Slashdot Mirror
Silicon Graphics Will Put Linux On Origin
deran9ed writes: "Silicon Graphics plans to introduce a version of its Origin 3000 series computer built around Intel's 64-bit IA-64 Itanium processor running Linux, according to SGI Chairman and CEO Robert Bishop. The current Origin 3000 computers from SGI are built around processors from MIPS Technologies and run SGI's proprietary Irix operating system. SGI has not decided as yet on the name for the new product line. Infoworld article."
Yes, I played long time ago with Irix (on Indy). It's a sweet OS. Great Multitasking, great MP support, great graphics etc...
So what made you think that the PEOPLE who DESIGNED the Irix cannot do it to Linux too?? it's not like they're taking your Redhat and changing it - it's IA-64, a whole new ball game.
Give them credit. If they wrote Irix and done what they did - they can surely make Linux something that no one have ever dreamed of that Linux can be...
Just wait
Hetz (Heunique)
Using my psychic powers, I predict 5 replys to this calling you not-nice things for removing Irix and putting Linux on a MIPS workstation.
They didn't say they were going to replace Irix with Linux on *all* of their machines. They just don't want to port Irix to IA-64.
SGI is becoming just another Packard Bell.
These are my friends, See how they glisten. See this one shine, how he smiles in the light.
Umm... have you played on the SGI 550's?
They aren't proprietary like the 320's were. From tests I have ran, the 550 was better than the Octane and almost as fast as the Octane 2.
It is expensive for an Intel box, and you could probably build one for 2/3 the price... but SGI support is still SGI support, and to many buisness that means something.
-I just work here... how am I supposed to know?
"Only" about twice as many if you beleve comp.arch leaks. I think those were mostly baised off of some parts of the IA64 compiler (mostly done by SGI). At least that was the roumor six months ago. Havn't seen anything else since.
Yes, but it also has way more registers, and modulo addressing of those registers so software pipelining can be used. If you look on comp.arch there are some really impressave code snippits that can make great use of those features. There is also a lot of head scratching about how to get compilers to do a good job cranking them out.
Worse yet there are a lot of code sequences shown where a IA64 follow-on with more functional units runs slower then the existing one because of how the explicit stops work.
Rock on! That's cool.
It is missing things needed to make NUMA systems useful (as opposed to "can boot and run"). For example:
I have no doubt SGI can add those things to Linux, IRIX does (almost) all of that allready.
Normally when being told one CPU only runs at 800Mhz and another runs at 1.4Ghz, so the 800Mhz one is crap, I have a lot of objections. Like "the 800Mhz one may do a lot more work per cycle", or "they could be designed for diffrent markets". However in this case both are for the same market, and McKinley is likely to do more work per cycle. The Itanic is a shammbling disaster, if anyone but Intel was behind it, they probbably would be bankrupt by now.
That said, there is a good chance that both IA64 systems have the same memory interface, making it a useful test run to design multi-way systems around this. It also has probbably been in the works a lot longer then it has been known that the Itanic is a dog.
There is also the chance that Intel is subsidising (or outright funding!) this thing. That makes it less of a risk to build.
Well the current Intel boxes don't have any more CPUs then you can get elsewhere (they do have more memory bandwidth, by a factor of around 3). The O3000 supports 100s of CPUs, if the O3000 IA64 does as well they will at least have a nice story to sell people on. I don't know if it something people are willing to pay a lot for, but time will tell that.
It is interesting that they will port Linux to it. As far as I know Linux isn't tuned to work in a large NUMA (Non-Uniform Memory Access -- local memory in maybe 5ns, memory from a few racks away at 500ns) with 100s of CPUs. It will be interesting to see what they change to make it happy. Esp if they don't go the simple route (treating it as a bunch of total different machines with a fast network).
I don't really see why they don't do that either. It isn't like there is anything special about the IA64 or MIPS that makes Linux or IRIX better on one or the other. Even if they think putting Linux on the MIPS boxes will scare off existing IRIX users, won't porting Linux to the IA64 O3000, but not porting IRIX to it will be even worse?
