Slashdot Mirror

And it's being done, with the Robocup league. According to This site, the goal is to have a fully bipedal team of autonomous robots go head-to-head against the winners of the year's World Cup by 2050.

Mine is bigger than yours :)

Linux XFS: 9 exabytes

Also supports extended attributes.

It's nice to see some SGI support so my eight Octane2 and twelve Fuels are more than a big doorstop.

It's nice to see some SGI support so my eight Octane2 and twelve Fuels are more than a big doorstop.

How will the 64 proc model compare to the new SGI Altix 3000? 2. Is anyone (now or planning to in the near future) scaling the Itanium2 up to that level? I have not heard mention of a 64 proc I2 production system ...

The SGI Altix 3000 is an Itanium2 system that scales up to hundreds of processors (currently up to 64 per node).

Check out some of the equipment from Sun Microsystems, SGI, IBM, and Stereographics.

A bunch of their equipment is designed for a 10 year obsoletion-cycle. Cost's a hefty penny, though. Designed for business and major research universities.

At the University, we were using Creator3D graphics cards from Sun Microsystems. That was in 1999, and the general consumer market still hasn't caught up with that tech. Me, I'm still looking around for auto-stereoscopic monitors. Sharp is coming out with a consumer model next year, I hear.

You obviously need to look better at how Linux scale on 8P machines and more before making such statements

If linux scalability is really an issue, beyond 8 processors, then i guess that the SGI Altix 3700 is just
vaporware/gare.

I suggest that you read the following articles that debunk the myth of the 8processor barrier :

SGI Busts into Linux with 64-Processor Scalability

NEC Calls Dibs on Breaking Linux Eight-Processor Limit

I personally hope that these benchmarks can be run against more recent kernels and a full description of optimizations and patches used disclosed.
Considering that SGI is using a [somewhat] standard 2.4.19 kernel to scale this well , I am certain that the results will be much better.

I was also suprised to see that they still use some of the old performance monitoring tools like looking at /proc, and other ascii tools, rather than something like PCP that collects all these statistics together so that you can look at any combination of subsystems on the same time line. Then they could have graphs showing the interraction and load on the disk, cpus, vm, network etc.

oh please, here's a nickel, go buy yourself a real computer ;-)

If you look at their developer platform for this machine, you see Fortran, C++, and C listed. No Java.

Just a thought for all the Java folk who got so defensive about my comparisons of their language to others. Java is a useful, powerful tool -- but if you want to develop for top-flight parallel hardware, you don't use Java.

I watched the Quicktime keynote with great interst, hoping that Jobs would finally introduce a 4-pound notebook. I've been waiting for one for a while, so I'm really excited that Apple finally introduced one!

Unfortunately, however, the notebook doesn't include DVI-out support, so my monitor would fall back to VGA mode if I tried to use the notebook with it. Does anyone know if Apple or a third party plans to offer a PC Card with DVI support? Margi had one, but it's only 4MB... not quite enough for this particular monitor.

Also, one thing Apple keeps failing to address is the #1 reason I haven't switched to a Mac. Steve, where are the software trade-in incentives? I own Photoshop 6 and 7, Dreamweaver MX, and Microsoft Office XP for the PC. What on Earth is keeping Apple and/or other vendors from offering trade-in incentives? Why can I not trade in my two boxed Photoshop-for-PC copies and receive Photoshop 7 for Mac OS X? The same goes for Dreamweaver MX. The cost to move to a Mac is almost doubled by the $1500 worth of software that I already have for my PC.

Here's hoping Apple will start to address this issue, especially since the platform is geared toward video developers and graphic designers -- two markets whose people invest heavily in expensive software.

Now that I've RTFM, let me correct my previous comment -- The Altix3000 runs a single Linux image over up to 64 processors and 512 GB RAM. After that, it's NUMA.

It can, however, do high-speed shared memory over all nodes in the cluster, allowing you to store HUGE shared data sets. Here's a link to the info on the memory.

