Tying Some Tidbits Together

Add Your Comments

There’s a report on an updated AMD roadmap over here.

Let’s hope this is an old and/or garbled one, because it certainly doesn’t reflect the optimism we’ve been reporting about 90nm processes.

There’s no faster Athlon 64 processor the first half of 2005. A few slightly quicker Semprons will show up, but that doesn’t really count. In the third quarter, the FX-55 will almost certainly turn into the Athlon 4200+, and we’ll presumably see a 2.8GHz FX-57.

This is none-too-aggressive.

There’s a fairly mysterious Athlon 64 3700+ which is supposed to show up in the second quarter. Best guess is that it is a spruced-up 2.2GHz chip with 1Mb cache. It could be instead a chip AMD thinks some tweaking and/or SSE3 is worth 200 extra PR points.

To the average person reading this, the importance of the 3700+ is not its existence, but rather WHEN it comes into existence. The earliest we can probably expect 90nm 2.0 chips is in the second quarter, 2005. It is possible that it might take until 3Q, when AMD will actually introduce faster processors than what’s available today, for these to become available.

DDR-2 Starts Growing Up

XBit Labs does some testing which reveals that DDR2 isn’t worse than DDR anymore when it comes to things like latency. Can’t really say it’s better, either,

That’s good to know, but what is most instructional about the article are the benchmarks here and here, which show what little difference memory actually makes in real-life use.

Of course, most reading this will not care less about DDR2 until AMD does (if even then). Many out there desperately want to know when that might be, but AMD won’t tell them.

There is absolutely no competitive reason why AMD would want to keep this secret. After all, Intel already uses DDR2, and they certainly know what it does and doesn’t do for performance.

Given AMD’s behavior over the years, the most likely explanation for the silence surrounding DDR2 is that they are having some kind of problem with the onboard controller for DDR2 to work right yet, for one reason or another. Intel has had some problems with running 266MHz standard on some of its DDR2 mobos, so this is understandable.

The real piece of evidence supporting this is the silence from mobo makers about any AMD DDR2 mobos. They’re less tight-lipped about what they’re going to come out with in the next six-to-nine months than AMD, but look at this handy accumulated roadmap website, and there’s not a word about such a beast.

Why? There’s a ton of potential reasons, but if I had to bet on one, it might be as simple as this being a low-priority item compared to getting 90nm SOI running, and running fast.

The less likely explanation is that the controller works, but AMD can’t figure out when to introduce new mobos with new sockets for it.

Say Yes, Say No, Say Something!

If you looked at the benchmarks linked above, you’ll see that DDR2 is no big deal when it comes to performance. AMD could certainly get through 2005 without it.

From a planning standpoint, it probably would be best not to introduce DDR2 until desktop dual-cores are ready in 2006. That gives the mobo makers a year to sell their current socket systems and get Hammer established. After that, then you can change the sockets all you like to accommodate the changes.

AMD ought to say, “No DDR2 until at least 2006,” soon. They can make up any excuse they like as to why, just give a period when it definitely won’t be around. That may rub against the geek grain of never admitting to bad news, but as you see, this really isn’t bad news, and for many “bad” news is far better than no news.

If people knew that DDR-only Hammer systems were going to be around a year, in all liklihood, some would wait a year, but more would probably buy now or soon. Not knowing at all just makes people concerned about this just wait.

I grant you the average Best Buy computer buyer isn’t writhing with angst over this issue, but more than a few of those who tell that Best Buy computer buyer what to buy are.

This is how Intel normally handles these kinds of things. When they decided to go to LGA775, they told the world about a year ahead of time. No doubt some didn’t like that news, but it’s a whole lot better knowing a year in advance rather than a month.

Why can’t AMD be that way?

Besides, giving that kind of advance notice to your customers shows them respect. It shows that you treat customers seriously by giving them the information they need to buy wisely, and that you have a plan and don’t run the company by the seat of your pants.

If you want respect, you must give it, and this is an easy way to do so.

Ed

Leave a Reply

Your email address will not be published. Required fields are marked *