The Athlon 64 3000+ has been released, and it is available for as low as $240.
Should you buy one?
The answer to that question is another question. Do you want an expensive quick fling, or do you want a (fairly) long-term relationship.
A Quick Fling
This Athlon 64 is going to be obsoleted fairly quickly. AMD looks like it plans to move away from socket 754 to socket 939 for mainstream systems in 2004, leaving socket 754 for its value line.
This is probably because the Athlon 64 platform is a paradox: a high-performance engine with small fuel lines. If you want a performance system, especially for gaming, you need dual-channel memory. If you don’t have it, that’s a performance anchor that will become heavier and heavier as time goes on.
If performance isn’t that important to you, or you do things that don’t really require memory bandwidth, the Athlon 64 is overkill (at least given the price for the processor).
Perhaps more importantly, a generational shift in motherboards is imminent. This will consist of the following:
Intel will implement the first three with its next-generation chipset, and complete the transition at the end of 2004 with its “Bigwater” motherboard for Tejas.
AMD looks like it will take smaller steps. The next generation of Hammer mobos will implement the first two changes. Memory will probably follow sometime in 2005, and we’ll just have to see about form factor.
Buy an Athlon 64 now, and you miss out on all of this. On the other hand, missing out on all of this means you may be able to recycle more of your current system (i.e., memory and video card), thus reducing your upgrade cost.
Then again, too, an overclocked Athlon 64 is also likely to generally do a little bit better than the average overclocked PIV system. Depending on how Prescott does and when socket 939 shows up, you’ll have a couple to a few more months of glory. But that’s it.
The (Fairly) Long-Term Relationship
Socket 939 boards should be out and plentiful within the next six months. That will give you dual-channel memory, and will also give you any performance tweaks those building a desktop board from scratch can muster over server-oriented socket 940 boards. (We strongly suspect there will be a bigger performance gap between socket 939 and 754 than currently exists between sockets 940/754.)
Along with them ought to come Newcastle chips. The smaller cache will reduce performance, perhaps 5% for non-office use, but that loss should be more than compensated by the dual-channel memory/mobo tweaks.
More importantly, it looks very likely at this point that socket 939 boards will accommodate future 90nm CPUs with just a BIOS change, giving socket 939 buyers more of an upgrade option more quickly than with socket 754 (where 90nm chips will be delayed, and may or may not include AMD64).
AMD’s decision to hold off on DDR2 until 2005, the introduction of a (presumably cheaper) Newcastle, and the delay in 90nm processors shifts the optimal time to make a platform shift from socket A from later to sooner. Those who need to get more than a year’s mileage from a new platform should probably buy a socket 939 Newcastle system fairly early (prices and platform stability permitting), figure a 90nm processor upgrade somewhere down the road, then tentatively plan the next jump to be the K9.
Those more patient can of course wait, but we may not see an affordable 90nm Hammer until early 2005, and that will be just too long to wait for many.
Newcastle systems will cost a bit more than socket 754 systems, but the cost item that will get most people’s attention is a PCI Express video card. We’ll probably see hybrid systems and workarounds that will continue to allow the use of an AGP card, but on the whole, at this point in time, if you want a hot video card, it probably would be better to spend your money on the next big jump.
I doubt that initially PCI Express cards will be any better than AGP cards, but that’s not why you would be buying the card. You’re buying the card that will work in your machine without any potentially-perfomance reducing kludges to get it to work.
Another factor that may or may not come into play is the ability to change the multiplier upward on a socket 939 chip. You can with an FX. You can’t with an Athlon 64. We don’t know if you’ll be able to that with an FX Newcastle, but when it’s “don’t know” vs. “definitely no,” you have to lean to the first, since you’ll be no worse off no matter what.
Pimping An Orphan
I would bet that if the thoughts and feelings of this audience could be distilled into a single voice, that voice would say, “There’s nothing to do!”
If one were to do a personality profile of the bulk of the hardware audience: it would be a teenager-young man, impatient and impulsive, and very susceptible to any ego-boosting pitch.
They haven’t had any fun lately, and are looking for some.
Marketing to such a group is easy. You just give them hints that they’ll be really smart/important/powerful/cool if they would just buy your product, and not too surprisingly given the bair, many then “decide” that they’ll be really important/powerful/cool by buying the product.
These are the same people who read this piece and then tell me, “Don’t tell me what to do, let me decide,” then proceed to buy the first product that makes them feel smart/important/powerful/cool enough. 🙁
I call it the “Kiss their asses then pick their pockets” approach to marketing. You take advantage of your potential buying audience’s vulnerabilities, then pander to their emotional wishes. Subtly, though. It’s sort of like cake mix, you let the buyer take the last step or two. That way, it becomes his idea, and he’ll angrily deny that he was manipulated into thinking that way.
To some, this is great marketing. To us, it approaches child abuse, and the fact that it is so common as to be almost universal doesn’t make it right.
What does this all have to do with Athlon 64s? This is a product soon to be orphaned from the performance market. The product that is going to orphan it won’t be around for a few months yet.
So what do you do in the meantime? You lower the price a bit and pitch it to impatient people who don’t look ahead. You may call that smart, and certainly the retailers and many
retail assistants reviewers will agree with you.
I call it being played like a deck of cards.
Is AMD being especially evil about this compared to others? No, but that doesn’t make it good, or make it something you should do.
We just don’t think going socket 754 at this point is a terribly good idea or value for most people in this audience. After all, the company that makes the product implicitly agrees with that by phasing it out of the mainstream in its roadmaps. You may think we hate AMD, but does AMD hate AMD, too?
At the very least, we think that anyone who is seriously considering buying one ought to do so with eyes WIDE open. There are certainly a few categories of people in this audience who can accept the drawbacks to buying this. The point is to make all aware of those drawbacks.
Nor do we hate Hammer. We just haven’t liked any of the price/performance packages offered so far (and frankly, given the lack of talk in forums from people with them, neither have you). If it comes in at the right price, we think a socket 939 Newcastle is going to be the first decent buying opportunity for the average person reading this, or at least one you won’t be kicking yourself for three-six months later.
We’re not against you buying a Hammer. We just want you to buy the right one.