Sun May Use Opteron Chips

Runnin_Rob writes "CNET Nets.com is reporting that Sun is likely (not definite, but likely) to start using AMD's Opteron in the near future. The article also discusses how Linux is pushing for greater acceptability of Solaris x86 because 'All of the sudden it is OK to (put) something other than Windows.'"

Comments

sun needs to drop sparc

[mike_scheck]

Sun is losing market share and fast. It says a lot when you take the top of the line sparc chip, and put it up against a chip a quarter of the price that kicks its butt. Granted sparc is scalable (Scalable Processor ARChitecture), but not every case calls for a 24 cpu machine. Even hp has hpux running on Itanium. Sure sun has solaris x86, but they even tried to drop that!

Re:sun needs to drop sparc

[cant_get_a_good_nick]

Even hp has hpux running on Itanium.

Thats not that suprising, since Itanium is explicitly designed to be backwards compatible with PA-RISC. From the initial drawing on napkins in bars, Itanium was to replace PA-RISC and run HP-UX. They helped design the chips. Itanium II is more HP than Intel design from what I remember. HP actually had 64 bit design experience with PA-RISC and it shows in Itanium II. Carly Fiorina's plan saw R & D as a cost, not an investment, and has essentially killed all R & D and HP will now use commodity parts. Many people think this is a mistake, and now HP is up shit creek because the wanted economies of scale aren't even close to materializing. Itanium the 1st was a dog and though Itanium II is better, it's still workable only with huge megacaches. The whole EPIC thing requires very smart compiler writers, and will take a while before it comes out. People aren't buying them

Oddly enough, HP increased some performance by doing the opposite of EPIC in some PA-RISC chip: instead of forcing parallelism at compile time, there's a dynamic recompilation at runtime where you know what paths you are taking and you can optimize. The Sun compiler can use some runtime data and feed it to the optimizer, but this was happening all in-chip.

Re:opteron form factor

[bsharitt]

Could this be the begining of the end for SPARC? Will Sun start to adopt a strategy that HP/Compaq has with Itanium and Alpha, and just keep SPARC around for current users.

Re:opteron form factor

[domninus.DDR]

Yes but the first ones wont ship with an agp slot :( I know arima and MSI have retail boards (soon?) availible. look here, about halfway down

Linux helping Solaris?

[stevens]

Not at my workplace. We're mostly a Solaris shop, but it's not buying us much. We have to load new boxen chock full of GNU software to make them comfortable to work on.

Much of our software is Java, C or Perl-based. The Solaris JavaVM sucks donkey dicks (it's no better than linux, anyway), we use GCC (not Forte), and our Perl is portable to linux with a single scp.

Solaris buys us performance on machines with more than 16 CPUs. But we don't have any! Anything that needs serious cycles goes on the S/390 or AS/400s.

When the leases come up, it will be interesting to see how many Solaris boxes go out, and linux boxes come in.

Re:Linux helping Solaris?

[Skapare]

Anything that needs serious cycles goes on the S/390 or AS/400s.

Either the apps you've deployed on those machines are more I/O hungry than CPU hungry, or you've wasted dollars on mismatched architecture. S/390 and zSeries (no comment on AS/400 since I don't really know that one) are great machines if you need absolute up time and fantastic I/O throughput. But for CPU power, while those machine do have some, they are not giving you the bang for the buck you can get with a farm of P4s or AMDs. So maybe the reason you do have those machines is for something other than, or in addition to, CPU power needs. Does your S/390 serve web pages? Is it running a database? Does it have a PCICA?

Re:Linux helping Solaris?

[Anonymous Struct]

We're in the same boat as you guys. We had a slew of Sun hardware that just didn't need to be Sun hardware. I like working with Sun, and I like working with Solaris, but I just can't recommend Sun for most things in our environment. The fact of the matter is that we get fantastic deals from Dell on x86 servers, and using linux, we can (and will) replace most Sun pieces in our infrastructure.

