Chipsets

It was recently made public that Intel's upcoming chipsets - Grantsdale and Alderwood - would be called i915 and i925X respectively. We talked about those chipsets back at CES but what we didn't talk about were 915G, 915GV and 910GL. We originally thought 925X would launch only for Socket 775, and 915P for both. However, even though the roadmap is slightly vague it appears 915P will effectively replace 865PE; it will have the same performance of 875P with some extra features: PCI Express x1 and x16, Azalia Audio (Intel's AC'97 replacement), DDR2 and of course the new Southbridge which we will talk more about in a little bit.

Even though 915P will support 533MHz DDR2, we will most likely see DDR2 support showing up only on 925X motherboards. Most manufacturers we talked to during CES did not see enough differentiation between 865PE/875P, and this actually hurt sales in the long run. Even with 925X and 915P, we once again expect a PAT-like interface on 925X. Those in the industry say Intel has also made it a much more difficult enable "PAT" on Grantsdale. Other than "PAT", the only other real difference we can see is support for ECC memory on 925X. Most manufacturers we spoke with with planned to use current DDR on 915P boards, reserving DDR2 for 925X.

The new ICH6 southbridge will make its appearance with the 925/915 and makes some interesting promises. First, in any ICH6 denoted with a "/R" suffix, we will see 4 SATA interfaces whereas ICH5 only had two. As kind of a bonus, ICH6/W will provide integrated wireless. ASUS's P4P800S-E turned out to be a surprising success due to its on board wireless capabilities. Maybe Intel has learned a thing or two from its component manufacturers.

915G has all the features of 915P, but with integrated graphics. While we have no details of the graphics capabilities of the new 915G, it is reasonable to expect higher performance but nothing too impressive based on Intel's prior track record.. Intel anticipates this to be the corporate chipset of choice, and will follow up on the 915G with the 915GV and 915GL chipsets in Q3. 915GV and 915GL are basically just stripped down versions of 915G.

The roadmap coverage continues, we hope you enjoyed this one as well...

The New Performance Socket - LGA-775
Comments Locked

20 Comments

View All Comments

  • The12pAc - Monday, February 2, 2004 - link

    Umm.. yah acctually the pipe-line is 31 stages #18. which would make it slower, even with a 1 mb L2, and a 2MB L3 and a 6 MB L4 and a 25MB L5 100MB L6 and a 150GB L7 weene!!

    pAc
  • Stlr22 - Thursday, January 29, 2004 - link

    Ummmm.........yeah
  • Pumpkinierre - Saturday, January 24, 2004 - link

    #15 I doubt that Intel would release a new cpu that is weaker than the old one. The initial sckt 478 Prescott release doesnt have much of a speed increase which you would expect if there were a pipeline lenght increase. So my bet is that prescott will be substantially more powerful than Northwood and may even show up AMD's stupid PR rating scheme yet again.
  • Reflex - Friday, January 23, 2004 - link

    Well, I felt I needed to clarify my point since you were so quick to write it off as a fanboy response. I do believe you are a bit hard on the platform problem of AMD vs. Intel, Intel's platforms are not rock solid like they were in the BX days anymore, and arguably the AMD platform offers a better guarantee(the largest issue with any chipset is the memory controller, and that is no longer a factor on an A64 platform).

    However, it is personal preference overall, neither product is a clear cut, blow the competition out of the water performance leader. But, in general given any amount of money you can build a superior AMD system, even if only by a few % points in various benchmarks. Also, the future looks brighter for the A64 platform as 64bit will be arriving this year, and that will show serious performance increases in several areas(including the weak points of media encoding and compression where it will get a massive boost from 64bit).

    If someone buys a P4 today are they making a horrible mistake? No, they are not. Are they making the best informed decision however? It depends. I tend to say they are not, especially looking down the road, however I would not say they are a moron or something either. To some people the name on a product means more than the product itself, for those people we have companies like Intel, Sony, Volkswagon, etc.
  • TrogdorJW - Thursday, January 22, 2004 - link

    The point is, Reflex, that I made some comments specifically on the article in question, and then a couple people bounced on my post with the typical "OMG AMD is so much better" stuff. (Yours wasn't as bad as MAME's. "Totally dominates and OC...." In realistic systems, it's a moot point, as either CPU is fast enough. And the 3000+ does not OC very well at all.)

    The initial post of mine is perfectly reasonable, and it wasn't made as an Intel fanboy, so it should be responded to in kind. Is AMD a better choice right now? That's NOT the question at hand! The question is, how does the Intel roadmap look?

