nForce2+ and a Tidbit

That’s all well and good, and if you have an older Athlon system and want to get your last socket A upgrade in soon, this will be worth waiting a couple weeks for.

What’s more interesting in the longer run, though, is a tidbit about Hammer they snuck into the review:

“At the very last moment AMD postponed the launch of their Athlon 64 processors from spring 2003 to fall 2003. This relatively strange (at first glance) decision was made because AMD had faced some technological problems. Although the yields for AMD Athlon 64 were quite high, the top clock frequency of the CPUs hardly exceeded 2GHz, so that the performance of the new solutions couldn’t exceed that of the currently available Athlon XP processors. To increase the core frequency AMD is now working on a new Athlon 64 core revision, while Athlon XP Barton will have to defend AMD’s current position in the processor market.”

Simple confirmation of what we’ve been saying all along. We’re not seeing Athlon64s because AMD can’t get competitive ones out.

Another Indirect Indicator

How much power do these Opterons chew up?

The AMD techdoc for the Opteron does not provide any breakdown as to what wattage each of the Opterons chews up. It only states (pages 1 and 75) that the Opteron chews up 80 watts max, and has a maximum thermal power of 84.7 watts. That not a rumor or speculation, that’s what AMD says.

In comparison, the power of a Barton running at 1833MHz is typical 53.7W, maximum 68.3W.(page 21).

For a TBred B at 1800MHz, the power consumption at typical 57W, maximum 62.8W (page 23).

There’s some ambiguity about the Opteron’s wattage, and at least in theory, it could be a lot busier processor in multi-processor mode than a single CPU, so let’s give AMD the benefit of the doubt and entertain other estimates.

If you go by the Ace’s Hardware estimates, you get fifty-something watts. Let’s give AMD every benefit of the doubt.

Hmmm. This still comes in around the same as the wattages for “typical use” for the Athlons.

Where’s all this power saving SOI is supposed to deliver? Just last September, people from AMD were writing about getting huge frequency increases/power decreases out of SOI chips. Read it yourself.

More conservative estimates talked about a 30% reduction in power at a given speed.

So where is it?

No, it’s not due to the additional cache. As the Ace’s Hardware article points out, the amount of electricity cache chews up is negligible.

As we pointed out a few days ago, the voltages on the current Opterons seem awfully high.

If the Opteron was meant to run at 1800MHz at, say, 1.35V rather than it’s current 1.55V, that would account for a bit over a 30% power difference.

Don’t take the numbers literally; these are back-of-the-envelope calculations, but they do indicate there’s something wrong with the SOI sauce at the moment.

And at least part of the problem is that the chips are chewing up too much power for their speed.

Email Ed