Slashdot Mirror
SGI launches R16000
nkrgovic writes " SGI has just launched a new CPU - the long expected R16000. The new CPU works on 700MHz, has 4MB secondary cache and more goodies.
For now the new CPU is only used in SGI's Fuel workstations, but we should expect to see it pretty soon in SGI's Origin servers as well. With new high density compute nodes this should make the Origin's the fastest supercomputing server per square foot."
So fast, it helped me get first post.
I'm confused. I thought SGI was dropping support on IRIX. Why are they releasing new Irix boxen?
Information doesn't want to be anthropomorphized anymore.
I'm confused. I thought SGI was dropping support on MIPS, and switch to Itanium. Why are they releasing new MIPS processors?
The new CPU works on 700MHz
taking a page out of apple's playbook I see..
congratulations on breaking the elusive 500MHz barrier!
*chuckle*
it doesnt run linux nor can it be imagined in a beowulf cluster.. but in soviet russia, r16000 launches sgi!!!
I thought enough material had finally invaded the net for people to realize Mhz means nothing... I guess I was wrong.
Let's play what if... cause I don't have any facts on this processor: What if the mov operation of said processor is 1 cycle, whereas mov of pentium is 7?...
Where does that put you?
Books are written on CPUs. pick one up, and you'll understand Mhz means nothing.
processor performance has never been SGI's strong point, except for breifly after the R3000 and R10000 were introduced.
SGI's workstation line is largely unimpressive, especially for the 99% case of computer users, hell, even engineers.
The problem is, for a small set of jobs, for a small set of people, nothing else is suficient - at any price. You're either using an SGI, or the work isn't taking place.
That market is continuing to erode, but i dont think it will ever dissolve completely. I think eventually SGi will effectively become a US govt subsidized entity. SGI continues to build the systems that only governments need and only government agencies can afford.
Clustering has nothing to do with the markets SGI sells in. Please don't mention it, it makes me think you don't know what you're talking about.
Do you ?
My opinions are my own, and do not necessarily represent those of my employer.
yeah, but i believe that the R16000 is a vector processor whereas the desktop proc's are scalar and don't have the same processing power let alone used for the same tasks. I think i read somewhere that Cray/SGI was going to be using this processor in one of their upcoming Supercomputers that's supposed to rival the Earth Simulator, which uses vector processors by NEC.
I'll be running my P4 at 3 Gigahertz, thankyouverymuch.
;)
True, but your architecture still sux
"Backups are for wimps. Real men upload their data to an FTP site and have everyone else mirror it." -- Linus Torvalds
Uh, can I have that in libraries of congress, please? (Or at least cubic foot of server space / "per 1U rackspace").
The Playstation also used a R3000A MIPS processor. You really don't know what you're talking about.
sgi has been caught up in the Intel fashion du jour by bringing out Intel workstations. Had they not done this, they would have fared much better, witness Sun and Apple.
Another great example of this is Intergraph, they bought the Intel crap as well...
btw, many Intel systems suffer from performance issues that are just there because of the 20+ years compatibility built in. Something 'proprietary' systems usually don't suffer from.
Remember, it's not just megahurtzen that counts!
I guess that makes it faster than my car. :)
Linux is only free if your time has no value. Windows is only free if you threaten to use Linux.
There's also a thing called efficiëncy. At my job we make use of a lot of Alpha's ranging from 500 to 700 MHz. Those speeds look outdated if you're used to intel, but the workload these babies pull is impressive...
sigh clock speed isn't performance. Work done per unit time is performance. Work done per watt is a plausible performance metric as well.
I don't know that the 16000 is the cat's PJ's, but complaining that the CPU clock speed isn't as fast as some other processor is --- well, naive.
Yeah.... Only that you forgot things like the R8000, which was the fastest FP processor when introduced, the R4000 which was the first mass production 64bit micro (true 64bit), etc... etc....
And then you finish with this gem:
"Clustering has nothing to do with the markets SGI sells in. Please don't mention it, it makes me think you don't know what you're talking about"
So please do apply your own advice first... and shut up.
32% Government & Defense
28% Science
21% Manufacturing
12% Media
7% Energy
http://www.sgi.com/company_info/investors/presenta tions/shareholder_meeting_2002.pdf
As far as I know, most(?) of the rendering farms they've been using in Hollywood lately have been running Linux.
And as another poster already mentioned, playstations also use MIPS CPUs. Including the PS2.
Did you ever consider that perhaps Sony and Nintendo didn't use the top of the line CPUs? Or why do you think they can sell a Playstation 2 for $200 when SGI workstations run at what? $5000 and up?
Professional journalists investigate their facts.
The word "also" shouldn't be in the above response. Typo. The N64 uses something like a R4300. The difference in graphics quality had more to do with storage. CD-ROM verses "64 Megabit" (8 MB) cartridges - with 32MB cartridges coming out later on it's lifespan.
How many people use SGI workstations anymore anyway? The specs for the new chip sound good, but what about application support? Its no use if this is a fast server, if my apps don't run on it. It seems to me that SGI is used mainly in the graphics industry, and there it seems to be losing ground quickly to Linux and BSD based solutions.
more about me
Sigh...
Come on people. You all root for the Athlon when it is clocked well under the P4, yet you believe that SGI's MIPS line is crap when it tops out at 700Mhz???
Sun's UltraSPARC III Cu tops out at 1.05Ghz last I checked. Does that mean that the P4 at 3Ghz stomps the hell out of it? If you said yes, you are a fucking idiot.
People, the Unix world is far far different from what you are used to in PC land. High speed backplanes, dedicated busses, huge amount of L1 cache, insane L2 cache, incredibly efficient cpu designs (where 1 clock per instruction is pretty much the norm and cache misses don't occur every 3 operations), hot swap damn-near-everything, upwards of 72 processors and 288 GB of RAM...
It all adds up to a fucking badass machine that smacks the piss out of any PC on the planet when it comes to getting its job done. Don't compare apples to oranges. The applications these machines are designed for do not include Quake 3. The benchmarks you have memorized don't mean a damn thing in this realm, so go back home.
Getting back to the article, I'm glad to see SGI coming out with a new CPU. I still see a few SGIs in the wild now and again. If they lock down Irix a bit more security wise and expand their target market, they might be a decent competitor for Sun within the next 10 years. I don't see them winning any shining star awards right off the bat, but if they are persistant they'll do alright in the long run.
You really cannot compare a 700Mhz MIPS chip to a 3000Mhz x86 p4.
You must remember, the R16000 is 64-bit, not 32-bit.
Also, it has 4000k of L2 cache, not 256k or 512k.
Also, out-of-order instruction execution, x86 chips can't do this.
you are trying to compare two things that are completely different.
Alcohol & calculus don't mix. Never drink & derive.
I worked all summer in an all-SGI shop.. And I call tell you how far behind they are. The place where I work is specialized in HPC, so when they started in 1992, SGI was probably a pretty good choice, but now for workstation, I wouldnt say its overkill, I would actually say that its underkill. We made a benchmark comparing an SGI Origin and a linux Ahtlon cluster, the athlon needed only two nodes to beat the origin and with all 16 nodes where about 10 times faster... SGIs are just overpriced, for 99.999% (that's 5 nines) PCs can do the job and even do it better and especially do it much cheaper. So their workstation market is being destroyed from under them.
On the other end, their HPC (super-computers) is being attacked from above. On that sector, price is not really a problem, its just pure performance. And there too they are being beaten, SGI just does not have the research power that
NEC or IBM can have. So they are starting to be pretty much behind, so they become not only more expensive (which does not really matter), but more importantly much slower...
