I only read about that earlier today on tomshardware.com, though it seems to be old news.

Should people still go out and buy alphas?

Would you recommend an ES-40?

Is the support still going to be there? Or is that just empty promises from Compaq?

Itanium seems to be the way for 64 bit computing, or is it?

(1). Use your common sense. (2). If you have to use it, then you probably can get a better price on system and chip being phased out. And the support could cost more. (3). If you decided to buy, then you will get the answer at the vendor, I think.

I work for Compaq in applications support for High Performance Technical Computing, which means that a) I have a vested interest in the quality of the processors inside our platforms and b) you should discount whatever I say as likely to be biased.

Personally, I'm relatively sanguine about this move, mostly because the initiative to move to Intel seemed to come from within the Alpha development group themselves. They could see that while Alpha has a performance lead now, it is also a shrinking lead and would likely be overtaken by an Itanium variant in 4 or 5 years time. The company preferred to make this move pre-emptively rather than defend Alpha down to the last gasp.

The competitve advantage provided by Alpha is performance, pure and simple. Should Alpha lose it's performance lead there would be no reason to preserve it.

That said, there will still be an EV7 Alpha, and speed-bumps of that as well, before the end of the Alpha road is reached. So Alpha will continue to be cutting-edge through the next 3 or 4-year buying cycle, before it will be replaced at that bleeding edge by Itanium.

The end of Alpha doesn't mean the end of High Performance Computing at Compaq. Far from it. Development of the AlphaServer SC 'supercomputer' class clusters is if anything being ramped up, though they'll have to change the name from 'AlphaServer' once Itanium is inside... Compaq are hoping to differentiate themselves by their platforms and their support 'expertise' (that's people like me, I think...) rather than by their processor. That could be hard because we're used to piggy-backing on the performance edge provided by Alpha.

For anyone who's thinking of buying an ES40, the move to Itanium is so far downstream as to be completely irrelevant. The ES40 will be rendered obsolete eventually not by Intel, but by the EV7 Alpha and the platforms built around that.

As for continuation of support for Alpha into the future, you should probably judge that on the basis of how well DEC (and now Compaq) supported VAX customers after the last VAX was developed.

Do any of you remember the PDP-11? It's heyday was around 1978, so I was surprised to receive a corporate email circular about 3 years ago announcing that the last one had just been sold! And that last one, presumably, is still being 'supported'.

I think you can be sure that Compaq will continue to support it's Alpha customers. It's much easier to keep a customer than win a new one, so it's in Compaq's own interest to support Alpha customers so that when they eventually move, it will be to a Compaq Itanium platform rather than a competitive one.

Many of us are lamenting the end of Alpha. Perhaps if it had reached 'critical mass' in the market years ago the resources would be have been available now and in the past to help it preserve its performance lead. But since it doesn't have the volume business, its economies of scale are lacking and it's development has slowed more than we would have liked. Given current performance projections and development costs, it was a pretty obvious decision for Compaq to drop it in favour of Itanium.

"I think you can be sure that Compaq will continue to support it's Alpha customers. It's much easier to keep a customer than win a new one, so it's in Compaq's ..."

I'm sorry!

"It's" is a contraction for "it is."

The possessive is "its" without the apostrophe.

This error is becoming more and more common. I know you do a better job supporting the Alpha than writing about it!

But it is good to know that Intel will continue to have competition. That'll keep their pricing honest!

Sorry Again!

I intended to post a private e-mail.

(1). In old days, service was important. So, IBM, DEC, VAX etc. were very good names. (2). In 90's, with down sizing, merger, etc., market becomes more important than the service. The goal becomes the market share, instead of the better service to old customers. (3). More good products and companies will disappear this way. (4). In my nearly 30 years of using various PC's, I had only a few times of professional service on my PC. One to fix the monitor on Radio Shack model III, one to upgrade the CPU on Amiga, one to fix the adapter plug on PC. So, basically, PC market is basically hardware-oriented. It is too rigid. (5). Anyway, such global economy is not very stable. Fluctuation in US market is having serious direct impact on Asian PC/IC manufactures already. (6). Like CFD, one can not calculate its value purely on the market value alone. (7). The hardware and the software alone are not going to do any good, if there is a lack of service between the company and its customers.

