Future Chip News

Discussion in 'General Hardware' started by Taggert_LOA, Dec 15, 2004.

  1. Taggert_LOA

    Taggert_LOA Modding Addict Political User

    Shavings From The Rumour Mill

    So now we’ve seen the future and it is as dual-core as you can get. Since the last column, a lot more information has seeped out into the mill pond just outside the rumour mill. It’s pretty clear that a lot of future Intel technology depends on Microsoft Longhorn, and it’s also now apparent that the three chip giant’s Smithfield Prescotts due next year are really stop gap chips. This is incredibly frustrating if you’re on the verge of buying or putting together a top machine; do you wait until 2006 before you put your money down in the hope that Microsoft will have launched Longhorn by then? And if it makes you cross, just imagine how livid the two major chip firms are.

    There are already signs that the home and enthusiast market is damping down in terms of sales; what will sustain Intel are sales of PCs to big businesses. All the signs are that people are finally switching after a long period of hunting through their pockets for a spare quid or two.

    The three Smithfield Prescotts are to be called the X20, X30, and X40. We don’t know what happened to the X10, nor does there seem to be any rhyme or reason for the names and numbers. But we do know that each of the dual processors will have its own stash of 1MB cache. In another leak from Intel, it emerged over the last few weeks that the company feels the cache will give it an edge over AMD’s dual-core processors. We do expect AMD’s chips to be out well before Intel’s, however. No one is saying yet what the performance is going to be like, but as both firms are now de-emphasizing the megahertz myth, we suspect that raw power will be the least important part of the puzzle. Nevertheless, we have megahertz numbers for the X series: They will run at 2.8GHz, 3GHz, and 3.2GHz, respectively. It’s no wonder that Intel is playing down the frequencies—just now Intel released the last of the old Pentium 4 family (the 3.8GHz with 1MB of cache; reviewed on page 22). And when it introduces the 6XX Pentium 4s in early January, in Japan at least, they will have 2MB of level 2 cache each and won’t hold out much in the way of additional megahertz numbers.

    While we all wait to see just how Intel’s dual-core chips pan out, and much of the infrastructure such as chipset support is already there, the strong rumour is that there are some exotic dual-core chips promised for 2007 which will really capitalize on Longhorn. We can’t wait. Or rather, we’ll just have to wait. They’ve got exotic code names, but we’d better not mention them yet; otherwise Intel might dispatch a mole hunter to Leak Central and shoot our source stone-dead.

    Just before writing this, I got some information from an AMD mole in Taiwan who tells me that next generation Opterons will have a whole different pin out, and by then just about every AMD processor will be dual core—we think that applies to Athlon 64s, as well. The server chips will have 1,207 pins, so that’s an end to the 939/940 pin out scene. But we won’t see these processors until 2006 when Longhorn is expected. The last we heard, Microsoft is preparing for a launch of Windows 64 for AMD and Intel processors at the end of the first quarter. That takes us to within maybe a year and some months before Longhorn.

    Bill Gates told me and some other people last year that Longhorn is costing Microsoft more to develop than the man on the moon project in the 1960s. It had better be good. What I’m personally wondering is where this leaves the real current oddity in the Intel pack: the Pentium 4 Extreme Edition processors. Intel will release another of these this year. By the time you read this, it will already have been introduced, barring accidents and emergencies. Is there any future for the EE chips at all? Judging from the reviews I’ve read, they’re not exceptional performers and they are very pricy indeed, with the top-end chip costing a staggering $1,000. You be the judge.
  2. dave holbon

    dave holbon Moderator

    London England
    I’ve also heard similar roumers, well not really roumers, more like facts. I think that the speed war as measured by processor speed is over; the focus will now be on system throughput and bandwidth. This just can’t be a CPU problem but a system wide problem. Supporting chipsets which allow more that two processes to operate concurrently and each CPU to use/borrow memory from another is where all the money is being invested and I’m not talking about North & South bridge designs which are a cheap way to implement motherboards anyway.

    The motherboard design is about to become more important than the CPU or even memory modules, its here that the real speed increases will be made in my view.

    Once into the multi-processor area using 64 bit multi-processor designed boards (which will initially start at twice as expensive as current high-end boards) hit the mainstream, CPU MHz speed will not be an issue and system throughput will be the yardstick and hence overall system performance. Even using today’s components expected increases of about a factor of ten could be achievable. The downside is that the motherboard will cost you about three times the current price even in mass production.

    Well you can't always win!

    I take the view that its best not to upgrade for two years if your current system is based on an Intel 3Gig processor or more.

    :) :)
  3. Steevo

    Steevo Spammer representing. Political User Folding Team

    /puts it on the scales

    HT still sucks.
    Hopefully the game writers will have dual processor or HT capable games out by 06' but there will be alot of people pissed when HL2 still performes better on a old P4 or Athalon than their new high end rig. Perhaps compatability with older software to allow full useage....

    Then again the whole problem of compatability needs to be done away with. Shoot it and bury it. Bloated software, slow hardware. In 2010 we will probably still have "X86" compatable HW and SW.