Also on the workstation market, their desktop SUCKS, its just a pain to use. They are still stuck in the pre-win95 era... It might have been good compared to win3.1 or twm, but it just is not in the same world as GNOME, KDE, WinXP or MacOSX.
Also, their other strengh where there graphics board, they invented modern 3D hardware. And for a long time the roadmap for the PC 3d hardware was simple, they just had to do what SGI already had, but we have now passed a point where the PC hardware has actually more features then the SGI stuff. The only difference now between the pro and game markets are the amount of ram/cache and those "pro" cards exist on PCs. They do cost $ 2000-3000, but they are nowhere near the cost of the SGI workstation that includes them...
SGI has no future. They have been losing money for years. I have been thinking for quite a while that they where a good target for an acquisition, but now that MSFT has bought much of their patents. It might be cheaper to wait for them to go bankrupt and to pick up the pieces. They where in a fast playing game and they have gotten slow.....
Some of our customers' highest-end workstations are SGIs. (There are also a couple Sun Blade 2000s and a LOT of IBM RS/6000s.) SGI will continue selling IRIX and the machiens it runs on until they no longer make money doing so.
- A.P.
"Remember when the U.S. had a drug problem, and then we declared a War On Drugs, and now you can't buy drugs anymore?"
Wha?
You, sir, are almost completely uninformed. The R16000 is an R10000 variant, just like the R12000 and R14000 before it. It is not a vector processor, and has no vector units. The R16000 is, furthermore, a desktop processor in its own right, because it's currently being used in the Fuel workstation.
Incidentally, SGI divested itself of Cray some time ago. Cray was bought by a company called Tera Computing, which then changed its name to Cray. They're building the SV2 vector supercomputer now, using their own processors, and they also have an arrangement with NEC to market the SX-6 in the United States with a Cray logo, but that's strictly a resale agreement.
I write in my journal
SGI workstations run at what? $5000 and up?
$10,000 and up, but who's counting?
I write in my journal
Instead of everybody saying "GHz doesn't matter, dummy" why doesn't somebody quote some real benchmarks? I poked around on the web a bit and all the benchmarks I can find either (1) are out of date, or (2) show Alpha, Intel and AMD blowing everybody else out of the water.
In my experience SGI's are slow but are extremely scalable. With IA32-based machines you'd be lucky to get 4 CPU's sharing memory, unlike the 64+ you get from SGI. Very good for scientific codes but not so hot for applications that are either not parallelizable at all, or embarassingly parallelizable such as Seti@Home or ray-tracing a feature film.
You must remember, the R16000 is 64-bit, not 32-bit.
For the record, the R10000 series can run either 32-bit or 64-bit code. All other things being equal, the 32-bit version of a program will run faster than the 64-bit version; you can fit more 32-bit ints into cache at once than 64-bit ints, so the 64-bit version of a program generally suffers more cache misses than its 32-bit counterpart.
On an SGI box, you don't compile for 64-bit unless you absolutely have to address more than 2 GB of virtual memory.
Also, it has 4000k of L2 cache, not 256k or 512k.
That's pretty puny for an SGI. The processors they use in the Origin servers have typically been equipped with 8 MB of secondary cache; the 4 MB version must be just for the workstations, to keep costs manageable.
you are trying to compare two things that are completely different.
On this point, however, you're 100% correct.
I write in my journal
I'm sorry to disapoint you.. but I have no problem agreing with you that the clock speed is not all... But its still important... On our CFD (Computation Flow Dynamics) the kind of thing that SGI super-computers are made to handle.. Our el-cheapo AMD Athlon based cluster kicks the ass of pretty much every SGI in the data-center where it is.. and I think it even kicks the ass of the NEC... So yes, I'm sorry but 3Ghz is more than 4 times 0.7ghz and it does heck a difference.. And if you look at operation per dollar, there is not even a comparison... And I wont tell you how much their OS sucks.. the latest Irix versions feels like linux for 8 years ago (I mean the userspace stuff, I dont know much about their kernel...)..
Nice troll, and if it isn't, wow. Everything in your post is false. the 700Mhz MIPS certainly does stand a chance against other processors and I would love to have one, but as for price/performance, it is probably a very poor option.
The N64 did well as a system, and had far more power than the playstation. The playstation just did incredibly well.
Hollywood is a city, not a company. I am assuming you are talking about 3D and compositing visual effects studios, of which a few are near Hollywood, California. They aren't going to BSD, they are going to Linux, not just for rendering, but for workstations. Irix is unix and it makes it a very flexible choice for an OS. Because Linux is so similiar, it is also a flexible and powerful.
This Wiki Feeds You TV and Anime - vidwiki.org
Before evryone assumes that this thing is fast here some numbers to keep in mind:
OK there are no numbers for 16K but here the numbers for 600Mhz 14K
SPECint2000 500
SPECfp2000 529
For comparison
UltraSPARC III Cu 1.015GHz
SPECint2000 576
SPECfp2000 775
AMD XP 2800
SPECint2000 913
SPECfp2000 843
INTEL P4 2.8
SPECint2000 1040
SPECfp2000 1048
Looks like China has some serious competition on the chip front. SGI is already reaching a whole 700MHz! The Dragon better catch up soon is China wants to stay in business.
Don't trust a bull's horn, a doberman's tooth, a runaway horse or me.
On another note, I'm not even going to begin to comment on your thoughts on clock speed etc. I'm sure everyone else will flame you over the whole Megahertz Myth®.
I wish there was some there was some way that I could be outside playing basketball, in the rain, and not get wet.
Sorry 'bout that. I must've eaten something bad on Christmas because I fart constantly. The smell of the trolls gets lost in the noise so to speak.
Oh? Quick, everyone with Radeon 9700 PRO graphics boards in your PCs, make sure you have them in tower cases, or something!
For reference, the ATI specs page states:
I guess SGI might refer to actual output precision, i.e. the RAMDAC D/A-converters... In that case, it seems they still have the edge, since the ATI boards only have 10 bits per component. Still, I think that's of lesser value than the actual precision image operations are performed at.
main(O){10<putchar(4^--O?77-(15&5128 >>4*O):10)&&main(2+O);}
i dont need a MIPS history lesson. I didn't "forget" any of those CPUs. The R8000 was almost non-existant across SGI's product line.
While it was the first implementation of MIPS4, and it was an FP monster, and had a huge TLB for the time, it really wasn't so hot as a general purpose CPU.
A far as "true 64 bit" in the R4000, which version of IRIX ran on R4k with 64 bit pointers ? 6.2 and 6.5 certainly don't on my IP22.
When the R3k came out it was the first real example of commercially FAST and successful RISC design. It was used in multiple machines from multiple companies. SGI didn't "really" up the ante again until R10k, which was their first offering that was superpipelined and superscalar.
Finally, regarding SGI and clustering:
SGI is not price-competitive with shared-nothing clusters of PCs or Alphas. Nor is it trying to be. You probably know what the O2k/O3k systems are good at and how they differ from any other system being sold today, othewise you wouldn't have responded to me. I think my statement is valid --- the SGI big iron solves problems that shared nothing clusters CANT. Furthermore, they're so much more expensive than shared nothings that if you need shared nothing and buy origin, you're silly.
My opinions are my own, and do not necessarily represent those of my employer.
"Also, out-of-order instruction execution, x86 chips can't do this."
Bull.
x86 has done this since the introduction of the Pentium Pro.
You mean Apple ads?
Seriously, what 'material' are you talking about? I know about SPEC, according to which the currently fastest CPU is the Itanium 2 1000 MHz, followed closely by the PIV 3.06 GHz. From that I would deduce that even if you've got a relatively slow CPU (in terms of computations per clock cycle), if you manage to run it at very high frequencies, you'll still have one of the fastest CPUs out there.
