in a single building? I am certainly one for open source alternatives, but there are other applications where having your own computing beast is better.

But besides all of that, I think the did it more because they could than anything else.

Re:Internet distributed computing

[Anonymous Coward]

Hmmmm:

PowerPC 970 G5 1800 rc572: 13,400,000.00

AMD Athlon64 1600 rc572: 5,771,251.00

Intel Pentium 4 3500 rc572 4,960,583.00

Re:Internet distributed computing

[gcondon]

Okay, I took your advice and looked at the distributed.net speed statistics. I looked for the fastest PowerPC & Intel scores in each project. Here's what I found ...

PROJECT OGR:

CPU @ MHz = Speed
G5 @ 2000 = 19,180,166.00
G5 @ 1800 = 17,100,000.00
G4 @ 1250 = 13,946,216.25
P4 @ 3200 = 12,155,245.00
Xeon @ 2800 = 10,251,811.00
PIII @ 1440 = 9,570,000.00

PROJECT RC572:

G5 @ 2000 = 15,058,974.67
G5 @ 1800 = 13,400,000.00
G4 @ 1250 = 13,084,678.25
P4 @ 3200 = 4,502,730.00
Xeon @ 2800 = 3,935,299.00
PIII @ 1440 = 2,927,187.00

Of course, these numbers need to be taken with a grain of salt since there is only a few (or even one) top-end machine of each class in the statistics. However, contrary to your assertion, it appears that the PowerPCs kick ass compared to the x86s.

My understanding was that, if anything, the distributed.net algorithms unfairly favor the PowerPCs - esp. those with Altivec. I believe the Apple has used that fact in their advertising much to the consternation of many Slashdotters.

Re:Internet distributed computing

[JohnsonWax]

My understanding was that, if anything, the distributed.net algorithms unfairly favor the PowerPCs - esp. those with Altivec. I believe the Apple has used that fact in their advertising much to the consternation of many Slashdotters.

And, of course, problems in computational fluid dynamics and the like tend to also favor Altivec. Not always, but it's pretty common. As such, the value of Altivec shouldn't be tossed off as some kind of parlor trick. For these applications, it's quite relevant.

Re:Internet distributed computing

[mlyle]

A lot of this worship of altivec is just plain wrong.

Sure, it's nice and fast.. but it's also single precision-- woefully inadequete for accurate CFD work except in the simplest of situations. Dilution of precision is bad.

Re:Internet distributed computing

[netwiz]

yah, and there's a reason. The Altivec vector instruction set specifically provides an instruction in hardware that the RC5 routines (and most encryption in general) make heavy use of. 1GHz G4 procs kill the Intel hardware by about 3:1 (approx 9mil on the above #s)

That said, the intel hardware should rip thru properly vectorized FP code about as well (for functions it directly supports in hardware), and should kill the G5 on integer performance. Mostly has to do with the G5 having fewer integer units than the P4/Xeon. This has been pretty much proven with the SPEC scores we've seen thus far, with both sides using top-of-class compilers (none of this GCC crap).

And now, back on topic, I think that this says alot about the IBM FP hardware with regards to Intel. Intel bent over backwards for an architecture that's not much faster per clock, and given the apparent lack of clock scalability in the Itanium, they're going to be hard-pressed to keep up. Their lack of FSB bandwidth is going to hurt them as well, esp. in the server market. I know of at least two scientific apps (nothing big, just school stuff) that the Itanium chews thru faster than the G5, but only for small datasets. The guy showing me his results mentioned that when the dataset goes from 10MB to 500MB, the G5 ends up around 2-3 times faster, as the memory interface becomes a bottleneck on Intel.

This may be one of the core reasons LINPACK on hugely parallel systems brings the Itanium and G5 so close to each other.

It doesn't help that the PPC970 looks to cost about a third what the Itanium costs.

Re:Internet distributed computing

[martinX]

Don't know about that:

Apple and IBM jointly developed the PowerPC G5, world's first 64-bit desktop processor. With two double-precision floating-point units , advanced branch prediction logic and support for symmetric multiprocessing, the PowerPC G5 processor builds on previous PowerPC designs, combining an optimized Velocity Engine with a superscalar, super-pipelined execution core that can execute more than 200 simultaneous in-flight instructions. This high-bandwidth core has over 12 discrete functional units that process massive amounts of instructions in parallel.

And it has got to be the most totally buzzword compliant CPU ever built.

All from here and here.

Re:Internet distributed computing

[waaka!]

Yes, the G5 does have two DP floating-point units, but the Altivec instructions are still single-precision. The Velocity Engine and the FP units are numbers 6 and 7 on the execution core section of Apple's G5 pages, so this is actually something totally different.

Yes but...

[Nefrayu]

Can the Big Mac play games like the WOPR?
How about a nice game of chess?

Re:Yes but...

[mbbac]

How about a nice game of chess?

Of course, Mac OS X comes with Chess preinstalled.

Scant details on supercomputers...

[Spoing]

The Top 500 supercomputer list does provide the basic comparison information, though nothing like OS used or IO speed (network and storage). For that, you have to dig through each site and even then it is not easy to find. (The Earth Simulator uses SUPER-UX for the OS -- another Unix tuned to this type of task.)

That said, for what is provided, the Earth Simulator seems to be the current king by about 2x. (Corrections appreciated.)

Re:Scant details on supercomputers...

[d_strand]

I, for one, welcome our new supercomputer overlords.

Re:Scant details on supercomputers...

[Jack Auf]

I was also wondering about OS's used on the Top 500. I really wish they'd include the OS in the list as well. What would the OS statistics look like, I wonder?

IBM: 31%
HP: 28%
Linux: 20%
Sun: 12%
Cray: 6%
Other: 2.8%
Apple: .2%
Microsoft: 0%

Err, or something like that.

Re:Scant details on supercomputers...

[Spoing]

Well, this isn't for supercomputers, though for clusters I did find that about 90% ran some form of Linux, with about 3-4 each for Windows NT, Solaris, BSD, Tru64, and Other.

Here is the query I ran (top 256 clusters).

Re:Scant details on supercomputers...

[d_strand]

or maybe someone reads to much slashdot.

Check the #5 and #6

[Jesrad]

The Top500 site lists two competing 64bits architectures-based clusters: the Integrity rx2600, with 1938 Itanium2 at 1.5GHz (must be pricey), and an 2816 Opteron 2 GHz cluster, that achieves only three fourths of Big Mac's performance. Now that's a defeat for AMD.

Also, the VirginiaTech cluster is the only "self-made" supercomputer in the Top50 (the next one is ranked 63th, based on SunFire V60). The original #3 slipped to the 7th position because of the new supercomputers. Competition for that third place was tough !

Now where's the G5 XServe ? It was supposed to be out when OS X Server 10.3 was released.

Re:Check the #5 and #6

[hattig]

I read elsewhere that the interconnect used on the Opteron cluster would be the limitation. I think that it uses Myrinet which can't compete with 4x Infiniband as used on the PowerMac cluster.

Re:Check the #5 and #6

[lederhosen]

Also, the VirginiaTech cluster is the only "self-made" supercomputer in the Top50

And I guess the only one not using ECC memory. This is a *major* problem. Doing computations on many nodes during a long period *will* cause memory errors. This of course is both cheeper and also a bit faster (ECC has a performance hit). You will though not be able to do long computations. And short comutations must be run at least two times in order to check the results

Re:Check the #5 and #6

[damiam]

If you had read anything about this project, you would know that they have implemented their own low-overhead software error correction to compensate for the non-ECC memory. Presumably the benchmarks were done with this enabled.

Re:Check the #5 and #6

[lederhosen]

Then, as you have read it, you can tell the world how you do this "low-overhead software error correction to compensate for the non-ECC memory".

I am *very* interested in it!

Re:Check the #5 and #6

[lederhosen]

I have made this smart software that makes my
calculator solve NP hard problems in no time,
it is realy great, it has error recovery built-in.

Do you not trust me? why?

Re:Check the #5 and #6

[Bo Diddly Squat]

It's called running things twice. Look at the bottom of this article.

Re:Check the #5 and #6

[The Infamous Grimace]

"...Then, as you have read it, you can tell the world how you do this "low-overhead software error correction to compensate for the non-ECC memory"..."

Taken directly from interview with Srinidhi Varadarajan:

Q: How do you deal with Error Correction in Memory?

A: There's a lot of traffic on Ars Technica and other places. We do failure recovery, memory doesn't report. One of the things we've noticed is that failures aren't an issue yet. The reason they can be competent is the LINPACK test, which is showing 16 digits of accuracy. We are planning on moving to ECC systems in the future. They may have to run things twice for a bit.

(tig)

Re:Check the #5 and #6

[penguin7of9]

and an 2816 Opteron 2 GHz cluster, that achieves only three fourths of Big Mac's performance. Now that's a defeat for AMD.

Keep in mind that VT did a lot of assembly-level hacking before they managed to reach the number 3 spot. Presumably, LANL didn't bother since they probably didn't feel like they had anything to prove. Perhaps AMD will invest some time in doing something similar now.

Now where's the G5 XServe ?

Apparently, the G5's run so hot that making 1U rack mounts is difficult.

Re:Check the #5 and #6

[lederhosen]

Running things twice is *not* low-overhead.

When running on thusands of machines for weeks,
both runnings will most likely not be the same, thus
you will have to do them *many* times until the
result is the same.

See the thread about ECC on aceshardware.com.

Re:Check the #5 and #6

[lederhosen]

Yes, that is what I sugested.

Running things twice is *not* low-overhead.

When running on thusands of machines for weeks,
both runnings will most likely not be the same, thus
you will have to do them *many* times until the result is the same.

See the thread about ECC on aceshardware.com.

Re:Check the #5 and #6

[justins]

Does your L1/L2 cache support ECC?

On any modern processor, yes. Usually, on a PC, you'll find an option for it in the BIOS, and it's typically on by default...

Re:Check the #5 and #6

[justins]

Presumably the benchmarks were done with this enabled.

Why would they? It's not a requirement for the benchmark.

The benchmark is purely a measure of performance, not reliability or anything else. Of course the benchmark might end up being tainted by the creation of systems that just plain ignore reliability issues, but I think everyone involved in supercomputing knows to take these figures with a grain of salt.

Re:Check the #5 and #6

[tychay]

No, VT did not do a lot of "assembly-level hacking" one man working two months did port a bunch of code and he did use the best compiler and LINPAK on the market (Professor Goto's libraries). If LANL didn't do the same or better, I'd be disappointed.

