Without announcing any new products or delivery timetables, Dell will begin to publicly share its new blade strategy today.

At least one ZDNet reader, however, has helped me to fill in the product-less roadmap. "In November," according to the reader, "Dell is releasing its next generation blade solution called the 2855MC. It supports Intel's 64-bit enabled Xeon processors and the pricing that Dell was discussing is ridiculous when compared to HP and IBM -- in the vicinity of 50 percent cheaper." So far, Dell has only confirmed the use of Intel's hybrid 32/64 bit-based Xeon processors (formerly code-named Nacona), the first of which was released this week and the approximate timing ("later this year" according to a spokesperson).

On the basis of market share and offerings (no Xeon-based blades), Dell must be considered a laggard in the blade server market. However, the company has a stern message for IT shops to consider before making any new blade investments. Dell cautions buyers that, as seductive as blades are, the promise of density and cost benefit when compared with the closest alternative -- 1U rack mountable servers (see my explanation of what a "1U" is) -- is not all that it's cracked up to be. The density advantage is overblown, Dell claims, and even where there are density advantages to be had, the premiums are usually unacceptable to most CFOs. In other words, you may be paying more for less when you should be paying less for more.

Though blade players large and small (IBM, HP, RLX, Egenera, Verari, et al) might disagree with Dell's theory, and perhaps regard it as self-serving given Dell's lack of progress on the blade front, my own back-of-the-envelope calculations reveal that Dell could be on to something. Whether or not Dell is on to something may be irrelevant. More relevant is what Dell plans to do about its supposed revelation --a game plan that, based on what I've heard, could rock the blade market in the same way that Dell has rocked other markets with its business model: a practice that Dell-insiders call "Dell-izing."

According to Dell vice president of worldwide marketing Paul Gottsegen, the Dell-ized, economies-of-scale benchmark for blade servers to meet or beat will be as follows: Using the latest in processor technology, a blade enclosure that's fully loaded with enterprise-class systems (blades) must offer at least 150 percent of the density available with comparably configured 1U systems at 25 percent savings over the cost of the matching 1U setup. So far, claims Gottsegen, no offering comes close to this benchmark-- but Dell will.

Dell isn't the only company that's drawing the 1U-factor into the blade reality check. In a recent briefing with Douglas Erwin, CEO of RLX (the company that invented blades), Erwin said that the company was having so much difficulty competing against 1U servers that it decided to launch one of its own just so it wouldn't lose business to 1U-committed IT departments.

To buy into Dell's theory, you must first approximate an apples-to-apples comparison of blades servers to 1U servers--which isn't easy to do. Performing such a comparison will require you to view all servers, regardless of form factor, identically from a systems management and provisioning point of view, as they probably should be. After all, to the person sitting at the management console, why should a blade server appear any different than a tower or 1U server? It will also force you to discount some of the attractive but difficult-to-quantify benefits of blades, such as hot-swapping and the elimination of cable nests. In fact, at the end of the comparison, Dell argues that even if you can quantify the additional benefits of blades -- such as potential power savings -- those benefits, even over time, will be negated by the premium acquisition cost associated with blade offerings over their closest 1U counterparts.

To gut-check Gottesgen's assertions about existing offerings, I requested blade configurations from several manufacturers (some through their Web-based configurators). To make sure I was extracting the most value out of the one-time cost of an enclosure, I made sure that each configuration included an enclosure that was fully maxed out with systems conforming to a basic enterprise-class configuration. My system configuration called for a dual processor Xeon-based system with two SCSI-based hard drives, 512KB RAM, on-board Ethernet connectivity, and any on-board management modules necessary to facilitate systems management and provisioning functions.