Did you know you can fertilize your lawn with used motor oil?
Hum... exactly what "flow dynamics" package are you using? I am asking that because this is what I do for a living, and well... the AMD clusters have a loooong way to go. 1st, you have shitty internode BW, thus you have to have the code segmented just right. Then you are doing multiprecission correlations, which the AMD blows (and intel) at due to the smallish caches. Then you have the fact that the AMD doesn't run at 3GHz like you claim... and then well most userland apps can and are ported to Irix....
In other words, shut up troll. Geez I guess now and then people think that by throwing big words like "computation flow dynamics" we may think you are not BSing! Nice try though... BTW, CFD stands for computationAL FLUID dynamics.
And then you bring in the NEC (SX5 I guess) and well jeez I DARE you to post numbers from any nice fortran fd package and compare both the AMD and SX5 results. Jackass....
I've heard this before, and the SpecCPU2000 benchmarks just don't support what you're saying. According to http://www.specbench.org/cpu2000/results/res2002q4 /
For Integer performance, top dog goes to the 3.0 ghz P4 with a score of 1085 base. The Sun Blade 2000 at 1.015 ghz gave a score of 516 base.
In single task FP performance the 3.0 ghz P4 with gave a score of 1092 base performance. The Sun blade 2000 at 1.015 ghz gave a performace of 682 base.
It's mostly the same story for the multiple tasks test (rates). An AMD MP 2400+ gives a score of 15 integer base with 2 CPUs, the Sun gives an 11.4 Integer base with 2 CPUs.
For FP, Sun wins its first benchmark. Scores are 10.5 for 2 AMD MP 2400s, and 14.3 for the 2 Sun Blades. The PC arena gets pretty close to this performance with a 13.5 on 2 2.4 ghz Xeon processors.
So, with the exception of running multiple FP tasks, PCs smoke Suns pretty badly. If you'll take a look at the Itanium marks, they utterly destroy the Sun multiple FP tasks. Itanium maybe isn't really fair, since it's not really exactly at a PC price point.
AccountKiller
The R16000 has Out-of-order instruction execution? Sweet! So what was SGI's plan when they made this?
1.???
2.Profit!
3.Build new processor.
I wish there was some there was some way that I could be outside playing basketball, in the rain, and not get wet.
I believe you are missing the point.
The P4 only allows 4-way SMP. Most often you only see 2-way. Big iron scales much much further. Well over 20-way, sometimes 50 or more.
Let's see.
P4 3.0Ghz = 4way x 1085 = 4340
UltraSparc III 1.05Ghz = 20way x 516 = 10320
Of course, the Sparc could scale much further than that even. Plus SpecCPU2000 doesn't account for caches. They only give raw number crunching results. Real world data widens the gap even further.
erm. Slight update on the Cray thing.
They also sell the MTA, a hardware threaded architecture - from the Tera days - and the SV2 is now called the X1. They are also doing an AMD Opteron derived one-off system for Sandia National Laboratories. Though, from what I am hearing, it might not be a one-off system - they're considering productizing it.
Do you know why the road less traveled by is littered with the bones of the unwary?
Two way systems are not data center solutions that IBM, Sun, and SGI are competing for with this kind of hardware.
Even if they were, you're ignoring the fact that you cannot physically pack as many CPUs with Intel or AMD as with MIPS, Power4, or Sparc into a chassis. Part of the reason they are clocked slower is because you need to balance heat management with performance density when you're dealing with the big servers.
These boxes are about aggregate compute and storage power per dollar, not about whether the individual CPU cores smoke. The only place you see these cores as singletons is workstations (Single-cored "servers" are usually just the same or similar motherboards as a workstation, but in a case that has a beefier power supply and room for a useful number of hot-swap cages.)
You try and pack 32 Intel cores at 3GHz into a chassis that will handle the same number of MIPS cores, and the only thing you're going to get is voltage underflow from an overloaded power supply. Beef up the power supply, and within minutes you're going to be getting that wonderful whiff of frying, overheated electronics.
Raw performance of a core is only one factor in engineering a complete server. Anyone who claims otherwise has clearly not been involved with the hardware end of this industry.
I do not fail; I succeed at finding out what does not work.
Actually, one "trick" to getting high clock rates is to INCREASE the pipeline. The huge pipeline is what's been used to jack-up the P3/P4 clock speeds. It's actually harder to have a high clock speed and a shorter pipeline (you seem to indicate that a shorter pipeline should make it easier to have a higher clock speed).
-psy
That's 16-bits per colour in the RGB plane, for a total of 48-bit colour. Most PC cards do 24bit (or 32bit with alpha/Z plane).
Not the width of the GPU!!!!
You're confusing apples and oranges.
-psy
"...output precision... of lesser value than the actual precision image operations are performed at."
Not true if you're doing real imaging work. How about that fancy LCD monitor you've been eyeballing (or just picked up)? Noticed any of the color problems, especially with dark shades? No?
Then you aren't doing graphics work that needs the display accuracy of an SGI or equivalent.
I do not fail; I succeed at finding out what does not work.
Bottom line, if you need high precision integer colours, you still need an SGI. Of course, there's not many people who do, and someone will probably be doing it on the PC in a couple of years, so it's looking pretty grim for SGI as that's one of their few remaining technical advantages in the graphics workstation market.
You know what I hate? Wait, what do you like? I hate that!
OK, what is the deal with "In Soviet Russia XXX launch YYY".
I've seen two jokes with Soviet Russia now. And I'm not laughing. Can someone let me in on the inside joke here?
The Internet is full. Go Away!!!
How about we bring in the {mips|flops|whatever) per watt, then we start to get even better. The USIII runs off what, around 25-30W, the p4 draws around 70W. Using this, we have the Sun box beating the P4 quite easily, as we can scale up the number of processors to a higher number before we have problems due to dissipating heat and drawing soo much power that we blow the breakers in the data center.
As you can guess from the above post, I don't like the x86 architecture, ugly_hack(){ ugly_hack(); }. There's something to be said about elegance in the design of a processor.
How about with memory throughput, I believe the cray C90 is still up around 4-5, and it's around 10 years old. What's the point of massive compute power if it has to sit idle most of the time waiting for memory access.
In fact, work done per cost is what counts.
Cost includes:
- Price of the computer
- Price for energy supply
- Price for rental of the room your computer fits in
- Salary for the people operating the system
Clusters are expensive, because they fill large rooms and consume much energy.
Clusters are popular because usually you have to jusitfy the cost for purchasing the computers, but not the energy it uses, and you put it in a room that's there anyway.
Stefan
was kept in the 4mb of cache. That's large enough to keep a small Linux distro btw.
You can't judge a book by the way it wears its hair.
Why make a beowulf out of something that already scales above the size of any beowulf out there.
moron. my 500mhz alpha (yes, i really do have one) will beat the shit out of your 2ghz intel chip. why?
more registers.
rotating registers to make function calls way faster.
shorter pipeline.
no need for cisc-to-risc translation layer.
64-bits (want to see who encrypts a block of data faster?).
8mb on die cache.
ass-kicking memory bandwidth.
and it's safer, too, with things like X-bit per-page memory protection, to stop many common exploit types.
most of these hold true for the mips chips as well. the R16000 has 128 general purpose registers (?). x86 has 4.
The R16000 has many significant architectural and memory-related improvements over the R14000A. However, you are correct in that it's not the speed demon that some folks are making it out to be.
