Los Alamos to Use AMD's Opteron in Linux Clusters

nuke-alwin writes "eWeek is reporting that Los Alamos National Laboratory announced it will use more than 3,300 Opteron chips in two of its Linux clusters. According to the article 'The key to Opteron, as it tries to gain traction not only against Intel Corp.'s 64-bit Itanium chip but also its 32-bit Xeon offerings, is its ability to run both 32-bit and 64-bit applications equally well.'"

Comments

The key

[mao che minh]

Not only can the Opteron power both 32-bit and 64-bit, but it also performs just as well as the Itanium in either environment. The Opteron is also far cheaper (especially when you compare the costs of 3,000 Opterons to 3,000 Itaniums, as most potential customers will).

Intel can't compete with the Opteron on merits alone. It will be interesting to see what they try next.

Re:The key

[Jason1729]

It will be interesting to see what they try next.

Maybe claim that since they created the x86 platform, the Opteron contains their IP and any Opteron users owe them $750/CPU in licensing.

Jason
ProfQuotes

Re:The key

[jerkychew]

But Intel has something in the server market that AMD doesn't: a reputation. AMD is just breaking into this market, and we'll see how many companies jump on the bandwagon this early.

I've worked for enough good-sized companied to know that a difference of a few thousand (even a few hundred-thousand) dollars isn't as important as reliability when you get into enterprise-level systems. The old saying, 'nobody ever got fired for buying IBM', rings true for Intel as well.

Not to mention that Intel's profit margin is HUGE when it comes to the server market - such is the luxury of being the only x86 gig in town. Intel can stand to slash prices dramatically and still turn a profit. Remember how they almost bankrupted AMD a few years back when AMD promised its 6x86 chips would always be 25% cheaper than a comparable Pentium.

Yeah, this will be very interesting.

Re:The key

[javiercero]

" The old saying, 'nobody ever got fired for buying IBM', rings true for Intel as well."

Are you for real? LOL, Intel entry in the server market was as cheap low end, nothing to do with "reliabitlity" or anything. All that intel has in corporate accounts is: well we are cheaper than sun. Now there is someone saying, well we're cheaper than Sun AND Intel. Intel is SOL, this whole new representation of Intel as quality server stuff is laughable. LOL!

Re:The key

[binaryDigit]

Not only can the Opteron power both 32-bit and 64-bit, but it also performs just as well as the Itanium in either environment.

Oh really, the Opteron performs just as well with natively compiled 64bit apps? Do you have links to 64bit benchmarks that show this. Everything I've seen shows I2 ahead in integer performance and way ahead in floating point. Don't know if the Opteron #'s are for "native" 64bit code though (and what difference it would make if it were, but it's still nice to do a apples/apples comparison).

Re:The key

[AvitarX]

I don't know the numbers, but the opteron should to quite well in floating point (by x86 standards at least) sinc ehtey added those extra SSE thingies. They left floating poit alone and recomend against it in favor of SSE, but it should give them a big boost in stuff written correctly.

I don't know how the IA64 compares to x86 though.

Re:The key

[jerkychew]

Have you ever had to design a Windows-based network from scratch? Have you ever had your boss come to you and tell you you have $100K to spend? I have.

No, it's not Sun. That's why I said in my original post that Intel is the only gig in town for x86, not for all servers. But, guess what? Not everyone wants a Sun solution. Tell me how you're going to build an IIS server farm on Sun machinery. I'd love to hear your solution.

Go price out an eight-way Dell 8450 server. Then tell me about low-end.

Re:The key

[drinkypoo]

Intel is trying to move into the space where you have powerful central computers. AMD will sell nicely into the cluster market, which sells more parts at a lower cost per part. After all, if an individual machine fizzles, you don't lose all that much compute time. This will give AMD a chance to build more of a positive reputation in the post-K6 age. :)

Re:The key

[evilviper]

But Intel has something in the server market that AMD doesn't: a reputation.

And what do you think a reputation really counts for exactly??? Especially in the "commodity hardware" market?

Intel's specs can't come anywhere close to AMD in ANYTHING. AMD has Intel beat in: price, performance, heat, power, maximum operating temperatures, and (lower) heat output.

we'll see how many companies jump on the bandwagon this early.

Lots of companies jumped on the Linux bandwagon quickly, and the operating system is really more crucial than the hardware. If companies can make that switch, a little jump to Opteron is nothing.

I've worked for enough good-sized companied to know that a difference of a few thousand (even a few hundred-thousand) dollars isn't as important as reliability when you get into enterprise-level systems.

Really? First of all, what evidence do you have that AMD is ANY less reliable than Intel? Secondly, the reliability of an individual component isn't anywhere near as important in a cluster... If it was, you wouldn't see even Intel processors being used, you'd only see Alphas, PPCs, Sparcs, etc.

Besides, I would hardly call just about ANY x86 system an "enterprise-level" system.

Not to mention that Intel's profit margin is HUGE when it comes to the server market

What does that have to do with anything? Not like your enterprise is going to care which company is making more money. Intel can't take enough of a loss for a long period of time to be competitive with AMD.

Re:The key

[Dub Kat]

quote:
I've worked for enough good-sized companied to know that a difference of a few thousand (even a few hundred-thousand) dollars isn't as important as reliability when you get into enterprise-level systems. The old saying, 'nobody ever got fired for buying IBM', rings true for Intel as well.

This is definitely right, but I wonder if the name-brand is worth it if they're paying twice the amount for Intel vs. AMD.

The article says Linux Networx got the contract for just under $10 million, and will deliver 2816 Opterons.

Now, I've no idea how much the hardware costs them, but let's assume $8 million; the other $2 million goes towards salaries/profit.

$8 million / 2816 Opterons = $2850.91 per Opteron.

Compare that to the PowerEdge 3250 (Itanium) from Dell: $6229 per Itanium.

I'm sure Intel has quite a mark-up on Itaniums and could get the price lower when trying to win a contract, but AMD is in the same position with the Opteron as well.

Either way, a nice win for AMD, as it seems that the Opteron really is the best x86 processor at the moment.

Colocated Linux Server - $60/mo

Re:The key

[javiercero]

Who in their right mind implements an IIS farm, and spends $100K to boot, jeezus!

"Go price out an eight-way Dell 8450 server. Then tell me about low-end."

Go and price a nice sun fire, or a nice IBM regatta and get back to us kid. Yes those Dells are low end in the realm of servers....

Re:The key

[Fallen_Knight]

I think the problem ther is that your going to use IIS....

Re:The key

[zealot]

Intel's specs can't come anywhere close to AMD in ANYTHING. AMD has Intel beat in: price, performance, heat, power, maximum operating temperatures, and (lower) heat output.

Not true. The opteron's pricing is similar to the xeons (depends which model you're comparing to (mp, dp, regular)). The performance benchmarks go back and forth, with the opteron generally leading in multiprocessor configurations. The power dissipation is a flat out win for opteron.

Not like your enterprise is going to care which company is making more money.

Companies will care. They need suppliers that have the cash to remain around to give support and take responsibility for problems. Sure x86-64, wave of the future... unless it dies with AMD a year from now.

Intel can't take enough of a loss for a long period of time to be competitive with AMD.

Not sure what this means exactly, but Intel so far isn't losing anything. They're still profitable (in fact, I've read that they've been profitable every single quarter since they went public 20-30 years ago). And for many quarters in a row AMD has been losing money.

The hammer chips are nice, but so far they're don't kill Intel in performance, they're just competitive. And while they have potential to do more, it hasn't been realized yet except in certain niche markets.

As you mention, there's more than performance (like power consumption), but it's still all wait and see... what will Intel bring out next, how well will AMD be able to supply chips and rally support?

Re:The key

[MoThugz]

Have you ever had to design a Windows-based network from scratch? Have you ever had your boss come to you and tell you you have $100K to spend? I have.

Hate to break it to you but most corporate networks are built from scratch if you're expanding to new physical locations. Well, if you're actually serious about your question, then I'd say yes... in fact this year alone I've done roughly 6 such projects and with budgets ranging from (approx) US$50k to a cool mil.

To go really low-end then build your own damn servers dammit... heck some of the custom servers we've built are at least at par with the 8450, if not better. 2U and/or 3U casings are not that expensive nowadays.

Well anyway, if IIS server farms (wtf?!) are your "thing"... there is still AMD... and don't tell me your "IIS server farm" (which of course implies Windows 2000) which costs under 100k runs Xeons?

Man if I'm the one in charge of whatever it is you've done... I'd fire your ass and kill that asshole of a boss of yours.

Re:The key

[Glock27]

But Intel has something in the server market that AMD doesn't: a reputation.

First off, Athlons and Athlon MPs have garnered a lot of respect around the industry.

Secondly, I'd say Intel's efforts have gained a reputation - however in many cases it's a bad one.

