Sun To Build Opteron Servers 287
geekee writes "According to an article at CNET, Sun is planning on creating Opteron-based servers. These are expected to include 2-processor and 4-processor models running either Solaris or Linux. This move isn't surprising, given the performance and cost gaps between the Opteron and UltraSPARC processors. A move to Opteron would allow them to be more competitve in cost and focus more on what they're good at, designing systems, not processors."
another dell/HP (Score:5, Interesting)
I dont know, help me out. (Score:2)
The only gap I see is Sun not being co-operative with free software writers. That's dumb, because they will be comming up with alternate uses of their hardware. I know someone who bought a surpluss Ultrasparc based system and and I'd be jealous if i
Re:I dont know, help me out. (Score:5, Insightful)
Look, Sun makes great hardware above the low end, but an old K6-2 beats a Blade 100 desktop in perceived performance and compile speeds. The IIe chip is low power -- in more ways than one. If you don't have a CPU-bound process, like say, a web server for mostly static pages, a Netra X1 or V100 works great, but it's not a fast CPU.
OK. Price/performance. Let's see. SPEC2000 results, Sun Blade 100 (650Mhz US IIe, fastest IIe available in a system) gets 246 integer, 276 floating point. An Opteron 146 (2.0Ghz), on an Asus SK8N board, gets 1262 integer, 1300 floating point.
Just in case you meant the US IIIi, as used in the new V210, V240, V250, and Blade 1500, the results on a V210 (server chassis, 1002 Mhz) are 555 integer, 841 floating point. If and when Sun can get the IIIi up to 2Ghz, that would not quite match the Opteron for integer ops, and just beat it for floating point. Of course, by that time, the Opteron will probably be up to 3Ghz and smoke any available IIIi.
Any more bullshit to sling about price/performance?
Benchmarks from www.spec.org [spec.org], as published by the vendors. Configurations of the boxes are detailed there.
Re:I dont know, help me out. (Score:2)
Re:I dont know, help me out. (Score:3)
Re:I dont know, help me out. (Score:2)
Re:another dell/HP (Score:2)
I don't think article said anything about Sun dumping sparc-based line on short term. So just like IBM builds all kinds of system (from Power - chip based servers to PCs), perhaps Sun is just expanding their product line, taking advantage of cheapness of commodity x86 processor lines. Dell does not sell non-x86 systems, and HP only sells those systems for legacy purposes (HP-PA, Alpha).
Re:another dell/HP (Score:2)
Re: (Score:3, Insightful)
Re:another dell/HP (Score:2)
Sun is introducing these Opteron servers to compete at the low end.
Sun also develops its own OS - Solaris - which it is able to offer cheaper than Linux on the same hardware.
Because of Dell and HP's close alliance with intel, do not expect to see them shipping any Opteron (or any other AMD for t
Re:Another non-Apple (Score:3, Funny)
proprietary hardware (Score:5, Insightful)
But if you then say ANYTHING about IA-64, I'm going to jump down your throat with lawn aerators on both feet.
Be cautious about what you call commodity and what you call proprietary.
Just because a lot of something is made doesn't mean it's not proprietary.
Just because it's low volume doesn't mean it is proprietary, or not a commodity.
IMHO, Intel is only kept in check pricewise, by the presence of AMD, to a lesser extent, Via and Transmeta, and to a still lesser extent by PPC and other 'non-commodity' processors.
IA-64 is simply THE MOST PROPRIETARY processer there is. It's IP is held by a separate company, licensed to Intel and HP, so that prior contracts those two have don't give anyone else IA-64 access. The PII bus was patented, the PIV bus is patented, SSE (and/or SSE-II_ is patented.
They're perfectly within their rights to do this. But then you have to watch what you call 'closed' and 'open'.
Re:proprietary hardware (Score:2)
IA-64 is patented by a seperate company.
PII, III, 4 busses are all patented by INTEL. SSE/2 is patented by intel. But oh, wait, doesnt the opteron have SSE/SSE2? Yea, coz intel and amd have a cross licensing agreement so they ca
Re:Another non-Apple (Score:2)
Re:another dell/HP (Score:3, Insightful)
Not so much in the 1 and 2 processor space. Sun's lower end has been your basic PCI Mobo grade stuff for many years now, save Sparc and Solaris. IDE drives, limited expansion, "the works" we've come to know and love from PC land.
Somewhat hard to tell, unless you look inside, because Sun controls Solaris and Solaris controls you.
As for the Sparc CPU, it is nothing particular to write home about, not bad, not great either.
