The first poster missed the point completely. VT has a working 17.6 Tflop G5 cluster in only 3 months for a smidge more than $5 million TOTAL! UT has a partially completed 3.7 Tflop Dell crash-cluster that will take 5 years to complete for an amazing $38 million and counting. Who's the a** now???

the keyboard on that thar Dell superpooter won't fail like my Inspiron did. Dude.

The "Lonestar" cluster more than triples the power of the center's "Longhorn" cluster, which features 1.16 trillion computer operations per second, or 1.16 teraflops. The "Longhorn" cluster features servers from Armonk, N.Y.-based IBM Corp.


I wonder if the predecessor is called Longhorn because this is the configuration you'll need to run Longhorn smoothly, if it ever will be released.

Well, the $38 million figure is for the life of the program. This undoubtedly includes the overhead for running the system (electricity, rent) and a sizable staff (insert TCO joke here). I'd be surprised if the actual cost of building the cluster is more than $10 million.

...that Migrainesoft didn't supply the OS for the Dell clusterfuck. Oh, wait, they don't have "Windows XP Professional Server 2003 for Large Computer Clusters - Enterprise Edition".

Educational Institutions are the ones considered to be using commonsense in its broad sense when its come to choosing and buying technologic products. But once again, this is the alarming proof that even schools or high educational institutions are not lead by smart people. They make a company like Dell s**** them up and get richer without puting on the table the best solutions in favor of the US students.
But the Virginia Tech case with Mac G5 cluster will put those mentally corrupted people to shame in a very near future.

While the G5 cluster may end up being better how about we wait for linpack scores for the VT cluster rather than THEORETICAL PERFORMANCE OF 17.6TFLOPS. You people give Mac users a bad f****** name by not actually learning the facts before you start comparing numbers. Quit being f****** morons and recognize that if the G5 cluster actually proves to be substantially superior IT WILL BE USED. Also, this cluster uses linux so the whole "longhorn" and "crash-cluster" jokes are pretty inane.

If you want your opinion to be taken seriously, do not engage in ad hominem attacks and misinformation. That means most of you who post on MacNN I suppose.

And as a UT student I can say there are still a high percentage of Macs in labs and on desks of professors and secretaries. Mostly flat panel iMacs, haven't seen any G5s around. More PCs, but the Macs make up a significant percentage.

No matter who is correct in figures, it's good to see that the general public will see numbers finally putting Apple ahead. Who knows how many times mere numbers and dollar signs have put a bad cloud over Apple. We will see who's who when the Linpack scores comes out. The deadline for those rankings were Oct 1st. Did UT make it in?

"If you want your opinion to be taken seriously, do not engage in ad hominem attacks and misinformation. That means most of you who post on MacNN I suppose."

"Oooh. I took a latin class and learned a pretentious expression that will impress all the kiddies. I just wish I wasn't so impertinent and all the smart people would listen to me, me, me."

If you don't think that IT decisions like these are heavily influenced by 'business politics' then you should get your head out of your books and check out the real world.

In Stalin's SU you might have been shot for wasting monetary resources like that.

Apple has a large branch in Austin too. I doubt location had to do with anything as everything comes from China and Taiwan anyway.

...Though the G5 is made in NY...

I have to agree with the OP.
From my reading of the article, the 38M was planned expenditure over five years..and I couldn't tell if that was only for boxes, or infrastructure as well.

Currently, they are touting a 300 box cluster, with 200 more in the coming year. The article does not say how big the cluster is eventually planned.

3.6 Tflops/300 boxen = 0.012 Tflop/Intel box.
At UV, 16 Tflops/1100 boxen = .014 Tflop/G5 box

There just is not enough info in the article to draw any conclusions, IMO, regarding cost-effectiveness. Dell did, however, bring Cray on board (their nickel ?). That sounds like an added expense :)

According to the Ars Technica forum the 17 Tflops are indeed the theoretical performance. Real world numbers should be around 50-70%. Pretty impressive no matter BUT if all Mac news services repeat these "false" numbers over and over again it GIVES "us" a bad name.

Lets wait 'til November when the new top 500 list is published.