There is no question that Opteron pretty well smokes all current Intel offerings, especially in 2P and above configurations. :-)

Re:The key

[evilviper]

The opteron's pricing is similar to the xeons

I really don't think you can compare an Opteron to a Xeon. Think Opteron v. Itanium.

but Intel so far isn't losing anything. They're still profitable

My point is that, on price, AMD's processors are far less expensive than Intel, while still being better than them in just about every way. For Intel to prevent Opteron systems from taking over the market, they are going to have to eat some of the cost of their own chips long enough to keep people from even thinking about using AMD products. As I said, I don't believe they have enough cash to make that happen... That means they either need to make a HUGE leap forward with their processors, and come out with something that is actually better than AMD has, or their market is going to quickly slide out from under them.

The hammer chips are nice, but so far they're don't kill Intel in performance, they're just competitive.

Again, you are comparing two entirely different product lines. I wouldn't expect the first few chips (that are almost completely redisigned) to come out and be completely competitive with old technology, established chips. It won't take long for prices to drop, performance to improve, etc., but don't expect everything from the very first of an entirely new processor line.

Re:The key

[arne]

Not like your enterprise is going to care which company is making more money.

Companies will care. They need suppliers that have the cash to remain around to give support and take responsibility for problems. Sure x86-64, wave of the future... unless it dies with AMD a year from now.

The problem with x86-64 is compilers. Currently I gain a factor of 2-3 using intels compilers of gcc (and several other compilers I tried) in AMD hardware. And obviously Intel compilers are not available for x86-64.

Re:The key

[tomstdenis]

"reputation in the post-K6 age"

Um? What was so bad with the K6 processors? I had a K6-2 350Mhz processor and it was not only cheap but vastly outperformed my MII 300 [233Mhz] of the time. Considering I spent around 100$ for it [as I recall] it was a good investment at the time.

Specially considering pentiums were nearly double the cost where I live...

Tom

Re:The key

[devnullify]

The biggest problem with them was really Via's chipsets of the time. Drivers were extremely buggy on Windows, to the point of making them virtually unusable (at least on 3 machines I was forced to use from time to time). The one machine I got my hands on with an AMD chipset never caused me any grief.

Thankfully Via has cleaned up their act and now makes decent products.

Re:The key

[Jeff DeMaagd]

The thing I can think of is that Intel's been working its way into the slightly bigger server (2P, 4P) market for nearly a decade, and AMD really hadn't had a serious contender until this year.

Maybe AMD is on to something, but winning a few supercomputer contracts rarely affects corporate purchasing. It will be an interesting show as they seem to have taken the a lot of the hardest work out of making a multiprocessor system.

Re:The key

[chunkwhite86]

"Go price out an eight-way Dell 8450 server. Then tell me about low-end."

Go and price a nice sun fire, or a nice IBM regatta and get back to us kid. Yes those Dells are low end in the realm of servers....

Or how about an Alpha Wildfire GS160 or GS320. Or the Tandem/Compaq/HP "Himalaya" systems which run the NYSE.

If you want one of these, you better have a seven digit bank account balance.

Makes those 8-way Dell's look like tinkertoys.

Re:The key

[Anonymous Coward]

Not only are Opterons cheaper than Itania,
they also proceduce significantly less heat.

This is not so much of a problem in a desktop
computer, but on the Beowulf mailing list
heat production and ways to cope with it
are a constant hot (!) topic.

Toon Moene.

Re:The key

[Junta]

> 'nobody ever got fired for buying IBM'

Well here you go.

Re:The key

[paitre]

Portland Group Compilers.
Seriously.

They are -almost- as good as the Intel compilers (on Intel) and they're putting out a new rev that fully supports Opty.

The scientific computing community is only using Itanium in those few instances that they -need- the higher FPU marks -AND- have the cash to spend. In general, they're saving 30% and getting almost twice as many processors, so it's still a net win.

The compilers aren't the problem, provided you actually grab the right one.

Re:The key

[paitre]

First off, Athlons and Athlon MPs have garnered a lot of respect around the industry.

Specifically in the scientific computing community. There's also a number of companies that have latched on to Athlon MP-based web servers, too.

Cost plays a major issue, but for the average web machine, the little bit of extra oomph you get from a high-end Xeon over a high-end A-MP isn't worth the 15-20% price premium in most cases.
*shrug*

Re:The key

[drinkypoo]

You're comparing the K6 to an MII, a budget CPU to a budget CPU. If you put a K6 up against a Pentium II of the same clock rate, the K6 always loses. The Pentium has vastly superior floating point in two ways: K6 and K6/2 have 24 bit FPU, not 32, which meant that lots of drivers had to be targeted at K6 directly (mostly video drivers.) Plus, the K6 is just bad at being an i386. It's a great CPU but only code compiled directly for K6, NOT for 386, runs smoothly on it. The MII, well, don't make me laugh. I had my brief run-in with Cyrix chips, they were worse than the K6. I ended up with a P2 and I was much happier.

Now, I prefer AMD to intel, because the Athlon is just as fast (in some cases faster) as P4s with a higher clock rate. The PR follows Intel clock rate pretty closely until you get into the highest-end CPUs. But when the K6 was hip and new, it sucked. It wasn't worth a crap until the K6/3 came out, by which time I was committed to staying far away from the K6.

Re:The key

[jerkychew]

Wow. I've seen zealotry before, but jeez.

Look. I'm not, nor was I ever, comparing Dells to Suns. I only mentioned it in my second post as a retort to that other person's post.

What I'm doing is comparing Intel servers to the forthcoming AMD servers. that's all.

As for the shock of building an IIS farm - believe it or not, people do it. It was not my decision. When the company's product is written in ASP, and requires IIS to run, explain to me what I'm supposed to do with Solaris, or Linux, or any other non-Windows solution.

To go really low-end then build your own damn servers dammit... heck some of the custom servers we've built are at least at par with the 8450, if not better. 2U and/or 3U casings are not that expensive nowadays.

I love how all these so-called experts come out of the woodwork to defend their half-assed solutions. My point, as far as spending goes, was that corporations pay extra for reliability. They pay extra for a name. They pay extra for guaranteed uptime, and guaranteed support. All our servers were Dell, because Dell provided all of that. When you have major Fortune-500 companies using your systems, and downtime means you lose your job, the last thing you're going to do is use some generic solution. And don't ask me to build my own machine. At its peak, my company was buying half a dozen servers a week. Building each of those from scratch would have been a nightmare.

Re:The key

[ccp]

Andy, we know it's you.

Please, stop posting here. Thanks.

Cheers,

Re:The key

[phoenix_rizzen]

Well, it's not much, but our local School District (10,000 students, some 75 servers) is slowly moving away from Intel to AMD AthlonMP and (eventually) Opteron dual-proc servers.

They are much less expensive, performace the same or (usually) better, and come with the same warranties et al.

Plus, these are all FreeBSD or RedHat Linux servers with everything compiled from scratch, so they really scream.

Re:The key

[phoenix_rizzen]

Back in the 386 and 486 days, the AMD chips were generally better/faster than the Intel ones. That changed with the release of the Pentium and all the clones. The 5x86 was actually just a souped up 486. Cyrix, AMD, NexGen, IBM all tried to compete with the Pentium, but couldn't quite live up to it (especially with the release of the PPro and the P5-MMX).

Compared to the K5 from NexGen (I think that's the name) that it was replacing, the K6 wasn't so hot. It lacked features the K5 had, and was actually slower in a lot of areas.

Things got much better with the K6-2, and the (sadly shortlived) K6-III was even better still.

But, it wasn't until the original Athlon that AMD shone once again. And, the rest, as they say, is history.

Re:The key

[WNight]

Is the project is to build an IIS server, you're probably in trouble from the start. Proper projects start out as "Build a web server", where the hardware and software are assessed based on performance.

Re:The key

[phoenix_rizzen]

Actually, SPEC (the company) did some benchmarking comparing 2-way and 4-way Opteron, Xeon, and Itanium2 servers.

In both SPECint and SPECfp, the Opteron edged out the other systems.

Somewhere around here is a link, but it's mentioned in a nice blurb in this month's CPU Magazine.

Re:The key

[WNight]

Depending on what you're serving, it might not even be CPU limited (well, these days when a 1.8Ghz CPU is $75...)

A lot of companies are going the cluster route, especially with things like web servers that work so well this way, many independent and fairly expensive transactions.

Re:The key

[Unregistered]

'nobody ever got fired for buying IBM'

Wrong. Apparently some guy at Sun got fired for buying IBM.

Re:The key

[ckaminski]

Yes, but reality is often:

I want to run this software so we can milk more money out of our users. Plug this onto the webserver by yesterday afternoon.

