Sparc Enterprise M server partners Oracle and Fujitsu have upgraded the entry-level Sparc Enterprise M3000 rack-based server with a faster Sparc64-VII+ processor. The two companies announced the quad-core Sparc64-VII+ processor last December. The chip is designed and fabbed by Fujitsu, which also designs and manufactures the …
as dead as the Itanium
Who in their right mind would still spend big bucks on SPARC when the writing is on the wall? Just about every SUN shop I know about is frantically moving everything off SPARC to x86_64. 10 years ago maybe SPARC was decent but speak to any recent developer using Solaris and they will tell you they dread having to deploy/troubleshoot on SPARC boxes (so slooooow).
re: as dead as the Itanium
Funny, asdf, you make Matt sound reasonable... You throw out unsubstantiated comments at every mention of SPARC. At least Matt acknowledged a personal animosity toward Sun Sales Reps. What's your problem? Panties a bit tight?
With the money that Oracle is throwing at SPARC, it sounds more and more likely that SPARC will have a comeback, starting with the T4. Fujitsu apparently has bought into it. Better than Intel and HP making their customers migrate once again to another architecture...
RE: re: as dead as the Itanium
Actually, I don't agree with asdf as I think this box will actually sell quite well - compared to the CMT offerings, that is. See, it's the only real option for all those Slowaris holdouts that haven't ported off SPARC yet. And it is getting the speed bump for the very same reason the M3000 was brought to market in the first place - CMT just doesn't work with the typical, single-thread-heavy applications the Slowaris base are using. Snoreacle are having to load the faster SPARC64s into it in an attempt to stop the migrations off all those old UltraSPANKed servers. Can't help laughing at the idea of consolidating two v440s onto one though, seeing as we can already do better (and cheaper) with Slowaris on Xeon, let alone the even better option of Linux! Larry is just praying there will still be some Slowaris base left by the time he gets a CMT chip out that can manage better single-threaded performance than an ARM core. Now, Slowaris on SMP ARM, that would be a better bet!
Also worrying to see Larry is fitting features into the Oracle software products to only work with Slowaris. I can't think of a better way to switch us customers off the products than to try and force SPARC on them.
fwd:RE: re: as dead as the Itanium
No defense of your beloved Itanic?
RE: fwd:RE: re: as dead as the Itanium
"No defense of your beloved Itanic?"
Oh, we're having that fun in this thread:
As I wrote in another thread, what do you say about my suspciion? Could it have some plausbiility?
"I begin to suspect this. What do you say about my suspicion?
Intel has never earlier built an enterprise cpu with RAS functionality. x86 is too buggy and bloated, says several experts:
So, Intel wanted to learn to build high RAS cpus, they worked with HP (which knows that). Intel learned RAS via Itanium, and incorporated RAS into x86. So know x86 has almost the same RAS as enterprise Itanium. So Intel were never really interested in Itanum, but only wanted to learn how to build RAS cpus. And now Intel has learned RAS from HP, so Intel is not really interested in Itanium anymore. Intel is betting on x86 cpus with RAS.
That is my suspiciion. What do you say?
(Too bad that x86 is too buggy and bloated to ever rival Itanum, no matter how much RAS Intel builds in. Only RAS will not give you the same stability as Itanium. The Itanium is much cleaner architecture and it is difficult to get stability with a shitty x86 instruciton set, no matter how much RAS you build in)"
The issue with Sparc is a decent roadmap. Sparc T has a roadmap and it's looking reasonable. It's also a technology people can believe in. It makes sense and you can see Oracle simply rolling out generation after generation and developing it into the future. The M class servers are another thing though. The roadmap is poor. The future is pretty unclear. It simply looks like a chip going nowhere. Everyone ran away from the Venus chip when in development. So, where's it going? Even if Fujitsu have bought into it, so what. The reality is that Oracle/Sun have no credable high end servers, especially for databases.
Oracle is relying on it's installed customer base and the time it will take them to move to something else to keep them on Sparc until they can come up with something sensible for the future. Unfrotunately, the future doesn't seem to be coming fast enough. Oracle has put the knife into Itanium, which is presumably part of a strategy to sun support Sparc. Now, I wouldn't say Sparc is in the same place as Itanium (which is basically dead), but it's mighty close. Unless Oracle can release something credible in the next 6 months, you'd have to be really brave to stay on that ship.
Long in the tooth...and also dead
"Sparc64-VII+ is looking a little long in the tooth, particularly with its DDR2 main memory. It is time for the M4 chip or whatever it will be called, but Oracle and Fujitsu are not expected to get the next Sparc64 chip out the door until 2012."
But we all know the T4 which is a CMT chip is the replacement for SPARC64 as there will be no more SPARC64 chips. So let's all admit it and move on. It really annoyed me that Oracle said Itanium is dead but wont admit to the facts behind their own (well maybe Fujitsu's) roadmap. Then went to great lengths to say they have renewed the relationship but devoid of any technology direction.
Now as far as the CMT chips, can Oracle make a decent T4 chip to compete with Westmere EX and Power7/7+ only time will tell.
People have their doubts because Oracle cannot undue in 2 years what Sun did for the prior 5, people have their doubts about every other architecture out there also.
If nothing else Oracle has made this market interesting to watch albeit painful to participate.
....and Matt B...and please no more peeing in public
There is no way that a CMT chip can compete with the single thread speed of x64, Power, Itanium or SPARC64. CMT chips are all about maximising the throughput of processor cores by presenting virtual CPUs and using up otherwise "dead time" spent waiting for main memory access. They simply are not optimised for single thread processing as to do so means allocating lots or silicon real estate to superscalar techniques such as out-of-order execution, concurrent execution and the like. To do so with the T series would be impractical - the entire processor design is based around on a completely different philosophy. Unfortunately there are many workloads out there where single thread speed matters. For example, to optimise response times or for large, high-throughput, Oracle databases. Indeed it is telling that the Oracle Exadata is built round Intel x64 architecture processors, not the T series.
Now that AMD and Intel are pushing adding many high speed cores into single chips (with two threads per core in the case of Intel), the only real advantage of Oracle's T series processor is on binary compatibility with SPARC code (important for a lot of legacy apps). In almost all other criteria the advantage will be with x64 architecture servers, and on price/performance there is just no comparison.
Looking forward to later this year!
I am looking forward to later this year, when the T4 arrives with faster threads, even though there are fewer threads per socket. A 300% increase in thread speed and 50% socket throughput is a welcomed advance.
It is bizarre how people are concerned about the M-Series or road maps for SPARC.
- M-Series jut received a speed-bump and is projected to get additional bumps over the next 3 years! SPARC has the clearest public road map of any computing architecture. With multiple vendors producing complementary processors (which are becoming increasingly competing.)
It is great to see the innovation continue in the marketplace!