Also you keep harping on the fact that it was "self-assembled." But then you go on to compare it to a system not provided by IBM, HP, NEC, or Cray but one provided by Linux Networx. Perhaps if VA Tech had gone to them, Linux Networx might have beat out IBM's Opteron bid of $9-10 million. But could have they gotten as low as $4.2 million--the list price from Apple?

You're going to have to face the hard reality that the Opteron may be an integer demon but the IBM 970 has it beat handily in floating point. The Rpeak of the 2Ghz Opteron (2 Gflops/s) is 1/2 that of the 2Ghz 970 (4 Gflops/s). Even accounting for the fact that the Rpeak->Rmax dropoff might be larger for the 970, that's too much to make up. Also, Virginia Tech considered the Opteron, but found that at the performance they wanted (specifically floating point performance) the systems would have cost twice as much ($9-10 million instead of $4 million) which is why the correctly opted for the 970 and which is why they're #3 instead of #6.

And that's without using Altivec/VMX/Velocity, since that unit can't do double precision add-mults.

As for heat issues. The 2Ghz 970 uses 47 watts which puts it approximately 1/2 the heat of a Pentium 4 and significantly less than the Opteron. IBM will be selling 2x1.6Ghz 970's in a blade configuration early next year and I'm sure "heat" isn't the reason for the delay. The issue with G5s in a 1U rackmount is that they won't exist until 1Q 2004 and it came down to availability. If you did any reading on the subject, you'd find that Virginia Tech's first choice was actually 970 systems from IBM, but they wouldn't be available in time--same thing happenned to the Dell Itanium 2 bid (it could have also been cost, Dell was "exploring pricing options"). The IBM Opteron bid was too high as was the HP Itanium 2 bid so they opted for Apple after the announcement. Smart move, two months of coding and several hundred pizzas later they have the #3 supercomputer. Any compromize, NCSA's gigantic P4 cluster would have beat them out.

Re:Check the #5 and #6

[Lars T.]

• Virginia Tech thinks they can handle.
• If it had ECC, it would still be faster.
• The point of having a fast computer is to have shorter computations.

Re:Check the #5 and #6

[damiam]

I'm not a benchmarking expert, but I can't imagine that LINPACK and other number-crunching benchmarks don't even include a single if statement to verify that they got the correct answers.

Re:Check the #5 and #6

[tap]

The lack of ECC is a very good point. I built a small beowulf cluster a few years ago, we had 20 nodes with 1/2 GB of ECC memory each. When running computations 24/7 I would get emails about single bit memory errors from the ECC reporting software I wrote about once a week. That was after I replaced a marginal DIMM that had errors about once a day.

My memory (128MB PC133 DIMMs from crucial, this was several years ago..) had an average of around .01428 errors per day per gigabyte. If bigmac's memory is ten times better than mine, they should see an error on the average of once every 3.82 hours. If their memory is one hundred times better, then the interval would be 1.59 days on the average.

Big mac does NOT have a magic software memory corrector. Such a thing is impossible anyway. Srinidhi Varadarajan said that they may need to run things twice. Well, if they're running stuff twice, doesn't that halve their effective GFLOPS from 10.28 to 5.14?

Re:Check the #5 and #6

[justins]

I'm not a benchmarking expert, but I can't imagine that LINPACK and other number-crunching benchmarks don't even include a single if statement to verify that they got the correct answers.

Well, sure. The point is that they don't need ECC to get the correct answer. They just need to keep running the benchmark until it completes successfully.

Re:Check the #5 and #6

[lederhosen]

Exactly! they have to run it at *least* two times. If they dont get the same answer, they have to make more runs.

My memory (128MB PC133 DIMMs from crucial, this was several years ago..) had an average of around .01428 errors per day per gigabyte. If bigmac's memory is ten times better than mine, they should see an error on the average of once every 3.82 hours. If their memory is one hundred times better, then the interval would be 1.59 days on the average.

Suppose Apple has this 100 times better memory that will fail every 2 days. How many times in avarage will they have to run a computation that takes a week? A month?

Re:Check the #5 and #6

[Hoser McMoose]

1. Go read for yourself.

I read it myself, it doesn't exist.

2. No, you explain it to me.

Nothing to explain because it doesn't exist. The poster is correct, the lack of ECC is a MAJOR shortcoming of this cluster. There is no software correction as some people seem to think, the "Deja Vu" software is something TOTALLY different designed to detect hardware failures, not soft memory errors on fully functional hardware.

ECC is a planned upgrade for the Big Mac. Until it arrives they plan on just doing every calculation twice and hoping that their calculations are short enough that errors rate will be sufficiently low/run times short enough that it hurt performance too much.

Re:Check the #5 and #6

[Hoser McMoose]

No, memory is not the only place that bit errors can occur. However, with 4.4TB of memroy, they are likely to get roughly 1 soft memory error an hour, and that's using pretty conservative estimates. Every other type of error is likely to happen at rates an order of magnitude lower and they STILL use error detection/correction for it.

Good chip, good performance, but there are reasons why the cluster was cheap. As long as Virginia Tech is aware of their limitations, they can get by. In the case of ECC, the limitation is that you can't trust your data, or even your instruction stream for that matter.

Re:Check the #5 and #6

[Hoser McMoose]

The code to run Linpack is very well known and well optimized for damn near every processor out there. LANL might be able to optimize their interconnect system a bit to squeeze some more performance out of it, but they aren't going to get much.

The reason that the PowerPC 970 does pretty darn well on this benchmark is the nifty little multiply-accumalate instruction. Since LINPACK is essentially nothing but multiply-adds, this instruction is VERY useful for the benchmark. A fair bit of real-world supercomputer code is also mostly multiplies and adds together (think solving large matricies), so the performance numbers are not entirely out of line. That being said, a new HPC benchmark suite is in the works to replace the rather simplistic LINPACK test, but it won't be here until '05 or '06.

As for XServe G5's, heat's a non-issue. The Opteron and the PowerPC 970 (aka the G5) consume roughly the same amount of power, and dual Opterons are all over the place. Heck, there are even some 1U dual Itanium servers! Now THOSE chips run HOT! An Itanium2 pumps out up to ~115W, compared with a maximum of only about 60-70W maximum for the PPC 970 and Opteron (though I should point out that documentation for Opteron is a bit weak in this regard and the PPC 970's documentation is basically non-existant to the public, so a bit of guesswork is required).

Re:Check the #5 and #6

[WoTG]

Does no ECC automatically mean there's no parity checking? If there are parity checks, then if no errors are detected, then chances are pretty good that there is no need to rerun computations.

Then again, I'm by no means an expert, heck, I don't even know if common memory still has parity bits. Does it?

Re:Check the #5 and #6

[noewun]

It's been posted before: the major reason for memory errors is thermal stress. As the G5s aren't 1U or blade designs where everything is packed pretty tight, are cooled well, and as the cluster has additional dedicated cooling, the chance of error is pretty low.

Re:Check the #5 and #6

[tap]

Suppose Apple has this 100 times better memory that will fail every 2 days. How many times in avarage will they have to run a computation that takes a week? A month?

Well, in order to answer a question like this you need to use something called the Poisson distribution. A rate of 2 days per error is the same as 3.5 errors per week. The probability of no errors in a week is then 3.02%. The average number of times you will need to try the calculation is then 33.1! And that's just the time to get a single correct answer! If you get 33 different answers, there is no way to know which one is correct!

So now you have the question of how many tries it will take to get two successful runs when the probability of a success is 3.02%. For this one uses a negative binomial distribution. The average number of tries until you get 2 successes is 66.2.

You can see why ECC memory is considered very important. Say their memory is 10,00x better than mine, then the average number of tries to get two successes goes down to 2.83. Divide big mac's GFLOPS by 2.83, and they drop to 16th place on the top 500 list.

How about one hundred thousand times better memory? Probably doesn't exist. The chance of a week long calculation failing is 3.44%. Small enough to not double check the answer? I doubt it.

Re:Check the #5 and #6

[justins]

ECC is a planned upgrade for the Big Mac.

Which is funny since there are not any Macs with ECC memory, and there never have been.

Re:Check the #5 and #6

[nusuth]

you will have to do them *many* times until the result is the same.

That is only true if the computer can't interrupt and reliably store the state of calculation. If it can, the running the same computation only(!) twice gives the correct answer. Say the average rate of 1 bit errors is 1 per hour, if the computer checks its state every one minute, one has to do triple calculations only once every sixty minutes. There should be an optimum value in which the sum of overhead of storing and retrieving data and overhead of occasionally doing more than two calculations will be lowest.

Re:Check the #5 and #6

[lederhosen]

Yes, I think ECC can find two-bit errors and
correct 1 bit errors.

Normal mem modules do not have any parity
(what I know of).

Re:Check the #5 and #6

[lederhosen]

Interesting, do you know what the chance is
that a job running on all nodes for a week
will terminate with no errors?

What I have read these errors are common.

Re:Check the #5 and #6

[lederhosen]

Yeah, thanks!

Re:Check the #5 and #6

[lederhosen]

Yes, that would of course be the "solution"
to the problem. Would not like to be the person
doing these hacks!!!

Re:Check the #5 and #6

[tap]

Back in the days of 30 pin SIMMs, you had 8 bit normal memory and 9 bit parity memory. Now in days of 168 pin DIMMs, you have 64 bit normal memory and 72 bit ECC memory. No such thing as 65 bit parity DIMMs, as memory got wider, parity advanced into ECC.

Re:Check the #5 and #6

[Luzumsuz Lazim]

This is all in theory. In practice life is not so bad. I often do long calculation using multiple nodes. And, I tell you this; memory errors is not the only source of problem when you deal with a supercomputer.

For this purpose we write programs that saves its COMPLETE state periodically. Thus, when an error occurs or we need to shutdown for some other good reason, we continue crunching from the latest good saved state forward. That way, the inefficiency is minimal.

And, all the serious programmers who writes codes for long jobs addresses this issue beforehand.

Thus, their benchmark at its face value is good for me. It is one hell of a machine!

And, an extra question for the bonus point: What would be the average time of completion of a job that requires 1 week of computation, and saves its state every 10 minutes with a "2 days per error" memory?

Re:Check the #5 and #6

[tap]

For this purpose we write programs that saves its COMPLETE state periodically. Thus, when an error occurs or we need to shutdown for some other good reason, we continue crunching from the latest good saved state forward. That way, the inefficiency is minimal.

