c0t0d0s0.org

The first news about the upcoming SPARC64 VII appear at the usual suspects: e.g. cnet. There is another report at TheRegister but Ashlee Vance seems to be a little bit bitchy, as Sun canceled his call

Every once in a while a blogger, sales rep of a competitor or a misinformed journalist write "The UltraSPARC T1/T2 cores are just UltraSPARC II cores running at 300 MHz". Well, this is a rumour consisting out of two misunderstood points of the architecture. Well informed people already know the stuff in this article but this time your are not the target group of readers. But i think it´s time to write a little bit about all this rumours as i write or tell this answer again and again and again.

As usual Paul brings an interesting perspective into discussion:

On the other hand.. the way the processors are coupled - done by replacing the the T2’s on board 10Gbyte facility - demonstrated that Sun can now produce highly customized versions of the core CPU set and suggests what I believe may be a unique performance opportunity for this product line.

Taking into consideration the additional transistor budget by the usage of upcoming process technologies, it should be feasible to integrate other interconnect technologies as well, for example Infiniband on die (just think about the latency advantages of an Infiniband port directly conneted to the crossbar) or the integration of additional support circuits for special tasks. I assume, there is only one limit ... the pin count of the proc ... at a certain point you can´t get all the interfaces out of the chip in an economical feasible manner.

Paul Venezia of Infoworld wrote a really positive review of the UltraSPARC T2 based Sun SPARC Enterprise T5120 - "Lab test: Sun’s octo-core SPARC is made to multitask". He even concludes at the end, that reports of SPARCs death are greatly exaggerated in the light of this system:

Overall, the UltraSPARC T2 and the T5120 build upon the hallmarks of the first-generation UltraSPARC T1-based servers, and remind us that although the SPARC CPU may have been marginalized in recent years, it hasn’t surrendered, and may in fact be making a comeback.

Do you remember my article about a presentation held by Rick Hetherington about future directions in SPARC with many confidential informations? One of the most interesting informations was the new foundry for the 45nm procs, as TI itself won´t invest in this manufacturing technology soon.

Now there is an official comment about this. EETimes writes in Sun taps TSMC for 45nm CPUs:

Sun Microsystems Inc. has chosen Taiwan Semiconductor Manufacturing Co. to make its multi-core processors at 45nm and finer geometries.[...]"We have had engineering teams working for months now on 45nm designs and we will have multiple 45nm products," Azhari said.

Yesterday was the flagday and putback for 6521412 PSARC 2006/525: new boot sparc. This is an important development as this is an important prerequsite for booting from ZFS.

An interesting article about TI´s future manufacturing strategy: TI tips foundry strategy; UMC wins Sparc business

This article by Ashlee Vance sheds light on something, that kept my mind busy the last month: How to explain the dense pack of IV+ 2.1 GHz, OPL and Rock. To make one thing clear, the conclusions are mine, not the one of Sun.

While reading Scientias good article about AMD Q4 earnings, i thought a little bit about the situation in the x86 market:

AMD touts the quad core K8L, Intel speaks about the 45 nm Core Duo, both tells the world that they are faster than the other. And both is correct for certain timeframes. Sounds familar. Yeah, it´s leapfrogging. When two architectures are rather similar (UltraSPARC and Power, AMD x64 or Intel x64) then the companies starts to play technical leapfrogging. At one point of time you lead the competition, at an different point of time it´s the competition.

I believe, that the press still isn´t getting the point behind the "Solaris Binary compatibilty guarantee". There are no problems for the user by having four paths of Sparc in the coming years: Sun guarantees that you can run the same binary on UltraSPARC T1 (and Niagara II of course), UltraSPARC IV+, APL or later on Rock. You can use the SPARC that fits best to your workload. No recompile. Simple moving it to the new platform. The binary compatibility guarantee was introduced for this very reason. To protect the investments of the customers . And to enable the customer to choice the system that fulfills the job at best.

So, dear ComputerWorld:Where are the issues for our customers by having more than one choice? As "his Steveness" said: "It´s good to have choice"

ZDnet reports in Sun boots server with Niagara II chip:

The Niagara II system booted on May 26, spokesman Alex Plant said, about three months ahead of the late August or early September schedule

and

In addition, Niagara II has built-in 10-gigabit-per-second Ethernet networking, multiple floating-point units to speed mathematical calculations, and new encryption and decryption abilities.

