Archive 1997–2024

WinHEC 2005: Coverage Wrap

Closing Thoughts

If we were to be brutally honest, there wasn't a whole lot of new or amazing stuff shown at WinHEC. Even a first time tradeshow attendee (Jarred) noticed that there were a lot of bored looks. To quote a representative from a major hardware vendor, "I think a lot of companies just come out of respect for Microsoft." Still, after the rather lackluster Keynote presentation, there was plenty of useful information when we dug a little deeper. With over 90% of the market, Microsoft pretty much defines the future of most computers right now. If they say it's time to have DX9 level 3D graphics in Windows - whether that be on a Desktop, Laptop, Server, or Workstation PC - then it's pretty much a foregone conclusion. It will be interesting to see how the server platforms make use of 3D (or if they even make use of it).

While we applaud Microsoft for bringing forth the advancements in Longhorn, we have to admit that they are long overdue. The advancements in metadata and file representation have been around in the Linux world for years, and Apple's OS X is still hands and feet above anything we've seen of MS' new look and feel. We are certainly not going to be the one to speak out against Microsoft for trying to bring their OS up to par with the rest of the world. Honestly, being the platform for which most applications are written today should place weightier expectations on Windows than other operating systems. Microsoft has a responsibility to live up to their position in the market place, and unless they make some huge unforeseen strides with Longhorn, they will simply be keeping pace rather than setting it.

One of the questions we liked to ask of the various hardware vendors was what they thought the real world use scenarios and performance benefits of 64-bit would be. Obviously, servers and workstations will benefit, as will anyone with a large amount of memory that wants to work with a large data set. Graphics artists that want to open up numerous large images at the same time will find it helpful, for example. But what about the more typical users that only have 1 GB of RAM in their system? We heard a lot of claims of "substantial performance increases," but there were very few people willing to quote a concrete number. AMD claimed that 20% was common, but several others felt 5% or so was a more realistic expectation, brought about by the additional registers. Obviously, opinions on what constitutes a substantial increase will vary, but if it's only 5%, we think most people are better off continuing with XP-32 for the present.

Speaking of AMD, we chatted briefly with Barry Crume (AMD's director of product planning for the server/workstation segment) about the future of processors. There were a couple of interesting things that came up in our conversation (and since Barry didn't say otherwise, we'll assume it's okay that we mention this). The first thing we talked about was CPU clock speeds. Where Intel in the past talked about reaching 10 GHz and beyond, it's telling that one of their sessions was entitled "Intel Enterprise Platform - Value Beyond GHz." One fact that came up is that at 5 GHz, light can only travel 6 cm per clock cycle, and electricity would be more like 3 or 4 cm. Barry said that he felt 4 to 5 GHz may be the maximum clock speed that we'll see without some fundamental changes in technology or really long pipelines. He also stated that CPU designs with 10 to 15 pipeline stages appear to be the optimal solution, although as many as 20 is perhaps reasonable. In other words, the days of 28 to 39 pipelines stages (Willamette, Northwood, and Prescott) are basically over. Intel is admitting as much with their increased focus on the Pentium M architecture.

The other interesting comment from Barry was that he stated AMD will be using DDR2 in a future processor. We had asked if they might simply skip DDR2 altogether and head straight to DDR3, and he said they would not be doing so. His explanation was that DDR2-400 didn't make any sense at all, and while DDR2-533 has some advantages, it's really only at DDR2-667 that performance benefits become useful. How useful? He stated that a 10% performance increase was likely. Whether this was just a guess or if AMD already has DDR2 prototype processors - remember that with the integrated memory controller, AMD will have to create a new version of their chips to support a new memory type like DDR2 - we don't know. We were actually surprised at his response, as we thought DDR3 was close enough that AMD would simply skip a generation. Then again, they might like the idea of allowing Intel users the ability to convert to AMD with the only required purchases being the motherboard and processor. Having a common memory platform certainly isn't bad for consumers.

At the end of the show, while we would definitely say that WinHEC doesn't have the flash or glitz of Computex, E3, or some of the other larger shows, there was still plenty of good information to be found. Hardware devices still need robust drivers to achieve maximum performance, and that's what WinHEC really focuses on. Will we go again next year? Maybe, or if not DevCon 2006 should only be a few months before the ship date of Longhorn. It's certainly useful to keep tabs on how MS views the industry. A better option, judging by recent MS release dates, is that we'll be sure to attend WinHEC 2007 for the actual launch of Longhorn. (The current official word is that Longhorn is due in mid to late 2006. However, given the slippage of XP-64 and features like WinFS that have been pushed back to a later update, we remain skeptical.)

Originally published at https://www.anandtech.com/show/1670 (Wayback Machine).