32% Government & Defense
28% Science
21% Manufacturing
12% Media
7% Energy

http://www.sgi.com/company_info/investors/presenta tions/shareholder_meeting_2002.pdf

Cause on IRIX you can run that super-cool file manager from Jurassic Park. Why would you want to run that super-slow piece of dog shit Konqueror or Nautilus anyway?

Agree on the cube logo. However, they still have it on their website. See it spinning.

I'm not sure if that's true. Looking over SGI's website, they don't seem to sell ANY linux based workstations any more. Only the Fuel and the Octane2 (both IRIX/MIPS machines.)

They do have a yet-to-be-released NUMA Linux system based on Itanium, but it probably shouldn't be thought of as a workstation.

I'm guessing you're probably right though that "SGI barely sells Irix machines". Not sure how many they're selling, but they're still cettainly losing money.

Moving a renderfarm to a Linux cluster isn't surprising. Since rendering is an "Embarrassingly parallel" computation and AMD/Intel has more FLOPS/$ compared to the MIPS processors, this is expected. When you need to pass a lot of data between processors, you'll need one of those Origin 3000 servers with 1000 processors. Linux can't do this yet.

What is interesting though, is that they moved the workstation applications from SGI to Linux. I didn't know that the SGI hardware was lagging behind that much.

LOL! Google doesn't offer much proof. Your very search query is flawed.

"nvidia tainted [bug]" also gets reports of "not tainted" kernels with lspci output containing nVidia. Not to mention leaving out other bug-inducing items like patched kernels.

Right:

If I try to use NVIDIA's modules (the latest available are 1.0-2960), I get a
kernel BUG at filemap.c:236 in...

Look, this is well known. NVidia's driver does memory managment, takes various locks and such. It is sensitive to changes in the core kernel. If NVidia provided the driver in source form, Linux developers would keep the thing maintained, but they don't, so it keeps breaking.

Try real research next time buddy.

Trying pulling hard, and that foot may suddenly pop out of your mouth.

The benefits of Open Source or Free software to its users are undeniable. If the software has a bug, or the software does not do something you want it to do, you can change it. There are many advantages, and they have been explained at length by various people. If you are going to be using software, you are definitely better off if you have access to the source code.

Trust

The fundamental difference between open source software and closed source software is the level of trust required. For a business to use closed source software, the level of trust required is enormous. It is not simply a question of whether the money spent purchasing the software is a good investment. The time invested using the software is far more significant. Almost inevitably your own business information becomes tied up in a format that is specific to the software you are using. In order to buy software from a closed source company, you have to take the following on trust:

• They have not left gaping security holes in the code.
• They will fix bugs in a timely manner.
• They will eventually add the features you want.
• They are not using your computing resources to do things which are not in your interest.
• They will not increase the price unreasonably once you depend on them.
• They will not go bust.

Business Models Having access to the source code makes good sense to the users. However the business case for the software vendor is far less convincing. In fact, the dangers of closed source from the user's perspective can be considered opportunities from the vendor's perspective.

The open source foundation proposes "4 ways to win" which is reproduced here: Four Ways To Win

Now for a higher-level, investor's point of view. There are at least four known business models for making money with open source:

1. Support Sellers (otherwise known as "Give Away the Recipe, Open A Restaurant"): In this model, you (effectively) give away the software product, but sell distribution, branding, and after-sale service. This is what (for example) Red Hat does.
2. Loss Leader: In this model, you give away open-source as a loss-leader and market positioner for closed software. This is what Netscape is doing.
3. Widget Frosting : In this model, a hardware company (for which software is a necessary adjunct but strictly a cost rather than profit center) goes open-source in order to get better drivers and interface tools cheaper. Silicon Graphics, for example, supports and ships Samba.
4. Accessorizing: Selling accessories books, compatible hardware, complete systems with open-source software pre-installed. It's easy to trivialize this (open-source T-shirts, coffee mugs, Linux penguin dolls) but at least the books and hardware underly some clear successes: O'Reilly Associates, SSC, and VA Research are among them.

In fact, the number of companies that have had success with any of these models is miniscule. This is hardly surprising, they are simply not very good business models for software companies.

