More on Virginia Tech G5 Cluster: 17.6 Tflops

Better links for Windows Media by Mwongozi · 2003-10-12 03:07 · Score: 4, Informative

You can watch just the report itself, no skipping required, by following the links on this page:

http://www.bbcworld.com/content/template_clickonli ne.asp?pageid=666&co_pageid=3

Heist by DarkHazard · 2003-10-12 03:08 · Score: 2, Funny

Surley they only need 1099 G5s.

Re:Heist by Wudbaer · 2003-10-12 03:12 · Score: 1

Yeah, I also think 1098 should be more than enough.
Re:Heist by Anonymous Coward · 2003-10-12 03:21 · Score: 1, Funny

for(i=1096; i=0; i--){
reply(i)
}
Re:Heist by LeoDV · 2003-10-12 03:29 · Score: 1

Surely they only need 950 G5's, cause I could sure use a G5 and a hundred thousand bucks.

You know, to buy 22" LCD displays, iPods and Powermacs. And a Ferrari.
Re:Heist by AKAImBatman · 2003-10-12 03:38 · Score: 1, Funny

> for(i=1096; i=0; i--){
> reply(i)
> }

Congrats. You've managed to snag yourself exactly ZERO machines. Try this instead:

for(i=1096; i >= 0; i--){
reply(i)
}

--
Javascript + Nintendo DSi = DSiCade
Re:Heist by toddestan · 2003-10-12 03:52 · Score: 1

1090
Re:Heist by lowmagnet · 2003-10-12 05:15 · Score: 1

One must assume a used Ferrari.

--
Heute die Welt, morgen das Sonnensystem!
Re:Heist by flossie · 2003-10-12 07:15 · Score: 1

Er, do you mean
for(i=1096; i > 0; i--){
reply(i)
}
?

--
flossie
Write now. Defend liberty
Re:Heist by AKAImBatman · 2003-10-12 13:33 · Score: 1

Hah! You got me! :-) Usually during programming, you'd count the 0 as an item, but for physical objects, that wouldn't work so well. :-P

--
Javascript + Nintendo DSi = DSiCade

Yes, but, by Anonymous Coward · 2003-10-12 03:11 · Score: 2, Funny

can it defeat an iMac in Apple's Photoshop benchmarks ;-)?

Re:Can the results be trusted? by sakusha · 2003-10-12 03:13 · Score: 5, Informative

They have previously discussed this, they use error correction algorithms, no ECC RAM necessary.

Re:Can the results be trusted? by Wudbaer · 2003-10-12 03:14 · Score: 1

Everybody and their dog use systems without ECC ram, yet the world has not come to an end yet. I would be more concerned of programming errors than flipping bits.

Twice as fast...? by suwain_2 · 2003-10-12 03:14 · Score: 2, Interesting

...considerably more than doubling the performance of the current number 3 1152-node dual 2.4 GHz Xeon MCR Linux cluster.

If I understand this correctly, it's saying that a G5 is more than twice as fast as a dual 2.4 GHz Xeon? (1152 dual 2.4 GHz Xeons vs 1100 dual 2.0 GHz G5s -- there are fewer G5s and they run at a slower clock speed.)

This is a pretty staggering statistic. I hadn't really believed the hype about how fast the new G5s were.

--
________________________________________________
suwain_2 :: quality slashdot p

Re:Twice as fast...? by Genady · 2003-10-12 03:21 · Score: 1

No, they're saying this cluster is twice as fast as the other one. You can't boil these numbers down to the processor level. Doing so is overly simplistic.

--

What if it is just turtles all the way down?
Re:Twice as fast...? by adam872 · 2003-10-12 03:36 · Score: 2, Informative

The two clusters are different enough that making accurate comparisons is difficult. The new G5's have a more recent PCI architecture, they use Infiniband as the interconnect and it's possible that they made use of the AltiVec (though I hear that this may not be the case because of 32 bit limitations). I believe none of these apply to the Xeon's. In high speed computing, the interconnect is vital, so that alone may push this cluster ahead for the time being. I don't doubt that the individual G5 processor are bloody quick (and as a Mac user and fan, I'm kinda glad) though.
Re:Twice as fast...? by ErnstKompressor · 2003-10-12 03:52 · Score: 1

Seeing as the Xeon was a 2.4 Ghz and that the fastest Xeon is now 3.2 Ghz, It would only be .8 Ghz faster or approx. 33% faster which would still leave it behind the 2 Ghz G5's, if they truly were twice as fast...which I will not begin to argue here as I have yet to use one myself...

Happy Sunday :) "~~~" you out there?

--
We apologise for the fault in this post. Those responsible have been sacked. -- Signed RICHARD M. NIXON
Re:Twice as fast...? by Halo- · 2003-10-12 04:10 · Score: 1

As a lot of others have pointed out, there is a lot more to "speed" that the frequency of the clockchip. I use a lot of IBM "P-series" (what used to be called RS/6000) machines at work. The clock speeds are generally low (sub 1Ghz) but in certain situtations these machines absolutely smoke. In other situtations, such are running Java, they drag ass. A lot has to do with how well the application using the processor takes advantage of the low-level capabilites.

Think of computers like cars. The Honda S2000 only has between 2 and 2.2 liters of displacement, but it will smoke most other cars which displace equal amounts because of how it is tuned, geared, built, etc...

More clock speed is in many respects a cop-out when it comes to performance. Faster chips give you faster code, but as the chips get faster, the code gets slower. Moore's law applies to hardware, not software.

Man, did I get off-topic or what? Anyway, in summary: yes, the G5 is faster clockwise, but there are a lot of other factors.
Re:Twice as fast...? by TheRaven64 · 2003-10-12 04:24 · Score: 2, Funny

I'd still take the G4 over the P4...
I'm not so sure. It's almost winter, and this house can get pretty cold...

--
I am TheRaven on Soylent News
Re:Twice as fast...? by AArmadillo · 2003-10-12 04:33 · Score: 1

I'm not sure exactly what software they used for this benchmark, but the G5's AltiVec vector unit is several orders of magnitude faster than the SSE2 vector unit on the Xeon. If they used software that is optimized for AltiVec, it doesn't surprise me in the slightest that the G5 beat out the Xeon hands down. What would surprise me is if this difference was in unoptimized code -- ie, code that doesn't use the vector units.
Re:Twice as fast...? by liquid+stereo · 2003-10-12 04:34 · Score: 1

I've done my own testing with a CFD code I wrote and the G5 is more 2x the speed of the P4 Xeon. Just my experience...
Re:Twice as fast...? by Rand310 · 2003-10-12 04:40 · Score: 1

Um... isn't this the same argument Apple used when they had the slower CPUs (that WERE slower).

how the table's have turned
Re:Twice as fast...? by MesiahTaz · 2003-10-12 04:57 · Score: 1

I doubt you have to rev your RS/6000 to 6k rpm before it becomes more than a glorified Civic.

--
Are you an open source warrior?
Re:Twice as fast...? by Jeffrey+Baker · 2003-10-12 04:58 · Score: 1

Was that an exaggeration or what? By your reckoning, the Altivec unit in the G5 is at least 1000 times faster than the SSE2 unit in the Xeon?

Uh, yeah ....
Re:Twice as fast...? by Jeremy+Erwin · 2003-10-12 05:00 · Score: 1

Please document your assertions.
Re:Twice as fast...? by TennesseeVic · 2003-10-12 05:07 · Score: 1

Your CFD code is probably of a completely different nature than the Linpack benchmark. Most CFD is closer to Blas 1 which has no cache reuse, while Linpack is completely Blas 3, which is lots of reuse if you can code it the right way. If I'm not mistaken, the G5 still has only a single precision AltiVec, while SSE2 is double precision. That means the G5 gets no help from the AltiVec (the linpack benchmark is double precision).
Re:Twice as fast...? by pastafazou · 2003-10-12 05:10 · Score: 1

I guess it is Rpeak.
Therefore, this machine is NOT twice faster than the Linux Networx cluster.
Drawing a definite conclusion from a guess about the setup? Perhaps you should research a bit more before adding your 2 cents.
Re:Twice as fast...? by mangu · 2003-10-12 05:13 · Score: 2, Informative

17.6 Tflops in 2200 processors results in 8 Gflops/processor. I don't know about the Xeon, but I have benchmarked my own 2.4 GHz Pentium4 at 6 Gflops, multiplying two 1000x1000 random matrices using Lapack. So, yes, 8 Gflops at 2.0 GHz is faster than 6 Gflops at 2.4 GHz, but only slightly. Also, there is the overhead in the matrix multiplication. The peak performance in the 2.4 GHz P4 would be 9.6 Gflops, so one can say there's no magic other than Apple marketing in the G5. The diference in performance between those two clusters probably comes from other factors than processor power.
Re:Twice as fast...? by AArmadillo · 2003-10-12 05:15 · Score: 1

Magnitude is base 2 in computers ;-).
Re:Twice as fast...? by ColMustard · 2003-10-12 05:32 · Score: 1

Now factor in cost. Then you'll find more magic.

--
Moof.
Re:Twice as fast...? by redJag · 2003-10-12 06:02 · Score: 1

The G5 is clocked at 2Ghz, the Xeon at 2.4Ghz. How is the G5 faster clockwise?
Re:Twice as fast...? by Blondie-Wan · 2003-10-12 06:15 · Score: 2, Insightful

You obviously haven't been paying attention. The Mac world stopped crowing about the G4's superiority over other systems once it became clear it no longer was (if it had been to begin with), i.e., as the MHz gap got wider and wider. Prior to the G5 intro at WWDC, Steve hadn't run one of his keynote Photoshop bake-offs in a while; even he wasn't claiming dual ~1 GHz G4 systems were faster than what the Wintel world could offer. Many people pointed out (correctly) that there was still a MHz myth and that the Macs weren't as slower than the Wintel systems as the numbers might imply, but on the whole, most Mac people were certainly conceding at the beginning of this year that the fastest PCs were faster than the fastest Macs.
Re:Twice as fast...? by Halo- · 2003-10-12 06:17 · Score: 2, Informative

Yeah, it might be the same argument, but either way it is fairly pointless. This is why benchmarking is such a controversial subject. Do you measure pure operations per second? If so, which ops? Or do you measure the actual wallclock time it takes for real world programs to execute a set of common tasks? Again, which programs and which tasks? (Which doesn't even begin to get into the all-to-real problem of vendors adding hacks to screw with the benchmark ala the video card arms race....)

I don't know jack about Macs, but it wouldn't surprise me if some marketing drone had claimed that. They might have even had a few examples to back it up.

Ultimately, "speed" isn't really a singularly quantifiable entity. I can think of at least three different ways to measure speed:

1) Pure CPU operations per second. Good for hard math, but only part of the equation

2) Hardware speed across the entire system. If the CPU screams, but the memory subsystem drags ass, the usage speed is slower. I doubt VT's cluster would be very cool if it was using 9600 baud modems for interconnects. :)

3) Perceived speed. How fast does it feel like it runs? For example the Transmeta chips which "learn" and optimize will feel slower or faster depending on if the code has been run recently, but the hardware speed hasn't changed.
Re:Twice as fast...? by Alan+Partridge · 2003-10-12 09:56 · Score: 1

"Think of computers like cars. The Honda S2000 only has between 2 and 2.2 liters of displacement, but it will smoke most other cars which displace equal amounts because of how it is tuned, geared, built, etc..."

Honda's own Civic Type R is faster than the S2000, your analogy is totally idiotic.

--
That was classic intercourse!
Re:Twice as fast...? by codifus · 2003-10-12 11:18 · Score: 1

OK, the S2000 analogy is nice, but when you get down to the real nitty gritty, the S2000 achieves its phenomenal performance by doing drop-clutch-burning launches. The cars with twice the engine displacement don't burn any clutches to achive fast results, they just have massive amounts of torque, which the S2000 lacks. So, it makes me wonder, what are the G5s compensating for? CD CD
Re:Twice as fast...? by Blondie-Wan · 2003-10-12 11:35 · Score: 1

There certainly was the meme in the Mac community during the dark days of the G4 450Mhz that a G4 was "twice as fast" as a Pentium of the same clock.

