Virginia Tech Supercomputer Ranks #4 [updated to #3]

[MacRumors]

The New York Times (http://www.nytimes.com/2003/10/22/technology/22SUPE.html) provides preliminary numbers for the fully assembled PowerMac G5 Cluster (http://www.macrumors.com/pages/2003/09/20030902190056.shtml) which Virginia Tech assembled this month.

The preliminary numbers rank the G5 Cluster at the 4th in the world in performance with 7.41 trillion operations per second . Official results will not be reported until November, and Tech reports that they are "still finalizing their results and that the final speed number might be significantly higher."

The relative cost of the entire assembly is significantly less then previous solutions.

[idkew]

WOW.

That is a lot of operations per second.

Gotta like a Mac being number 4. Not quite number one, but 4 is nice, too.

[DrGruv1]

maybe 2nd?

[Kirtus]

I heard on your mac life that it would be number 2 when it was all said and done.

[Mr. G4]

What happen to #2.
I guess it's still in top5

[michael666]

The following article puts it on second place on the list:

http://www.wired.com/news/technology/0,1282,60821,00.html

The brand new "Big Mac" supercomputer at Virginia Tech could be the second most powerful supercomputer on the planet, according to preliminary numbers.

Early benchmarks of Virginia Tech's brand new supercomputer -- which is strung together from 1,100 dual-processor Power Mac G5s -- may vault the machine into second place in the rankings of the worlds' fastest supercomputers, second only to Japan's monstrously big and expensive Earth Simulator.

The Big Mac's final score on the Linpack Benchmark won't be officially revealed until Nov. 17, when the rankings of the Top 500 supercomputer sites are made known at the International Supercomputer Conference.

[XnavxeMiyyep]

I wonder whether this number will become higher in the rankings when Panther comes around. Or is this one of those things purely measured by the processor itself?

[hugemullens]

the number-4-baby department ? :) this is great news, cheap, ultra powerful, great time to be a mac user !!!

[Oirectine]

This is great news :)
And it is running OS X. I wonder how many of the top 10 super computers are running Windows...

[dukemeiser]

Now that's fast.

[arn]

Originally posted by michael666
The following article puts it on second place on the list:

http://www.wired.com/news/technology/0,1282,60821,00.html

The Wired article has very preliminary numbers. The NYTimes article is more accurate.

I wouldn't assume the rank will change. I don't know the spread.

arn

[MasterMac]

This is pretty good nonetheless :)

Hopefully the final numbers will place them a little higher

[Edot]

They are only a couple hundred billion off from 3rd.:D

[howard]

who's number 1?

[chazmox]

Isn't number one the Japanese Earth Simulator ( or some close approximation of that name.. )?

[QCassidy352]

yes, that is #1, and it's #1 by a lot.

[voicegy]

Earth Simulator Center in Yokohama, Japan, rated at 35.86 TFlops

http://www.top500.org/

[TrenchMouth]

From my knowledge the number one supercomputer is the earth simulator. i read an article about IBM getting the rights to build two new ones this year that will blow the earth simulator out of the water, but those are super huge and super expensive.

i think this is great for Apple. if this thing had placed in the top 100 it would be great. what this means is that the new G5s are great for running things in parallel fashion and that they scale incredibly well. the price performance ratio of this machine is off the charts. Kudos to VT for having the balls to put it together. so this thing has 2200 processors. anyone out there care to go for 4000? how about 6? you know they are thinking about it.

[MOM]

This is BIG news!! An article in the New York Times stating that one of the world's fastest supercomputers is made from Macs-for cheap, WOW! So many people will read this and think "Macs are cheap and fast." How often has that sentatment come from the mainstream press?

[Blaaze]

wow, i had no idea this big mac would be placed so high. It's pretty amazing comparing the costs and time to the big mac's counterparts.

big plus for apple. with newly released iTunes, and the supercomputer, their name is really being thrown around the market place.

[Powerbook G5]

So where can I pick up one of these babies? I wonder if I can get a good Edu discount on it :D

By the way...imagine all the free copies of Panther those guys are getting. Can you say eBay? ;)

[Rower_CPU]

Hmmm, they jumped up a couple of places since this thread (http://forums.macrumors.com/showthread.php?s=&threadid=42445).

Did a couple of those AlphaServers drop off the list?

[Sol]

4th fastest is not bad, especially when you look at the overall cost and the time that it was put together. Just imagine what will be possible in the same space when the G5 XServes become an option.

[ryanw]

Originally posted by Oirectine
This is great news :)
And it is running OS X. I wonder how many of the top 10 super computers are running Windows...

Uh, if I remember right, it's running Yellowdog Linux (http://lists.terrasoftsolutions.com/pipermail/yellowdog-announce/2003-June/000032.html) so Panther shouldn't help much in this situation ...

