Intel has had a great IDF thus far; the attendance is great, the products are innovative and honestly it's an exciting time to be involved with the tech industry. The first developer forum of 2002 is testament to the fact that even when companies are hurting financially and the sun doesn't seem to shine just as bright as it used to, engineering and innovation will prevail. After all, it's a love for this sort of technology that attracted most of the brightest minds in this industry to the sector in the first place.

In spite of all of that and in spite of the excellent keynote speech by Craig Barrett, the one thing that has been lacking from this year's IDF are the hard hitting product demonstrations that we're used to. Where's the air-cooled 4GHz Pentium 4? Where's the talk of the Pentium 4's successor? While IDF week is only halfway over, there has been a lack of some of the cool demos that we're used to seeing at these things - until we visited AMD.

As you've undoubtedly heard, AMD has announced that they have a working version of their upcoming Claw Hammer processor. Not only did we get a chance to see it running two different operating systems, but we actually got to hold the CPU (how geeky can we get?) and take a look at AMD's first reference board based on the AMD-8000 chipset.

Twice the pin-count of Athlon

The first thing we noticed about the CPUs is that they look a lot like the Socket-478 Pentium 4 processors. In fact, AMD did borrow quite a bit from Intel in designing the packaging of the CPUs as both the ClawHammer and SledgeHammer parts feature integrated heat spreaders (IHS), a technology which Intel has been using for almost two years now. This means that there will no longer be any horror stories of crushed cores from poorly installed heatsinks or badly manufactured heatsink clips.

The next thing you'll notice, which is difficult to gauge using these words, is that the CPUs are very heavy. They are easily the heaviest socketed CPUs I've ever held, with the weight being due to the IHS; it's simply interesting that they are heavier than their counterparts from Intel.

Next you'll note that there are no capacitors on the bottom of the CPU package to aid in power delivery to the core; again there's not much information as to why this isn't necessary but it's interesting to note nonetheless.

The 754-pin ClawHammer - Top featuring the IHS and bottom exposing the 754 pins.

By far the most interesting thing about the CPUs from a physical standpoint is their pincount. The ClawHammer has 754 pins (up from 462 on the Athlon and even up from 603 on the Xeon) and the Sledgehammer has a whopping 940 pins which is just over twice as many as the current generation Athlon.

The majority of the pin increase when going from the ClawHammer to the Sledgehammer is apparently due to the two additional Hyper Transport links and the dual channel 64-bit DDR memory controller vs. the single channel controller on the ClawHammer. Needless to say that manufacturing these things should be interesting.

The two SledgeHammer samples

The CPUs both the same basic footprint with the difference being that the pin-side of the Sledgehammer is almost completely filled with pins while the ClawHammer has a bit of room in the center.

Flying Solo

AMD's reference board was present at the demonstration and like the ClawHammer CPU being run, the board featured A0 steppings of the two parts of the AMD-8000 chipset that was used. These chips were not more than a month old and were already up and running which is the type of accomplishment which is very rare for AMD, in fact this is a first. While we're almost used to seeing this sort of thing from Intel, it's definitely commendable to finally see it from AMD.

The AMD reference board goes by the name Solo, and is a pretty standard ATX motherboard. There is a single 754-pin Socket on the board for a ClawHammer CPU and it introduces what will most likely become the new (or something very similar to it) retention mechanism for AMD's Hammer line of CPUs. If you'll notice, it's again much like the Pentium 4 heatsink retention mechanism except it only uses a single clamp in the center instead of two along the sides.

For the sake of not creating standards that unnecessarily compete with one another, we'd hope to see AMD just adopt similar specifications to what the Pentium 4 processors use so that the cooling manufacturers don't have to continue to make multiple lines of coolers.

Click to Enlarge

Click to Enlarge

You'll also notice that the board features an ATX12V power connector which was originally introduced with the first Pentium 4 platforms. It is still unclear as to whether or not this will become a requirement as it is on most Pentium 4 systems although it would make sense in order to prevent a repeat of what happened with the launch of the original Athlon and poor power supply support.

