Silicon Investor (SI) -- The First Internet Community

STOCKTALK

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.

Technology Stocks : Advanced Micro Devices - Moderated (AMD) -- Ignore unavailable to you. Want to Upgrade?

To: Road Walker who wrote (57137)	10/4/2001 9:08:34 AM
From: fyodor_	Read Replies (2) \| Respond to of 275872

John: Just trying to get a straight answer. Is there, or is there not, a standard for measuring MHz? There is a standard way of measuring frequency, yes. There is not, however, a standard way of rating computer chips in terms of frequency. It used to be very simple. The whole chip ran at one single frequency. This frequency was supplied by inducing oscillations in a crystal on the motherboard¹, which was then passed on to the CPU. Typically, all the busses (ISA, in the x86 world at the time) ran at this frequency as well. Around the time of the 50MHz chips (486), it was found that the busses really couldn't keep up and motherboard manufacturing was becoming prohibitively expensive, so the "DX2" was invented. It was cheaper and more reliable to have the busses ("FSB" running 25MHz, CPU running twice that) and performance only suffered "slightly" (actually, a 50MHz 486DX was about as fast a 66MHz 486DX2 - and faster in graphics, since the graphics bus was essentially overclocked). Still, aside for some simple (but still space consuming) circuitry to handle the CPU <-> motherboard bus, the whole CPU ran at a single frequency and this frequency was then specified as the MHz-rating of the chip. You'll note that CPU designations originally contained information about the clock-doubling or tripling or quadrupling that took place - by way of the DX2 / DX3 / DX4 designations - this information has since been dropped. Today, things are even more complex, with processors running internally at a wide variety of frequencies. There is, of course, still the circuitry necessary for the operation of the CPU <-> North Bridge bus, but there are other "anomalies" as well. Take the VIA C3 processor (IIRC): The floating point unit actually only runs at half the "MHz-rating". So while the chip claims to be a 500MHz chip, the FPU actually only runs 250MHz. This is certainly noticeable in floating point benchmarks! With the P4, a small part of the chip actually runs at twice the "MHz-rating". (There have been some claims on this board that other parts actually run at half the "MHz-rating", but I don't know enough about the internal workings of the P4 to be able to claim that with any level of confidence). Regardless, there is no standard describing what the "MHz-rating" of the CPU should be, but it has always been fairly straight-forward, since the vast majority of transistors were running the same frequency (and this frequency was then used). In the future, however, I think we will see this trend breaking down even more than is the case with the P4 and C3 processors. Why should the optimal solution be to have all parts of the processor run at the same frequency? -fyo ¹ Slight simplification here. The most common type of crystal in early PCs oscillated at a frequency of 14.318MHz, which was then immediately divided by 3 to get the 4.77MHz clock of early CPUs (+ISA bus). A further division by 4 yielded the 1.19MHz of the system clock (time). Furthermore, it should be noted that the ISA spec only required the ISA bus clock to be betw. (IIRC) 8 and 12MHz.