Silicon Graphics will start showing off the Altix 4000 Monday, the second generation of the company’s technical computing machines based on the Linux operating system and Itanium processors.
SGI to Show Next-Generation Linux Machines
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
I really hope this is a success for SGI, they are a great company and I just hope it isn’t to late for them. If I had the need for a supercomputer or a ultra high end graphics workstation SGI would be my first choice.
Best of luck to you SGI
> If I had the need for a supercomputer or a ultra high end graphics workstation SGI would be my first choice.
If you need a supercomputer IBM would be a better and a more stable option. IBM does much better than SGI on TOP500 list. As for the high end graphics workstations, well it is soon to become history since SGI is about to EOL pretty much all MIPS/IRIX gear in their lineup starting next year. SGI is pretty much dead as we know it. This Altix crap powered by Linux/Intanic junk is not going to help them a bit. SGI is currently in a very tight niche lacking pretty much any competitive differentiator — pretty much the only piece of technology that gives SGI any edge is NUMAflex, the rest (Itanic and Linux included) are just dead weight. SGI would have done much better if they just held on to the much larger market that used their MIPS/IRIX machines (content creation, 3D, broadcast/fil postproduction, etc.). Honestly I would have been much more impressed if they released an updated version of IRIX than this Altix garbage.
I liked irix better to but I still like SGI, I really hope they pull out of this
SGI would have done much better if they just held on to the much larger market that used their MIPS/IRIX machines (content creation, 3D, broadcast/fil postproduction, etc.).
Now if only someone will go back in time and convince everyone to not ditch MIPS/IRIX. It’s not like SGI choose to dump MIPS/IRIX, no one was buying it anymore. That’s the reason they switched to Itanium/NT (they switched to NT first, remember), if they didn’t switch to something they were going out of business anyways. They might have done better if they had started working on Linux then and released Linux workstations instead of trying to move people from *nix to NT.
Honestly I would have been much more impressed if they released an updated version of IRIX than this Altix garbage.
They already donated all the technology that was worthwhile in IRIX to Linux, why bother updating IRIX now? (I have a feeling this is going to descend into a flamewar where you try and disparage Linux for something or another)
SGI was dead when it became cheaper to buy NT or Linux workstations for design and Linux clusters for rendering.
> Now if only someone will go back in time and convince everyone to not ditch MIPS/IRIX. It’s not like SGI choose to dump MIPS/IRIX, no one was buying it anymore.
That is a load of crap. I was running a fairly large IRIX back in the days when things started to “change” at SGI. SGI could have more than maintained itself if they chose to maintain direction, instead they ran scared shitless from their own products to MS and Intel just because there was too much pro-Windowz/pro-Intel and anti-Unix propaganda at that point in time. Customers could sense the fear and lack of direction at SGI at that time and this is why a lot of customers abandoned SGI (just like our shop, we switched most of the SGI gear to Sun). Again SGI could more than maintain itself if it continued developing MIPS/IRIX and catering to their old mainstay markets, they could have lowered the prices and increased volumes, instead they chose to use somebody else’s technology just because it was hyped as the way of the future.
> SGI was dead when it became cheaper to buy NT or Linux workstations for design and Linux clusters for rendering.
Again SGI simply abandoned the workstation market and did not compete. There was opportunity for SGI, they just did not have any balls to compete and just decided to chicken out whoring themselves out to MS and Intel. For some reason Sun which sold workstations in the same space as SGI is still able to sell their Unix workstations and Sun is still #1 technical workstation vendor out there beating all odds against Windows. The reason Sun is still out there is because they always stuck to their own products and core competencies not chasing after hype and this is why customers love Sun and this is why things are starting to really look up for them. SGI could have done the same.
> They already donated all the technology that was worthwhile in IRIX to Linux, why bother updating IRIX now? (I have a feeling this is going to descend into a flamewar where you try and disparage Linux for something or another)
Ha ha, SGI donations to Linux have been so miserable it is not even worth noting. As for IRIX to Linux comparison, IRIX is still more advanced than Linux pretty much across the board and if given choice I would pick IRIX any time of the day. I’m saying that resurrecting IRIX is going to change things in any way though, it is way too late for that, SGI f*cked it all up too much…
That is a load of crap. I was running a fairly large IRIX back in the days when things started to “change” at SGI. SGI could have more than maintained itself if they chose to maintain direction, instead they ran scared shitless from their own products to MS and Intel just because there was too much pro-Windowz/pro-Intel and anti-Unix propaganda at that point in time.