I think the official meme has always been "up to twice as fast" (insert alternative value of choice for "twice"), which keeps it fair. The two platforms perform differently for different applications, obviously, and a given PC might outperform a given Mac at one thing, while the same Mac outperforms that same PC at something else.
Moreover, lots of Mac people I've come across were certainly conceding PCs were faster on the whole long before the beginning of 2003, though. Not in 1999, mind you, but back then, it was "up to twice as fast" (or up to thrice as fast, or as fast, or whatever) as the fastest PCs - depending on what one was doing with it. ;)
Re:Twice as fast...? by pdykstra · 2003-10-12 16:05 · Score: 1

The first thing you should understand about bench marking with TFLOPS is that this isn't a good measurement of performace. Modern processors have a seperate unit for floating point operations which means two processors can have the same clock speed and process flops at different speeds. It depends on what the processor designers thought was more important. I don't know a whole lot about the G4s, but intel focuses mostly on instruction sets that improve speed with multimedia programs. Because most multimedia programs don't use a lot of floating point operations, they don't focus on improving the speed of the flops. The biggest execption to this is 3D games, but most of the flops are done on the graphics card now with specialized hardware. The reason supercomputers are measured in TFLOPS is because most of the applications running on supercomputers are scientific simuations. These generally have a lot of complex math done with flops. But this still doesn't account for a lot of the bottlenecks that might occur in the calcuations. The major bottlenecks would be the bus speed which in 1GHZ on the G5 and 400MHZ on P4s and the speed of the network. The P4s just came out with a 800MHZ bus and a 1GHZ bus, but older supercomputers might not have these. The network on VT's terascale supercomputer is using an infinaband network, which is the fastest avalible today without using a custom built network. I worked on the terascale server and Jason Lockheart said the speed was calculated, but they weren't releasing any information on the benchmarking done on it until later so it wouldn't affect their placing in the top 500 list. So I don't know where slashdot got their information, but I wouldn't trust it completly. On a side note, those G5s are looking really nice right now. Dual 2.4GHZ and 1GHZ bus speed, not to mention OSX. Almost makes me want to go out and buy one. Almost, they're still macs, but mabe in a couple years. It's apple's first attempt at selling computers based on their power and not on how good they look.
Re:Twice as fast...? by Alan+Partridge · 2003-10-13 02:32 · Score: 1

Those specs are pretty meaningless as the Civic has consistently proved itself the faster machine on the road and track.

Both nice cars, mind you.

--
That was classic intercourse!

Re:Can the results be trusted? by IM6100 · 2003-10-12 03:18 · Score: 1

So they bog down the software doing something that could be done in hardware?

--
A Good Intro to NetBS

Re:Can the results be trusted? by OS24Ever · 2003-10-12 03:20 · Score: 1

ECC memory doesn't solve world hunger, it just corrects single bit errors when a request is read from memory. If ECC memory isn't there and an error occurs, the system reboots.

Data isn't returned corrupted, it don't come back at all

if you used computers in the early 90s, it was AKA 'PARITY ERROR'

--

As a rock-in-roll Physicist once said, No matter where you go, there you are.

oh my god... by jx100 · 2003-10-12 03:25 · Score: 1

The BBC page of the devil!!

(look at the link URL)

There was a reception yesterday by mcknation · 2003-10-12 03:26 · Score: 1

For all the people involved in it's construction. I helped in the construction. Wouldn't have known about it if not for the Slashdot article. I don't doubt the Linpack score. Allthough the e-mail I got about the reception said the score wasen't supposed to come out for another week or so. Anybody know what went on in the reception?

mck

Re:There was a reception yesterday by knothead99 · 2003-10-12 07:12 · Score: 1

I helped out with the construction of the machine as well. FYI, the reception isn't until NOVEMBER 11.

How fast can it copy a 17mb file? by Anonymous Coward · 2003-10-12 03:26 · Score: 1, Funny

That is the true test.

Re:How fast can it copy a 17mb file? by Mod+Me+God · 2003-10-12 05:17 · Score: 1

Seems nero-online is offline right now so I cannot find the BSD copy a 17MB troll, but i dug up another quality troll:

when you go out to your car, i will be in the backseat. i will not let you
know this until you are miles away from home. i will hold a Glock to your
right temple, and tell you to keep your fucking hands on the wheel. i will
direct you at gunpoint to a secluded parking lot behind a failed dot-com's
warehouse.

i will take you out of the car and instruct you to tape your legs together
at the ankles. halfway through, i will tire of your sniveling and blubbering
and pistol-whip you. this will wake you up, and sharpen your mind to the
fact that you are as good as fucked. i will force you to touch your toes and
hog-tie you with cat-5.

then i will cut your clothes off. if my knife slips and cuts you, that's not
my problem. i'll spit on my hand and rub it on my cock, to get it a little
bit slick.

then you will take every millimeter i can give, all at once, and if you
scream i will just laugh and stab you in the leg. i will fuck you as though
you are the object of all my spite and assfucking is the worst punishment i
can give you. i will fuck you until you bleed, and i will bash you on the
head with a rock at the moment when my cum mixes with the oozings from your
tattered colon.

then i will drive away, and drive your car into a river. your humiliation
will be complete, public, and nagging. i will get away scot-free.

--
-- FreeNET user? Comfortable with the adverse selection?

Dude.... by iamacat · 2003-10-12 03:29 · Score: 1

Windows Media and Real Player about a G5 cluster? Don't you mean a nice Quicktime movie? (yes I have both players on Mac, sans spyware and other undesirable "system integration". But Apple will be sad)

Re:Dude.... by Ohreally_factor · 2003-10-12 07:30 · Score: 1

>But Apple will be sad)

I read this out loud to my dog in my most plaintive voice. =)

--
It's not offtopic, dumbass. It's orthogonal.
Re:Dude.... by capmilk · 2003-10-12 12:05 · Score: 1

Eat this. It's an mpg version of the clip.
Re:Dude.... by norkakn · 2003-10-12 13:28 · Score: 1

OT, but.. I just got a G5 after using bsd and *shudder* windows for a number of years, so I am still in the resedit days and I'm really curious as to how the spyware was stripped.

would you be able to tell?

Cluster problem. by Anonymous Coward · 2003-10-12 03:38 · Score: 1, Funny

I don't wan't to start a holy war here, but what is up with you G5 cluster zealots? Ive been sitting at my freelance gig of a G5 cluster (1200 Dual G5s) for twenty minutes as it attempts to copy a 17 megabyte file from one folder to another! Twenty minutes! At home my Beowulf cluster of 100 Celerons running Linux, the same operation takes just 2 minutes, if that. While this is happening, Itunes won't work, and everything else grinds to a halt. Even vim is struggling to catch up as I type this.

Yes, I do have DMA enabled, and I am using 25K SCSI disks. My old 3.2 Ghz 64 way xeon runs faster than this G5 cluster at times. G5 zealots, flame me if you like, but I'd rather hear some intelligent reasons why I should use G5s over faster, cheaper clusters.

Re:Cluster problem. by the_2nd_coming · 2003-10-12 03:54 · Score: 1

and I am one of the privileged that can say "I was there when this famous troll was first posted!!"

it is nice to see it being used so often :-)

--

I am the Alpha and the Omega-3
Re:Cluster problem. by Morky · 2003-10-12 04:15 · Score: 1

This is not a troll. This is some funny shit.
Re:Cluster problem. by MuckSavage · 2003-10-12 04:50 · Score: 1

That is funny. Damn, I wish I had some mod points.

Re:Imagine... by the_2nd_coming · 2003-10-12 03:44 · Score: 1

oh wait.

--

I am the Alpha and the Omega-3

New top-500 list will be announced around Nov 18 by slyfox · 2003-10-12 03:46 · Score: 3, Informative

The new "top 500" list will be announced right before SC2003 and discussed in detail at a session of SC2003 on November 18.

Look for another (less speculative) story on Slashdot around then.

Re:Can the results be trusted? by Ed+Avis · 2003-10-12 03:46 · Score: 1

If ECC memory isn't there and an error occurs, the system reboots.

How is the system to know whether an error has occurred? If the value at some byte address should be 202 but it comes out 201, how to determine that the result is wrong? For that you need parity memory, which is able to detect (though not recover from) single bit errors. As you say, the detection is called a parity error - but this requires memory with parity information (usually one extra bit for every eight).

I believe that common ECC memory is able to recover from single bit errors and detect two bit errors.

Or does the Mac G5 use parity memory? It seems hard to believe, hardly anyone manufactures parity memory these days, it's either cheapo no-checking sticks or full ECC.

--
-- Ed Avis ed@membled.com

Are they all running Panther? by failedlogic · 2003-10-12 03:49 · Score: 1, Flamebait

What's been left behind is wether or not these systems are using Panther as the OS. It would seem that with this kind of performance, an Apple supplied OS -- as opposed to Yellow Dog would -- only be capable of performing well on the G5 since Panther has processor optimizations for the G5.

If the original XServers were too costly and low performance (since they came with a G4) wouldn't a G5 server (since the performance is apparently much better) be a great option for small/medium size businesses for a web/mail/database server?

Re:Are they all running Panther? by ivan256 · 2003-10-12 04:10 · Score: 1

-- as opposed to Yellow Dog

This bothers me as much as people synonymizing Red Hat with linux for x86, and Internet Explorer with the internet; and yes, you sound just as dumb when you jump into a discussion on PowerPC based supercomputers running linux and equate the entire PowerPC world with Yellow Dog as somebody who jumps into a conversation about the internet and equates it with AOL version 7.0.

Just FYI, Red Hat is the only major linux vendor out there that doesn't support PowerPC. Very few people use Yellow Dog when they run Linux on their Mac. Most, in my experience, tend to run Debian or Suse. The proper way to describe Linux that is run on a PPC processor is "Linux", "LinuxPPC", or "PowerPC Linux". Now, armed with this knowledge you can at least pretend to know what you're talking about in this discussion without being ignored.
Re:Are they all running Panther? by Maserati · 2003-10-12 04:16 · Score: 1

No, the G5s shipped with 10.2.7, it's recompiled for the G5 to take advantage of the new architecture - should be reasonably optimized although that probably isn't a priority yet - and has some 64-bit support. Yellow Dog Linux is ready for the G5, but their support is only as good as gcc at this point.

--
Veteran, Bermuda Triangle Expeditionary Force, 1992-1951
Re:Are they all running Panther? by sylencer · 2003-10-12 04:29 · Score: 1

First of all, it's XServe, and many of those should be called XServes.

About G5 in an XServe - we will have to wait for that to happen, because you can't cool one or two G5s in a 1U case, and a G5 PowerBook won't hit the shelves too soon because of the power consumption...
Re:Are they all running Panther? by StarmanDeluxe · 2003-10-12 05:00 · Score: 1

I wouldn't find it all that unlikely that they shipped with a specially-made version of OS X. Perhaps not a Panther pre-release version, but I don't think Apple would just slap 10.2.7 on all these computers (whose performance needs to be very high, as I'm sure Apple would love the advertising).
Re:Are they all running Panther? by mithras+the+prophet · 2003-10-12 05:30 · Score: 1

If they shipped with 10.2.7, can they buy the Panther family pack? Or 220 family packs?