And you go to the CEO to complain, because he's an open kinda guy, and he says:

Oh yeah, I told "PHB" to make sure you had that done by Friday...

And you give up and go dig ditches in Jamaica smoking all the pot you can get your hands on, because it's better than interfacing with these dipshits. :-)

Re:The key

[roger_and_out]

Reliability? kill the chip fan on an Intel and see what happens. Now do the same on an AMD. Any difference? Hell, yes! The Intel shuts down and stops. The AMD fries. Fix for Intel = 1 new chip fan. Fix for AMD = 1 new chip fan + 1 new chip.

Re:The key

[evilviper]

The Intel shuts down and stops. The AMD fries.

Mine doesn't. If you'd like, I'd be happy to provide you with Asus' video of what happens when you pull the whole heatsink+fan off the processor while it's running. Long story short, the system shuts down...

The only time an AMD chip fries is when it's put into a mobo that is so cheap that it has absolutely no protection what-so-ever. In fact, no processor of any type in any decent mobo should ever fry (due to heat).

SCO will bill them for...

[ihummel]

$2,306,700 and that is if they order their SCO liscenses before mid-October. I wonder whether the state of Utah won't suffer an un-accounted for nuclear accident in close vicinity to SCO's offices.

Re:SCO will bill them for...

[bersl2]

Maybe we should sign a treaty with SCO, so that we can have a hotline to Darl's office. It'll be just like Soviet Russia...

Think about it. SCO is essentially telling us "We will bury you!" every day. Of course, look what happened in the long run...

Re:SCO will bill them for...

[ihummel]

Somehow I do not think that SCO is going to be around for the next 45 years (about how long the cold war lasted).

Re:SCO will bill them for...

[pherris]

Somehow I do not think that SCO is going to be around for the next 45 years (about how long the cold war lasted).

Somehow I don't think SCO will be around after the next 45 days.

Re:SCO will bill them for...

[Henry V .009]

Just imagine SCO winning all of its lawsuits and becoming the next Microsoft. That'll help you sleep at night, won't it?

Re:SCO will bill them for...

[grep_who]

Well... at least we won't have to patch things on daily basis.

Re:SCO will bill them for...

[EmagGeek]

In other news, the Nuclear Regulatory Commission has changed the location of the future nuclear waste dump from Yucca Mountain in Nevada to the small town of Lindon, Utah.

A spokesperson for the NRC, who spoke to us on condition of anonymity, said that, "the new site along Highway 520 will be excavated in late 2003 via our new Ballistic Outward Fissionary Haul process, or BOFH, which is a process for rapidly removing large quanitities of land by blowing it sky high with a nuclear missile."

The Ballistic Outward Fissionary Haul process uses a high-velocity nuclear projectile to deliver the nuclear ordnance to the Selected Clearing Object, or SCO. A demonstration of this method was given last week near the proving grounds in southern Nevada.

The project is to be completed sometime before the end of the year. Once completed, the NRC said, there will no longer be any barriers to the storage of nuclear waste nor will there be any impediment to the widespread adoption of sensible free software.

Re:SCO will bill them for...

[DrFrob]

Funny thing is the state of Utah just got it's own 1000 processor AMD Opteron cluster.

Re:SCO will bill them for...

[pavera]

LinuxNetworx is building the cluster,
they are another Canopy Group member,
I bet they are exempt from licensing fees

They didn't create x86....

[Fallen Kell]

IBM did that.

Re:They didn't create x86....

[tyrnight]

"IBM designed the PC around the Motorola 68000 cpu"

It was actually the 6800 cpu.. the purple 40 pin package with the gold square on it

Re:They didn't create x86....

[IM6100]

The 8086 architecture was around LONG before IBM used it in any volume. There was a period before Intel produced any processors at all (before the 4004 and 8008) where Intel was mostly making RAM memory for IBM, and when I believe they were mostly owned (or outright owned?) by IBM.

Re:They didn't create x86....

[Tuxinatorium]

now that would REALLY overload the power grid...

Re:They didn't create x86....

[javiercero]

LOL Do you have any freaking idea about IC and board layout and design? Jeezus, do you have an idea about what you are saying? You do not design something and find out at the end that your CPU is too big for your process. I mean that is the CENTRAL element to your design! In fact in those days CPUs were socketed using standard TTL sockets, and the 68K was no bigger than the 8086.

The 68K was a 16bit bused processor, and IBM were producing mainframes and other systems with far more demanding production technologies. Believe me IBM had the facilities for dealing with the 68K socket!

LOL, jeezus tap dancing christ that has to be one of the most moronic comments I have ever read, and the fact that it was modded interesting means that you are not alone out there.

BTW IBM was using the 68K in other products.

Re:They didn't create x86....

[Jason1729]

A quick websearch leads to a lot of different explainations for why IBM picked the 8088 over the 68000. Some of the reasons I just found are IBM having a large supply of the intel chips on hand, IBM having a pre-existing arrangement with intel where they gave intel rights to their bubble memory in exchange to rights to the 8088 (that doesn't make much sense), and 68000 required more expensive support hardware. I didn't find any sites mentioning my explaination, but that's what I read in the early 80's when it was happening.

Jason
ProfQuotes

Re:They didn't create x86....

[Jason1729]

I'm pretty sure it was the 68k. I did a lot of work with the 6802 (basically the same thing as a 6800 but it just needs a crystal instead of an external clock). It's a pretty nice CPU, but it's not in the same league as the 8088 and 68000. A quick websearch seems to confirm that it's the 68k.

Jason
ProfQuotes

Re:They didn't create x86....

[mpthompson]

IBM planned on running a variant of CPM by Digital Research on their new 8088 based PC, but through a dramatic serious of screw-ups by the executives at Digital Research, IBM ended up choosing a CPM clone from Microsoft. The fact there was a functioning operating system and a stable of software (Wordstar, DBase, Supercalc, etc.) easily ported to the 8088 from the existing 8080/Z80 based CPM OS probably had more to do with the selection of the 8088 than anything else.

Re:They didn't create x86....

[Tuxinatorium]

Only if you don't watercool them with niagara falls.

Lightning computer system from Linux Networx

[Chyeburashka]

Here is the LANL link to the story.

Still plenty of floor space in the new building.

After having received notice from SCO,

[pair-a-noyd]

That they were to pony up millions of $$$ for free software, Los Alamos has decided to change the mission.
The new object of the project is precise targeting of the
Manhatten Project II on SCO HQ so as to cause as little collateral damage to Utah residents.

Remember, when you hear the siren, duck and cover.

*/waves buh bye to SCO/*

The Real Story is Here

[LuxuryYacht]

Los Alamos to get Lightning computer system from Linux Networx

LANL

For more specific technical detail on the supercomputer Linux Networx is building for Los Alamos, go to
LNXI Newsroom

Re:The Real Story is Here

[X]

That isn't too surprising since both companies are Canopy Group companies.

Other Interesting Info

[heli0]

Microsoft pre-beta Windows for AMD64 available(to resellers)

Opteron 246 Overclocking Benchmarks

Re:32 bit swapping to 64bit

[pair-a-noyd]

"Doesn't the 32 bit compatibility need a reboot to use, then another reboot back to 64?"

Only if you are running Windows...

Wasteful

[Anonymous Coward]

What people don't realize is that while the Pentiums they used before might not have been as quick, they were nevertheless a better buy because they served a dual purpose of, besides computing, producing enough heat to trigger small fusion reactions simply by placing certain radioactive isotopes close to the computer

This doubling of utility made the Pentiums far superior, i think, to the AMDs for Los Alamos' use, but i guess some suit at a desk ignorant of the technical issues just wasn't aware of that

Your tax dollars at work i suppose

-- roast beef

Re:Wasteful

[UtucXul]

You do realize that your average athalon can easily double as a heater, don't you? I can't imagine that any Intel technology can keep up with the heat produced by the dual athalon machine in my office without melting the plastic in the case.

Re:Wasteful

[sirsnork]

Actually Xeons use more power than Athlons, so any dual Xeon will wasily beat that. In saying that Opterons use even more power than Xeons....

Re:Wasteful

[Pharmboy]

You do realize that your average athalon can easily double as a heater, don't you? I can't imagine that any Intel technology can keep up with the heat produced by the dual athalon machine in my office without melting the plastic in the case.

Intel chips have been running hotter (mhz to mhz) than AMD for a few years now. This is well documented and has been talked about ad nausium here and elsewhere.

Re:Wasteful

[evilviper]

Not only do Intel P4 processors produce more heat, and use more energy for the same performance as their AMD counterparts, but Intel processors also have a lower operating threshold than similar-performing AMD processors, by about 20 degrees celsius.

Let's review shall we?

Intel P4: Runs Hotter, Can't handle high temperatures, more expensive.
AMD: Runs cooler, Easilly handles much higher temperatures, and are far cheaper.

