We’re starting to get overclocking reports on Prescott.
Here’s one from some of the forum folks at Xtremesystem.org.
It’s a bit disconcerting that a Vapochill can only keep a Prescott to about room temperature.
The real hardcorers are doing better than that, but the numbers aren’t yet as good as what they got out of Northwoods.
This is an important point. Pretty much across the board, the Prescott results haven’t been as good as the later Northwood results.
ZDNet UK provides some very interesting total heat dissipation figures for various platform (AMD as well as Intel), and the heat figures for Prescott simply blow everybody away. Mind you, that’s at default.
Finally, HardOCP provides some numbers with a 2.8E using standard and water cooling.
More important in my book, though, are the comments made in the article about what all this heat does to the rest of the system:
“The IC7-MAX3 did do something that we have never seen before. If you have noticed, we have used square blocks of plastic foam to elevate our boards for years now while we are testing them. The IC7-MAX3 got so hot that it actually melted the plastic block into a liquid form.”
The one thing that bothers me most that we have yet to see and will not be seeing for a while is the toll that will be taken on these motherboards by having to deal with all of this heat. It was not uncommon for us to see capacitors at 235F. That is certainly enough to burn you quite quickly. How are those capacitors and other motherboard components going to react to being baked for days at a time…or better yet, months at a time. I have a gut feeling that we are going to see a lot of Prescott related failures on the current motherboards that are in the market even if they are “Prescott Ready.”
Are You Ready To Test To Destruction?
Given all the following, anybody who wants to seriously overclock a Prescott (i.e., anybody who has the words “4GHz” in their heads) had better be serious about the whole proceding.
If you think the Intel fan is good enough, you are certainly not serious.
If you think any air cooling is good enough, you are certainly not serious.
If you think the MOSFETs and capacitors on your motherboard can take care of themselves without serious cooling, you are certainly not serious.
Actually, there’s a simpler test to determine whether or not you’re serious enough. Are you ready and willing to lose it? CPU and/or mobo?
If the answer is “No,” you have no business buying this stuff.
This is a particular problem with Intel overclockers because many have gotten spoiled. Many are used to buying an Intel CPU with the standard fan and just changing the FSB and voltage numbers in the BIOS to get a 30%, 40%, 50% overclock.
It’s not going to happen like that here. We’re going from risk-free to risky. More importantly, it’s not just a matter where the worst that can happen is people fall short of their goals. Like the HardOCP article said, we’re talking about a good chance of a lot of people breaking a lot of stuff.
What makes it even worse is that for right now, the results for which you’re taking these risks aren’t likely to be even as good as they’d be for the “old” Northwood you’re likely replacing.
Overclocking Prescott is a risky proposition. There are three ways to handle it. Two are smart, one is dumb.
The first way is to say, “OK, this is risky, but I’m willing to take the chance. However, I’d better do everything I can to shift the odds in my favor by getting some serious cooling into my system, and not only to the CPU, but to all these power circuits.” That’s smart because you accept the reality of the situation and are taking the steps necessary to meet the challenge.
The second way is to say, “Since it doesn’t look like I’m going to gain much if anything by it, I’m not going to bother. It’s not worth the risk nor the extra effort.” That’s smart because you’ve accepted the reality of the situation and decided it isn’t for you.
The third way is to say, “It’s not going to happen to me.” This is dumb. D-U-M, dum!! That’s denying reality in favor of fantasies. If the only way a person learns anything is to have reality punch them in the face (aka “experience”), that person is either a few brewskis short of a six-pack, or (more likely) just a kid (no matter his or her age) who still believes in a different version of the Tooth Fairy. Either way, that person is going to be one bruised individual until he or she smarten up.
The dumb thing isn’t taking risks. Show me a winner, and I’ll show you someone who took calculated risks and prepared for them.
The dumb thing is thinking there are no risks when there are, or thinking that risks apply to everyone but you. That’s the sign of a loser.
The point is not to insult, but to tell people (perfectly anonymously, no one will or even can ever know) how to stop losing and be winners.
A Little Light
Within the next ten days, we will know what Intel plans to do with 64-bit on the desktop when they describe it at the Intel Development Forum.
This is something everyone ought to wait to know about before they buy anything new.
What the “CT” technology probably will be is some form of x86-64. Intel will say that it is incompatible with AMD64, but in all likelihood, the “incompatibility” won’t be anything programmers can’t work their way around.
However, it is not inconceivable that Intel might be talking about CPUs that have some elements of IA-64 technology inside them: call it IA-64 Jr. I wouldn’t bet on it, but don’t be shocked if that’s the case, or even if they have both.
Frankly, I don’t think Intel could get an IA-64/32 hybrid out there in a year’s time.
It is the less likely outcome that would have the most impact on the average member of this audience. If Intel really is going to try to bring IA-64 to the desktop by 2005, that ought to make even AMDers hesitate about buying Hammer in 2004, especially if Intel tosses in x86-64, too.
If Intel just essentially says, “Me, too” on x86-64, though, I see no reason why AMDers ought to hesitate about buying Hammers down the road. Intel’s version may well have a few extra goodies a la SSE2 or PNI, but AMDers have lived without those kind of extra goodies in the past with little harm. 90nm Hammers are supposed to be out about nine months before Tejas, and the time to buy will probably depend far more on what AMD does with its prices than what Intel will do with Tejas.
A “Me, too” on x86-64 will have a lot more impact on Intel lovers. That will make it almost impossible to justify the average purchase of Prescott for anybody outside of someone with a PIII or worse (and if they need it right away, I wouldn’t even say them: Northwoods would probably be better).
Ironically, should that happen, what will become more important in the Intel field will not be what current systems will or won’t do, but rather how fully Tejas-compatible anything someone might buy in 2004 will be. I wouldn’t expect to see such equipment available until at earliest sometime in the second half of the year, probably late in the second half.