FSB800: first impression and a look into the future

FSB800: first impression and a look into the future

Granite Bay?

While the Asus P4PE with i845PE was able to address the processor and RAM with 200 MHz without any problems, we had our Granite Bay (E7205) mainboard unfortunately not so lucky. So it was 'already' over with a memory clock of 185 MHz. But that wasn't the only problem. In addition, in the current BIOS version 1002, the bus timings in the background when the frontside bus is raised above 133 MHz is significantly reduced. Even with a theoretical memory and frontside bus throughput of 5.9 GB/s, the system could not benefit from any performance advantages. The bus timings, which cannot be set in almost every BIOS, were simply too bad. Fortunately, Sandra's memory benchmark is completely unimpressed. Therefore, at this point, a little preview of what we could expect with the upcoming dual-channel chipsets.

Sandra memory throughput
  • Int Buffered:
    • P4 2.4 GHz FSB800 2xDDR400 *
    • P4 2.4 GHz FSB533 RIMM4200
    • P4 2.4 GHz FSB800 DDR400
    • P4 2.4 GHz FSB533DDR355
    • P4 2.4 GHz FSB400 RIMM3200
  • Performance gain:
    • P4 2.4 GHz FSB800 2xDDR400 *
    • P4 2, 4 GHz FSB533 RIMM4200
    • P4 2.4 GHz FSB800 DDR400
    • P4 2.4 GHz FSB533 DDR355
    • P4 2.4 GHz FSB400 RIMM3200

The dual-channel chipset sets itself at 4.4 GB/s clear to the top before Rambus. Compared to the original system, the performance has been increased by over 60 percent. With FSB800 and single-channel DDR400 it was only 10 percent, but in the end this was converted into 4-5 percent more performance. With a bit of luck, we can really expect 20-30 percent more performance from Intel's upcoming chipsets. And all that without turning the CPU clock screw up again!


However, it seems obvious that in this way a good 20-30% increase in performance is possible despite the identical CPU clock.

We were astonishedIn addition, the behavior of the two boards used. On the one hand, that the P4PE accepted the FSB800 without any problems. On the other hand, the Granite Bay Board seems to massively defuse the timing from an FSB of 140. Often nothing remains of the record-breaking 4.4GB memory bandwidth in the applications - in most cases there is even a drop in performance. We will try to run similar tests on boards from other manufacturers in order to find out whether it is a bios-dependent setting or perhaps a uniform chipset variable. It could well be that Intel has 'trimmed' the Granite Bay with regard to this feature in order to be able to sell corresponding chipsets better next year. We do not know this.

Be that as it may. No matter which motherboard will unleash the performance of the FSB800 in the next year: CPUs based on this memory connection in combination with Hyper-Threading will make life really difficult for the possible counterpart AMD Hammer.

This article was interesting , helpful or both? The editorial team is happy about any support from ComputerBase Pro and deactivated ad blockers. More about Ads on ComputerBase .