The reviews up so far on the 2.8GHz PIV seem to indicate that one can achieve roughly 3.2GHz without too much effort.
Before We Go Any Further
There will be very little difference between the relative performances of the C1 stepping PIVs and the revision B TBreds overclocked as opposed to unoverclocked. The ratio between the two running at spec 2.13/2.8 is virtually identical to the probable overclocking ratio of 2.4/3.2.
So if the benchmarks you see unoverclocked favor the Athlon, the overclocked speeds very likely will too, and vice versa.
Increases in FSB will be a wash since overclockers will be running at over 166MHz FSB anyway. Dual DDR will give Intel a small but not negligible boost over the Athlon that will be unanswered until the cache of the Athlon gets increased to 512Kb in Barton.
There is no single obvious choice between the two; what programs you personally use should be the main criteria, followed by your personal preferences, prejudices, ideologies, or, in very extreme cases, which side of your head you got dropped on as a child. 🙂
Onto the C1 Stepping
The 2.8GHz chip is one of the first representatives of the C1 stepping for the PIV, which we told you about a while back.
You want one but don’t want to pay $400-500 for it? That’s what this article is about.
These aren’t available yet. When they will be is our next topic.
A Three-Step Intro
The chips just introduced by Intel will all be C1 stepping chips. They are:
2.5 GHz (100MHz), NOT the 2.53
2.6 GHz (100MHz)
2.66 GHz (133MHz)
2.80 GHz (133MHz)
Outside of the 2.5GHz, these will all be over $400 at inception. The 2.5GHz should be more like $250, but expecting to get it to 3.33GHz at 133MHz looks to be a bit too much without extreme cooling: probably easy with a Peltier/vapor evaporator, very iffy with water, very doubtful with air.
The new 400MHz chips should be avoided overclocking them will be an awkward affair. The likely overclock won’t permit either synchronous FSB/PCI ratios or provide for ideal asynchronous memory use at current available ratios (you really would need a 3:5 ratio to do that).
For instance, let’s presume we have a 2.5GHz processor, which we can get 3.2GHz from. That gives you a 128GHz FSB, which means you either attempt to run with a PCI divisor of /3 (which will likely fail), a /4 (which is a small underclock), or you lock the PCI speed at 33MHz and lose any benefit from a moderate PCI overclock.
Per memory, current asynchronous ratios will leave you with a slower memory speed than if you did so at 133MHz+. A 3:5 ratio would be better, but you’d still be better off running at 150Mhz with a 3:4 or 4:5 ratio.
A much better case will be able to be made for the slower 400MHz CPUs when they go C1, but as you’ll see, that won’t be for a while.
Step two will be the 2.40B and 2.53GHz CPUs. Intel stated their earliest available date would be the end of August. That doesn’t mean you can order one at the end of August.
What Intel does when they have new steppings is to clear out their old inventory first. This usually takes from one to two months; it has taken longer.
Unless the website flat out says that they are selling you a C1 stepping chip; it would be foolish to order one blind the next few weeks. You almost certainly won’t get one, and if you send it back, you’ll usually get charged a restocking fee, after all, you ordered a X, they sent you a X.
When this situation occurs, I personally go to computer shows and verify the CPU I buy is whatever it says, but most people don’t want to do that.
If you don’t want to make the effort, it probably would be best to wait until we hear about a source that is selling them.
The remainder of the PIV will go to C1 at very earliest the beginning of October. Realistically, that will be November, and if the rumors about inventories of PIVs building up are true, it may be a good deal longer than that.
Whenever it does happen, the killer Intel bargain chip will be 1.8A SL6LA. That should be able to do 3GHz easily for about $130.
An Example of Intel Hang Time
Intel is not going from the original Northwood stepping to the C1 stepping. There was an step inbetween, what they called a “B-0 optical shrink.” See the link towards the beginning of this article for a further discussion of those chips.
The earliest release date on those was supposed to be July 22, and they haven’t shown up in the channels yet. XBit Labs reported a few days ago that they’re on their way; but this gives you an idea of the wait between “earliest” delivery and “actual” delivery.
Since we don’t know how well the optically shrunk ones will overclock compared to what we now know the C1s can do, it would be best to plan to wait for the C1s unless we get evidence to the contrary. There are a few structural improvements in the C1 which should help performance a couple percent.
Other Reasons To Wait A Bit Longer
There’s been a number of report stating that since the high-speed PIVs need higher wattages than are called for in the specifications for current PIV mobos, mobos have to be redesigned to accommodate them.
While terms like “incompatibility” are too strong, nonetheless, there is the chance that a current mobo may prove unstable if you’re cranking hard at 3.2GHz or so.
In any case, if you’re out to build the ideal Northwood system, it would be very short-sighted to settle for single DDR when you should be able to get a dual DDR board fairly shortly, which will certainly have any mobo fixes.
If you can wait until 2003, there’s always Hammer to consider.
If you can’t wait quite that long, perhaps the prospect of further prices cuts might help.
It’s possible we’ll see dueling price cuts over the next ten days or so. More likely, we’ll see unofficial ones over the month of September, especially on AMD’s side.
One possibility is that Intel may push the 3.06GHz processor onto the stage September 1. A couple non-American websites have offered the product. Should that happen, we may see the price of the 2.66 GHz processor go down to the $250 level, with smaller changes for the lower rated processors.
That would mean getting a C1 stepping chip at a somewhat reasonable price without hassle right away (though getting it to 3.33GHz at 166MHz is by no means a gimme).
If sales don’t go well in September and October, expect to see “cut, then tell” price cuts from AMD mid-October anticipating a likely Intel cut October 25.
Look at it this way. Any CPU price cuts will help to finance the additional cost of a dual DDR board and two smaller DDR sticks over one bigger one.
Based on the initial finding, Intel isn’t leaving itself a lot of room for the 3.06GHz. This is hardly going to break O/Cers hearts, but you should note that the first few people with these chips may be able to post high WCPUID numbers, but are reporting lower stable numbers.
We are nearing the end of the Northwood product cycle with these chips. While 3GHz should now be easy with C1 stepping, 4GHz without extreme measure won’t.