Yes, for your typical office PC, 3.06ghz is more than adequate. That's current available tech so if intel and amd both stopped making faster chips today that market wouldn't be affected for a long time.
But I didn't mean that I think current CPUs are more than adequate in a more general sense. Quite the contrary. Like that mpeg example, even as fast as CPUs are today video encoding takes time. Just nearing realtime for some encoding methods, well past realtime for others. But when I encoded my first mp3 it took 5 hours. Now it takes, literally, about 5 seconds.
At the time when it would take me 5 hours to encode a mp3 there were already DirectTV sat receivers covered with dust from sitting on top of TVs. MPEG2 decoding on the fly at a level that PCs didn't see for some time after when the Hollywood DVD decoder boards came along. Now we can watch DVDs on our computers without them stuttering without specialized decoders, the CPUs are fast enough. But even that took till what, P3-500 and faster? With DivX at high res with filters maxed a CPU of over a ghz is needed. This even after MMX has become standard, MMX2, SSE, SSE2, etc.
So decoding we're ok for as long as your CPU is fairly new. But what about encoding on the fly? Oh sure, you can, at pitiful quality. How come we can't get the same quality on the fly with a 2700+ and a dumb caputure device as a much slower CPU with a mpeg encoder caputure device can do? (argue here whether we can or not today, might be close)
Because the chips on that mpeg2 card just know how to do one thing. They can't run MS Word. They're of no use in the business style computer. But what about a CPU specificly tuned towards media encoding? Is there a market? Is it technically feasible?
In the graphics card arena we see speeds coming up. But more important than the speed increases to game play are the changes to architecture. Pixel shaders, vertex shaders, T&L units, all that stuff I just see talked about and don't really care to read what it means. Games vary a lot from one to another, but under some set of standards (DX/OGL) vastly different hardware (ATI/nVidia/intel/Trident/etc) manages to do the same thing, but at different performance levels. Some have better fillrate, which benefits one type of gaming. Some have something else that benefits another.
We've seen the attempts made with CPUs to balance. Back to the MMX and other instruction sets, the goal is to make the CPU perform better in multimedia tasks without having to make the entire unit faster. Add an instruction set to a 200mhz CPU and make it perform at the same level it'd take that same CPU 800mhz to attain without it. Well, I'm saying that approach is a smart one. Look at the Pentium-M 1.6ghz, it's Content Creation score was way below a P4-1.6, but in the General Usage benchmark it tops a P4-2.66, and in gaming it scores extremely well. And that was comparing notebooks to a coal burning full bore desktop.
That's what I'm looking at. P4 is a wonderful CPU and it's architecture will allow it to keep going for quite a bit more. But when I see a low power chip of half the mhz doing some tasks even faster I'm impressed. When I see that same chip do other tasks half as fast I'm not dissapointed in the least. I'm hoping it sparks more new designs, new ways of thinking about the tasks and solutions to the CPU role.
Facebook
Twitter