RE: [TCLUG:9472] bios

["Schoenhofen, Kelly" <KSchoenhofen@Carlson.com>]

-----Original Message-----
From: Scott Dier - dieman [mailto:dieman@ringworld.org]
<snip>
Hate to kick your thought down, but perhaps you should check out
tomshardware.com  the Anthlon kicked the 733's ass.
<snip>
--------------------------

That's it - you're toast :)

First of all, you've got to quit restricting yourself to one reviewer. 
Second of all, Tomshardware.com has really been slanted in the past towards
the company that's paying him the most thru advertising on his site. I
always take his reviews with the world's largest grain of salt, and this
review is no exception.

The hardware setup is flaky at best - the benchmarks he actually ran gave
the Athlon a nice Athlon optimized mobo, and the P3-E (coppermine) got a
Via133 mobo. In fact, in various benchmarks a Coppermine in an off-the-shelf
440BX mobo outscored the Coppermine in a Via133 mobo. That's one of the
biggest stumbling blocks in his benchmarks, regardless of his attempts to
tell you otherwise. The other issue I have with his setup is that every
benchmark but a couple direct timedemo gaming benchmarks, he used a TNT2,
not a T&L video card. Which was my entire focus on my earlier post. You put
t&l in both, watch the benchmarks change.  But we'll get to that later.

I take umbrage with two things on his review. 
Regarding FPU performance of the p3-e: "Athlon is still way ahead of its
pursuers from Intel. Coppermine hasn't got the slightest chance." 
He does a *real good job analyzing why Coppermine stunk so bad on FPU in his
rendering benchmark (sarcasm). To be honest though, the biggest reason has
nothing to do with his testing procedures, it's the 256k on-die L2. Intel
made a conscious decision on this one - they chose on-die L2 rather then
much larger off-die L2, for the much better all-around performance it gives.
For a straight, lights-out box in a rendering farm, mhz for mhz, in a choice
between a p3-e and a Athlon, you'd want the Athlon. This is the same reason
a P-Pro 200 would whip an early P2-233 at rendering, despite 3+ years of
engineering difference.
Once we see real i820 boards, with rev'd chipset driver sets, I wonder if
FPU performance will change - there are so many kinks in the pipe, as it
were, you really can't gauge pure anything anymore like you used to be able
to (more so). I'm just tired of the blanket statements he makes when he
tries to prove a point. He's a paid reviewer, I'm not.

The biggest thing I have a problem with is the one benchmark he totally
disses - the Q3 bench. The Coppermine outperforms the equivalent Athlon by
10fps across the board. In fact, it takes a 900mhz Athlon to catch the
750mhz Coppermine, in _his own_ benchmarks, skewed as they are. 
I LOVE his explanations: "Coppermine scores a respectable 10% more in office
applications and for some reason in Q3Test as well. I fully understand the
gain in Sysmark98, but the only explanation I've got for the high
Q3Test-scores is GeForce's driver. I wouldn't be too surprised if NVIDIA
already used an early version of Intel's upcoming compiler for their
drivers. It could be Id's software too though." 

And once again: "I mentioned my surprise already before, Coppermine scores
incredibly well under Q3Test. This is the one benchmark where Athlon is
looking bad, but I wonder if this hasn't got something to do with GeForce's
driver. I am sure that Athlon could look a lot better with some software
optimizing as well."

We'd do well to analyze this abit. He runs Q3 v1.08 at 640x480 at 16bit
color, normal graphics. 
First off, Q3 1.08 is NOT optimized for the Geforce in any way, shape or
form. It's actually deoptimized. Q3 1.08 uses less than half of the card's
features. Q3 deliberately does not allow the t&l card to do any l, lighting
that is. 
Second off, he runs it in a low res. I'm not asking for 1600x1200
benchmarks, but 1024x768 would have been nice. Make each machines work abit,
please. 
Third, he keeps the graphic detail level low. This smacks of the
3dfx-fansite "benchmarks" pitting the TNT2 Ultra's vs. the V3's, then
repeating the same test with the Geforce when it came out. "Hey, we keep the
effects down to a minimum, and look! The V3 keeps up! Yaaaay 3DFX! Yaaaay
fill rate!"
Fourth, he used v3.47 of the reference tnt/tnt2 drivers - I believe 3.48 was
the first rev that supported the Geforce. In fact, they have done 5
revisions since then, with 3.53 coming out a week before he published his
review. Maybe a week is pushing the envelope (I think not), but there's no
reason he had to go back that far.

Given the previous four points, the fact that it takes a 900mhz Athlon to
catch a 750mhz Coppermine (or, if those cpus seem stratospherically high in
speed and forever out of your reach, it takes a 700mhz Athlon to catch a
600mhz Coppermine), I'd have total doubts about purchasing an Athlon for a
personal machine. 

Fifth point, for us people (and not paid reviewers who get everything free).
If you go the Athlon route, keep in mind a point that seemed to escape
TomsHardware. The Athlon 700 runs HOT. It's completely unoverclockable
without a Kryotech cooler system. Will AMD be able to get substantially
higher clock speeds out of the Athlon without changing the architecture of
the chip? How much will a smaller die help? 
The Coppermine cpus put ice cubes to shame. Intel is committed to the flip
chip PGA architecture - Slot1 will be phased out entirely by Christmas of
next year.

We could debate fine points for a year over this. All I want to say is
neither kills each other at the moment, but in the near future I really do
see the Coppermine beating the life out of the Athlon. I thought the price
difference was far larger in my previous post - I now know that's not the
case. With only a 10% difference in price, mhz for mhz, there is no way I'd
recommend an Athlon over a Coppermine to a friend or colleague.