Paul Murphy speaks it out: Apple and Sun would be a perfect match. Imagine the best operating on earth with the best usability on earth. Imagine a SMP-out-of-die capable Niagara X with a FPU on every core, with an Altivec-Unit on every core, with 2.5 GHz. Imagine Solaris with the Desktop of MacOS X. Or MacOS X on a 880z with this ultrafat-badass-sonofthehell graphic board in a processor slot. The world could be such a awesome place.

Am Ende kommen sie doch wieder zu uns, wenn es wirklich was zu tun gibt: Dell turns to Sun in battle against, er, Sun

Ich wollte schon immer einmal in einer chilenischen Silbermine arbeiten, dieses Gefuehl bekomme ich zumindestens, wenn ich die Einstufung von Moodys ueber unsere Firma höre: Moody's cut Sun Microsystems' debt ratings to junk. Für Steve Millunowich gibt es anscheinend einen Nachfolger was schwachsinnige Analysen angeht: Herrn Enderle. Ich frage mich, wie er auf diese Kommentare kommt. Das mit der Aufgabe der "poison pill" hört sich eher danach an, als würde sich ein Aktionär darueber aergern, das er keine von diesen extrem ueberhoehten Angeboten in Rahmen von feindlichen Uebernahmen bekommt.
Aus dem Niagara-Umfeld gibt es interessante Neuigkeiten: Anscheinend liegen wir vor dem Zeitplan(Bitte dann denken: Ich dokumentiere nur, ich bestätige garnichts). Damit steigt der Yield bei diesen Prozessoren erheblich. Ich muss halt nicht mehr die ganze CPU wegwerfen, wenn ein Core nicht funktioniert.
Der Register berichtet ueber die UltraSPARC IV+-Prozessoren, allerdings gilt es zu bedenken, das die Leistungs des Prozessors nicht notwendigerweise in der Applikation ankommt, und genau das ist der Grund warum Solaris auf Sparc immer noch so gut bei anderen Systemen bei echten Applikationen mithalten kann.
Einen weiteren oft vergessenen Vorteil der Sparcs wird von IT Jungle vorgebracht:

Those chips dissipated about 57 watts of power as the clock speed for the Cheetahs was cranked up to 1.3 GHz; moving to dual-core with the Jaguars jumped the heat up to 90 watts. But in moving to the Panthers, even with boosting the clock speed to 1.5 GHz and adding on-chip L2 cache and in-package L3 cache, Sun has been able to keep the Panthers in the 90-watt thermal envelope. This is a significant feat, especially considering that Sun's competitors are cranking out a lot more heat in the current designs.

Weniger Wärmeabgabe gleich weniger Verbrauch mal zwei. Denn der Chip braucht Strom um sich aufzuheizen und ungefähr das gleiche noch mal um es wieder loszuwerden, naja so Pi mal Daumen. Und Strom kostet. Und je mehr Wirbelstürme die Staaten heimsuchen werden, desto mehr.

Positiver Artikel in Serverwatch bezueglich Niagara und Solaris 10:

Far from doom and gloom in the RISC camp, Sun and IBM continue to fly the flag and have development roadmaps for their chips stretching years into the future. Both already have dual-core products on the market, and multicore is on the immediate horizon.
[..]
"Sun has been a bit mired down with little talk in terms of processor or system performance, although their new OS [Solaris 10] is a great technical achievement and seems be enjoying a rapid adoption rate by customers," says Olds. "Sun's Niagra project might change all this in the near future.
[..]
Sun stands to gain much from Niagra's success. "If this works as well as they say, it will get Sun back into the performance race in a big way," says Olds.

Achja, und auch zu diesem Artikel gilt wieder: Es ist nicht so das in der Auseinandersetzung zwischen RISC und CISC letzeres erfolgreich war, wie es die Presse gerne umschreibt. Die modernen CISC-Prozessoren sind nichts anderes wie RISC-Prozessoren mit extrem schnellen CISC-to-RISC-Konvertern. Aber diese Feinheit ist den meisten Journalisten wohl zu fein.