John,

My knowledge of computer history is not bullet proof, but I would hazard a guess that the PC has not been around for 30 years (IBM introduced their first PC in 1981). Do you really have 30 years experience working on them?

Enrico

(1). Before IBM, there were other PCs on the market already. (2). Among the most popular are Apple(II), Radio Shack Model III (monitor, keyboard, computer, two floppy drives were all integrated in one case, with user memory of 48k bytes) (3). IBM really was the late comer in PC field. And was not the main player until the success of Window3.1 in early 90's. And even in mid-80's, the Amiga computer can handle multi-tasking, graphics, multi-media sound and music, voice. I had used it for 2-D computer graphics and 3-D modeling and ray-tracing. (4). The popularity of PC actually was the result of poor economy, because the mainframe computers and the mini-computers used to be the main-stream of computer resources. and the more powerful option of the workstation was much too expensive than the PC.

(1). Obviously, it should be 20 rather than 30 years. So, you are right in math.

Enda,

I have always admired the compilers, debuggers and profiling tools that come with Alpha systems. They just seem to do the job better than thoses available elsewhere. Are they going to be phased out too?

- Steve

Hi Steve,

Your question about compilers is a key one and of immediate relevance, since Itanium won't replace Alpha for another 3 or 4 years while a sizeable fraction of compiler people are re-badged as Intel employees already.

The short answer is yes: compiler, debugger and profiler development for Alpha will be phased out along with the chip itself. That's only to be expected. Still, lots of people (including myself) want to know how compiler issues that might arise today or in 3 years will be resolved differently from say 3 months ago. The verbiage and "body language" coming from the corporate suits are reassuring, but the proof will be in the pudding, i.e., when a real issue actually does arise, we'll see how it will be handled.

As I understand it, those working on the 'GEM' compiler (i.e., the 'back-end' behind Fortran, C, and C++) have moved already or will move to Intel, while those working on the front-ends remain, and everyone is supposed to continue to interact.

In the past, there was a semi-bureaucratic, semi-informal method for getting something done in a compiler. I don't know how the 'culture' at Intel differs from at DEC/Compaq, though I suspect that since we'll be dealing with a different company things will get somewhat more formal.

Here's a quote from a manager in the compiler group (sorry, I don't know her exact title) in an email about a customer with just this question:

"(The customer) was particularly concerned about whether we would continue to support the Alpha compilers. I told him that we are committed to doing so. (The customer) asked what was happening to the compiler group. I explained that some of the compiler engineers will become Intel employees, to bring the Compaq compiler technology to IA-64. Intel will be opening a compiler engineering facility here in Nashua, near the Compaq facility. I expect the engineers to continue to collaborate closely."

For the moment I'd ask for the benefit of the doubt, pending real-life verification or otherwise of these promises.

Of course, since we'll be needing decent compilers and tools for Itanium processors in a few years, it's good to know that something of the DEC/Alpha tradition will be going into them...

Finally, there is more information about this Alpha/Intel deal at, e.g., http://www.compaq.com/hps/ipf-enterprise/tru64unix.html . Anyone with a question about the deal can email to: IPFAlphaHotline@compaq.com (IPF stands for Itanium Processor Family).

-Enda

(1). It is more a Compaq issue. Both are in PC business. (2). And PCs are mostly made in abroad now. (3). For the CPU, I think, there will be more different CPUs on the market in the future. (4). Who knows whether they will still be around five years from now.

I heard the I-64 is not living up to it's name (like most new chips I suppose).

AMD is going to have a 64-bit processor some time next year. Some say it will be better than the Alpha.

Also, look for the G4s. They could end up being the best processor around.

John Van Workum TTI "Online and On-demand"

> Also, look for the G4s. They could end up being the best processor around.

From the numbers I have seen, the G4's are the best processor around, but performance is unfortunately not the determining factor in this game.

Adrin Gharakhani

(1). What is clear in the phase-out of Alpha is, PC is gradually phasing out the traditional computers and couputer companies. (2). Soon, workstations will be replaced by PCs. The huge market size and the low cost of PC will change the face of the earth. (3). In the process, the traditional engineering companies using mainframe computer, mini-computers, workstations all will be replaced.(most likely disapper through merger because of inability to change due to lack of experienced engineers) (4). Persons from the super-computer era will think that a PC is too small and too slow, but this is because they are still thinking in terms of the super-computer. (5). "thinking" means the "software". The evolution is still continuing. The computer will determine the software, and the software will determine the fate of a company.