[MasterMac]

Originally posted by ryanw
Uh, if I remember right, it's running Yellowdog Linux (http://lists.terrasoftsolutions.com/pipermail/yellowdog-announce/2003-June/000032.html) so Panther shouldn't help much in this situation ... I'm pretty sure that the VT comps are running 10.2.7 not Linux, I'll try and find an article confirming it.

Edit; http://www.macrumors.com/pages/2003/09/20030903135718.shtml

[Rower_CPU]

Ha, and I heard it's been Panther beta from the start...probably Panther GM pretty soon, if not already.

[nickmcghie]

Originally posted by howard
who's number 1?

The Earth Simulator in Japan with a cost of USD$350 million compared to just over USD$5 million for the G5 cluster.

[AndrewMT]

Originally posted by Blaaze
big plus for apple. with newly released iTunes, and the supercomputer, their name is really being thrown around the market place.

Apple users really do live in their own world, don't they?

[Rower_CPU]

Originally posted by nickmcghie
The Earth Simulator in Japan with a cost of USD$350 million compared to just over USD$5 million for the G5 cluster.

Almost 5 times the performance and at 70 times the price. :eek:

Go Apple. :)

[Abstract]

Yeah, that's what I don't get. Why not just build a $25 million behemoth full of Dual G5's right now and blow that Earth Similator (its for predicting weather on a large scale, right?) out of the water by next month?

Sure this is great news, but it doesn't prove that $5 million worth of Intel Xeons or Opterons or whatever wouldn't be faster. Does anybody know how fast such an Intel system would be? Anyone want to take a wild stab in the dark?

[Edot]

Originally posted by Abstract
Yeah, that's what I don't get. Why not just build a $25 million behemoth full of Dual G5's right now and blow that Earth Similator (its for predicting weather on a large scale, right?) out of the water by next month?

Sure this is great news, but it doesn't prove that $5 million worth of Intel Xeons or Opterons or whatever wouldn't be faster. Does anybody know how fast such an Intel system would be? Anyone want to take a wild stab in the dark?

If you read the article, and by the looks of some of these posts many people haven't. The #3 computer is just that.

By contrast, the fastest cluster machine, the Lawrence Livermore system consisting of 2304 Intel Xeon processors, is capable of 7.63 trillion operations a second, at a price estimated at $10 million to $15 million.

However, I am not sure when this computer was built and whether the estimated prices are in current USD or when it was purchased.

[Potus]

Originally posted by TrenchMouth
i think this is great for Apple. if this thing had placed in the top 100 it would be great. what this means is that the new G5s are great for running things in parallel fashion and that they scale incredibly well. the price performance ratio of this machine is off the charts. Kudos to VT for having the balls to put it together. so this thing has 2200 processors. anyone out there care to go for 4000? how about 6? you know they are thinking about it.

Amen! It's the price performance ratio that is most amazing! Kudos to the students who worked for pizza!

[ryanw]

Originally posted by Abstract
Yeah, that's what I don't get. Why not just build a $25 million behemoth full of Dual G5's right now and blow that Earth Similator (its for predicting weather on a large scale, right?) out of the water by next month?

Sure this is great news, but it doesn't prove that $5 million worth of Intel Xeons or Opterons or whatever wouldn't be faster. Does anybody know how fast such an Intel system would be? Anyone want to take a wild stab in the dark?

I think origianlly they thought they WERE going to hit #2. But whats nice with cluster software is you could add another 1100 systems when you get more funding ...

[macnews]

The price is the big point here. You don't see any plain, off the shelve computers in the top list. What a great thing for Apple. Don't need to be the fastest but being the cheapest while still being competitive on speed should help all those MS trained IT people who want one OS culture realize the benefit of the mac.

[avus]

Originally posted by nickmcghie
The Earth Simulator in Japan with a cost of USD$350 million compared to just over USD$5 million for the G5 cluster.

I am p*ssed off with the comment like this (and Jack Miller at AtAT) for missing the point that the cost for The Earth Simulator INCLUDES the whole facility that protects the hardwares from earthquakes and lightenings. It is a very elaborate system, and it is not fair to compare this with just the cost of the G5s at VT. Go read about it:

http://www.es.jamstec.go.jp/esc/eng/ES/facilities.html

Plus, The Earth Simulator was a 6-year-old project that finally became operational last year. Its root is ancient compared with the current computers - YET NOBODY HAS BEATEN IT. Go figure:rolleyes:

Edited: spellings

[alandail]

#4 fasted in the world, far less expensive than the others in the list, built in a matter of months. This will get the attention fo the whole supercomputer design world.