Quit parroting BS that some dude on Slashdot came up with for christ's sake!

17.6 TFLOPS is NOT a linpack test result but just a THEORETICAL MAXIMUM that everybody can easily verify by calculating himself using the following standard Rpeak-Formula:

4 FPops/cycle (2*IBM-PPC Multiply-Add) * 2000 Cycles/Second (MHz) * 2200 (CPUs) = 17.6 TFLOPs

Go on! Open Calculator.app and verify it for yourself!!

All of a sudden everybody seems to be bent on propagating hugely overblown expectations, and when the cluster shows up "only" on #6 with 6 or 7 TFLOPS everybody is "very disappointed" at the performance, just because SOME guy on Slashdot guessed the BBC was talking about Linpack scores when it said "..capable of 17.6 Trillion FLOPS"!

If it makes it into the Top5 it will be a h*** of a performance! Quit expecting the impossible! A Rmax/Rpeak Ratio of ~90% like the Earth Simulator is simply NOT POSSIBLE with a commodity cluster! Especially when OS X and Apple are both complete newcomers in supercomputing and this thing was built in WEEKS only!..

Be realistic and stop propagating false assumptions as "facts"!

"Be realistic and stop propagating false assumptions as 'facts'!"

Um, dude, I don't know if you've noticed, but we live in the Bush Era now. Facts are irrelevant. Waiting or thinking about things is verboten.

If the conservative news media ever got wind of your commie-like desire for "truth" and "analysis," why they'd have to fold up overnight.

No sleep till the first Tuesday in November 2004.

Mike Dell and his Boxes from h*** are screaming ME TOO!
AGAIN!

They can make a cluster just like Apple.
They can make an iPod wannabe.
And, let's not forget, they invented wireless networking, too!

Waiting for hard numbers is a sensible course before we all start tooting the G5 horn TOO loudly. A side by side comparison of costs AND projected TCO would be interesting to see, as well.

However, on the surface of things, it sure makes that Apple shine!

Remember that these are research computers, do you really think they would replace IBM computers with OS X? I'd like to hear you tell a tenured professor that he has to re-program all of his research applications because Unix is "better" or costs less.

They are trying to get more grants by upgrading the computers, that doesn't mean they physically need the fastest system in the world. They need good PR (not getting it from the Mac sites, but probably will get it from regional media) and good research facility. I'd think switching OS's would be a good way to jerk around your facility.

And then if you needed donations/grants would you really try to jerk around one of your most well known Alumni?
iolaire

First: I find it great that there is now a Mac Supercomputer, and I'm sure, it beats the one
in Texas. However:
It clearly does say: "The cost of the five-year project is about $38 million" So, that is not just
the hardware, this probably includes the salaries
for people running/administring the system, overhead,
and who knows what else. If anybody
can find the hardware costs, that would be the
right comparison.

$38M is for 5 years...
$5M is just purchase price...

Add the $129/6 mo apple service pack price and the VT cluster is 100 million gazillion dollars

Last I heard the Apple cluster cost 5 million over the next 5 years.

Hello again,
just look up the web site of the texas supercomputing
facility and you will get better info, here is an
excerpt that shows that it's not just the hardware,
but a lot more:
"The TeraGrid award is another important achievement in TACC’s rapid growth in the past two years. In 2001, TACC
began operation of the cutting-edge ACES Visualization Laboratory. In early 2002, TACC became the first center in
the U.S. to deploy IBM Power4 Regatta-HPC systems for academic research. In December 2002, TACC was a key
participant in obtaining a $38 million award of university, industrial, and private funding that is now enabling The
University of Texas at Austin to build and operate a terascale grid computing infrastructure (UT Grid), including the
enhancement of the IBM Power4 system and the construction of the Cray-Dell cluster. The same funding created the
Institute for Computational Engineering and Sciences (ICES), enhancing the former Texas Institute for
Computational and Applied Mathematics as a world-class center for computational science and technology
research. The funding also made it possible for TACC to hire additional experts in high-performance computing,
scientific visualization, and grid computing. In early August, the university’s board of regents approved a new facility
for TACC with a much larger machine room and many more offices, which is expected to be completed in mid-2005.