I’d be scared too if one of my biggest customer groups (movie studios) started switching en mass to Intel/NT and Intel/Linux.
Again SGI could more than maintain itself if it continued developing MIPS/IRIX and catering to their old mainstay markets, they could have lowered the prices and increased volumes, instead they chose to use somebody else’s technology just because it was hyped as the way of the future.
You’re kidding yourself if you actually think that SGI could have kept up with Intel. You’re kidding yourself if you think SGI could have kept up with AMD for that matter. They made the right choice by switching to Intel, they made the wrong choice in going with NT instead of Linux and not pricing their system much cheaper.
Again SGI simply abandoned the workstation market and did not compete.
Well, they tried to compete. Problem is, the market was moving to Intel and SGI tried to sell Intel systems at MIPS prices. Neither their Intel or MIPS systems were worth the cost.
For some reason Sun which sold workstations in the same space as SGI is still able to sell their Unix workstations and Sun is still #1 technical workstation vendor out there beating all odds against Windows.
Sun is the #1 workstation vendor? According to who’s numbers? Dell probably sells more workstations a week then Sun sells a year. Or are you going to come up with some category of “technical workstation” which exclude everything but Sun systems?
The reason Sun is still out there is because they always stuck to their own products and core competencies not chasing after hype and this is why customers love Sun and this is why things are starting to really look up for them.
Yup, that’s why Sun doesn’t sell Opteron systems with Linux preinstalled. Oh, wait they do. Sun is still out there because they haven’t finished thrashing around while dying. And if Sun’s current state is considered “looking up,” what would have to happen for things to look down? Once the shareholders get rid of the poison pill Sun will become a big buyout target. Their executive staff has no long term plan. And their big plan to make Solaris a little bit relevant is to open source it. We all saw how well that worked out for Netscape.
Ha ha, SGI donations to Linux have been so miserable it is not even worth noting. As for IRIX to Linux comparison, IRIX is still more advanced than Linux pretty much across the board and if given choice I would pick IRIX any time of the day.
The only features worth having from IRIX were XFS and support for their various NUMA systems. Both of them were donated to Linux. There is no reason for SGI to spend the money on a large, in house engineering team to maintain their own OS when they can share the costs by using Linux.
> I’d be scared too if one of my biggest customer groups (movie studios) started switching en mass to Intel/NT and Intel/Linux.
It is not like studios started switching to Intel/NT out of the blue. SGI customers and studios in particular would be more than happy to stay with MIPS/IRIX if SGI could proove that they could deliver value and they could if they were committed to it — there are still loads of applications that were primarily designed for IRIX and poring them to other architectures is a true pain in the ass. SGI failed to provide competitive MIPS based products and failed to provide any sort roadmap to proove that they can do it. Instead they decided to stop developing MIPS altogether and strongly hinted at the long-term partership with MS and Intel in particular (this is when the first line of SGI 320/230 workstations came out). SGI themselves proved that there is no future in MIPS and that customer might as well look elsewhere. SGI and Richard Beluzzo being the CEO at that time called the doom on the company.
> Sun is the #1 workstation vendor? According to who’s numbers? Dell probably sells more workstations a week then Sun sells a year. Or are you going to come up with some category of “technical workstation” which exclude everything but Sun systems?
According to IDC’s numbers, you idiot. If you’re idiot enough even not to know there such a category, it does not mean it doesn’t exist. I can spell it out for you — “technical workstations” are now generally considered to be 64-bit workstations used for highly demanding tasks such as visualization, fluid dynamics, etc. Now to my best knowledge there is not a single true 64-bit worksation in Dells line-up, so Sun is really outselling Dell in that space! Same thing applies to other vendors (IBM, HP, SGI). So yeah, Sun is the #1 technical workstation vendor out there. Talk to IDC if you don’t believe me.
According to IDC’s numbers, you idiot. If you’re idiot enough even not to know there such a category, it does not mean it doesn’t exist.