--
four nine eighteen twenty-7 thirty-nine forty-7 fiftyeight sixty-nine seventy-9 eighty-8 one-hundred-and-nine one-twenty
Re:Are they all running Panther? by Ianoo · 2003-10-12 05:49 · Score: 1

But we don't say "x86 Linux", do we. The operating system is called Linux (or more correctly GNU/Linux) whatever processor it runs on.
Re:Are they all running Panther? by Paradise+Pete · 2003-10-12 05:56 · Score: 1

But we don't say "x86 Linux", do we. The operating system is called Linux (or more correctly GNU/Linux) whatever processor it runs on.
Just plain "Linux" was the first suggestion in his list.
Re:Are they all running Panther? by rsmith-mac · 2003-10-12 06:19 · Score: 1

No, the family pack is only for home users; commercial and education users need to look elsewhere. Likely, Apple gave them a special contract(especially if they built a customized version of the OS) which cost a fair bit more.
Re:Are they all running Panther? by Midnight+Thunder · 2003-10-12 06:31 · Score: 1

They had a 1st of October deadline to meet for the cluster. If they had waited for Apple to bring out their G5 XServe, then they would have missed the deadline. The G5 towers were available and had been tested to a fuller extent, and they probably worked out to be cheaper. If you look at the photos, on the site, then you will notice that space was not an issue, so enclosure size was probably not important.

I would not be surprised if the first generation of G5 XServes are 2U cases, so that there is enough space for the heat sink and proper cooling.

--
Jumpstart the tartan drive.
Re:Are they all running Panther? by arivanov · 2003-10-12 08:02 · Score: 1

which cost a fair bit more

Or less...

The power of this err... is nothing compared to the power of the force (marketing).

Just think of all the free publicity they have got so far.

--
Baker's Law: Misery no longer loves company. Nowadays it insists on it
http://www.sigsegv.cx/
Re:Are they all running Panther? by jespring · 2003-10-12 08:15 · Score: 1

They are not running Panther, but they will be, hopefully soon.

The reason they are using G5 towers instead of Xserves is not price or perfomance, but reusability. VA Tech has a huge lab of macs (mostly FP iMacs and older G4 towers). As the cluster machines age, they can be swapped out with newer hardware, and the G5s can then make excellent lab workstations for students.

Very smart thinking by VA Tech, IMHO.
Re:Are they all running Panther? by 11223 · 2003-10-12 08:15 · Score: 1

No, it's not. Yellow Dog is /not/ shipping a G5 Linux, and as far as I know there's no distro out there that does. (The earlier report of Gentoo booting was a hoax.)
Re:Are they all running Panther? by Lars+T. · 2003-10-12 09:14 · Score: 1

In a recent interview (in German) (actually in a future one, since it's dated 13th of October 2003 ;-), Suse Boss Richard Seibt says "In the future, Computers from Apple will run Linux. I can envision that in 10 years their comfortable user interface will be standard on all Linux systems."
No idea why I brought that up now ;-)

--
Lars T.
To the guy who modded me down from perfect to terrible Karma - Apple haters still suck
Re:Are they all running Panther? by jtrascap · 2003-10-12 09:32 · Score: 1

According to Apple, **all** G5 systems will get Panther free (well, $19.99 for a shipped CD) - they've been grandfathered into the program:

When you fill out the form to check your G5 serial number, be sure to leave out any dashes.
Re:Are they all running Panther? by n8_f · 2003-10-12 11:15 · Score: 1

That is interesting, but it is the first time I've heard it. Could you provide a source? Thanks.
Re:Are they all running Panther? by n8_f · 2003-10-12 11:23 · Score: 1

Why don't you think Mac OS X is very high performance? Don't you think that if Apple could gain some performance benefits in OS X, they would roll it in to the OS so all their customers could benefit rather than trying to make a bunch of special builds?

The GUI doesn't suck up any additional resources, except perhaps a small bit of RAM that is wired. Everything else will get swapped to disk and the GUI only sucks up CPU if you use it. So why is OS X a poor choice?
Re:Are they all running Panther? by StarmanDeluxe · 2003-10-12 11:27 · Score: 1

Oh, I don't think OS X is a poor choice, at all. I do, however, think it likely that Apple streamlined the OS to remove eye candy and other features that wouldn't be particularly pertinent to the cluster.
Re:Are they all running Panther? by rsmith-mac · 2003-10-12 13:16 · Score: 1

True, but still, the education/government prices are nowhere near the family pack prices. The family pack is $40/machine; no education contract I've seen gets below 2x that amount, although the contract does stipulate 3 years of free upgrades.
Re:Are they all running Panther? by n8_f · 2003-10-12 13:45 · Score: 1

Well, that is reasonable, but they are using a stock install according to Dr. Aref, as I note here. And isn't that much better advertising than using a custom-built OS X or Linux?

BTW, if someone was interested in building a custom OS X, they can just use Darwin, which omits all of the Apple niceties and for which, of course, source is available.
Re:Are they all running Panther? by jespring · 2003-10-12 14:32 · Score: 1

Well, I'm not affiliated with either VA Tech or with Apple, but I heard it from someone who has reason to know. Take that for what it's worth... sorry to be cryptic.

I would imagine that this will be discussed when Varadarajan speaks on the 28th.
Re:Are they all running Panther? by n8_f · 2003-10-13 12:54 · Score: 1

Well, that would be smart planning. But I wonder if it is making lemonade from lemons? After all, there was no rackmount option with the G5, so they were pretty much stuck with the desktop form factor. I wonder if given the choice, the reusability of the desktops would outweigh the extra space required (less density also means less stringent cooling). It would be an interesting study, especially since one of the design goals for this project appears to be low cost.

Thanks for the info!
Re:Are they all running Panther? by marcinjeske · 2003-10-13 13:00 · Score: 1

Uh... no.

Standard educational pricing is $69 for the OS.

Looking at my university's internal pricelist, Mac OS X is $53 for a single-user license.

While not $40/machine... it is better than 2x that amount.

(and I must point out, that's really $200/5 machines, and further, that assumes you have a household with 5 OS X compatible computers used by people residing in the household. My household, although very Mac centric, currently only has 4 Macs running OS X, three of which came with their own license, so that family pack won't do much until Panther, even then, only 2 or three of those computers will get upgraded, so that $100-$66/machine.

Marcin
Re:Are they all running Panther? by jcr · 2003-10-13 17:36 · Score: 1

What's been left behind is wether or not these systems are using Panther as the OS.

They're using Jaguar (10.2.7) at the moment. They'll be using Panther shortly.

-jcr

--
The only title of honor that a tyrant can grant is "Enemy of the State."

Doubling the second fastest? by }{@wkmooN · 2003-10-12 03:49 · Score: 1

considerably more than doubling the performance of the current number 3 1152-node dual 2.4 GHz Xeon MCR Linux cluster.

Humm, isn't the second fastest ASCI Q at Los Alamos has now been measured with 13.88 TF/s?
As mentioned here: http://www.top500.org/lists/2003/06/top5.php

Re:Doubling the second fastest? by frostjoe · 2003-10-12 04:18 · Score: 1

considerably more than doubling the performance of the current number 3 1152-node dual 2.4 GHz Xeon MCR Linux cluster.

It reads number 3, not number two. 17.6TFlops is about twice 7.6TFlops from your link.

Ergo, the G5 is more than twice as sweet as the P4
Re:Doubling the second fastest? by }{@wkmooN · 2003-10-12 06:56 · Score: 1

and that's why kids you shouldn't take drugs like me

High speed interconnect? by draziw · 2003-10-12 03:53 · Score: 1

MCR uses Quadrics - What does this cluster use?
http://doc.quadrics.com/Quadrics/QuadricsHom e.nsf/ DisplayPages/7C18E51DBC215D3E80256D5900535959

Re:It runs MacOS X !!! by paulthomas · 2003-10-12 03:56 · Score: 3, Informative

Ugh.

It has been said thousands of times by now I'm sure.

Running Mac OS X does not mean running FreeBSD Mac OS X is a system of frameworks running on top of a Mach Kernel. The only thing that relates Mac OS X to FreeBSD is the userland. In addition to the userland you have: Cocoa, Carbon, Aqua, Java, etc. The FreeBSD portion is minimal.

And yes, if you want you can run this lower level unix without the rest of Mac OS X. It is called Darwin. It runs on Intel and PPC if you're wondering. No, this doesn't mean that Mac OS X runs on both or ever will.

Here is a short description of the BSD families.

5.2 million by cacheMan · 2003-10-12 03:56 · Score: 1

Could someone who knows the going rate of these top 5 supercomputers please tell me how much less expensive 5.2 million is? I know that it sounds like a lot of money to me, but I'm guessing that it is orders of magnitude cheaper than the other top computers.

Re:5.2 million by cacheMan · 2003-10-12 04:02 · Score: 1

From this article: zdnet.co.uk
"Sun Microsystems, for instance, is designing a supercomputer under a $50m (31.6m) grant from the US Defense Advanced Research Projects Agency (Darpa) that will contain 100,000 processors, according to Jim Mitchell, who heads up Sun's labs. The whole thing could conceivably fit into a room."
5.2 million is a lot less then 50 million.
Re:5.2 million by Kevon · 2003-10-12 04:17 · Score: 1

yes but 100,000 is a lot more than 2200
Re:5.2 million by PingPongBoy · 2003-10-12 04:23 · Score: 1

The whole thing could conceivably fit into a room

It strikes me that the macs are still in their own cases. How fast can information travel between machines at the far ends of the racks? It all comes down to the size of the room that the computer is built in...

Since they went to so much effort to assemble the entire system, they should put the motherboards closer together. Is that harder to do with Macs as opposed to PCs?

How accurate are the benchmarks for supercomputers? If I want to compare this Mac cluster to a futuristic CPU running 17 THz who will appear faster?

--
Know your pads. One time pad: good for cryptography. Two timing pad: where to take your mistress.
Re:5.2 million by Gilmoure · 2003-10-12 06:59 · Score: 1

They kept the Apple cases as they're designed for the cooling the G5s need as well as keeping the cost down. Adding in 4U boxes or whatever it'd take to fit all the fans in the G5 cases would have been really expensive as well as time consuming.

--
I drank what? -- Socrates
Re:5.2 million by bdow · 2003-10-12 16:45 · Score: 1

Electricity does NOT travel at the speed of light. I don't recall the exact numbers, but we're talking about cramming electrons into a tube until it pushes some out the other end, not photons traveling through vacuum.
Re:5.2 million by geekboy_x · 2003-10-12 21:11 · Score: 1

They keep 'em in the cases for cooling purposes. This way the macs cool themselves and you only have to cool the room, as opposed to cooling every board that you pull out of the case.

--
-- There are two kinds of motorcycles. 1: German. 2: Crap.
Re:5.2 million by jcr · 2003-10-13 17:48 · Score: 1

Another factor to consider here is that in a couple of years, when VPI wants to update the cluster, they can part out the dual G5's and put them on people's desks. Two years hence, it's still going to be a pretty spiffy machine.

In fact, they could probably e-bay them and recover half or more of their cost.

Compare that to most other cluster nodes, which are typically built without display cards, sound cards, etc.

A two-year-old intel box that needs a couple hundred dollars worth of parts and six hours of labor to make it useable on your desk is probably headed for a landfill or metal recycler, rather than five more years of use.

-jcr

--
The only title of honor that a tyrant can grant is "Enemy of the State."
Re:5.2 million by Hagen · 2003-10-14 14:55 · Score: 1

An article from last year says the price for the current top of the heap, the NEC Earth Simlator, cost roughly $350 million to build, and IBM's ASCI White, another contendor, was around $110 million.
Re:5.2 million by bdow · 2003-10-16 00:31 · Score: 1

Yes, the electricity moves much faster than the electrons (which is what I was getting at with my analogy of pushing them through a tube). But still nothing approaching the speed of light. Optical IS faster, not that it usually makes a measurable difference.

Infiniband by Anonymous Coward · 2003-10-12 03:58 · Score: 1, Informative

10 Gbps Infiniband from Mellanox

Each machine has a PCI-X Infiniband card, interconnected with several 96-port switches.