So tell me again why anyone should buy from Intel.

Re:Wasteful

[El Cubano]

So tell me again why anyone should buy from Intel.

Duh...to get "Intel Inside" sticker on the computer's case. Sheesh, some people just don't get it.

Re:Wasteful

[MonMotha]

http://users.erols.com/chare/elec.htm

Check it out for yourself. That's a handy dandy table compiled from the datasheets for all common procs (the opterons are on there, and I will compare both 32 bit and 64 bit offerings from Intel and AMD).

The Athlon XP 3200+ (2.2GHz Barton core) is said to dissipate a maximum of 76.8W of heat (that's a lot), normally about 60.4W. This appears to be one of AMDs hottest chips (the famed "melting thunderbird" puts out a maximum of 72W according to the datasheet).

The Intel P4 3.2Ghz (presumably comparable to the 3200+ if you believe the speed ratings, which overall they seem to be at least close) is designed to dissipate a whopping 82W of heat, and no maximum is given. 82 > 72.

The Opteron-246 (2.0GHz) is designed to dissipate 80.6W of heat in normal operation (check the spec sheet for that definition).

The Intel Itanium-2 900 (slowest, lowest heat 64 bit offering from Intel) is designed to dissipate 90W of heat. 90 > 80.6. The fastest Itanium from Intel listed on this page (Itanium-2 1.5G) dissipates a humongous 107W! Go find a 100W light bulb for a quick comparison. Of course, they are dissipating it over an area of about 1 cm^2 (not sure of the exact core size on the itaniums, may be bigger).

Basically, the Intel procs actually put out MORE heat than comparable AMDs. That's not to say that AMD chips run cool. Indeed, my tbird system keeps the area nice and toasty (then again, I can use my P4m laptop as a hand warmer too!). x86 chips just tend to run very hot these days.

Re:32 bit swapping to 64bit

[Fallen Kell]

To my knowledge this is NOT the case. Now, it might be an issue with the OS that they are using (i.e. the OS might not be able to run 32bit applications in 64bit mode, and vice-versa).

Sorry to be bringing up a MS product, but their new 64bit Windows will be able to run 32bit programs with in 64bit OS mode, but not 64bit programs in 32bit mode (at least from my current understanding of the new product line). However, there was some performance hits, at least at the time that I read about these features (a few months back), so it may or may not still be the case now.

I am sure it won't be long for Linux to be able to run 32bit applications and 64bit applications within the 64bit OS version. Especially since MS figured out a way to do this, it shouldn't be too hard for Linux to be able to do so as well.

Re:/.ed at 11 comments...

[ihummel]

Alas that most websites don't set their servers up to survive a slashdotting. Perhaps the web would be a better place if they did.

Maybe eWeek could use ...

[DogIsMyCoprocessor]

some Opteron chips in some Linux **web servers**. Active Server Pages error 'ASP 0126' Include file not found /article2/0,3959,1220701,00.asp, line 891 The include file '/display_industry_brains/0,4302,c=Ziff+Davis+Seni or+IT+%3E+Linux+Unix+and+Open+Source,00.asp' was not found.

Re:Imagine...

[Stephen Samuel]

Actually, what's stopping Intel from shipping both the 32 and 64 bit chips on one board?

You mean besides the fact that they keep melting the board?

Re:Chinese and Los Alamos

[t0qer]

Actually, intel doesn't manufacture motherboards in USA. I think most of their motherboards are manufactured in Malaysia or in Ireland. (I know this because I used to work for a Taiwanese grey market dealer I.e. their token english speaking salesman)

If I remember right, the pentiumII's (and probably most of the modern P4's) are made in the phillipines. Phillipines is notorious for having Taliban sympathizers that have kidnapped and killed US tourist there.

why?,

[Bob The Lizard]

so as to cause as little collateral damage to Utah residents.

Why?. Its not as though Utah residents haven't been sucking up the rads for decades now.

Re:Chinese and Los Alamos

[jimhill]

Wen Ho Lee had been fired from the lab and lost his access to secure areas before the hard drive incident, so it's not too surprising that he was found not at fault for that. He is also not at fault for the Cerro Grande fire that sent us scurrying from town in the dark of night, nor is he currently thought to be a suspect in the local Blockbuster Video's inability to keep an unscratched DVD of "Shanghai Knights" in stock.

Sheesh.

Errm...

[pr0ntab]

Yeah, and you also need to reboot your Pentiums to run 16-bit code.

It is just the state of a flag in a control register. In particular, see page 68 of
AMD's Opteron System Programming Guide.

64-bit mode is enabled with the flip of bit 8 of the EFER Model-Specific Register. Otherwise it defaults to 32-bit mode. OS designers should test/set this bit just before running a thread in the scheduler, or jumping into system code as it can only be modified by code running in ring 0. This is the same way people treated the Virtual-8086 (16-bit) mode bit in CR0. In fact, you can combine the protected-mode, virtual-8086 mode, and "long mode" bits to have a variety of register-size and memory addressing modes per thread.

Re:Errm...

[DrFrob]

Scrawled on public bathroom tile grout: White Line Fault Line Tan Line

I NEED A BLO JOB LINE Is that the one is Hauser? I thought they cleaned that.

How Many?

[Prien715]

Quoth the author:
nuke-alwin writes "eWeek is reporting that Los Alamos National Laboratory announced it will use more than 3,300 Opteron chips in two of its Linux clusters...." (emphasis mine)

From lanl.gov:
Lightning includes 2,816 Opteron processors from Advanced Micro Devices (AMD), making it the largest Opteron system delivered in 2003 and the first 64-bit Linux supercomputer in the ASCI program.

Last I checked, 2816 is less than 3,300, though I can't get to the e-week article. Are the extra 500 chips in case of failure? But an 18% failure rate seems rather high....

Re:How Many?

[realdpk]

"in two of its Linux clusters"

"making it the largest Opteron system"

Two is also greater than one. :)

Re:How Many?

[Vyce]

"two of its Linux clusters..." Nuff said.

Re:How Many?

[Anonymous Coward]

Linux Networx is building a second 256 node (512 Opteron chips) cluster (LANL Orange) that will use the Infiniband interconnect from Mellanox technologies. LANL Orange will be the largest Infiniband cluster to date.

therefore: 2816 + 512 = 3328 Opteron chips

Re:Chinese and Los Alamos

[Anonymous Coward]

I don't understand why it received *interesting* rating. This is more of a flamebate.

First, Intel makes their chips where it is the cheapst. Pentium was designed from a group in Isreal. Consider the history of Isreal collabrating with Jews in US spying and steal info out of US military and companies, would it be possible that the Isrealies gov might thought about pressure the researh group to put some special circuit into the chip so they can steal more US secrets. In addition, Intel regularly ship the material to overseas for final assemblies in places like Malay. Remember that IRS wanted 600mil from Intel because Intel ship material overseas for final assembly? The IRS considered it as not totally made in USA therefore they don't get the tax credit.

Besides, if you look at motherboards, I can't find anyone making them from the US. The market is simply too compatitive. The labor along is too expesnive in US. Companies like Intel is a globalized company and they will ship stuff to where it is the cheapest. It is no better then AMD.

Personally, I believe that Intel's marketing people and Berret need to get their heads out of their collective asses and quit telling people what they need and don't need. The current slogan at Intel is "One Generation Ahead". Well, it's time to put the slogan to practice and produce some 64bit chips that is backward comp. to the existing 32bit apps and beat everything that AMD can dream up.

Re:32 bit swapping to 64bit

[Anonymous Coward]

This is already possible gcc -m32|-m64; granted both the 32-bit and 64-bit libraries need to be available. 32/64-bit compatibility is supported in SuSE Linux Enterprise Server 8 (SLES8); and will be available in RHEL 3.0.

Debian developers are also working on a port to Opteron that does/will support 64 bit Kernel with 32-bit and 64-bit compatibility.

Re:Chinese and Los Alamos

[heli0]

" It is well-known that the Opteron chips and mobos are manufactured in mainland China"

Not according to AMD: AMD Opteron(TM) Processor
All Opterons are made in Germany.

The only two fabs they they have are Dresden, Germany(fab 30) and Austin, Texas(fab 25).

Re:32 bit swapping to 64bit

[enos]

There is no need to reboot. Both 32 and 64 bit apps can run at the same time, as was demonstrated a while back with a linux system running both 32 and 64 bit versions of the same app side by side on the same machine.

Re:Chinese and Los Alamos

[Sokie]

All the troll/flamebait moderators must be without power.

Someone already pointed out that the Opeterons are fabricated at the Fab30 in Dresden. AFAIK, AMD has no fabs in China and I'm not aware that they even do assembly there.

