Please create an account to participate in the Slashdot moderation system

 



Forgot your password?
typodupeerror
×
Technology (Apple) Businesses Apple Technology

Xserve Outperforms Sun, SGI, Windows 127

Pahroza writes "Xinet has released their 2002 benchmark configurations, with tests including output generation and AppleShare file serving. Xserve was only bested by machines sporting at least twice as many CPUs as the two it was using. MacCentral is also running a story on the results, and you can download a PDF of the benchmarked configurations."
This discussion has been archived. No new comments can be posted.

Xserve Outperforms Sun, SGI, Windows

Comments Filter:
  • No shit (Score:4, Funny)

    by Starship Trooper ( 523907 ) on Monday July 01, 2002 @04:00PM (#3803105) Homepage Journal
    Apple servers work faster for AppleShare? Say it ain't so!

    Show me some less biased benchmarks, please.

    • Re:No shit (Score:5, Insightful)

      by dbrutus ( 71639 ) on Monday July 01, 2002 @04:34PM (#3803350) Homepage
      Since the XServe is aimed, first of all, for those Macintosh islands of creatives in corporate america as well as apple using scientific firms like genentech, Appleshare connections are very relevant indeed. I would expect that when Apple updates its UFS support to the modern spec and it gets a lot of other small details nailed cold, they'll branch out in their marketing but they're shooting for a specific target market right now and the benchmarks are relevant for that market.

      I expect that when Samba 3 integrates perfectly with Active Directory, Apple is going to go after the workgroup file and print market that finds Linux too technically challenging and is sticking for the higher priced Windows solutions for that reason alone. At that point they'll have a track record and IS organizations won't be so nervous about the Apple label anymore.
      • Re:No shit (Score:1, Insightful)

        by Anonymous Coward
        Excellent comment and summary. Some people don't seem to get it. Apple has clearly stated what their market is and they've set the benchmarks up in this case for that market.

        Original comment about bias is totally off target. Look at the specs for the tests, look at what the parameters are, then criticize the results if there's something phony going on as alledged.

        Thanks for thinking before leaping like the original poster in this thread.

    • Re:No shit (Score:5, Informative)

      by garren_bagley ( 413546 ) on Monday July 01, 2002 @04:46PM (#3803436)
      Xinet writes software for the publishing industry, magazines, newspapers, advertising. This includes AppleShare servers for the Macs in these shops. Thier AppleShare software was written for Solaris and SGI machines and was quite mature before the Xserve even came out.

      Apple may have targeted their design to this kind of thing since these are shops that would most likely be open to trying their servers. I don't know. If they did it sounds like a pretty good plan to me.

      I'm actually pretty impressed. The SGI 300 box is pretty sweet and incudes Ultra3 SCSI Drives. I wonder how much cheaper the Xserve really is once you've got the ATA Raid setup on it like the benchmarked machine had for the tests.
    • No, Schmuck. They work faster period.

      Aaaarrrggh.

  • Will corporations really take Apple seriously as a contender in the server market? This market has traditionally been dominated by Sun, among others. It seems by these performance graphs that Apple is really at the same level as the other major contenders as far as speed. Given that it has fewer CPU's, it is probably cheaper than some of the better-performing ones tested.

    Will companies really start using Apple's Xserve to replace Sun or other servers?
    • Given that it has fewer CPU's, it is probably cheaper than some of the better-performing ones tested.
      Have you seen the price of Sun computers? To get anything decent you need about $10,000, and those weren't the kind of Sun computers in this competition. Sun's cheapest 2 processor machine (that I could find on their crazy site) is $17,995. It lacks DDR RAM, Gigabit ethernet, and has a max of two hard drives. Of course I would like to see how this setup compares to the Xserve in other tests though. My point is that Sun is killing itself with really grotesque prices -their great OS now has a lot of competition in the lower end -Linux and now Xserve, and that competition is soooo much cheaper. It's a good thing for them that Microsoft's OS sucks so badly, or I imagine corporate types would've dumped Sun to go to X86 hardware a while ago. Luckily Microsoft seems incapable of putting out a decent OS so the server market still has innovation and competition. I'm willing to give the Xserve a chance, there's definitely room for competition here.
  • So the benchmark only tested them on Photoshop, which is a AltVec optimized applicaiton from the beginning? That and where is the rest of the benchmark data? Doesn't seem to be a real fair comparison.
    • Re:Photoshop Opens (Score:4, Informative)

      by Lars T. ( 470328 ) <.moc.liamelgoog. .ta. .regearT.sraL.> on Monday July 01, 2002 @04:49PM (#3803456) Journal
      No, they used Photoshop to open and save large images from/to a network server. Can't you read?
    • by Saint Fnordius ( 456567 ) on Tuesday July 02, 2002 @06:41AM (#3806470) Homepage Journal
      This wasn't a test of all-around server tastiness, but of the area Xinet is most interested in- publishing. That means pulling files down local with Photoshop, collecting for a print job (whether QuarkXPress or InDesign), opening the darn EPS file straight from The Server because the Boss doesn't want you saving it on the local drive, and so on. It's the client that needs the AltiVec optimisation; the server just needs to "shovel" the files here and there.

