“Gunningham’s departure dredges up lingering questions. Can Apple ever escape its desktop perch and march into the data centers of America Inc.? And what does Apple have to show for the past two years of efforts to sell powerful servers to businesses?” Alex Salkever is wondering for eCommerceTimes.
I am assuming Apple is taking its time improving Mac OS X so it’s a production-quality Unix. The achievements in performance they made with Panther are a good start; prior to this, Mac OS X (Jaguar and earlier) was not a good server platform.
Apple still doesn’t seem to get it, however. Rumor sites seem to indicate that Apple has this “enterprise group,” yet isn’t going at it full-bore. Apple is still way too conservative (albeit profitable, out of debt, and gaining momentum).
I think that Apple should buy Sun. This would make sense in many ways:
– Apple would get Java, and Java is good.
– Apple would get the Pixo stuff, if they don’t have it already (iPod OS).
– Apple would get Sun’s customer list.
– Apple could integrate any good parts of Solaris that it sees fit.
– Apple would have more workstation/Unix/server development knowledge.
Seems like a pretty good fit while Sun is cheap. Of course, there would undoubtedly be ego clashes, etc.
I would very much love to see Apple become a great research organization again.
Steve
years ago when cobalt began selling cubes, i wrote to apple stating that the small business market needed a small appliance server. options such as internal tape back-up and workgroup solutions would be required for such a market. sure the xserves are great, but why does a small business need such hardware? apple has a great core business, and i’m a client, but they don’t listen to their customers, they simply strike out at virgin territory without having a solid base. If you want to get past 3-5%, grow your base. such a base will provide the best marketing through referrals.
I am assuming Apple is taking its time improving Mac OS X so it’s a production-quality Unix. The achievements in performance they made with Panther are a good start; prior to this, Mac OS X (Jaguar and earlier) was not a good server platform.
Apple still doesn’t seem to get it, however. Rumor sites seem to indicate that Apple has this “enterprise group,” yet isn’t going at it full-bore. Apple is still way too conservative (albeit profitable, out of debt, and gaining momentum).
Well, I wouldn’t say their too conservative. The fact is, and I don’t blame them, they would much rather go in and make a good impression rather than rushing in resulting in customers regreting the purchases. Just look at the number who regret moving to NT4 under the illusion that it was “UNIX stability with out the price!”.
They simply don’t care much about Computers much anymore. A huge cashcow for them is the consumer electonics sector, where things like the ipod have done wonderfully well. I’d like to see Apple focus on that.
Apple buying sun wount happen in this millenia. (the other way round could be interesting but again will never happen)
IMHO, OSX is nice, but for an enterprise… I’m sorry but no.
X.3 is a nice, but X.0-2 was… a joke. Maybe by X.5 it’ll settle down enough for enterprises to consider them in 10Km of their data centers.
The issues that Apple need to address to get into data centers is not that they don’t run the software. OSX is close enough to other Unices that if vendors wanted to port software they could. I think the issue is more on monitoring and production settings. How many SNMP hooks does OSX have? Can I run security updates with out rebooting the server farm? When can I stop reinstalling patches every day of the week and build a ‘stable’ enviroment? Can I do full text only configuration? The list could grow very long with stupid and smaller questions like those.
Don’t get me wrong I love MacOSX. But I would not want to configure 500 MacOSX servers and keep them going 24×7.
I continue to be confused about what XServe does for Apple strategically. It doesn’t help them sell iPods. It doesn’t help them sell iTunes. It doesn’t even really help them deliver their own content or services – they could either custom-brew G5 1U cases internally, contract out the work, or just serve content on Dell boxes…no one would really care.
No one is disputing that they are a decent product, but Apple is nowhere in the serving market. You can buy POWER rack boxes from IBM, if you were stuck on the architecture. More to the point, OSX Server is a not even a footnote in the breakdown of server OSs in terms of market share.
I would be surprised if they are in production two years from now.
Apple makes consumer computing devices aimed at the creative set and the home user. Sun makes boxes for people who explicitly do not want to go low end. Sun has $5 billion in cash that you have to pay for too in any acquisition, making them an expensive target. Yes you get the money back the day you acquire them but that isn’t the point – Sun is for many reasons a very expensive acquisition target considering their prospects.
In any case this is framed in the outdated logic that all of Microsoft’s competitors should acquire each other. This line of thought has been dead for two years.
To the person who suggested Apple should buy Sun: Apple has a market cap of 8.8 billion, while Sun even at the current depressed price has a market cap of over 17 billion.
If a merger like this happens, you would call it Sun buying Apple, not the other way around.
The current patch cluster for Solaris 8 contains 323 patches. This is for an OS that is over 3 years old. Please do not tell me that OSX has too many patches for a production environment. You must obviously not work in IT but comment on ‘what you think is right’. That is okay, because comments are welcome here. But when someone is wrong someone also needs to step up and correct these statements.
According to Apple, Xserve was built in response to explicit customer requests.
It’s only with the Xserve G5 and the introduction of hardware RAID and ECC RAM that the Xserve has become even close to “enterprise ready”. Even now, it still lacks niceties like redundant power supplies and SCSI (even just as an option) that competing products have had for _years_.
