An NEC server with 32 of Intel’s Itanium 2 6M processors and running the Windows Server 2003 takes the top spot in a widely watched performance measurement by displacing a Unix server. […] NEC is top of the heap right now, but HP has aggressive plans for Itanium. It will release 64-processor Superdome systems initially, then through a technology code-named Hondo that plugs two Itanium processors into a single socket, will sell 128-processor Itanium systems. HP supports three operating systems on its Itanium systems today–Windows, Linux and HP-UX Unix–and will add support for OpenVMS next year.
NEC Windows Server Wins Speed Test
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
This event it no big deal. Sun doesn’t like this particular benchmark, and it looks like only SMP systems are considered (so a few racks full of dual Athlons in a cluster won’t count).
It seems from reading the article and it’s related articles that this benchmark *is* a big deal.
The current version of the NT kernel (Win2003) is already 128-way capable right now. As soon as these massive new HP boxes get built, Windows can be running on them.
Sun’s excuse for not posting their benchmarks like every other vendor is just a lame cop-out. Oracle pulled the same crap when SQL Server started kicking them out of all of the top database benchmark rankings.
RE: Kobold – MSFT now owns the top spots for both clustered and non-clustered systems.
Gilly, do you imply that a cluster twice the size of the MSFT one couldn’t be running under Linux? Intel and MS may be building new shiny big toys to market Windows 2003, but production clusters are still staying away from MS for a good reason.
And, as the article mentions, MSFT system would soon be replaced by a multi-OS one from HP. And you can be sure that no one would be running Win2003 on that box.
You say no-one uses Windows clusters or multi-proc systems?
Ever heard of Exchange? It runs the email for probably 70% of the Fortune 2000. Ever wonder what kind of hardware these Exchange deployments run on?
So a specially tuned version of the non-shipping Windows 2003 beats a machine that has been on the charts since August 2001 using a chip that itself is not even formally shipping yet?
And let us notice that this new machine, using software and hardware that will be essentially TWO YEARS newer than the old system, barely beats the old system.
Seems like a hollow victory to me, especially considering Windows Server 2003 won’t be fully debugged for at least another year.
Some in the UNIX crowd might look at it as a defeat on the part of Microsoft. Let us see in two years if the top UNIX machine can’t defeat this NEC system, okay?
Who would want to utilize Windows 2003 and why? What compelling features does it offer to the enterprise besides being able to run on a massive Itanium system?
With Microsoft’s zero-responsibility anti-customer attitude and folding in of all key UCITA provisions into their licenses, running an enterprise on Microsoft doesn’t seem like good business sense to me.
maybe you should have a look at the top 500 list (http://www.top500.org).
If you’re going to compare, do it with similar hardware and operating systems.
regards
Alex
[quote]And let us notice that this new machine, using software and hardware that will be essentially TWO YEARS newer than the old system, barely beats the old system. [/quote]
Barely, huh? Regarding that the old system has 4 times as much CPUs…
EOM.
Let’s at least get the facts straight. There were 64 processors (unreleased processors) on an unreleased version of the OS with an unreleased version of the database. Sure you will be able to buy it within a few months, but not now.
This does seem to prove Microsoft can make the numbers the Enterprise demands, but I would doubt they will hold the crown for long.
[i]RE: Kobold – MSFT now owns the top spots for both clustered and non-clustered systems. /i>
So what? A fast piece of crap is still a piece of crap!
Check http://www.tpc.org/tpcc/results/tpcc_perf_results.asp now. An HP Superdome running 64 Itanium2 procs (not the new 6M ‘Madison’ NEC is using) has surpassed this NEC system by more than 20 percent tpm’s.
Now MSFT holds the top 4 rankings in non-clustered TPC-C. Probably won’t be long before they get all 10.
AllMighty – you might want to look at those results yourself. The executive summary disclosure report states right up front that HP used 1.5GH Itanium2 cpus with 6M of cache, just like NEC did.
Given that they used twice as many cpus and got less than a 30% performance increase, something is messed up. I suspect it is MS Windows as the superdome cell-based architecture has pretty good scaling at the hardware level. Although you never know, HP may have hobbled the Itanic implementation of the superdome cells with poor memory-bandwidth or something silly like that.
That’s a great achievement from MS, I’m very happy they managed to push their system to that point. I wish they’ll keep up the good work, so that other OS builders will try to catch up.
Now, raw power is an interesting benchmark, but what about robustness ? Is there a rating or benchmark to evaluate a system capacity to sustain the workload as long as possible ?
Microsoft probably cheated on all the benchmarks anyway. Because of Microsoft’s no-benchmarking clauses in their software licenses, nobody can independently test Microsoft’s scores and report the results to the public. So the test scores are meaningless.
Any vendor that refuses to allow third-party benchmarking while using benchmarks to tout their systems superiority is stacking the deck.
Put up or shut up – that’s the bottom line.