Why the T-bird won't ride in Slot A

As we reported earlier, the next revision of AMD's Athlon processor, code-named Thunderbird, won't be widely available in anything but a socketed layout, due at least in part to an incompatability with Via's KX133 chipset. Even the folks at ZD have picked up on the story by now, but we've seen some interesting developments over the weekend.

If you'll recall, last time we were left pondering how Kyle at the HardOCP had gotten a hold of purported T-bird benchmarks using a KX133 chipset. I mean, the two won't tango, right? Well, now Asus has released a BIOS update for their KX133-based K7V mobo that claims to support the Slot A T-bird. So it appears the KX133 and T-bird will indeed work together.

How well the two work together, however, is the big question. The blokes at the Register weighed in with their own report on the situation. It seems the KX133 and T-bird don't always play well together thanks to engineering problems:

It is believed the problems stem from longer traces on the Slot A cartridge which lead to the timing inconsistency. A similar problem with trace lengths led to Intel reducing the number of Rambus RIMMs supported on its Vancouver VC820 mobo late last year.
Which raises the question: who is to blame for this problem? AMD or Via?

This post on JC's BBS may very well be worth listening to; the poster claims AMD had to take the PR bullet on this one, even though it wasn't their fault:

No, they just don't want to get sued for saying that via is incompetent. KX is barely functional with k75, won't work with TB. This has nothing to due with slot/socket, etc. Look at the huge problems via had getting agp 4x to work. ev6 is a lot harder than agp.
Having wasted part of my weekend watching a KX133 board fail in various horrible ways with a perfectly good Athlon 800 processor, I'm inclined to believe it.

Meanwhile, the HardOCP's secret source for T-bird benches has his own theory:

What we think is that AMD simply wants to move from Slot A to Socket A as quickly as possible ...

Also, keep in mind that AMD has stopped producing the Athlon ...

That means , after the stock is sold out , there will be no more SlotA processors in the retail market ...

That means , SlotA will be dead very quickly ...

Indeed. Not only would Slot A owners be locked out of the T-bird game; they wouldn't have access to any new "classic" Athlon 800s, either.

A solution might be coming, however, from those crazy slotket makers, PowerLeap. The Nerd pointed me to a post on the PowerLeap forums. Could a Socket A-to-Slot A adapter work?

Hi, Chris,

You're at the right place and time to ask this question. AMD & VIA's comments are the same " It's impossible ".

Well, Martin & I are working on this project, we've just done the Beta circuitry design this morning, it's under PCB Layouting now, however, it's too early to predict if it can be done.

One thing is for sure : If it works, just like the Neo-S370 PPGA to FC-PGA converter, we'll be the world's first and possibly the only source of such Slot-A upgrade. It's pretty difficult to design the Socket-462 converter, too many uncertain pin signals.

Wish we luck, we'll keep you guys posted.

Paul at PowerLeap

Looks like they're going to give it a shot. If they succeed, current Athlon owners may have an upgade uption after all.

So what really happened here? I'm not sure, but I'll venture a guess. I don't think AMD's recent moves are solely motivated by cost cutting, but don't underestimate the power of the profit motive. The reported problems with the T-bird and the KX133 chipset are probably very real. However, I doubt they're insurmountable. At the very least, a new revision of the KX133 could be adapter to support the T-bird. Heck, that's pretty much what Via's KZ133 chipset will be, right?

Via's KX133 problems probably converged rather neatly with AMD's own interests. After all, why risk harming the Athlon's reputation with these sorts of problems when the socketed Athlons will be cheaper to produce?

Tip: You can use the A/Z keys to walk threads.
View options

This discussion is now closed.