On CBS.com: Watch Exclusive CSI Cross Over Video
BNET Business Network:
BNET
TechRepublic
ZDNet

August 16th, 2007

Analysis of Sun's 'Niagara 2' UltraSPARC T2

Posted by George Ou @ 6:28 am

Categories: Energy efficiency - green, Hardware, Intel, Networking, News, Processors, Servers

Tags: T2, Performance, Sun Microsystems Inc., Sun UltraSPARC, Server, CPU, Analysis, Intel Corp., George Ou

Sun Microsystems launched their brand new “Niagara 2″ UltraSPARC T2 8-core CPU last Tuesday August 7th 2007 as the successor to the UltraSPARC T1 8-core CPU launched in November of 2005.  During the launch Sun’s executive VP David Yen took a few shots at big iron competitors IBM and Intel by noting that they didn’t try to crank the clock speed to 4.7 GHz (referring to IBM Power6) or try to shove 24 MBs of Cache (referring to Intel Itanium 2) though this was a bit strange since the UltraSPARC T2 isn’t really aimed at the big iron market at all.  Of course Sun’s launch page does take a shot at Intel’s x86/x64 quad-core CPUs by saying that they didn’t resort to “packaging gimmicks, such as MultiChip Modules (MCMs)”.  At the end of the launch event during questions and answers, one of the event highlights was when CEO Jonathan Schwartz came to the defense of Linux in response to someone saying that Linux has a “checkered past” with scalability by saying that Solaris has a “checkered past” with usability.

Market for UltraSPARC T2
The UltraSPARC T2 processor is designed to go in to a single CPU server from Sun Microsystems and it should not be compared to Itanium2 or Power6 that are designed to scale to 32 or 64 CPU “big iron” servers.  Sun’s Niagara series chips are designed to consolidate up to 64 slower legacy SPARC-based servers or host up to 64 lightly-loaded logical servers using Solaris 10 containers or LDoms (Logical Domains).  But because container and LDOM technology only supports Solaris 10, legacy servers that are likely running older versions of Solaris will need to be validated to run on Solaris 10.  While that validation process of migrating legacy servers should usually go smoothly, the cost and time required is non-zero and it isn’t as simple as Virtualization technology such as VMware or Microsoft Virtual Server which allows you to basically import legacy machines as is without changing the OS version or OS vendor.  But once the migration is complete, there can be massive annual savings just in the hardware maintenance cost of the old legacy SPARC servers, rack space, and power consumption.

UltraSPARC T2 Architecture
The UltraSPARC T2 is an 8-core CPU built on a 65nm process and is based on a single 342mm squared die.  That’s relatively small considering the fact that eight Intel Core 2 65nm cores (two processors) would measure 572mm squared and eight 65nm Barcelona cores (two processors) would measure 566mm squared.  While that might sound like it makes the UltraSPARC T2 cheap to make, reality hits when you realize that 342mm squared is a huge single die.  Intel’s 65nm dual-core dies are only 143mm squared and AMD’s upcoming 65nm Barcelona die is 283mm squared which is already big.  Any tiny flaw in that massive 342mm squared die means the whole chip is bad or you have to give up two or four of the cores though it’s not known if Sun will offer 4- or 6-core versions of the T2 like they did for the T1.

The UltraSPARC T2 has 8 1.4 GHz cores with 2 pipelines per core and 4 threads per pipeline giving it a total of 64 threads that it can support.  The T2 CPU has eight crypto processors designed to offload symmetric and asymmetric encryption as well as hashing functions.  The UltraSPARC T2 also has eight fully pipelined floating point units making it substantially better than its T1 predecessor.  The T2 die also houses four memory controllers that can support up to 512 GBs of fully buffered DIMMs with an aggregate memory bandwidth of 64 gbps.  Even a 10 gigabit Ethernet controller and a PCI-Express controller are included in the UltraSPARC T2 processor.