But... keep in mind that it consumes far less than 20 watts of energy (and thus gives off little heat) and will eventually find itself packed in with other CPUs into Origin servers/supercomputers. The CPU bricks for the O3900, for example, have 16 CPUs in just 4U of rack space.
SGI's ccNUMA MIPS/IRIX machines are typically used for tasks that are severely I/O bound, that is, their strong point is chugging thru massive amounts of data where raw per-node CPU power is important, but not the largest factor. Somewhat like a mainframe, but with less redundancy and more CPU power.
The Itanic draws somewhere around 130W, the P4 ~70W and (for the r14k I believe) mips processors draw ~20W. What's the point of some high speed if you draw soo much power that the lights dim with an add operation and the street goes black during a sub operation.
and iirc you windows can't officially have higher than 32bit(24bit+8bit alpha/wasted/whatever) currently.
the matrox card that does 48(? or was it 42), is actually using a dirty hack of some sort to get that depth on windows..
world was created 5 seconds before this post as it is.
pentium 1, yes... pentium4's mov takes 1 cycle too... if data is in cache of course and if there is no page miss.. otherwise it takes even 20 cycles. (but that's same on cisc processors, except there is 4MB cache, where P4 has 512k)
.. and so on.. in modern x86 1 cycle everything... itanium will be able to do 3 such instructions per cycle with ILP... differences between risc & cisc are fading...
Take other instructions for example... add, sub, shr, shl,
A coworker of mine has a pretty wild SGI box... but keep in mind, he hasn't even modded it yet!
Octane workstation
24" HD monitor, 21" monitor
dual R12000 @ 400 MHz
two internal scsi drives
internal DDS4 tape drive
two XIO gfx cards
fibrechannel XIO gfx card w/ external ciprico fibre raid
video capture XIO card
scsi pci card w/ assorted external drives
two weirdo data capture pci cards
Oops, now that I think of it, he does have sort of a mod... he bought an LED lightbar from reputable.com to replace the incandescent bar after it burnt out.
The machine is used pretty heavily to analyze video signals from various bits of broadcast and closed-circuit sources.
Another odd tidbit... he runs a much older version of IRIX 6.5.x, not the more recent 6.5.17 or 6.5.18. (IRIX and its applications and freeware CD sets are updated quarterly). Does the job, I guess, so no major reason to upgrade.
SGI is still in trouble. I love the company, their concepts, their hot rod machines and the supercool names they come out with..... BUT they are in trouble. And Linux is one of the primary reasons for SGI getting into trouble. A large number of design studios seem to have succumbed to the temptation of a cheaper yet stable machine (i.e Linux Boxes). As some other slashdotters pointed out, these guys are using Macs for the artwork and Linux boxes for the actual bull work. I wonder if SGI can reconquer their old customers and charm even more people.
|/________
|\A|ALYS|
The 128 bit precision you mentions is really a marketing trick.
If you look at the pixel shader architecture, you'll find that each instruction operates on a vector of 4 values (corresponding to RGBA). Four 32 bit floats don't really equate to 128 bit precision (more like 24*4, or 96).
In reality, of course, the output resolution (or frame buffer format) is what counts.
As of now, DirectX (including 9.0) supports a maximum total of 32 bits per pixel in the output buffer. Although that will certainly change in the future.
I'm not as annoyed at your ill-informed comments as I am with the person who tagged them as insightful. When exactly did insightful become a synonym for retarded?
As for the content of your post, none of the chips you mention as replacements for SGI's cpu allow for any meaningful single image multi-cpu scaling. It wouldn't matter if a P4 was indeed twice or thrice as fast as an R16000 if someone's application requires 128 cpus to work on the same data in a single computer (not a cluster); since P4 based (or your other examples) architectures do not scale beyond two or four cpus. You don't have to be mesmerized to realize that, it's just a fact.
Somehow I don't think you actually care about the facts and you were trolling. If you were, congratulations and consider this a late xmas gift.
Isnt MIPS the company of the RXXXX chips? AFAIK SGI only uses them for their Machines (and Cobalt used to for their Qube/Raqs)
------- The last Sig. got fired.
OMG It happened!
/me dies
Finally someone compared a desktop (not pro, I know some ati pro level cards as matrox ones) Gfx card, optimized for gaming... With... SGI...
Yes, folks this is history.
I think you're underestimating ATI's and nVidia's latest generations. If you go and read this technical document (PDF, ~4 MB) on the nVidia GeForce FX, you'll see that it does, in fact, support full 128-bit floating point all the way up to, and including, the frame buffer. I'm not certain why 12 bits of integer should be better than 32 bits of floating point, per component, maybe you could educate me a bit there? Also, I don't know what the actual analog output quality of the FX chipset is, maybe SGI has a (small) lead there, still.
main(O){10<putchar(4^--O?77-(15&5128 >>4*O):10)&&main(2+O);}
Well a server Blade does exactly this. Check out IBM site or Dell. I think dell calls it a PE1655MC
just two??? where the hell have you been?
typical headline reads:
Noun Verb Noun
aka subject and predicate
i.e.:
IBM announces some spiffy thing
Alan Cox washes his nuts
SO, the whole soviet russia thing, wherever the hell it came from, goes like this. they exchange the first noun with "you", flop the nouns around, and prepend the headline with "in soviet russia".
so, here we go:
SGI launches R16000
- In Soviet Russia, R16000 launches you!!!!
cant believe i wrote this post.
Yes those facts are all correct. And they would have been impressive in the high end workstation world about , oh , 5 years ago. Today they're pretty lame. I suggest you take a bus ride out of your toy computer ghetto and find out some facts about high end systems.
Have you read the SPECint and SPECfp results posted above? The Pentium4 runs at 6 (six) times this cpu's speed, yet only scores twice. Talk about good cpu design.
You should also keep in mind that SGI has some ass kicking technology when it comes to cpu and memory interconnect. NUMAFlex makes it possible to have a penalty as little as 1.5 vs 1 for memory accesses outside the local ram banks. Now try doing that with commodity x86 hardware. For problems that aren't easily broken down in small parts and, that have huge datasets, nothing touches SGI.
Kudos to the SGI engineers for their great job.
A long time SGI fan :)
Yes, does'nt it feel good to know your tax money is poured into SGI?
A linuxfarm may be much cheaper performance/dollar but it will suck bigtime on tasks that require closely-coupled systems like fluid-dynamics and material stress calculations, etc.
How about the slashdot reader who IS employed, and is reading and moderating on a day off from work?
Way to cover all the bases.
"Hey! Unless this is a nude love-in, get the hell off my property!!"
It's not the run-single-task-as-fast-as-possible. Intel/AMD rule that part, with Intel up a slight lead.
It's not the very parallell computing, like movie rendering. Clusters, usually Linux clusters do much better compared to cost.
It's not most kinds of servers, that are usually IO bound and it's the disks, controller, NIC and mobo (backplane) that make the server. Few of those need more than dual MP cpus to do well.
I know roughly where the SGIs still shine. But how many really have those specific needs? Not many that I can think of.
Kjella
Live today, because you never know what tomorrow brings
Honest Answer: not if they share the same heatsink a la Powermac G4 MDD! That's a big fucking fan, man!
That aside, why doesn't everyone here understand that the more competing architectures we have the BETTER - and why compare this R16000 to a P4 when it's obviously a competitior for the (800Mhz) Itanium?
That was classic intercourse!
Once upon a time SGI realized it bought lots and lots of MIPS cpus... so SGI bought MIPS.
As time went on, SGI noticed that the MIPS market was fragmenting... high end R1x000 series CPUs for workstations and supercomputers and low end embedded cpus for the consumer market. So SGI spun off MIPS, Inc but kept the R1x000 for itself.