And it'll put Apple on this page as clearly the least expensive option on there -the intel machine at #3 was build last year for $10 to $15 million.

http://www.top500.org/lists/2003/06/top5.php

It's a top 500 list and on the first try, an Apple system hit the top 5.

Here's a quote from the article "Word of the low-cost supercomputer, put together by faculty, technicians and students at Virginia Polytechnic Institute, is shaking up the esoteric world of high performance computing, where the fastest machines have traditionally cost from $100 million to $250 million and taken several years to build"

Steve should have fun with this.

[acj]

The price comparisons floating around are retarded, literally. In 18 months a computer twice as fast at 90% of the cost can be built.

[Wardofsky]

Hmm, 1100 G5's is nice but I wouldn't mind it when they get even faster, this supercomputer is one very savvy thing.
What was the cost of the while thing again?

[neutrino23]

Originally posted by avus
...it is not fair to compare this with just the cost of the G5s at VT. ...

The price for the VT system also includes the cost of the cooling systems and such. Not sure what else it includes. It is not just the cost of the G5s.

Originally posted by ACJ
The price comparisons floating around are retarded, literally. In 18 months a computer twice as fast at 90% of the cost can be built.

Certainly faster, maybe not cheaper. I'm hardly an expert, but my impression is that until now most of the fastest systems were highly customized (like the
#1 system) and highly customized means expensive.

I think this is the first system to reach the top ten at such a low cost and so quickly using off-the-shelf computers, not just readily available processors.

[MacSlut]

What a crock, it's only 2GHz, heck my PC is faster at 2.8GHz, you don't see me getting any articles about being in the top 5!

Just kidding, couldn't resist.

[MacBandit]

Originally posted by michael666
The following article puts it on second place on the list:

http://www.wired.com/news/technology/0,1282,60821,00.html

If i remember right the wired article published 17 teraflops.

In my opinion the NYTimes is notorious for publishing a new article but taking old data from previously published articles.

That being the case I wonder if when they were digging for there data they forgot it was 17 and not 7 and then did some research and determined that would be number 4.

I guess we'll know for sure in the next few weeks.

[macphoria]

I imagine this will show up as Apple TV ad some time in November?

[Wardofsky]

Originally posted by macphoria
I imagine this will show up as Apple TV ad some time in November?

Hmm, well if one G5 can blow one man through many layers of wall, I suspect 1100 will eith blow all the students out and perhaps destroy the entire school...

[MacBandit]

Originally posted by Wardofsky
Hmm, well if one G5 can blow one man through many layers of wall, I suspect 1100 will eith blow all the students out and perhaps destroy the entire school...

It's a WMD!!!

It could turn a small country into a smoking crater.

:D

[Potus]

Originally posted by MacBandit
...
In my opinion the NYTimes is notorious for publishing a new article but taking old data from previously published articles.
...


...and your opinion is based on what? And comparing the New York Times to which other daily?

[sethypoo]

I think I want one.:D :D :D :D :D

[bt9]

Cluster Supercomnputers lose a huge performance boost as you add more nodes.

So, No, making a supercomputer with 2200 G5s would not blow anything out of the water, it would probably score a teraflop or so above the current 1100 node cluster.



Plus, i think avus is absolutely right, the Japanese earth simulator is not comparable, the cost of the surrounding infrastructure which was factored into the price is far in excess of any equivalent in the G5 cluster (such as cooling).

Having said all that, it's still cool, I imagine when the G5 Xserve comes out a lot more Colleges are going to be thinking about this idea.

[MacBandit]

Originally posted by Potus
...and your opinion is based on what? And comparing the New York Times to which other daily?

It's based on reading about something and then 2 weeks later the NYTimes posts an article about the same thing and some of the facts are obviously from the previous article I read. No, the previous articles are not from dailies.

[Potus]

Originally posted by MacBandit
It's based on reading about something and then 2 weeks later the NYTimes posts an article about the same thing and some of the facts are obviously from the previous article I read. No, the previous articles are not from dailies.

Which articles? About what? Can you site an example? If it's from professional journals it's not unusual for print media to publish later nor for that publicatin to be the source for the news article. Which daily would you recommend for up to the minute reporting and accuracy?

[MacBandit]

Originally posted by Potus
Which articles? About what? Can you site an example? If it's from professional journals it's not unusual for print media to publish later nor for that publicatin to be the source for the news article. Which daily would you recommend for up to the minute reporting and accuracy?

Sorry I don't have specifics. That's why I said it was in my opinion. If I could remember the specifics I would have stated it as fact and posted the instances.

I don't have any specific dailies. One of the main places I get news is right here in the MacRumors forums. There are lots of posts about new science discoveries and technology. It just seems to me that some non-mainstream site will publish a news article (which I read from a link here or at a handful of other sites) and then a few days or a few weeks later the NYTimes publishes the same thing just written differently and often missing facts.