But the most farciful (Note to grammer Nazi's: Yes I did just make up that word.) statement is your post is this one:

What would stop them from putting data-wrangling code into the Opteron chips?

So even if the Opteron was fabbed in China, you think that the Chinese James Bond is just going to slip in an entirely new chip design into the assembly line and none of the automated or manual microscopic inspections each chip undergoes is going to notice that there are an extra 25,000 transistors over there and these other 30,000 transistors are in the wrong place?

And as another person pointed out, Intel does very little manufactering in the USA these days. At least AMD has a fab in Texas. I couldn't find any info on Intel's fab locations quickly, but I don't recall that they have a large scale one in the USA anymore, but that's complete guestimating.

You sound as bad as the lady at work that thinks buying Microsoft is her patriotic duty!

Re:Chinese and Los Alamos

[Pompatus]

I was waiting to reuse this link I found on here the other day. What you need is one of these.

No reboot, but you do need 64 bit OS.

[Anonymous Coward]

You need the 64 bit OS, as you might expect, because allowing an application to address a 64 bit memory space while the OS is only capable of managing a 32 bit one, well, wouldn't work out very well. Hint: It's really tough returning from system calls made from the upper few terabytes of memory.

So, a 32 bit OS starts out "seeing" your standard 32 bit X86 CPU. This is the legacy mode, and looks like any 32 bit CPU out there today.

On Opteron, a 64 bit OS will also know how to set the new "Long Mode". Bzzt...A bunch of new instructions and handy registers magically appear.

From there, the OS can mark each code segment (your program) as being in "Compatibilty" (32 bit) or "64 bit" mode.

For more than you ever wanted to know...

Opteron Tech Manuals can be found here.

just toured Linux Networx today... wow

[dougnaka]

I saw quite a few of their new clusters ready to ship out. I had to constantly wipe the drool off my face while I watched them assemble 2GHZ dual Opteron boxes with 2GB RAM per processor. Their tech is impressive. They have their Ice Box control units and quite a bit of custom control/monitoring hardware that makes building your own cluster seem less advantageous.
They boast #3 on the worlds fastest super computers, so questions about Linux on the "Enterprise" should be easily resolved.
This Tom's hardware review of Opteron vs. Xeon is quite interesting to give a better feel for comparison to todays speeds.
It will be nice when we have some numbers to compare Itanium II direct 64 bit to Opteron, although it doesn't seem much can save the Itanic IMHO.

The real key...

[DeathPenguin]

...is that AMD opened their platform well enough to the LinuxBIOS developers while Intel basically told them to screw off and live with EFI. Here is what Ron Minnich had to say earlier on the LinuxBIOS mailing list.

Re:The real key...

[Anonymous Coward]

Excellent point; it is 100% true that AMD cooperated with the LinuxBIOS developers but one might also say that the real key is that Eric Biederman (an LNXI developer) was the work horse behind making LinuxBIOS on Opteron a reality.
BTW, most of the LinuxBIOS clusters that were mentioned in Ron Minnich's LinuxBIOS ML post are Linux Networx systems... emm gooood

G5

[useosx]

Don't forget the "G5" (PPC 970). It runs 64/32. Not sure how to compare the price since you can't buy G5s alone (damn Apple). But as I recall, a dual MB with two of the fastest Opterons on the market ran to about $2,000 USD on Pricewatch. The dual 2Ghz G5 is 3 grand. So 1 grand difference but you get a sweet case, DVD-R, 160GB HDD, 512 RAM, Radeon 9600, keyboard, mouse, OS, all their onboard stuff, and 9 fans.

Re:G5

[Mr. Arbusto]

Don't forget the "G5" (PPC 970). It runs 64/32.

Apple doesn't make the G5 and as you mentioned it is a PPC 970 which is made by IBM. IBM is happy to include it in other machines with the Non-G5 name. I'm not sure if you can purchase just the processor from IBM, but they are very happy to included it in there new machines and possibly build for custom applications like a Linux Cluster.

Re:G5

[Durandal64]

That's a little different, since the PowerPC spec was designed to be 64-bit from the start. The MPC74xx, PowerPC 75x, and 60x series utilized a 32-bit subset of the 64-bit spec. So everything you do with the 32-bit subset is valid with the 64-bit spec.

Re:G5

[MSTCrow5429]

If that is true, it is quite interesting, from the standpoints of planning and mechanical design. I have never heard of this before though, and was hoping you could offer some reliable sources via URL so I could further my knowledge on this subject. Thanks.

Re:G5

[Pharmboy]

Apple doesn't make the G5 and as you mentioned it is a PPC 970 which is made by IBM. IBM is happy to include it in other machines with the Non-G5 name.

IBM will be making servers with the 970 once they catch up with demand for apple. They are anticipating Quad 970s starting at $3500 for an empty box, which is very cost effective. All of this on IBM's site.

I submitted a couple of stories, plus several other /.ers as well, on this subject. Including benchmarks, etc. My understanding is that Linux is pretty much 970 ready, although some tweaking will go on. The 970 is based on their Power4 chip, and so far has shown to be a real badass. Even Apple's commercials are claiming "the most powerful home computer in the world" about their dual 2.0 box, and the benchmarks seem to back them up.

Re:G5

[Durandal64]

Of course. Check out this link for details.

Re:G5

[MSTCrow5429]

Are you aware of any other CPU architectures that had 32-bit implentation of a 64-bit spec? Might they have had 4-8-bit, 16-32-bit, or similiar?

Re:32 bit swapping to 64bit -NOT- Just FUD

[Anonymous Coward]

FUD and FUD alone. I bought a dual Opteron and am using it right now.

The Opteron's lower 32 bit registers look just like an Athlon's or P4 registers. When code wants to run in 32 bit mode, it just uses those registers. When the second process wants to run in 64 bit, it uses the lower 32 bit registers as well as the upper 32 new registers.

It's the Itanium chips (64 bit from Intel) that you are thinging of. :) It has a horribly slow software emulation layer to run 32 bit apps.

AMDs 32 bit implentation is in the silicon. Can't beat that.

Re:Opteron, MADE IN MALASIA

[dougnaka]

well the idiots in Germany apparently don't know how to mark their chips correctly

look at the close up... you may notice the MADE IN MALASIA markings...
Now, maybe that's just the engineering sample, but a tech at Linux Networx told me today that they just got a large shipment of 2GHZ Opterons direct from Singapore...

I can't help it oh no here we go Sco again!

[ratfynk]

Can you say SMP can you say SCO can you say pay lets see how much can you ask from the OS licensee to the mother of all clusters. Dig deep Los Alamos you are very close to where Davy Crocket died and just might find that Utah has some wild SCO tribal war chant starting up. The Sco Lawyers are passing around the war pipe (filled with crack) right now thinking about how to attack the Long Knives over in New Mexico. Maybe this might turn into a US Agencies new version of Clusters Last stand!

Al Gore never claimed he invented the Internet.

[C3ntaur]

What he *actually* said was, "During my service in the United States Congress, I took the initiative in creating the Internet." Granted, it's debatable as to how much initiative he took in its creation, but he was in fact involved in legislation and funding that helped to shape it.

Whenever I see this twisting of words and facts perpetuated, it reminds me of the fools who just can't say nuclear (it's "noo-clee-ar", not "noo-kyoo-lar", damn it!!!).

Re:Al Gore never claimed he invented the Internet.

[Tuxinatorium]

DARPA started the internet, and DARPA is funded as congress damn well pleases, and funds can be earmarked for certain projects. So yes, Al gore did have a role in creating the internet.

Re:Al Gore never claimed he invented the Internet.

[Pharmboy]

DARPA started the internet, and DARPA is funded as congress damn well pleases, and funds can be earmarked for certain projects. So yes, Al gore did have a role in creating the internet.

DARPA was started in 1969, while Al Gore was a photographer in Vietnam. He had no role in funding DARPA/Internet at that time, obviously, and anything significantly after that time would not be considered origination. All that Algore created was a great running gag for late night TV.

Re:Al Gore never claimed he invented the Internet.

[Jaysyn]

But he did have a role in turning it over for civilian use.

Jaysyn

Re:Al Gore never claimed he invented the Internet.

[Pharmboy]

But he did have a role in turning it over for civilian use.

I am guessing that Jesse Helms also voted to turn it over to civilians, but I don't see anyone praising him for it.

Re:Al Gore never claimed he invented the Internet.

[Rostin]

From http://www.webster.com Main Entry: nuclear Pronunciation: 'nu-klE-&r, 'nyu-, /-ky&-l&r

Re:Al Gore never claimed he invented the Internet.

[evilviper]

and anything significantly after that time would not be considered origination. All that Algore created was a great running gag for late night TV.

No, no. You are obviously missing some facts.

