Slashdot Mirror
Athlon Xp 3200+ 400FSB is Coming
SoDaLaS writes "Athlon 3200+ Coming:
According to CNET The Athlon 3200+ with a 400MHz FSB is on the way in the next two weeks. It'll be interesting to see how well the processor overclocks at that high of a bus speed...it didn't seem to hamper the new 800MHz FSB Pentium 4, which many people were worried about too."
The Athlon chips have not been super-sensitive to changes in FSB. The performance impact of the Athlon XP moving from 2x133MHz to 2x166MHz was significantly less than the P4's gains going from 4x100MHz to 4x133MHz. The P4 gains have been incredible with the jump to 4x200MHz.
It seems that AMD is trying to squeeze every bit of performance out of an architecture that would be better squeezed by being optimized, i.e. Opteron. It's a shame that AMD's yields of Opteron have proven to be dismal, but if I was a motherboard manufacturer I'd be pretty mad at AMD right now. More motherboard manufacturers are going to have to qualify their boards and more chipset manufacturers will have to qualify their products as well, even if they can already meet 400MHz operation. Will the performance impact really justify the costs that all parties incur by moving to yet a new FSB in less than, what, 6 months?
Unless mankind redesigns itself
With faster processors we get less efficient programs. 10 Years ago you could do internet/email/word processing/spread sheets with just a 33MHz Intel 386 with 16 MB RAM. Today you need for the very same things a Pentium IV with 2 GHZ and 128 MB RAM. There are some niece applications which need a lots of CPU Power like Quake or Nurmerical Simulations, but must Joe Adverage apps don't really need it. The programs need it due to sloppy coding. And the faster CPUs gave rise to the OOP paradigm. While it primarily is a nice theoretical concept for safer and more secure program, it's used these days just for code-bloat and GUI overload. Inpedendent studies show that in fact 73 percent of all "OOP" code is just imperative with C++ class bloat added.
Further the higher compiler and debugger speeds introduced much more sloppy coding styles. In the 60/70ies the computers of the Apollo program hadn't a single computer crash, which is completely unthinkable these days. The reason why the NASA is keeping old 8080 Intels in their shuttles is that they won't get decent code quality form modern processors these days.
Personally I think that the whole CS community must rethink their position towards computers speeds. Instead of the todays faster-is-better point we need a paradigm change towards just-as-fast-as-necessary.
Owner of a Mensa membership card.
I work from home, and have a network of 5 PCs of various specifications. You can tell when they are on because the floorboards in the hall vibrate, never mind the noise they make.
I have recently invested in a VIA EPIA-M10000 motherboard. It is very, very quiet.
Sure, it isn't as fast as the latest P4 or Athlon, but it plays DVD (with hardware support), DivX, and MP3 media without any problems. Quake 3 runs well.
More importantly, I can run all my business applications without any noticable loss in speed.
I'm going to ditch my other boxes and buy some more of these EPIA systems. It's the quiet life for me.
Any fool can talk, but it takes a wise man to listen.
What's the real clock speed of this beggar?
Purchasing the right part WILL get you a massive overclock though. The prime example is the Athlon XP 2100+ Thoroughbred B part. This CPU is actually running at 1730Mhz, however many overclockers are getting these CPU's up to 2500Mhz on stock air-cooling and 100% stable. It's unknown why these particular parts are doing such a high overclock, but anyway you slice it a 44% overclock is pretty significant.
Also, my own CPU which runs at 1830Mhz and overclocks to 2200Mhz is still getting a 20% overclock.
I try to make everyone's day a little more surreal.