The AMD-8151 HyperTransport AGP3.0 Graphics Tunnel

There are two chips present on the Solo board that make up the AMD-8000 chipset; the chip closest to the CPU that looks like a mini-CPU is in fact the AMD-8151 HyperTransport AGP3.0 Graphics Tunnel with AGP-8X support while the chip in the position of a conventional South Bridge is the AMD-8111 HyperTransport I/O Hub. As you can clearly see in this picture, the 8-bit HyperTransport links make trace layout much simpler.

The AMD-8111 HyperTransport I/O Hub

The motherboard was outfitted with two DIMM slots and was built on a 4-layer PCB; it will shortly be making its way to Taiwan to be the basis for third party manufacturers to work off of.

The Demo

AMD had two identical ClawHammer systems running side-by-side in their hotel suite. Now remember that these CPUs are no more than 30 days old and are the very first steppings to be released from manufacturing (A0). This demonstration of A0-stepping next-generation CPU technology has never been done in such a short amount of time by AMD. Again, we're used to seeing this type of a demo from Intel but with no such demonstration at the conference AMD stepped in to fill the void.

Click to Enlarge

The CPU was manufactured on AMD's 0.13-micron Silicon on Insulator process out of FAB30 in Dresden. Unfortunately it was not running at full clock speed (which is something that Intel's demos usually do consist of), but we were told that the CPU was running as least as fast as the "other" 64-bit CPU out there (meaning at least the 800MHz - 1GHz of Itanium/McKinley). AMD informed us that they were still on track for a Q4-02 release at full clock speed which we've always hypothesized to be at around 2GHz.

Click to Enlarge - Note the lack of an AGP card in the slot

Neither test system was running with an AGP card mainly because of issues with the 8151 Hub however the hub was obviously functioning since it was connected to the 8111 I/O hub and that was working fine. The AGP issues are apparently working in lab but the systems they setup weren't able to run with AGP support.

The first system was a ClawHammer running 32-bit Windows XP, straight from the box with no modifications. For those of you wondering what OS you'll need to have in order to run the Hammer processors, your current 32-bit OS will work just fine. Granted that with a 32-bit OS you don't get any of the benefits of the x86-64 architecture, one of which happens to be the additional registers made available in x86-64 mode, but it will work. The Windows XP ClawHammer system was running Microsoft Word and Microsoft Excel test scripts nonstop without a hitch.

The next system was an identical ClawHammer system running a 64-bit Linux port. This setup was a bit more interesting; the demo consisted of two windows each with a a ball bouncing inside the window. The window on the left was running a 32-bit version of the ball-bouncing demo, while the window on the right was running an x86-64 compiled version of the same demo at the same time. This was a simple demo used to show that working with 32-bit and 64-bit recompiled applications concurrently was possible on such early hardware and it worked just fine. This system had apparently been up for the past 24 hours without a single crash, granted it was only bouncing balls all day but that's still an accomplishment for 30-day old silicon.

With that said, what we'd really like to see from AMD is more in the way of compiler support for x86-64. While there are currently tools available, what's necessary is a version of Visual Studio with support for x86-64 so that the majority of application developers simply have to use a switch in order to produce x86-64 binaries in their current development environment. The end result for producing x86-64 binaries is that the applications now gain use of the additional x86-64 registers thus improving performance from having to go to memory to fetch data much less. We've heard from developers stating that the performance improvement, especially when dealing with today's object oriented code, could be anywhere as high as 10 - 20% from a simple recompile. Maybe it's time for AMD to produce their own compilers to plug into Visual Studio much like Intel does with their C++ and FORTRAN compilers.

Final Words

This has been, without a doubt, the most exciting thing at IDF thus far and we just thought you'd get as much of a kick out of it as we did. For now, it's a bit of rest before it's back to the show for another day of cool technology and interesting announcements - stay tuned.