Al Gore lead the process of allowing commercial interests to use the Internet... Previously, it had been non-profit use only. In that process, Gore is quite correct that he helped to create the (commercial) internet, and more specifically, the internet we know today. Without that change, there would be no Amazon, no Yahoo, no slashdot, and you would not have discs mailed to you that allow you to quickly connect... You would have to arrange for access through a university, or someplace similar.

Al Gore's statements are quite correct, in context. It's only when taken out of context (which is what is happening constantly) that it becomes a gag.

Re:Al Gore never claimed he invented the Internet.

[rumpledstiltskin]

mehehehehehe
It's "noo-kyoo-ler" dummy; the s is silent
--Family Guy

Re:Opteron, MADE IN MALASIA

[aka1nas]

The CPU packaging is done in Malaysia or Singapore, the actual dies themselves are all fabbed in Dresden, Germany.

Re:32 bit swapping to 64bit

[evilviper]

I am sure it won't be long for Linux to be able to run 32bit applications and 64bit applications within the 64bit OS version

Why would you even want to mix-and-match with Linux (or BSD)? With Windows just about all programs are propritary, so you have to run whatever you can get, but with Linux, you just have to recompile and get everything to 64-bit, no problem.

Yes, I know some people run propritary Linux programs as well, but those are rare (and it's their own fault anyhow... *nag, nag, nag* )

While it's nice to see the 264 OC'd win...

[pr0ntab]

...they didn't even try running anything in 64-bit mode. I want to know how much faster 64-bit native stuff will run vs. 32-bit stuff.

When will we have 64-bit native q3a? Come on Carmack, stop fiddling with your rocket and get cracking! The benchmarking websites demand it.

Re:While it's nice to see the 264 OC'd win...

[heli0]

" I want to know how much faster 64-bit native stuff will run vs. 32-bit stuff."

Check it out:
http://www.tomshardware.com/cpu/20030422/opteron-1 7.html

Start there and read the next couple pages. Native 64-bit runs up to 30% faster.

As far as 64-bit native q3a probably not until the end of the year when a more stable Win-64 build is available. Although there is already a Half-Life2 64-bit build running on Win-64 Alpha, which may lead to some leaked figures earlier.

Re:Chinese and Los Alamos

[spotlight2k3]

With the hat, even sco can't see you. every one run out and get the hat before sco claims they own that too!

Application specific Kernels

[bstadil]

"Doesn't the 32 bit compatibility need a reboot to use, then another reboot back to 64?"

On a serious note since it's based on LinuxBIOS they have a system where you can launch another Kernel without taking the system down. It is quite clever and you can optimize alternative kernels for the program you are running, incl 32 vs 64 bit.

Re:32 bit swapping to 64bit

[Weirsbaski]

Sorry to be bringing up a MS product, but their new 64bit Windows will be able to run 32bit programs with in 64bit OS mode, but not 64bit programs in 32bit mode (at least from my current understanding of the new product line).

In Opteron, 64-bit apps cannot be run under a 32-bit OS. Opteron doesn't recognize code as being 64-bit code unless long-mode is enabled, and once long mode is enabled the OS must be 64-bit (because all switches to more privileged code also switch to 64-bit mode).

Being able to run 32-bit apps under a 64-bit OS was one of the absolutely required features for Opteron, however.

(yeah, I work at amd)

PPC 970

[batura]

When I read stories such as this, I really start to wonder what type of dumbfucks run IBM. By closing the hardware support for the 970, they've bascially prevented their chip from becoming a player in these massively parallel computing environments.

Who wants to buy 3000 IBM or Apple branded boxes when you can get 64-bit Opterons with whatever box maker you want? Doesn't that make a lot more sense to the bottom line? The most annoying part is I am sure you are bound to using IBM service contracts as part of the deal.

Re:PPC 970

[Viceice]

They Didn't buy PWC Consulting for nothing.

Re:Chinese and Los Alamos

[labradorx]

Intel has fabs in Arizona(fab12), New Mexico(fab 11), Massachusets(fab17), Colorado(fab23), Oregon(fabs 20, D1C, and soon to open D1D) and at least 2 fabs in California(don't know thier names). There is a fab in Ireland and one in Israel. I'm not aware of any Intel fabs in Asia at all. The facillities in Malaysia, China and Phillipines are assembly. The bulk of Intel's manufacturing is in the US.

Another link here

[Zog The Undeniable]

At The Register. Sad that all this power is being used to simulate nuclear weapons (and presumably speed up GWB's adoption of battlefield mini-nukes), rather than its proper purpose of getting more fps at UT2003.

Re:Another link here

[lewger]

Would you prefer they did real testing in some poor countries backyard like another nuclear power?

Re:Chinese and Los Alamos

[MoThugz]

Intel:
- Malaysia and Ireland, you're right.

AMD:
- Malaysia and Germany (Dresden if I'm not mistaken).

But labelling any country as a terrorist sympathizer is just plain dumb.

Re:SCO will bill them for..

[Netlink]

They are buying the cluster from Linux Networx who are a Canopy Group company Like SCO. Perhaps Linux Networks customers are immune to the SCO claim. I think it would be interesting to ask them, and Canopy who obviously support the SCO claim.

Re:SCO will bill them for..

[X]

Yeah, I think this is the interesting aspect of the whole thing. My guess is Linux Networx may quickly become one of the big SCO licensees, at which point the money from the Linux gig becomes funnelled in to the anti-Linux legal fund. :-(

That explains it.

[Kynde]

$2,306,700 and that is if they order their SCO liscenses before mid-October. I wonder whether the state of Utah won't suffer an un-accounted for nuclear accident in close vicinity to SCO's offices.

SCO hq in Utah ? I didn't know that, and I'm guessing if more people had known this from the start I may have never even heard of SCO. /. might have mentioned that "some mormon commitee is full of hot air and it" instead of all this "trying to sue the world"/"take over computer industry"/"overthrow GPL" crap...

The Hegemony Is Broken

[tgrasl]

The age of WinTel ends, and the age of LinAMD begins

Re:Chinese and Los Alamos

[MSTCrow5429]

But labelling any country as a terrorist sympathizer is just plain dumb.

Iran, Syria, North Korea, Cuba, Libya...

Imagine a Beowulf Cluster of These

[patch-rustem]

Okay,
I'm sorry,
next time I'll rtfa,
I promise.

Intel Screwed Up

[Bruha]

I dont know who was asleep at the switch over in intel land but surely that guy is one job less after rejecting the idea to use the same backwards compatibility idea that made things so successful in the 16 bit to 32 bit switch all those years ago.

Now I get to watch a chip company I've supported (AMD) for years finally succeed. I just hope they stand by and keep their good prices for performance and not start to charge more for their chips if they become top dog or at least get a lot closer to knocking Intel off their high horse.

Re:Intel Screwed Up

[hackstraw]

Huh?

You still want to be using a CPU from the 70s for how much longer? You still desire 8 and 16 bit apps do you? You still run a.out binaries under linux?

Granted, if you plan on using an Itanium on a Windows platform, you may have some creature comfort programs unavailable, but as for Linux and HPUX, everything you need is there. Also, backwards compatability with parisc 32 and 64 bit apps are available under HPUX. From http://www.hp.com/products1/itanium/infolibrary/wh itepapers/archives/parisc.pdf :

1)Execution without conversion With very few exceptions,HP-UX 11i and HP-UX 11.00 32-bit and 64-bit PA-RISC application binaries execute on HP-UX 11i Itanium systems automatically and transparently without modification or conversion .Binaries from earlier versions of HP-UX (HP-UX 10.20,10.10,etc.)qualify for Itanium compatibility if they are certified on HP-UX 11.00 or HP-UX 11i.

2)Recompiling without modification With very few exceptions,HP-UX 11i and HP-UX 11.00 PA-RISC application source code can be recompiled without modification for native execution on HP-UX 11i Itanium systems.

3)Data transfer without modification or conversion Existing PA-RISC data can be transferred and used without modification or conversion on Itanium HP-UX 11i systems.

Oh, and with your Opteron, which compiler are you going to use to get performance anywhere near the performance of the Intel compiler for the Itanium? You know those spec numbers were most likely done with handcrafted assembler from AMD and unreproducable by you and me. I was able to measure _exactly_ the memory bandwidth of the Itanium as reported by Intel. And, yeah, its fast.

Power consumption problem

[DV]

One of the very serious problem related to building
Itanium clusters is their very high power
consumption and the associated heat removal problem.
It's okay for a few server in a room, but for
cluster trying to pack boxes is a key point of the
architecture. Apparently Opteron is not too bad
since there are dual Opteron in 1U server format
design commonly available, and it was overheating
that would be known by now, but for the Itanium(2)
cluster I know off, they never managed to get the
full cluster running without bringing either the
power supply down or the air conditionning down.