These days MIPS Inc has nothing to do with SGI. And SGI's R16000 etc have nothing to do with MIPS Inc. I believe NEC fabs the R1x000 series for SGI.
and it was Intergraph that killed SGI (and Quantel) in post production - unfortunately they killed THEMSELVES in the process, so there are stil plenty of SGI and Quantel boxes out there getting the deadlines met...
That was classic intercourse!
A far as "true 64 bit" in the R4000, which version of IRIX ran on R4k with 64 bit pointers ? 6.2 and 6.5 certainly don't on my IP22.
64-bit support was first supported with IRIX 6.0 running atop R4000 and up.
However, certain platforms do not support 64-bit pointers. IP12/IP20 (Indigo) IP22 (Indy/Indigo2), and IP32 (O2) are among those that don't. This is due to memory contraints and other assorted issues.
Most, if not all, Onyx and Challenge (L and XL only) machines support 64-bit pointers with IRIX 6.0 and up.
Onyx2, Origin, Octane, and Fuel certainly do.
Getting back to the article, I'm glad to see SGI coming out with a new CPU. I still see a few SGIs in the wild now and again. If they lock down Irix a bit more security wise and expand their target market, they might be a decent competitor for Sun within the next 10 years.
Hear Hear. I'm also happy to see SGI pushing some new kit. It sounds like they've been quite busy lately. Rumor has it there are even some revolutionary (not simply evolutionary) MIPS cpu changes due soon.
IRIX security isn't too bad, it's certainly way better than it was just a couple years ago. If you dig around the software section of their website you'll see that they've even been working with the IPFilter author on some pretty serious IRIX packet filtering.
A lot of us out in academia/research hope SGI decides to drop their per-cpu price soon. Their individual CPU performance is still pretty decent but certainly not cutting edge. It's their I/O and thruput that's amazing... and we'd like to make better use of that. Shucks, the IRIX kernel can easily support 512 CPUs in a single machine (1024 if you use the IRIX XXL kernel). It's been tweaked every which way. But as it stands, we can't afford more than a 64 CPU machine. Still pretty nice, though. Even when working on a 6-CPU job, our (already somewhat old) Origin 3000 stomps all over our Myrinet-based cluster for anything that uses a significant amount of I/O. When shared memory is involved, the differences are even greater! (To compare, the newest Myrinet interconnect is 4 gbit/sec full duplex... SGI's NUMAlink3 is 25.6 gbit/sec [3.2 gbyte/sec]).
I'm looking forward to working with the new MIPSpro compilers too. Our SGI sales rep is supposedly going to bring the newest version and some demo licenses soon.
I see you failed to understand my ironic comment, let me elaborate the issue.
American tax payers is the ones that have keept SGI alive thru the companys resent crisis and struggle, goverment investing in bigiron SGI machines at the major national labratories and military facilities. Which is in fact your hardearned money that have paid for these systems. One part of the reson is that it is the US goverment interest to have a strong national supercomputer industry, and not being dependent on foreign countries for key technologies which supercomputers are (you dont see many Japanese systems at the national labs, do you?). The US goverment has managed serval govermental tax funded joint projects of commercial nature with the supercomputer maker Cray for example. The Cray SV1 was the result of one.
Check out:
8 87 .html
http://www.brainyquote.com/quotes/quotes/y/q128
"Sun's UltraSPARC III Cu tops out at 1.05Ghz last I checked. Does that mean that the P4 at 3Ghz stomps the hell out of it? If you said yes, you are a fucking idiot."
UltraSPARC III Cu 1.015GHz
SPECint2000 576
SPECfp2000 775
INTEL P4 2.8
SPECint2000 1040
SPECfp2000 1048
Doh!
Apparently, you're convinced that a P4 3GHz would do worse than the P4 2.8GHz quoted here. I hope not, because that would make you a genuine blithering fucking idiot, as opposed to a desperate anti-x86 zealot.
Maw! Fire up the karma burner!
Out of sheer boredom, I blew about $80 on an SGI from eBay.
Pretty old machine. 133MHz, 64MB, dual 2GB SCSI drives.
IRIX is pretty neat. It had some pretty decent 3D hardware. The installed Demos were equal to about a 400MHz Pentium II with a graphics card that didn't come out until 1998. No idea what year this SGI came out. early 90's?
I've also used O2's at some customers sites, and they're about as fast as the fastest wintel desktop you can get right now, at least in the CAD programs and such that they use. They tell us the Windows boxes can't handle the file sizes and datasets they use, and a lot of the software isn't even availble under 2K/NT.
believe it or not people will not buy a console until they can copy games [or have someone do it for them].
This is not because people are inherently criminal [although the something for nothing element can't be denied] but because for most casual gamers £40 for a game they may only play once is just too much.
Here we have a folder with literally hundreds of copied ps2/xbox titles. 99% of them don't get played for more than a couple of hours on the day they got downloaded. [the pile of non-pirate games is larger than most people's collection too]
There are places where the networks are not touching,and there are places where they are-Boeing's Lori Gunter
Sun's UltraSPARC III Cu tops out at 1.05Ghz last I checked. Does that mean that the P4 at 3Ghz stomps the hell out of it? If you said yes, you are a fucking idiot.
:-{
Then you have never experienced a UltraSPARC III Cu side by side with a decent x86 system. While it has a huge amount of cache, it is SLOW SLOW SLOW. Matlab (with float heavy code) is at worst 50% faster on a lowly 1.67GHz (MP2000) Athlon compared to the 900MHz USparc III. It gets smoked in integer performance, but does better clock for clock in FPU usage. Alas, overall, it is about 40-50% "better" on a MHz basis. In my experience, the 3GHz PIV (Xeon, mind you) stomps HARD on the USparc. It's embarrasing.
The 1.0GHz is about on par with a 2GHz PIV. And that's the uber-blest 8MB of L2 cache version. Cheaper ones don't fair so well. High memory bandwidth on Athlon MP and PIV Xeon systems help make up for cache shortcomings.
I like the SPARC instruction set and the impressive FPU speed attainable on low GHz cores, but it is entirely too expensive and it doesn't pack enough oomph for smaller (4-way) servers and workstations. The same can be said for the various iterations of the R10000 series in workstation environments (too keep from going entirely offtopic here).
And in bigger configurations, IBM's Power4 architecture makes it look like a toy too.
Fuck Beta. Fuck Dice
Isn't it MIPS that make the CPUs?
(This is not sarcasm, I really wanna know.)
4MB L2 cache => *huge* die => low yield => huge cost. Yeap, it's that simple.
The Raven
Looks like SGI is a Poindexter and TIA/Bush whore.
This is from their website concerning their "Information Dominance Solution" (Jesus that sounds fucking creepy).
"Today, governments worldwide are inundated with an abundance of data but a shortage of information that supports the decision-making process. In the future, a variety of sources and sensors will continue to generate ever-increasing amounts of such data. That data includes text-based information and records along with more complex media types such as video, audio, imagery, scans, electronic emissions, and other geo-referenced data."
Translation: Governments are finding it harder and harder to spy on their citizens.
"SGI has developed a variety of visualization and computing technologies ideally suited to allow government decision makers to rapidly assimilate the growing amounts and diverse types of data being collected. Finding the few important bits of information out of mountains of data is the job of the SGI Decision Support Center (DSC) solution, which uses large-scale visualization, high-performance computing, and the management of complex data to provide mission-critical information to support rapid and confident decision-making cycles. "
Translation: SGI wants to help.
"Decent gcc-3.x support"
Uhm, it's the compiler developers task to make sure that their compiler supports the architecture, not the other way around. And, GCC is absolute crap on MIPS, even the latest version.