Then there’s the price – Xserves are nice machines, but they’re priced well into the upper end of the 1U server market (with precious little to offer that justifies the high price) and they have nothing to compete with lower end models like Poweredge 650s or IBM x305s.
I was always under the impression that the XServe was really about letting people like Pixar (if they moved to an Apple platform) build a render farm easily.
The enterprise market is such a different market to what Apple usually targets that I don’t think they are really after that.
Anyone went to fosdem this year? Did you notice the massive ammount of ibooks and powerbooks? It was near 40% of all the laptops I think, and most of them were running mac os x. The year before, it was more something like 4%. So there is a large ammount of interest in the unix community for mac os x. Looking at the rather large growth this year, I’m thinking Apple can get a +70% market share in the unix desktop market. (kinda like the graphical market)
The unix server market is a different beast. There you have pretty tough competition. On the unix desktop it’s pretty “lame”. You have the linux desktop, which is pretty much immature. (cheap tough)
Any company that cannot even put a high-quality power supply in their top-of-the-line workstation is not ready for the enterprise. Or for any other business market either. If things go the typical Apple way, it will take a class action lawsuit to make Apple fix the G5 power supplies. Which, by the way, have to be 600W to power the PPC970 which is a heat/power monster like like Pentium 4E.
And what is IBM going to do? Sell chips at a loss to Apple so Apple can compete with them? The absurdity just makes me laugh.
Apple + Enterprise = appetite for destruction. Apple is a niche player for artists, creatives, and richie elitists. Apple has never made a business computer in their history. I don’t think they’re going to start now.
Any company that cannot even put a high-quality power supply in their top-of-the-line workstation is not ready for the enterprise.
As far as I’ve heard, there have been no functional issues with the G5 power supplies whatsoever, only acoustic, which is hardly an issue in an enterprise system.
Or for any other business market either. If things go the typical Apple way, it will take a class action lawsuit to make Apple fix the G5 power supplies.
Do you know anyone who’s had a problem getting their power supply replaced? As far as I’ve heard Apple has been perfectly willing to replace the power supply for any dissatistfied customers.
Which, by the way, have to be 600W to power the PPC970 which is a heat/power monster like like Pentium 4E.
Oh come now, this is ridiculous hyperbole. The heat dissipation of the 1.8GHz PPC970 is documented at 42W, compared to 68W for a 2.8GHz P4. The IBM eServer JS20 only needs a 300W power supply for two 1.6Ghz PPC970 procesors.
The 600W power supply and fan setup in the PowerMac G5 are just overengineering by Apple.
Why would anyone in IT buy a MAC when you can build a perfectly good celereon server for $199 as long as you dont want a lighted case or shiney eye candy metal like appples. Apples will never be good for servers until they let you build your own and compUSA offers some good rebates so you can buy Windows XP for servers.
a celeron server?? thats the biggiest oxymoron ever, not only can a 1 user workstation pound a celerons performance into the ground, i’d hate to even think what any real amount of users would do to it.. and also, the ‘perfectly good celeron server for $199’ is a pretty funny line as well..
$199.. for a celeron server.. funny funny…
lol. slam dunk.
The cream on the cake was when he mentioned getting Windows XP as a server OS. Very funny
“Apple has never made a business computer in their history. I don’t think they’re going to start now.”
Actually they’ve made quiite a few in the 80s and 90s, just nothing since Jobs came back.
Also, the XServe was originally introduced for rendering. They had the QuickTime team introduce it and they showed off zero configuration networking for doing cluster renders.
… when you can build a perfectly good celereon server for $199…
I think I’ll address this in some less than a sarcastic tone, because many people make statments about such and such being useless for servers – although very true – without explaining why.
Servers need several things above and beyond desktop computers, and almost the complete expense of everything else. I could spend all day writing about whats good in a server, and what isn’t, but I’ll cover the basics to your question here. The main things servers provide over desktops is reliability and expandability.
Networking – assuming that the $199 even has networking, it will most likely have some intergrated generic junk like a Realtek or VIA based network. Servers will have a decent network card/chip, most likely from 3Com or Intel (the most common in my neck of the woods) that do a lot of pre-processing for the CPU, reducing it’s load.
Mass storage – Once you get beyond about two IDE drives (or eight tops, if you use something like a 3Ware hardware RAID controller – not those $25 highpoint/promise software RAID cards), it all becomes about SCSI. Reduced load on the CPU, hotswapping, and the magical ablity to quickly go from 10 GB of single HD storage to 10 terrabytes of RAID 5+1 storage.
Bandwidth – 64bit/66 MHz PCI versus 32bit/33 MHZ PCI, gigabit versus 100 mbit ethernet, 400 MHz 64 bit Northbridge versus 133 Mhz 32 bit… Servers have much much more, and often – it’s massivily parrellel versus
CPU – Cache (64kbytes versus 8 Mbytes)and bandwidth (see above)
ECC – error checking and correction. Found in the RAM, cache, CPU, hard disks…
And so on, and so on. Do a google search, you’ll pick up quite a bit.
