The Inquirer-Home

All Athlon64s will be dual channel soon

Mists lift from AMD's roadmap
Thu Dec 04 2003, 08:02
AS OFTEN HAPPENS when I write an explanation that is incomplete or confusing, someone who knows corrects me in a hurry.

Yesterday's roadmap story left some questions unanswered. But brave souls who know have slapped me down. Now I've managed to get to my feet again, I can tell you that the 939 pin mists are starting to clear.

Let's start out with the obvious problem, the difference between the FX and the vanilla Athlon64. That one is easy to explain, it's the difference between cache and cash.

Newcastle is the 512K cache version of the K8 core, which costs much less cash than the 1MB cache FX. Basically, twice the cache costs around twice the cash, so the differences are more to do with dollars than transistors.

This still leaves a puzzle about the difference between single channel Athlon64s and the dual channel FXs. That's easy because all will be dual channel chips in about two months. All A64s will soon have dual memory channels, so you won't have to worry about buying a single channel CPU with a dual channel mobo, or the other way around.

So, where does this leave the numbering scheme also known as PR ratings? When the FX goes to 939 pins and loses the registration mandate on the DIMMs, it gains a few per cent in speed. And the FX line is not speed rated. An FX-53 will be an FX-53 with or without a few percentage points gain from latency.

Vanilla A64s are much trickier beasts, with hard numerical ratings. How many points would you lose from going from 1MB to 512K cache? How much would you gain from going to dual channel memory from single?

The answer is about the same. A 939 pin A64 with 512K will end up about the same PR-wise as a 754 pin 1MB.

The only question remaining is what the heck happens to the 754 pin socket models listed on the roadmaps as Athlon XPs. Are Paris, and the later Palermo taking a trip on the short bus ride to the K8 school? And who is going to buy these things anyway? HP? µ


Share this:

blog comments powered by Disqus
Subscribe to INQ newsletters

Sign up for INQbot – a weekly roundup of the best from the INQ

INQ Poll

Heartbleed bug discovered in OpenSSL

Have you reacted to Heartbleed?