Taking each in turn... Selling Support The better documented and more reliable the product is, the less support it needs. A business model where the more perfect your product, the less money you can make has got something fundamentally wrong with it. Loss Leader The very fact that this can be advanced as a viable business model for OpenSource shows desperation. What it comes down to is an admission that the best way to make money from software is by selling it. Widget Frosting This makes perfect sense if you are a hardware company, or when the software is a side issue. However, its no use at all for a business whose main product is software. Accessorizing Selling accessories is fine, but there is no pressing need to actually develop the software when one is in the accessories business.

There are of course other business models for Open Source. For instance, the one adopted by the Perl foundation and several others is begging. This is not a business model that many companies would find appealing though.

The basic problem is that for a business whose primary function is to make software, then the primary reward has to come from selling the software. We need a business model that actually works and we have one, it's called capitalism. It works like this: make something that people want and sell it to them. This model works for software too, and there is no reason why this model cannot work even when source code is available. Closed source vendors are relying on something a little closer to the business model of a heroin pusher. It starts off like capitalism, but there is the added feature that the user gets addicted and has to carry on buying the same thing even if he does not really want to. The more he uses the same vendor, the more reliant he is upon it.

The Solution Community software is software where the vendor can be paid a fair price for the software he creates, but where the buyer does not end up in a similar position to a junkie.

Community Source is software that guarantees the following:

1. The right to see what the software is doing, ie access to unobfuscated source code.
2. The right to add enhancements.
3. The right to fix bugs.
4. The right to sell his enhancements to other companies. This does not mean the right to the sell software without the original vendor receiving any money. The buyer still needs a license from the original vendor, but he does not have to rely on a single vendor for upgrades and enhancements.
5. The right to buy enhanced versions from 3rd parties.

IIRC one of the biggest advantages of stippling in rendering surfaces is that you can get a fast simulation of transparency. Check out here. So maybe the same applies in 3D. The 3D stippling might allow you to simulate complex semi-transparent volumes - perhaps also avoiding some z sorting or alpha blending.

Also, maybe you WOULD see more of this in games if it could be done in real time. Just because all we have now is polygons doesn't mean that's the way it has to be.

It runs on Irix and Massive is being ported to Linux. Quote: From the beginning of preproduction, Weta Digital has also used the IRIX OS-based Octane visual workstations to write extensions to Maya and create proprietary technology. This technology includes Massive, a custom-built crowd animation or "artificial ecology" system developed on IRIX and now ported to Linux that draws from a huge database of motion-capture data. (see here).

I bet Fresco will be finished before Xrender has image transformations, true hardware alpha channel, etc.

Why bother with XRender when there's the proven GLX?

No idea. You can learn more about sprocs by starting with the man page, here.

Maybe someone can describe the bricks?

Yeah, maybe.

Here
Maybe someone can describe the bricks?

You want to see images: http://www.sgi.com/features/2002/nov/hpc/images/lg _origin_3900_out.jpg

I think that sgi has had this going on for a while now...
http://www.sgi.com/visualization/van/

These days, I still have it for strangely similar reasons. I wouldn't say it's my main OS, as I have it solely to run games without a linux version (yet -- nwn should have a linux client any time now). But since I mainly use my home computer for games, well...

Oh, and I also have an X-Box, which, in a certain way, Microsft pays me to use. I'm okay with that.

If I add up my computer time, then I spend most of it on this guy.

SGI dropped Intel after the PIII

SGI finally saw the light. They returned to the MIPS architecture after failing to compete successfully on cost with their Intel "solutions."

None of the OSes (only Windows versions) it works with are certified for TOP SECRET data

Yeah, and Trusted Solaris, and Trusted Irix, and a bunch of other OSs you've probably never head of. Look at this if you don't believe me.

When I first read the post, I got pretty excited. Dreams of cheap clustering for scientific applications danced in my head. No more need for Myrinet, no Dolphin, just Firewire and Beowulf!

Then, I read some performance metrics on Firewire. High bandwidth. High latency. Doh! The fairies stopped dancing for joy.