...as long as it makes profits for dell, so be it...
..take from the apple, and give to the Dell..

UT used to be about 80% mac and only 20% pc, but over the last 5 years, that has changed to about 60% pc to about 40% mac. price of desktops had a lot to do with that. sadly, no one considers support.

apple is down the street also though. i wonder if they were even given a chance to bid

an "alum" of UT, he is a drop-out.

He may have bought himself some popularity there recently, but before that he wasn't always on the best of terms with the University. The business school has always been troubled by Dell's lack of education and subsequent success (which he has not been shy in rubbing their noses in.)

You may look at the business school as a bunch of pretentious eggheads trying to maintain their control of the tools of business success. On the other hand, you can view Dell as a one-in-a-million lucky fluke and his disrespect for education harms most people for whom it is the most reliable tool to attain success.

From an InfoWorld Article

"Dell's list price of a configuration similar to Lonestar is $1.9 million, with services and installation charges expected to bring the total cost to around $3 million, a Dell spokeswoman said. "

"The university received educational discounts off the $3 million fee quoted by Dell, but is not disclosing the actual price it paid for the cluster, said Tina Romanella, communications and development manager for the TACC."

The $38 million is for the 5 year project.

UT has announced that it's tuition will increase 31% next semester... (thanks also to a new law deregulating state-run university tuition)

You're gettin' dope slapped by the provost and the comptroller and the Texas lege...

Not to mention the next site visit by NSF - that should be a hoot.

This should be on the Apple Hot News page.

1. Oklahoma
2. Miami
**3. VaTech**
.
.
.
20. Texas

The cluster gets a steady 257000 FPS in Quake 3.

please, y'all. learn to spell "definately" by the time you leave college.

What a pathetic partisan crowd who rally around unsubstantiated numbers to proclaim their supposed intelligence. Sad.

According to the Ars Technica forum the 17 Tflops are indeed the theoretical performance. Real world numbers should be around 50-70%. Pretty impressive no matter BUT if all Mac news services repeat these "false" numbers over and over again it GIVES "us" a bad name.

17.6 TFLOPS is NOT a linpack test result but just a THEORETICAL MAXIMUM....
All of a sudden everybody seems to be bent on propagating hugely overblown expectations, and when the cluster shows up "only" on #6 with 6 or 7 TFLOPS everybody is "very disappointed" at the performance, just because SOME guy on Slashdot guessed the BBC was talking about Linpack scores when it said "..capable of 17.6 Trillion FLOPS"!

Well, if you go by the other guy's expectation of 50-70%, then the expected range should be about 8 to 12 TFlops.

Oh, and the Dell cluster's number is also a theoretical maximum. I don't see anyone mentioning that. So they are comparing two theoretical maximums here, which is about as OK as you can get with theory.

Or should I say to the dud who said:

"...if the G5 cluster actually proves to be substantially superior IT WILL BE USED. Also, this cluster uses linux so the whole "longhorn" and "crash-cluster" jokes are pretty inane."

1. Not so. What is superior is usually NOT what is used. As a Mac user, it is clear that the following tells it like it is with some humor.

"DOS Computers made by companies such as Compaq, HP ...and millions of others are by far the most popular... Macintosh fans, on the other hand, may note that cockroaches are far more numerous than humans, and numbers alone do not denote a higher life form."
-The New York Times

2. And the fact that the Dell cluster opted for Linux is not exactly an endorsement for the Microsft OS, if you know what I mean.

is the clear winner here. I wonder if Mr. Dell can sleep at night?

The educational price for the Dual 2GHz G5 is $2,699. The full cost of the G5's (without any volume discount) would be 1,100 x $2,699 = $2,968,900.

plust 4GB or RAM per computer...

I just can't get the reasoning of "part of that is for maintaining the cluster".

you know, 30m can hire 480 research assistants or 100 faculty to take care the machines for 5 years.

now we all know where to apply for jobs.

