Bad News for 975X

Several weeks ago we briefly touched on the new Intel 975X chipset, a revision of 955X with a PCIe lane configuration that would work with either SLI or Crossfire. However, after talking with manufacturers in Taiwan, we hear that this new chipset will not support NVIDIA's SLI. Furthermore, the original November launch schedule has been abandoned in favor of a launch very close to Intel's Napa (Centrino) platform.

975X will feature two PEG slots in a dual 8 lane configuration, more than ample bandwidth to run two NVIDIA 7800GTX or two ATI X800XT cards in tandem. Originally, we expected the 975X launch date only a few weeks out; manufacturers have working samples they were able to show us in Taiwan this week. Unfortunately for Intel, this launch has been pushed back to January. One product manager mentioned to us that the reason for the delay was actually the immaturity of [ATI and Intel] drivers rather than the silicon.

Virtually all manufacturers we visited this week mentioned that they will have a high end, 975X solution in their roadmaps. Particularly interesting was ECS's plans to put the new Yonah socket on a 975X platform, making it the only option for dual graphics and Yonah. Gigabyte and MSI also had new 975X motherboards to show, both with interesting cooling features that will show up on new high end boards from both companies.

Yonah Tidbits

Intel's next generation Centrino, dubbed Napa, launches early January, 2006. The core component of Napa is Intel's 65nm Yonah processor, a Dothan successor. As we mentioned a few days ago, Yonah will have a new product naming scheme compared to Dothan and Banias, but it will also use an electrically incompatible 479/478 pin socket. Thus, existing 915GM and 855GME motherboards will not work with the new processor.

Fortunately, many manufacturers are already working on two different chipsets to succeed the existing mobile on desktop (MOD) motherboards. The first of these, 945GT, is nearly an identical revision to Intel's 945G, but will feature the new Yonah specific socket. As with other Intel CPUs, the 945 and 955 North Bridges are required to enable both cores. 945GT will show up for several small form factor and HTPC PCs, but vendors tell us there are no full scale ATX motherboard designs in the works.

Intel's 945GM is a follow up chip to 945GT, but will feature SO-DIMM DDR2. 945GM will be used mainly for laptops and ultra portables, but we will also see set-top DVRs based on the 945GM as well due to the profile advantages.

Other interesting Yonah tidbits we've picked up over the last few days include:

  • There will be single core and Celeron versions of Yonah - but Celeron M won't show up until H2'06.
  • 945GM and 945GT will cost about the same as 955X does on the chipset level.
  • Centrino will have a slightly newer logo (due apparently to legal counsel).
  • Most Yonah models will feature VT, but the ultra low voltage and low voltage designs geared for ultra portables will have it disabled.
  • Merom (Yonah's successor) will feature 64-bit extensions, but Yonah will not.

The upcoming Yonah Celeron M "delays" aren't too surprising. Current Celeron M chips are already very competitive with equivalently clocked Pentium M chips; the difference in performance between 1MB and 2MB of L2 cache really isn't that large. Rather than devaluate their high-end, high-cost Pentium M chips, Intel is simply refusing to release faster versions of the Celeron M. The fastest current Celeron M is clocked at just 1.6 GHz, with a 1.7 GHz version launching in January '06. Of course, while Intel has a couple 65nm fabrication plants getting ready, they'll still have 90nm plants that won't be needed as much for desktop CPUs once the new 65nm chips launch. There's no sense in rushing the conversion of Celeron M to 65nm if Intel isn't going to increase the clock speed much anyway.

Comments Locked

12 Comments

View All Comments

  • rinthos - Tuesday, October 18, 2005 - link

    Nay, Celeron M will not last longer than Pentium-M on battery.

    Why? Pentium-M supports Speedstep. Celeron M does not.

    Therefore, Pentium-M will use more power when at max-performance, but during 'normal use', will use less power (due to speedstep lowering clock etc).

    Thus you get better battery-life with a Pentium-M :)

    The only scenario would be if you are running the systems at max capacity the entire time, then celeron M (or a slower Pentium-M) would naturally get better battery life.
    ---
  • JarredWalton - Tuesday, October 18, 2005 - link

    Of course, even at max power the Celeron M (and Pentium M) don't use that much. 20W, give or take, and cutting to 10W with EIST helps but isn't going to double your battery life. The screen is the biggest power draw these days, I think.

Log in

Don't have an account? Sign up now