Pains a Mac user by mariox19 · 2003-10-12 03:59 · Score: 1

The video is available for either Windows Media Player or RealOne Player -- but not QuickTime!

Ouch!!!

--

quiquid id est, timeo puellas et oscula dantes.

Re:Pains a Mac user by zpok · 2003-10-12 05:35 · Score: 1

So?

Download RealOne.

--
I think, therefore I am...I think.
Re:Pains a Mac user by mariox19 · 2003-10-12 05:51 · Score: 1

God forbids Mac users ever install the MacOS versions of Real One Player or Windows Media Player...

I have both on my iMac; and that's not my point.

I just thought that fellow Mac users would feel the same way. I like to see QuickTime being used, that's all. And for a story specifically about Apple hardware, I just sense a kind of irony.

--
quiquid id est, timeo puellas et oscula dantes.
Re:Pains a Mac user by capmilk · 2003-10-12 12:08 · Score: 1

Here is an mpg version of the clip.
Re:Pains a Mac user by MoneyT · 2003-10-13 05:28 · Score: 1

Some of us prefer to avoid using Real and WMP is a fairly crappy product (think PC version of quicktime)

--
T Money
World Domination with a plastic spoon since 1984
Re:Pains a Mac user by repetty · 2003-10-15 07:47 · Score: 1

"Yes, God forbids Mac users ever install the MacOS versions of Real One Player or Windows Media Player..."

Why are you jumping to conclusions? I tried Microsoft's product and, frankly, it sucked. Badly.

Re:How much power ?!?! by Rand310 · 2003-10-12 04:06 · Score: 1

Air Conditioners, switches, etc....

more than just CPUs

Re:Can the results be trusted? by Usquebaugh · 2003-10-12 04:11 · Score: 4, Insightful

Yep, and they're going to be top 5. Between you and them I wonder who has the best knowledge of how to build a cost efficient cluster?

Re:I don't give a flying f*** about you by cshotton · 2003-10-12 04:11 · Score: 1, Informative

A "real" troll realizes that the term derives from the verb "to troll" and not the noun for a hairy creature that lives under a bridge and eats farm animals. The operative definition of "to troll" as it applies to Internet messages is #3.

Main Entry: troll
Pronunciation: 'trOl
Function: verb
Etymology: Middle English
Date: 15th century
transitive senses
1 : to cause to move round and round : ROLL
2 a : to sing the parts of (as a round or catch) in succession
b : to sing loudly c : to celebrate in song
3 a : to fish for by trolling b : to fish by trolling in (troll lakes) c : to pull through the water in trolling (troll a lure)
intransitive senses

If you want to use the term to refer to a message that invites one to respond or otherwise lures you into a discussion, you want this definition of the noun, from Webster's:

Main Entry: troll
Function: noun
Date: 1869
: a lure or a line with its lure and hook used in trolling

For those of you who were in diapers when the Internet was created, a "troll" is a message designed to lure you into responding, to rise to the bait, so to speak. Please learn this bit of Internet lore before we have to start the canings again.

--

Shut up and eat your vegetables!!!

But it doesnt add up...? by skinfitz · 2003-10-12 04:13 · Score: 1

I didn't see anywhere the source of these benchmarks - I believe what they have done is multiply the SPECfp_rate_base2000 for a dual processor 2GHz G5 (which according to Apple is 15.7) and multiply it by 1100.

The thing is, that only comes to 17.27TFLOPS, and in addition it does assume that the original spec scores were accurate.

Would anyone care to shed some light onto this?

Re:But it doesnt add up...? by Aardpig · 2003-10-12 04:37 · Score: 2, Informative

Would anyone care to shed some light onto this?

I can shed light to this extent: a linear scaling between processors and processing power is only realized in the most idealized of situations (those known as 'embarrasingly parallel'), where each job is small and completely independent of other jobs. The funny thing about embarrasingly parallel tasks is that they do not need a fancy parallel computer; they can just as easily be accomplished on N separate 486 machines, if N is sufficently large.

The upshot? If they claim a purely-linear scaling, they are either lying, or only considering those jobs for which one can get by on a (large) Beowulf cluster of shit machines. My head is not turned by this news...

--
Tubal-Cain smokes the white owl.

Project leader speaking at conference Oct 28 by daveschroeder · 2003-10-12 04:20 · Score: 4, Informative

The project leader, Dr. Srinidhi Varadarajan, will be speaking at a session entitled Building Virginia Tech's G5 Supercluster on Oct 28 at the upcoming O'Reilly Mac OS X conference.

He'll probably reveal some of the technical details, such as the version of Mac OS X used, at that session.

Also, according to a blog at O'Reilly:

Next year, all the little known details [about the cluster] will be revealed in a new book. By that time we'll know what the project means for supercomputing and for Apple.

Re:Project leader speaking at conference Oct 28 by _fuzz_ · 2003-10-12 10:06 · Score: 1

... some of the technical details, such as the version of Mac OS X used...

According to this press release (at the very bottom), they're using Linux, not OS X.

--
47% of all statistics are made up on the spot.
Re:Project leader speaking at conference Oct 28 by n8_f · 2003-10-12 11:12 · Score: 1

No, Dr. Aref, the Dean of VT's College of Engineers, has said they are using the stock install of OS X, which at the time was 10.2.7. See my post here for references.

The quote is from the Mellanox CEO and he is probably using Linux to appeal to a wider audience than Mac OS X would.

The most important benchmark by pwiebe · 2003-10-12 04:22 · Score: 1

How many seti packets can it do in an hour?

Re:The most important benchmark by mjmalone · 2003-10-12 05:22 · Score: 1

Funny you should ask, my friends and I went to the planning meeting and when they opened the mics up my friend asked if they would run Seti on it and see how fast it was and they just laughed :(

--
Visualize the world of wine
Re:The most important benchmark by Thomas+Wendell · 2003-10-12 09:43 · Score: 1

I just got my dual G5 Thursday. I run two instances of the command line version via cron tasks. So far, it has cranked out 39 units in a bit less than three days, averaging 2.93 CPU hours per unit. My 2.5 GHz P4 Dell does one unit every 2.78 CPU hours running the Windows command line client.

The dual G5 puts out nearly 16 units a day, which is about the same as my other machines combined (dual 500 MHz G4, dual 800 MHz G4, 500 MHz PowerBook G4, 2.5 GHz P4, 266 MHz G3 iMac).

If the VT cluster was working on nothing but SETI, it should be able to do about 750 work units per hour, requiring something like 250 MB of transfer per hour. So, that and a decent DSL connection and you're set!

Re:17.6 Tflops sounds more like an Altivec number by Rosyna · 2003-10-12 04:25 · Score: 1

And which part of the score was released by apple?

Re:Can the results be trusted? by Waffle+Iron · 2003-10-12 04:38 · Score: 4, Interesting

So they bog down the software doing something that could be done in hardware?

Just because it's in hardware doesn't mean it's free. The ECC logic is going to add a small delay to each of trillions of memory accesses. Plain memory can most likely be tuned to run faster than ECC memory.

If you're running a constrained problem and can verify the results at the end, a single error check in software could consume far less overall time than the continuous ECC hardware checks. The software check would probably catch other types of errors as well (including many errors caused by software bugs).

I wonder... by Alpha_Traveller · 2003-10-12 04:39 · Score: 1

what the frames per second would be on Q3A...Has anyone tried it yet?

--
"Love is like pi - natural, irrational, and very important." (Lisa Hoffman)

How does it not add up? by daveschroeder · 2003-10-12 04:40 · Score: 3, Interesting

Easy: you yourself point out that 1100 * 15.7 = 17.27 ... not 17.6.

Since the call for papers for the new Top 500 list was Oct 1, and the BBC show aired on Oct 9 with a companion BBC News story dated Oct 12, you'd hope that VT was simply regurgitating the figure that has already been sent to the Top 500 organization.

And why are you trolling around with one of those super-old benchmarking stories? We've already established that every manufacturer does what they can to show their products in the best possible light. At least Apple documented their test results and methods in full.

So acually, your logic doesn't make any sense: you jump to the conclusion that it's not real results - even though real results already exist and have been submitted, and the entire story is pretty much about that process, making performance figures a critical piece to get accurate - and that they must have just multiplied some benchmark number by 1100. Then, even though the subject of your own post indicates your recognition that "it doesn't add up", you still apparently assume that the results are somehow doctored, this time for the worse, and you manage to weave in one of the stories that tries to make it look like Apple lied with its benchmarks - which it didn't - which is unrelated to the current issue! How does it "assume" the original scores were accurate?? YOU are assuming that they're just multiplying. You might have been onto something if the multiplication actually came out, but it doesn't, meaning that is NOT what they did.

Bravo, +1 Troll.

Re:How does it not add up? by skinfitz · 2003-10-13 07:07 · Score: 1

Easy: you yourself point out that 1100 * 15.7 = 17.27 ... not 17.6.

Since the call for papers for the new Top 500 list was Oct 1, and the BBC show aired on Oct 9 with a companion BBC News story [bbc.co.uk] dated Oct 12, you'd hope that VT was simply regurgitating the figure that has already been sent to the Top 500 organization.

Um.. that's not an explanation.

Bravo, +1 "I have nothing useful to say but I must say something".

Re:17.6 Tflops sounds more like an Altivec number by Jeremy+Erwin · 2003-10-12 04:41 · Score: 1

And why would it be misleading? If Xeons would be used, surely the researchers would optimize for SSE2, as x87 on the Pentium 4 has been crippled.
Scientific Computing has always emphasized numerical linear algebra. An entire strain of supercomputer processors was developed to support such requirements.

Besides, the final "score" will be produced by benchmarking with LINPACK. It's not merely a matter of taking manufacturer supplied numbers, multiplying them together, and claiming a spot.

For those on APPLE computers, the REAL file by sbwoodside · 2003-10-12 04:41 · Score: 1

http://www.bbcworld.com/content/template_clickonli ne.asp?pageid=666&co_pageid=20

Skip ahead to 13:00

Since the freakin' Windows Media files won't play on OS X.

simon

--
home page

Re:For those on APPLE computers, the REAL file by damiam · 2003-10-12 05:38 · Score: 1

Since the freakin' Windows Media files won't play on OS X.
Why not?

--
It's hard to be religious when certain people are never incinerated by bolts of lightning.
Re:For those on APPLE computers, the REAL file by Quobobo · 2003-10-12 06:18 · Score: 1

WMV9 video won't play on OSX, not in the official player, not in MPlayer, not in VLC. I haven't checked out the video, but I'm assuming it's WMV9.
Re:For those on APPLE computers, the REAL file by sbwoodside · 2003-10-12 06:28 · Score: 1

Since the freakin' Windows Media files won't play on OS X.
Why not?

Because the most popular audio codecs used for Windows Media files are not supported in the OS X Windows Media client. Nor are they supported by MPlayer, or VideoLan Client (VLC) for that matter.

Perhaps when Windows Media Player 9 comes out for OS X (supposed to be "soon") this will be fixed.

simon

--
home page
Re:For those on APPLE computers, the REAL file by bo-eric · 2003-10-12 06:30 · Score: 1

Well, for one because the current mplayerosx can't decode the sound. I tried.

--

-- Free speech is only free if your time is worth nothing.
Re:For those on APPLE computers, the REAL file by bo-eric · 2003-10-12 08:21 · Score: 1

No, only that the latest release of mplayerosx can't play it. We still have the options of compiling a newer mplayer ourselves, using WMP, RealOne or viewing the Quicktime version posted here somewhere. You know, mplayer under OS X isn't much different from mplayer for linux, apart from the Win32 codec support.

--

-- Free speech is only free if your time is worth nothing.

Power consumption by Mr.+Spectre · 2003-10-12 04:42 · Score: 1

The power supply was another huge challenge. The Supercomputer uses the same amount of electricity as 3000 average sized homes.