Man, what zany group. Like comparing cockroaches to computer users make a valid comparison? Or an one-time purchase (no consideration for added cost of RAM, cable, rack, etc) or the maintenance/support contract or the buildining/energy cost etc is comparable to a figure reported by a biz journal. Idiot zealots.

Nobody was comparing cockroaches to computers, you moron. If you could read better than a 1st grade level, you might have been able to pick up on that. However, if you feel like you know better than everybody else, why don't you prove that the Dell cluster is both faster and cheaper than the G5 cluster at VA Tech. I mean if you're going to call us idiot zealots, you must have some way of backing that up, right? Right?

Sure there are some zealots here, but at least read the comments closely enough to understand their point. The previous reader did not compare computer users to cockroaches; he compared DOS machines to cockroaches. The point he was alluding to is that whatever product is most popular or successful is not necessarily the best product available. Markets are fickle, and can be driven by forces other than logic, despite what Free Market purists believe. The fact that the National Enquirer is the most widely purchased newspaper in the country does not make it a quality publication.

Your other point is valid. We don't have comparable numbers to compare cost/performance costs at this time, just make wild-eyed conjectures. However, what is clear is that the new G5 processors have pulled the Mac platform from the Motorola doldrums and put the Mac back into being, at the very least, in the same category of performance as the best Intel has to offer.

And by the way, calling people "idiot zealots" is about as valid as calling computer users cockroaches.

Back in the early 90s, UT Austin had the largest installed base of Mac's in the world. The Texas Student Union had a great computer store through which students got their Macs at a more than 40% discount. Once those great Edu discounts went away, and Dell opened his checkbook, the campus was slowly assimilated. Now, if one attends the business school, one is required to buy a Dell laptop.

Course, it might help to read the little notice at the bottom, which kind of puts everything said in it's place.. yes, G5's are certainly better as individual computers, but if performance is what matters, then it's which is faster, for the price.

Saying 300 is "small potatoes" and 1100 is huge, is a total distortion of truth.. 300 computers is a rather large cluster, I'm afraid, and the way you talk about it you make it sound like 1100 is an order of magnitude larger, which it certainly isn't.

As for the "how many FPS does it get in Quake III" comment, I really hope that's a joke, as it would be no faster than a single G5 with a Radeon 9800- the Video Card is the bottleneck. Think before you speak. Actually, don't speak at all.

for all the idiot posters that make all Mac users look like idiots. This is clearly not an apples to apples comparision. Read Slashdot for more insightful discussion.

Thanks, Dave Shroeder, for making Macnn and Mac users look like idiots.

http://www.theinquirer.net/?article=12114

Performance and price are indeed what we are talking about. Unfortunately, the data we have at hand is sketchy and preliminary, but it hints that the G5 cluster may rate very high. In previous articles, the researchers building it have spoken of how little effort it is taking to configure the system, much less than similar clusters with other operating systems.

And yes, 1100 is not an order of magnitude larger than 300. It's only 366% larger.

And good heavens, of course the Quake FPS comment was a joke. If you have a hard time recognizing that, reading online forums may be a troublesome experience for you.

I'm the person who was quoted in the original article. BUT I DID NOT SUBMIT IT TO MACNN. MacNN picked it up, itself, from Apple's scitech mailing list. Secondly, MacNN had the full capability, one would hope, to read the stories. They published it, not me. Third, FYI, I did not submit this story to slashdot. But it has spread like wildfire. They took my comment, where it was immediately pointed out that the $38M included other things, and printed it. From here, it has been linked everywhere.

Here is the followup message I sent to the scitech list:

--------------

http://www.theinquirer.net/?article=12114

So, the value of UT's 3.7 Tflop (Rpeak) asset itself is $3.0M.

The value of VT's 17.6 Tflop (Rpeak) asset itself is $5.2M.

Let's try this again:

UT cost per Tflop (Rpeak): $811,000
VT cost per Tflop (Rpeak): $295,000

The UT cluster costs 1.7 times as much for almost 5 times less theoretical maximum performance.

Or, the UT cluster costs 2.7 times as much per Tflop (Rpeak).

Some notes:

1. If Rpeak isn't important and/or is useless, than neither institution should be quoting the figures. It's just as valid for one as it is for the other. Take it at face value.