[Telomar]

Originally posted by MacBandit
That being the case I wonder if when they were digging for there data they forgot it was 17 and not 7 and then did some research and determined that would be number 4. The 17 teraflops was Rpeak. The 7.5 teraflops is Rmax, which is what they rank on. The supercomputer was always destined for around 3rd or 4th.

[MacBandit]

Originally posted by Telomar
The 17 teraflops was Rpeak. The 7.5 teraflops is Rmax, which is what they rank on. The supercomputer was always destined for around 3rd or 4th.

Thank you I had forgotten about the other benchmark numbers numbers.

[Belly-laughs]

Originally posted by nickmcghie
The Earth Simulator in Japan with a cost of USD$350 million compared to just over USD$5 million for the G5 cluster.

The proof that Macs aren´t expensive. And what´s the maintenance cost? Close to nothing compared to the other top-ten I would assume.

[Bakey]

Originally posted by ryanw
Uh, if I remember right, it's running Yellowdog Linux (http://lists.terrasoftsolutions.com/pipermail/yellowdog-announce/2003-June/000032.html) so Panther shouldn't help much in this situation ...

If I remember rightly the US Navy had ordered a number of Macs [Possibly Xserves, I can't remember?!] to which it was quoted they wouldn't be installing X and indeed would be whacking one of the many flavours of Linux on there...

PS. This is my FIRST post... whoooh!! ;)

[edesignuk]

IMO this is a huge achievement for both Virginia Tech & Apple. To come in at #4 with these preliminary results is amazing for a supercomputer costing just US$5m!!! It is great PR for Apple and the G5 PowerMac that the CPU, and in fact the entire machine that is available for everyone to buy and have on their desktop at home, can hold it’s own in the supercomputing world.

[yamabushi]

I think it is great that a Mac cluster is likely to be in the top 5. I would be even happier if many groups started making smaller clusters of Macs and knocked a bunch of those HP clusters off the list. It's great to be on the list, but claiming just one out of five hundred spots isn't so great. I guess it's nice that quite a few are made by IBM.

[Catt]

I really like the sound of "The Earth Simulator"
don't know why just sounds impressive.
The G5 cluster sounds like a good way to go in many ways. I think Apple should concern themselves with producing medium power Supercomputers to allow research projects to utilise them. A new generation of more affordable but still powerful SCs could push science IN GENERAL forward a little bit faster. But there will be areas, such as weather prediction, that will require the sheer power of SCs like the Earth Simulator.

[biscuit]

I read somewhere (mebbe on MacRumors even) that this wasn't the full 2200 processors but 2100 or so. Still, a hand-full extra won't take it to number 3. Maybe after a few more optimisations?

Still, great to be top 5, top 3 would've been the icing on the cake :)

biscuit

[crenz]

Well, seems that people are still comparing the Virginia cluster to the old list (June). Let's see how well the other supercomputers that have been added in the meantime perform... I thought IBM also wanted to launch a supercomputer that they think will make top 5?

[AidenShaw]

Originally posted by TrenchMouth
what this means is that the new G5s are great for running things in parallel fashion and that they scale incredibly well. the price performance ratio of this machine is off the charts.

Scaling on a massive cluster is largely a function of the network interconnect, it needs to be very fast and very low latency.

The VT cluster uses InfiniBand, an Intel-developed high speed interconnect fabric (http://www.intel.com/technology/infiniband/).

The Xeon (and Alpha) clusters in the top 500 list are using slower interconnects, which results in lesser efficiency and poorer scaling.

Y'all should prepare yourselves for Xeon and Opteron clusters using the InfiniBand networking.... You'll see that the network is what's special about VTech's cluster, not the CPUs.

[telos]

barely half the speed it was expected to go at. huge disappointment :(

[Telomar]

Originally posted by telos
barely half the speed it was expected to go at. huge disappointment :( It's around the expected speed just nowhere near the peak numbers. They were expecting a touch more but if these are preliminary numbers that will come.

[jamilecrire]

See slashdot article
http://www.apple.com/ibook/

[benoda]

why is no one talking about the new iBook G4's? I have a friend who's about to buy a new laptop and can't decide between the 12" powerbook and the new iBooks...!

[jamilecrire]

Originally posted by benoda
why is no one talking about the new iBook G4's? I have a friend who's about to buy a new laptop and can't decide between the 12" powerbook and the new iBooks...!

Tell your friend not to be stupid, get the iBook. I have a G3 900MHz and it's every bit as fast as my best friends 867MHz 12" Powerbook (and it cost about $500 less at the time). Now that it has a G4 why bother with the 12" Powerbook? The only real difference is the Powerbook keyboard is nicer.

This is a prelude to G5 Powerbooks by the end of the year in my opinion (maybe after the first of the year).