How do 1100 dual G5s end up using the same amount of eledtricity of 3000 average sized homes? Even with the cooling and networking, this number seems way too big.

Re:Power consumption by NSash · 2003-10-12 04:51 · Score: 1

That'd mean that a single G5 would account for more than 1/3 of the electricity consumption of an "average sized home." To be fair, at home you generally don't count climate control as part of the cost of owning a computer. If you just bought an air conditioner to keep your computer room cool, I'm sure that the AC + the desktop would suck up quite a bit of juice.
Re:Power consumption by rigau · 2003-10-12 05:00 · Score: 1

actually i think 1100 G5s consuming the electricity of 3000 homes means that each G5 would be consuming the energy of about 3 houses. that is simply bizzare.
Re:Power consumption by Mr+Bubble · 2003-10-13 15:19 · Score: 1

That struck me as odd as well. However, I think they are talking about the total electricity usage, including the cooling system - they must be.

--
"The world is a construct of forceful imagination. Those who don't know walk around in the reailties of those who do"

Re:Can the results be trusted? by IM6100 · 2003-10-12 04:45 · Score: 1

Clearly they do. As I have no interest in having 'one of the five fastest clusters' for however many weeks they remain in that spot. I mean, gee whiz.

However, it's still an interesting topic. Making memory error correction a processor intensive task seems like a kludge. Kind of the 'Win Modem' of the whole Apple design.

Sorry if raising it as an interesting point to ponder gets you all in a defensive fluster.

--
A Good Intro to NetBS

What will they do with it afterwards? by chiph · 2003-10-12 04:45 · Score: 1

As I've asked before, what will they *do* with it once they've posted their performance benchmarks?

Japan's Earth Simulator will be used for climate studies. What will VT use it for?

Chip H.

Re:What will they do with it afterwards? by Jeremy+Erwin · 2003-10-12 05:11 · Score: 1

It's a general purpose supercomputing facility. From the press release:

irginia Tech researchers are already active in a number of areas that will benefit from the new supercomputing facilities, says Kevin Shinpaugh, director of research and cluster computing for the university. These include: nanoscale electronics, quantum chemistry, computational chemistry, aerodynamics through multidisciplinary design optimization, molecular statics, computational acoustics, and the molecular modeling of proteins.

The single-purpose nature of the Earth Simulator is somewhat rare, among supercomputing facilities

Re:Can the results be trusted? by treat · 2003-10-12 04:48 · Score: 1

Everybody and their dog use systems without ECC ram, yet the world has not come to an end yet.

And they have random data corruption ALL THE TIME.

It's amazing that no one seems to care about random data corruption. Random, unexplainable crashes.

Guaranteed to happen. Your non-ECC memory flips bits all the time.

Solaris doesn't even log ECC corrections until it it happens repeatedly to the same bit. That's because bits get flipped all the time. Physics and all that.

Re:How much power ?!?! by m_niessner · 2003-10-12 04:54 · Score: 1

They meant 3,000 prison-cell sized dorm rooms.

What? by m_niessner · 2003-10-12 04:59 · Score: 1

1 Hollywood = 100 Oscars 1 Oscar = 100 flops therefore 100 Oscars = 10,000 flops Therefore 1 Hollywood = 10,000 flops

Re:How much power ?!?! by Midnight+Thunder · 2003-10-12 05:06 · Score: 1

What I would like to see is the excess heat being used to heat the rest of the building, instead of simply being junked outside. By using this heat in the winter they would reduce the amount of heat the need to generate using their conventional heating system.

That would be smart use of technology.

--
Jumpstart the tartan drive.

Re:It runs MacOS X !!! by JamesP · 2003-10-12 05:09 · Score: 1

Running Mac OS X does not mean running FreeBSD

Yes, but no one is going to use Acqua, Java, Cocoa and stuff...

Its gonna work mostly w the kernel (theres no need for graphics and stuff), hence it's going to matter the most.

--
how long until /. fixes commenting on Chrome?

BJs for Geeks by bstadil · 2003-10-12 05:11 · Score: 4, Interesting

My favorite comment:

In fact the heat is so intense that ordinary air conditioning units would have resulted in 60 mph winds

--
Help fight continental drift.

Interesting math by lexcyber · 2003-10-12 05:11 · Score: 2, Interesting

The VT cluster cost about $5.2 Million and get approx. 17TFlops - The NEC Earth Simulator gets 35TFlops and cost one billion dollars. That makes it 192 times more expencive. So you can build 192 VT Clusters. And then in theory get. 3.2PFlops for the same amount of money. If you detract performance for cable lenght etc. - You will most definitly get around 1PFlops.

So, you supercomputerusers out there - build a 1PFLOPS cluster NOW!

--
- To understand recursion, we must first understand recursion -

Re:Interesting math by damiam · 2003-10-12 05:40 · Score: 1

I doubt Apple has the production capacity for that many G5s.

--
It's hard to be religious when certain people are never incinerated by bolts of lightning.
Re:Interesting math by imsabbel · 2003-10-12 05:54 · Score: 1

Yeah, but you forget between these 2 computers lie nearly 2 years of moores law, and if i remember correctly ES was more like 360 mio$.

And 5m Bucks for the cluster? 2000 machines a 2.500$.
SOmewhere something doesnt fit there. Sure, they have gigabit ethernet on board, but the switching architectuer alone should cost at least a few millions.
Not to mention the building.

--
HI O WISE PRINCE. WHT TOOK U SO DAM LONG?
Re:Interesting math by TheRaven64 · 2003-10-12 08:14 · Score: 1

And 5m Bucks for the cluster? 2000 machines a 2.500$.
First of all, the cluster only has 1100 nodes, not 2000. Mind you, the price for the default config is $2,999 which gives us $3,298,900.
On the other hand, the education price is 2,699, giving us $2,968,900.
Mind you, they probably don't need mice, keyboards, DVD-R drives, or even graphics cards on most of them. In fact they only really need motherboard, CPU, and RAM (although a lot more than the default config), case and a small hard disk. Factoring in the propaganda value and the fact that they are buying in bulk, I think that Apple would do them a good deal...

--
I am TheRaven on Soylent News
Re:Interesting math by iomud · 2003-10-12 08:58 · Score: 1

All the machines have 2gb of ram, which is much more expensive than the default config if apple is selling you the ram. I'm sure they got a fat educational discount on top of some other deals for the sheer visibility of such a powerful cluster.

numbers ok....reading is wrong... by djupedal · 2003-10-12 05:15 · Score: 4, Informative

The 'project' uses the same amount of electricity as 3,000 average sized homes. There are many more devices deployed than just the 1100 G5s. The cooling system alone is a major power eater. Read the articles :)

Re:Quicktime by FosterKanig · 2003-10-12 05:15 · Score: 1

Excellent use of the blur filter to pretend that some of the footage was shot poorly. Wait a minute...it WAS shot poorly.
Thanks for the link. I have no immediate plans to install WMP and Real.

Bleaksburg by PhilipMatarese · 2003-10-12 05:17 · Score: 1

Supercomputer is #2 & Football team is #3
-Go Hokies!

Re:Bleaksburg by k1ng3r · 2003-10-12 15:00 · Score: 1

You of course realize that the 650mb upload limit is from on campus to off campus. There are no restrictions on transfers that remain on campus.

As for running servers, i ran one last year, with no problems. it's realy more of a don't ask don't tell policy then a no you can't or else policy

Re:How much power ?!?! by Daniel_Staal · 2003-10-12 05:17 · Score: 1

Only semi-smart.

Places have done this, but to do it you need to design (or re-design) the heating system to do it. And then, in five years, when you upgrade the computers to newer, smaller, cooler (overall) models the entire system fails.

It's a good idea, but computer speed/heat needs to stabilize first. Which isn't going to happen anytime soon.

--
'Sensible' is a curse word.

VT Power Plant by morcheeba · 2003-10-12 05:22 · Score: 1

Yep, the numbers must be wrong.

And if not, they have a 6 megawatt coal-fired power plant within spitting distance of the engineering buildings. That's 5 kilowatts per computer; plenty for either a house or a computer.

Incidently, after years of putting up with coal dust, I thought we would finally see the benefits of living next to a huge coal pile. We had a severe ice storm and most of blacksburg lacked power. but so did campus!.

--
HIV Crosses Species Barrier... into Muppets

Re:VT Power Plant by AsmordeanX · 2003-10-12 06:08 · Score: 1

I think someone added a zero in there.

300 average homes sounds more sane than 3000.
Re:VT Power Plant by Graff · 2003-10-12 06:51 · Score: 1

We had a severe ice storm and most of blacksburg lacked power. but so did campus!

It's probably the same problem the North East had a few months ago. One set of power generating equipment goes out and several closely related systems do the same. This is because when one part of a grid goes down the demand suddenly shoots up in another part, overloading circuits. These circuits automatically shut down to avoid burning up. When enough gets shut down the power plants also shut down to avoid mucking up the situation further.

That power plant you are talking about was probably still able to generate but since the transmission system was mucked up it was not able to provide power. The thing is, if these systems were designed a little better and were better able to split up into smaller independent grids then we would have less problems when one of them goes down.

--
Sapere aude!

Re:It runs MacOS X !!! by paulthomas · 2003-10-12 05:22 · Score: 1

Indeed.

But, as I said, the kernel is not of FreeBSD lineage. XNU is the Darwin mach kernel. Not that it really matters.

What is the link to the Real One Player version by CarlBenda · 2003-10-12 05:32 · Score: 1

Install Windows Media Player all you want. The BBC site uses a codec that M$ doesn't support in the Mac version of Media Player. God forbid someone installs it and gets disappointed by M$ Mac support. Any who M$ plans on bringing out the next version of Media Player for the Mac so maybe they'll fix that issue.

I see no link at all for the Real One Player version of the video. Will someone post that link or just reply to my post.

Re:Can the results be trusted? by TheSunborn · 2003-10-12 05:33 · Score: 1

They do? Then how come nobody ever sees it? My linux setup does newer crash. At at the university we got something like 100 indy/Octane/O2 workstations widtout ecc memmory and they all got a uptime which is equal to the last time someone cut the power. But we might just be luckey.

Martin

Re:17.6 Tflops sounds more like an Altivec number by Jeremy+Erwin · 2003-10-12 05:34 · Score: 1

Indeed, I forgot that the top500 site requires 64 bit precision or rather, specifies an error bound that requires such precision.

Re:Can the results be trusted? by Goldfinger7400 · 2003-10-12 05:36 · Score: 1

Dude, I love macs, but something tells me they didn't order all those computers with DVD-Burners in EACH ONE. And I don't know how much of a workout the PCI bus will be getting, unless they've got some new networking cards.

Re:Nice by Aardpig · 2003-10-12 05:38 · Score: 1

I said nothing about Apple, I merely commented on a benchmark procedure. It is you who are carrying the baggage of preconceptions; release yourself, my friend!

--
Tubal-Cain smokes the white owl.

SUPERCOMPUTER for the rest of us by danigiri · 2003-10-12 05:47 · Score: 1, Flamebait

Slap together 4-5 G5 (conceivable to be together in an office) and you have your home-brewn supercomputer cluster.

Then there's only the need of coming up with the applications to use it (besides research), apart from gobbling up seti@home data snippets for breakfast.

However, XCode and Rendezvous enable something in this front, enabling distributed seamless compilation of big projects.

Non-upgradable? by invisik · 2003-10-12 05:48 · Score: 1

Supercomputers, like PCs, become old in technology terms within about 3 to 5 years. But whereas you or I would go to a shop and buy one, Virginia tech has to start again from scratch. That's why they are building a new multi million dollar facility opening in 2006 for the next generation of Supercomputer.