2. The Apple cluster is *clearly* cheaper. I don't see how anyone can argue with that, though I'm sure some will try. This will include operational, infrastructure, and staff costs: because I'm sure that the Apple cluster isn't going to be arbitrarily higher, and running academic supercomputer centers will likely have very similar costs associated with them.

3. The VT cluster *will* be the most powerful 'homebuilt' supercomputer, *and* one of the least expensive, most powerful super computers ever, with likely significantly less cost than anything that will rank near it, including possible long-term operational cost and maintenance advantages by running Mac OS X out of the box as opposed to staging and maintaining Linux.

So while it's *still* admittedly not that simple, who can argue against the price/performance of the Apple cluster? That "Cray support" for the UT installation comes at a cost.

What if the Apple cluster handily outperforms, by large margins, the UT cluster in real life research problems? Then who's got the short end of the stick? It seems like everyone was resting comfortably with whitebox Intel systems and systems integrators building everything with Intel and Linux. Now someone tries Apple, and it looks like not only is it going to be cost-competitive, it might be a LOT BETTER. But the hegemony won't give in to that so easily...

In closing, I regret posting the original story as worded. It was intended to be more of an off-the-cuff statement on Apple's entry into this realm, not a hard and fast statement carved in

(continued)

stone (that ended up apparently getting quoted and linked all over the internet). But I stand firmly by the spirit of the original post, which is that Apple looks like it's going to be a cheaper and "better" (subjective, I know; in this case, "faster for the practical tasks at hand") than some given alternative. Can I prove that yet? Nope. But I'm sure that will come to pass with Rmax, and with real world use.

Additionally, I might say that many news stories talked about the "project cost" of the VT cluster being $5.2M. Only after looking at the presentation did I find that is really only the initial asset purchase. So upon seeing "UT's $38M supercomputer", again quoted as a "project cost" in Atlanta Business Journal, well, ... I apologize, because I hadn't found out what the $5.2M encompassed - just reading two different news articles, about the same topic, with statements about theoretical max performance, and talk of "project costs" (which in this case referred to entirely different things).

Regards,

Dave Schroeder
University of Wisconsin - Madison
das@doit.wisc.edu
http://das.doit.wisc.edu/

From: "John R. Boisseau"
Date: Tue Oct 14, 2003 1:40:22 PM US/Central
To: davidedwards@mac.com
Cc: Tina Romanella de Marquez
Subject: RE: Cost of new Dell/Linux cluster

Dr. Edwards:

Simply put, the article you read (I'm guessing on MacNN.com from your
email address, but possibly on Austin Business Journal's web site or one
of two others that picked up the misleading ABJ article) was grossly
incorrect. The $38M was the total for a five-year package that included
endowed professorships, building construction, research endowments,
increases in operational funding, and new equipment for computing,
storage, networking, and visualization. The retail value of all of the
equipment was about $14M, of which over half was discounted by the
technology partners. The cluster retail value was about $3M, again
discounted by over half by our generous technology partners, Cray and
Dell. In short, we paid less per processor than VaTech and about the
same price per Tflop.

We are now one of the leading academic advanced computing centers in the
nation. I assure you that we did not achieve this standing by spending
35x the cost of equipment (or misusing funds in any other way). In fact,
our total of donations and discounts from hardware vendors over the past
2 years is about $15M.

Thank you for your concern re: this issue. I hope that you will continue
to read about us and let us know if you have any concerns.

Regards,
John R. Boisseau

How do they figure they paid the same per Tflop as VT?