Apples will never be good for servers until they let you build your own and compUSA offers some good rebates so you can buy Windows XP for servers.
Well… no. First off, Apple isn’t interested in letting people build their own, that’s not what they do. For anyone running a serious server (that is, one that people who are paying you complain when it’s down), it might cost $1000 to get a 99.99% uptime rather than a 99% uptime, but if that extra 0.99% costs, or pays them an extra $50,000 it’s money well spent.
Oh, and XP (even though a server edition exists) isn’t for servers. It might exist, but anything that wastes memory on anything but the most basic GUI is nothing but a toy.
(Yes, I run linux, and no, I don’t have a GUI on my servers, but that’s not the point)
sorry to be dumb…. but you can’t run any of MS servers without the GUI can you? I do understand why you run your linux in text mode.
thanxs
Anyways, yeah, Apple’s a no-go for enterprise and business, but the activities of artists, serving, or clustering converge just a bit — It’s not just for IT professionals. And it’s only going to converge more in other areas too.
This is not really for Enterprise admins, Business computing, etc..And it never was meant to be! Though it very well could be sooner or later. But everyone I work with are either audio and video artists like myself or 3D and Animation artists. We don’t need an IT department, but we can put alot of power to use — Which just happens to be power typically reserved for servers.
Apple’s not trying to shift it’s target (yet) — They’re just providing something useful for the market they’ve already been involved in. With their entire lineup they’ve attracted and/or regained some of the scientific and education market too, but even that was just a plus. It’s not like they’ve shifted focus. It just happens that their products come in handy for all kinds of jobs.
Anyways, on a sidenote: I see an Apple/SGI partnership (if at all) far more likely than Apple/Sun. Well, at least that would make more sense.
but you can’t run any of MS servers without the GUI can you?
I believe it is possible to run Windows “headless”, but from my understanding, it’s a lot more difficult to do, and some functions are essentially locked out, or incredibly difficult to access/make changes.
My point about Windows XP server is that if it’s going to have a GUI, it shouldn’t be pretty. Pretty uses more ram, and means more things to go wrong. This is a Bad Thing[tm] when it comes to servers.
You do realize that there are scenarios where a $199 celeron server is totaly acceptable such as redundant web servers.
You do realize that there are scenarios where a $199 celeron server is totaly acceptable such as redundant web servers.
Redundant server for whom is the question. How’s Windows XP and a Celeron (or two) even relevant? Check this guy’s line of reasoning again:
Why would anyone in IT buy a MAC when you can build a perfectly good celereon server for $199 as long as you dont want a lighted case or shiney eye candy metal like appples. Apples will never be good for servers until they let you build your own and compUSA offers some good rebates so you can buy Windows XP for servers.
OK. Windows XP, IT, Celeron, Shiny Apples, CompUSA?
I wonder if they might be better off not trying to launch Apple into the enterprise, but to set up another company to do it as a joint venture with say IBM where Apple could provide OSX, hardware designs, and industrial design expertise, and IBM the chips, market knowledge, and credibility.
Competing in the WindowsNT space (easy to manage low end servers) you’d combine the ease of use of apple, and the market muscle of IBM, but without competing directly in eithers main markets.
Apple has a OSX license they can trust not to eat their market share (because they control them) and IBM gets another user of PPC chips.
They could even call it NeXT2
Just a thought.
Hello?! Doesn’t xserve run as the basis for the virginia tech supercomputer cluster (top 3 currently and will still remain in the top ten when the opteron clusters get finished)? That’s surely Enterprise level, no?
And if not, that surely defines a role for the xserve, with
its ECC memory (that the workstation G5’s don’t support at
the moment), it is probably used by many in the scientific and academia environments.
Apple will definitely make inroads into the server market, just look at xgrid for their investment in utility data center/on demand computing.
Does OS X have cluster capability?
My thoughts exactly. I thought apple was targeting scientific, education, and creative markets with its servers not really enterprise.
Apple was also talking about targeting small business, which again is not enterprise. My guess is that apple is trying to expand within its existing markets (those mentioned above) before going after enterprise in earnest.
I’d have to agree, with some IBM/Novell co-branding Apple could get LOTS of XServers in small businesses with limited IT departments and costs.
My last shop was an AS400 shop. They’re nice, but really expensive for what you get horsepower wise… A standard Xserver for like $3-$5K would save nearly $10K in upgrades for company intranets, email, and gateways rather than using the AS400. Most importantly, it keeps MS out of your back office! Linux is nice but not user-friendly enough for “older” IT guys…So everybody fills their backroom with special purpose MS boxes for email, directory, etc… When a XServer can do all that on one box, more stable than windows, and cheaper too! For IBM it wouldn’t hurt to help them out…after all, most of the low-end servers they sell all run Windows. They already lost that fight…but like people have said, there’s a market for easy to use “appliance” servers. That’s what XServer seems to do nicely at…and they are competitively priced for the horesepower too!