So, they are just going to junk this setup in 3 years and start over? Doesn't make sense to me. The PowerMac is always had the better processor upgrades then PC's. I'm sure PowerLogix or Sonnet would love to make 1000+ CPU upgrade cards for them. Sure, the bus speed would be slightly behind a new system at that time, but the cost of buying an upgrade versus a whole new machine is worth looking at. Maybe after the first round of upgrades they should consider scrapping it.

-m

--
http://www.invisik.com

Re:Non-upgradable? by redJag · 2003-10-12 06:20 · Score: 1

Well consider that they wanted performance bad enough to spend $5 billion dollars to get it. Also, they are going to use these machines in the lab, where they would normally have to buy new computers. So basically, they are getting a new supercomputer instead of new lab machines.
Re:Non-upgradable? by arivanov · 2003-10-12 08:10 · Score: 1

More likely: swapping fans, power supplies and disks, putting a memorabilia sticker and handing out to faculty and into the labs. After all it is a piece of history and as such the fact that it is a bit out of spec will not be offensive for many people out there. If I worked there I for once would have liked to have it on my desk once it has been taken out of the machine room.

--
Baker's Law: Misery no longer loves company. Nowadays it insists on it
http://www.sigsegv.cx/
Re:Non-upgradable? by dbirchall · 2003-10-12 08:17 · Score: 1

Er, that's $5 million, not $5 billion. Big difference. :) $5 billion would get you the 1,100,000-node G5 cluster that would crank out, er... 7-15 petaflops or so.
Cooling would be a bit*h, but maybe you could rig up some sort of cogeneration plant, using the immense amount of heat generated to produce electricity, heat your dorms in winter, etc.

Re:Can the results be trusted? by imsabbel · 2003-10-12 05:49 · Score: 1

Yeah, sure. How do you thnik that they did implement that magic software correction in the linpack benchmark?
Simply run the matrix multiplies twice and check if the result is the same? (no, and i dont need to explain why)

--
HI O WISE PRINCE. WHT TOOK U SO DAM LONG?

Re:17.6 Tflops sounds more like an Altivec number by Jeremy+Erwin · 2003-10-12 05:51 · Score: 1

Wow, those GRAPE 6 machines must be great for modeling tertiary structure of proteins...

FINALLY!! by piecewise · 2003-10-12 05:54 · Score: 1

Virginia Tech's new 1100-node dual 2.0 GHz G5 Terascale Cluster. The report quotes the performance as 17.6 Tflops.

Finally I'll be able to play SimCity 4!!

--
The next comment I write will be ready soon, but subscribers can beat the rush and see it early!

Re:Can the results be trusted? by bash_jeremy · 2003-10-12 05:54 · Score: 1

If you order the g5 with 8 gigs of RAM, then it costs a lot more than 3,000 dollars.

Yea but... by LnxAddct · 2003-10-12 05:59 · Score: 1

does it run windows?

17.6 TFLOPS is Rpeak, not Rmax! by DeeKay · 2003-10-12 06:15 · Score: 3, Insightful

Simple equation: 4 FPops/cycle (IBM-PPCs) * 2GHz * 1100 G5s * 2 CPUs/G5 = 17.6 TFLOPS!

No *real* Rmax linpack scores are known yet, and from what i figured the submissions on Oct 1st are just for *inclusion* in the list, real Linpack scores can be submitted till shortly before (or even on!) the conference mid-November..

This article is BS and should be removed...

P.S.: 4 FPops/cycle per clock with 2 FPUs i hear you scream - Impossible! - That's due the Multiply/Add FMAC thing that counts as 2 FPops!

Re:17.6 TFLOPS is Rpeak, not Rmax! by addaon · 2003-10-12 11:39 · Score: 1

Should be 8 FPops/cycle, no? Four-element parallel MAC with altivec, pipelined to one a cycle?

(Yes, we all know Rpeak is meaningless. Still.)

--

I've had this sig for three days.
Re:17.6 TFLOPS is Rpeak, not Rmax! by DeeKay · 2003-10-12 12:05 · Score: 1

Well, unfortunately linpack is 64bit! ;-)
But if we take 32bit-FPops as a measure, the G5 could do a theoretical peak of 8 FPops/cycle (FMAC) using Altivec *plus* 4 FPops/cycle (FMAC) using the FPUs, so that's 12 FPops/cycle, which would give us a theoretical peak of 52.8 TFLOPS! ;-) Altivec only: 35.2!
Hell, the Earth Simulator has "only" 40.9 TFLOPS Rpeak!
Ofcourse it's near impossible to feed BOTH the FPUs and Altivec every single cycle, but pure Vectorcode works very well! Virginia Tech is also planning to do streaming simulations and gene-sequencing tasks for this thing from what i read, and these are 2 areas where Genentech and NASA's Craig Hunter proved that Altivec works our really really well! Craigs Vectorcode for Jet3D using Altivec is 10-12 times (!) faster than the scalar code!

Now, just imagine what would happen if the managed to "abuse" the 1100 Radeon9600s for numbercrunching puroses! ;-) Yes, there is already research underway in this direction..
How many GFLOPS does such a Radeon have again? ;-)
I remember the GeForce4 had 51 GFLOPS one year ago just using the pixel shaders...
Re:17.6 TFLOPS is Rpeak, not Rmax! by fiontan · 2003-10-12 16:15 · Score: 1

Reading this article, I wondered... how much time are these supercomputers around the world actually putting into useful calculations, and how much time into benchmarks to prove how fast they are?

I want to see another benchmark... how many TFlops per year are being spent on benchmarks vs useful applications!
Re:17.6 TFLOPS is Rpeak, not Rmax! by Wesley+Felter · 2003-10-12 16:19 · Score: 1

I would imagine they benchmark a supercomputer once and then devote the rest of its life to real work.
Re:17.6 TFLOPS is Rpeak, not Rmax! by acorsair · 2003-10-13 04:16 · Score: 1

You are correct in that 17.6 TFLOPS is an Rpeak figure. That is two FPUs, each returning 2 64-bit results per cycle, at 2 GHz time the 1100 G5s with two processors each. What will be interesting to find is the actual Rmax result. That will, of course be dependent on the number of actual compute nodes that can be used (likely less than the full 1100), compiler and network efficiency. Cluster effieciency has improved with the improvements in software and networking, as well as CPU and memory performance. The better ones are getting over 50 percent efficiency (my term for Rmax/Rpeak). Just a couple of years ago that rate was under 30 percent. Really good clusters are approaching 60 percent. What will be exciting will be if the VT G5 cluster can gain an Rmax of over 10 TFLOPS. For the price/performance and short timeline of installation, that would be impressive.
Re:17.6 TFLOPS is Rpeak, not Rmax! by DeeKay · 2003-10-13 23:32 · Score: 1

What will be exciting will be if the VT G5 cluster can gain an Rmax of over 10 TFLOPS.

It won't. 6-7 TFLOPS is a realistic guesstimate, Top10 would be really good.

a) it isn't even known if Mellanox has the OSX-drivers ready, so they might have to use GBit-EN, which would have a huge impact on performance.
b) This is the first Supercomputer cluster using Infiniband, OS X and G5s. Apple has no Knowhow in the area whereas Linux and Myrinet/Quadrics have been optimized and tested for years! Do you really think they can match that all on their own in a 3-months timeframe?
c) Apple isn't the only one submitting new clusters to the Top500 in November. The US is bent to beat the "humiliation" of the Japanese Earth Simulator, did you already forget? ;-)

If you think about it by JazzPhanatic · 2003-10-12 06:16 · Score: 1

It's amazing what a bunch of turkeys can accomplish when we set our minds to it. This makes us #3 in the nation in football and #2 in the world for computing power. Dont mess with the Hokies ;-)

MOD PARENT UP by Mark+of+THE+CITY · 2003-10-12 06:37 · Score: 2, Interesting

I'm glad someone out there thinks these things through...

BTW, an acquaintance told me of her ILLIAC IV days. With 64 independent processors it was the fastest pre-Cray machine, but sometimes did produce wrong values. Standard practice was to have 3 processors run the same problem and compare the results at the end, deciding that the 3x performance hit was worth it, if the results actually meant something...

--
The clearance system sounds logical. It is not. It is completely arbitrary. -- John Bolton

Re:Can the results be trusted? by cosmo7 · 2003-10-12 06:46 · Score: 1, Flamebait

They do? Then how come nobody ever sees it? My linux setup does newer crash.

Either that's a typo or the most ironically appropriate ascii bit error in the history of computing.

Re:Can the results be trusted? by cosmo7 · 2003-10-12 06:54 · Score: 1

ECC memory uses a set of extra bits - 7 bits to cover each 32 bits or 8 bits to cover each 64 bits. Each error-checking bit works like a parity bit for a different subsection of the memory it is correcting. By working out which subsections contain an error, the errant bit can be deduced and then corrected by simply flipping that bit.

ECC can also detect (but not correct) 2, 3 or 4-bit errors in each block, though that would be unusual.

When's a Tflop not a Tflop? Dan does the math by dbirchall · 2003-10-12 06:56 · Score: 1

The only skepticism I have regarding this figure (and I have this skepticism as a current G3 owner who's waiting for a dual G5 that had #$%#% better ship by Tuesday like they said it would! ;) is that in the past, Apple has published figures based on single-precision floating point performance, and TOP500 rankings use a benchmark that requires double-precision.

I had this pointed out to me in May of 2002 when Apple introduced the Xserve... in this Slashdot thread.

Apple's page about the G5's execution core states that its two FPU's are double-precision, and that it can thus "complete at least two 64-bit mathematical calculations per clock cycle." That's obviously an improvement over the G4, but let's see...

They've got 1,100 machines. Each has 2 CPU's for a total of 2,200 CPU's. Each CPU has 2 FPU's for a total of 4,400 FPU's. Each FPU is capable of doing a 64-bit (double-precision, I presume) floating-point calculation each clock cycle. The machines are running at 2GHz, or 2 billion clock cycles per second. So... 4,400 * 2 billion = 8.8 trillion 64-bit floating-point calculations per clock cycle.

To me, that's 8.8 TeraFlops, which is conveniently precisely half of the 17.6 figure. Did I forget to multiply by two, or did someone else multiply by two an extra time? Or... is the situation (most likely) that there are situations under which the G5's additional mathy bits can actually turn out a couple more FLOPs per cycle, but that's not going to be the case most of the time, and probably won't be the case running LINPACK?

Of course, it is important to point out that this is a theoretical maximum, and that things like interconnects, RAM limitations, chip failures, entropy and unauthorized use of the cluster to render hot babes in "Poser 5" will detract from it.

So... I don't think this will be the second-most-powerful machine in the world. Even relative to the previous list. I would not be at all surprised to see it in the top 5, though, and would be surprised if it didn't make the top 10.

Re:How they calculated the 17.6 tflops number.. by VojakSvejk · 2003-10-12 07:04 · Score: 1

Cool post. This is almost always what goes into numbers
quoted by IT groups that don't know anything about the machines they're running (I know; it's a redundant sentence). Just take the marketing drivel from the manufacturer, multiply by how many you bought, and quote it! Everything I've read from VT appears to have been synthesized that way.

I'm running a G5 at home now. I'm still wondering how a machine with a 32-bit OS "breaks the 4 GB barrier", given that it can't do anything that a 32-bit Xeon with more than 4GB of RAM can't do.

But it hasn't quadrupled my home energy bills, at least, and that include air conditioning.

Re:MOD PARENT UP by sakusha · 2003-10-12 07:11 · Score: 2, Interesting

Yep, that's an accepted practice in mission critical real-time systems. I recall reading about the IBM computers used in the Space Shuttle, they have triple redundancy, all 3 computers operate in parallel, and they "vote" on all results. If one computer doesn't agree with the other two, it is outvoted. Of course this is an extreme oversimplification of the software design, but you get the idea.