$3M for 3.7 Tflops

or

$5.2M for 17.6 Tflops

john r boisseau, no credentials or affiliation listed, is relying on the fact that much of the equiptment was heavily discounted or donated to achieve a comparable price performance point.

tflops / price = ratio
17.6 / 5.2 = 3.385
03.7 / 3.0 = 1.233

ut-austin would need
03.7 / 1.1 = 3.364

so someone gave them a free 2 million to equal vt as far as ratios go. i'd like to see the discounts given (and the bean counter's faces) if the cluster was anywhere near the size of the vt cluster.

of a larger package. Obviously, the draw wasn't just a supercomputer. I bet UT was looking to build a Institute for Computational Engineering & Sciences (ICES) anyway and found the most funding through Dell, which probably required them to purchase a Dell cluster. Not a cutting-edge cluster, but still a great deal overall, it seems.

just a note, i go to ut-austin and dell has been sticking their fingers everywhere. the entire business school was for all intents and purposes bought up by dell and i hear constant problems and headaches from the lab tech to have to run the computer labs. and about half of the people in the business school hate their mandatory laptops (also required to be dells, not just a pc) that they pay for out of pocket. it is curious though because the education college requires people to buy ibooks (also out of pocket)

John Boisseau is the director of TACC:

http://www.tacc.utexas.edu/general/contact/boisseau/

...but that's still beside the point. At retail, the Dell cluster costs almost 3 times as much per Tflop.

Dr. Boisseau points out that the $3M retail price was discounted by OVER HALF...which brings me to this:

We already know how razor thin Dell's margins are, maybe less so in servers, but certainly not anything near 50%. If Dell has discounted this by over half, to the point where UT can claim they're paying the same per Tflop as VT, then that means Dell is LOSING MONEY on this cluster. Yes, that's their perogative and all, but how is that honest? It smacks of the same kind of predatory tactics that MS used with IE. Sure, if Dell gave everyone a cluster for "free" then they'd all be "cheaper" than the alternatives; but don't you think there's some ulterior motive there? Even if it's an out-and-out charitable donation, there's STILL always ulterior motive...

FYI, the word was that VT really didn't get any discount from Apple on their equipment...

UT thinks they do everything better than any other university in the country, just look at the c*** they were dishing out before last week's game against OU. VaTech is kicking their butt just like OU did.

Cockroaches have proven to be a viable and robust platform for the past 350 million years. The technical specs are quite impressive:

Background info: The cockroach was introduced 350 million years ago. Since its initial introduction, the cockroach has proven to be extremely versatile, and has been implemented succesfully in home and office environments.

CPU: 400,000 neuron processors operating in parallel. The CPU is capable of several Petaflops and can process visual and audio signals simultaneously in real-time.

Static memory storage: 40 GB of DNA. Error-correcting mechanisms and massive redundancy ensure the integrity of the data.

OS: Regular upgrades of the operating system are made every few weeks. Through use of genetic algorithms in its regular updates, the cockroach platform able to resist most viruses and can adapt to changes in the environment.

Support: The cockroach platform is very robust. It is so robust in fact that ZERO support is required. The cockroach is self-configuring and self-correcting. In the 350 million years of the cockroach platform, not a single reboot has ever been required.

Cost: The cockroach is a very inexpensive platform and compares favorably with PC, Mac, and Unix platforms in terms of purchase and maintenance costs.

Proposed projects: An archive of all of human knowledge. See
http://people.advanced.org/~jaron/roach.html

"Even if it's an out-and-out charitable donation, there's STILL always ulterior motive..."

No doubt about it. One must always consider the funding source and what the patron stands to gain from the contribution. In this case it could be merely mindshare...but it sounds like Dell already has that at UT.

I won't proclaim to understand the motives behind such acts of "altruism", but I can almost guarantee that Dell will be receiving [u]something[/u] in return.

If you believe the cost is discounted by half, you must also believe in the tooth fairy and paying MSRP.

Considering both Apple and Dell will discount for publicity's sake, compare the "Do-It-Yourself" cost of the clusters.

From Apple.com: G5, 2Ghz, 4G RAM, Fibre Channel - $5849 each x 1100 = $6,433,900

From Dell.com: PowerEdge 2650, 3Ghz, 4G RAM, Fibre Channel - $5243 each x 500 = $2,621,500*

*The article mentions that cheaper PowerEdge 1700's would be used as well, but doesn't mention what the breakdown would be, so the dollar amount is high.

Not included are the networking and cooling costs, which of course would be cheaper for the Dell cluster with fewer machines.

It seems worthwhile to use G5's in clusters and it would be nice to really compare the two platforms apples to apples. Unfortunately this is next to impossible.