[Potus]

Originally posted by MacBandit
Sorry I don't have specifics. That's why I said it was in my opinion. If I could remember the specifics I would have stated it as fact and posted the instances.

I don't have any specific dailies. One of the main places I get news is right here in the MacRumors forums. There are lots of posts about new science discoveries and technology. It just seems to me that some non-mainstream site will publish a news article (which I read from a link here or at a handful of other sites) and then a few days or a few weeks later the NYTimes publishes the same thing just written differently and often missing facts.

Got it. Although not immune from error (duh.) the great dailies (The Wall Street Journal, The New York Times, The Washington Post, The LA Times) usually do not print rumors. Although there is the "unnamed sources" stuff. Moreover there is an editing process that goes on in any newpaper which results in not presenting the unmediated/unedited "facts". Although, recently, at least in the political arena. the fact that there are "rumors" has actually become news so that the press begins reporting on the fact that there is rumor and that becomes a news story. I recall, the later disproved, "ossuary of the brother of Jesus" and the "Bill Clinton murderd Vince Foster" stories. I'm sure that there are other rumors and facts that don't get published immediately by the respectable press. Of course, I read online just this week that review from the WinDoze people on iTunes was mixed: I guess that was true--only 87% loved it. :) :)

[Tim Flynn]

I hope they tweeked it for the final numbers. If they get to #3 spot, they beat all the Intel machines.
The only machines above it are then IBM and Earth Sim.
Apple wins, IBM wins & OS X wins !
Any of the top 100 running Windows?

[Rincewind42]

If you look at the Rpeak vs the Rmax of all the other supercomputers in the Top 5, you'll notice that they are WAY closer to their Rpeak than the score for the VT Cluster is. For comparison the current Top 5 are at 87.5%, 67.7%, 69%, 59.4%, and 73.1% efficiency. In fact, in the Top 40, none of them go below 45% efficiency. The VT Cluster, given these numbers, is at 42% efficiency.

I'm willing to bet that this number comes from one of the first times they've run the benchmark on the cluster as a whole. They haven't shaken all the bugs out yet, and the certainly haven't optimized the benchmark for the cluster as a whole if this is true

Assuming the VT Cluster follows the trend of it's fellow super computers in the Top 40, it should perform at at least 7.92 Tflops once all the bugs are shaken out - and that is still less than 50% efficiency. If they can get efficiency to the level of the Top 5 computers, it should post over 10 Tflops. Personally, I think that they can easily move up to 3rd and score over 45% efficiency - 3% is well within the limits of optimizations.

The real question to ask is what other Super Computers have been built and are in the running for Top 5 spots this year...

[Rincewind42]

Originally posted by bt9
Cluster Supercomnputers lose a huge performance boost as you add more nodes.

So, No, making a supercomputer with 2200 G5s would not blow anything out of the water, it would probably score a teraflop or so above the current 1100 node cluster.

Well, the VT cluster posted some very prelim results using 128 nodes and was projected at only 7.055 Tflops when complete. With all 1100 nodes it actually scored (prelim) 7.410 Tflops. So it looks like extra nodes DID help a lot. But as AidenShaw pointed out, it is often more a function of the communications fabric than the number/speed of the CPUs themselves (this of course depends on what the cluster is doing - if it's a massively parallel job then the com fabric may end up idle for most of it).

If they did double the size of the cluster, it probably could challenge the #2 slot and win it.

Edit: Oops, wrong number of nodes in the final cluster.

[edesignuk]

Originally posted by jamilecrire
See slashdot article
http://www.apple.com/ibook/
Originally posted by benoda
why is no one talking about the new iBook G4's? I have a friend who's about to buy a new laptop and can't decide between the 12" powerbook and the new iBooks...!
Why are either of you brininging this up in a thread about the Virginia Tech Supercomputer!?! :rolleyes:
Please stay on topic :)

[sushi]

Originally posted by Belly-laughs
The proof that Macs aren´t expensive. And what´s the maintenance cost? Close to nothing compared to the other top-ten I would assume.
What is really cool about this system, is after a year or two when the G5 chips are much faster, they can upgrade the entire system easily by just purchasing new G5 Macs. At the same time they can sell off the old ones offsetting the cost of the upgrade somewhat. Sweet.