666 by __aafkqj3628 · 2003-10-12 07:26 · Score: 1

I suppose I should mention that the pageid for this thing is '666' :P

Re:MOD PARENT UP by canajin56 · 2003-10-12 07:33 · Score: 1

I believe the space shuttle has 6 computers that vote. I read somewhere that if they have one failure, they are fine. But they have to abort if they have two, because another failure would being them down two three, the minimum required for a proper voting scheme.

--
ASCII stupid question, get a stupid ANSI

1100 G5 dual 2Ghz 2*1152 Xeon dual 2.4 Ghz... by generationxyu · 2003-10-12 07:37 · Score: 1

...and considerably more than doubling the performance of the current number 3 1152-node dual 2.4 GHz Xeon MCR Linux cluster.

It's a good day to be a Mac man.

--
I mod down pyramid schemes in sigs.

Re:MOD PARENT UP by sakusha · 2003-10-12 07:41 · Score: 1

Hmm.. a quick google reveals we are both wrong, they have 5 computers. How odd (literally).

Re:When's a Tflop not a Tflop? Dan does the math by dbirchall · 2003-10-12 07:56 · Score: 1

If the G5 can do a fused multiply add at double precision in one clock cycle, the 17.6Tflop figure would then presumably depend on 100% of the code executed consisting of fused multiply adds, yes? :)

(Of course, there may be other fused 2-operation things it can do in one cycle... but it seems unlikely that code would consist of them so completely as to attain that theoretical peak.)

Ow, my electric bill! by dbirchall · 2003-10-12 08:07 · Score: 1

Hmmm... according to my electric bill, over the last year, I've used anywhere from 377 to 469 kwh of electricity per month, at a cost of US $84.89 to $105.15. My daily usage (averaged by month) has ranged from 12.6 to 14.7 kwh.

The dual G5 that should be arriving soon (are you listening, Apple?) has, if I recall, a 600 watt power supply. 600 watts * 24 hours = 14.4 kwh per day.

Yowza.

Um... yeah, I think I had better leave all the power-saving features turned on, and put it to sleep at night and all that other good stuff. Don't want to have to explain to the wife why the electric bill is suddenly double...

Re:Ow, my electric bill! by FueledByRamen · 2003-10-12 14:21 · Score: 1

Just because the PSU is rated at 600w doesn't mean it's drawing 600w all the time. Keep in mind that the PSU is designed to power hard drives, removable drives, fans, memory, PCI(-X) cards, external firewire devices, monitors (via the ADC connector), etc all at once. I imagine there's a fair amount of excess capacity built in.
You also have to remember, though, that 600w is the output rating of the supply - not necessarily how much it draws from the AC line. Switching power supplies (especially those made by the lowest bidder) are not 100% efficient - so even though there may be less than 600W drawn from the regulated outputs, it's pulling in more than that (and converting it to heat). A perfectly efficient switching PS wouldn't need a fan - since Apple's (and every other computer manufacturer's) does, it's wasting more than just an insignificant portion of energy during the conversion. A quick glance at a 400w power supply for a PC that I have rates its input at 120V 10A (for the full 400w output). I'd hope that the G5's power supply is more efficient than that, because then it'd be using (10 / 400 * 600) 15 amps, which is 1800 watts at 120v.

--
Every cloud has a silver lining (except for the mushroom shaped ones, which have a lining of Iridium & Strontium 90)

2006 Supercomputer by ShishCoBob · 2003-10-12 08:08 · Score: 1

Anyone have any idea where he was standing at the end of the video and mentioned a building would be built to house a new supercomputer? I go to VT. I know all the MACs are over at the corporate research center but I can't figure out exactly where he's standing at the end of the video.

--
http://www.maximum-cars.com - My little hobbie.

Re:2006 Supercomputer by AragornCG · 2003-10-12 09:12 · Score: 1

He's standing on the sidewalk between McBryde and Holden Hall, in the road towards the handicapped access lot in front of Randolph. The camera's pointed at Durham Hall (New Engineering Building, if you knew it in the early years after it was built in 1995.) The new building is going to go where Durham Lot 11 is currently, and will be multiphased with lots of research space for the College of Engineering and (rumored) a small dining hall to replace Shultz for lunchtime food operated by HDS (Housing and Dining Services, for those of you who don't know Virginia Tech consistently has the highest ranked dining halls in the nation, and we run it all in house instead of contracting to scum like Aramark.), and targeted at the engineering students (perhaps themed in some way). This explains why they're getting projects like the Randolph wind tunnel air-abatement out of the way early.

The project was approved by Virginia voters last year in the bond referendum. (Thanks, if you voted for it in that referendum. You had good company, Virginia supported it by a huge margin.) The building is described as "VTRI Phase I" on this page: Bond Referendum It will cost $31 million, slightly less (but not much) than Torgersen Hall. They're definitely going all out and doing it right.

For those of you not from Tech, the building is sited behind McBryde in square L3: Map This land was hotly contested for some time - a local developer owned it and wanted to put apartments there, and I think Tech finally gained the land via either eminent domain and a lawsuit. (It was definitely sticky.) That explains why it's one of the few chunks of land left on core campus undeveloped.

Re:Behold The POWER of Apple and tremble Penguinis by displaced80 · 2003-10-12 08:11 · Score: 1

blimey ... someone got hold of some cheap speed, didn't they?

--
What's the frequency, Kenneth?

Re:When's a Tflop not a Tflop? Dan does the math by TheSunborn · 2003-10-12 08:35 · Score: 1

You did forgot a factor of 2. First you say

Each processor can do 2 fpu instructions/clock. So each computer can do 4 fpu instructions/clock.

benchmark of mathematica by xahlee · 2003-10-12 08:40 · Score: 1

if you have g5 and mathematica, can you guys go here
http://www.scientificweb.de/mathstef3.html
http://smc.vnet.net/timings40.html

download the benchmark file and report the result?

thanks.

--
Xah
xahlee.org
http://xahlee.org/PageTwo_dir/more.html

Re:When's a Tflop not a Tflop? Dan does the math by dbirchall · 2003-10-12 08:50 · Score: 1

I don't think I missed that factor of two. But let's take it from what you just said:

Each computer can do 4 fpu instructions/clock.
(and those are double-precision 64-bit flops).

There are 1,100 computers. Times 4, that gives us 4,400 flops per clock.

(Note: given 1 DP flop per clock per FPU, this ties in nicely with my use of 4,400 as the number of FPU's, in my comment.)

4,400 flops per clock * 2 billion clock cycles per second = 8.8 trillion flops per second.

I'm still getting the same number. Where's the factor of two I missed?

Re:Can the results be trusted? by treat · 2003-10-12 09:07 · Score: 1

They do? Then how come nobody ever sees it?

Because nobody is checking to make sure their data isn't silently corrupted. And the corruption might be extremely minor - one bit flipped in audio or video might never be noticed. Or it might corrupt data that won't ever be looked at, or at least not for years. Or it might introduce a wrong answer one someone's homework, and they never realize why they got it wrong. Or it crashes a random application and no one suspects a thing.

The error rates on good hardware are pretty small, but guaranteed to be nonzero.

How 'bout $ per Tflop? by betasaur · 2003-10-12 09:31 · Score: 1

Seems to me what matters most is simply how fast is it and how much did it cost?

$300,000 per Tflop - I doubt any other on the list is even close...

Re:Can the results be trusted? by afidel · 2003-10-12 09:52 · Score: 4, Informative

No, ECC ram typically is just made with faster internals. As an example most ECC comodity ram is CAS2 latency whereas most generic ram is CAS3, so the ECC ram will perform exactly the same as the non-ECC ram. You can buy CAS2 non-ECC ram but it's nearly as expensive as the ECC ram. If you have a simple idiot check at the end of a complex calculation then saving the cost of going with ECC may be worth it but most clusters this large will be used on too many different projects to assume that all of them will have such checks. For an idea of how important ECC is read (a href="http://www.ibm.com/servers/eserver/pseries/c ampaigns/chipkill.pdf">This IBM whitepaper on their chipkill ECC scheme. Even normal SEC ECC ram (what most ECC ram is today) will have aproximately 900 failures per 10TB per three years. I think that IBM is right and that eventually all ram will be RAID-M, that is a RAID5 style array of redundant memory banks that are composed of ECC banks. At future densities this will be necessary because a single high energy particle will have the ability to scramble an entire memory word including it's ECC checking bits.

--
There are 4 boxes to use in the defense of liberty: soap, ballot, jury, ammo. Use in that order. Starting now.

Re:How much power ?!?! by afidel · 2003-10-12 10:12 · Score: 1

The simple way to do it is to continuously pump the waste heat into the heating ducts during the winter and have the main heating system on a zoned thermostat system. That way your main heating system is only producing (n-m) btu's where n is its normal btu output and m is the waste heat output from the cluster. If you later remove m then the system still has the ability to produce n btu's. Besides that most datacenters have INCREASED in btu output and power consumption over the years not decreased. The old systems were large and hot but the new systems are small and REALLY hot. Trust me I have been in quite a few datacenters that have had their HVAC systems upgraded from the days of room sized mainframes.

--
There are 4 boxes to use in the defense of liberty: soap, ballot, jury, ammo. Use in that order. Starting now.

Well, you reckon wrong... by djupedal · 2003-10-12 11:59 · Score: 1

Read the article...they don't just plug them in and turn them on.

"The Supercomputer, unofficially nicknamed Big Mac, was built in just three months. Right from the start there were major hurdles that could only be overcome with significant construction in and around the building. Running 1100 computers in a 3000 square foot area sends the air temperature well over 100 degrees Fahrenheit. In fact the heat is so intense that ordinary air conditioning units would have resulted in 60 mph winds. So specialized heat exchange cooling units were built that pipe chilled water into the facility."

Re:When's a Tflop not a Tflop? Dan does the math by stephentyrone · 2003-10-12 12:30 · Score: 1

yeah, but scientific computation *does* consist entirely of multiply-adds (ish). How do you computationally solve a problem? You find approximations in some convenient linear space - you reduce the problem to a whole lot of linear algebra problems. And linear algebra problems are solved by a whole lot multiply-adds. i'm being incredibly simplistic here. really mindbogglingly simplistic. and anyway, why bother to speculate when there will be real numbers in a month?

Re:Not the double speed myth again. by SynKKnyS · 2003-10-12 13:08 · Score: 1

If they went to RackSaver (http://www.racksaver.com/shop/UserCustomize3.asp? lSystemID=70) they could have created a super computer with over 20 TFlops of peak performance for that same 5 million.

Re:Can the results be trusted? by Hoser+McMoose · 2003-10-12 14:02 · Score: 1

It's been previously discussed, and their "error correction algorithms" are going to do dick-all for them if they can't trust their data, and without ECC, you can't trust your data.

The first poster is correct, results from this machine can not be trusted as being 100% accurate. It may be that they can live with the lack of accuracy, but it's definitely something that they will have to figure into their work. With 4.4TB of memory, they are going to have soft memory errors on a VERY regular (daily?) basis, and they can NOT be caught by some sort of software algorithm unless you store every bit of data twice and do every single calculation twice. If you're doing that, you might as well save yourself a few million and buy a 550 node cluster with ECC and get the same result.

Re:Not the double speed myth again. by Ffakr · 2003-10-12 16:02 · Score: 1

1) your link doesn't work

2) 5 million is the cost of the project.... that is 1100 Macs with (I think) 4 GB of ram and 160 GB HDs, GigE and FW800 on board, Infiniband cards, Customized 19" racks, a customized cooling solution.... The Macs were just a fraction of the overall cost, a sizeable fraction but there were a lot of other significant costs.

--

I'm not feeling witty so bite me

Re:When's a Tflop not a Tflop? Dan does the math by DeeKay · 2003-10-12 20:47 · Score: 1

I'm still getting the same number. Where's the factor of two I missed?