Yea, because I actually care enough to see what IDC’s various categories are.
I can spell it out for you — “technical workstations” are now generally considered to be 64-bit workstations used for highly demanding tasks such as visualization, fluid dynamics, etc. Now to my best knowledge there is not a single true 64-bit worksation in Dells line-up, so Sun is really outselling Dell in that space!
So I was right, it was a category which was rigged to exclude the large scale x86 vendors.
The Dell Precision Workstation 670, 470 and 380 are all true 64-bit systems (Intel EM64T processors). Now that Intel has the EM64T processors out the door I doubt you’ll see Sun on the top of the “Technical Workstation” pack much longer, unless they manage to redefine the category to exclude Opterons and EM64T processors.
Same thing applies to other vendors (IBM, HP, SGI). So yeah, Sun is the #1 technical workstation vendor out there. Talk to IDC if you don’t believe me.
I completely believe you. IDC will make up a “special” category for anyone that pays enough so that they can be #1 in something. I mean, isn’t Microsoft IIS the #1 web server?*
* Where #1 web server is defined as the most used webserver by Fortune 100 companies on their hompage.
> Now that Intel has the EM64T processors out the door I doubt you’ll see Sun on the top of the “Technical Workstation” pack much longer, unless they manage to redefine the category to exclude Opterons and EM64T processors.
Well, we’re starting to really deviate from the core topic, which is SGI, but as a counterpoint to your statement above please keep in mind that Sun is #1 Operon workstation and server vendor at the moment and the prefered vendor by AMD, so chances are that Sun will still be at the top of the heap.
If you need a supercomputer IBM would be a better and a more stable option
Except most of those a clusters and not real super computers. If you need a 1024 CPU single image machine sgi is still the company to call. The problem (from sgi’s point of view) is that these clusters are getting better and better doing jobs which before where the domain of single image super computers.
SGI would have done much better if they just held on to the much larger market that used their MIPS/IRIX machines (content creation, 3D, broadcast/fil postproduction, etc.).
Although those areas where the big ones from a PR point of view and the reason why most people have heard the name sgi, that is never where they made the big money. The real money came from industry and defence who bought huge visulaization systems and 500+ CPU Origin boxes. What really hurt sgi’s bottom line was not losing the content creation business, but the birth of clustering technology, which could compete with their Origin systems for many applications.
SGI had made tons of mistakes recently and while they transition from the SGI we once knew to a very niche Linux company, they behind a world class Unix and a devoted user base. Kind of sad.
i think it is over for sgi. cray might be next.
ibm servers are a better option right now.
-2501
> i think it is over for sgi. cray might be next.
I think Cray is actually more spot-on than SGI. Cray didn’t go the Itanic route and instead chose to use AMD Opteron, which is in my opinion a much better alternative. Perhaps Cray should also adopt Solaris as the OS of choice for their servers — Solaris is an excellent fit for high performance computing taking into consideration that Solaris DTrace, ZFS and Zones could provide phenomenal advantages compared with any other OS out there.
I think Cray is actually more spot-on than SGI. Cray didn’t go the Itanic route and instead chose to use AMD Opteron, which is in my opinion a much better alternative.
It’s not like the Opteron existed when SGI started working on the Altix architecture. Or maybe whoever goes back in time to convince everyone to stay with MIPS/IRIX and bring a bunch of Opteron chips with them so that SGI could use that instead of the Itanium.
Not that the Itanium is a bad choice, it’s neck and neck with Power5 for the most powerful processor available.
Perhaps Cray should also adopt Solaris as the OS of choice for their servers — Solaris is an excellent fit for high performance computing taking into consideration that Solaris DTrace, ZFS and Zones could provide phenomenal advantages compared with any other OS out there.
Oh, now I see why you were making the Linux comments you were before, you’re a Solaris troll.
> Not that the Itanium is a bad choice, it’s neck and neck with Power5 for the most powerful processor available.
Yeah right, may be on some synthetic (SPEC) benchmarks Itanic appears to be a close match to Power5, but in reality it is quite a bit behind, just check any more meaningful database or ERP benchmark. If you factor in the price, the price/performance for Power5 is still quite a bit better than Itanic. Anyway, SGI should be starting to think about dumping Itanic and replacing it with AMD Opteron perhaps — that could really slash the prices that way.