      Xinet needs to know where its software will be best used, so that they can plan accordingly. Other 'benchmarks' aren't interesting to them.

      Y'see, until now, WinNT box sellers were trying to muscle into the publishing server market, extolling their rack-mountability and cross-platform compatability, and Linux box manufacturers weren't that far behind. You could say that Apple's xServe is going to win back those shops first, then go for the mixed-OS networks, securing the flanks before launching the main offensive into Serverland...
  • nBLAST performance (Score:1, Informative)

    by Sleepyguy ( 12339 )
    Let me start by saying that I'm an apple fan, I own two recent macintoshes (a g3/400mhz pismo laptop and a g4/550mhz desktop). I like apple, i like osx... I'm crazy in this way... that being said...

    I've been itching to get my hands on one of these to test with, but since Apple couldn't get me one I went ahead and had them send over a dual 1ghz g4 tower to test nBLAST with. My feeling being since most of nBLAST does is cpu dependent the distinction is minor.

    I went ahead and set up a number of machines, but I'm only going to talk about two.

    a 1u dual 2.4ghz intel box with 1.5gigs of ram (custom)
    a dual 1ghz g4 tower with 1.5gigs of ram

    I created a number of standard BLAST queries and ran them against both machines multiple times and compared the results. To summarize the dual 2.4 edged out the mac for queries going to nBLAST (which apple optimized) and trounced the mac for other queries (non-optimized). This is still impressive ... i mean the executives look and say "gee this machines is only using two 1ghz processors, and its almost as fast as this one using 2 2.4ghz processors". I mean as far as a breakdown by mzh apple did great almost twice as many requests per mhz. however when the price/performance breakdown was computed apple came out with a price problem, it cost almost twice as much...

    apple, when will you learn, you might be able to charge more for those fancy cases on the desktop, especially during a boom, but during a bust, noone cares what their rackmounts look like.

    _
    • The G4 tower does not have DDR main RAM. Unless your data is fitting in 2 MB, the path to memory is operating half as fast as the Xserve.
    • Are you sure you were using the Apple/Genentech optimized BLAST? (from apple's site?) It sounds like you were not.
    • I mean as far as a breakdown by mzh apple did great almost twice as many requests per mhz. however when the price/performance breakdown was computed apple came out with a price problem, it cost almost twice as much...
      Hmm... my experience at work has been that once a scientist has finished decking out a dual Athlon whitebox with decent components, it costs about the same our 2 x 1Ghz G4 price of $2500. Of course, as an IT department it then costs us nothing, because we won't support it unless they buy an off the rack Dell or Mac. Users get to support it themselves out their grant money.
  • I hate reading that such-and-such has released statistics showing that xyz is faster than abc, when I know next to nothing about such-and-such.

    So wondering what I could find out quickly:

    According to their homepage:

    "...Xinet is the leading developer of prepress networking software. ..."
    (http://www.xinet.com/ [xinet.com])

    and if you go looking for it - '2002 testing details' (deeplink: http://www.xinet.com/benchmarks/benchmarks.2002/bm . etails.html [xinet.com]):

    "...Xinet's Benchmarked Configurations measure network throughput by Macintosh clients opening and saving Photoshop files stored on the server..."

    Given who they are (a company who makes software for an Industry Apple has a significant presence in) and how they test (using Macintosh clients) how they obtained their benchmark results (and why photoshop gets in to them) is a little easier to understand.

    (I have to wonder if this is why Apple Keynotes have photoshop demos to show off their hardware - they're focusing on /selling to one of their key markets...)

  • thoughts (Score:2, Insightful)

    1. They used an old dual-750Mhz Fire 280R when Sun doesn't even have base configs for 280Rs with that CPU anymore. How bout using up-to-date machines?

    2. Strangely, they chose to use third-party Gig-E cards rather than Sun's own, quite good Gig-E hardware. This alone could be enough to ruin the validity of the benchmarks. This is probably because they chose to focus on Gig-E over copper, a strange choice in and of itself.