The problem is there is no sure way to detect soft memory errors. If a node fails because of a bad fan or powersupply, you can detect that and restart your code. But if a bit gets flipped and changes some floating point value, how can you know? You may save your state every 10 minutes, but the only way to detect a flaw would be to run each 10 minute checkpoint twice and compare results. In the best case, that will reduce your thoughput by half!

As to the bonus question, I won't calculate it exactly, but I can say that it will take more than 2 weeks.

Re:Check the #5 and #6

[Luzumsuz Lazim]

Ooo! There is no doubt that ECC memory is useful. What I tried to stress is the fact that poison distribution doesn't apply here if the program you're using is smart enough. And, when there is ECC, and we have it by the way, the overhead of non-correctable memory errors is zero for all practical purposes.

And, without ECC when an error occurs in the executable part or in the logic-decision data, it is more likely that the application will crash, which is perfectly fine as we can continue from the last saved state. And, just for amusement, even the error occurs in the data part of the memory, at least for our case, it would be just another source of randomness we appreciate for a Monte Carlo simulation :)

Thus, the worse case requires a factor of two correction even with the worse memory chips that one can get, not the poison statistics. And, my guess is that, most of the users will prefer to run the application only once, as the executable part of the application often requires small amount of memory (means smaller possibility of errors), and the large data is mostly prone to single bit errors. Though, I understand that it may depends on the application. And, the thermally well design G5 system makes this risk worth to take.

And, although I can't imagine how, Varadarajan says that "We are planning on moving to ECC systems in the future." Does anyone know if the G5 memory subsystem supports ECC modules? Or, did he mean to replace the existing G5 boxes with newer generation ones that can support ECC, G5-server maybe?

Re:Check the #5 and #6

[lederhosen]

I belive they have at least to change motherboards.
I think the memory controller is on the northbridge
so the cpus can be reused. But I do not know how
they package these processors; maybe the processor is
on some sort of package, with the memory controller
on it and then the processors have to be replaced too.

What I think they will do is taking them to normal
computer usage and buy new macs for the cluster!!!

Oh, oh yeah?

[Faust7]

The sort of stuff that will make Big Mac look like that old TI-85 collecting dust in your drawer.

Cluster a billion TI-85s together and then we'll see who's collecting dust.

Re:Oh, oh yeah?

[Kr3m3Puff]

Well... that would make it only about a TFlop where Big Mac is about 10 TFlops...

Re:Oh, oh yeah?

[FuzzyFurB]

Seriously though, my TI-85 is not collecting dust in a drawr. No, I don't do heavy computation on it, but I still use mine on a regular basis. Miss programming assembly games on it too. :(

Re:Oh, oh yeah?

[JohnsonWax]

(Having owned a TI-85 many years back)

I'd love to see a chart that would show the combined computational power of every single TI-85, or VIC-20, TRS-80, etc. that was manufactured up against single modern systems.

Would a dual G5 be able to take a hypothetical cluster of every single Apple II ever made?

Re:Oh, oh yeah?

[Gabrill]

And a another chart showing how fast a system you need emulate the T-85 in real time . . .

BigMac isn't the only new addition to the top 10

[RalphBNumbers]

There are some other interesting semi-commodity hardware based new additions to the top 500 right under VT's #3 slot.
BigMac is certainly impressive, but even if these systems can't quite match it's scores, they deserve a mention.

4
NCSA
United States/2003
Tungsten
PowerEdge 1750, P4 Xeon 3.06 GHz, Myrinet / 2500
Dell
9819 Rmax
15300 Rpeak

5
Pacific Northwest National Laboratory
United States/2003
Mpp2
Integrity rx2600 Itanium2 1.5 GHz, Quadrics / 1936
HP
8633 Rmax
11616 Rpeak

6
Los Alamos National Laboratory
United States/2003
Lightning
Opteron 2 GHz, Myrinet / 2816
Linux Networx
8051 Rmax
11264 Rpeak

THe bandwidth demands are high.

[mindstrm]

Interconnect is very important.

This is nothing like distributed.net.

For a problem that can be broken into millions of discrete, independent chunks, sure, distributed.net's model is fantastic, and works really well... (seti, folding, distributed.net, etc)

For something where you need lots of feedback from nodes, (like these benchmarks, and lots of simulation work), bandwidth is everything.

Of course.

[Jesrad]

Since it's running MacOS X, you can play the games available for this OS. Search for any title here.

Re:Of course.

[IM6100]

That would be the few games in that short 2/3 length aisle in the back of the store that has everything the computer store offers for Macintosh, eh?

Re:Of course.

[Jesrad]

Yes, that part of the shop where software stays at $50 apiece even after sitting one full year on the shelf.

Re:Of course.

[IM6100]

A few years back I remember on repeated visits to Best Buy seeing that sad, tattered copy of Office 98. It grew more, and more shopworn, but the price never dropped.

Re:Of course.

[More Karma Than God]

The same thing happened to a few copies of Windows 95 (and later on Windows 98) at some local stores.

There seems to be a common factor here. B^)

Re:Of course.

[IM6100]

Generally the Linux boxes got remaindered out quickly. I remember the big bin of Corel Linux boxes.

Re:Of course.

[More Karma Than God]

You're right. I recall buying an old copy of Red Hat out of Officemax's bargin bucket.

What's the difference?

[mindstrm]

I mean, I understand reasonably well the benchmarks used... but my question is this:

In the past, we always looked to the DoE or DoD for who had the fastest computers... they had stuff we could only dream of.. huge, fast clusters of funky computers we've never heard of.

Now, a university built one out of macs... and it competes with the same benchmarks.

What I wonder is, are there applications the old-style supercomputers are still better at, or has technology simply advanced since then? (Things like 10gig ethernet and ghz processors and memory busses, etc)... have we simply surpassed them? Don't just feed me some line about I/O either....

Re:What's the difference?

[CrayHill]

There are numerous applications in the applied sciences where it is very difficult to take advantage of 10's or 100's of thousands of processors. There are a few applications that are termed "embarassingly parallel" because they are easy to run, but most are not. It is much easier on the harder applications to get some degree of parallel efficiency on a few fast cpus (the NEC/Fujitsu-type of vector supercomputer), as opposed to trying to be efficient on a cluster of G5's.

Re:What's the difference?

[pvt_medic]

But there are quite a few computers out there that actual power is classified.... how do we know this is really the fastest computer out there. we dont.

Re:What's the difference?

[Jeff DeMaagd]

have we simply surpassed them?

"we"? Are you on a "side"? I'd guess that you have nothing to do with any manufacturer or group involved with making new computer technology. I always thought it odd to take sides in something that has nothing to do with the person.

But I digress.

I think it just might be economies of scale. Billions of dollars are spent in CPU design and improved manufacturing to make a cheaper product, and to make that cheap product better. When one can ammortize the cost over millions of units one can fund a lot of research.

Re:What's the difference?

[cosmo7]

Also this list doesn't count alien supercomputers on other planets, or those in alternate dimensions.

Re:What's the difference?

[Hoser McMoose]

There is no one answer to this question, but there are a few things to consider.

First off, commodity processors simply have WAY larger economies of scale. NEC is just about the only company left designing actual processors for supercomputers (their SX series, as used by Earth Simulator, among others). The simple fact is that even if you sell 5,000 processor for a supercomputer, you're probably only going to sell one or two of those a year if you're REALLY lucky. For comparison, Intel sells ~130 million x86 chips a year. Simply put, there's a LOT more money going into the R&D for commodity chips.

The story is somewhat similar for high bandwidth/low latency I/O. Ethernet still isn't going to cut it except for pretty small clusters, but things like myrinet, quadrics and infiniband get you at least within the same ballpark as traditional supercomputer designs. They aren't quite there yet though, and this is still a major shortcoming of cluster design. However, there are at least two supercomputers I've seen that use a sort of hybrid design with AMD Opteron processors. AMD's Opteron has really nifty "hypertransport" point-to-point I/O connections right on-chip. A couple companies (most noteably Cray, but also a little guy called OctigaBay) are using the mostly-commodity hardware of AMD Opteron systems and a custom interconnect chips that hang directly off hypertransport links to create massively parallel processing supercomputers. This may be the way of the future for supercomputers, not exactly a cluster but still built from almost entirely commodity parts.

Next is the simple benchmark itself. LINPACK is a relatively simple benchmark that is fairly easy to run in parallel. Some supercomputer applications closely mirror this (Linpack is just solving linear equations after all, and that's the main thing that a lot of supercomputers are used for), others do not. In many cases though, I/O bandwidth and latency really become your limiting factor. Unless your data can be very easily cordoned off into little chunks that run on each node, you tend to spend all of your time waiting for data from remote machines. In the case of the Big Mac with 10GBit ethernet, you're looking at a best case scenario of getting 1/5th of the bandwidth of local memory (1.25GB/s vs. 6.4GB/s) and at more than an order of maginitude higher latency (~100ns vs ~2us; note of course that these are quick 'n dirty estimates, but the ratios are reasonably accurate). What's even worse, even on the nodes themselves the memory bandwidth is fairly low and latency rather high for most tasks (always has been for supercomputers, regardless of the arechitecture), so your taking a performance hit in an area that the system is already weak. Linpack doesn't necessarily show this weakness too much, but actual applications run on the supercomputer might. As always, YMMV.

I hope this doesn't sound like I'm just feeding you some line about I/O, but.. umm.. I/O is really where you differentiate the men from the boys in supercomputers.

FWIW, you might want to check out this PDF. It's Cray describing their new Red Storm architecture, and among other things it talks a lot about the challanges that supercomputers face.

Re:What's the difference?

[More Karma Than God]

> Also this list doesn't count alien supercomputers on other planets, or those in alternate dimensions.

Why not?

Re:What's the difference?

[esalathe]

Yes. LINPACK says very little about many real-world simulations, and the top-500 list says very little about who has the best computers for certain tasks. Climate modelling, one of the principal super computing applications, is a good example. Climate models do not parallelize over more than a few hundred nodes. Thus, machines with a few NEC SX-6 nodes are far better than those with many cheap nodes. The weapons of mass destruction research that dominates US super computing is much better suited to massively parallel computing than more useful science. Unfortunately, that application has dominated US super computer design to the detriment of areas like climate modelling where European centers with Japanese super computers are leading. Another thing to consider is that these machines don't run a single process that uses all nodes. Most have hundreds of processes running, each using a small subset of the available nodes. The gloming together of many nodes to run a single benchmark is pretty much an artificial excersize. The only real advantage of a large cluster is 1) bragging rights and 2) administrative. -Eric

Re:Still use mine...