As Admin of a small business, I’ve was watching Xserver for a while. The IBM or MS stuff is just too F*n expensive with processor levels and CALS for everything. If Apple could really show that the servers could “play nice” as domain controllers and gateways with maybe some light web hosting duties they’d have a good niche. The trend I see in very small businesses [<100 people] is that moving to “appliances” makes more cost sense because money is wasted paying for “full-time” admins of server farms. The bar for the midrange market is AS400…with about 1 admin per 200 employees & 99.9% uptime![and able to do other stuff too!] Windows will never match that for small business. The time to simply keep up with patches on window boxes was 5-10X more than I ever spent with the 400…and admining is about helping the company…not playing with toys!! The great part is that you can learn to use OSS tools on your XServer too! Letting or even recommending people use OSS tools on Apple hardware wouldn’t hurt them at all…because the hardware is decently priced and better than Windows boxes. So once you get people off wintel it shouldn’t be hard to keep them off…as long as you keep up committment to the platform!!
“Does OS X have cluster capability?”
Very much so. Zero config, actually. You plug a bunch together and they do the rest for you.
Actually, no, the VT cluster was built from a whole lot of consumor grade G5 machines. However, the VT team is now replacing them with XServe’s built from G5’s. Why? Because they can put a whole lot more of them in the same space. A 1U form factor is a hell of alot better than the gigantic cases the G5 machines come in.
Why would anyone in IT buy a MAC when you can build a perfectly good celereon server for $199 as long as you dont want a lighted case or shiney eye candy metal like appples.>>
1) Because maybe I want to run a program like BLAST where the altivec unit in a G5/G4 processor allows it to cream the x86 competition.
2) Because maybe I want to render & serve video.
3) Because I’ve taken a look at the capacity/cost ratio of SCSI hardware and realize the IDE or SATA actually gives me more bang for my buck.
4) Because I want to run the same OS on my desktop as on my server — without needing hours of IT config time or running into dependency hell when a software update is run. (Happened where I work. A server went down for a week while Systems staff tried to sort out what happened. [Ain’t Linux grand?]).
5) Because I’m also the IT person for my small business/school department and I need an OS that I don’t need an MIS/Comp Sci degree to run, and that doesn’t have every back door on the planet open to script kiddies.
The VT cluster now takes up about 2/3 the space it used to take up so that leaves a lot of space for more XServes so VT juat may retain its number 3 spot just by purchasing more XServes in 2004. The ECC debate concerning the memory has been put to rest.
The XServe is a strategic product for Apple. It gives Apple customers an alternative regarding server offerings. So as a product it is important because think of those shops that run ASIP or NT but a majority of their clients are Macs. XServe gives them alternatives.
Apple’s enterprise efforts have so far been weak and I think part of it is they do not have the right leadership in that division to drive it forward. I wonder if Joel Koecher is still around. They need to be more aggressive. Apple also needs to foster better relations with developers.
Did someone mention that Celerons running WindowsXP and built from scratch or parts make good servers? I don’t think Apple is competing for the mindshare of these “server admins”
I was always under the impression that the XServe was really about letting people like Pixar (if they moved to an Apple platform) build a render farm easily.
Steve Jobs (CEO, Apple Computer) is also CEO of Pixar.
They only render on Mac’s. (That’s why it looks so damn beatiful!)
>> Hello?! Doesn’t xserve run as the basis for the virginia tech supercomputer cluster (top 3 currently and will still remain in the top ten when the opteron clusters get finished)? That’s surely Enterprise level, no?
I don’t think a supercomputing cluster at a university need necessarily be categorized as enterprise.
Bottom line, the cluster can be down for extended periods of time and not sink the mission of the lab. The same cannot be said for a business. Offline servers means offline business.
I am not stating this to imply the Xserve boxes are inappropriate for business, I am simply stating that research labs can in fact implement many types of solutions that would be totally inappropriate for a business.
“However, the VT team is now replacing them with XServe’s built from G5’s.” – how stupid is that? Just few months after cluster was build? From the beginning it was known that in spite of whatever software VT plan to use they need ECC (In general at that time G5/OS X was not ready to make usefull clusters). So total price for VT cluster will double. Now there are two questions: knowing that, they decided to pay twice as much (that makes VT cluster more expensive than anything else), or VT had a deal with Apple which makes highly possible that someune will sue apple for setting prices below system value.
Beside clusters have nothing to do with enterprise. Neither with BLAST or video rendering.
1) Because maybe I want to run a program like BLAST where the altivec unit in a G5/G4 processor allows it to cream the x86 competition.
And BLAST is a lot faster on Apples because Apple marketing says so? Yeah right. What those pretty bar graphs show is that Apple’s version of BLAST works better for absurd word sizes. And, while written in PPC assembly, it does not use any of Altivec neat features (like fast add/multiply), and can and was ported to any other arch.
Honestly speaking, Apple benchmarking now has a lot less credibility then a random bozo from the street for me. Way to go, folks!
2) Because maybe I want to render & serve video.
And linux and windows boxes can’t render and serve video… because?
3) Because I’ve taken a look at the capacity/cost ratio of SCSI hardware and realize the IDE or SATA actually gives me more bang for my buck.
Virtually every budget x86 server satisfies this requirement.
4) – 5)
Subjective, but could be an important point for you.