The problem is that in scientific computing, the time it takes for one node to say I need that data to another node, and actually get that data determines the performance of many more apps than does the speed of the CPUs.

So, until a cheap, low latency solution for communications comes by, real clusters will be communicating over Dolphin, Myrinet, or some other propietary technology.

Tony

Anyways, what I'm trying to point out is that it is actually becoming very convinient to build a super computer with lots of PCs that just lie idle. I am not sure if Saddam has heard about cheap linux systems. But what if he could build a super computer cluster?

Well, it depends. A Linux cluster is a good way to render a movie, because you can easily parallelize that task - send a frame to each node you've got, wait for it to come back, send out the next one, then when you're done composite them into an animation. That's easy, because you can make each task essentially stateless. For example, you don't have to wait for frame 1 to rasterize before you know how to light frame 2.

But in many scientific computations, there is a limit to how you can subdivide a task. Say you are modelling the movement of a gas in 3 dimensional space, you cannot partition your space 3x3x3 and send it to 27 compute nodes, because what happens in each partition both influences and is influenced by what happens in adjacent partitions. If you did try to do something like this on a cluster designed for rendering movies (or brute forcing a cipher, or serving web pages) performance would be terrible because of the overhead of communication between nodes. For that, a Single System Image machine has a vast advantage.

So the question is (and I don't know, I didn't study nuclear physics beyond A-level), are the significant computational problems associated with the development of nuclear weapons easy to parallelize, or do they require a real supercomputer?

The Japanese made an order of magnitude increase in processing power and you think this toy from SGI is leading edge? LOL

SGI will happily sell you a 512-processor machine if you want one. The innovation in the 3900 is compute power/m^3, not raw power. It just so happens that the Origin 3000 has got the raw power too.

SGI is rapidly becoming the transmeta of super computer manafacturers. There product fills a very small niche, yet all the stupid kids like you think they're so neat.

Don;t write them off so quickly. There are plenty of things that only an SGI can do. There are circa-1993 Indigo2's still on people's desks (being used for things like Gladiator), because even a 2002 PC can't do some of the things they can do. SGI are a niche vendor, true... but so is Mercedes.

As an aside the open critical component in a supercomputer is memory, fast memory, whith out that it matters not a jot how quickly your processors work. So what is the memory bandwidth of this baby?

Put it this way: internal bandwidth in an SGI workstation is 3.2Gb/s. Can your peecee do that?

POV-Ray has a deathly slow renderer though.

If you're interested in comparing some renderers, check out the Internet Raytracing Competition (IRTC)

POV-Ray is good for a number of things, though I would recommend using a modeller if you plan on rendering anything 'organic'. (Defining bezier patches by hand, while possible, is not reccommended.) Animation support has come a little further lately, though mainly you're going to be rendering a number of stills with some computations in the scripts varying based on a clock value which gets passed in.

Oh, and here's one big advantage -- the source code is freely available! Hack to your hearts delight, just follow the rules if you plan to distribute. Binaries are available for a number of platforms, and the generic UNIX source should be fairly easy to compile. Note: If you're using IRIX, check out SGI's freeware site for a binary. I've had fairly good luck compiling it on Solaris, but I don't have the dev libraries for Irix so I use the precompiled version.

If anyone does end up using POV, please at least give a thought to donating a couple of bucks to them, I'm sure they could use it.

Perhaps SGI workstations running FSN? :) FSV for other unixes...

The SGI 1600SW. It's getting old, but it's still the king.

why would you want to buy your hardware and software from a specific vendor? Sure it might work better

You answered your own question. Please see another product solutions manufacturer for more details.

When Linux runs on the smallest devices all the way up to boxen like this 10 tera-FLOPs beast and this SGI supercomputer that just set a memory bandwidth speed record with 120GB per second (faster than a Sun SunFire 15K, Cray C90, IBM p690, etc) on a single system image? Scale it any which way you like ;-)

Rock on with your bad self.

If you want performance you've got to know where to look.
Unless you're playing games, real workstations blow away the fastest desktops.

