Click Here for previous A Guru's World columns.
Hurrah! Hurrah! AGP 8X is in the works and I will run right out and buy one as soon as I can.
It seems that the marketing gurus at Intel and NVIDIA are on their toes with this "breakthrough" in technology. Just when ALi makes available a DDR compatible chipset (the ALiMAGiK1 for the Athlon and the Aladdin Pro 5 for the Socket 370 platform) that will actually allow us to finally appreciate AGP 4X for what it is, if and when anyone actually gets around to writing software for all of the extra bandwidth we will have available, even at the risk of alienating all prior technologies, Intel and NVIDIA drop the bomb shell that implies that that just isn't fast enough.
Now don't get me wrong. Faster is better and moving forward quickly is a major plus for this industry. But look at the track record AGP has. AGP was great for the 66 MHz bus. What a great development to have you video card capable of shuttling information at the same bus speed as the rest of the system, instead of PCI's typical 33 MHz bus. PCI had to die as a means of interfacing a video card as soon as it's time begun. Even Vesa Local Bus was invented to get the bus speed of a video card up around the neighborhood of the system bus, but as bus speeds went from 33 to 40, from 40 to 50 and then 50 to a steady and long used 66 MHz, it was time for AGP.
Then came Intel's Deschuttes (the 100 MHz bus Pentium II) and AMD's Super Socket 7. System bus speeds started to "normalize" at 100 MHz. Looks like it's time to raise the bus speed of the video card as well. Enter AGP 2X. For some of us, AGP 2X was the first time we had ever heard of the principle of a "double data rate". Information is being sent on the rise and the fall of the 66 MHz signal, thus effectively DOUBLING it's bandwidth. Essentially we end up with an effective 133 MHz AGP bus that easily accommodates the day's 100 MHz front side bus PCs.
AGP 4X came out. A little premature, but Intel Coppermines with 133 MHz FSBs and Cyrix Joshuas with 133 MHz FSBs were coming out and we would also have chipsets with asynchronous bus capabilities that allowed us to run 133 RAM with 100 MHz FSB CPUs. So the time for the next technology to be introduced was not too early. Better to have the technology ready for when we need it, then too late. So what exactly is AGP 4X? Well, somehow they managed to produce FOUR TRANSFERS per clock cycle. Phenomenal. At least it looked good on paper. The bus speed for AGP 4X was equivalent to a 266 MHz FSB far exceeding the bus speed of any PC made… yet.
Now, ALI has come out with the first chipset to support the DDR SDRAM standard. Using the same technology as the Alpha and Athlon's EV6 bus, and yes, the same technology as AGP 2X, the RAM now is capable of two transfers of data per clock cycle, but unlike AGP, but very much like EV6, the bus speed of DDR is 100 (using PC1600) and 133 (using PC2100 RAM) and not just 66 MHz. This gives us clock speeds equivalent to 200 and 266 MHz on the memory bus!
For an Athlon zealot like me, this sounded like a PC in perfect harmony. Think about it. A 100 MHz CPU using a DDR, soon to become a 133 MHz CPU using a DDR. Pair this with an AGP slot using a 66 MHz FSB using QDR (quadruple data rate, I guess) and a motherboard capable of DDR SDRAM!
I've always used the same PC for the last 5 years, only replacing one or two or three parts at a time as technology dictated. Time for me to rebuild myself a new PC from scratch and not look back for a long, long time.
But then… AGP 8X! Bastards!
So now, they're saying that they can write eight transfers to a single clock cycle, thus giving us a bus speed equivalent to 532 MHz just for video. Like I said, it's not a BAD thing, but slow down for a minute JUST TO LET ME CATCH UP. For god's sake, if my Athlon wasn't overclocked, the AGP 8X bandwidth would be MORE than the clock speed of my CPU! If they waited until the ALi DDR chipset actually showed up on some boards before announcing 8X AGP, that'd be fine. But, Noooooooooooooooo!