SGI does lots of good and interesting stuff, but they really are inscrutable sometimes.
Actually, the whole "We are now SGI, not Silicon Graphics" thing was a policy from the much despised ex-CEO Rick Belluzzo. Bob Bishop, the current CEO, interchangeably refers to the company as Silicon Graphics and SGI.
On their OSS pages they have a port to MIPS, even with X-windows on the Indy series. I think SGI is
more committed than even other companies to free software, which is evident if you look at their free software pages.
Engineering and the Ultimate
What's even better is that the page mentions that they will do any necessary development to bring Linux to a robust and scalable operating system if the community doesn't do it beforehand.
Engineering and the Ultimate
Yes they do. In fact, SGI has been a great contributor to the scalability of Linux. Look at their free software pages. They aren't holding anything back. They've release (just to name a few), XFS (journaling filesystem), a high-availability clustering solution, a Linux kernel profiler, an OpenGL toolkit, a system management framework, a Linux kernel debugger, and a host of other things that you can find on their website. Make no mistake, SGI is all about free software these days.
Engineering and the Ultimate
They won't. 7.1 features are pretty much frozen at this point AFAIK.
But please do write to them and ask for it in their next release.
Also, there will be an unsupported XFS-root installer available from SGI that works with the RH 7.1 release, like there was for RH 7.0.
---
High Performance Computing
MFCF has received a gracious donation from Silicon Graphics of an Origin
2000 supercomputer(http://www.sgi.com/origin/2000/). This machine has 8 CPUs and 2 gigabytes of RAM with plenty of disk. Its ccNUMA design offers excellent scalability while maintaining the ease of programming in a shared-memory SMP model. SGI's comprehensive suite of compilers and related
tools are installed. SGI is interested in our help with developing Linux for this platform, and in the meantime, we can use it with its native IRIX operating system. Researchers interested in making use of this high performance computing resource (tuxor.math) are welcome to send their proposals to dabrown@math. Also, MFCF has established a mailing list called "hpc" as a forum for those interested in high performance computing. This mailing list also serves as a
means of circulating announcements from the C3.ca group(http://www.c3.ca/).
I smell victory. The OS skirmish is over. Unix won. Linux will be everywhere that matters.
:-)
Windows will never be able to compete with Linux on the 64-bit architectures. They've already failed to migrate it before. (Okay, arguably, they've never been able to implement one properly on 32-bits either.
The racks get better, wider and faster. Linux grows along with them while NT 4.0 SP4 is an aging, bug ridden, insecure, closed-source and expensive option.
Nobody trusts SP beyond 4 or ME. My employer has one ME evaluation machine while the servers are Alpha's running VMS or NT racks and we have NT boxes on the desktops. (but our clients are starting to ask for Linux for servers and desktops. As soon as IBM ports VisualAge Smalltalk, we go.)
The internet connected desktops need to be biometrically secure, crypto-secure and 32-bits just doesn't cut it. Look for desk-top Alphas, Sparcs, Itaniums and G5s to win that market share too.
MSBPodcast.com The opinions expressed here are my own. If you don't like 'em... Think up your own stuff.
I question SGI's overall commitment to Linux.
They agreed to host our local Linux user's group meetings, and right after that our founder declared the LUG dissolved and went away.
Between the confusion on that and the change of location (they're way off the beaten path here and a lot of people couldn't find the place), we had two meetings with light attendance. This is after years of 30 to 50 people coming, and was during the summer when many of our college-student users weren't around anyway.
Based on these two small meetings, SGI told us they weren't interested in hosting our meetings anymore, because we weren't bringing enough people for them to advertise to.
That's right, their concern wasn't helping the group, it was getting people into a room where they could make sales pitches.
So if they decide next week that they can make more money on Irix than Linux, Linux is gone.
If you're going to deal with SGI, take a page from Ronald Reagan:
Trust, but verify.
-
* Alta Vista (I see that a mountain is back in the logo)
Yes, but that mountain is their debt.
-
They switched OS GUIs from home-grown to Sun-NEWS
to XWindows to keep up with the time.
They toyed with NT, but that couldn't be as
competative as the clone makers.
