With Sun Microsystems and Oracle hogging all of the debate about the Sparc architecture these days, you can't blame Fujitsu for wanting to get a word in edgewise. So today, somewhere in Japan, Fujitsu reminded everyone that even though it's getting out of the chip manufacturing racket, it does have an eight-core Sparc64 chip in …
Bootnotes? Come on, Reg! You're slipping.
Apparently, all they found was Uranus.
> Apparently, all they found was Uranus.
Sparcs in Uranus ? Ouch...
Fujitsu Venus ..... Mars Godisagoddess
..... Shining Parallel Paths Providing Immaculate Pleasure Protocols ..... RAW Love Algorithm.*
"with SIMD extensions aimed at boosting performance for parallel supercomputing workloads. The extensions are known collectively as HPC-ACE, apparently, and what they do isn't clear."
And that's the Magic Bit of the Equation which Makes Everything Work Perfectly in an Imperfect World...... Bringing Stability for Deep Rooting Deep Seated Peace.
* Which you may notice is not shared as doubtful question.
Bad news for Itanium
Let's see, Fujitsu outsources chip fabrication, but continues SPARC64 design and has successful prototypes of 4.0 GHz, 8-core, scalable SPARC64 VIII chip. This seems like a direct competitor to IBM's 4.0 GHz, 8-core POWER7 chip.
What does this mean for Fujitsu's Itanium PRIMEQUEST computers?
With Oracle now owning Sun, and both Sun and Oracle being strong partners with Fujitsu, my bet is SPARC has another five strong years at the high-end, and Itanium is going to continue to shrink to being an HP-only chip. Unisys has bailed on Itanium, SGI has bailed on Itanium, Fujitsu may be next, leaving NEC as the only second tier Itanium vendor. Tukwila is a joke, just being a four-core Montvale, and who knows if Poulson will ever see the light of day.
Itanium is dead...
I wont buy a SPARC64 VII let alone a Venus because of Oracle Pricing
Remember the backlash of USIII to USIV because the performance per core went down?
Well here we are again with VII cores being 80% of the performance of a VI core.
Sorry, I will not buy quad core ships from Sun or Fujitsu that deliver less performance per core.
Oracle may be buying Sun, but their pricing model favors Power, Nehalem and Itanium over SPARC and I don't think Oracle wants to lower their prices to sell some Fujitsu OEM crap.
I didn't believe it at first but then I saw the Sun M-Values. M5000 with 16 cores (8 dual core chips) is 345,000. M5000 with 16 cores (4 quad core chips) is 255,000.
Ouch there is no way I am going to "upgrade" to quad core and get 20% less performance when my full Oracle software stack on this machine at 16 cores (12 licenses) is $1M just for the first year.
Make sure you get a full (non-confidential) report of all the Sun systems M-Values before you sign any purchase order. REFUSE to accept an NDA version, your rep wont care because he wants to make a sale and odds are will be looking for a job in July. Thats how I got mine.
Sun's Historical Role
People are very funny sometimes, when they think about who Sun really was from the beginning.
Sun was a systems manufacturer, from the beginning. They used off-the-shelf 68K CPU's, used 386 CPU's for a short period, made their own CPU's, used other vendors SPARC CPU's, resold x86 and x64 systems later, used Fujitsu SPARC CPU's again - but Sun was always a software company, programming to some type of specification, to stick on top of their boxes. Open Firmware, NFS, desktop applications, X, etc.
When Sun collaborated with the industry for the SPARC specification, they were not the manufacturer, but they teamed with TI. Other vendors made SPARC chips that Sun leveraged.
People talk about Oracle getting Sun out of the hardware business, or out of the manufacturing business, or out of the design business. Sun has traditionally been very comfortable with other companies doing CPU design & manufacturing. Sun has also been very comfortable with their own design houses and leveraging other design houses more recently.
I am sure that Venus will be another option on the joint Fujitsu & Sun SPARC lines if/when this CPU will be released.
No venus on Sun HW
Venus is a specific chip intended for Fujitsu HPC business and is not planned to be put on SUN OPL HW .
usual disclaimer ...
SPARC64 VII, M-Values, POWER, Oracle, and Single Benchmarks
Anonymous Coward posts, "I will not buy quad core ships from Sun or Fujitsu that deliver less performance per core"
Few vendors release multi-core CPU's where individual core speed is equal at the same clock rate or even higher clock rates. It will usually take 2 or 3 generations for a multiple core CPU to overtake fewer core CPU's in single threaded throughput. (i.e. Intel 5500, UltraSPARC IV+, etc.)
Anonymous Coward posts, "Oracle... pricing model favors Power, Nehalem and Itanium over SPARC"
POWER and Intel traditionally use multi-chip modules, which means pricing is not necessarily favorable while SPARC is a single chip module. Oracle pricing for the "standard" licensing normally charges per-socket on up to 4 sockets - but in the case of multi-chip module, each chip is considered a socket.
In the case of Enterprise licensing, the Intel Nehalem (5500) series processors get a scaling factor of 1 instead of 0.5 and the POWER6 gets bumped to 1. Increases of 25% and 100% in Oracle license fees for POWER and Intel while SPARC licenses remained the same - at least this was the case as of March 16th (before Oracle announced the purchase of Sun.)
Anonymous Coward posts, "I saw the Sun M-Values. M5000 with 16 cores (8 dual core chips) is 345,000. M5000 with 16 cores (4 quad core chips) is 255,000"
You can see from other comparisons, M-Values are not necessarily an accurate measure, you can see in this comparison where M-Value shows up with other benchmarks.
Any single benchmark is not a good indication of workload performance of real applications. For example, a single socket T2 processors will score equivalent performance anywhere from 1 to 4 Intel sockets, depending on the benchmark cited. Understanding your workload is just as important as understanding the CPU performance.
Sun has never been able to move to sw from hw
just look at Sun's financials. Sun is still stuck in hardware land. Their revenue for JAVA could not pay for 10% of the people they have working on it.
M-Values are Sun's most accurate guess at average workloads
Sun is forced to have M-Values because customers who are upgrading need to understand the capacity they need. You can see that the new quad core chips suck in the M-Values. As much as you can try to point to standard edition no one buys that limited thing. But I will guess from your statement that if you want enterprise edition you don't want to buy SPARC64 VII.
Face it SPARC64 VII at .75 per core so 12 licenses at $47K each for a 4 chip box is insanity.
I guess that is why Sun's "enterprise systems business" are -27% in fiscal 1Q, -32% in fiscal 2Q and -28% in fiscal 3Q.
Don't be an idiot.....Don't buy an M-Class!
M-Values are Sun's most accurate guess at average workloads
As far as Oracle Standard Edition - a lot of larger companies have Oracle applications deployed all over the place - and the Standard Edition works quite well.
If someone is going to spend $10,000 on a server, I would suspect they are going to look at more benchmarks than M-Value and have a clue about the workload specifics.
There are choices in the SPARC architecture, and depending on the workload: SPARC64 VI, VII; UltraSPARC T1, T2, T2+ -- and they have subtle differences which impact workload & licensing significantly.
- Vid Hubble 'scope snaps 200,000-ton chunky crumble conundrum
- Updated + vids WHOA: Get a load of Asteroid DX110 JUST MISSING planet EARTH
- 10 years of Facebook Inside Facebook's engineering labs: Hardware heaven, HP hell – PICTURES
- Very fabric of space-time RIPPED apart in latest Hubble pic
- Massive new AIRSHIP to enter commercial service at British dirigible base