[vox-tech] processor shootout: new intel, amd offerings

[Bill Broadley]

> availability
> ============
> intel's chip is already on the market.  AMD's chip is not available.
> the media has coined AMD's release as a "paper release".

Sounds like FUD to me, lag of weeks from announce to parts in your hand is
normal for both.  intel just happened to release the p4-2.54 a bunch of
weeks before AMD released the XP2600.  So yes intel has a a bunch of
weeks head start

> heat
> ====
> reviewer said the new P4 runs pretty cool review said the new P4
> operating temp was at 34C.   ... oh my god, becky!  that's cooler than
> my athlon 1.4!

Umm, sounds like unscientific bull.  Fact is according to spec they are
just about identical, in the 60-75 watt expected power usage range.

> overclocking
> ============
> the new intel reached 3.15 reliably.

I consider that irrelevant, it depends on many more factors then the cpu,
including motherboard (both the brand, model, and actual sample), chipset
(brand, model, and actual sample), ram (brand, model and actual sample),
cooling fan (brand model and actual sample), heat sink compound (brand,
sample, skill, amount, lapping, etc), FSB (clockspeed, multiplier,
crystal quality, etc), etc. etc. etc.

The "glory" days of overclocking are IMO over, the margins are smaller,
gone are the days when a 300a celeron can run reliably at 1.5 times
the speed, not overclock the memory bus, not overclock the motherboard,
and be reliable for years of uptime.  Sure if you want to tinker
and go for say oh a bit 10-15% and give up reliability go for it, not really
an option for those who just want it to work.

> benchmarks
> ==========
> the new P4 outperformed the new athlon by 10% - 15% on each benchmark.
> the new athlon is about equivalent to a 2.53 P4.  reviewer points out
> that you don't want to skimp on the RAM for the P4.

Actually the benchmarks I've seen show the xp2600 winning most benchmarks
against the 2.53 Ghz p4.  This radically depends on the memory system
(rambus vs ddr), fsb (400 vs 500), and application.  Not to mention
details like chipsets, bios, etc.

In general the athlon ratings hold true, not that they aren't off by
a factor of 2 for a particular case.

Keep in mind the next athlon to come out is rumored to have a 333 Mhz
FSB, I expect the perf increase to be substantial since all the way back
to 1.0 Ghz or so the athlon has had the same memory bus (which has been
increasingly the bottleneck), increasing the FSB by 25% should help alot
across the board.

Of course later this year I expect another giant increase when the
hammer (clawhammer and sledgehammer) comes out.  Which fix my 3 biggest
beefs with athlons/p4's.  To few registers (upgraded to 16), 32 bit
pointers (upgraded to 64), and a totally lame stack based FPU (now
a 16 register FP).

Other fun stuff will be included as well (bigger cache, SSE2, other
tuning, better memory bus, awesome interconnect for up to 8 cpu's, etc).

> pricing
> =======
> pretty much what you'd expect.  the new P4 is $508, new athlon is $300.
> major drag.  i'm not sure $200 is worth a 15% gain in oomph.  but if i
> had a real job...   :)   when the 2600+ reaches the market, hopefully
> the 2.8 P4 will drop a bit.

Bleah, why get a 2.8 Ghz p4?  Don't forget the fastest p4's are rambus
which increases the price even more, oh and you need one of the fancier
dual channel motherboards (another few $100) to really take advantage.

The one I'm watching for is the nforce 2 + 333 Mhz fsb athlons, cheap,
everything on board, decent 3d graphics, dvi on the motherboard, network
on the motherboard, killer 128 bit wide memory interface @ 333 Mhz.
Perfet for a thin client to provide a state of the art linux environment,
run everything locally, and fit in a case the size of a large textbook.

-- 
Bill Broadley
Mathematics
UC Davis