Itanium 1 was notoriously power hungry and
a common source of joke about this, Itanium 2 is
certainly better in this respect, but the clock
speed has been multiplied by nearly 3, I really
doubt they could compensate the initial problem
enough to get the new high speed chip to get back
to a decent consumption.

On the other hand Opteron seems quite better
probably getting the benefit of all the power
consumption research that AMD did during the 90's
where AMD chip were at the time consuming significantly more than Intel equivalents.

Now if someone has the time to make a search
for the advertized power consumption of both chip
that would be a really interesting post :-)

Daniel

Re:32 bit swapping to 64bit

[salimma]

Why would you even want to mix-and-match with Linux (or BSD)?

In any case it's already possible - you need the 32-bit libraries from libc upwards installed in parallel to the 64-bit libraries. Kind of like the way Linux emulation works in FreeBSD I guess.

It was the 8086...not the 6800

[voss]

The 8086 was the 16 bit chip...
but it was too expensive to use
at the time...so the 8088 was a compromise.

The 80286 was the next progression
on the 8086. The main difference
between the 8086 and the 80286
is that 286 went faster.

RTFM

[Chep]

where the Fine Manual = The Athlon 64 and Opteron optimization guide (PDF from AMD).

It's spelt out in full, page 225:

Future processors with more or wider multipliers and adders will achieve better throughput using SSE and SSE2 instructions (Today's processors implement a 128-bit wide SSE or SSE2 operation as two 64-bit operations that are internally pipelined).

(...)

The SIMD instructions provide a theoretical single-precision peak throughput of two additions and two multiplications per clock cycle, whereas x87 instructions can only sustain one addition and one multiplication per clock cycle. The SSE2 and x87 double-precision peak throughput is the same, but SSE2 instructions provide better code density

(emphasis mine)

So, in other words:

1. SSE is twice as fast as single-precision x87
2. their initial implementation of SSE2 is actually just an implementation of the instruction set, the actual execution units are borrowed from the x87 and have the same throughput limitations
3. if you are writing really large code, you might get some benefits from SSE2 because of L1 savings (but your case has to be pretty borderline for that to make a significant difference, or you have to be able to interleave a significant amount of integer code with your FP code that using SSE2 frees enough decoder bandwith).
4. In the future, they might bother with spending more space for more double-precision FP units (I'd think that if they can sort out the manufacturing problems, and the HPC segment seems like the area they're selling to, then it may make sense to spend a few mm to double SSE2 performance. Depends on who actually buys Opterons once the software/OS front lines are settled)
5. Today's AMD64 SSE2 performance in double-precision is not going to be leaps away from their x87 stuff (unlike the P4), but you have to remember that AMD's x87 is really good.
6. Don't forget that in long mode, you have 16 128-bit registers, which may help your compiler make more out of SSE2/64 code than x87 code; some benchmarks with SSE/SSE2 enabled compilers would help clarifiy whether that's the case or not.

So in short, yeah, they claim a somewhat faster SSE, but their SSE2 is mostly a "yep we can run that code too" item rather than "yep, we can run that code too, and faster to boot".

NEWSFLASH

[aurelian]

Nuclear weapons have already been developed, tested, and used. They worked. Lots of people died. So why exactly do we need to spend money and resources developing even more lethal versions?

You call people naive and then refer to 'the supremacy of good over evil'.. think about it dumbass.

Re:NEWSFLASH

[covertlaw]

Actually, I think the point is to make "less-lethal" nuclear weapons. Weird, yes, but if you can develop nukes with no radioactive fallout, better ground penetration, and lower overall yield, you will save civilian lives.

I'd be willing to bet that 80% of our stockpile isn't even useable anymore. All the chemicals used in their construction has some sort of expiration date.

Nuclear weapons are the shittiest form of battlefield technology we've ever invented, aside from chemical and biological weapons. I believe that the WORLD should disarm itself of such devices. However, with the threat posed by these tertiary dictators and despots all over the world, not to mention Islamic extremists, it's hard to justify giving up deterrence. Maybe some day we can.

Re:NEWSFLASH

[charnov]

Low yield nuclear weapons (fission bombs) that generate a nice bang and flash with nearly no residual radiation have existed, but it was deemed too "usable".

The idea is that the weapon should be so horrible that it is strictly a last resort. Nukes are rapidly entering the territory of being able to be used all the time for any reason. That is the scary thing.

OBTW, the US government is very much into the devlopment of stuff that FAR surpasses the power of nukes, but can be smaller and directed in it's use (think ray gun type weapons...aka directed energy). These weapons are safe (relatively), cheap(er) and don't fall under the current treaties for nonproliferation.

Thanks Bush.

Al Gore helped make the public utility.

[Futurepower(R)]

That's right. In several email conversations, I questioned Vint Cerf about this, and he said that Al Gore was extremely important in making DarpaNet, a research tool at a U.S. government organization, into the Internet, a public utility available to all.

DARPA is the U.S. government's violence research department. DARPA is devoted to finding more efficient ways to kill people. In the beginning of networking computers together, there was no intention of benefiting anyone.

According to Mr. Cerf, Al Gore recognized the importance of a public computer network long before other public officials knew anything about computers, and made sure the public network had funding.

Some have called Mr. Cerf, "The Father of the Internet", but, as his biography says, many people were involved. Mr. Gore was the main promoter, "father", of the public utility we now call the Internet.

Re:Al Gore helped make the public utility.

[ajp]

Furthermore, Al Gore, as is common amongst politicians, was trying to capitalize on his father's legacy as the senator who secured funding for the interstate highway system. Note that the senior Gore did not by any means invent nor claim to invent the interstate highway system. It was, like the internet, a military project meant to help to efficiently kill people (provided they had landed on American soil with big nasty guns.)

Al felt that by securing funding for the information highway (though, I believe, that buzzword was years away) that he could capitalize on daddy's good juju.

Unfortunately, this only works for Republicans, who seem to be in control of the media people actually pay attention to: talk radio.

Re:Al Gore helped make the public utility.

[Jaysyn]

The interstate highway system was developed to get people out of the way of hurricanes & other natural disasters faster. A good friend of mine's grandfather was one of the engineers who designed it.

Jaysyn

Al Gore helped make the public utility.

[Futurepower(R)]

Software failure: Slashdot's system posted my comment in the wrong position. Hopefully this will be posted as an answer to Jaysyn, under comment #6704329

My comment: That's right. In several email conversations, I questioned Vint Cerf about this, and he said that Al Gore was extremely important in making DarpaNet, a research tool at a U.S. government organization, into the Internet, a public utility available to all.

DARPA is the U.S. government's violence research department. DARPA is devoted to finding more efficient ways to kill people. In the beginning of networking computers together, there was no intention of benefiting anyone.

According to Mr. Cerf, Al Gore recognized the importance of a public computer network long before other public officials knew anything about computers, and made sure the public network had funding.

Some have called Mr. Cerf, "The Father of the Internet", but, as his biography says, many people were involved. Mr. Gore was the main promoter, "father", of the public utility we now call the Internet.

National Dedicated Game Server Farm

[Anonymous Coward]

As a taxpayer, I demand that any spare computing power on government hardware be put to use for dedicated game server time.
It's worth the security risk.
My vote is for Halflife 2.

they're out there

[smatt-man]

I think they're using Opteron processors to be more compatible with the aliens' OS.

Re:Chinese and Los Alamos

[evilviper]

Not according to AMD: AMD Opteron(TM) Processor All Opterons are made in Germany.

Hmm, I'd sure like to hear the explanation for the "ASSEMBLED IN MALAYSIA" stamp on Opteron processors then.

Opteron Photo

One way or another please

[MarcQuadra]

Simulating nuclear weapons is a heck of a lot better than finding out 'the old way'. If we're going to HAVE nukes (and there's no turning back now, I'm afraid) we may as well have the right ones for the job. I'd rather have them develop 'right-size' weapons for the post-cold war era than use giant 1970's warheads on third-world dictators. I find it foolish to demand reductions in 'collateral damage' while simultaneously demanding limits on harmless weapons simulations aimed at reducing said damage.

I'd rather have 20 very accurate, very small nukes in our arsenal than the thousands of ICBMs we have now, it would be just as effective a deterrent; getting there means money and (simulated) testing.

AIST also going for Opteren Cluster

[Junta]

See link here

This one is an IBM made one. Pretty interesting.

They want to bill their own customers

[TheConfusedOne]

At least according to the SCO licensing FAQ. They're saying that people who bought SCO's Linux are going to have to buy an IP License as well.

Re:32 bit swapping to 64bit -NOT- Just FUD

[Mojo Trolljo]

"32 bit" is more a function of the OS and compilers than it is of the processor. For example, I use HP-UX on itanium which is a hybrid 32/64 OS and whose compilers generate both modes natively in the instruction set. Don't blame shortcomings in popular operating systems and legacy code on the Itanium.