Like I said, I've got nothing against Sun. I like working with their hardware. But when my boss asks me 'how much will it cost?', Sun leaves me in a real bad spot. Now they've realized how much they've priced themselves out of markets like ours, and they're working on putting out reasonably-priced systems these days with things like this amd deal or by using standard-registered dimms that you can buy from third parties on the cheap. But it's too late, I think. We started migrating away from Sun a while back, and we're not going to swap vendors again now.

The bottom line is this: What is Sun going to offer me in the linux/x86 world that Dell (or insert your vendor here) isn't? Better support? Lower prices? Better hardware integration? Sorry, we've got all of that. Our core applications will continue to run on Sun hardware for the forseeable future, but the low/mid-range stuff is already long down the road of x86 and linux, and Sun was just way too late to the game.

Re:Linux helping Solaris?

[b0r1s]

Same situation, but we went with FreeBSD.

The OS is cleaner, the speed better (especially when you push the system, using close to max RAM, because the FreeBSD swapping alg. is smarter than Linux), and the ports and upgrading systems make the systems much easier to maintain.

Finally, FreeBSD has much better system documentation (manpages for EVERYTHING), and all of those 'linux only' applications can run (quite quickly) under emulation (even NVidia is finally catching up, with their binary drivers).

(I guess I'm a new resident FreeBSD fanboy - so be it)

This would

[abhisarda]

mean that Sun would probably phase out Intel chips in the next 1-2 years in its lower end Linux systems. They will move entirely to AMD for their 32 bit lower end Linux and 64-32 bit mid level systems.
Given that so many companies: Sun, IBM, Dell want to increase their 64 bit x86 offerings, Microsoft *will* have to work double time to speed up their version of 64 bit Windows.
Already 5 varities of Linux, 3 BSD's, IBM's DB2, CA Ingres and Oracle have confirmed firm support for Opteron. Delaying Windows for this segment will mean that as Opteron becomes popular in the coming months, Linux will become the dominant operating system. This will mean a further boost to Linux.
A few months back Sandia National labs signed up to put 10,000 Opteron's in a supercomputer named Red storm which is supposed to become operational in 2004.

Re:sounds concrete

[Jahf]

FWIW, Sun has a long-standing behavior of taking extra time to test new hardware, losing the cutting edge in favor of higher stability. "Near future" in this case probably means a year or so away.

The upcoming round of x86 servers that John Loiacano alludes to, which by the definition of "near future" are coming out in the "extremely near future", are definitely not going to be based on the Opteron. It has already been leaked that the servers will be Intel Xeon processors running at least 2.8Ghz speeds.

Just trying to clarify without being too specific :) since this reminds me a lot of the "leaked" specs on the Cobalt RaQ XTR, which about 2 weeks before it was released with Pentium 3 chips, was reported here that it would be using AMD Athlons (mostly because the RaQ 4 and Qube 3 had used AMD K6 chips and someone extrapolated Sun / Cobalt would continue to use AMD chips).

The x86 groups within Sun are relatively processor agnostic. They try to choose the best match for the product's price/performance. Sun currently has various product lines that use both the Intel and AMD x86 processors.

Re:I thought Sun already had a 64-bit CPU

[mimmm]

Personally I'd love to have a CPU architecture that fully departs from the x86 designs, whether it be 32 bit or 64 bit

Have you ever heard of these machines called...ahh...what was it...ummm...APPLE!?

Seriously though, OSX is already a *nix, and from what I understand you can run a number of flavours of Linux ( Yellow Dog is one...) in/under/over/though OSX which means you can still do things like use the dvd reader/burner and use the firewire port etc, etc. If you're not with in Apple's price market, you could always buy an old cheapy one and run Linux on it without the OSX overhead. Or even build your own.

Also it looks like you would have the option of 64bit architecture within 6 months or so (courtesy of IBM's PPC 970). And although there wouldn't be any 64bit apps to start with, how long would it take the Linux masses to fix that?

ps - please don't give me any grief like "we don't need 64bit - it's only for GODS - not meer mortals like us" because you people can just go a use your AI/particle renderer/speech recognition/hand writing recognition/bloody mp3-4 player on that AT286 over there - ok?

I wonder...

[DragonWyatt]

How are they going to build 106 cpu boxes with opterons?

Maybe somebody more familiar with the architecture can chime in here...?

