Things were quite strange in the month before the Athlon64 launch. That was about the time reviewers received first details of FX51 — the Enthusiast version of Athlon64. For those who had been following AMD's slow progress to A64 launch, it certainly looked like AMD was worried and wanted to make absolutely sure that A64 would outperform anything Intel could throw at the launch. Perhaps the EE chip info leaked to AMD, and the FX was the insurance just to make CERTAIN Athlon64 would maintain a performance lead. Perhaps AMD was concerned that lack of Dual-Channel memory support on Athlon64 might make consumers think A64 was a slower chip. It isn't, but perceptions sometimes matter more than facts. We don't know the real reason for the FX launch, but whatever the reason, it is clear that it was a last minute decision.

FX51 is basically an Opteron with a different name. It works with a BIOS update on many Socket 940 Opteron boards, and will not work on any of the Socket 754 boards being introduced right now. Like Opteron, it also suffers (or benefits, depending on your perspective) from the requirement for Registered Dual-Channel memory to operate. The only concession is that while it can use Registered ECC memory like Opteron, only Registered is required on FX51.

The even stranger part was AMD's positioning of FX51. They refused to even sample Athlon64 chips, forcing review sites into a position of begging their Advertisers for Athlon64 chips to review anything. Athlon64 launch partners were frustrated because many review sites could not review their new boards because they had no Athlon64 chips. The only sampling that was done by AMD was the very limited sampling of FX51 chips in an AMD prepared system. AMD was clear in their actions and words that they wanted the Launch to be with Athlon64 FX51 processors.

Even now, we have to wonder why. Athlon64 is fully competitive with Pentium4 3.2GHz. It enjoys both a price advantage over the P4 3.2 and the Athlon64 FX51, and, maybe even more important to buyers, it is available while there are almost no FX51 chips to be found. There were also many, many Athlon64 boards at launch, and only the Asus SK-8N board for the FX51 — a board first released as an Opteron motherboard. In addition, AMD was clear in its plan to move FX51 as quickly as possible to Socket 939 in 2004, which is capable of using regular unbuffered memory like most enthusiasts already own. AMD also announced very small production numbers for Socket 940 FX, leaving little incentive for manufacturers to produce Socket 940 boards for the desktop.

Thankfully, the major players — like Gigabyte, Asus, and MSI – have delivered or will deliver Athlon64 FX boards. We were glad to see Gigabyte's K8NNXP-940, which continues Gigabyte's recent stretch of high-end “Dual-Miracle” motherboards. The name is the same as Gigabyte's nForce3 Athlon64 board, with the addition of “940” to the name. As you will see, however, the differences are much deeper than just a different socket.

Gigabyte K8NNXP-940: Basic Features
Comments Locked

35 Comments

View All Comments

  • Anonymous User - Saturday, October 11, 2003 - link

    Hmm.. something just occured to me. (This is #24 again.) Anyone else remember the days of the Pentium and Pentium Pro? Well, it seems like we may be reentering the whole "high-end CPUs are different from midrange ones in ways other than clock speed" thing.. except this time around, the Macs aren't faster (the G5 and its super-deep pipeline can kiss my ass, thanks.. and probably the Hammer's while it's at it), and there are two companies in the game. This is going to be fun.
  • Anonymous User - Saturday, October 11, 2003 - link

    Hey, why isn't the Nehalem in this review? So what if it doesn't exist? They've got like 80% of it planned out now anyway, it's unfair to have this review biased towards AMD.

    Well, SOMEONE had to be ignorant and stupid, and hell if I'm going to say a thing about the Pentium 4 Xeon MP Edition.

    Uh. Anyway. The Athlon FX may just be a rebranded Opteron, but it's cheaper than the rebranded Xeon MP and much better at its job, so who cares what's a rebranded what? Not that I'd ever buy an Athlon 64 at these prices, but it seems the only market sector Intel has left is the low high end :D
  • Anonymous User - Friday, October 10, 2003 - link

    Excellent review! I'll be reading all of your writings from now on. :D
  • sandorski - Friday, October 10, 2003 - link

    Sweet motherboard, makes me think that as Athlon 64/FX motherboards mature, more performance will be acheived.
  • Anonymous User - Friday, October 10, 2003 - link

    Haha, good point #20!
  • Anonymous User - Friday, October 10, 2003 - link

    #12, perhaps because P4EE does not exist...
  • Anonymous User - Friday, October 10, 2003 - link

    When will this board be released?
  • Reflex - Friday, October 10, 2003 - link

    #11: I am not reffering to Quad-Channel DDR as I think you believe. I am reffering to Quad-Data Rate SDRAM. It uses the same pin count as DDR but sends information four times per clock, resulting in twice the bandwidth as DDR. If AMD supported it in their on board controller it would not require a higher pin count.

    However there must be some technical reason for QDR not appearing by now since its been 'just around the corner' for over two years now...
  • Anonymous User - Friday, October 10, 2003 - link

    Mostly I meant that running hand-compiled 64-bit apps would be irrelevant. I'd love to see another article in a few months, when Linux apps start actually arriving in 64-bit versions. But until then, it would be akin to Tom's OC'ing the P4EE. It may be interesting to a few people, but it would appear biased to almost everyone else.
  • Anonymous User - Friday, October 10, 2003 - link

    #15, 64-bit tests running linux would not be relevant? what about those of us who are running linux right now? I for one would love to see a 64-bit set of linux benchmarks included.

Log in

Don't have an account? Sign up now