Solaris has it's points,
Re:another dell/HP (Score:2)
System level design (Score:3, Interesting)
Individual components and pieces of performance (CPU clock and IPC, for instance) are only part of the issue. System balance is important, and only learned with
Re:another dell/HP (Score:5, Interesting)
I have a few Dell servers. All and all, I am pretty happy considering what I spent, but they still do not compare in quality and reliability to my IBM servers (pc325's from 1997, running 24/7, still running). Its the difference in 99% vs. 99.9%. It may only seem like
I will still use my Dells for non-critical web serving, routing, backup dns, etc. but the average Sun box would blow these lower/mid level IBMs away, and there really IS no comparison to Sun boxes. Apple is also much more reliable and robust than the average Dell. Same for HP, and others.
I like my Dells, but realistically, they are decent boxes thrown together from off the shelf desktop grade parts.
Re:another dell/HP (Score:2)
I've got ~20 or so Dell rackmount boxen running websites, and only one has ever shown problems, with the gbit nic card (supplied by dell) failing after
Re:another dell/HP (Score:3, Insightful)
Wasn't. Was comparing them to my IBM servers. As I said, they are fine for non-critical systems, and are good for the money, but my *experience* has shown that they are not as robust as IBM. This is based on years of using both brands, not only in the server room, but on the desktop as well. In the server room, they are adequate for many tasks, but there is NO comparison to IBM in quality. Its not bashing Dell, its just real world experience.
Ren
Re:another dell/HP (Score:2)
Actually, Sun takes a Dell-OEMed 1U rackmount system and sticks a purple faceplate on it.
Which makes the people saying Sun machines are so much better than Dell machines even funnier.
Sun is simply adapting to survive (Score:5, Insightful)
Re:Sun is simply adapting to survive (Score:4, Insightful)
hahaha. Sorry, I just had to laugh.
Isn't this the same ship that SGI is sailing out on?
And by out, I mean off the edge of the world, into the abyss.
Competition is great (Score:5, Interesting)
- G5 vs. Opteron
- OS X vs. Windows
- Linux vs. Windows
- Mozilla/Firebird/Thunderbird vs. IE/Outlook
It is a good time for computing. Although, with Longhorn so far out (and no further IE improvements until then) I think the competition is going to be a little bit one sided.
great when it works (Score:3, Insightful)
Re:great when it works (Score:2)
Re:great when it works (Score:2)
Re:great when it works (Score:2)
Re: (Score:2)
Re:Competition is great (Score:2)
Re:Competition is great (Score:2)
Re:Competition is great (Score:2)
The way I get them to switch without question is by saying "Not only does it block popups without modification, it will also prevent any browser-based exploits such as the 500 peices of spyware I just removed from your computer." I may also add "and don't install Kazaa, ever. With Mozilla and no Kazaa you'll never have a problem again."
Even if people don't blindly click YES on the b
Re:Competition is great (Score:2)
Re:WARNING: Karma whoring (Score:2)
And OS X is rapidly moving in to the low-mid server space, with Apple's excellent XServe 1U boxes.
You couldn't PAY me to use Sun! (Score:5, Funny)
Let's prorate this non-usage (Score:2)
check out this video: (Score:2)
OP is Flamebait (Score:3, Interesting)
So what does the 20+ years' lineage of the SPARC architecture represent, if not Sun's ability to successfully design, implement, market and deploy processors? Hello? McFly?
Re:OP is Flamebait (Score:4, Insightful)
If they're so successful, why does a $2000 Opteron system outperform a $10000 Sun system? SPARC has fallen behind on the performance curve, and yet they still charge a fortune for their machines. They are surviving only because people still need legacy apps, but as more stuff is ported to Linux, they're losing that market too. They have no choice but to compete in the x86 market since they have no better solution currently, either in performance or cost.
Re:OP is Flamebait (Score:5, Insightful)
That $10,000 will still be running, with full hardware support 5 years from now.
You'll be lucky to be able to buy ram with a warantee for that Opteron. Which will probably ahve bit the dust from component failures anyways.
Now a $5,000 Opteron or G5 on the other hand...
Just remember your cheap-ass desktop components may stand up to 3-4 years of mild desktop use, but the same components will die much quicker in a server.
Re:OP is Flamebait (Score:3)
The CPU is a minor part of the issue. As you say, desktop components will die quickly in a server. But put an Opteron in a server with server-quality fans, cooling design, power supply, and all the rest, and you've got a decent server.