[Rude Turnip]

I've had my TI 85 since high school and I use it to this day (going 10 years now). It's been used for everything from chemistry to physics, calculus and finance and is full of custom equations I made to handle each situation.

my switcher story

[caffeine_monkey]

I used to run an Intel-based supercomputer, but then one night, I was modelling a nuclear explosion on it, and all of a sudden it went berserk, the screen started flashing, and the model just disappeared. All of it. And it was a good model of a nuclear explosion! I had to cram and remodel it really quickly. Needless to say, my rushed model wasn't nearly as good, and I blame that Intel supercomputer for the fact that DARPA yanked our funding.

Re:my switcher story

[adrianbaugh]

Dare I suggest you should blame your evident lack of backups? Real Men may well not use backups, but fake men don't get fired :-P

Re:my switcher story

[jjohnson]

*whoosh*

Re:my switcher story

[adrianbaugh]

Oh well.. I thought I was making an insightful point about backups and personal responsibility and it gets modded twice, once as funny and once as overrated. It is National "Cretins Get To Moderate On Slashdot" week or something?

Re:my switcher story

[adrianbaugh]

Yes. Thank you for providing such a complete demonstration of this fact, Mr. Anonymous Retard.

Re:TI-85?

[ducomputergeek]

I'm still using my HP 48G from circa 1993. A new set of AAA once every couple years and the onling thing it won't do is calculate IRR...had to break down and purchase an HP business calc for that...

Re:Yeah, but..

[Apiakun]

Yes, I remember... Big Mac, McDLT, a quarter pounder with some cheese, filet o fish, a hamburger, cheeseburger, happy meal, mcnuggets, ??? fries, a salad, chef or garden, or a chicken salad oriental... something along those lines

Re:Still use mine...

[_Shad0w_]

Mine doesn't get used for much beyond general purpose calculation now days - I too have had it since I was about 16, nearly 10 years now. I used it while I was studying Physics, Chemistry, Maths and Computer Science at A-Level. Funny thing is I no longer have anything to do with any sciences if I can help it. Come to think of it, this week is the first time it's been used for "real" calculations for quite some time (I was playing around with Mersenne Numbers and the Lucas-Lehmer Test, was using the ol' TI-85 to double check my program).

miniatyrising

[gl4ss]

**IBM has managed to cram 1,024 PowerPC 440GX processors into a slanted cabinet the size of a dishwasher. The unit -- described by IBM as a small-scale prototype of Blue Gene/L -- is already ranked 73rd in the new Top500 list.**

now that's intresting.. you could fit one hell of a cluster in your basement.

Re:miniatyrising

[KillerHamster]

Or it could go in our kitchen, in the spot where the dishwasher used to be.

Re:miniatyrising

[Hoser McMoose]

Only problem with that is that you'll probably have the cops beating down your door thinking your got some huge dope-growing operation!

"No officier, I swear it's not a hydroponics setup that is eating up 1MW of electrical power, it's just my the supercomputer in my basement!"

"Sure son. Book him boys!"

Its also the CHEAPEST

[goombah99]

Mac has never been known for being cheap, just a good value. Now even this seems to be fading: macs are a cheap too!

Now this system is the cheapest of the top 10. its cheaper than many it beat by a factor fo ten (more than that considering some of the building infrastructure are in that figure). Even more interesting these were stock mac at full price loaded with DVD-roms, firewire, blue tooth, the OS, etc..---not some stripped down model.

Its a good bet too that this thing is going to have lower maintainence costs and higher up-time given the macs attention to cooling, the use of high quality hard drives and power supplies, and high end memory chips. (on our cluster a tenth that size we blew 60 hard drives in the first 6 months and had to replace 10% of the motherboards.

Re:Its also the CHEAPEST

[tychay]

goombah writes:

Now this system is the cheapest of the top 10. its cheaper than many it beat by a factor fo ten

Let's not get ahead of ourselves, it's more like a factor of two, not ten (it's a factor of 20-30 cheaper than #1 and #2 and some of the older members of the list, sure). The Mac system price cost $4.2 million educational list. However, #4-6 were built using the Pentium 4 Xeon, Itanium, and Opteron respectively. The systems cost for these would be around $9-10 million if they were built today.

Not that there isn't a lot to say Mac isn't "back". After all, of the four, it offers the highest flops/cpu, the second most flops/cycle (Rpeak is the same/cycle as Itanium but it's not as efficient so it's Rmax gets edged out), and the best price/performance (by a factor of 2 as noted above).

I hope when IBM introduces 970-based blades next year, a lot of Linux users take a serious look at it. With 90 nm next year and Power5-based GPUL's coming down the pike, Big Blue-based Linux compute nodes should be looking mighty impressive as the most bang for your Linux buck. I can't wait for Opteron and Itanium blades. Combined with the P4 Xeon blades they already have, you could have 4 different CPU families in the same BladeCenter!

Re:Its also the CHEAPEST

[Anonymous Coward]

The fans operate at low RPMs, and are pretty high quality (compared to some of the shit that's out there).

None of the Beige G3's that our company is still running (24/7) have blown a single fan. Compared to our Dells (blown motherboards, power supplies, and fans), and even our Suns (two of which are making some very bad fan noise because the bearings are about shot)--which our macs outnumber handily-- they're virtually indestructible.

Re:Its also the CHEAPEST

[Billly Gates]

True.

Most of the 1984 macs still run! I heard stories on slashdot about Apple laser suddenly not working. Upon inspection the guy found an old macII with an inch of dust on it hidden behind a bookcase. Running since 1987 and forgotten untill recently when the nic failed. Heheh

How many of you have old 486's or pentium's that still run? Mine have old died.

But a mac will run forever.

Re:Its also the CHEAPEST

[scrod]

They used to, anyway, back when they were still being manufactured in Cupertino by Apple themselves. Now that Apple doesn't make all their own components the average life span of a modern Mac is nothing compared to what you found in the 80's. Though I suppose one could say the same thing to some extent of VCRs and televisions, too.

Re:Its also the CHEAPEST

[Feztaa]

How many of you have old 486's or pentium's that still run?

I do, I do! :)

I have an old 486, it still works fine. I leave it off most of the time, simply because it's just not useful to me. It runs a recent LFS system (maybe a year old). Slow as hell, of course, but I think it's impressive that a very recent version of linux is running on an old 486. That impresses me, just try running XP on it :)

Re:Its also the CHEAPEST

[Cecil]

I would guess that your guess is wrong.

Have you seen a G5? And do you know why your typical fans fail? Could it be perhaps because they are undersized, underpowered, cheap fans being pushed to their limits to try and brute force enough air churning around to keep your system cool? The G5s have a properly engineered case with a specific path for airflow, and abundant, high-quality fans mounted on rubber shock absorbers to dampen vibration. I suspect that these fans will have few problems. And if they do, won't it be a painful process to replace them: unlatch side panel, remove plastic airflow enclosure panel, put hand on fan assembly, slide out, slide in new fan assembly. Click, done. I wish my servers were that easy to replace things on.

Cheap hard drives? Yeah, sure, whatever. They use Seagate Serial-ATA drives. I don't know about you, but Seagate stopped fitting into my definition of 'cheap and crappy' about 8 years ago.

I have never dealt with ECC memory, so I grant you that point, however.

Re:Its also the CHEAPEST

[lederhosen]

From personal experience the hardware on
our Sun boxes at school has little problem.
The macs and the PCs have much more problems.

Now, the harddrives are the same used in PC:s,
if you think these drives are near the quality
of high end scsi drives then I think you are wrong. I did *not* say 'cheap and crappy' I
said cheap. This is a supercomputer, and I
think it is worth having great uptimes.

Having 9 (right?) fans per node, and having so
many nodes, many will break.

Re:Its also the CHEAPEST

[Alan Partridge]

Nonsense, mass produced items are almost always more reliable than custom jobs. When you tool up for mass production, you make sure that you get it right - that's why a Ford Focus is more reliable than an Aston Martin DB7.

Re:Its also the CHEAPEST

[TrancePhreak]

I have an IBM XT from the 1984 era that still runs.

Re:Its also the CHEAPEST

[lederhosen]

I do not agree.

Most important, what has this to do what I said?

Re:Its also the CHEAPEST

[slashnull]

that's why a Ford Focus is more reliable than an Aston Martin DB7

Can you cite a reference for this or is this not a fact? I'd sincerely like to know.

Re:Its also the CHEAPEST

[RockClimbingFool]

the ford focus has had MANY recalls and service bulletins since it has come out. i know the point you are trying to make, but god, the ford focus?? check you facts man...

Re:Its also the CHEAPEST

[be-fan]

I've got a 486/33 that's running just fine.

Re:Its also the CHEAPEST

[nocomment]

Anyone have a price comparison of Bic Mac vs #1, and #2?

Re:Its also the CHEAPEST

[ZackSchil]

Have you ever even looked at a G5, seen its internal design, read the technical documents outlining airflow zones and the large, high quality, low RPM fans, the controller that measures temperature and keeps them moving only when they need to? No? Ever looked inside a generic PC box (as in technical slang and literally a box) with little to no airflow and 2 or 3 noisy fans drilling away as fast as they can the whole time trying to circulate air in and out of the little holes in the back? The G5 has been designed to keep cool with as little noise and wear as possible.

Re:Its also the CHEAPEST

[JDWTopGuy]

Me too! I have a 90Mhz pentium and a 166Mhz pentium.

The 90Mhz is running Slackware 9.0 and a bunch of other crud (see me latest journal entries, mate) and it works pretty well. If I had disabled external SSH access (not running the latest version :-O) I'd post a link to it right now... it'd be quite interesting to see it slashdotted... although it's dial-up, so I 'spose it wouldn't get enough bandwidth to do much more than clog my connection.

The 166Mhz I'm using right now, and I use it to play descent and stuff like that. Again, see my journal if you really care.

Re:Its also the CHEAPEST

[MrTangent]

"The machine is the first supercomputer based on Macs; it is one of the few supercomputers built entirely from off-the-shelf components and it cost a bargain-bucket price -- only $5.2 million. By comparison, most of the top 10 supercomputers cost about $40 million and up. The Earth Simulator cost $350 million."

The Earth Simulator is #1 on the Top 500 list as seen here. Quote taken from here.