Note: Surprisingly, Sun actually shortchanged their own T2 processor at the launch event by saying that it would shave $190 from the cost of the motherboard because you don’t need a 10 gigabit Ethernet controller and crypto offload engine.  $190 might be the cost to the motherboard manufacture to integrate those components but the end user will see a multiple of that cost when they buy the actual server.  However, even 8-core UltraSPARC T1 servers have been known to be twice as expensive ($30K from Sun and $15K from HP) as an 8-core x86/x64 servers from Intel so the true savings may be dubious.  Given the steep pricing on the UltraSPARC T1, I don’t expect the T2 to be cheaper.

<Next page - UltraSPARC T2 performance and power consumption>

Pages: 1 2

George Ou is Technical Director of ZDNet. See his full profile and disclosure of his industry affiliations.

  • Talkback
  • Most Recent of 99 Talkback(s)
If that is consoling to you, then believe what you will
"You're not defending your article, you're defending Intel's competitive position! Fanboy."

If that is consoling to you, then believe what you will. I've given you all the proof to show how In... (Read the rest)
Posted by: georgeou Posted on: 08/27/07 You are currently: a Guest | | Terms of Use
WTF  Erik Engbrecht | 08/16/07
What you expect is wrong  georgeou | 08/16/07
Dependent on Assumptions  Erik Engbrecht | 08/16/07
Here's the problem. T2 has moderate aggregate performance but lousy single  georgeou | 08/16/07
Was that your purpose, to show off Clovertown?  Uber Dweeb | 08/16/07
Sun made the comparison, I answered the call  georgeou | 08/16/07
wtf  richvball44 | 08/16/07
You have the right to your opinion, but you can't dispute the facts  georgeou | 08/16/07
Are these facts...  zkiwi | 08/16/07
They're SPEC.org facts  georgeou | 08/16/07
Well...  zkiwi | 08/16/07
You're saying you don't like the facts, I got you  georgeou | 08/17/07
No George  zkiwi | 08/17/07
Expect the wrong thing  Bob.Kerns | 08/18/07
Just when I thought George was improving...  Uber Dweeb | 08/16/07
what he said...  flounder141 | 08/16/07
Sun made the comparison to x86/x64 server world  georgeou | 08/16/07
What the hell are you talking about?  Uber Dweeb | 08/16/07
Ok, you tell me the advantages  georgeou | 08/16/07
But Mr Ou, he just did -  mhenriday | 08/17/07
Yes, and that's a worthless advantage  georgeou | 08/17/07
Only because you say so?  CFKane | 08/18/07
I hear you talk, but you never provide data  georgeou | 08/18/07
That's your problem  CFKane | 08/18/07
Poor Sun gets George's fanboy treatment  CFKane | 08/16/07
When a T1-based server costs $30K and x64 server costs half  georgeou | 08/16/07
Threads, George, not cores.  Uber Dweeb | 08/16/07
You can get 64 threads working on x64 architecture just fine with 8 cores  georgeou | 08/16/07
Why is this so difficult?  Uber Dweeb | 08/16/07
it's so difficult because...  zkiwi | 08/16/07
The question is....  jragosta | 08/17/07
The overhead on time slicing is very efficient  georgeou | 08/16/07
You're kidding right?  t_mohajir | 08/16/07
George, you don't know what you're talking about.  Uber Dweeb | 08/16/07
Don't quote out of context  georgeou | 08/16/07
No running system has 64 threads ...  s_souche | 08/17/07
Learn, George  jragosta | 08/17/07
That's funny that you think you need one physical thread  georgeou | 08/17/07
How about backing up that claim with some numbers?  CFKane | 08/16/07
That's calling for 8 threads per core, and I'll be happy to test.  georgeou | 08/16/07