Sparc would still likely be more reliable, because there are things you do can inside the CPU, but a well-designed box around an Opteron would still be very good.
Re:OP is Flamebait (Score:2)
The real question is how close to the standard x86 PC these things are going to be. The x86 platform isn't terribly well designed for Servers, although it's much better than it used to be. Especially the craptacular firmware restrictions that we're stuck with for compatibilities sake.
Image an Open-Firmware based Opteron. Oh, just imagine the possibilities.
Re:OP is Flamebait (Score:2)
Re:OP is Flamebait (Score:2)
Wouldn't bet on it.
The only box I own that I would bet on is my SGI Challenge S.
Of course, it's been up 24x7 since 1996, first for the original owner, and since June for me. It did lose a hard drive in the process, but th rest of it is original.
Re:OP is Flamebait (Score:2)
This has been 10 years now. it's still running and relaying the data back as well as acting as a ham radio Packet radio relay/bbs station. batteries get changed every 2 years by the local ham radio club.
Yeah
Re:OP is Flamebait (Score:2)
I'm not betting the business on luck.
And there's a huge difference in quality between pc104 boards and your generic clone (Or even your Asus clone). pc104 boards aren't consumer.
It all comes down to QC and engineering.
Remember, that critical apps are generally heavily engineered and tested, so that they can handle failure safely.
Re:OP is Flamebait (Score:2)
They're engineered nearly as well as the Sun boxes. And they work pretty well.
But their lifetime is half that of the Sun's. You don't see too many 6 year old compaq's in Datacentres. You do see plenty of high-end sparcs from that era.
Re:OP is Flamebait (Score:2)
Without fail, thein windows/x86 boxes are the worst, the Linux/x86 boxes are middling and the sparcs are tops fro reliability.
And some of those sparcs are twice as old as any of the x86 boxes.
Of course, support contract levels do matter. I work for MCI. wou could probably have atech sitting 24x7 in our office if we asked nicely.
Re:OP is Flamebait (Score:3, Informative)
Not best on performance but it is best on price/performance. According to that Sandia paper, an Alpha EV7 at 1.25GHz is about 50% faster than Opteron at 2.5GHz. Obviously, using Alpha likely costs more than 50% more than an equivalent Opteron implementation.
Not so in my case... (Score:2)
For what it's worth, I use both a Sun Blade 1000 (2x950MHz) and a dual processor (146) white box Opteron. Both of 'em have 8 GB of RAM and fast wide SCSI drives. The Blade 1000 churns through a SpecctraQuest transmission line simulation about 40% faster than the AMD system.
The Sun system cost a lot more than the AMD system, but the payoff is that the time that I have to spend waiting for a simulation to complete
Re:OP is Flamebait (Score:2)
In the long run, zip. Sun didn't start with SPARC, and it looks likely they won't end with them. Sun Microelectronics may be wedded to the Sparc, but Sun Microsystems appears ready to cast them off if they can sell storage -- their most profitable line. It's not an easy choice, but if the choice becomes "SPARC or Sun", there's no dou
64-bit question... (Score:5, Interesting)
Re:64-bit question... (Score:3, Informative)
Re:64-bit question... (Score:2)
I see no need, strategically, Opterons are only scaleable 8-way so far (the 800 series chips), and the article says Sun only plans on making 2 and 4-way boxes.
I think this allows them to "upsell" anyone who wants to keep their Solaris/Linux investment and move into the 16 or higher processor boxes that only the big iron platoforms (not x86) can p
Re:64-bit question... (Score:2)
Sun's missed move (Score:5, Interesting)
First mistake was in not encouraging 3rd party vendors to adopt the higher-end SPARC's, and ignoring the low-end SPARCs that used to dominate the embedded space. They had a strong position when they moved the SPARC architecture into the open, but lost it when they failed to support that initiative with bare-bones development machines.
Next mistake was creating Solaris for x86. Sun's logic was to hook folk on Solaris in order to get them to move over to their profit-making SPARC's. BIG MISTAKE. Instead, those SPARC vendors decide that they can instead move off of SPARC and keep using Solaris on the lower-cost x86 machines.
Final Mistake was Sun ignoring the low-to-mid range workstation market that they dominated during the 80's. Sun's focus on extreme-high-end servers cost them the middleware support that made Sun boxes worth purchasing in the first place.
This move to Operton might be the only step left for them if they are going to survive outside of a vertical market.
not entirely related (Score:5, Funny)
Re:not entirely related (Score:4, Funny)
Re:not entirely related (Score:2)
No, you see, the reason Opteron is so powerful is that he is protected by ITANIUM ARMOR!! If the forces of evil ever get their hands on the secret of Itanium, we're doomed!