So why not Linux?
I own a Irix Based O2 workstation, an Indy (R5k based). (I administrate a Origin 3k (only a 4 CPU system, so don't get to excited)
Irix is great, but I believe that Apples OSX is going to eat their shorts as far as graphics workstations are concerned. Unfortunately you don't get a lot of performance boost on a MIPS R12k processor than you do on a PPC-G4. It's the backplane architecture (ccNUMA) that gives the interesting twist to their architecture.
For example, our application pushes a LOT of image data to lots of machines (actually to linux based clusters). I am not using the machine so much for it's 'crunching power' but it's sustainable IO and reliability.
The fact that XFS allows us to scale a growing image database to multiple terabytes without having to ever take the system off-line was pretty surprising.
Also the fact that I can add NUMA bricks as as my business grows is quite interesting.
To make it even more intersting, you can get into a Origin 3000 for as little as 40-50k$. To buy into that scalability with HP or Sun and have the feature set that my applications require was not economic.
To have one monster Origin system for 100k dollars that can scale and be managed with minimal human resources (total cost of ownership) is much better than having 20 linux machines with 4-5 administrators and engineers to keep the things running.
They are neat toys to say the least, but it's crazy. I have compiled some open source software with my O2 and it runs great on the Origin 3000.
Anyway, this is a bit of a rant, but the reason it would be nice to see intel based hardware on SGI numa architecture is the level of scalabilty without having to cluster. Some applications you can have clustering, some it just doesn't make sense.. (Where inter-process communication, backplane bandwidth, etc is important).. that where the Origin 3000 shines.
not every application is going to benifit from such a server. For us a HUGE image store was justification in such a machine. if you are going to run a web server or applications such as that which will cluster, you will get a lot more bang for the buck out of a Linux ia32 system.
Anyway, it's late here and this is a bit of a rant. previous poster is a troll anyway.
Cheers
--------------------
Would you like a Python based alternative to PHP/ASP/JSP?
There is no such thing as "Silicon Graphics" anymore. They changed their name to SGI a good while back.
Surely a multi-CPU alpha system would make a great Linux box - why would SGI use Intel when the Alpha, clock for clock, kicks almighty ass on the Pentium4. A 1.5 GHz 21264 would almost certainly destroy a P4, but nobody seems to be shipping them outside of Compaq's high-end server department.. So theres no real incentive for the price to come down either.
The 21364 and 21464 sound like monster trucks in a world of SUVs from Intel, and i imagine they would kick ass when coupled to a fast OpenGL system and an array of U160 disks.
I'm pretty sure i heard that Alphas have special instructions for MPEG encoding too, but i've never really had the chance to do much with an Alpha.
I gots ta ding a ding dang my dang a long ling long
Apple? Hey, and maybe you'll be able to get 'The world's fastest graphics workstation' in the Flower Power colour scheme too.
All of Apples machines are completely outclassed in terms of clockspeed, memory bandwidth, SMP capability, memory capacity, expandability and flexibility, reliability (can you rack mount a macintosh, or order one with a hot-swap power supply?).
Sure, the G4 is a fast chip, and Altivec is interesting only because its puzzling why you would bother hooking such a fast vector unit up to such a slow memory system.
The G4 certainly has potential, but I very much doubt you'll be seeing anyone who currently depends on either SGI or x86 hardware to move to the Macintosh until MacOS X has proven itself up to the task, and Apple has either licensed another vendor to produce G4s for the server market, or starts producing servers itself.
You can't run a shop on workstations alone, and Apples track record for interoperability isn't exactly stellar.
I gots ta ding a ding dang my dang a long ling long
See: http://oss.sgi.com/projects/
If you read the notes under some projects, you'll see that they already have a mips64 port in-house that they are running today on the O3K which they are using as a testbed for the stuff they want to run on Itanic.
What seems odd to me is that it is pretty clear that Itanic will not be cost nor performance competitive when it finally ships - all the other big boys have said they aren't going to bother with Itanic for anything but 1-4 way type boxes. McKinley (the successor to Itanic) is looking pretty good, recent reports say that it will debut at 1.4GHz around the end of the year (whereas Itanic can barely do 800MHz today).
