Ouch, those poor Celerons

First things first, yes it has been entirely too long without a blog update. You know how one thing always leads to another and you end up putting something off for way too long? That's where updating the blog got and I apologize for it, that's something I'm hoping not to do again. I'm quite hard on myself so sticking to updating it on a consistent basis shouldn't be a difficult thing.

With that out of the way, there is a lot that has happened over the past two months that I've got to update you on - some hardware related and some not so hardware related.

Happy belated Thanksgiving to those who celebrate it, and for those who don't, Happy belated Thankful-to-be-alive-and-well day. My girlfriend came down from Connecticut (she's a 1L at Yale law...now do the questions about CT make sense? More on her later...) and we cooked our first Thanksgiving meal for my family, her family and the family of a mutual friend of ours (we call him Manny). The fact that neither one of us had ever cooked a turkey resulted in quite a bit of skepticism but in the end, cooking for 12 people turned out quite well. The process wasn't really all that difficult, just very time consuming and very tiring - I have a newfound respect for my mom who used to cook Thanksgiving dinner for us all the time mostly by herself, it's a tough job. My biggest issue with the cooking of the turkey (aside from the whole stuffing process, which as one beautiful lady put it - "is just bizarre") is the fact that if you're cooking in a covered pan, you really have no idea what's going on with the turkey for entirely too long. The light button on my oven is my friend and he lets me check on all that's going on in the world of the oven, but lighting up a closed lid pan does nothing but worry me about what the turkey is doing on the inside. For all I know it invited its cousin the chicken and nephew the steak and they're all sipping gravy instead of cooking. The meal was great though and I slept like a rock that night.

The logical progression from turkey is to talk about...Prescott. Many of you caught a little easter egg I left in one of our Video Card Roundups, where if you highlighted the right area and stood on one foot you'd reveal that our test bed had a 2.8GHz Prescott. Yes, we picked up a Prescott in Taiwan earlier this year and have been playing with it internally ever since. Why have you not seen any real benchmarks of it? Well, it seems as if Intel is tired of the leaks and has taken some serious measures to prevent important internal information from hitting the web. It also appears that the Prescotts that are being sent out, aside from having no higher than a 14x multiplier, have an artificially slower L2 cache than Northwood. As far as I know the latency on Prescott's 1MB L2 cache is not supposed to be any greater than on Northwood's, but these initial Prescott samples do show a higher L2 access latency. If Prescott were still 130nm then I would believe an increase in L2 latency, but the fact that Intel has moved to 90nm and are using a more efficient cache architecture for the L2 in Prescott leads me to believe that latencies should be more similar than they are. The increased L2 latency is why current benchmarks of Prescott that are on the web show it being slower than the Northwood in some cases. I've been trying to get my hands on something a bit more representative of the final core but so far I haven't been successful.

As far as how you should expect Prescott to perform, just look at it this way - it'll be slower than the Extreme Edition on a clock for clock basis and it should be faster than Northwood. So while that will keep Prescott on top of the Athlon 64 in encoding and 3D rendering applications, gamers will still want to go with an Athlon 64. The true potential of Prescott will be exposed through exploting the 90nm process to reach higher clock speeds and with the new instructions that will eventually be utilized in programs, but today's benefits will be limited.

Since we're talking about CPUs, I had a meeting with AMD at the last Comdex I'll probably attend (sheesh it was horrible). It looks like you can expect no more than two speed bumps for the Athlon 64 FX in 2004, the FX53 and FX55 (2.4GHz and 2.6GHz respectively). If we are assuming that AMD would not want to release a faster non-FX, you can expect to see the fastest Athlon 64 in 2004 run at 2.6GHz (Athlon 64 3800+). Intel is planning on being very close to 4GHz by the end of 2004, so in 12 months we'll see how well Prescott can scale and how well AMD's on-die memory controller lets the Athlon 64 scale with clock speed. AMD may end up being held back a bit by the fact that they're not going to support any higher bandwidth memory architectures other than DDR in 2004, although there is talk of doing a higher-than-DDR400 memory controller for the 90nm Athlon 64 processors.

The 90nm shrink for the Athlon 64 will bring it nothing more than a smaller die, so don't expect a larger cache or any new features. In fact, the Athlon 64 as you know it today will be phased out in the first half of 2004. AMD will stop producing the 1MB L2 parts early next year and you'll only be able to get 512KB L2 parts in an attempt to make the Athlon 64 more affordable. It's difficult to say what the negative impact of a 512KB L2 cache will make on the Athlon 64, but given the on-die memory controller it might not end up being all that bad. The introduction of the 512KB core (codenamed Newcastle) will give enthusiasts more of a reason to purchase the Athlon 64 FX, which will retain its 1MB L2 cache and should now offer a more tangible performance advantage over the Athlon 64.

The Athlon 64 will continue to be a locked processor, while the Athlon 64 FX will not. AMD is very concerned about remarking so they have employed an Intel-like system to lock their Athlon 64s. It seems that they have not disabled PowerNow however, so you can trick the processor into running at a lower multiplier if you'd like with some software tricks - there is no way to increase the clock multiplier though.

If you haven't already checked out our budget CPU roundup, I would strongly suggest you do. We haven't looked at the new Celerons since the 1.7GHz parts first hit the streets, but it seems like the latest batch are just a pure disappointment. Take a look at the review and see for yourself.

I'm working on storage stuff right now as well as training a possible new peripheral editor (you'll see the first of her trial articles this weekend with any luck). I've got more to say but a meeting in 30 minutes so that'll be all for now.

Take care :)