"The $200 million (US$) computer is the fifth in the DOE's nuclear weapons simulation program, called Advanced Strategic Computing Initiative (ASCI), which allows the U.S. to keep its nuclear stockpile while complying with the nuclear test-ban treaty. Named "ASCI Q," the newest supercomputer will be capable of performing 30 trillion calculations per second.*"

Taken from here.

So, the #1 Supercomputer, the Earth Simulator, cost $350 million. The #2 Supercomputer, the ASCI Q, cost $100 million. The #3 Supercomputer, the Big Mac, cost a mere $5.2 million.

*Looks like ASCI Q got only 13.88 trillion calculations per second, not 30.

Re:Its also the CHEAPEST

[Alan Partridge]

The Germans seem to think it's pretty reliable.

Re:Its also the CHEAPEST

[ZackSchil]

I know why he wrote that, but Mac zealots, from years and years of such comments, are extremely knowledgeable of Apple's product line. Even more so when Apple touts the G5's design as one of the nicer things about it both aesthetically and technically. They tout the cooling zones, etc right on that page.

Obligatory slashdot gag

[adrianbaugh]

Imagine a beowulf cluster of..... oh, wait.

Re:Obligatory slashdot gag

[More Karma Than God]

Amazing. The old Beowulf cluster joke has become funny again.

Re:Kudos to the Mac (don't forget the others)

[tychay]

Hmm, guess this means my submission a couple hours ago won't go through (dangit, Wired!)...

Here is the official press release and the list.

There is a lot of good points to note all around. The first is the G5 Terascale cluster at Virginia Tech at #3 (10.28 Tflops/s, 2200 CPU, Infiniband) is the first academic computer to break 10 teraflops/s. This extra performance was promised at Mac OS X Developer's conference last month. Not to sure if the price is a testament to Infiniband ($1.5 million cabling, cards, and routers) or the Macs ($4.2 million list).

Good thing too because in a surprise move the NCSA cluster made the list at #4 (9.82Tflops/s, 2500 CPU, Myrinet). This cluster is built using Dell's running Pentium 4 XEONs and Red Hat Linux! One subtle point to note is that they didn't get all the systems online in time (there should be 2900 CPUs, not 2500). I bet some programmer at PSC and an ex-Chief Scientist of SDSC is appreciating having a hand in edging out NCSA for #3--not to mention Apple beating Dell for #3.

The fastest Itanium cluster is at #5 (8.63 TFlops/s, 1936 CPU, Quadrics) which is looking like the odd man out boxed in by a PC based systems using Myrinet, the P4 Xeon above, and the most powerful Opteron system at #6 (8.05 Tflops/s, 2816 CPU, Myrinet). Another point of similarity:did I mention it's also using Linux?

And finally, It's easy to overlook #73, a single compute node of BlueGene/L (1.44 Tflops/s, 1024 CPU). Imagine 128 of these connected together and you have something that will easily take #1 when it's completed even if we handicap it 20-40%. As noted on SlashDot earlier, this will be running Linux.

Re:Kudos to the Mac (don't forget the others)

[tychay]

I miffed some of my links in the parent post:

Here is the reference to the ex-SDSC scientist.

Here is the link showing that the Opteron cluster is using Linux Networx.

Finally in the interest of full disclosure and to pre-empt the anti-Mac zealots, I should mention that the $4.2 million for the G5 machines is probably the education list price, because when you go to Apple Store, putting 2GB of RAM into 1100 2x2Ghz G5's will cost you $4.4 million (+ a little more for having some spare machines).

The TI-85 lives in my Zaurus

[tropicflite]

Texas Instruments Ti85 emulator

Where have all the g5 gone?

[n3z0rf]

Well now we know where they are being stockpiled. ;)

Re:Kudos to the Mac (don't forget the others)

[Anonymous Coward]

Good thing too because in a surprise move the NCSA cluster made the list at #4 (9.82Tflops/s, 2500 CPU, Myrinet).

That's pretty funny. A Pentium cluster with 2,500 CPU's is slower than a Mac cluster with only 2,200 CPU's.

What's also funny is that the hottest competition in supercomputing today is between Apple and Dell. Remember when it was between SGI, NEC, and Cray?

It's interesting...

[zeux]

that more and more of the new supercomputers use the same processors than everybody's computer...

Xeons, Opterons, PPC 970s, ... and I think this trend will continue in the future.

I think this is because of the low cost of these processor.

But I still can't understand how a so expensive G5 (for personal use) can blow everybody in term of performance/cost. Amazing, it's obvious that we gonna have more supercomputers powered by G5 in the near future.

Re:It's interesting...

[wankledot]

I think you'll find that the G5 is very competitive with Xeon systems, when it comes to workstations. Opterons... I think that remains to be seen.

Re:It's interesting...

[zpok]

That's because for the offered infrastructure it is an extremely cheap machine.
It's expensive as in "Costs a lot of money, more than I want to - scrap that - can pay".
It's cheap as in "extreme bang for your buck".

On another note:

This is normal: "I want one, I'm jealous, I hate you!"

This is also normal: "For the same buck I can build me a machine just as impressive AND be in game heaven. Ahahahahaaaaahaaaaaaarghl!"

This is irrational: "I'm absolutely certain, despite all the evidence to the contrary that this machine simply cannot be priced right. It's an Apple, so it's got to be expensive. Whatwhat?"

This is pittyful: "Oh, you know, speed isn't everything. Just as with size..."

Congratulations VTech

[gsdali]

Now to get on with the research. It's a credit to them that this computer got from the drawing board to fruition in the tiny amount of time that it did. It's raised the bar for price/performance in the research computing world and hopefully many less wealthy institutions (I'm looking at UK universities especially here). At the end of the day its about the research they put into it and the results they get out of it.

Re:BigMac isn't the only new addition

[gcondon]

Yes, I agree - these machines do deserve a mention. However, not necessarily a favorable one.

For example, the Tungsten machine at #4 uses 14% more processors at a 53% higher clock rate to achieve 95.5% of the Rmax and 87% of the Rpeak.

The Lightning machine at #6 uses 28% more processors at the same clock speed to achieve 82% of Rmax and 74% of Rpeak.

I'm not impressed.

I only have a TI-83+

[Unregistered]

you insensitive clod

I'm confused...

[Spoing]

Aren't there supposed to be comments about Microsoft here? This is Slashdot -- isn't it?

Well, to get the ball rolling, here is a query on the top 500 supercomputers using Microsoft Windows. Corrections and insight are appreciated.

Re:I'm confused...

[You're All Wrong]

I have a correction to make. Your post is presently marked (Score:1), which is obviously a grave error.

Alas I can't do the honours.

YAW.

Re:I'm confused...

[You're All Wrong]

I find it odd how someone can _still_ whore by posting a single obvious URL, and get to 5 immediately; likewise it's a shame that genuine wit like the above isn't rewarded more swiftly.

Glad to see it on 3 now.

(I'd love a '-1' whoring option..., yes I know I'm OT, sorry.)

YAW.

Real world??

[Zane Edwards]

Yes, it might make some fast scientific calculations really really fast, but I want to know how fast it does some real world stuff. Give me some Quake framerates, or Photoshop gaussian blur benches.

Re:Real world??

[saddino]

Give me some Quake framerates, or Photoshop gaussian blur benches.

Or if you're PC World, some Premiere or MS Word benchmarks.

Re:Real world??

[Daniel Dvorkin]

I'm assuming you're joking, but here's a serious answer: for a supercomputer, scientific calculations are real world stuff. A significant chunk of the processing power in the world goes to things that the average desktop user will never see directly. I like a fast framerate on Quake or a fast Photoshop filter as much as anyone, but as a comp. bio. grad student, I also really appreciate a system that can run my bioinformatics apps in a reasonable time.

Re:Real world??

[Zane Edwards]

Yes, it was meant as a joke. I was going to originally use just Quake framerates, but creating gaussian blurs are equally useless for me. Seriously, though, setiathome kicks ass on my G5, about 1 unit every 2 hours. I wonder if Berkeley should be using a G5 supercomputer and forget everyone else!

Re:Real world??

[commodoresloat]

Or at the very least, let's see how fast it copies a 17M file!

Infiniband, not G5

[binkleym]

The G5 is a cool processor, but it isn't the reason the VT cluster is so fast, the Infiniband interconnect is. The LINPACK benchmark that is used to determine position on the Top 500 list depends very strongly on the latency of the network connection.

Infiniband has ~ 8-12 us latency (probably even less by now), while ethernet is an order of magnitude slower. In real-life applications it's actually worse than this suggests.

We have tested a real-life application (socorro) using both gigabit ethernet and Myrinet (slightly slower than Infiniband), and gigE took 600 seconds to finish a run, while Myrinet took 4.

VT's cluster is using the largest Infiniband network yet built (or at least announced). The previous largest Infiniband network was O(100) machines. VT could have built the cluster using Xeons, Itaniums, or Opterons and arrived at roughly the same level of performance.

Re:Infiniband, not G5

[afantee]

>> VT could have built the cluster using Xeons, Itaniums, or Opterons and arrived at roughly the same level of performance.

This is not true at all. VT clearly has stated in their presentation that G5 has the best performance / price for what they do.

The 1.5 GHz Itanium 2 costs over $3000 per chip, and even the 32-bit Xeon 3.06 GHz is about $1000, while the 2 GHz PPC 970 is about $300 or $400. In addition, VT wants 64-bit chips, so Xeon is a nonstarter.

Re:Infiniband, not G5

[binkleym]

Note: I am completely processor/OS agnostic. If I could get the best performance for my cluster by by running Windows 3.1 on a Z80, that suits me. "Religion" doesn't sway me, only real-world application results.

I have inside contacts on the VT cluster. The money for it came from state funds, and was going to vanish within a few weeks. When the deal with Dell for Itaniums fell through, they didn't have time to wait for Opteron systems to come out because the money would be gone before they came out, so VT went with G5's because they were the only other 64 bit system available at that time. If they had another month, there's just as good a chance that they would have gone for Opteron.

According to Pricewatch:
3 GHz Xeon $473
Opteron 242 $297

So your Xeon price wasn't quite accurate. Admittedly, you lowballed the Itanium. I couldn't find it anywhere for that low. :) But Intel is coming out with a "low-price" Itanium next year.