When information is power, privacy is freedom.
I don't think SGI's hardware does it, so that would need OS support, but other vendors, like HP, support page caching, or at least cache-line levels of page-caching in the hardware.
Page Migration
With a page cache, then you probably do need page migration. But, for a large number of applications, it is sufficient to default to 'local allocation' such that that whatever thread allocates the memory gets it from node/cell local memory. Sure, there are pathalogical cases where that doesn't work at all, but for the general case, it works quite well. I thought I saw in there that SGI's work included local-allocation.
User level hints
Being able to specify location at allocation goes a long way for this, I don't know, but I would suspect that the work going to local-allocation would include the ability to hint for remote-allocation too.
Partitioning and seperate OS instances
I haven't looked at it for while, but I think the current 2.4.x kernel supports linux-on-linux. E.g. you can boot and run a second (or third, or fourth) linux kernel as a process on the first linux kernel. I believe the impetus for this development was to make kernel debugging easier, but it seems like it would be just a couple of steps away from a poor-man's OS partitioning system. Now, if you want fault isolation and online-replacement, you'll probably have to go quite a bit further than the current implementation, but I think Sun with their E10k is the only common box short of mainframes that support this kind of thing today - although most vendors have announced that they want to get there (SGI had their Cellular IRIX project canned a few years back, I don't know what they are doing today, probably nothing because it is more of a feature that the commercial world cares about, than the technical world which is SGI's primary market).
800MHz vs 1.4GHz
In this case, comparing Itanic to current RISC cpus on a frequency basis is pretty valid for floating-point (SGI's market) as almost all current RISC cpus have the same number of functional units as Itanic does (2 FP). Itanic to McKinley is a harder comparison to make because the people who know how many functional units McKinley has, aren't allowed to post that information here on Slashdot due to signing NDAs, but it would be really messed up if McKinley had less units than Itanic does...
When information is power, privacy is freedom.
This is happening on IA-64? Guess that means it will be running the 2.6 kernel...
Intel should call it the Inanium. Few really like the thing, even inside Intel, but their marketing operation is good enough that they're getting quite a number of design wins. Whether it will sell is something else.
What are they using for a compiler, I wonder? The VLIW beast needs an optimizer with near-omniscience to get halfway decent performance.
The university of Waterloo has an SGI origin which they have apparently been given to develop Linux for. Some information can be found here.
Great, the NT workstation had that stupid number nine card as a bottleneck (i.e. you couldn't plug that nice LCD screen on anything else, so you were stucked with a cheezy card) Now it's gonna be the processor, IA-64 is dead on arrival (compare it to alpha, AMD's solution which is way better for backward compatibility + performance, and heck, I'm sure even MIPS got something better. The only way I do see intel's pulling an "intel" out of this one would be with their marketting "lies" again.
Then again both companies needs prestigieous PRs to please the shareholders
--- Metamoderating abusive downgraders since my 300th post.
Howabout 'Gin4Lin?
Hahaha! Yeah... too many. Uh-huh.
O2, Octane, Origin, Onyx
VisualWorkstation 320/540
At least in 1999. I had no problems keeping that straight.
Given this strategy, the rest of the system is very predicatable. They're not going to abandon their other supercomputer hardware technology, so the architecture has to look like an Origin with Itania instead of MIPS. And it's a lot cheaper to subsidize efforts to scale up Linux than to port IRIX. IRIX die-hards will mostly stick with the MIPS systems anyway.
What's new here is yet another round of SGI rebranding. They dropped "Silicon Graphics" so their server customers would stop thinking of them as the Jurasic Park company. They phased out most model brands because there were too damn many of them. There were going to phase out all model brands (one extreme to another!), but somebody realized that just putting a "2" in the model number wasn't enough to distinguish their servers from their workstations.
__________________
they do a better job with these then with the Intel based workstations they sell now. Which while pretty sweet machines when they are working tend to break (hardware) *much* more often than they should. What I would really like to see is a port of Linux to the MIPS based machines though. If SGI tweaked it all out that would be sweet.