"From Dell.com: PowerEdge 2650, 3Ghz, 4G RAM, Fibre Channel - $5243 each x 500 = $2,621,500*"

Hmmm... I went to the Dell Higher Education institution site and configured a dual processor 3.06 Ghz PowerEdge 2650 with the above specs and it cosy over $10,000. This was with the free Red Hat Linux, but the dual optical Fibre channel. I don't know where you got your numbers, but something smells.

However, it is true they will also be using lower end 1750s.


- Maclectic

Only ignorant a******* and possibly Macletic go to a web site and price out a big purchase. What are you, stupid? Do you not know what how business works? Do you pay MSRP?

Sheesh, just because you are used to paying Apple-dictated price for decades does not mean the PC world is like that.

Maclectic - you are an idiotic zealot.

Can't really respond to hypocritical and senseless comments...

I suggest you read this: http://www.macdailynews.com/opinion_comments.php?id=P1943_0_2_0_C

Maybe you'll learn something. You also may find that alot of your "zealots" work for Dell and MS...


- Maclectic

UT is a pretty big Mac campus... DeLL has less influence there than you might expect... IBM is also big in Austin/UT... Remember, Michael flunked out, and I think the adminstration still holds it against him...

so are you saying dell doesnt dictate their prices? well if they dont then who does? and its called RESEARCH when someone goes to a website to dig in to the costs of thing. you must be a bush supporter too and still beleive that millions of pounds of weapons might be found that were an imminent threat to us. none yet guy. not some yellow cake uranium that was a huge forgery. back to teh point, dont be scared because someone calls a value or a bluf into account. if someone told me something i would research it too just to keep things in cgheck. so YOU sir are the zealot. so it would seem that "Only ignorant a******* and possibly TROLLS DONT go to a web site and price out a big purchase" as you obviously are scared to do. dont be scare dof the truth, live in it...

All the previous poster commented on was what pricing is publicly available. The higher ed price he found IS a discounted price. Certainly Dell discounted well below that, but that isn't public info. Since Apple certainly made a similar discount, a comparison of Apple's public higher ed price to Dell's public higher ed price is probably a good ballpark estimation of how the costs compare.

If you have the actual costs that Apple and Dell used in selling to the two universities, then by all means share that information. Don't accuse others of being ignorant or lazy because they haven't been able to find it.

The thing is, you can go to Apple's store and pretty much figure out what you are going to pay for it. This is likle Saturn, you can haggle until your face is blue but you ain't going to get a discount. With Dell and any other major vendors, sure you can go to the web site and do what you call "RESEARCH" but vendors almost always offer a discount to their listed price to their corp customers especially if they know you can go somewhere else to get what you need e.g. the other car dealers.

If someone in my company (major international tech comp) got a quote from a web site for a major purchase he/she would be laughed at. Who is scared of the truth???

cheap c***

Being an IS employee for a company with ~50,000 employees (admittedly, not all with computers) that uses Dells, I actually tend to disagree that the Dell website is a bad place to get pricing. It's not too bad for ball park estimates. The problem comes in when your company actually buys a 'custom' configuration, as I am sure UT did. When you have 300 computers 'custom' configured, it isn't really all that custom anymore, certainly not a one-off at least, and that is where Dell can save you some money.

As for Dell's razor margins, I think the server margins are way higher. Just like they are for all the other manufacturers. So they aren't necesarily losing a ton of money on this Cluster. Especially when you consider the PR value.


P.S.: Loved the cockroach quote. And the DNA encoding. Sounds like a great idea! ;^)

The UT guy--John Boisseau--said: "In short, we paid less per processor than VaTech and about the same price per Tflop."

Before he uttered that ridiculous statement, he must have been smoking his socks. According to another UT spokesperson, Tina Romanella de Marquez, as quoted in The Inquirer (http://www.theinquirer.net/?article=12114), the capital cost of the Dell cluster of 300 machines is $3 million. The capital cost of the 1,100 G5s at Virgina Tech is $5.2 million.

That makes the VT cluster less than HALF the capital cost per node and yet it delivers 33 per cent higher performance per node (16 gigaFLOPs per node vs. 12).

