|
Installation
Like
just about every GeForce product we've ever looked
at installation was as simple as simple as it gets.
The installation CD includes the required drivers,
in fact it comes with two driver sets, Detonator 40.72
and 31.40. To their credit, although Albatron haven't
gone to the effort of designing their own overclocking
interface they do include the CoolBits registry hack
in script form making applying it a lot simpler than
having to poke around in the registry yourself, particularly
if you're a little nervous of this as many are.
If
you've ever used a GeForce card before you'll pretty
much know the score. Installation is as simple as
it gets. The only potential problem you might run
into is when using the 40.72s which are a bit touchy
about old driver remnants. If the 40.72s refuse to
activate you can use a handy little app called "Detonator
Destroyer" if you're running Windows 9X or
ME.
Performance
Ti4680P-Turbo
Ti4200P-Turbo
Before
we kick off I should mention that this isn't going
to be our usual performance assessment, I doubt many
of you aren't familiar with Ti4200 levels of performance
and although you'll get a fair representation of what
this level of graphics card is able to do the review
kind of evolved in a different way to usual. What
the hell, it makes a change eh! Initially I was just
going to run a few loops of 3DMark and throw up the
results but after getting some unusual results on
the first run I decided I may as well turn it into
a full on comparison session and at the same time
try to clear up some of the questions we've been asked
about AGP aperture sizes and the benefits of enabling
Fast Writes in the BIOS. I realise some of you don't
have a lot of faith in 3DMark but for the purpose
of this kind of like for like comparison it's the
perfect tool, and look on the bright side, if you
get fed up of looking at 3DMark scores during this
review imagine how I was feeling after having sat
through them all!
Test
System :
Motherboard
- KT400: Albatron
KX400 8X
Motherboard - KT333: Albatron
KX400 Pro
Memory: 256MB Corsair XMS3200 CAS2
CPU: AMD Athlon XP2100+
Video: Albatron
GeForce4 Ti4200P-Turbo
Video: Albatron GeForce4 Ti4680P-Turbo
3DMark2001SE
Build330
In
the words of MadOnion.com, 3DMark®2001
SE (build 330) is the latest installment in the
popular 3DMark series. By combining DirectX®8.1
support with completely new graphics, it continues
to provide benchmark results that empower you to make
informed hardware assessments.
To
begin with I ran up some results using 3DMark at the
default 1024x768. Memory was running at 300MHz and
timings were set manually to 2.2.2.5.1T. The Ti4200P
being an NV25 part ran using AGP4X while the Ti4680
being based on the NV28 part, or NV28.2 to be more
accurate, was running using AGP8X.
All benchmarks on this page are run using the VIAKT400
chipset Albatron
KX400 8X
|
Ti4200P-Turbo
|
 |
|
Ti4680P-Turbo
|
 |
You
don't have to look too hard to see that something
seems a little odd here. 10,751 3DMarks is an admirable
score but it's some way behind its older brother the
4200P-Turbo. Figuring I'd missed something I retested,
then retested again. I set the memory to run from
its own SPD timings which did lower the 4200P-Turbo
timings to around 10,700 points but alas it also lowered
the Ti4680 score to around 10,500 points in the process.
Time to fire up UT and see what it had to report.
Unreal
Tournament 2003 :
I
ran UT2003 using the script created by the guys at
[H]ard|OCP. Brent has coded a very nice script with
a simple, easy to use front end that belies the effectiveness
of what it can actually achieve. From this simple
GUI you can set one of three resolutions or specify
that all three be run. You can set high or low quality
with a click of the mouse or you can isolate CPU performance
by forcing the benchmark to run at 640x480 with most
of the detail options turned down or off. If you don't
already have this terrific little app I suggest you
head straight over and grab it HERE.
Please remember this is for use only with the full
version of UT2003 and that because Brent has tweaked
some of the detail options under [D3DDrv.D3DRenderDevice]
these results aren't directly comparable to those
you'd get using the inbuilt benchmarking options.
|
Ti4200P-Turbo
|
 |
|
Ti4680P-Turbo
|
 |
Here
the results are less clear. The Ti4200P-Turbo seems
to demonstrate higher maximum and minimum framerates
in many cases but whyen looking purely at the average
framerate the Ti4680 has a very slight advantage,
and it is very slight.
With
this in mind it was time to see what OpenGL performance
was like and feeling like a change I loaded up VulpineGL.
VulpineGL:
Vulpine
GLmark v1.1p was run at 1280x1024 with detail set
high and standard OGL 1.2 used.
|
Ti4200P-Turbo
|
 |
|
Ti4680P-Turbo
|
 |
Again
the performance advantage swings back to the Ti4200P-Turbo,
not by a huge amount it seems but 1 FPS is actually
fairly significant in Vulpine. Lastly I ran a few
loops of AMD's reasonably demanding but pitiful to
watch NBench. This is version one in case anyone is
wondering.
NBench:
Ti4200P-Turbo
Ti4680P-Turbo
At
this point it's becoming fairly evident that the Ti4680
can't compete with the Ti4200. It should also be kept
in mind that none of these benchmarks, and so far
as I'm aware no video system benchmark currently available,
is designed to utilise the AGP8X standard. Is what
we're seeing the results of inefficient use of the
AGP bus by AGP8X cards when the software they're running
isn't specifically coded for AGP8X?
At
this point the sensible thing to do would have been
to drop the AGP bus from 8X to 4X and compare again
but neither this nor indeed any of the KT400 chipset
motherboards I have available here allow this option.
In
a moment I'll switch out the KX400-8X (VIA KT400)
motherboard and replace it with the KX400-Pro (VIA
KT333CE) and see how they shape up when both are constrained
to run at AGP4X.
Before
this though I was curious as to what was causing the
discrepancy and remembering a few recent questions
we'd had about the effect of AGP aperture size on
performance I got to wondering whther the AGP8X NV28
was more sensitive to AGP aperture than the AGP4X
NV25. We knew that AGP aperture sizes over 16MB made
very little difference on AGP4X boards but perhaps
the new NV28 was a lot more choosy? No harm in looking
anyway.
|