Having a machine that is a bit easier to administer gives value to the XServe, particularly in an Apple shop. The pricing is competetive with other 1U machines, and then you have the single sourcing angle which can make support easier.
Windows excels in areas like this because it’s fairly easy to piece together a network of a few machine and tie them together with a back end server machine. With some plug-n-play functionality, the rough details of a small network are set up quickly.
Apple could do that easily with yet another tower, but they obviously felt that a more specific form factor would be a better deal. For those who want to, the Xserve is more easily colocated than a G5 Tower.
Then, throw in their RAID device, and you have a nice form factor.
I see the XServe rather than lead of an attack on Corporate America, rather it’s a fairly simple measure to let Apple users keep a pure Apple solution.
“The VT cluster now takes up about 2/3 the space it used to take up so that leaves a lot of space for more XServes”
Actually it’s down to 1/3 the space.
“Steve Jobs (CEO, Apple Computer) is also CEO of Pixar.
They only render on Mac’s. (That’s why it looks so damn beatiful!)”
It’s been a recent switch (at the intro of the G5). Previosly it was a mixed user network and an Intel server farm. Nemo and all future movies are Apple only stuff, though.
“And BLAST is a lot faster on Apples because Apple marketing says so? Yeah right. What those pretty bar graphs show is that Apple’s version of BLAST works better for absurd word sizes. And, while written in PPC assembly, it does not use any of Altivec neat features (like fast add/multiply), and can and was ported to any other arch.
Honestly speaking, Apple benchmarking now has a lot less credibility then a random bozo from the street for me. Way to go, folks!”
Try benchmarking it yourself, I think you will be surprised that it comes out in the G5s favor. Forget what Apple says, do it yourself.
“And linux and windows boxes can’t render and serve video… because?”
Linux can render, yes, but when you use FCP or Shake (they only run on Macs now), then you will be using their built in Distributive Rendering, which Linux doesn’t do, and OS X does. As for media (ie- video) serving, Linux is way behind there. Ever tried to stream video from Linux? It’s many times easier and more powerful on OS X Server. Fun Fact: Apple’s QuickTime movie ad site (http://www.apple.com/trailers/) is the largest video site on the internet (measured by size of content and number of hits). Not to mention the power they have behind the iTMS downloads…last I checked it was well over 2 million sales (sales, not songs, as in an album of 15 songs is still 1 sale) a week.
“Virtually every budget x86 server satisfies this requirement.”
Check your prices again…the XServe RAID gives more band for your buck than Dell’s best server offerings.
“Subjective, but could be an important point for you.”
Hardly subjective. A single platform in a small business is much cheaper and simpler to maintain than multiple networks. Business 101.
Perhaps in my day job (today is a sick day) I’ve been over to the biology department of the University where I walked and seen a dual G4 go head to head with a dual xeon x86 box?
Try benchmarking it yourself, I think you will be surprised that it comes out in the G5s favor. Forget what Apple says, do it yourself.
Testing in biology apps? I can’t do this, and don’t need this. However, if SPEC is good enough for you, we could have a look at scores for PowerPC 970 (G5) from IBM whitepaper from http://www-306.ibm.com/chips/techlib/techlib.nsf/techdocs/7874C7DA8… –
Performance 1.8GHz
SPECint2000 828 (est)
SPECfp2000 1036 (est)
Dhrystone 2.1 MIPS 5800
Doesn’t look that nice compared to Intel or AMD.
Linux can render, yes, but when you use FCP or Shake (they only run on Macs now), then you will be using their built in Distributive Rendering, which Linux doesn’t do, and OS X does.
FCP is not really the kind of app for which people use distributed rendering. And Shake is not the only app one can use for video rendering, right?
Ever tried to stream video from Linux?
Never. However, from a google search, there is Apple’s own Darwin Streaming Server available for it. Maybe something else will work, too.
Check your prices again…the XServe RAID gives more band for your buck than Dell’s best server offerings.
I thought we were talking about the XServe itself. If you meant XServe RAID, then I agree with you. However,
1) you can connect it to any machine, not just a Mac
2) Apple does not have a lot of reputation for being a good storage solution provider, so they have to sell their products a lot cheaper.
The guy says, “Maybe you’re in a specific market and need a specific app.”
You claim APple’s benches are bull, he replies that the benchmarks do reflect real world performance, adn you say…
bleh, I don’t need to do biology apps. I want general useless SPEC benchmark scores???!!!
bleh, to your reply. It is a valid application in the scientific market. Scientists like ease of use as much as artists. Just because you want to ignore that doesn’t make it valid.
“Even now, it still lacks niceties like redundant power supplies and SCSI (even just as an option) that competing products have had for _years_. ”
SCSI is available as an option. They don’t sell it directly through the web site, but you are whining about being enterprise-class,aren’t you? If you talk to a service rep, Apple is happy to equip an XServe with SCSI. But really, it’s about time to move onto S-ATA anyway.
http://uptime.netcraft.com/up/graph?site=www.apple.com
http://uptime.netcraft.com/up/graph?site=www.army.mil
http://news.netcraft.com/archives/2003/06/10/recent_changes_at_nota…
“”However, the VT team is now replacing them with XServe’s built from G5’s.” – how stupid is that? Just few months after cluster was build? From the beginning it was known that in spite of whatever software VT plan to use they need ECC (In general at that time G5/OS X was not ready to make usefull clusters). So total price for VT cluster will double. Now there are two questions: knowing that, they decided to pay twice as much (that makes VT cluster more expensive than anything else), or VT had a deal with Apple which makes highly possible that someune will sue apple for setting prices below system value. ”
MP, don’t be stupid.