Sushi

[fourthtunz]

So the final #'s are due in November, should we expect them to be much higher?
Sure there are lots of processors coming that will be even faster but for now if it turns out that the Mac is #4, I think that's quite good, being that it is the first Mac on the list?
6 months from now should be even more interesting
:D
daniel

[dongmin]

Originally posted by Rincewind42
Well, the VT cluster posted some very prelim results using 128 nodes and was projected at only 7.055 Tflops when complete. With all 2200 nodes it actually scored (prelim) 7.410 Tflops. So it looks like extra nodes DID help a lot. actually I think the original 128-node sample put it at 80% of the theoretical maximum of 17 Tflops. That's where the 14 Tflop figure came from (17 x 0.8 ~= 14). So this latest result of 7.4 is drammatically lower than that earlier projection. As another have noted 42% efficiency is A LOT worse than the other top 5. So while the cost-performance ratio is excellent, it's still a little disappointing.

[MacBandit]

Originally posted by dongmin
actually I think the original 128-node sample put it at 80% of the theoretical maximum of 17 Tflops. That's where the 14 Tflop figure came from (17 x 0.8 ~= 14). So this latest result of 7.4 is drammatically lower than that earlier projection. As another have noted 42% efficiency is A LOT worse than the other top 5. So while the cost-performance ratio is excellent, it's still a little disappointing.

The 7.4 teraflops number you are tossing around is indeed the maximum but it is not the peak. The peak is around 17 teraflops so everything is scaling exactly as they expected.

[sketchy]

Originally posted by Potus
...and your opinion is based on what? And comparing the New York Times to which other daily?

well -- they did have a little incident where they had a writer making up stories and an editor who was changing statistics...

[AidenShaw]

Originally posted by MacBandit
The 7.4 teraflops number you are tossing around is indeed the maximum but it is not the peak. The peak is around 17 teraflops so everything is scaling exactly as they expected.

The "peak" number is a mathematical calculation made by looking at the spec sheets for the chip (not even the system) and multiplying.

The "maximum" is what the system actually is measured at.

The "peak" number isn't useful, but "max" is. That's why the Top500 are ranked by "max", not "peak".

Your statement more or less implies that "peak" is the useful number, was that what you meant?

[MacBandit]

Originally posted by AidenShaw
The "peak" number is a mathematical calculation made by looking at the spec sheets for the chip (not even the system) and multiplying.

The "maximum" is what the system actually is measured at.

The "peak" number isn't useful, but "max" is. That's why the Top500 are ranked by "max", not "peak".

Your statement more or less implies that "peak" is the useful number, was that what you meant?

No just an uneducated statement. Thanks for clearing me up on the what RPeak and RMax mean.

[legion]

As pointed out by an earlier poster, everyone seems to be comparing the new VT cluster to the June list. Of special note in the June Top 500 list is the information concerning ASCI White (The SC currently being referred to as the #5 machine in this thread.) It needed to be retested in June and based on credible information possesed at that time, it should rise in rankings to #3 or #2, thereby pushing the Apple VT to, at the highest, a #5 ranking.

For the June list, they weren't able to retest the ASCI White because it had too many projects scheduled for it to place any time to the side to run LinPack numbers. (Go figure, a SC being used for it's intended purpose rather than running tests for bragging rights...) However, at that time, a similar SC with a smaller build than ASCI White was retested with the new LinPacks and rose dramatically in the rankings even above its "big brother" ASCI White. Obviously this can't be true which is why the June list had statement qualifiers about White on its ranking lists. In the last few months, there's a good chance that White has been retested (I think the expected numbers were in the mid 8Tflop range.)

[Potus]

Originally posted by sketchy
well -- they did have a little incident where they had a writer making up stories and an editor who was changing statistics...

yes. the reporter and the managing editor resigned. I'm sure this was the only paper in recorded history where it happened.

[Frunobulax]

The performance in now close to 8.2 TFlops. Does anybody know how often this
file (http://www.netlib.org/benchmark/performance.pdf) is updated?

[Rower_CPU]

Originally posted by bt9
Cluster Supercomnputers lose a huge performance boost as you add more nodes.

So, No, making a supercomputer with 2200 G5s would not blow anything out of the water, it would probably score a teraflop or so above the current 1100 node cluster.



Plus, i think avus is absolutely right, the Japanese earth simulator is not comparable, the cost of the surrounding infrastructure which was factored into the price is far in excess of any equivalent in the G5 cluster (such as cooling).

Having said all that, it's still cool, I imagine when the G5 Xserve comes out a lot more Colleges are going to be thinking about this idea.

Do you have some sort of formula that expresses this, with diminishing returns after a certain point?

Because that's certainly not borne out by the data in the Linpack scores in the PDF in this thread (http://forums.macrumors.com/showthread.php?s=&threadid=42445).

A 256 CPU G5 cluster gets 861 GFlops at about 3.4 GFlops/CPU, while the 2112 CPU VT cluster gets 8.16 Tflops at about 3.86 GFlops/CPU.

What gives?

[tychay]

Originally posted by Telomar
The 17 teraflops was Rpeak. The 7.5 teraflops is Rmax, which is what they rank on. The supercomputer was always destined for around 3rd or 4th.