I've been following the opteron's and..

[Visaris]

I've been following the opteron's progress for quite a while now (3-4 months). Well, to be honest, I type "AMD" in the google news search and read anything that comes up everyday before I actually start doing any work. While SUN seems the be the largets potential supporter, Newisys, MSI, Appro International, smaller MB makers, and various vendors do seem to be giving good support. Even Microsoft has a working (I've seen some screen shots of a beta) version of windows for x86-64 called "Anvil" (Not yet released of course). Linux is up to speed, RedHat and SuSe and I don't know who all else has support. The chip isn't even out yet! I think things are going to work out for AMD and the Opteron/Athlon64. So far it seems like a good product, and I can say for meself at least that the delays only make me want one more.... If SUN jumps on the AMD wagon from the start as it looks like they're planning on, I think it may be what saves them.

w00t for AMD!

Solaris 9, the best Unix of 1995

[jeramybsmith]

So, I haven't really used Solaris much since the 2.5 days. I decided to give Solaris 9 a go on my Sun Ultra2 box. Guys, installing and setting up Solaris once will teach how far Linux has come. The fancy "webstart" install starts with an xterm with a text based disk partition tool. I haven't seen something like that since on Linux since 1999 (and the distro that did it was acknowledged pure crapola). Package selection during install was horrible. Not to mention it let me remove something that cuased kernel panics so I had to reinstall with a default package set. Once you get it installed, you got get it up to date. Of course, Sun's main update system is the "look around on our web site and pray what you download is for your version" like Windows and MacOS had 10 years ago. There is also a utility called patchpro with withh automagically download patches and apply them to your system. Of course, it comes in a tarball, requires you to alter your path, and finds weak excuses not to install patches thus requiring you to do a bunch of research to find out the reason. Installing patches on Solaris could only get better if they sent them to you on 9 track tape. Seriously, I was a big user of AIX, Solaris, and HPUX back in the mid nineties and I am sad to say that for Solaris not much has changed. I downloaded the Sun branded gnome2 for Solaris 9. Ich! Ximian really fleeced these guys. You get a clunky dual-toolbar desktop thats actually not much better than CDE. Someone sould have told Sun that gnome was all about the apps. In short, I still think of Linux sometimes as the distro I first installed by picking kernels with the correct CD driver to make boot floppies with back in 94. However, Linux has gotten to the point where its install and administration are superior to windows! However, the big Unixes are still behind the curve and wondering why they're losing market share.

Good thing for Sun

[moduc]

Long before Sun evaluated the first AMD chip for blade, I was thinking this would be a good idea for them. Regardless of many things, prices is just the tipping point. Using AMD, Sun does not have to put too much effort developing chips. AMD and Intel has quicker CPU cyle because of the Desktop root. This makes Sun's offer more competitive in term of price, publicity, and performance (at whatever cost they charge).

I am surpised that people compares Sparcs, and AMD (interm of its vitality), while they don't bother look at a real life example. IBM sells Intel chips for years, and doing very well with it, but they still sell their power line of chips. Does it really matter if the power line failed and their intel server succeeded? Sounds like a negative impact, but not really since if not supporting x86 line, and their Risk/Unix line is dead, they are worse off.

So, for a company, Sun supporting the AMD and Intel is no brainer. Cost is everything to business. Whether Sparc dies or not is irrelevant as long as you execute things right.

Time change, things change, your business must change, and sometimes your processors must change too.

People don't by Sun's hardware because 3 factors:

1) High cost (compare to the commodity of x86 HW)
2) Close (closed hardware and software environment)
3) Company and industry images (people tends to be nervous if everyone else support linux, and they also fear if Sun goes down).

So, lower the cost, open up with widely support x86 hardware, show that you're more competitive, and you'll survive and thrive.

USIIIi finally available

[Asdex]

Sun is now offering UltraSparc IIIi processors:
http://www.sun.com/processors/UltraSPARC-IIIi/

They do have some similarities to AMD's opteron processor:
- 1 MB on-chip L2 cache
- integrated memory controller
- 128bit DDR Ram
- large L1 cache

It should be interesting to compare those two processors.