VT paid FULL educational price for the initial G5s and received the 1U Dual G5 XServes when they came out. They did not buy them again. This news is 2 weeks old.
Who is going to sue Apple? You?
VT wanted the fastest cluster they could get their hands on so they approached Apple, not the other way around. Varadarajan wasn’t even a Mac user at the time, he was evaluating sheer processing power. VT probably had advanced knowledge that an XServe would be coming down the pipe.
You may think VT is stupid but they have bragging rights no other university in the world has as well as the fastest computing cluster outside of the military/gov. That doesn’t sound stupid to me.
http://www.eurekalert.org/pub_releases/2004-01/vt-vtm012704.php
http://www.macobserver.com/article/2004/02/12.1.shtml
@kobold
“And Shake is not the only app one can use for video rendering, right?”
Right but it didn’t hurt Weta Digital at the Academy Awards.
“We could not have done ‘The Lord of the Rings’ trilogy without Shake’s fast compositing speed, quality and extensibility,” said Joe Letteri, Weta Digital’s Academy Award winning visual effects supervisor. “Shake was the cornerstone of our visual compositing pipeline.”
I take it Weta could have used systems costing 10X Apple’s solutions or systems costing 10X less. I’l leave it up to you why they use Shake. I am sure the speed of Shake on their G5s is imaginary as well also(sarcasm)
Why don’t you guys stick to the topic of Apple in the enterprise where they are not doing too well at all. This is the wrong day to be picking on their pro-apps.
Testing in biology apps? I can’t do this, and don’t need this. However, if SPEC is good enough for you, we could have a look at scores for PowerPC 970 (G5) from IBM whitepaper from http://www-306.ibm.com/chips/techlib/techlib.nsf/techdocs/7874C7DA8….. –
Performance 1.8GHz
SPECint2000 828 (est)
SPECfp2000 1036 (est)
Dhrystone 2.1 MIPS 5800
Doesn’t look that nice compared to Intel or AMD.”
Well if you knew anything about benchmarks you would understand how those tests work. SPEC tests are extremely geared toward x86, and floating point tests are extremely geared toward PPC. Hence how any half intellegent person knows that benchmarks are for the most part useless as a generalization. You benchmark something based on what you are going to do with it. And typically SPEC tests show the advantages of applications in which the user is the bottleneck, such as word processors, while floating point tests show the advantages of applications where the computer is the bottleneck, such as rendering.
“FCP is not really the kind of app for which people use distributed rendering.”
FCP is exactly the kind of app to use distributive rendering with. That’s what the XServe was original introduced for. Just try rendering a high def video (that’s 23 MB a second) on your single processor Pentium box.
“And Shake is not the only app one can use for video rendering, right?”
No, not the only, but one of the best. It’s made all the movies that have won Best Special Effects for the past 7 years.
“Never. However, from a google search, there is Apple’s own Darwin Streaming Server available for it. Maybe something else will work, too.”
OS X Server is the only platform QuickTime Streaming Server has a GUI with, and the only platform it is preinstalled on. If you are streaming video, you will want a GUI. It’s amazing how much easier it is to manage a large number of video with a GUI.
“1) you can connect it to any machine, not just a Mac”
True, but given that Apple makes money on hardware, it’s not very important what OS controls the storage hardware, as long as Apple sells it.
“2) Apple does not have a lot of reputation for being a good storage solution provider, so they have to sell their products a lot cheaper.”
Actually they have a great reputation for storage products, just ask any video professional. I know quite a few people who have been using Apple hardware for storage and rendering for the past few years without any problems. Enterprise servers aren’t the only reason to have a lot of storage.
bleh, I don’t need to do biology apps. I want general useless SPEC benchmark scores???!!!
SPEC is built from code from real-life applications. Is gcc real-life enough for you? or perl? I certainly see them in MY life, and those are subtests of SPEC.
“We could not have done ‘The Lord of the Rings’ trilogy without Shake’s fast compositing speed, quality and extensibility,” said Joe Letteri, Weta Digital’s Academy Award winning visual effects supervisor. “Shake was the cornerstone of our visual compositing pipeline.”
I take it Weta could have used systems costing 10X Apple’s solutions or systems costing 10X less. I’l leave it up to you why they use Shake. I am sure the speed of Shake on their G5s is imaginary as well also(sarcasm)
I just meant that there’s more then Shake, and if Apple, in it’s infinite wisdom decided to cut all non-mac versions of it, other systems can still be used for rendering.
Well if you knew anything about benchmarks you would understand how those tests work. SPEC tests are extremely geared toward x86, and floating point tests are extremely geared toward PPC.
