Judging by the presentation I received at TechXNY from AMD Computation Products Group Ed Ellett, AMD is on the verge of making a 64-bit server bet that so far appears to be a risky proposition.

There's no question in my mind that AMD's forthcoming Opteron processor (formerly code-named "Hammer") has innovative and groundbreaking technology. Without ever having tested the processor, the only attribute (and its chief selling point) that I can vouch for is the hybrid design that allows the chip to run 32- or 64-bit operating systems natively. Ellett claims this feature is extremely important to IT managers because it means they won't have to reinvest in new 64-bit machines (the way they'd have to if they stuck with Intel-based systems once they decide to start making the switch).

Another reason AMD says Opteron should be the prime processor architecture (and one that I can't vouch for) is that, in its 32-bit mode, it will offer twice the performance of its Intel-based counterparts. In other words, one AMD processor could handle the equivalent workload of two closely rated XEON processors thereby affecting the overall costs of multiprocessor systems and the pay-by-the-processor licenses of enterprise applications like Oracle's. Whether this is true or not remains to be seen. So far, no independently run and audited benchmarks have been published.

But let's just assume that everything Ellett says is true. It's an impressive story. Unfortunately, as I told Ellett and his posse of AMDers, IT managers don't buy 64-bit processors--they buy 64-bit servers. What popular server vendors (tier one preferably) are going to make and sell Opteron-based servers? The answer to this question is the real indicator of whether Opteron is going to get any traction on the server side.

How do you actually get an Opteron server?
Don't get me wrong. I'm drinking the Opteron Kool-Aid. AMD's Opteron story rocks. If the Opteron's 32-bit price/performance is competitive, I have no doubt that it will get some decent traction--as AMD always has--in the desktop market where it will play under the Athlon name and maybe even in the 32-bit server market against Intel's Xeon. My only concern is how do you actually get an Opteron-based server? At the time of our meeting, Microsoft was the only tier-one hardware or software vendor (that I'd heard from) planning Opteron support. Microsoft announced that it would ship an Opteron-compliant release of the 64-bit Datacenter version of Window .Net Server, but the company has not committed to a date.

According to AMD spokesperson Whitney Otstott, "All of the tier-one server vendors have seen Opteron and are very interested." So, being the journalist that I am, my next request was rather obvious. "In that case, please give me the names and numbers of contacts at those tier-one vendors so that I can verify that interest." All I wanted to do was explain to our audience where they could buy AMD's technological marvel. AMD's response? "We can't. Nothing's been announced, so that's confidential."

I subsequently talked to Hans Mosesmann, an analyst with Prudential Securities who upgraded AMD last month. Mosesmann was evasive when I asked about who he thought would adopt the Opteron. "I can't tell you who it was. But what I can tell you is that Dell has come out and publicly stated that they aren't going to use Itanium 2 and are looking very closely at the Opteron. So, you can connect the dots. AMD is doing well on the design win front, and it's becoming more evident that Hammer is gaining more momentum on both the desktop and server. If Dell were to use it, the stock could go to 20 or 30bucks. It's real technology from a real company. But I can't give you specifics in terms of who is looking at it. The top three for sure are looking at it as are all the traditional clone vendors."

No problem. I called the potential Opteron systems vendors myself. But since there are only three tier-one system vendors (Dell, HP, and IBM), I decided to expand my inquiry to include some major software vendors too. (My ad-hoc survey should in no way be considered exhaustive.)

Unlike AMD's other processors, the 64-bit Opteron is not 100-percent compatible with Intel chips. This means that all software that runs on Opteron will need to be redesigned and recompiled in the same way that software vendors must recompile their products for other non-Intel compatible processors like Alpha or SPARC. In other words, it will take a bit more effort than usual for a software vendor to support Opteron.

From the hardware vendors, here's what I learned.

IBM IBM xSeries Server Products (the Intel stuff) media reviews analyst Michael W. Haley basically said there's little or no chance that IBM will build an Opteron-based server. It's easy to see why IBM would be cautious. Not only does IBM already have its hands full with two processor architectures (Intel and PowerPC), the company has so far sunk millions of dollars into its Extended X Architecture; a proprietary chipset that's optimized to work with Intel processors.