Every engineer using a PC has a "supercomputer" at their desk nowadays. You have to look at the state-of-the-art. The PC's we're using today are orders of magnitude faster than the "super-computers" such as CDC we used to use up to early 80s. In terms of megaflops, some of the pc's today are actually faster than the Cray YMP I used to use at late 80s early 90s.

And of course, today, there are no super-computers, per se. Commercial cost is not exactly the reason. Even way back when, companies such as Cray survived because they were practically subsidized. Because, for some reason, the US gov't decided in mid-90's that pc's can do the job of supercomputers (supercomputers based on the standard of today not yesteryear) there were no more subsidies to go around for developing computers that are an order of magnitude faster than today's pc. Ideas of teraflop computing are all based on the massively parallel paradigm, which in itself is not bad.

Clearly, people can solve problems on their pc's that were impossible to solve on a "supercomputer" just a few years ago! So I see no reason to complain as vigorously

I believe, market or not, the computer industry will make yet another full circle (that will be beneficial to people like us). I was thinking just the other day, that "common folks" used to think that their PC DOS was it and that Unix was a waste of space, useless and only good for a bunch of geeks. Microsoft went through cycle after cycle after cycle of "improvements" and larger and larger and larger "wastes of space" to finally come to the same point us geeks were at all along. Now, Mac has OS X, MS has XP, (all thanks to the free availability of linux) and all the "common folk" and the market they represent finally end up using what they should have from the very beginning. I see the hardware end of things making similar progress. Why would anyone need gigaflop operation for word processing? And that was a legitimate question just a few years ago. But ask anyone to use an IBM XT or AT today (remember those?) even for word processing and you can't pay them enough to use those junks. The hardware may always be behind what the engineer wants, but not too far behind (even if the market dynamics are determined by the "common folk")

Adrin Gharakhani

(1). When I say super-computer, I mean the environment around the super-computer. (2). When I say PC, I mean you can call in and order a computer from a local computer store. (3). So, the speed is not the issue. (4). When a company decided to bring in PCs to replace the workstation, right away, he will be facing the problem as what to do with the old analysis codes and future new codes in the engineering department. (5). If suddenly you have several hundreds codes to worry about, then you will understand why the software determines the fate of a company. (6). If the simple minded users can not feed the supercomputer code directly into the PC and get the identical result, then I guess, it will become a problem. This is the issue I am talking about. (7). In other words, the failure of a company comes from the failure to switch to the new system to compete. And then the company will finally decide to sell the company rather than trying to fix it. (to convert the software so that it will work in the same way or better on the new computer system) (8). More large companies will run into this problem because it requires large investment to do so, and it is cheaper to cut back and sell it in the open market.

Please allow me to try to clarify what is really happening with Alpha, and with technical systems from Compaq in general. I am the business manager for engineering simulation applications at Compaq.

While we are transferring Alpha technology and engineers to Intel, new Alpha systsms will be released through multiple speed bumps and the next processor generation, EV7. Support for Alpha systems will continue well beyond that.

Alpha systems are the highest performance systems for CFD and will continue to be for a number of years. ES40 is an ideal system for CFD, as it combines excellent floating point performance with a crossbar-based system architecture that provides the memory and I/O bandwidth necessary to keep the CPUs fed with data.

We have begun to discuss the eventual migration from Tru64 UNIX on Alpha systems to later generation IPF systems with CFD software vendors, and their response has been quite positive. The migration process will be quite simple, with the same robust OS on both architectures, and full source code compatability. Since all Alpha applications are inherently 64-bit, while most other vendors moved from 32-bit to 64-bit leaving 32-bit vestiges in their systems, we think the best way to create a robust 64-bit application on IPF will be to start on Alpha and Tru64 UNIX, and move to IPF and Tru64 UNIX later.

For further details on the Compaq/Intel announcement, see http://www.compaq.com/hps/ipf-enterprise/index.html. Terry Shannon's article, which is referenced on that page, provides a good third-party analysis of the announcement.

I would be happy to respond to emails with any specific questions.

Ed Turkel