2 CPUs in each Mac!

FYI: The formula FPops/cycle * MHz * number of CPUs is the formula for all Rpeak-Values in the Top500! Try it yourself by dividing the listed Rpeak-numbers down and you'll end up with the FPops/cycle for each CPU-Type! ;-) Intersting: IBM-PPCs (Power4, Power3, G5) have 4 FPops/cycle, Itanic dito, x86 has 2 FPops/cycle and Alpha (!) aswell! Yes, Alpha, the FPU-Monster! ;-)

This 4 FPops/cycle "FMAC-feature" (besides the price and Altivec!) probably explains why they went with G5s in the first place!

17.6 TFLOPS NOT Apple-BS-Marketing but just Rpeak! by DeeKay · 2003-10-12 20:54 · Score: 1

To all the ones who like to scream "cheating" and "lying" trying to blame Apple:

The formula FPops/cycle * MHz * number of CPUs is the formula for all Rpeak-Values in the Top500! It's the way Rpeak is calculated! Try it yourself by dividing the listed Rpeak-numbers down and you'll end up with the FPops/cycle for each CPU-Type! ;-) Intersting: IBM-PPCs (Power4, Power3, G5) have 4 FPops/cycle, Itanic dito, x86 has 2 FPops/cycle and Alpha (!) aswell! Yes, Alpha, the FPU-Monster! ;-)

This 4 FPops/cycle "FMAC-feature" (besides the price and Altivec!) probably explains why they went with G5s in the first place!

P.S.: Did you know that test by the german Magazine C't using IBMs XLC/XLC and P4s/Xeons with Intels ICC basically proved what apple claimed with their GCC comparison? That the G4 is faster than a 3Ghz P4/Xeon Dual 3GHz in FPU but slower in Integer? And the IBM-Compiler is still in beta, mind you! ;-)

So much for cheating lying Apple, hehe! Ofcourse, Slashdot, The Register, the Inquirer, Wired and all the rest would NEVER write an article about that, a supposed "Mac-User" who's accusing Apple of cheating without having the glance of a clue is just so much better, and who are they to admit they made a mistake?

Re:Video? by coolgeek · 2003-10-12 21:20 · Score: 1

Posting video to slashdot, isn't that a bad idea?

And soon to be considered a felony in a courtroom near you.

--

cat /dev/null >sig

Oversimplified - I don't get it by alexdewaal · 2003-10-12 22:15 · Score: 1

I really don't understand how a voting scheme can enhance reliability.

What a majority says is not right per se, it's just less likely to be questioned...

But really, how different need the voters to be, to get the same results independantly?

Different hardware?
Different OS?
Different programming language?
Different programming model?
Different requirements? Oops!

The best way to popularize Open Source Software is to support the BSA.

Re:Can the results be trusted? by Marovingian · 2003-10-12 23:44 · Score: 1

In the slashdot spirit of things, I would like to differ- Apple has historically (at least for the last 3 years or so) shipped it's computers with 2-2-2 spec'd RAM, OS X typically doesn't like to install on a computer that has 3-2-2 latency RAM installed. A workaround that many people with the slow (read: cheap) RAM must deal with is uninstalling their non-OEM RAM for the OS X install, then reinstall the RAM after the OS installation. The reason being the RAM tests performed during bootup from the install CD/DVD are more stringent than the normal booting-from-the-HD startup RAM tests. There are multiple threads on macintouch, macfixit, [and probably even] apple.com's message boards et el to document this.

--
Cursing in the French language is like wiping your ass with silk.

Re:Imagine... by LondonLawyer · 2003-10-12 23:49 · Score: 1

Damn. Beaten to it.

the real question... by Box+Checker · 2003-10-13 04:16 · Score: 1

is it snappier?

Re:I don't give a flying f*** about you by darby_smeed · 2003-10-13 04:46 · Score: 1

First, I am amazed that you support the annihilation of small livestock. (That's what your post means to me.) Second, check who you're replying to. It's Chuck Shotton. Stop reading Slashdot and read some internet history.

Re:When's a Tflop not a Tflop? Dan does the math by dbirchall · 2003-10-13 07:21 · Score: 1

Umm, no, I didn't miss that there were 2 CPU's per computer. There are 2 CPU's and each CPU has 2 FPU's and each FPU can perform "at least" one double-precision 64-bit floating-point operation per clock cycle. Thus my quote of:

There are 1,100 computers. Times 4, that gives us 4,400 flops per clock.

Or, going back to my far more comprehensive original comment:

They've got 1,100 machines. Each has 2 CPU's for a total of 2,200 CPU's. Each CPU has 2 FPU's for a total of 4,400 FPU's. Each FPU is capable of doing a 64-bit (double-precision, I presume) floating-point calculation each clock cycle. The machines are running at 2GHz, or 2 billion clock cycles per second. So... 4,400 * 2 billion = 8.8 trillion 64-bit floating-point calculations per clock cycle.

It appears (based on these comments) that the only "factor of two" I missed would be the G5 FPU's ability to perform a certain pair of double-precision 64-bit floating-point operations (multiply and add) in a single clock cycle -- a pair of operations which apparently come up frequently in scientific programming. (Something that, not being a scientific programming sort, I didn't know.)

(Gosh... it's almost like they designed it to be really fast, or something!)

So... hmm. I guess like everyone else, I'll have to wait and see what the numbers are on the TOP500 list. That's an interesting formula you've got there, by the way.

(I wonder how fast it'll run The Sims?)

Here's a link for a Quicktime version of the clip by LionMage · 2003-10-13 08:42 · Score: 1

A buddy of mine is hosting a clip he re-encoded in Quicktime for those of us who don't want to use WMP or RealPlayer. The link was posted earlier on the Accelerate Your Mac website. The direct link is here.

Re:Not the double speed myth again. by Vitriol+Angst · 2003-10-13 09:37 · Score: 1

Perhaps. But there might be a bit more involved to getting all those computers to work on the same problem. I'm sure Rendesvous had something to do with that. It's also interesting to note that they chose to go with OS X, rather than Yellow Dog Linux. Again, probably a configuration issue.

Possibly the high bandwidth bus may have more effect with a network of cpus than the individual performance of CPUs. Perhaps an expert would know.

Also, that 17.x terraflops is for Peak. No figures on sustained performance. But perhaps more than 5.7 TF, putting it in top 5.

--
>>"ad space available -- low rates!!!"

Re:parent is right, lame choice of players by mzajac · 2003-10-13 09:45 · Score: 1

I do, but Microsoft's out of date WMP for Mac won't play the audio in the BBC files.

Re:Can the results be trusted? by OS24Ever · 2003-10-14 00:21 · Score: 1

The system knows because the parity bit is checked and if there was an error it's not in parity, causing a parity error and rebooting

--

As a rock-in-roll Physicist once said, No matter where you go, there you are.

Re:The press release mentions Linux for some reaso by gabebear · 2003-10-14 00:49 · Score: 1

The slide-show says MacOSX which sounds more likely, They could even mean just Darwin(The command-line totally open source part of MacOSX). I imagine that the press people meant GNU software. The slide-show also says that they are using IBM's compiler that was just released for MacOSX.

http://don.cc.vt.edu/tcfslides.pdf
page 13 of the slide has the interesting stuff.

On a side note, without ECC Ram this thing can not be trusted to give the right answer all the time, I wonder what the researchers will do about that.

Re:Can the results be trusted? by Ed+Avis · 2003-10-14 07:15 · Score: 1

Are you saying the Mac uses parity memory? I assumed it did not (as mentioned above).

--
-- Ed Avis ed@membled.com

More tech info in previous interview by Hagen · 2003-10-14 14:18 · Score: 1

Hassan Aref, Dean of Engineers at Virginia Tech, answered a lot of questions about the setup of the computer during a 20 minute interview last month. Go to the Your Mac Life archives page and pull down the MP3 or AAC archive of the Sep 10 show. The interview itself is 1:15-1:35.

Re:Can the results be trusted? by DAldredge · 2003-10-18 03:27 · Score: 1

They do not use parity memory on the G5.

Japanese MORON or White racist in disguise by sinsoldier · 2003-10-23 08:56 · Score: 1

"Your Identity is Filipino. Think about what that means.... " I SAY IT MEANS AWESOME!!!!!!

You little, pathetic, ugly, piece of shit. Filipinos requested NOT TO BE CLASSIFIED WITH OTHER ASIANS on racial/ethnic classifications, so how could you say they want to be like the fucking Korean culture ripping Japanese or the Chinese???????? YOU ARE A ROACH that badly needs to be SQUASHED.

Here is a list of Philippine Achievement in the U.S.

WRITER: Jessica Hadedorn

COMEDIAN: Rob Schneider

ACTORS: Lou Diamond Philipps, Nia Peeples

ACTRESSES: Tia Carrerre, Tamlyn Tomita, Phoebe Cates

SINGERS: Leah Salonga (Broadway star & Tony Award winner), Julio Iglesias Jr., Prince

MISS UNIVERSE: Gloria Diaz, Margarita Moran

MISS U.S.A.: (can't recall her name)

ATHLETES:
* Pancho Villa - World Champion Flyweight Boxing Champion 1924
* Gabriel Elourde - World Champion Boxing 1962
* Vicki Draves - U.S. Olympic Diving 1984, 1st woman in Olympic history to win springboard and platform diving, America's sports sweetheart in 1948
* Bobby Balcena - Cincinnatti Reds Baseball Player 1948-1963
* Roman Gabriel - Quarterback NFL MVP 1969 Hall of Fame
* Tai Babilonia - Olympic Ice Skating 1979 Champion
* Benny Agbayana - Baseball Outfielder 2002 Colorado Rockies
* Jerome Williams - baseball pitcher, S.F. Giants

POLITICIANS & BUSINESS PEOPLE & PROFESSIONALS
* Governor of the State of Hawai`i is a Filipino-American named Benjamin Cayetano
* Co-founder of Los Angeles. In 1781, Antonio Miranda Rodriguez Poblador, a Filipino, along with 44 other individuals were sent by the Spanish government from Mexico to establish what is now known as the city of Los Angeles.
* Loida Nicolas Lewis is chairman and CEO of TLC Beatrice International Holdings, Inc., a multinational food company with sales in 1995 of $2.1 billion. Working Woman magazine hailed her as the top businesswoman in the country for 1994. Mrs. Lewis was the first Asian woman to pass the New York State bar exam without having studied law in the U.S
* Ruben Aquino and Cynthia Ignacio, two Filipino-Americans, were instrumental in the creation of the Disney animation film "THE LION KING."
* Eleanor "Connie" Concepcion Mariano - President Bill Clinton's physician
Virginia Rep. Robert Cortez-Scott, a Harvard alumnus.
* Philippines Herald war journalist Carlos P. Romulo in 1941. (He was also the first Asian to become UN President

****
(Not an American history, but I wanted to include) my Philippine hero, the great revolutionary leader of the rebellion against Spain before U.S. colonizatoin:
Jose Rizal could read and write at age 2, and grew up to speak more than 20 languages, including Latin, Greek, German, French and Chinese. What were his last words? "Consummatum est!" ("It is done!")

INVENTORS
* Nikola Tesla - inventor of fluorescent light
* Eduardo San Juan designed the Lunar Rover or "moon buggy" which was used by the Apollo astronauts to explore the moon?
* Dr. Abelardo Aguilar, a Filipino-American, discovered the now widely used antibiotic known by its generic term as erthyromycin. Chances are, you've already used this antibiotic.
Kali martial arts - unknown inventor in Philippines

Re:Japanese MORON or White racist in disguise by sinsoldier · 2003-10-23 08:58 · Score: 1

CORRECTION: Inventor of fluorescent light: AGAPITO FLORES, not Tesla. Tesla was Serbian.

Slashdot Mirror

More on Virginia Tech G5 Cluster: 17.6 Tflops

222 of 390 comments (clear)