BTW, Even the Earth Simulator was around 80% efficiency so it was expected to drop off to around 60% efficient (not the 40% efficiency of the quoted number which even the article noted was preliminary). Nobody has an InfiBand cluster on the top500 before so we don't know what to expect. In any case, this is all based on a Wired article where the author could be accused of some wishful thinking.

In any case, the New York Times data was probably old. This isn't Apple here so I went to the top500 maintainer's website and downloaded the latest PDF benchmarks (http://www.netlib.org/benchmark/performance.pdf). Using Panther you can search for "G5" and find the Virgina Tech cluster on page 53. RMax is 8.164 TFlops (as of today).

However, that while that would put it in 3rd place on the June list. It'll be in 4th place in the November one. However, the good news is that the new 3rd place is a 1900 processor Itanium system and the Apple machine has beat out all Pentium clusters in performance, price/performance and efficiency.

Take care,

terry

[Potus]

Originally posted by tychay

In any case, the New York Times data was probably old.
Take care,

terry

True. What I like about the NYtimes tech writing is that they report Apple news with regularity. The other paper I subscribe to (which shall be nameless...and deservedly so) reports Apple news rarely. In fact their tech editor seems to be actively hostile, if the tone of his writing is any indication. But let Dell put out dog poop and he raves about the shape and smell...

[nek]

The Virginia Tech supercomputer will rank high and very high if you consider the cost, but it will only place in 4th if you compare it to last year's list. There will be other computers which are new to the list.

For example, in june IBM said they were building a computer for Japan's National Institute of Advanced Industrial Science and Technology (AIST) which is supposed to have an rPeak of 11.168 Tflops. I don't know if it was completed and tested, I'm just saying that there may be other supercomputers competing for the top 5.

When IBM finishes the Blue Gene/L supercomputer, next year I think(?) it will put the whole list to shame with its rPeak of 367Tflops. It has the obvious advantage of having 130000 Power5 processors though. Its hard to say what the Rmax will be, but it won't have any problem claiming the #1 spot.

[MacBandit]

Originally posted by Potus
Which daily would you recommend for up to the minute reporting and accuracy?

I previously said I don't have a recommend but I was wrong. I just forgot because it's not a daily news site I check. They send me news alerts.

www.Forbes.com

You can get free email alerts. I'm set up to receive anything that has Apples name in it and a number of other specific topics.

It's really great up to the minute accurate news from what I have seen.

[Potus]

Originally posted by MacBandit
I previously said I don't have a recommend but I was wrong. I just forgot because it's not a daily news site I check. They send me news alerts.

www.Forbes.com

You can get free email alerts. I'm set up to receive anything that has Apples name in it and a number of other specific topics.

It's really great up to the minute accurate news from what I have seen.

Thanks. I'll look into it.

[tychay]

Originally posted by nek
The Virginia Tech supercomputer will rank high and very high if you consider the cost, but it will only place in 4th if you compare it to last year's list. There will be other computers which are new to the list.

For example, in june IBM said they were building a computer for Japan's National Institute of Advanced Industrial Science and Technology (AIST) which is supposed to have an rPeak of 11.168 Tflops. I don't know if it was completed and tested, I'm just saying that there may be other supercomputers competing for the top 5.

Other than the Earth Simulator very few computers are 80% efficient (Rmax/Rpeak) so the IBM AIST computer will not threaten Virgina Tech's--that 16.9 TFlops Rpeak means that the Rmax can continue to inch up.

As for other computers appearing on the list, my previous post has addressed it: the only one new one that competes with Apple is HP's Itanium system. Traditionally large computers such as those have taken many years to set up so it is very predictable.

If you look at the list today (http://www.netlib.org/benchmark/performance.pdf), you'll see that the VATech cluster has jumped to an RMax of 9.55 Tflops from 8.17 earlier and 7.5 in the original post! That's approaching a 60% efficiency.

This is significant because it places the Apple cluster 3rd on the list, ahead of the HP cluster. Now, only the ASCII Q, LANL's 8160 processor Alpha system, and the Earth Simulator, a custom NEC monster, are more powerful! Too bad that this probably didn't make the Fall Top500 but very good news indeed and probably deserves some reporting (Arn?).

I'm inclined to believe that Infiniband opened the door so that a small university could build a top10 supercomputing cluster in a short time at a small cost. Virginia Tech jumped on it and Apple happened to be the best price/performance with a delivery date within the timeframe to make the Top500 (in fact, it looks like they may have cut it too close).

Now that the cat is out of the bag, we may be witnessing a sea change toward clusters based on PC CPUs over the next couple years. 970-based will still have the edge over the Itanium, Pentium Xeon, and Opterons because it's Rpeak is so large and there is little doubt that IBM be hitting their target speeds, but I don't expect these servers to necessarily be coming from Apple.