And SPECFP is geared towards… what?
And perl or gcc from SPECINT are geared towards… what?
Check this out: http://www.spec.org/cpu2000/CINT2000/ . This should tell you, which applications SPECINT simulates.
Kobold, when you don’t specify a certain test, the term SPEC is generally thought of as an integer, non floating point test. They are usually used to represent the relative speed of one certain function, so as not to take into consideration any bottlenecks.
“And SPECFP is geared towards… what?
And perl or gcc from SPECINT are geared towards… what?
Check this out: http://www.spec.org/cpu2000/CINT2000/ . This should tell you, which applications SPECINT simulates.”
In your original statement you did not specify a SPEC test that tested the overall speed of a system, bottlenecks and all. If you want to get into the overall speed of a system, then SPEC tests fall drastically short of being relevent. By saying SPEC was geared toward x86 I meant that it focuses on one thing (hence the integer test)…usually clock speed. That brings up the fundamental differences in architectures, x86 focusing on clock speed and PowerPC focusing on effiency (btw, gcc is certainly not optimized for PPC. Being cross platform doesn’t mean being cross optimized). The effectiveness of PPC comes from how well it works, so when you test one specific feature, as SPEC does, you often come up with results pointing toward x86, and when you test overall speed, PPC usually comes out on top. I say often and usually because you can spin any benchmark to point to any result you like if you know how.
Ok, so drool with me… A Sun/Apple MERGER. Sun gets to replace it’s gui of gnome/whatever the older use they have, along with the advanced auto-configuring protocols and easy admin use, while Apple gets a rock solid kernel to run it’s system on.
Java, however, would be ‘sold’ in a sense, set up an open standards body of apple/sun/ibm/other_guys.
Sun migrates from SPARC to PowerPC, leaves linux as a low-end option, and moves to MacSOL v XI or somesuch name for the Solaris + MacOS X merger.
It could be amazing.
When I don’t specify the SPEC CPU test, it means considering results of both . It looks like we had a miscommunicaton.
If you want to get into the overall speed of a system, then SPEC tests fall drastically short of being relevent.
Well, SPEC CPU measures how fast your CPU and memory is in interger apps. Macs seem to have the same hard drives as PCs do, same PCI bus, etc, so the rest doesn’t really matter, right?
By saying SPEC was geared toward x86 I meant that it focuses on one thing (hence the integer test)…usually clock speed.
Opteron and Athlon64 processors are quite competitive with Intel processors in SPECINT tests, and their clock freqencies are just a bit higher then those of G5 (2.2 vs. 1.8). However, G5 fares a lot worse. You can’t blame clock focus here, can you?
That brings up the fundamental differences in architectures, x86 focusing on clock speed and PowerPC focusing on effiency (btw, gcc is certainly not optimized for PPC.
Replace x86 with NetBurst, and you are right. You can’t say “x86” – AMD’s processors break your argument. And, BTW, gcc is the Apple’s compiler of choice, isn’t it?
I still consider SPECINT to be quite a good perfomance measurement for integer tasks. Of course, if you want to archieve your “overall perfomance” measurement holy grail, just get a weighted average of it with SPECFPU. As I said, disks are the same on Macs and PCs. What other components you care about in overall perfomance figure?
“Well, SPEC CPU measures how fast your CPU and memory is in interger apps. Macs seem to have the same hard drives as PCs do, same PCI bus, etc, so the rest doesn’t really matter, right?”
Actually, a lot of other things matter. Macs and PCs do have the same drives, but certainly not the same bus. The G5 has more bandwidth (due to a high system bus and many smaller dedicated buses, the latter of which no PC system has). Then take into account the software you are focusing on. Integer optimized applications. Those are often used on PC’s because of Intels dominance and their focus on clock speed. Mac applications are written to take advantage of altivec (vector engine, floating point 128 bit vector processor, whatever you want to call it). These apps would run unbearably slow on x86, yet the scream on PPC. Hardware is only half the equation of a systems speed.
“Opteron and Athlon64 processors are quite competitive with Intel processors in SPECINT tests, and their clock freqencies are just a bit higher then those of G5 (2.2 vs. 1.8). However, G5 fares a lot worse. You can’t blame clock focus here, can you? ”
Guessing the test you are refering to is on the IBM link you posted, which I can’t see because I am not a member of their website (hate it when that happens). But judging from tests of that sort that I have seen before, I would assume you are once again running integer software tests. The Opteron and Athlon64 are very nice processors and I think AMD is great at what they do, but they still focus their hardware on running integer heavy software, and as I’ve said before, I don’t consider that a worthwhile thing to test because the bottleneck is almost always the user…making the computers speed irrelevant.
“Replace x86 with NetBurst, and you are right. You can’t say “x86″ – AMD’s processors break your argument. And, BTW, gcc is the Apple’s compiler of choice, isn’t it?”
Yes, I can still say x86 because while AMDs processors are more powerful than Intels, they still focus on integer stuff. No, gcc isn’t Apple’s compiler of choice, it’s the compiler they used to show cross platform benchmarks. That doesn’t make it the best suited compiler for that platform. Like I said before though, I ignore Apple’s benchmarks, as well as those of any computer maker, becuase they are always biased.