Re:Infiniband, not G5

[foniksonik]

Maybe the G5 has the cheapest CPU that is interoperable with infiniband and that is why they chose it? Remember that the G5 is a computer not a processor and you have to look at it as an integrated solution not a selection of components. PPC970 would be far less useful in a poorly designed computer.

Re:Infiniband, not G5

[afantee]

>> So your Xeon price wasn't quite accurate. Admittedly, you lowballed the Itanium. I couldn't find it anywhere for that low. :)

A quote from http://www.cbronline.com/current-issue/cdd51c7c909 59c1680256d7c0018cf5e

"But a quick glance at Intel's price list shows how anchoring Opteron's price to Xeons could boost AMD's top line. AMD's Opteron single processor 100 Models range from $229 to $438. Intel's Xeon runs from $198 for its 2.4GHz with 512Kb of cache part to $690 for its 3.06GHz with 1MB of cache. Intel's multiprocessor Xeons start at $1,177 and run up to $3,692, while AMD's dual processor 200 Models range from $256 to $794 and its multiprocessor 800 series runs from $749 to $2,149."

>> But Intel is coming out with a "low-price" Itanium next year.

G5 / PPC 970 will reach 3 GHz next summer, and IBM is coming out with G6 / PPC 980 soon after.

Re:Infiniband, not G5

[binkleym]

Um, no...

They chose the G5 processors because it was the only affordable 64 bit processor they could buy before their money vanished. This month, there's an equally good chance it would be an Opteron system.

Ask them.

Re:Infiniband, not G5

[Hoser McMoose]

Actually I would tend to disagree, in a manner of speaking. Yes, high-bandwidth/low latency I/O is termendously important in real-world supercomputer applications, but Linpack doesn't always show this, Linpack can be run in parallel pretty easily. The reason why the cluster does so well in this test is very much related to the processor itself, though it's performance in other applications may end up being much more closely connected to it's I/O performance.

The real key here is that the PPC 970 has a multiply-accumulate instruction and can decode and retire two such instructions each clock cycle. Since Linpack does mainly multiplies and adds on data, this instruction is ideal and makes things go REAL fast. Notice that the Big Mac has an Rpeak value of 8 GFlops per 2GHz PPC970 processor as compared to the 4 GFlops per 2GHz Opteron processor for LLNL Lightning.

FWIW there are Infiniband clusters using Xeons and Opterons in the list. They don't really do any better, clock for clock and processor for processor, than the Myrinet clusters do in Linpack, though they probably would in many real-world supercomputer applications.

Re:Infiniband, not G5

[gumbi west]

VT could have built the cluster using Xeons, Itaniums, or Opterons and arrived at roughly the same level of performance.

Yeah, or even comadore 64s. And the racks would have looked even cooler if they were filled with comadore 64s.

Re:Infiniband, not G5

[crawdad9]

I'm sorry, but your information on the source of funding is not correct. The money for the cluster is coming from multiple sources entirely within the University, including the Colleges of Engineering and of Science, Research Computing, and others. This amazing project was not the result of any special support from the state, or a "donation" from Apple or any other company. The VT cluster happened because enough people on campus realized it was important and found the funds to make it work.

Re:BigMac isn't the only new addition

[Johnny Mnemonic]

I agree. And that's before cost considerations: do you wanna bet that the NCSA at #4 is in fact more expensive than Big Mac?

Wishful thinking...

[Anguo]

At first, I read this:

An anonymous readers noted that according to wired, It will be announced officialy on monday MacOSX is the third Operating System on the market, behind Microsoft Windows and GNU/Linux. The article notes that due to an impressing growth rate in rate years, the Free OS has taken over Mac OSX [... snip rest of article].

Re:Wishful thinking...

[Anguo]

...an impressing growth rate in recent years...

I did preview but I missed that one... [grrr\]

Apples and Oranges

[penguin7of9]

Now this system is the cheapest of the top 10.

You may notice that the BigMac is listed as self-assembled. It just doesn't make too much sense to compare a self-assembled system against company-delivered systems. You could do exactly the same thing VT did with G5 Macs using dual-processor Opteron or Xeon systems and end up paying significantly less.

Why don't people? Because most places that build machines that large want something turnkey and something that is supported as a cluster.

Its a good bet too that this thing is going to have lower maintainence costs and higher up-time

Actually, the opposite is probably true: desktop machines, like the G5, do not make good cluster components. They require lots of space, are difficult to access when sitting on racks, etc. And Apple's history on quality is not uniformly good anyway.

Re:Apples and Oranges

[beelsebob]

You could do exactly the same thing VT did with G5 Macs using dual-processor Opteron or Xeon systems and end up paying significantly less. Actually, the exact reason that they went with the G5 was that they contacted lots of other major suppliers and found the G5 to be the cheapest in terms of power/pound. Bob

Re:Apples and Oranges

[prichardson]

Actually, you are horribly horribly wrong about two things.

You definitely could not do that with Opteron or Xeon systems. VT was in negotiations about price and delivery time with Dell and Apple. Apple beat out Dell's prices (shocking!!!).

Also, the G5 makes a great cluster computer. It comes standard with gigabit ethernet and has very easy access to parts (no screws required to install anything).

Finally, the Apples make a good cluster because in 5 years or so when they disassemble it they have 1,100 really nice desktop machines. PC's need to be upgraded more often to serve as a desktop computer (that's why Macs have awesome resale value compared with PCs).

Re:Apples and Oranges

[daknel]

>And Apple's history on quality is not uniformly good anyway.

I disagree

Re:Apples and Oranges

[penguin7of9]

You definitely could not do that with Opteron or Xeon systems. VT was in negotiations about price and delivery time with Dell and Apple.

Because Dell can't put a cluster together at a good price, you conclude that "[one] definitely could not do that"? That's absolutely idiotic. Dell is overpriced and they didn't even sell Opterons at the time.

Apple beat out Dell's prices (shocking!!!).

That's "shocking" only to people who don't have a clue what's going on in the PC industry. Of course, Apple beats Dell on price. They probably also beat IBM and Sun. That's not where anybody who is cost-conscious would buy their cluster machines.

Also, the G5 makes a great cluster computer. It comes standard with gigabit ethernet and has very easy access to parts (no screws required to install anything).

Yes, and if you buy a dual-Opteron 1U rack mount, not only do you get two Gigabit Ethernet ports on the motherboard, you also get something that provides even easier access to parts.

If you think that a G5 is easier to maintain than a rackmount, you really have never touched a rack mount unit.

Finally, the Apples make a good cluster because in 5 years or so when they disassemble it they have 1,100 really nice desktop machines. PC's need to be upgraded more often to serve as a desktop computer (that's why Macs have awesome resale value compared with PCs).

PCs running Windows need to be upgraded frequently because people want to run the latest version of Windows. If you run Linux on a PC, it has an even longer lifetime than Macs (and, in fact, five year old Macs will not run current versions of OS X particularly happily).

Re:Apples and Oranges

[penguin7of9]

I have owned half a dozen Macs. They all looked nice and they all felt solid. But some of them have had a lot of hardware problems. Others worked like a charm. As with any computer company, Apple does not always get it right. The G5, being a new architecture and design, is something I would have been somewhat wary of until it has proven itself. As it is, we haven't heard of any major problems with it, so it seems to be OK.

But if you want quality PC hardware, you can certainly easily get that, too.

Re:Apples and Oranges

[linuxpyro]

It's a big complex setup. Therefore, in some way or another, it is going to give someone hell.

Re:Apples and Oranges

[ivan256]

If you think that a G5 is easier to maintain than a rackmount, you really have never touched a rack mount unit.

Up until about a month ago I worked for a linux clustering company. I've worked with upwards of 50 different styles of rackmount computers ranging from 1/3U to 8U designs. I've also had a PowerMac G4 cluster (4 nodes, dual 500Mhz CPUs). Let me tell you that being rackmount doesn't make a system easier to maintain than a Mac. Sure, some rackmount systems are a joy to work on. Most aren't. The ones that are aren't cheap or 1U, either. In fact I've never seen a 1U system that I would prefer to service over my PowerMac. Top of the line IBM Xseries, Dell, and HP machines included.

If you think any old rackmount machine is easier to maintain than any recent PowerMac, you really have never touched a modern Mac.

Brute Force supercomputing = meh

[Boone^]

Run down the list and look at processor counts. We've got 5120 at the top (vector), but number 2 needed 8192 to get the job done. BigMac at #3 drops to 2200 and the processor counts hover in that 2000+ category. Until #19, when Cray's X1 jumps in at 252 processors.

Having a fast computer is cool and all, but if you can do it with 252 CPUs instead of 1024 (#22, P4 2.4), isn't that a win?

Besides, LINPACK doesn't stress interconnect latency and bandwidth, only cache and memory performance. When you run a "real" codes on these Mac/Xeon clusters and get 5% efficiency, suddenly the Earth Simulator (and the small Cray X1's) look good when they blow well past the 50% efficiency mark.

Re:Brute Force supercomputing = meh

[2nd Post!]

It means you solve different problems with different computers. Isn't that obvious?

There are solutions where multiple, numerous, nodes are fairly efficient. Something with coarse granularity and high compute effort, so that you can allocate per node infrequently. In those situations, something like the VT cluster is cheaper, more cost effective, and more capable than the Earth simulator because you can built many of them for the same cost.

Different tools for different problems, I think is the conversant idiom.

Re:Brute Force supercomputing = meh

[Jeremy Erwin]

Having a fast computer is cool and all, but if you can do it with 252 CPUs instead of 1024 (#22, P4 2.4), isn't that a win?
It really depends on whether you think liquid cooling is a win.

Besides, this article implies that the ORNL machine is only half finished.

The rest of the initial setup of eight cabinets won't arrive until ORNL completes construction of its new 40,000-square-foot computational sciences facility.

At 64 processors per cabinet, the "initial installation" will be 512 processors.

G5, not Xeons, Itaniums, or Opterons

[2nd Post!]

Yes, you're quite right, the networking hardware is important.

But as researched by the VT folk, the G5 is significant: It was cheaper for their needs than the Xeons, Itaniums, and Opterons of similar performance and energy consumption!

So both component choices were critical to their achieving number 3.

Can someone tell me...

[Whatah]

Why does every product apple make have its own icon ?

And yet equally, if not more, important products like amd64 don't have their own icons ?

Additionally, why does this CPU have a G5 icon? And not a PPC970 icon ?

Has slashdot sold out to apple ?

Re:Can someone tell me...

[2nd Post!]

1:) Because the editors of slashdot have decided it should be that way.