The Linux blade and rack vendors who supply datacenters and renderfarms should seriously be looking at the 970 to complement their Opteron and Itanium systems.

Looks like IBM has a winner of a chip in the HPC market.

Take care,

[revenuee]

Originally posted by ryanw
Uh, if I remember right, it's running Yellowdog Linux (http://lists.terrasoftsolutions.com/pipermail/yellowdog-announce/2003-June/000032.html) so Panther shouldn't help much in this situation ...

YDL!, really? that sucks, definitely not my first choice, ran really slow compared to to OS 9 and OS 10 when i tried it...

i know were talking about a completely different setup here, not to mention that a single computer in the cluster blows my machine out of the water, but still.... oh well... what do i know

[Rower_CPU]

This PDF (http://don.cc.vt.edu/tcfslides.pdf) says it's running OS X, just not which version. I'm still trying to dig up a reference to it running Panther beta...

[revenuee]

Hey Rower_CPU, hows the YDL working for you?

[Rower_CPU]

Originally posted by revenuee
Hey Rower_CPU, hows the YDL working for you?

Well, I installed it...checked it out for a little while, and then wiped the drive and installed Panther. ;)

It seemed to work OK, but I wanted to get a system back on my laptop I would actually use. Since I'm still in the middle of a semester I don't have a ton of time to play around with that stuff right now. But I did get it working...

[alandail]

9.555 tflops - wow - now it's clearly #3 in the world. Anyone know the rules for the top 500 list? I know there was a deadline for having the system up an running for the next list. But what is the deadline for the benchmarks used for the official rankings? It already is only the 3rd supercomputer to break 9 tflops - I wonder if they can still tweak it enough to break 10 tflops.

What a coup for apple. The first Mac based super computer blows em all away for a fraction of the cost - you have to think it's just a matter of time before the list is dominated with Apple supercomputers. Looks to me like a larger supercomputer with 3 GHz G5s could challenge the earth simulator for #1.

The G5 supercomputer is so fast now that it's Rmax is faster than the Rpeak values for all but 9 supercomputers. Another optimization that gets Rmax to 10241 cuts that number to 6.

[alandail]

the best thing about the new list is that the G5 system now outperforms the titanium system every way you look at it. Rpeak and Rmax are both considerably higher. And Rmax per CPU is higher as well. And cost/performance blows them all away.

And it has definitely been confirmed that they are using MacOS X on these machines. I don't think any of the supercomputers are using Windows. I think Apple wins the scalability argument now.

[pdickins]

[legion]

It looks like IBM's POWER processors (not the PowerPC processors) and AMD processors are destined to dominate the Top 20 for the next year (all you have to do is review all the current projects from Blue Gene/L from IBM, Cray's Red Storm, or HP's systems for Japan.)

Apple had/is having its day in the sun, but there are some serious computer about to go online. (Plus, the real story of the VT cluster is Infiniband and its first implementation which is now infiltrating other SC projects; even replacing old connects for projects like the ASIC colour series with Infiniband could propel them back to the top level status.)

[alandail]

#1 - cost $350 million
#2 - cost $200 million
#3 - cost $5 million

Don't underestimate the impact of this. Systems planned may not currently be G5 systems, but you have to think people building these things will have to sit up and take notice. And the connections can only do so much to speed things up - the Rmax value of the G5 system is faster than the Rpeak values for all but 9 supercomputers, and a little more tweaking would get that number down to 6.

[fourthtunz]

Originally posted by legion
It looks like IBM's POWER processors (not the PowerPC processors) and AMD processors are destined to dominate the Top 20 for the next year (all you have to do is review all the current projects from Blue Gene/L from IBM, Cray's Red Storm, or HP's systems for Japan.)

Apple had/is having its day in the sun, but there are some serious computer about to go online. (Plus, the real story of the VT cluster is Infiniband and its first implementation which is now infiltrating other SC projects; even replacing old connects for projects like the ASIC colour series with Infiniband could propel them back to the top level status.)

Sour Grapes= bad whine:(
Right now the first Mac ever is in the top five at 1/50th(or less) the cost of the others. Does anybody expect it to stay there forever?
I don't think so. Will the systems you are quoting beat the price/performance of the Mac? Maybe
So I guess So your point is that things change in the computing world? Thanks:rolleyes:

[Potus]

Originally posted by fourthtunz
Sour Grapes= bad whine:(
Right now the first Mac ever is in the top five at 1/50th(or less) the cost of the others. Does anybody expect it to stay there forever?
I don't think so. Will the systems you are quoting beat the price/performance of the Mac? Maybe
So I guess So your point is that things change in the computing world? Thanks:rolleyes:

good point