I doubled-checked with IBM eServer spokesperson Tim Dallman. "eServer" is the mother brand of all IBM servers--xSeries (Intel), iSeries (AS/400), pSeries (PowerPC-based AIX), and zSeries (the mainframes). Dallman told me: "We continue to evaluate new technologies to ensure that we are going deliver the best value to our customers. With respect to Opteron-based servers, we have not made any announcements regarding Opteron and IBM's family eServer solutions. Beyond that, we don't comment on our future product plans."

While it's difficult to get a consistent reading from IBM--a company that's often mum on products until they're announced--I wouldn't expect to see IBM make a huge commitment to the Opteron any time soon.

Dell
Dell is the systems manufacturer that Prudential's Mosesmann hinted was the tier-one vendor that was seriously interested in Opteron Ten years ago, when I interviewed Michael Dell in front of an audience of about 250 people, I asked him why he didn't shave the prices of his systems even more by going with AMD. So he turned to the audience and asked, "If I could save you 50 bucks, how many of you would be interested?" About half the audience raised their hands. Ten years later, Dell has never sold a product with an AMD processor. Never.

Referring to Opteron, Dell spokesperson Bruce Anderson told me "Our strategy is 100-percent customer driven. If you talk to any number of our enterprise customers, our strategy is relatively easy to determine and so far, we haven't seen the interest in Opteron. We are, however, aware of AMD's plans, we talk to them as well as all other processor manufacturers all the time, but have not announced any plans ourselves. If there's sufficient demand for that technology, we'll bring a product to market, and we'll bring it to market faster than anybody else and apply the efficiencies of our direct model in a way that makes the cost manageable."

Based on my experience with Dell, which is rightfully ultraconservative with its business decisions, when it says its 100 percent customer driven, I take that to mean a significant share of customers have to be asking for it. For example, the company still doesn't sell a handheld system because it thinks the market has not matured enough. And, the company doesn't see enough justification from customer demand to start pumping out Itanium 2 machines. So, imagine what sort of demand would have to exist before it started to include sell AMD Opteron servers?

HP
Staying respectful of some important relationships, Hewlett-Packard Business Critical Systems worldwide product line manager Brian Cox said "AMD is quite a big partner of HP's. We do a big business with them in laptops and desktops primarily in the consumer space so they are a very important partner. But, I haven't had a customer say they won't buy a server from me unless it had an AMD chip in there, and frankly we are a customer-driven company."

But instincts about the market and AMD's resources also seem to drive Cox's decisions. "I like the AMD guys. They're gutsy. But compared to Intel's scale, it will be a daunting challenge for AMD to come up with yet another 64-bit architecture. There are already 20 different hardware vendors producing servers and workstations based on Intel's architecture and the software vendors like this. For them, it's less attractive to see another 64-bit architecture emerge. As the co-developers of Itanium, HP and Intel have both spent billions of dollars to bring Itanium to market and it has a design that can last 20 years. Again, AMD is a great partner of HP's, but you have to look at it honestly. Are the deep pockets there and do they have the resources to develop this ecosystem and make it thrive? It could happen but it will be a challenge."

So, currently, as far as I can tell, no tier one systems vendor is ready to express any serious interest in Opteron. For good measure, I asked the folks at Fujitsu. They make refrigerator-sized multiprocessor systems and they didn't express interest in the Opteron. .

I called back AMD to share my findings and get their response. I spoke with Kevin Knox, AMD's enterprise segment marketing and business development director, who said: "I can assure you that we are working very closely with IBM, Dell, and HP on Opteron and are seeing strong interest from them as well as other OEMs," Knox said. "We're just further along with some than others. Obviously we have to have confidentiality with our OEM partners, but we're not seeing what you're hearing as far as our interactions with them are concerned."

An entire ecosystem that must be developed before another 64-bit processor can succeed and part of that system includes the major software vendors. After all, even if some hardware vendors were planning to build Opteron-based systems, it helps to know what software they could run on it. As it turns out, the software community is Opteron's ray of hope. For starters, both Windows and Linux will support Opteron's 64-bit mode. Of the two, Linux will probably be available first. But on the application side, support for Opteron was mixed so far.