2:) Because the importance and equality of AMD64 is not your decision, nor is it's icon.

3:) Because all 1,100 nodes of this computer are powered by Apple's PowerMac G5; it just happens that Apple's PowerMac G5 is also the IBM PPC970, and to call it the PPC970 would detract from Apple's achievement. To be fair, it is also IBM's achievement, but they have their own supercomputers on the list.

4:) No, of course not. To sell out is such a close minded and antagonistic viewpoint. The slashdot editors just happen to like Macs more than you do. You just haven't kept up with how cool Apple is nowadays.

Cool article, but Wired expands on 'future bashing

[danigiri]

The article is cool enough (first official confirmation of the 3rd spot) but after stating this in simple terms, they go on bashing Big Mac with the 'future' clusters that will beat it big time.

It might be well true that BM will be beaten, but please, a more positive spin on the present achievements would be in good order. If in 6 months it gets beaten, yeah, cool, but it will be *then*, not now. Pleaaseee, wasn't it over, that pointless subliminal Apple bashing?

dani++

The name "Big Mac"

[mslinux]

I don't know much about trademarks and copyrights, but I'm wondering if McDonald's can sue Virginia Tech if the university begins using the name "Big Mac" to refer to this cluster? Obviously it's a lot of Macs that take up a lot of room and power in order to do 10+ teraflops/sec (hence the name "Big Mac") but is it legal to use that name?

Re:The name "Big Mac"

[saddino]

Generally, a trademark offers protection in a market segment. For instance, Domino's Pizza and Domino's Sugar each have a respective trademark. You cannot start a confectionary supplier called "Domino's" nor can you start a Pizza place called "Domino's."

Big Mac is certainly a trademark in the food industry, but that may not preclude its use for other products, e.g. the name of a supercomputer. Especially if there is not likelihood of confusion on the part of the consumer (nobody is going to think the supercomputer has anything to do with sandwiches).

But I say "may" because McDonald's could conceivably argue dilution of trademark...it's an extremely famous mark.

Re:The TI-85

[TheRaven64]

The TI86 was, for the most part, even nicer. Basically, it was a TI85 in a nicer case with more RAM (128K Vs 32K). I'd have to check in the manual if it could determine if a pixel was set (and the manual was about twice as big as the machine itself...), but the most useful upgrade was a trivial feature; a multiple entry history buffer (the '85 could recall the last thing you typed, the '86 could recall the last few things).

Unfortunately, addressing the 128K of RAM required paging, and so programs written in BASIC were a lot slower than when run on the '85. Still a great machine though. Mine's still on my desk, although since I got my PowerBook I've spent more time working in comfy chairs than at a desk...

Actually not

[aliquis]

My TI-85 isn't collecting dust in my drawer. This summer I had a small accident with flavoured oatmilk. What started as a tilted backpack ended as a TI-85 with some IC connectors magically erroded away after 5 hours or so of traveling.

Kids, remember this: Oatmilk with salt = ionized water. Batteries = electricity. Ionized water + electricity isn't healthy for those small metall pieces of yours.

Whoa, do I smell a intresting+informative moderation?

according to Top500

[BensonLeung]

According to Top 500's little blurb about the G5 cluster that has been called "Big Mac" or "SuperMac", the cluster is designated as "Virginia Tech's X"

I'm sure that no one has trademarked the letter X yet...

Linkie

Re:The TI-85

[Lehk228]

Yes, the Ti86 does have a way to check that, PxTest(), i once wrote a maze program that used pxTest() and PxChg() that could use any image as a maze as long as the two corners it used were clear (never bothered to make it autoclear the corners) I also implememted mine sweeper with everything except the timer and auto-clearing large blocks of zeroes (trying to write recursive functions on a platform where all variables are totally open is not usually worth it)

Re:Cool article, but Wired expands on 'future bash

[saddino]

I suppose it's no worse than Slashdot bashing: The sort of stuff that will make Big Mac look like that old TI-85 collecting dust in your drawer??? Regardless of the success of future supercomputers, it is complete hyperbole to compare Big Mac's eventual demotion in the SC ranks to today's utility of a TI-85. Slashdot ediotrializing at its best folks!

T-85's

[Billly Gates]

I still use mine. heheh

I left highschool in 96 and now this year rejoined college since the tech jobs now require a college degree.

Hell the el cheapo Ti's have more memory then my high end model but at the same time mine has alot of functions reserved for the TI-89.

Re:10.28 teraflops with Jaguar

[afantee]

They have been working on it, so expect some improvement shortly.

a little perspective please?

[dhananjay]

Blue Gene/L will be capable of performing 360 trillion calculations per second, or 360 teraflops

are there any rocket scientists here who can put this into perspective? Earth simulator does 35 teraflops; this new one is going to do TEN TIMES as many?

Re:a little perspective please?

[dapyx]

> Let's call it Solar System simulator.

Re:a little perspective please?

[2nd Post!]

Hmm, maybe this would help:

Welcome, to the Matrix.

Re:a little perspective please?

[nusuth]

300 teraflops is the upper bound of human brains' computational capacity based on neuron count and a very generous estimate of how much a neuron can represent and communicate, assuming no redundancy and no calculation outside neurons occur (both assumptions are known to be false, they are there for simplification.) 500 teraflops is the processing power of human brains' based on pretty accurate estimate of computational capacity of retina and extrapolating that to whole brain.

No, that's the name of the facility

[daveschroeder]

The facility's name is Terascale Computing Facility. The cluster itself is called "X".

This is similar to NCSA. The facility is called the National Center for Supercomputing Applications (NCSA). Their new #4 cluster is called "Tungsten".

Re:BigMac isn't the only new addition

[dlapine]

It's possible, but I do have to ask some questions- how much storage did the VT guys get with it? Our Tungsten machine has access to over a 100 TeraBytes of SAN that was delivered with the cluster. I've not been able to see anywhere how much storage the Big Mac came with.

Perhaps they used this, he said with a grin: Mac Raid Storage Solution

The other question I have is how much of network pipe do they have to the outside world? Tungsten has a 40Gb connection to the Internet.

The focus of these questions being- if you can't get your data onto and out of the cluster, what use are all those teraflops?

Big Mac does prove a valid point, but it's purchase smacks more of a stunt than a real attempt to provide users with a useful computing environment.

Actually, my TI-85...

[crashnbur]

...is not collecting dust. It sits with all my other calculators. They are so tightly crammed into a little shoebox, along with their instruction booklets, that there is no room for air to get in or dust to settle. It's actually kinda sad that the very machines that bore my appetite for tech toys and knowledge are reduced to boxdwelling on a cabinet shelf directly below my stereo.

On the other hand, it may even be more sad that they still sit so close to human touch, considering that no one has gotten any use out of them since my freshman year in college, some three years ago.

That TI-82 was good to me. It saw my first programming efforts come to fruition: from simple BASIC math applications (for which I became a seemingly proficient developer) to minor assembly applications (which I never got deeply into). Since getting my TI-83 in high school I haven't used the TI-83, and since getting the TI-89 my senior year I haven't used any of the others (except occasionally the TI-83 for its financial calculating prowess!)...

Oh the history of calculators. I still remember the days when I worked on the web site Calc.org, which was then called Dimension-TI (heh), and later the TI-Files... and of course we all can remember submitting to and downloading from ticalc.org, the only of all the TI calculator web sites with just the right look and feel to endure the many years of evolving technology and still remain on top.

Sorry for this post. I think I was reminiscing. Hmm...

Bad news for Itanium 2

[afantee]

The 1.5 GHz Itanium 2 costs over $3000 per chip, and even the 32-bit Xeon 3.06 GHz is about $1000, while the 2 GHz PPC 970 is about $300 or $400.  In addition, VT wants 64-bit chips, so Xeon is a nonstarter.

Excluding the Earth Simulator, the 2 GHz G5 has the highest Flops per CPU, even 5% higher than the 1.5 GHz Itanium 2 and 10 times cheaper:

#2 Alpha 13880 / 8192 = 1.69

#3 G5 10280 / 2200 = 4.67

#4 Xeon 9819 / 2500 = 3.92

#5 Itanium 8633 / 1936 = 4.45

#6 Opetron 8051 / 2816 = 2.85

Re:Kudos to the Mac (don't forget the others)

[ruiner13]

"Finally in the interest of full disclosure and to pre-empt the anti-Mac zealots, I should mention that the $4.2 million for the G5 machines is probably the education list price, because when you go to Apple Store, putting 2GB of RAM into 1100 2x2Ghz G5's will cost you $4.4 million (+ a little more for having some spare machines)."

From what I remember about the press release, they did not purchase all the RAM through the Apple store. They got the standard 2x2GHz model and bought the RAM elsewhere, at a cheaper price.

Fastest, but...

[yroJJory]

can it run Maya AND Photoshop at the same time?

The G5 is even better for FFT

[afantee]

http://www.fftw.org/speed/

Two posts up...

[payote]

...another writer claims LINPACK DOES stress latency and bandwidth, so one of you guys has it wrong. The point is, they have a fast supercomputer, TODAY, for under 5 million. And 5% efficiency? Uh, let's wait til they run some actual code before guesstimating.

Re:Two posts up...

[Boone^]

From Netlib:

What is the Linpack's "Highly Parallel Computing" benchmark? The third benchmark is called the Highly Parallel Computing Benchmark and can be found in Table 3 of the Benchmark Report. (This is the benchmark use for the Top500 report). This benchmark attempts to measure the best performance of a machine in solving a system of equations. The problem size and software can be chosen to produce the best performance. http://www.netlib.org/benchmark/hpl/

Please note the words "highly-parallel" and "best performance" and the following phrase from the link from the quote:

Nonetheless, with some restrictive assumptions on the interconnection network, the algorithm described here and its attached implementation are scalable in the sense that their parallel efficiency is maintained constant with respect to the per processor memory usage.

In a round-about way those quotes mean "if it does X FLOPs on 1 processor, it'll do 10X FLOPs on 10 processors because it's embarrassingly parallel. We're talking SETI- or RC5-type parallel. Regarding the 5% efficiency, please reference the following paper to see some numbers. In STREAMS Triad, the Cray X1 has 24% efficiency, while a P4 had 3.4% of a peak rated at 2x the X1 processor. The paper was done by the Army HPC Research Center in Minneapolis, MN: http://www.ahpcrc.org/publications/X1CaseStudies/C luster_CrayX1_Comparison_Paper.pdf

BigMac i/o

[dlapine]

Reading carefully through the presentation .pdf that covers what information they give out, I see that they have no external disks or SAN. The specs state that they have 160GB of serial ATA per machine, probably in 1 drive.