SGI: 48 bits
Or to put it this way, if we use the alpha channel as well, the PC has 4294967296 possible colours, the SGI has 281474976710656 possible colours.. (since WinDOS (and AFAIK Loonix/XF86) won't support more than 32 bits for colour).
Clear enough yet?
/Mikael
Greylisting is to SMTP as NAT is to IPv4
"700MHz, has 4MB cache"
:-P
Hey, according to intel, this processor, at 700MHz, is about 4 years old, and has no hope of competing with intel's True MHz processors!
Many factors limit the performance of systems. One thing if I remember correctly that makes SGi machines so fast is that components talk to each other directly and at full speed. They have a fancy name for it, like interweave connect or something like that. So who cares how fast your processor is when you are using legacy designs for components that can't keep up with your processors cycles. Also the IA64 is solid, but there are a lot of better choices out there. Mainly becausse MS and Intel are pushing hard to move into that high tech market. I personally feel that AMD is going to corner the market early with their x86-64 processors and linux as their prefered OS. Plus SGi still alive because of the high tech industry, which is a small portion of the market, but a great deal of the money, ie people paying millions for a 500+ processor super computer, or $60,000+ high end servers. They are dead in the graphics industry, 1st because of no Photoshop 4 for IRIX (due to adobes poor coding and MS) and finally linux taking away the 3D market. They are probably going to get bought out by one of the larger High Tech CORPs like IBM, NEC or Terracomputing...
But who can resist a curvy indigo case? These are every /.ers dream date! A supermodel and an SGI? You know which one youd pick..
It's 11 bits for blue, 11 bits for green, and 10 bits for red. People don't like red as much as the other colors.
*sigh*
I thought enough material had finally invaded the net to realize sarcasm.
Note the original posters helpful use of an emoticon at the end of his post. Emoticons can symbolize many tones in situations where the text would otherwise be unclear.
Where does that put you?
Your Comment is wrong. There is a RISC - CISC conversion, since the core of the chip is RISC, but the instructions comming in is CISC. Also, 64 bits are extremely important. Even Intel and AMD think so. That is why they are spending TRILLIONS to develop the 64 bit architecture, which has been present in my R10K Indigo2 since 1994. The biggest problem with X86 (IMHO) is the fact that it only has 4 general purpose registers for programmers to play with. This MIPS has 128. Also, even if most instructions do take exacly one clock cycle to execute (which is doubtful) until recently, only ONE instuction did get executed at a time. Intel's SuperScaler Instructions fixed this, but once again my 1994 Indigo2 does this already. Hell, even ones made prior to my machine does this. Another big advantage for the MIPS is the fact that it has an 8 Megabyte _ON-DIE_ cache. This is an absolutely insane amount, and something like 2-3 times what a Xeon has. From my point of view, the biggest advantage that X86 hardware has is cost.
P4 != IA64
Bother to read a bit will you. I bet you're thinking of the SGI Itanium2 NUMAflex system(s) that have made the news recently.
on 64bit archs, ints are STILL 32bits. only longs and pointers become 64bits.
> Translation: The government is finding it harder and harder to spy on terrorists
He had it right first time. How do you know who the terrorists are without spying on them.
Personally I think I have much more to fear from the US government than from a bunch of beardy islamic nutters. Not that the average federal employee is more evil than the average fundamentalist muslim, but they have their fair share of fucktards, religious loons included, and they have more and larger weapons.
http://rareformnewmedia.com/
* 700 MHz
* Four-way superscalar, 64-bit architecture
* Out-of-order instruction execution
* Five separate execution units
* MIPS 4 instruction set
* 32KB two-way set-associative on-chip instruction cache
* 4MB fast secondary cache
Key Architecture Features
* 3.2GB/sec main memory peak bandwidth
* 1.6GB/sec system-to-graphics interconnect
* VPro graphics
* MIPS RISC processing, 64K primary cache, 4MB secondary cache
* Optimized 200 MHz front-side bus
* 32- or 64-bit binaries
* Priority I/O
* Integrated PCI
Blades are clusters-in-a-box, not integrated SMP systems. Like clusters, they hide the fact that you're dealing with multiple boxen, and don't have the shared system image and devices of a large system chassis.
Assuming those are 2-CPU blades, you'd need 16 of them to equate a 32-core system chassis, and would still need to add RAID arrays to the rack (unless a virtual SAN will do for your application -- it won't for large database servers.)
Bottom line is you need to know what the system is going to be used for, and compare the features that support those needs. Even identically-cored systems based on SMP vs. cluster vs. blade are going to have radically different performance characteristics and benefits for different uses.
Blades are great for things like web server hosting, where you want a lot of isolated processes. Clusters are good when you need shared storage, but don't need shared memory. SMP chassis can handle all of the above, plus deal with the large IO caches and shared memory that database and application services require, but at a higher dollar/benchmark cost than the first two. (Not surprising -- SMP backplanes require far more complex engineering than clusters or blades of off-the-shelf SMP systems relying on GNet or other non-backplane interconnects.)
I do not fail; I succeed at finding out what does not work.
on 64bit archs, ints are STILL 32bits. only longs and pointers become 64bits.
Of course, you're right. I don't know what I was thinking. I mean to say "pointer" but typed "int" anyway. Oops.
I write in my journal
The output framebuffer colorspace is _NOT_ what matters, 32-bits is plenty. The problem that you run into with 32-bit precision math is that you get rounding errors that affect the quality of the output when layered calculations are being done in the graphics pipeline. Every transparency, pixel shader, and lighting effect you apply to the same 32-bit number, the more lossy it becomes. The 128-bit colorspace alleviates this problem, because the rounding errors are made so much more minute.
<Amanda`> I just went out to the parking lot in my bathrobe to exchange warez CDs.
with calligraphic lights you hit the screen with an extra, independent electron beam. This produces intense light points. These light points are useful in flight simulations, and I believe, required for some levels of certification of a sim.
"A good example of the failure of mips is to look at the nintendo64, a video game system developed which uses a mips processor. It was so slow and underpowered that the playstation, which has been out for 2 years before the n64, still had better and smoother graphics. The n64 should have been the final nail in sgi's coffin."
Are you kidding? The reason that the PS seemed 'smoother' than the N64 was that it had almost no graphic features turned on. Don't believe me? Fire up Ridge Racer and watch the road beneath you. Notice that it turns all zig-zaggey when it gets close enough to you? That was one of the limitations of the PS hardware. It wasn't doing anywhere close to the number of calculations per pixel that the N64 was doing.
The real reason that the PS appeared 'smoother' was that it used minimal graphics tricks and pumped around 300k triangles on the screen. The N64 had all the features turned on and was getting around 100k triangles. So the result was that the N64 had fewer triangles to work with, but much MUCH better texture quality.
As for being a nail in SGI's coffin, I agree with you, but not for the reason you suggest. The N64 was both quite powerful and quite popular. The PS may have done better, that doesn't mean that the N64 didn't do well. The SGI processor did just fine, but they pretty much designed themselves out of business. Why are they charging a premium for their hardware when they can get a slimmed down system crammed into $150 box? In reality, that may or may not have directly affected their credibility. But it did significantly lower the value of the effects they were able to accomplish. Suddenly, consumer hardware can do what SGI does. Hrmm Why do I want this expensive box again?
"PC's running bsd are still a far greater value than expensive sgi hardware."
No argument here. Though I believe SGIs have their place, I think your comment's right on the ball. SGI didn't isn't doing enough to wow customers. Let me give you an example, when they launched the Intel based NT Line, there was only one real major difference between that machine and any other PC on the market was that it had a much faster bus between the RAM and the graphic chip. The problem is, what do you do with that when everything's designed around a 1x AGP bus? (this was 2-3 years ago..)