    3. The one PC platform box was a dual PIII 1.4Ghz. Not exactly the performance leader in dual CPU PC servers.

    4. The benchmarks were all runnning one server app, Xinet's own fullpress.

    All these benchmarks show, is that one app, from one developer ran faster on the Apple Xserve than on some selected, out-of-date, hardware from other vendors. To all appearances, Xinit was not doing a platform cometition so much as a random benchmark with hardware that they happened to have one hand. The number of variables that could invalidate the results entirely is just silly.
    • What's so weird about copper GigE? It works just fine for me.
      • I suspect the reason they used third party Gigabit cards is because Sun wouldn't sell them a GigE/copper card. As recently as three weeks ago my Sun rep told me they did not make such an animal. Turns out they had recently introduced them. As for the whole concept of GigE/copper, that is what the target customers run.
    • Re:thoughts (Score:2, Informative)

      by pyrotic ( 169450 )
      The one PC platform box was a dual PIII 1.4Ghz. Not exactly the performance leader in dual CPU PC servers.

      Actually, for 1U boxes, it is. See IBM [ibm.com] for example. Their top speed is 1.2mhz PIII. Fast Pentium IVs aren't out for servers, unless you count Xenons, and those things need a good 4U if you don't want to toast your rack.
    • Re:thoughts (Score:4, Interesting)

      by dhovis ( 303725 ) on Tuesday July 02, 2002 @08:14AM (#3806776)
      3. The one PC platform box was a dual PIII 1.4Ghz. Not exactly the performance leader in dual CPU PC servers.

      Actually, when it comes to 1U servers, it is. Go check Dell's website. The Poweredge 1650 is Dell's fastest 1U machine and it offers 1 or 2 1.4GHz Pentium IIIs. PIIIs are still commonly used in servers, because the P4 is a power hog and doesn't actually give you a speed boost. Remember that the PIII is actually about 50% faster than the P4 at the same clock speed. Intel's server optimized processor is the Xeon, and that is just a Pentium III with boatloads of cache. You can't find them in 1U servers, either.

      4. The benchmarks were all runnning one server app, Xinet's own fullpress.

      What exactly do you expect Xinet to use for benchmarking? They need to have an answer when people ask "what hardware runs your software best?" Read the article, it does not go around drooling over the XServe. It says how big of a shop each server is good for. It specifically says that the XServe is a good choice for small to medium size graphic shops. Nothing more. It is of significantly more interest how much the 2x1GHz XServe outperforms the 2x1GHz Powermac G4. To quote from Xinet's site:

      Why Publish These Tests?

      Xinet believes it is important to publish current, accurate, detailed information concerning the performance of our software on the server platforms we support.

      By publishing this data, Xinet hopes to assist customers in choosing the systems that are appropriate for their needs, and to ensure smooth integration of Xinet products into their workflows. Xinet's Benchmarked Configurations are the result of extensive testing in the dedicated Xinet lab. Xinet hopes customers will use this data as a guide to the performance they can expect from various server-based workflows.

      • Actually, when it comes to 1U servers, it is.

        What exactly does this have to do with anything? Only two of the systems being benchmarked in this study are 1U servers. I see no reason why the only PC server represented needs to be a 1U system, particularly since this means using slow CPUs.

        What exactly do you expect Xinet to use for benchmarking?

        Well, for the kind of information they want to give their customers, I'd expect exactly what they did. For meaningful benchmarks to give a real sense of system capabilities, the answers are quite different, and I don't feel like going into it. I think Xinit had good intentions, and is not not responsible for other people trying to extract more meaning than is present in their data.
        • because Xserve *is* a 1U server. no point in comparing it to anything but 1U servers.

          Additionally, a 1U server fits very specific business needs which is what Xinet is catering to.

  • by Anonymous Coward on Monday July 01, 2002 @10:32PM (#3805392)
    We use xinet software on SGIs and Solaris machines long before they ported full-press to OS X. We have an Irix box that is 4 years old running xinet. For a long time, to get better Appleshare than Apple was to run on a unix machine running Full-Press.

    So this is not a matter of Xinet writing software optimized on OS X. That is not true. Full-Press does not rely on the built in OS's capabilities for AFP but its own. OSXs version is about less than 2 years old. In fact, I believe most of FullPress implementations run on SGI.

    The benchmarks are quite impressive because you are dealing heavy I/O (pushing 600 megs and 1 gig files) are normal in this business.
  • people, I just feal the need to point out one thing,

    IT'S A GOD DAMMED MOTHER FUCKING BENTCHMARK!!!!!!

    That means its like politics, unless its something good for the side your on it doesn't mean anything.

    as far as I can tell this bentchmark says that a Mac is good at Photoshop.

    what a shock. get over it. move on.
    • by Anonymous Coward
      as far as I can tell this bentchmark says that a Mac is good at Photoshop.