IOW, Virginia Tech is getting about 2.8 times the bang-for-the-buck with its G5-based system than UT is getting with its Dell-based system.

"... less per processor than VaTech and about the same price," my a$$... Whoever signed the dotted line on the UT deal should be hung out to dry...

interesting that you mention the capitol cost. IF people would read before they speak they would see that the 5.2 million dollars included the comps, racks cooling, etc. its all listed out. They were even projecting 10 tflops for the performance, and yes they will be running osx. which i thinhk is interesting because i could have sworn that infiband didn't have drivers for osx b4.

When did we trust the spokesperson instead of the person actually running the thing? Oh yeah, right, this is Macnn.... I forgot.

There is no itemized price for the Dell cluster. The deal UT made with Dell included other items and services such as support from Cray. Dell sold the cluster/service/support as a package and Dell will gladly tell you that the cluster only cost $10, but the service and support cost $37,999,990. We will never know the truth. Get over it.

-d*** Burnes

There is relatively little information available about the system. Based on my background of 15 years in supercomputing, I have written some impressions for readers to share and comment. I'll update the text when more details become available.

There are a lot of open questions, such as

How easy it is to port computational codes to this system?
How well the communications network scales in actual parallel applications?
How much work is required for maintaining the system in a production environment?
Is there support for parallel I/O? What is the performance when using, e.g., a NFS-mounted disk on several nodes simultaneously?
How well the cooling facilies and power supply work in the long run?

"...we paid less per CPU..."

He wasn't talking about the the price you'd pay. He was talking about the price he paid, meaning, including Dell's undisclosed discount.

"He was talking about the price he paid"

What part of "[according to] UT spokesperson, Tina Romanella de Marquez, as quoted in The Inquirer (http://www.theinquirer.net/?article=12114), the capital cost of the Dell cluster of 300 machines is $3 million" did you not understand?

To further show the apples to dell-oranges comparison.

The Dell cluster only has 1GB RAM per processor, and uses primarily 1U PowerEdge 1750's.

According to:

http://www.npaci.edu/Resources/Systems/cgi-bin/compute_all.cgi?arch=parallel

The TACC Cray-Dell PowerEdge Xeon Cluster contains 600 3.06GHz Xeon processors within 282 Dell dual-processor PowerEdge 1750 compute nodes, 16 Dell dual-processor PowerEdge 2650 compute-I/O server-nodes and 2 Dell dual-processor PowerEdge 2650 login/development nodes. Compute nodes have 2GB of memory; and the I/O nodes, as well as the login/development nodes, each have 4GB of memory.

Where the Dell cluster clearly surpasses the Apple cluster is in engineering and appearance.

How's that for irony.

"Where the Dell cluster clearly surpasses the Apple cluster is in engineering and appearance"

Huh? Appearance, maybe, but in engineering? I don't think so.

new math - MacNN.com Reader 128226
10/14/03, 10:26 pm EDT "interesting that you mention the capitol cost. IF people would read before they speak they would see that the 5.2 million dollars included the comps, racks cooling, etc. its all listed out. They were even projecting 10 tflops for the performance, and yes they will be running osx. which i thinhk is interesting because i could have sworn that infiband didn't have drivers for osx b4."

You beat me to it. I wish people on both sides would read before posting. UT's responses were somewhat convoluted and verbose, but when distilled down, we may never know what they paid for the hardware between discounts and grants. Methinks the Cray involvment caused them a few chuckles along the way though.

new math - MacNN.com Reader 128226
10/14/03, 10:26 pm EDT "interesting that you mention the capitol cost. IF people would read before they speak they would see that the 5.2 million dollars included the comps, racks cooling, etc. its all listed out. They were even projecting 10 tflops for the performance, and yes they will be running osx. which i thinhk is interesting because i could have sworn that infiband didn't have drivers for osx b4."

You beat me to it. I wish people on both sides would read before posting. UT's responses were somewhat convoluted and verbose, but when distilled down, we may never know what they paid for the hardware between discounts and grants. Methinks the Cray involvment caused them a few chuckles along the way though.