    The "new" Celerons may OC very well and may atually be able to compete against the Duron/XP. *MAY*. As I said initially, we need to wait for the product to be released to find out. You reiterated that point.

    One final point is that while there is a lot going on in the chipset and CPU world right now, my take is that P4 and A64 is still a close race. The A64 is generally a little faster, but reliability, upgradability, etc. is up in the air. At least we all know that the P4 socket-478 line is going to end at 3.4 GHz. So you've got a case of trading the beast that you know and understand in for a new and unknown beast. Should you do that? There is no correct answer to that question; it's personal preference.
  • Corsairpro - Thursday, January 22, 2004 - link

    Any word on how long the prescott pipeline will actually be? I've heard 22, 30, and 32 stages (listed in order of feasibility). And just how much will this reduce the prescott performace say in comparison to a NW at same clockspeed?
  • Reflex - Thursday, January 22, 2004 - link

    TrogdorJW: Its not fanboyism at this point to point out that at the moment, AMD pretty much has Intel beaten on every front. The Opteron is a much better server CPU than the Xeon, the AthlonFX is ahead of the P4EE, the A64 is ahead of the P4, and the AthlonXP and even the Duron line trounces the Celerons(and even the P4 in many cases). And thats before factoring in the fact that these chips generally cost considerably less than their Intel competitors.

    The largest argument against an AMD64 system right now is the uncertainty around the platform. And that *is* a valid point. Thats why I stated that once socket 939 showed up the A64 would be pretty much the clear cut leader. Now, if Prescott arrives and blows the doors off of what AMD's top offerings are then I will revise my opinion, but at the current time I see no reason to do so. The competing platforms both have only a small amount of headroom(both the Intel and AMD platforms will change by the end of the year for their top end CPU's), and in general the AMD offerings are better performers than the Intel equivilents, and the price is lower on an AMD system, it becomes *very* difficult to justify purchasing Intel AT THIS TIME.

    Now, in a couple months when Prescott arrives(actually arrives, not paper launches) and we see whether or not AMD can scale quickly and when 939 is on the table and all thier CPU's standardize on it, then we will be able to re-evaluate the situation. But for now we can only see the cards that are actually on the table, so anything else is speculation.

    As for 64bit, as far as I am concerned its a bonus. The A64 is already the fastest 32bit CPU on the market, 64bit OS and apps will only make it better. If they arrive this year, great, but if not you still were not cheated seeing as you got the best bang for the buck possible with the apps on the market currently.
  • TrogdorJW - Thursday, January 22, 2004 - link

    Yes, the Athlon 64 is fast. That does not mean it automatically has to be used in every system. Get a grip, #5 and #6. All I said is that it will be interesting to see what happens with the next Celerons. If they cost $70 and OC to 3.5 GHz or more, they could be very interesting. We won't know for sure until they're released.

    Technically speaking, you can get an 865PE motherboard and a 2.8C Pentium 4 for about $93 and $212, respectively. It's a mature platform, and the 2.8C STILL overclocks better than the Athlon 64. Or you can get a K8T800 motherboard and an Athlon 3000+ for $128 and $227, respectively. (That's using an MSI motherboard in both systems, with retail CPUs from Newegg.) Needless to say, socket 754 is not as mature as 478. So almost $50 more, and 64-bit software isn't going to become important for at least another 6 months, possibly longer. (Unless you run Linux? Didn't think so....)

    As for the statement, "Not only that, [Athlon 64 3000+] totally dominates the p4 in everything except the encoding department, in which case it's pretty much even," that's not entirely true. It dominates in games ***provided you have a 9800 Pro or XT and run without AA/AF***. Otherwise, it's relatively close. With a 9700 Pro or less, neither system is going to be much faster than the other.
  • Icewind - Wednesday, January 21, 2004 - link

    Yes, but thats PRESCOTT, NOT the current P4c model.

    Two different things, so its good they corrected it.
  • Oxonium - Wednesday, January 21, 2004 - link

    #9, if you read the article it says that there WILL be a 533 MHz FSB Prescott and several other sites like Xbit have been saying for weeks that there will be a 2.8 GHz Prescott with 533 MHz FSB but with no HT support. In any case, the authors have fixed the table and it now shows a Prescott 2.8 GHz with 533 MHz FSB.

    I'm sitting pretty with my Athlon64 3200+. While there will likely be faster CPUs from both AMD and Intel this year, I don't really see a reason to upgrade from that chip for quite a while.

Log in

Don't have an account? Sign up now