Performing such a gut-check requires some close reading of each vendor's fine print. For example, IBM's 7U-sized BladeCenter enclosure appears to hold 14 of IBM's HS20 systems (which appears to exceed Dell's 150 percent benchmark). But, upon further examination, that's 14 systems with IDE drives. If you want to go the SCSI route, which most companies consider to be the enterprise choice for on-board storage, IBM's SCSI add-on option consumes an additional slot per system. In other words, the density drops from 14 systems per 7U (2:1) to 7 systems per 7U (1:1). Compared to a 7 1U setup, the BladeCenter's density improves slightly due the fact that IBM's Ethernet switch can go into the back of the enclosure (it doesn't occupy a system slot). With the 7 1U setup, such connectivity would take up an additional U, bringing it to 8U (a comparison that, admittedly, sets aside the advantage of the blades' cableless nature).

But, again, that 1U improvement using the BladeCenter comes at a premium. Whereas IBM's Ethernet switch costs $2,199, a separate, enterprise-class 1U-sized switch from another vendor (a switch that supports more connections) can be found for half that.

Even considering the density advantage (seven systems with connectivity in a 7U space vs. seven systems with connectivity in an 8U space), it's not much to speak of. On the cost front, using IBM's online configurators, both the blade-based HS20 and IBM's comparably equipped 1U-based xSeries 335 server (which actually was better because it had two 3.06 GHz Xeons) rang the cash register at almost the same price (approximately $7,100). Throw in the cost of the BladeCenter enclosure (approximately $2,000), and now the price of the blade configuration exceeds that of the slower 1U configuration with a very minimal gain in density. In other words, based on the IBM example, Dell's theory isn't so far fetched.

I also checked with RLX. Based on my specification, RLX furnished me with two configurations: one based on a fully loaded blade enclosure, and one based on a comparably equipped 1U-based configuration. Fully maxed-out (based on my specification), an RLX 600ex chassis can accommodate nine RLX 2800i ServerBlades. The chassis consumes 6Us of rack space, which means that, at a 9:6 (3:2) ratio, the comparative density over a nine 1U configuration meets Dell's minimal acceptable density improvement of 50 percent. But again, when compared on a cost basis, the cost of RLX's blade-based configuration ($51,289) exceeds the cost of its 1U-based configuration ($49,088) by $2,200. Here, RLX fails Dell's savings-over-1U benchmark of 25 percent. To pass, the price of the blade-based configuration would have to come down to $36,816.

Verari Systems (formerly known as RackSaver), actually comes closer to Dell's benchmark. There's a slight catch, however: To get full economies of scale out of a Verari enclosure, which is a 7-foot tall rack, you need to load it with 88 systems. According to Verari CEO Dave Driggers, the list price of a fully tricked out 88-system blade-based setup (according to my specification) is $275,000. In contrast, an 88 1U server-based setup from Verari is $225,000. But Driggers warns that the street prices are much more in line with each other, which means, as least as far as Verari is concerned, there's a negligible premium involved in going the blade route. The density advantage however, is significant. According to Driggers, whereas the 88 blade configuration takes one 7-foot telecommunications rack, the 88 1U configuration takes two (and the racks have to be a little bit taller). In other words, a better than 100 percent improvement in density can be had for nearly the same price, which puts Verari's solution much more in-line with Dell's benchmark than are IBM's or RLX's solutions.

If you're a proponent of blades, you're probably welling up with emotion right now because it appears as though, in an attempt to make this an apples-to-apples comparison, I'm tossing aside important advantages of blades over their 1U siblings. Where the density improvement is significantly better than 1:1, for example, blades may have rescued you from a physical space problem.

Indeed, everyone's mileage will vary, perhaps significantly. It will vary in terms of the pricing differential between blades and 1Us that you'll pay. It will vary in terms of how important space savings are to you. It will vary in terms of the value you assign to those difficult-to-quantify benefits like the elimination of cable nests and hot-swapping. In fact, you can throw away all my logic if you want. The bottom line is that if Dell enters the market with an offering that meets or beats the benchmark, it will send IBM and HP back to the drawing board, as it should you.

You can write to me at david.berlind@cnet.com. If you're looking for my commentaries on other IT topics, check the archives.