This is apples and oranges IMO.

Re:Chinese and Los Alamos

[Cleveland Steamer]

The chips are fabricated in Dresden and packaged in Malaysia.

This will finally answer the burning question...

[Call Me Black Cloud]

...what happens when a beowulf cluster of SCO lawyers attacks a Linux cluster of Opertons. Film at 11.

Coals to Newcastle!

[spineboy]

Oh that's great, the scientists are putting 3000! AMD chips in the desert!. As if it wasn't already hot enough there already! They could move it to Chicago in the winter and raise the downtown temp to a balmy 80 F. Chicago, the windy city, has the proper amount of wind flow for over-clocking these babies.

Fantasy time! If I worked in Los Alamos I would; 1 definately be in the lead in SET@home, and 2 My Quake FPS would be pretty good I suspect...
later.

Re:Coals to Newcastle!

[Nynaeve]

If I worked in Los Alamos I would; 1 definately be in the lead in SET@home [sic], and 2 My Quake FPS would be pretty good

You forgot: 3) Unemployed because of misappropriation of government property.
:)

Re:32 bit swapping to 64bit

[tugrul]

You want to mix-and-match on Linux/*BSD because some apps gain nothing and lose out on increased memory usage from going 64-bit (think memory references and decreased code density), hurting both bus and cache utilization as well.

ppc vs. pc

[devphaeton]

Probably off-topic, but when might we start seeing IBM-branded computers with the 970 available? Before the Apple Party Line gets going, please understand that i'm not looking for a cheap box to run OS X on.

I'm more interested in a decent, low-power-using (iirc) 64-bit package for GNU/Linux and FreeBSD.

I will admit right up front that there are many people more clueful about this than me, but from what i understand, AMDs new chips (and mobo designs) are great and all, but they are continuing the legacy of "kludge on top of kludge on top of kludge" in hardware design that goes back to the 8088. I thought that Intel was trying to wipe the slate clean and start from scratch, to "clean up" the PC hardware architecture, so to speak.

I dig AMD and all, but they are sure to undermine this effort with their "64-bit Kludge" processors.

And as for performance, yes i understand that 64-bit hardware does you no benefit if you run all 32-bit apps. But the apps will come. And also, my entire post here is completely beside all the "Performance and Benchmark" debates going on between AMD, Intel and Apple's benchmark announcements. Performance-wise, my 1533Mhz Athlon is quite sufficient for all the things i do (hell, my 300hp K6-II is still 100% useable every day).

I guess i just want better engineered hardware that puts out less heat, noise and has fewer internal performance mismatches and bottlenecks.

Disclaimer: I have all AMD systems, I dislike OS X, I hate Windows, I'm not trolling and i wasn't the one that spray-painted your cat.

Re:Chinese and Los Alamos

[Sokie]

I was able to find this document (PDF warning) that doesn't speak about volume, but does tell exactly what each Fab does and on what process. Fabrication of logic products (which I assume to mean processors and such) is done at the following facilities:

Fab 22 - Chandler, Arizona
Fab 12 - Chandler, Arizona
D2 - Santa Clara, California
Fab 10/14 - Leixlip, Ireland
Fab 24 (under construction) - Leixlib, Ireland
Fab 8 - Jerusalem, Israel
Fab 18 - Qiryat Gat, Israel
Fab 17 - Hudson, Massachusetts
Fab 11 - Hudson, Massachusetts
Fab 11X - Rio Rancho, New Mexico
D1C - Hillsboro, Oregon
Fab 20 - Hillsboro, Oregon

Like I said, it doesn't say anything about volume, but about 1/3 of their logic fabrication facilities are international.

Interesting stuff.

Re:32 bit swapping to 64bit

[jumpingfred]

Actually lots of propritary are starting to get ported to linux. Lots of CAD tools are being ported to linux.

Re:Chinese and Los Alamos

[ralphclark]

It's "Grammar Nazis", not "Grammer Nazi's".

impressive

[Trailer Trash]

AMD makes one single sale that's bigger than Intel's combined Itanic sales...

Why not debate the point with Vint Cerf?

[maynard]

Whatever you may think of Al Gore and his policies, the fact is that core Internet developers recognize Gore's contribution to the Internet as a policy-maker in both Congress and the Senate as funding source. Really, this Republican talking point ought to die on the vine for the vindictive misrepresentation and lie that it is.

Here's what Vint Cerf had to say on the matter as forwarded by Declan McCullaugh.

Date: Thu, 28 Sep 2000 17:43:58 -0400

From: vinton g. cerf (vcerf@MCI.NET)
To: Declan McCullaugh (declan@well.com), farber@cis.upenn.edu
Cc: rkahn@cnri.reston.va.us
Subject: Al Gore and the Internet

Dave and Declan,

I am taking the liberty of sending to you both a brief summary of Al Gore's Internet involvement, prepared by Bob Kahn and me. As you know, there have been a seemingly unending series of jokes chiding the vice president for his assertion that he "took the initiative in creating the Internet."

Bob and I believe that the vice president deserves significant credit for his early recognition of the importance of what has become the Internet.

I thought you might find this short summary of sufficient interest to share it with Politech and the IP lists, respectively.

===================

Al Gore and the Internet

By Robert Kahn and Vinton Cerf
Al Gore was the first political leader to recognize the importance of the Internet and to promote and support its development.

No one person or even small group of persons exclusively 'invented' the Internet. It is the result of many years of ongoing collaboration among people in government and the university community. But as the two people who designed the basic architecture and the core protocols that make the Internet work, we would like to acknowledge VP Gore's contributions as a Congressman, Senator and as Vice President. No other elected official, to our knowledge, has made a greater contribution over a longer period of time.

Last year the Vice President made a straightforward statement on his role. He said: 'During my service in the United States Congress I took the initiative in creating the Internet.' We don't think, as some people have argued, that Gore intended to claim he 'invented' the Internet. Moreover, there is no question in our minds that while serving as Senator, Gore's initiatives had a significant and beneficial effect on the still-evolving Internet. The fact of the matter is that Gore was talking about and promoting the Internet long before most people were listening. We feel it is timely to offer our perspective.

As far back as the 1970s Congressman Gore promoted the idea of high speed telecommunications as an engine for both economic growth and the improvement of our educational system. He was the first elected official to grasp the potential of computer communications to have a broader impact than just improving the conduct of science and scholarship. Though easily forgotten, now, at the time this was an unproven and controversial concept. Our work on the Internet started in 1973 and was based on even earlier work that took place in the mid-late 1960s. But the Internet, as we know it today, was not deployed until 1983. When the Internet was still in the early stages of its deployment, Congressman Gore provided intellectual leadership by helping create the vision of the potential benefits of high speed computing and communication. As an example, he sponsored hearings on how advanced technologies might be put to use in areas like coordinating the response of government agencies to natural disasters and other crises.

As a Senator in the 1980s Gore urged government agencies to consolidate what at the time were several dozen different and unconnected networks into an ?Interagency Network.? Working in a bi-partisan manner with officials in Ronald Reagan and George Bush?s administrations, Gore secured the passage of the High Performance Computing and Communica

Re:Who cares, remember, its Algore

[maynard]

I didn't even vote for Gore. But facts are facts and this ridiculous meme keeps getting posted over and over again. Gore did not lie, or fib, or say anything inappropriate in this instance. Deal. --M

Philippines

[meehawl]

Phillipines is notorious for having Taliban sympathizers that have kidnapped and killed US tourist there.

Don't believe everything your leaders tell you.

Both the Philippine president Gloria Macapagal Arroyo and the Indonesian president Megawati Sukarnoputri have embraced Bush's crusade as the perfect cover for their brutal cleansing of separatist movements from resource-rich regions - Mindanao in the Philippines, Aceh in Indonesia ... the soldiers were not the first to accuse the Philippine government of bombing its own people. Days before the mutiny, a coalition of church groups, lawyers and NGOs launched a "fact-finding mission" to investigate persistent rumours that the state was involved in the Davao explosions. It is also investigating the possible involvement of US intelligence agencies.

MMMMmmmmm??????

[vandan]

An IIS server farm?
Now THAT sounds like a intelligent thing to do!
You make me laugh.

Ebay

[jjp5421]

Does this mean that eBay will be flooded with stolen Opterons?

Re:Who cares, remember, its Algore

[sidmystic]

Let it go, man -- 2000 is rapidly approaching a whole term ago. Al could have, however, framed his response a little differently.

Re:F*#k that

[sidmystic]

And it's a particular hang-up that people like yourself insist on carrying around to this day. Good Lord, I think most people realize that he was referencing his "legislative" lead taken in creating a publicly-available world-wide network.

Sheesh... grow a sense of humor and register already.