Do we still have the best cpu?

Question

hello,
with pentium 4's hyper threading out, is athlon still the best solution?  i mean, pentium 4's prices are already dropping...they're already catching up with athlon.  and athlon is not catching up with the speed.
thanks,
edison

Answer 1

Quote
Originally posted by AlexM
Hi Edison,
Actually your cousin was making a legitimate point being made here w.r.t. Athlons being 'overclocked' (although he probably didn't know it    :D).
This doesn't mean that the processor is operating at higher clock speed than rated, but rather that AMD is having to push their process to get good bin splits for the higher clocked models.
In the same way that an overclocker will sometimes have to nudge voltage upwards to get stability at overclocked speeds, the Thoroughbred requires a 10% boost from 1.5v to 1.65 volts to increase speed by 400mhz. This isn't an overclock as such because this is within the process specification, but in a sense AMD is having to eat it's future to stay semi-competitive with Intel in terms of clock speeds. The datasheet specifies a max voltage of Vcore +.5v (i.e. 30%), so there is still plenty of overclocking 'headroom'.
I'm not sure what the corresponding voltages are for the PIV, but I know it is less clock-for-clock when comparing chips fabbed at 90nm, which implies a process advantage for Intel (but wait for SOI!).
Cheers,
Alex
Well, Intel's done the same thing on numerous occasions--upped the voltages for MHz speed bumps when people expected them to stay the same or in some cases reduce.  Most egregiously it was done with the PIII, but several production runs of the P4 are included. Most notorious of the lot is Intel's HYPEr-Threading scenario (lets pretend we have two cpu cores so we can get the one we actually have to operate more efficiently) which, in the HT-enabled 3.06GHz version (which is no great skates in yields yet), can displace up to 105W. So AMD is nowhere near being unique in this category.
In fact, with each succeeding generation of Athlon AMD has refined its processes far better than anyone looking from the Intel side of the fence imagined was possible, and so a lot of Athlon's "heat" reputation is more of a hangover from the earlier days than it is accurate in comparsion with current Intel processes--in which the 3.06Ghz HT cpu is stretched as far as Intel can make it go under their current .13 micron processes. (EG, that's the reason Intel waited for 3GHz before implementing working HYPEr-Threading circuitry--they knew about the increased power and thermal requirements and weren't prepared to put those anywhere else in their chip offerings for yield considerations.)
Yes, AMD has stretched the longevity of its current design through process improvement--but Intel has certainly done no less.  I consider that in both cases this has zero correlation to "overclocking" since all such chips made by Intel and AMD are validated rigorously at their released MHz speeds. Overclocking by definition is operating a chip at over it's validated MHz speed--something that AMD, again, has never done.
As far as "eating up their future" is concerened, I think it's about a zero concern at this point, since AMD is on the eve of transisitioning to a new architecuture.  As such it really matters little what they do with AXP/Barton in that regard, since in a couple of months those designs will move to an economic position displacing Duron, and the Hammer designs will take the place of the current AXPs in the AMD chip lineup.  And of course we reset all counters to zero in terms of MHz and processes when that happens, at least in the mid-to-upper markets where Hammer will thrive.  In displacing Duron, AMD's current  AXP/Barton line will have no trouble moving Celeron completely out of its current market position (which is why Intel is designing Banias [sp?] with such gusto at the moment--a chip design which, ironically, places more emphasis on IPC than raw MHz ramping.  I love these kinds of ironies... Who *says* Intel can't learn a trick or two from AMD?  *chuckle*)