“I still consider SPECINT to be quite a good perfomance measurement for integer tasks. Of course, if you want to archieve your “overall perfomance” measurement holy grail, just get a weighted average of it with SPECFPU.”
SPECINT is a quite good performance measurement for integer tasks, no argurement there. The arguement was the relevance of those tasks in a cross platform comparison, when one platform focuses on them and the other doesn’t. Hardware’s still only half the equation. If I was comparing AMD vs Intel, SPECINT would be great to use. But that’s not what we are comparing. SPECFPU tests are not accurate for the reasons I stated above.
“As I said, disks are the same on Macs and PCs. What other components you care about in overall perfomance figure?”
Other components include software (the big one), bandwidth, bandwidth dedication, I/O speed, and the situation you will be using the system in.
The G5 has more bandwidth (due to a high system bus and many smaller dedicated buses, the latter of which no PC system has).
Bzzzt. Wrong. Opterons beat G5’s bandwidth. In a dual config, a processor has 2 HT links – one to other processor, the other one to I/O. Each processor also has their own integrated memory controller. Compare to G5, in which both processors do I/O and access memory through the single northbridge. Your statement applies to Opterons (x86) a lot better then to G5 ;-P.
Mac applications are written to take advantage of altivec (vector engine, floating point 128 bit vector processor, whatever you want to call it).
And Intel and AMD have SSE/SSE2. Which is also a set of SIMD instructions. So, what?
The Opteron and Athlon64 are very nice processors and I think AMD is great at what they do, but they still focus their hardware on running integer heavy software, and as I’ve said before, I don’t consider that a worthwhile thing to test because the bottleneck is almost always the user…making the computers speed irrelevant.
There are more then enough integer heavy applications that don’t have the user as bottleneck. Development, servers, databases… Those things seem to be worthwhile to me.
No, gcc isn’t Apple’s compiler of choice, it’s the compiler they used to show cross platform benchmarks.
http://developer.apple.com/documentation/DeveloperTools/Tools.html
http://developer.apple.com/darwin/projects/compiler/
Looks like they do point developers at gcc. So, it’s more then relevant as a Mac app – Mac people compile code with it!
SPECFPU tests are not accurate for the reasons I stated above.
I kinda missed this. Could you please explain, what is wrong with SPECFP?
Other components include software (the big one), bandwidth, bandwidth dedication, I/O speed, and the situation you will be using the system in.
Memory bandwidth – covered by SPEC. I/O speed – same on both platforms. What else?
The link to IBM brochure is getting mangled by OSAlert. Here is my local copy: http://24.80.15.91:32132/PPC_QRG_2-22-04.pdf
Anyone who says OS X is not enterprise grade is stoned. Go to netcraft and check out what server platforms the U.S. Army is using…OS X. If it was unreliable or unsecure they would shy away from it quickly. They use it because it is rock solid, secure, and easy to maintain.
MacOSX 4D_WebSTAR_S/5.3.0 (MacOS X) 15-Jun-2003 140.183.234.10 Army Belvoir Research and Development Center
I am rather amused by the fact that every time people attempt to put down Mac OS X Server and the X Serve they use the excuse that if it were ready for the “enterprise” than it would be in use at whatever business they are familiar with instead of education or creative markets. Let me tell you something, higher education has been in IT longer, has greater requirements, and qualifies more than any business as a “enterprise”. A given public university has hundreds of thousands of people in their directory system, complex security requirements, countless web and database applications, and incredibly complex HR and admin systems for dealing with government grants. In fact, higher education is where many of the technologies that business IT departments live by were born. (MIT Kerberos, Umich LDAP, CMU SASL, etc…) Mac OS X Server/XServe may not be the only server OS/Hardware used at these institutions but it is gaining ground, and for good reason too. It has the stability of Linux, without the pain in the butt configuration issues, and is cheaper than Solaris.
As for speed? I hardly think that your typical business, which runs Office and lives in excel has higher requirements for speed and bandwidth than a biology cluster, animation studio, or ad agency. These people push gigabytes around the network and crunch huge amounts of graphics and video data all day long. Did you know the XServe RAID + XServe can be reliably used to deal with HD video? Thats a lot of word documents…
The fact is, X Serve/OS X Server is a compelling choice for many enterprise applications. It has the stability and security of a UNIX OS with the “ease of management” of a windows server OS. About the only thing it doesn’t have is an Army of MCSEs that want to recommend it to you so they can charge you up the yin yang to fix it when it breaks every day.
SCSI is available as an option. They don’t sell it directly through the web site, but you are whining about being enterprise-class,aren’t you? If you talk to a service rep, Apple is happy to equip an XServe with SCSI.
They certainly weren’t offering it as an option for my prior employer when we enquired – and that was one of the biggest Universities in Australia.
Heck, I’ve never even *heard* of an Xserve with SCSI. You’d think if it were available at all, there’d at least be a few people who had bought them.
But really, it’s about time to move onto S-ATA anyway.
Who sells 15k RPM SATA drives ?