There was no mention of network bandwidth external to the cluster.

Collecting dust in the corner?

[ernst_mulder]

Please ship your dust-collecting calculator to me, I will happily clean it up and put it on my web site. As you can see (user info) I collect calculators. Darn I forgot I'm on a 12kB/s upstream connection and now I'm gonna get /.ed... My mistake.

Or was it a joke? :-)

Heh

[autopr0n]

You ought to look into the '89. It's really amazing that TI's products have been so static over the years, but also somewhat cool. Ti's most advanced calc is now the voyage 200, which is basicaly a suped up '92, from what I can tell. I remember using the 92 back in highschool.

Re:Cool article, but Wired expands on 'future bash

[greenstork]

Seems like Apple's ranking will be a springboard for supercomputer sales, especially given their relatively inexpensive supercomputer price tag.

I would imagine that the cost per Teraflop is one of the cheapest in the top 100 at least.

I'm an insensitive clod...

[MikeXpop]

... you insensitive clod!

Hmmm

[autopr0n]

Well, I've never had something I've run continuously for years and years, but in my experience you can almost always get old hardware assembled and booting. For a couple years I was running a mail server on a P200 board shoved into a 386 case and held in with just one screw. It worked.

Any kind of PC will last for a very long time, after all they have no moving parts. There are some issues with capacetors leaking, and things like that, but if they're well made you should be fine.

Re:Kudos to the Mac (don't forget the others)

[mduell]

They had 4GB RAM, not 2, which raises the total cost of just the towers to $5.3M. Add in $1.8M for networking equipment, and $2M for facilites upgrades and it comes to $9.1M.

how much the NCSA one cost?

[cyfer2000]

And #5 and #6? Is there anybody has some ideas? Thank you.

Re:pff

[rolocroz]

Why, yes, a 5 year old Mac can certainly run OS X. Right now I'm using Jaguar on a beige G3 from 1997, and it's perfectly acceptable. It's not the fastest machine ever, sure, but it gets the job done.

The real reason the Big Mac is so good...

[psyconaut]

...it's the special sauce! ;-)

-psy

Re:The real reason the Big Mac is so good...

[jo42]

...and: "That's NOT beef."

Re:BigMac isn't the only new addition

[Johnny Mnemonic]

fwiw, the information that I've read about the Big Mac cluster is that each G5 shipped with the standard-build hard drive; so that's 160G of SATA per node. 1100x160=176000G, which I believe is also 176T. So that trumps your storage--unless each node of yours also comes with it's own HD in addition to the SAN that you mentioned. And, I'll concede that a single SAN is more useful than 1100 drives in different machines in some circumstances--but each node in Big Mac will be able to cache information locally rather than having to call it over the network.

As for your other question about network pipe to/from, I haven't seen that.

Just Do It.

[BigBadBri]

As the officious lawyer for Nike, I demand that you cease and desist from using our trademarked slogan in your comments without delay.

Otherwise, we'll send you to one of our Vietnamese sweatshops as an indentured slave.

Darl McBride's pants explode with pleasure...

[payndz]

...at the thought of a tennis court's-worth of tightly-packed IBM processors all running Linux. Blow the dust off your TI-85 and work out the total per-processor fee SCO will be demanding for that lot!

BlueGene/L

[frenchs]

If the BlueGene/L interests you, take a look at the next member of the family BlueGene/P (the P means Petaflop). If I recall correctly, the Petaflop version is going to have more than a million processors in it. These computers are pretty much used for biological applications, and are going to benefit from some serious hardware, software, and networking.

Here is the project update from a while back, talks a bit about each level of the blue gene project. It also talks about the biological motivations for supercomputing.
http://www.research.ibm.com/bluegene/BG_External_P resentation_January_2002.pdf

And more generally, the blugene homepage: http://www.research.ibm.com/bluegene/

-SF

Re:pff

[pohl]

The "Rev B" iMac that I'm responding from is maybe a month away from turning 5 years old, and I think it runs OSX just fine. It's a 333Mhz G3 with 160MB of RAM, which is an abundance of computational resources compared to the 66Mz i486 (with 16MB of RAM) that I formerly ran NeXTstep on.

Re:Kudos to the Mac (don't forget the others)

[bursch-X]

Bzzzt. It's running OS X. Nice try.

Shame on /., no one got the joke!

[dasboy]

10.2.8 = current Jag version 10.3.1 = current Pan version

not in LINPACK

[green pizza]

Dig around the Top500 list and you'll see that for this benchmark (LINPACK), Myrinet and Infiniband don't do much better than plain GigE. (Which is one reason why the Cray X1 systems aren't ranked higher).

In fact, there are some nearly-identical setups in which there is no difference between GigE and Myrinet.

LINPACK is a good benchmark for generating big numbers for clusters, but it's a pretty poor supercomputing bechmark in general. The faster your machine can multiply and add fp numbers, the better its LINPACK score. This isn't SPECfp_rate. (Notice I said SPEC rate, not SPEC base).

Re:The TI-85

[toddestan]

The reason the TI-82 had it, but the TI-85 lacked it is because the 82 is actually a newer calculator. The TI-85 was the 2nd graphing calculator TI produced, with the TI-81 being the first. By the time they made the TI-82 TI had figured out a few things like PxText(), multiple line history, and a much improved statistics package.

The TI-86 was the successor to the TI-85. I was pretty disappointed with it myself. It was basically a TI-85 with more memory. Some things like putting ", and other characters on the ALPHA-shifted keys wasn't done (which was desperately needed), some buttons were moved around for no reason, some functions renamed for no reason, the memory paging caused a significant performance hit, the TI-86's built in assembly language support was pretty buggy and clumsy while the TI-85's hacked assembly support was smooth and stable. The thing that pissed me off the most was the fact that the TI-86 did not include features that the already released TI-83 did, which was mostly statistics stuff. Even the add-on TI offered didn't include all of it, but most of it.

And I still use my TI-85 pretty much daily. I also use my HP-48G a lot too. I basically understand the TI-85 inside and out, I wish I could say the same for the HP (which is a much more complicated and harder to use calculator, but that is because it is far more powerful than the TI-85).

TI? TI?

[wfolta]

I didn't think techies used non-RPN calculators. Oh well, I guess they could get on a daytime talkshow. ("I was a techie who couldn't understand RPN.")

Slow in the future

[fullofangst]

"The article also talks about some of the amazing supercomputers in the planning stages. The sort of stuff that will make Big Mac look like that old TI-85 collecting dust in your drawer." Well no, not really... they may be fast relative to the mac cluster but in absolute terms, where it matters, the mac cluster is still -fast-.

Wrong...

[Anonymous Coward]

The five year old iMac that I am typing this on is running OS X.3 (Panther) very, very well thank you. And, mind you, this is an all-in-one, blueberry iMac. Not a, at the time, top of the line PowerMac.

Your facts are quite off.

While I tend to agree with you that a rack-mounted cpu is generally easier to maintain than a typical PC, I am not so sure about the PowerMac. With the right rack mount you will get the same benefits that you would get from a dedicated rack-mount unit. Slide out the box, pull a switch and drop the side, do the work, raise the side, and slide it back in. The process is the same for both.

Now a typical Intel box set-up is rarely like that (there are exceptions). Their engineering sucks. Getting to parts and pieces is a real pain.

Re:pff

[Hawthorne01]

Five years ago, Pentium III chips blew away anything that Apple had.

From Low End Mac .com "Offering up to twice the performance of the G3 and three times the power of a Pentium III at the same clock speed, the G4 was Apple's first serious pro computer after Steve Jobs became iCEO."
I could find more, but it's late, I'm tired, I have work tomorrow, and I assume you can use Google.

Hokie Pride

[GreatTeacherMusashi]

Anybody actually from VT postin here? I think it's kinda crazy how few ppl know about the supercomputer even on campus, I mean everyone knows our football ranking but tell the average joe on campus we have the 3rd fastest supercomputer in the WORLD, and they're like, really? where? In point of fact I dun even know where they're storing this thing on campus *wants to go and drool* oh well, too bad hokie pride's only allowed to apply to sports eh?

Re:pff

[pohl]

Excuse me, but your crack habit prevented you from understanding my post. I did not compare to the i486 machine for the purposes of contrasting what Intel had available at the time. Thanks for playing.

Only part of it.

[Llywelyn]

>The G5 is a cool processor, but it isn't the reason the VT
>cluster is so fast, the Infiniband interconnect is

You are half right and completely wrong.

The Infiniband interconnect is fast, it is cool, and it is a big part of the reason that they got so much out of the G5s, however, the G5s had to have something there to give. Rpeak is the telling statistic on this point--that is purely based on the processors.

The "Big Mac" has the third highest Rpeak as well as the third highest Rmax. This says that yes, the G5 /is/ responsible for the speed of the cluster. Infiniband may be responsible for its efficiency, but it is not the only determining factor in this any more than the processor is.

Re:Kudos to the Mac (don't forget the others)

[geoffspear]

Is teraflops/s supposed to be the rate at which the processors can accelerate in speed? Or did the industry just redefine "flops" like they did with "gigabyte"?

WarGames

[natefanaro]

If the cluster were named WOPR, I'd be scared.

Read the article, IBM is working on it:

[John Harrison]

IBM makes the G5 for Apple. It also uses similiar processor in its own machines. And yes, they can cram a lot of them into a small amount of space and still deal with the heat. If you had read the article you might have noticed the following:

Meanwhile, IBM is working on a monster supercomputer that will easily rank as the world's fastest supercomputer when it comes online next year. Blue Gene/L will be capable of performing 360 trillion calculations per second, or 360 teraflops.

Commissioned by the Lawrence Livermore National Laboratory, Blue Gene/L will be based on 130,000 processors.

Not only will it be the fastest, but Blue Gene/L will also be the most compact, IBM said.

IBM has managed to cram 1,024 PowerPC 440GX processors into a slanted cabinet the size of a dishwasher. The unit -- described by IBM as a small-scale prototype of Blue Gene/L -- is already ranked 73rd in the new Top500 list.

When finished, Blue Gene/L will be about the size of half a tennis court. "That's very small considering how powerful it is," said IBM spokesman Adam Emery.

By contrast, the Earth Simulator's 5,120 processors would fill four tennis courts.