Shelton: Covington Reincarnated?
Those of you who have been around a long time remember Intel’s Covington chip. It debuted March 1998, and was the initial Celeron chip.
It was most noted (and derided) in the mainstream media for having absolutely no L2 cache, which caused a major hit in performance. However, it could also overclock close to fifty percent and thus overclocked, approach the performance of far more expensive PIIs in at least some things.
Covington began the first fairly heavy flow of high-percentage, cheap overclocking, a flow that became a flood six months later when Intel introduced a Celeron with some cache (Mendocino) which could compete against PIIs in most things.
People often forget about Covington, but it looks like some folks at Intel haven’t.
The Inquirer reports that Intel is introducing a processor called Shelton in a few Asian countries which looks like the great-grandchild of Covington.
Speed? 1GHz. Made with 90nm technology. No cache.
Should you care about this?
Why You Should Care: Level One
In the short run, there’s a number of aspects about this chip which are rather interesting.
You have a 1GHz chip made using technology good for 3.5GHz. In the sport of overclocking (like basketball), you always look for the mismatch. For overclocking, it’s advanced technology, low clock speed.
There’s never been a potential overclocking mismatch like this before. If this were basketball, this would be like Shaq facing Mini-Me.
In theory, the Shelton is a chip that could triple its clock speed without breaking a sweat.
Granted, in practice, this is unlikely to be the case, if only because (assuming a multiplier lock) you’re just not going to be able to triple the FSB of any motherboard that can run this chip (at least not right away).
We know practically nothing about the chip; we don’t even know what platform it will use, but somehow I doubt it’s PIV-compatible; it’s probably a socket 370 chip.
Granted, even if you get this chip up to 3GHz, it probably won’t perform as well as a 2GHz PIV, so it’s not for performance-hounds.
However, that isn’t so much a factor for two other computing groups.
The Quiet Ones
Those who want, who need quiet ought to take a look at the power consumption figures for this chip when they become available. It’s possible that Intel’s current electrochemistry yields processors that leak a lot of power no matter how slowly they run, but if that’s not the case, we may be looking at a chip that can get by with just passive cooling, even with a bit of an overclock.
The Cheap Ones
How much can this chip possibly cost? The original Covington used a 250nm process technology and was 130 sq. mm. If this were simply a process shrink, a 90nm Covington would be less than 20 sq.mm, or about a quarter the size of a current Athlon Thorougbred XP chip.
Manufacturing costs per chip for such a chip would be less than $5, might be more like $3. This is potentially one cheap chip; Intel could sell this chip to OEMs for $10-15 in bulk.
Match it up with old, shrunk down socket 370 technology, and you have a platform that will make Sempron look expensive.
But It Will Suck!
Granted, much like great-granddaddy Covington, Shelton will run well behind any AMD designs. Even the Via chips could well end up giving it a beating, so it probably won’t be a good choice even for the groups mentioned above.
However, remember that Covington gave way to Mendocino within six months, and Mendocino showed that a little cache can go a long way in boosting performance. Adding 128K cache to a Shelton would add less than 5 sq. mm to the chip’s size; so it will still be tiny.
Such a chip would definitely give the Via Edens and Nehemiahs a run for their money, but what is far more intriguing is how somebody else responds.
What Does AMD Do?
A tiny, cheap Intel processor will give AMD fits. Why?
It’s not that they don’t have a design to match it. Call it Athlon XP, call it Sempron; that product line can certainly more than handle some shrunk-down Celeron on steroids.
The issue becomes “Can they afford it?”
If Intel pumps out these little wonders and sells them at $15-20 a pop, what does AMD do? It can’t match Intel’s production costs, certainly not at 130nm, so it won’t be able to match Intel’s prices and even hope to break-even.
They can reduce the problem by shrinking socket A to 90nm, but that may be a cure worse than the disease.
Racing To The Bottom
An ultra-cheap 90nm socket A chip would no doubt be very popular in the Third World. That’s not AMD’s problem.
AMD’s problem is that an ultra-cheap 90nm socket A chip would be very popular in the First World, too.
About a year ago, we asked this audience what they would prefer, 90nm XPs or Hammers, and a majority said XPs.
Yes, it was done a year ago, and the audience is probably more Hammer-friendly today. On the other hand, though, a $30 90nm XP could undo a whole lot of that, and not just for this audience, but for OEMs, too.
AMD is especially vulnerable to this because it’s not much of an exaggeration to describe its product mix as selling one CPU for $500 and ten for $50. The vast majority of units sold by AMD are cheap XPs, and will remain so for some time to come.
Yes, Hammer prices will get more reasonable fairly shortly, but how does it compete against (almost) free in the minds of the average computer buyer?
Right now, AMD can’t make enough money from socket As alone because the price is too low. They can’t make a lot of Hammers alone because the socket As offer far better bang for the buck. So they’re muddling along selling a few for a lot, and a lot for a little.
A hyper-cheap Celeron is a trap for AMD. If they match Intel, they get stuck in the low-price arena because their bang-for-the-buck problem intensifies. If they don’t match Intel, they forfeit both current and future volume.
A nasty push into the quicksand, isn’t it?
However, there’s one problem with this trap.
The Global Problem
Put simply, anybody in the world who wants a computer and has money already has one. The only people left are those people who want a computer and don’t have money.
Leaving out those who need to get running water and electricity first (an appreciable chunk of the world’s population), that still leaves a realistically potential audience rather bigger than the current one for whom price is the principal problem.
Manufacturing techniques and capacities have reached the point where good-enough CPUs and platforms can be delivered to such folks. That’s the only way CPU companies can continue to grow at the usual rates. They can’t wait for the have-littles to get rich enough, that will take quite a few years, even decades. The CPU companies have to come to them with much lower prices.
However, what’s good enough for the have-littles is also good enough for most of the have-lots. How do you keep the have-lots from finding out what the have-littles are getting, and demanding it, too?
Joe Sixpack can afford to pay $500-$1,000 for a computer, but he’d rather pay less for something he doesn’t much like, anyway.
Yes, Intel has crippled this Shelton so much that it’s unlikely to become a hit at the Best Buys of the world, but given the likely consequences of introducing such a chip; it’s hard to see how Intel can push AMD into the quicksand and stay out of the pit, too.
This wouldn’t be such a big deal if people needed more and more computer power, and the CPU manufacturers could easily provide more and more, but that’s hardly the picture nowadays, is it?
In The Long Run, China?
Intel and AMD may have a vested interest in not killing each other, but there’s another emerging player that couldn’t care less about their continued health and well-being: China.
Right now, AMD and Intel are falling all over themselves trying to get a foothold there, and I bet their Chinese partners are asking for and getting a lot of information on just how to make a CPU.
China still has a native CPU program, you know. It’s not terribly good at the moment, but something tells me that by hook or by crook, it’s going to get a lot better in the next few years.
Again, these efforts might be laughable in a world where the desktop computing boundaries get extended a lot every year, and people want to pay for it. In such a world, copiers are always behind the curve. But what if those days are over, and price rather than performance becomes the criteria?
And really, if that’s not what’s happening, then just why is Intel releasing this Shelton to begin with?
We have long thought that the CPU industry would eventually become a commodity industry, and that somebody else would eventually break up the Intel-AMD duopoly (at least as we know it). For us, the issue has never been “If,” but rather “Who.”
We’re just suggesting a new potential “Who.”