BEA
BEA is one of the first software vendors to publicly get behind Intel's Itanium 2 in a major way. I spoke with BEA product marketing director John Kiger to see if that interest might extend to Opteron. The company's flagship product--WebLogic--is Java-based. Therefore, it only requires a J2EE-based Java Virtual Machine and has no specific dependencies on the operating system or processor. But BEA also makes a J2EE-based virtual machine (JRocket) that has dependencies on the operating system, and it would have to be redeveloped for Opteron.

According to Kiger, "BEA has partnered with Intel extensively so that together, we can offer optimized solutions that run on the Intel architectures. We have not looked at what it might take to implement JRocket on the AMD architecture." While that's not a "yes," I'll be fair and say that's not a "no" either. The partnership with Intel seems a bit too cozy for BEA to warm up to AMD any time soon.

IBM, again
While the hardware side of the company appears to pretty cold to the idea of building systems based on AMD's processors, the software division actually appears to be excited. The company is going to announce and demonstrate an Opteron-based version of its DB2 database for Linux at next month's LinuxWorld. IBM spokesperson Sean Tetpon told me that "one of DB2's unique selling points is the number of platforms it supports. It only took two days to port DB2 to Opteron on Linux. So we went ahead and did it using SuSE's distribution of Linux, which was tuned for Opteron, and we will be showcasing that solution at LinuxWorld."

As for the rest of IBM's software portfolio, which includes products like WebSphere (IBM's competitor to BEA's WebLogic) and the collaboration tools from Lotus, Tetpon says the company is looking into Opteron support but has nothing to announce DB2-style just yet.

Of course, even the announcement about DB2 begs an important question; "What system is it going to run on?" IBM itself doesn't have an Opteron-based system to run this on. Tetpon cited Newisys, a relatively unknown manufacturer that, judging by its Web site (www.newisys.com), is still in startup mode. Hmmm.

J.D. Edwards
According to Lenley Hensarling, J.D. Edwards' vice president of product marketing, "We don't have plans at this time for AMD's Opteron processor. Support for specific platforms is driven by customer demand within our target markets. If we see significant customer demand our position might change."

The more I hear the "customer-driven" answer, the more I realize that part of AMD's problem is the old chicken-and-egg conundrum. It's hard for customers to get excited about Opteron without first seeing some systems and software that run on it. But it's hard for systems and software vendors to get excited about Opteron without first seeing customer interest. This can't be a good situation for AMD.

Oracle
Oracle was a bit more evasive on the issue of Opteron. About all I could get was a quote from the company's platform alliances vice president Doug Kennedy. Says Kennedy, "Currently, work with AMD is at the R&D stage." That could mean "yes" but it could also mean "maybe." Unfortunately, after a week of trying, I was unable to get any more information about what that means.

It's not surprising to me that, in terms of software, the primary interest in Opteron is coming from the database community (Oracle and IBM's DB2) where 64 bits of addressable memory can really come in handy. If AMD's promise about performance vs. Itanium turns out to be true, then Opteron may find a home in the niche of high-performance databases

I'm skeptical at this point about Opteron servers. Until there are some verifiable benchmarks to show the price/performance benefits, and until I see at least one of the tier-one system vendors commit and there's more support from the software community, I remain unconvinced about the long-term prospects for Opteron as 64-bit server processor.

On the other hand, AMD has everything to gain and nothing to lose by betting that its 64-bit technology will garner interest at a time when other 64-bit technologies are dropping like flies. As long as AMD can keep Opteron's 32-bit performance and compatibility on par with Intel's 32-bit offerings while keeping a lid on its cost, Opteron should do well in the 32-bit market. Any traction for its 64-bit technologies will be icing on the cake. Also, AMD's mainstream processor business is but one of three major markets the company is active in. The company participates heavily in the flash memory and embedded systems markets.

Says AMD's Knox, "In terms of Opteron's importance to the company's future, it's definitely a strong piece of our future as we go forward. I wouldn't say it's a bet-the-company product. We also have the flash and embedded stuff. But it's an important piece of our story. We're betting our server on Opteron but not immediately on 64-bit. We think Opteron is well-poised for enterprises that want two, four, or eight processor systems."

What do you think? Can Opteron penetrate the 64-bit server market or does it have a huge uphill battle against other 64-bit architectures like IA-64, SPARC, and PowerPC? Let me know. TalkBack below, or write to me at david.berlind@cnet.com.