Cypherpunks: Civil Liberty Through Complex Mathematics. Those who live by the sword die by the arrow.
Only if they distribute it ,selling machines with the code installed on it would count, besides SGI has shown many times that they get the GPL and like it. Look at the code that until recently they where sending back to the Apache project. You won't be able to afford one of these for awhile if you can't afford it now. The Octane under my desk still costs in the 10s of thousands of dollars range.
Cypherpunks: Civil Liberty Through Complex Mathematics. Those who live by the sword die by the arrow.
I asked this of our rep a few months ago, asking about the port work of Linux to Mips. What he said was that Irix has a buch of special magic going on in conjunction with the Mips CPU. i.e. make sure code gets to the cpu through at the same time, special read ahead things, tweaks with the bus and cpu, etc. to put them into a Linux system would fundamentaly change how things are currently done in most of the kernel . Think how hard they've been trying just to get big memory support into the kernel, could you imaging trying to get a completely different core change that would only help Mips cpu's.
They are doing porting work, and Linux is running on Mips currently (so is NetBSD), but the special performance enhancements aren't in the kernel. Currently running Irix on a Mips Origin will be much faster than with Linux, but on the smaller workstations O2, Indigo II, etc. Linux will be faster (for non gui work) because it's so light-weight.
Don't plan on getting a speedy Linux on your Onyx, Origin systems anytime soon, unless you want to take a big performance cut.
SGI's daftness is beginning to be offensive. Linux fails pimping pixels on IA32 to WindowsNT, where Linux would claim to be a better OS. Now we have a 64bit proc and a 64bit OS, written by the Linux people, and we have Win2k for 64 bit procs. I'm betting Win2k will pimp pixels faster. SGI will always fail on a platform that is not IRIX, and is not proprietary, because you can get anything other than both of those for less.
--
SGI has not decided as yet on the name for the new product line.
Maybe they should have a contest, with a free one to whomever comes up with the best name? Here are my suggestions for an "Origin" box running "Linux":
We could have a lot of fun with this... Any other ideas?
...about as good an idea as changing your logo from an awesome three-dimensional cube, to some lame initials.
I realize you Linux fans are happy... but we SGI fans are crying right now!
"And like that
--
A feeling of having made the same mistake before: Deja Foobar
--
A feeling of having made the same mistake before: Deja Foobar
Personally though, I suspect this is yet another attempt at pulling their heads above water for a little while.
SGI used to be about innovation and advanced technology, and seriously high-end graphics. Now their relegated to making servers, and ones based on foreign, Intel chips at that.
I love Linux and I love SGI - but none of this will be what gets SGI out of the gutter.
Interestingly, there's a hotbed of activity in the Apple/G4/Mac OS X world that the Mac might be the platform to watch for 3D. Without OS X though, the G4 lags pretty far behind the Pentium in performance, as per a recent CGW article.
Apple at this moment, beleive it or not--at least in terms of price/performance. They have a 128 bit vector processing unit attached to their processors, that provides many software programs with now 7 gigaglop performance. They are banned from export because they are classified as supercomputers, unlike intel crapola. Their OS was way behind until OS-X, which is a UNIX operating system. As far as brute force, Sun is way ahead of SGI. Check out their Elite3D graphics attached to their SunBLADE 1000 (not the 100). Expect to pay over $10,000 for this beast, but you get to have an UltraSparcIII processor that has been in development for many long years. All Sun's stuff is 64 bit, and has been for a long time. Sun had a lead over PC performance with their 400mhz 63 bit Ultrasparc processor, which outruns 1ghz Intel Crapola. Now their stuff runs at 900mhz. Sun has special on chip instructions that are similar, but superior to MMX and for the 64 bit world. Im sure you really dont give a crap, you're not worried about the best "Graphics Workstation" Im sure all you want to do is play games. In that case, Wintel is certianly superior with its vast array of games.
- Coke (eventually removed stupid gray stripe from ribbon)
- NBC peacock
- Alta Vista (I see that a mountain is back in the logo)
I think that