My company has a particular application today (but not back when we had the machines) to get ludicrious amounts of data to the graphics processor, but that's a very specific need. Not something you can build a whole company around.
They should have done more than just having the fast graphic bus if they were going to cater to the Wintel crowd.
So yeah, I basically agree with what you said, but your details about the N64 were significantly wrong.
Clusters are popular because usually you have to jusitfy the cost for purchasing the computers, but not the energy it uses, and you put it in a room that's there anyway.
A big adavantage of clusters is their inherent fault tolerance: a cluster can keep on working if one of the clustered machines spontaneously combusts. With a proprietary supercomputer, you usually have to wait for a service droid to arrive with the needed parts.
As for the cost of energy consumed, as long as you're talking about fully indistrialized countries that's going to be negligable compared to the costs of personnel and space.
Gentlemen! You can't fight in here, this is the War Room!
Actually, they have a few 1024 CPU's out there. I'd like to see such a linux cluster working as a single image.
:)
Don't get me wrong, I work with both technologies and some systems are better at some tasks while others are better at other tasks.
My philosophy is why use a F1 in downtown when you can use a scooter instead? (and you can probably carry more stuff on a scooter anyway).
Yeah, SGI is pricey, yes it doesn't apply to anything. But so what? I run applications that require some of the SGI "way of thinking" while other applications work fine on a cluster.
Let's not forget that some people don't mind about money and SGI will give them peak performance for their money
One thing most people don't seem to know is that you don't go out to sgi.com or hp.com and buy a $5M super computer. Your application is usually benchmarked and you buy from the vendor which gives you the best data. Then the contract certifies the benchmark. If you're not happy, they have to take the machine back...
That was my rant of the day
-- Leeeter than leet
Let's not forget that every machine has it's limitations. Otherwise, everybody would be using the same machine right?
You have to see the machine as a whole, not only the CPU.
Sure, a F1's engine is fast, but can you let it run 24/24 365/365? Probably not. How come we're not all driving the same car? Why do we have SUV's, car's, scooters, trucks? It's for the same reason. Each machine has an "edge" in a certain way...
-- Leeeter than leet
The only thing your pathetic 500MHz alpha will be a 2GHz P4 in is running hand-compiled Alpha code. Even then, I wouldn't be so sure.
You are very stupid or very naive. What you claim simply isn't true.
We have a cheerleader! Wonderful - I assume you work in the scientific visualization field?
You seem to be confusing mainstream P4/PPC/even _Alpha_/Power4 chips with high end systems based on them.
Do you really think SGI's competition is Dell or Gateway? IBM, HP, and many other companies sell very high end systems, and there are supercomputer class systems based on all of the above processors.
And how many single-quad processor systems does SGI sell to suckers out there for 5 times the price of a better performing system just because people are so enamored of the SGI name? Lots. We're not always talking about the heavy iron, here.
PC: 32 bits
:)
SGI: 48 bits
Not with the latest hardware. As the post you replied to mentioned, the Radeon 9700, as well as Nvidia's upcoming chip, support render targets and internal data paths with 32 bits of PER CHANNEL for a total of 128 bits of color data per pixel. SGI's hardware supports 12 bits of precision per channel for a total of 48 bits.
In fairness, those are 32-bit floats, not integers. The actual precision of the color data is "only" 24 bits per channel, and the remaining bits are taken up by the sign bit and the 8-bit exponent which allows a much greater range of magnitudes than if this was just a 32-bit integer.
Clear?
(And before anyone points out that 24 + 1 + 8 = 33, the top bit of the mantissa isn't stored because it's always implied to be a 1.)
Maybe so, but much of the graphics technology we take for granted was pioneered by SGI. Their engineers and technology built 3DFX and Nvidia and revolutionized the PC industry. Silicon Graphics can still easily build systems to run circles around PCs for real-time graphics. Their buses, memory and design are usually far more efficient. We'll forget they ever made the O2, for now. But they could never compete with the PC price/performance ratio. Cheap PCs have mass bandwidth, but are really inefficient, buggy and hard to scale (GB+ clusters and the right software might strongarm even this obstacle). Although I bet it is extremely difficult to outclass SGI's supercomputers. Perhaps IBM or NEC have a chance, but any cluster of PCs would be laughed at today.
The primary motivation for more color bits isn't the final output. Many monitors don't even have 8 bits per channel in their color resolution! But when you're performing many layers of compositing and texturing, if your intermediate results and your mathematical precision have an error of 1 part in 512 (the best you can hope for with perfect rounding and 8 bits per channel), the accumulated error in your final pixel can be very visible.
Furthermore, going to higher color precision makes it practical to have a linear color space. Your eyes are much more sensitive to subtle changes in dark colors than light ones, and the color distribution on most monitors is skewed accordingly (i.e. the difference in intensity between a green level of 13 and 14 is much smaller than between a green level of 240 and 241). That eliminates visible banding in dark areas, but it makes it a bitch to add things together and have it come out right (i.e. an color value of 40 isn't twice as bright on screen as a color value of 20). With insanely high precision, you can just use a linear color space and still have enough precision at the low end to avoid banding.
Well, last time I checked CFD (especially FLUENT) was one of those borderline embarrassingly parallelizable applications that are ideal for the shared-nothing architecture of a cluster, therefore I would be surprised to hear that your SGI/NEC machines get their butts kicked by a cluster of Athlons.
SGI, with it's ccNUMA architecture excels at handling problems that need the performance boost of multiple processors but that require a single-system image because they are inherently only poorly-parallelizable. These are usually problems like 3D seismic modelling, crash test simulations for vehicle manufacturers and CAT scan visualization in medical computing. This is the sort of computing that would slag a cluster of Athlons into a puddle of glowing silicon if the Myrinet cables didn't melt first. Origins handle them with ease.
utter rubbish
utter rubbish
> for 99.999% (that's 5 nines) PCs can do the job
> and even do it better and especially do it much
> cheaper.
I don't buy it. With ix86 PCs, it's not just the software that's crap compared to legitimate enterprise solutions, but the hardware too. Linux is nifty and all, but it only improves the software side. The hardware is still shit.
I've used ix86 boxes from most every builder... from solidly well-built IBM machines, to crap boxes built by dell from commodity parts. Not a one of them has achieved five nines. Remember, that's only five and a quarter minutes of downtime PER YEAR. With most OSs, if you reboot two or three times, that eats up all of your downtime right there, assuming NO other problems.
ix86 boxes just are NOT up to the "five nines" standard. OTOH, I've seen more than a few Sparc, SGI, and RS6000s that can do it.
Remember... just because you CAN do something on the cheap with crap hardware doesent mean that you should. And it doesn't mean that enterprise hardware doesn't have its place.
cya,
john
Imagine all the people...
Why the fuck am I trolled? This is an honest question! The alpha used to run at 333 Mhz when Pentiums were 100 Mhz, regardless of the actual performance advantages. I remember in my computer organization class (I'm a computer science major) said that RISC chips can run at a higher clock since they don't have a microcode decode step and there aren't so many ridiculous multi-cycle instructions (like PUSH/POP all registers, String processing intstructions, etc). Are all the moderators a bunch of Mac fanatics?
Hey, I'm just your average shit and piss factory.
CISC is even more of MHz increaser, because the decoding of the instruction gets chopped up into many parts too...