Yeah right, may be on some synthetic (SPEC) benchmarks Itanic appears to be a close match to Power5, but in reality it is quite a bit behind, just check any more meaningful database or ERP benchmark.
What does a database or ERP benchmark have to do with a computer designed for scientific computing? For what SGI is selling the Altix for the SPEC fp benchmark is one of the best available. And there is no good reason for SGI to switch to Opteron, it does not provide the performance that the Itanium does for the workloads the Altix is designed for. People buying a system which scales to hundreds of CPUs aren’t going to be concerned about the price difference that Opterons would get them.
> What does a database or ERP benchmark have to do with a computer designed for scientific computing? For what SGI is selling the Altix for the SPEC fp benchmark is one of the best available.
Well, this is sort of my point, Itanic shows well on SPECfp and LINPAC and that pay be fine for a very narrow band of scientific applications. On anything else Itanic is less than stellar. Which in other words means that by choosing Itanic SGI is painting themselves in to a corner dealing only with HPTC and nothing else, which could be a very tight spot to anyone’s liking. The uncertainty around the future of Itanium could also be a big factor — with Itanium sales doing so poorly for so long no one can guarantee that Intel won’t pull the plug on it or at least stop developing it. With SGI the whole company is riding on Itanic, so if Itanic goes down, SGI will go down with it. I’ve said it before and I’ll say it again — Itanium is a very poor choice for SGI and the sooner they change it, the better they will be.
nice answer,0xbadbeef. that is why i said that cary might be next. cray is doing a lot better but still, cray has to be very careful too. but i agree with you.
-2501
No, really. I’m not trolling and I hope nobody starts. Frankly, my eyes glaze over at the sight of Itanium marketing literature and white papers. Is there an Itanium in my future? I have an AMD 64. Software’s maybe a bit behind the curve, but it’ll catch up. I’m not sure I need another 64-bit proc. Based on what I’ve heard so far, I gather that Itanium is unlikely ever to occupy my personal desktop, so where will it be, if anywhere? Servers? Vertical markets? Does it’s performance live up to the hype? Will it live up to the hype in the future? Are the new SGI’s the ideal use of Itanium? Can I talk without asking silly questions (don’t answer that one)?
From what I’ve read, Itanium’s major market these days is high-end scientific computing, cases where the Itanium’s way of doing 64 bit processing is a big advantage and 32-bit compatability isn’t as important. (It just sucks very badly on running 32-bit code, and I guess price, compared to AMD64?)
The Itanium’s problem is that high-end scientific computing isn’t quite as big a market as they wanted the Itanium to have.
Does anyone know any actual details to correct my impression?
If SGI does end up going belly up, hopefully they will open source IRIX. They’ve released a decent amount of software as open source. http://oss.sgi.com/projects/
OpenIRIX could very well be a reality.
Will I’m not sure other projects such as Linux and the BSDs will have much to gain from it, or if it will have a thriving community like Linux and the one OpenSolaris is starting to gain; I’m sure that there is a lot that could be learned from IRIX.
Oxbadbeef wrote:
well it is soon to become history since SGI is about to EOL pretty much all MIPS/IRIX gear in their lineup starting next year.
rhavyn wrote:
It’s not like SGI choose to dump MIPS/IRIX, no one was buying it anymore. That’s the reason they switched to Itanium/NT (they switched to NT first, remember), if they didn’t switch to something they were going out of business anyways.
I’m not understanding this. Why did SGI spend the time and money to buy and develop MIPS, only to spin it off in 2000? And now to migrate away from it completely?
I’ve always heard that MIPS seriously kicked butt. Whenever someone would tell me about an SGI machine in their computer lab, it was only in hushed tones and raised eyebrows. Almost as if to say, “don’t tell another soul we have one of these firebreathers or I’ll *never* be able to get back on it!”.
Is it just another case of the best tech losing out to the best-marketed tech? (A la Alpha?)
They sure have fallen from their heady days in the early 90’s! They will need more than Linux and Intel or AMD to get out of the mess they are in!