There are a great variety of cross-platform libraries and frameworks that you would find useful. For example, for a humble JPEG coded, the Independent JPEG Group's JPEG library works really well and runs on everything from DOS to a Cray. It is portable beyond belief. For a lossless graphics format there is libtiff. (I don't know what's available for cross-platform video format software, but I'm sure there is some.)

If you're going to write in in C++, my favorite framework for GUI, file API's, TCP networking, multithreading and database is ZooLib. (But note that presently the best code to use is what's in CVS because it hasn't had a release in a long time (Real Soon Now, really!).

I've started writing a book about ZooLib that is released under the GNU Free Documentation License.

But if you don't like ZooLib, it's important to use some cross-platform framework. There are many to choose from.

Another important framework is the Simple DirectMedia Layer. You will want this for your rendered scenes and for sound (ZooLib does 2-D graphics, it's not a 3-D API).

If you write in C++, do as much as you can with the Standard Template Library. There are some excellent books that teach how to use it.

For a long time, the STL has got a bad rap, in part because the template definitions in the header files are hard to read, and in part because of poor compiler implementations of the C++ ISO standard, or poor implementations of the library itself. But by now there are excellent implementations for every OS that is in common use. For example, on Windows, don't bother with Visual C++ - use Metrowerks CodeWarrior or Comeau C/C++.

Even if you choose to work with a broken compiler, the STLPort library provides a compliant standard library that will work almost anywhere.

I was rather intimidated by the STL when I first encountered it but once I got a good book and learned how to use it, I thought it was the best thing since sliced bread.

Boost has many portable C++ libraries that are of excellent quality.

Finally, I am (slowly) building a website devoted to educating developers in cross-platform and portable programming called ByteSwap.net. Read my first article there Writing Cross-Platform Software - Getting Started. More articles will appear when I get more free time!

Please stop sending messages that do not have any support. Obviously you're talking without having the slightest clue.
For your information, if you go to www.sgi.com, in just one click (on Products/Servers) you can see how mistaken you are.
According to the webpage http://www.sgi.com/servers/ the SGI Origin 3000 have "Up to 716 GB/sec" internal bandwitdh. How do you compare that to the 1GB/s of your PC?

I agree, the Origin 3000 is not a graphical workstation, but SGI's focus shifted from graphical stations to supercomputers.
You were comparing PCs with SGI workstations, which indeed do not have a large technical superiority over PCs nowadays.

Please stop sending messages that do not have any support. Obviously you're talking without having the slightest clue.
For your information, if you go to www.sgi.com, in just one click (on Products/Servers) you can see how mistaken you are.
According to the webpage http://www.sgi.com/servers/ the SGI Origin 3000 have "Up to 716 GB/sec" internal bandwitdh. How do you compare that to the 1GB/s of your PC?

I agree, the Origin 3000 is not a graphical workstation, but SGI's focus shifted from graphical stations to supercomputers.
You were comparing PCs with SGI workstations, which indeed do not have a large technical superiority over PCs nowadays.

.haeger

.haeger

.haeger

.haeger

Just download the tardist and add it as another distribution. freeware.sgi.com rocks. ;-)

Show me where on the Irix support policy page it says that it's not supported, and I'll believe you.

so all of you fan boys who say "oh my $900 dollar linux boxen is as good" can shut the hell up cause you have obviously never layed your hands on a real workstation.

if i could afford the price tag, there would be no way that i would even consider buying a mac or a pc, i would go straight to SGI, and im seriously thinking about taking out a loan for an SGI fuel.

but anyways, relevent links are here
sgi octane 2
sgi fuel
studiotools

so all of you fan boys who say "oh my $900 dollar linux boxen is as good" can shut the hell up cause you have obviously never layed your hands on a real workstation.

if i could afford the price tag, there would be no way that i would even consider buying a mac or a pc, i would go straight to SGI, and im seriously thinking about taking out a loan for an SGI fuel.

but anyways, relevent links are here
sgi octane 2
sgi fuel
studiotools