[quote]Originally posted by HOS:
The thread's been hijacked!
This is what I get for having just been on jury duty!
Sorry, I don't mean to be jumpy. It just seemed that the tone had changed, and it wasn't clear to me where you were being facetious and where you were being serious.
Anyway, the 2¢ version of my brand argument is that Apple's and SGI's brands don't overlap, and are considered extremely different, a la Mack trucks to Jaguar sedans, rather than Bentley sedans to VW sedans. There's clearly an added cost to an Apple/SGI merger here, since it either dilutes existing Apple and SGI brands, or else the new SGApple has to create a new brand...
Just for fun, to maybe steer this thread back a little bit, here's a neat link I found. Just read the first few pages- this background detail repeats what I've seen elsewhere, and I'm willing to accept it as being real.
Ignore the stuff about Sun's V480, just read the bits about how IDC expects the worldwide server market to move. Especially look at that graph that shows traditional servers selling fewer and fewer, while smaller "rack optimized and blade servers" will sell more and more.
Accepting this, it's a really bad sign for SGI, since SGI's core market (those big servers) is shrinking, while it's not a bad sign for current Apple strategy (Xserve).
It's also a really bad sign for Apple buying SGI- why buy a company that's shrinking so quickly?
And I thought this thread had run out of steam!!
Blade servers are an interesting development: But they have a segment and a place in the market just like every other product segment (handheld, sub-notebook, laptop, desktop replacement......, high-end corporate server, mainframe).
My particular belief is that blade computing is a really great solution for a particular kind of problem, but the constraints on I/O in most implementations limits their applicability to roles such as Java servers (potentially the basis of Sun's interest), thin-client serving, or web serving.
As for the high-density 1U or 3U rack-optimised server. They are great "in their place", namely SMEs, workgroups or similar. A 1U server with 3 PCI slots does not give the kind of redundancy/fault-tolerance or expansion that larger companies often require.
A render farm or compute farm executed using what are effectively densely clustered 2-way xServes and Ethernet (at whatever speed) is a fantastic way of solving a computing problem in a modular, cost-efficient method for SOME marketplaces. However, there comes a point at which the combined capital cost and TCO can no longer compensate for what is essentially a "workaround" solution for many computing problems.
The next technology step: Highly expandable 4-ways or 8-ways connected by a high-speed fabric, such as Infiniband, is exactly what I'm proposing. My next proposal, the usage of a coherent cache manager, would allow the results of the computation on one x-way node to influence the computational action of any other x-way node. Whatever you say about Ethernet-connected compute farms, managing all of the memory across the farm at the kind of performance that SCI + Infiniband would offer is a pipedream.
I wish to differ on the perception of Apple's and SGI's brands!
My perception of AAPL is that it is an excellent "evolution factory" for ideas which are awaiting the mainstream; GUI, WIMP, Postscript, USB, CD-ROM, and several dozen other technologies (including DVD-R) finally found their potential when Apple found a way to bring them to a volume marketplace.
However, I don't believe that anyone with an objective eye could deny that Apple has lost ground in the hardware marketplace. We all hope that the ground will be regained with a technology hike at the forthcoming MWNY or (more likely) MWSF, but - even if that technology hike happens - Apple will still, rightly or wrongly, be a vendor with a very thin product catalogue.
I perceive SGI as the Alpha male of high-end post-production, with hardware - even at a CPU level - that has been progressively evolved to very skilled at a) floating-point math and b) graphics I/O.
This evolution has been both a virtue and a vice for SGI. with the specialisation denying SGI any kind of marketplace in "traditional" IT environments such as data-warehousing, client/server business applications or even the bread-and-butter file/print serving marketplace.
In many ways, Apple's "perceived" specialisation in a (comparatively) narrow market segment also leads to parallel "vicious" and "virtuous" circles: A vicious circle of being unable to re-enter the "conventional" SME or enterprise markets, as much because of the (mainly spurious) WinXXX apps/peripheral portfolio argument as the (completely spurious and now completely outmoded) "open systems" argument; and a virtuous circle of having nearly every major Creative audience awaiting your next offering.
In these respects then, SGI and AAPL remind me of the TNG episode where an "alternate" Picard from six hours in the future has crash-landed a shuttle onto the deck of the Enterprise.
My point being that SGI is almost certainly a "view into the future" for AAPL, self-selected marginalisation ultimately leading to erosion and entropy until all that remains is the computing equivalent of a pulsar, sending out cyclical reminders (every major expo) of one of the greatest centres of innovation that any industry has ever seen.
Assuming (and I say this with all the relevant fine print) that the URL
I referred to in my previous posting (which was a leech from another thread that has since been closed) is in any way a reflection of a solid truth, both SGI and Apple can save each themselves and each other from extinction (medium- and long-term) respectively.
If G5 is seriously beiong considered by SGI, then logic dictates the chip must be a viable floating-point performer when compared to their owen MIPS processors. Even more importantly, it says something about the untit when compared to x86 or Itanium or even Clawhammer, either with regards to floating-point or bandwidth or a combination of both.
However, AAPL's more mainstream requirements would also demand that the chip is capable of relatively good integer benchmark, because an Oracle or a Sybase or even a 4D are going to focus on that parameter for performance.
So, assuming The Rumour is true, SGI will deliver IRIX onto G5, thus becoming technically capable of pursuing large-scale DBMS implementations were it not for the fact that (from what I can tell) there is no Oracle or Sybase implementation for IRIX at the moment.
Conversely, Apple will ultimately deliver OS X onto G5, which will be the first mainstream 64-bit capable CPU to be released at any kind of cost which could be considered attractive.
The word "mainstream" in this context means destined for a wide range of marketplaces including consumer PCs (something which neither Sparc nor Itanium can claim): This is overwhelmingly important - consumer market means economies of scales, economies of scale mean cheaper to buy, cheaper to sell.
So now, Apple "popularises" the 64-bit capable platform which means that SGI can take advantage of economies of scale and sell its' own hardware more competitively, none of which I mind at all.
But AAPL now has something that SGI need even more than AAPL: a broad mainstream applications portfolio, including the dreaded 800-pound gorilla of the business workplace, MS Office.
Conversely, SGI have something that AAPL would (or certainly, should) like: a genuine reputation for building "big iron" (or at least, "bigger iron" than Apple currently can claim) that has penetrated major R&D facilities like BP, Ford, Volvo and hundreds of others. Likewise, that "big iron" - with the prospect of an integer-friendly G5 and license to use cc:NUMA design - is exactly the kind of iron that Oracle or Sybase can use to do serious damage to the world of DB2 and the S/390-class mainframe.
The key here is the political link between Apple and Oracle: If Larry Ellison chooses to support Apple, it happens - no ifs, no buts. So the argument that AAPL would have no reputation as a vendor of DBMS iron would only be partially true at that point; it would have whatever reputation Jobs, Ellison, Schiller and all of their wizards and familiars chose to endow.
To reiterate the summary of my position: -
My point here is that the potential sum of these two companies becomes greater than their parts conditional on the validity of The Rumour: It assures SGI and their customers of a future with a broader applications portfolio than they could ever expect; it assures an enormous proportion of SGI personnel a future in an uncertain world, even the basic act of purchase gives Apple greater weight and mindshare than could currently be the case, especially in Hollywood - the company becomes an "across-the-board" vendor of highly capable computing systems, scalable from classroom to corporate, from digital media to datawarehouse.
I would point out that I think that all of this is about as likely as Spiderman not generating a sequel (i.e. slim, bordering on non-existent), but I can't help wondering...