Thank you. I was not being clear enough, it seems.
main(O){10<putchar(4^--O?77-(15&5128 >>4*O):10)&&main(2+O);}
Actually, good guess, I have worked on scientific visualization, simulation, manufacturing and content creation on both SGI and other platforms. That doesn't mean I'm a cheerleader or that I'm on some kind of "SGIHAD".
As far as I'm concerned there have always been three distinct reasons to buy desktop systems from SGI:
1. You need to develop an app and you want access to specialized development tools. Let's not forget they invented OpenGL and a few other useful apis.
2. You need to run an app that only runs or runs better on SGI because of special hardware or because the people who wrote it came to discover reason #1.
3. You want to develop or run an app that you want to scale to the mega multi-cpu systems we were discussing. You do realize that many O2 and Fuel workstations are sold to support the development or use of Onyx apps.
Price comparisons are only useful when comparing systems with equivalent specs. You always pay a premium for unique technology differentiators, partly to support the costs of the R&D that led to them and partly because the manufacturer can get away with it due to lack of competition. I believe this addresses your comments on both big iron and desktop systems.
Now I will in turn make the assumption that you are not in a position to be making real purchase decisions about high-end systems. This does not undermine fact that you are right about one thing: that they are less situations nowadays where the decision to buy SGI is a nobrainer since others have now caught up with them on many fronts. However, please do not presume that those of us who make these purchase decisions, do so because the spinning SGI logo is cool.
I will admit that once in a great while I get nostalgic for the good old days when SGI was Silicon Graphics and groovy was groovy. But then I remember how much 256MB of RAM and a decent graphics card used to cost.
x86 has done this since the introduction of the Pentium Pro.
Even before that. The Cyrix 6x86 chips and AMD K5 and K6 chips had out-of-order execution too.
Here I'd think sane people would think such horrific annhiliation of beowulf mentioners is funny.
I guess that makes moderators insane...
(or too lazy to understand the thread)
Someone set us up the bomb, so shine we are!
Used to design SGI based systems in 1997 of so.
We used graphicless origin 2000s.
All our workstations were wintel using humingbird eXceed to run our graphics on the SGI.
And we loved them.
Why?
It's the I/O silly. Too date I don't see anything in the intel world that is keeping up with SGI's multi gigabyte packetswitched IO.
That's 1996/7 technology kicking intels 2002 but.
Why oh why hasn't SGI capitized on these strengths????
This post reminded me of a poem. Others, I Am Not the First [...] Agued once like me were they, But I like them shall win my way Lastly to the bed of mould Where there's neither hot nor cold. [...] Alfred Edward Housman
Only if you don't care about heat dissapation or power density. At some point watts per cubic nanosecond*c becomes the limiting factor. P4's are not optimal if you want to pack a bunch of them in a box small enough that the interconnect distances are short.
This is not a nintendo. Why waste cycles rendering nice boxes?
Besides, if you want to waste money on your investment, go ahead and download either KDE or Gnome on their freeware page.
-- Leeeter than leet
Actually, you could make one. And it would be nice to see a cluster of 1024CPU machines...
-- Leeeter than leet
AFAIK, MIPS processors run on pretty low wattage compared to Intels. For example, SGI has 128 MIPS CPUS in a single rack while on the Intel side, it can only offer 16 due to heat factor. Those Itaniums are just getting too hot..
Don't forget about your A/C ratio as well... on in the end, it cost more to run Intel stuff..
-- Leeeter than leet
well, during my time coding gfx effects, from my dos days with vesa modes(2.0 vbe was sweet)...
i have never enountered such mode in actual use.
the 8 bits got wasted in every mode available on any card(that had any sense to use)..
world was created 5 seconds before this post as it is.
The accuracy is neccessary for complex lighting, transparency effects and so on, 8-bit alpha sucks. The output is 10-bits per component.
s .h tml
http://www.sgi.com/workstations/octane2/graphic
Right, except that you're still oversimplifying.
First, five 9s does not make the mistakes that three 9s makes.
Second, five 9s recovers smoothly from the mistakes that three 9s makes.
Third, five 9s does not let errors go by unnoticed like three 9s does.
Fourth, five 9s has a much more critical sense of what constitutes an error.
Fifth, when something does break, it has to be fixed. That counts.
Misquote from Dijkstra: "A baby crawling and a jet plane from JFK to LAX are both means of transportation".
Probably stays up 100%. That's what? About -1 or -2 9s, methinks.
Your post here got me to thinking. I had wanted and Indy back in the early 90's when I was working with Indigo's at a NASA contractor in Hampton, VA. Did not have the $$ tho. Now seems a good time!
...for $400.
After doing some digging, several things became clear:
- For a little more $$ than you'd need to spend for a used Indy (maybe $150 more on eBay), you can get a used O2 or Octane which are both much more powerful and viable today.
- The Octane is notably more powerful than the O2, but the market is flooded with them so the Octane is oddly cheaper than the O2. It's also much louder and larger than the O2--less of a "personal workstation" (see O2 and Octane photo here -- O2 on left)
So I have just grabbed an O2 with:
- R10000 CPU @ 175MHz, 1MB L2 cache
- 256MB RAM (unified memory architecture)
- 4GB HD
- A/V module (audio & video in & out)
- O2 cam
- keyboard / mouse
So I shall add to the list my first IRIX machine. Hope my OS X box does not get jealous...
blakespot
-- Heisenberg may have slept here.
iPod Hacks.com
While most of the VFX industry has been trying to drop SGI like an unwanted stepchild, SGIs are still very common in the world of digital color grading for feature films.
This is because an Octane2 can play back a movie at half of film resolution (1024x778) at 24fps, UNCOMPRESSED.
Being able to play back a 24fps movie uncompressed at 1/2 of film resolution (as opposed to playing it back at something like 640x480 or 720x480) is necessary for digital color grading because you need to be able to see if the correction operations you are applying degrade image sharpness and how they affect the appearance of grain. The movie has to be uncompressed because maximum image quality is a must. Even if the final images are uncompressed, the proxy images (the source images and final images are at 2048x1556 uncompressed, the proxy images used (in place of the full resolution images while the artist does his/her work) to keep test rendering time down to an acceptable level are "only" 1024x788 and uncompressed) must also remain uncompressed, because you don't want compression artifacts affecting your judgement of the picture quality. Compression artifacts in the work/proxy image can affect your evaluation of the grain appearance in a shot, etc.
How are Octane2s able to do this? Simple: SGI systems are designed to have massive I/O bandwidth. SGI's Octane (and, thusly, the Octane2) uses a crossbar-switch to send data between the components (RAM, CPU, graphics, hard disks, etc.) instead of using a system bus. The crossbar-switch can give a component a dedicated data channel to another component. For example, the CPU can be writing stuff to the hard disk (let's assume that the CPU is generating it's own data and doesn't need to access RAM to get the data) while the graphics card fetches stuff from RAM, and they can all work at maximum speed. With a system bus, the whole computer has to share that bus, so the components can only operate at full speed some of the time.
THE LESSER-KNOWN PROGRAMMING LANGUAGES #8: LAIDBACK
This language was developed at the Marin County Center for T'ai Chi,
Mellowness and Computer Programming (now defunct), as an alternative to
the more intense atmosphere in nearby Silicon Valley.
The center was ideal for programmers who liked to soak in hot tubs while
they worked. Unfortunately few programmers could survive there because the
center outlawed Pizza and Coca-Cola in favor of Tofu and Perrier.
Many mourn the demise of LAIDBACK because of its reputation as a gentle and
non-threatening language since all error messages are in lower case. For
example, LAIDBACK responded to syntax errors with the message:
"i hate to bother you, but i just can't relate to that. can
you find the time to try it again?"
- this post brought to you by the Automated Last Post Generator...