*ahem* yes well, back to doing important things now...
NASA's right... (Score:2, Funny)
Will they be called... (Score:2, Funny)
Re:Will they be called... (Score:2)
No, we tried that here at Sun and the sun hired a lawyer and sent us a cease & desist order for using their Sun Flares(tm) trademark.
The sun also sent us a lengthy list of other trademarks we can't use. One trademark, Sun Go Nova(tm), makes us a little nervous.
Bravo, Sun. (Score:5, Insightful)
Opteron is a great choice. Not only is it technologically superior to Itanic, but it allows Sun and AMD to work together to keep Intel at bay. What's good for Intel usually ends up being good for Dell and Microsoft -- not Sun. Plus, Sun gets to save face by not having to turn around and say "uhhh... ok, maybe Intel isn't so bad after all."
All Sun has to do now is execute this properly, sell the products at a reasonable price, and stand behind a solid dual Linux/Unix strategy the way IBM and HP are doing. The toughest part will, of course, be keeping McNealy's big mouth closed.
Re:Bravo, Sun. (Score:2)
I'm not processor architecture expert, but it would seem to me that exteneding a 20+ year old design would be technologically inferiour to a design developed in the last 10.
From a business perspective I think 64bit x86 is a better bet, a more gradual progression, but not technologically superiour.
Re:Bravo, Sun. (Score:2)
signs of Intelligent life found in the SUNW! (Score:3, Insightful)
Sun reminds me of Atari or Amiga from days past...great company with lots of innovative ideas, piss poor execution.
They really need to spell out the future for their customers, will they adopteron the Operon for all servers eventually or is this just a little hack to keep the analysts off their back.
If they treat this like their x86 servers with annoucements like:
"We'll sell you this x86 junk if you really want it, but if you want to do anything serious give us a call about our UltraSPARC servers running Solaris!"
Comments like that don't incite confidence that as a customer I'm going to get support. Or long term roadmaps.
Re:signs of Intelligent life found in the SUNW! (Score:2)
Sun's x86 servers right now are durn cheap, considering you get dual SCSI drives, dual ethernet, etc, etc. Just as cheap as comparable Dell boxes, for example.
This was started last year (Score:5, Interesting)
There are a bunch of boxes on the drawing board, the ones they announced are just the first of many. The delay is that there is no real support for Opterons until they ship Solaris 10, which is due in the not to distant future. Until that OS hits, the Opteron support will be pretty half baked, just Xeon code, and no real use of AMD64 extensions.
That said, without trying to sound to much like a whiny martyr, I have been writing this stuff up for the last year on the Inquirer, just no one believed me
-Charlie
Re:This was started last year (Score:2)
Seems like it might actually be something that would at the least be good for Sun, especially if they're intent on moving into a little more of the x86 space.
Re:This was started last year (Score:2)
They need to abandon the SPARC ship, but they still have a lot of good engineering talent in that division so it would be a waste to just let them all go. They need to move that talent to working on AMD's chips, and compete with Intel not on pure price, but new features as well. Intel is a slow moving giant, if they c
Re:This was started last year (Score:2)
Perhaps the most important thing for Sun is to keep software vendors supporting Solaris - especially 64 bit Solaris. Hardware isn't worth much without useful software and having a large market for Solaris x86-64 software can be beneficial for teh Sparc market. The effort of porting from x86-64 to Sparc and back should be pretty small.
One of the chief selling points for Solaris is
Re:This was started last year (Score:2)
Re:Yes, but what is the target Operating System? (Score:2)
BTW, Sun's website had directions on how to set up dual boot Solaris/Linux on a Sun Blade.
Sun really is good at designing processors (Score:5, Insightful)
Sun really is good at designing processors. It's just that because Intel won the volume war because it happened to be the processor for the peecee, it was able to scale up manufacturing to cut prices even more, and sell to PHBs who care about price, not quality. Had IBM gone with the Motorola 68000 back when the first PC came out, which almost happened, we would see a totally different landscape today, where Intel would have probably gone the way of companies like National Semiconductor or Zilog. Imagine the first Linux kernel could have been written for an architecture with 4 times the registers. But alas, today, perhaps our only hope to remove the x86 plague is the PPC.
Re:Sun really is good at designing processors (Score:2)
No, they aren't. The UltraSPARC III's ass is kicked around the block by both the IBM POWER4 and Intel-HP's Itanium 2. There's no rerason to belive the UltraSPARC IV won't be handed its ass by POWER5.