      As far as I can tell you didn't bother to read the article. The machines were serving photoshop files to the clients, not running photoshop.


      What this tells us is that the Xserve is great for small and medium prepress and design shops. And that's all it claimed to tell us.


      As an admin for a medium sized ad agency, these benchmarks are telling me exactly what I want to know.

    • as far as I can tell this bentchmark says that a Mac is good at Photoshop.

      Apparently you can't tell very far. Photoshop was not involved.

  • You call these benchmarks? Who cares about file sharing & serving? And batch print jobs? Why can't these guys get on the planet & use the industry standard?!

    Quake 3 frame rates!
  • The accepted standard for CPU benchmarks are the SPEC benchmarks [spec.org]. They aren't perfect either (no benchmark ever is), but they are well understood, reproducible, and published.

    Apple is notably absent from SPEC's list--they never submitted results.

    However, third parties have run the SPEC benchmarks [heise.de]. A 1GHz G4 seems to perform about as well as a 1GHz Pentium III: decent but not overwhelming. See also this Register article [theregus.com].

    Apple should move to the G5 quickly. Or, perhaps, Apple should even switch to some 64bit Intel or AMD processor--Motorola is likely going to keep remaining behind the curve a bit.

    • And testing a server with the SPEC CPU benchmark is a good idea why exactly? Esp. when SPEC has several server benchmarks?

      As for the Heise tests: they also show that a 800MHz G4 is faster than a 1GHz P3. Which brings us to the issue of compilers...

      • And testing a server with the SPEC CPU benchmark is a good idea why exactly? Esp. when SPEC has several server benchmarks?

        Because Apple makes big claims for their CPU peformance, because that's what many people are using servers for. But if you like, fine, run SPEC's server benchmark. Running PhotoShop, however, is bogus, and it definitely isn't anything one would use OSX server for.

        As for the Heise tests: they also show that a 800MHz G4 is faster than a 1GHz P3. Which brings us to the issue of compilers...

        So what? That shows that you can make the Intel chip perform bad with a bad compiler. For PPC, they pretty much used the best compiler there is.

        • For the last time (or rather, for the people without a brain) they did not test Photoshop on the Xserve.

          As for the compiler, it may be bad, but it's the compiler almost all Windows software is compiled with. And where did you get the information that gcc 2.95 is "pretty much the best compiler there is" for PPC?

          • For the last time (or rather, for the people without a brain) they did not test Photoshop on the Xserve.

            Come on, you are quibbling. What matters is that they failed to use standard benchmarks, not what particular oddball choices they made for comparisons.

            As for the compiler, it may be bad, but it's the compiler almost all Windows software is compiled with. And where did you get the information that gcc 2.95 is "pretty much the best compiler there is" for PPC?

            What's your point? Even under the most charitable assumptions, the G4 is in the ballpark of a PIII with similar clock frequencies.

            I hope Apple has something up their sleeve in terms of performance. I and others are willing to put up with a bit of performance lag relative to Intel and AMD, but at some point, it is going to start hurting them. Where are those G5's?

            • You don't get it, do you? This was a test by Xinet, testing their own product FullPress on several servers, including the Xserve, to give their customers a hint how those servers would perform. How would any "standard" benchmark be better than something that is nearly identical to what you will be doing with your machine?

              Even under the most charitable assumptions, the G4 is in the ballpark of a PIII with similar clock frequencies.

              For varying sizes of ballparks. On average. Unless you use AltiVec.

              • You don't get it, do you? This was a test by Xinet, testing their own product FullPress on several servers

                It doesn't matter what Xinet's original intentions were, what matters is how this gets portrayed in the press. I'm simply pointing out that announcements like "Xserve Outperforms Sun, SGI, Windows" are misleading.

                Apple and the Mac press have a history of greatly overstating the performance of their systems. This is just more of the same.

  • Apple Ships First Xserve Rack-Mount Servers to Customers [apple.com]

    Includes some benchmark results:

    Apache Web Server--Xserve can support 60 percent more connections on an Apache Web Server than an IBM eServer x330. Under industry standard WebBench* performance benchmarks, an Xserve running Apache on Mac® OS X Server can support 4,051 web connections per second compared to 2,547 connections per second on an IBM eServer x330 running Apache on Linux. Xserve provides an affordable and robust server platform for even the most industrial strength web applications.
  • Good
  • by axehat ( 590580 )
    It seems to me that if Apple spreads the word about how well their Xserve performs, possibly doing more benchmark tests, they could take control of the small server market. Heck, they might even branch out to larger servers. Its Apple, anything could happen.

There is no opinion so absurd that some philosopher will not express it. -- Marcus Tullius Cicero, "Ad familiares"

Working...