Ok, did quick test with WPrime. Also a link to someone testing 128 threads  georgeou | 08/16/07
You have to run WPrime with 64 Threads  Burana | 08/16/07
No, it's 8 threads per core  georgeou | 08/17/07
Nothing surprising there  CFKane | 08/17/07
Great, you ignore the Apache web server test  georgeou | 08/17/07
Was that meant for me?  CFKane | 08/17/07
Over the last several...  bjbrock | 08/16/07
Tell that to LostViking who accused me of being an Apple shill  georgeou | 08/16/07
Are you sure...  zkiwi | 08/16/07
He wasn't the only one complaining about my praise of the new iMac form fac  georgeou | 08/16/07
Tada!  zkiwi | 08/17/07
George is wrong  rweiler@... | 08/16/07
WHAT! Who's the vaporware?  georgeou | 08/16/07
Geore, please re-read your article  rweiler@... | 08/17/07
L5335 is Clovertown, not Penryn  georgeou | 08/17/07
L5335 has same status as Niagara 2  rweiler@... | 08/17/07
L5335 is same as E5335, only lower power and better architecture  georgeou | 08/17/07
wow a lot of inflamed readers...once again...surprise!  jjarman | 08/17/07
Nothing inflamitory about this article  georgeou | 08/17/07
Wrong  CFKane | 08/17/07
I havn't been unfair in my critism, nor do I have a chip on my sholders...  jjarman | 08/17/07
You are not being honest  georgeou | 08/17/07
And yet...  zkiwi | 08/17/07
Wrong, who is being dishonest now? *link to thread*  jjarman | 08/17/07
Latest ZDNet reporting trend inspired by Simon from American Idol?  jjarman | 08/17/07
This is the kind of professional trolling I'm talking about  georgeou | 08/17/07
Wow, two thread make me a stocker? You think pretty high of yourself!!!  jjarman | 08/17/07
Single thread NOT big deal  Bob.Kerns | 08/18/07
And you think an Intel or AMD x64 processor can't run 64 or 512 threads?  georgeou | 08/18/07
Not at the same time  Bob.Kerns | 08/18/07
Tell that to VMware and Xen  georgeou | 08/18/07
What???  Bob.Kerns | 08/18/07
To be clear on this  CFKane | 08/18/07
You make a good case...  Bob.Kerns | 08/18/07
Looking at 32-way Unisys box  THEE WOLF | 08/18/07
An analysis of the "analysis"  CFKane | 08/18/07
The L5335 is faster and you admit it.  georgeou | 08/19/07
Clutching at straws now?  zkiwi | 08/19/07
Back at changing the rules again  CFKane | 08/20/07
Last time I checked, a CPU doesn't constitute a computer.  georgeou | 08/20/07
Right  CFKane | 08/20/07
You forgot the cost factor.  georgeou | 08/20/07
You couldn't have been much clearer  CFKane | 08/21/07
If that is consoling to you, then believe what you will  georgeou | 08/27/07
Speaking of Java performance and SPECjbb  georgeou | 08/19/07
Forgot to include link to SPECjbb results  georgeou | 08/19/07
You don't get it  CFKane | 08/20/07
we're going nowhere  s_souche | 08/20/07
I'm not headed there  CFKane | 08/20/07
Nope, I said multi-thread and single-thread is important  georgeou | 08/20/07
It's always the same: You're shifting the rules as you see fit  CFKane | 08/20/07
The Dual Intel X5365 is the confirmed SPECjbb champ, T2 is a no show.  georgeou | 08/20/07
So that's your analysis  CFKane | 08/20/07
You're taking the classic weasel tactic  georgeou | 08/20/07
It's futile  CFKane | 08/20/07
Last time, it's Sun's job to provide proof  georgeou | 08/20/07
Sun's job?  jragosta | 08/21/07
Don't just make accusations without backing it up  georgeou | 08/21/07
Everything is made up  zkiwi | 08/22/07

What do you think?

SponsoredWhite Papers, Webcasts, and Downloads

Click Here
advertisement

Recent Entries

Top Rated

    advertisement

    Archives

    ZDNet Blogs

    White Papers, Webcasts, and Downloads

    Meet Doc