Hell, Sun isn't really good at designing SPARC processors; the Fujitsu SPARC64 V kicks Sun UltraSPARC III's ass, and the dual-core SPARC64 VI will kick the dual-core UltraSPARC IV's ass.
Sun builds complete systems, and the complete systems are pretty good. But th
Why Wait for Sun? (Score:4, Informative)
And it comes preconfigured with Linux (Score:2)
6GB of memory can be installed, an unusual number. I would have expected a larger upper limit. If you really need to break the 4GB barrier, 6GB seems low.
I think we can finally write off the Inanium.
Humm... (Score:2)
Designing systems? Well, if they're just going to be come another motherboard manufacture, why would you ever buy an overpriced Sun System? I've seen sparcs handle loads that bring expensive x86 systems to their knees. Can opterons come in 106 processor [sun.com] combinations? I see they want to intoduce some lower end options, but their lower end models give them a lot of
Re:Humm... (Score:2)
Can opterons come in 106 processor combinations?
Yes. Will 10000 [com.com] do?
No, it's not available yet. But the point, I think, is that it is feasible to put lots of opterons in one box.
Sun is doomed (Score:2)
I think Sun is basically doomed. Unlike the restaurant industry, the computer industry doesn't have much of an expensive vanity market. Sun has nothing to offer in their hardware product line that AMD and Intel-based solutions don't do better and cheaper, and S
Eternal Sunshine of the SPARCless mind? Nah. (Score:2)
Like in the past, they'll find their customers don't want cheap x86 processors. They want Sparc processors that will run all of their existing apps and tools without having to port it. If they want su
Re:SGI tried this and failed miserably (Score:2)
So you are saying Sun should focus on making the Sparc the best processor money can buy. Period. Cost is no object. But if you factor in price/performance they would suck, but if you have infinite money and don't care, its marginally better?
OK. I guess so.
Re:It's Friday night... (Score:3, Insightful)
that depends... which one (if either) of you is posting from work???
Re:pardon me (Score:3, Informative)
Re:pardon me (Score:2)
Re:pardon me (Score:2)
I've not tried Solaris, but my top (bottom?) pick would be Windows ME
Who said OS? (Score:2)
2.6? (Score:2)
Re:LINUX IS A TOY (Score:2)
IANAL? (Score:2)
What on earth does that have to do with you being a lawyer?
Re:IANAL? (Score:2)
Re:LINUX IS A TOY (Score:3, Informative)
Sure Solaris is a dog on a lightly loaded system. But when your load average is sitting at 30, it's still performing near the same level. x86 boxes would fail under the load that Sparcs can hold up under.
And they're bloody reliable, and when they break, Sun's support contracts are excellent. Only HP and IBM compare on the support side, and only HP and IBM's RISC boxes compare on the reliability side
Re:LINUX IS A TOY (Score:2)
But I believe its hardware and not OS that sets them apart from pc's. Yes, the os is important but load is really backed up instructions that are not executed. Its all i/o. A fast processor can certainly help because it finishes processing faster. However sun machines have multiple busses that are backplained and specific memory that is wired to several cpu cores.
Very different architecture. To prove my point, the fastest servers mo
Re:LINUX IS A TOY (Score:2)
On the same hardware, FreeBSD will handle more load, but Linux will handle the lighter load faster.
And Itanium boxes aren't even in the running for the fastest servers, unless you are talking Altix 3000 Clusters. Itanium is a bloody dog performance-wise. Now for a trulay fast server, you are going to spec high-class Power5 or Sparc hardware.
Re:LINUX IS A TOY (Score:2)
I do suspect it's more of a platform issue than a CPU issue.
Re:no kidding. (Score:2)
It aint.
Re:pardon me (Score:2)
My linux boxes don't freeze on large file transfers, not even close. I can transfer large multi gigabyte files over gig-e from a scsi device, and play UT2003 at the same time with no noticable performance hit. Or transfer large files from one drive to another while mythTV runs without dropping a frame.
Don't you think "machine freezing on large file transfers" would be a bigger issue if it were true?
Re:Funny (Score:2)
SGI... (Score:2)
Apperantly they were all diffrent inside, and had signs of manual re-wireing. We may very well have the only one in existance
Re:departure from R&D (Score:2)
Or perhaps a Java programmer. Would that be Forte4Forte, Forte4Java or Forte4C (now all known as Sun ONE, with the later formerly known as Sun Workshop). Erm. Maybe I shouldn't have admitted I know that...