ti4200 vs 8500 = no difference? Grrrrrr!

  • Thread starter Thread starter Li'l ol' me
  • Start date Start date
The GeForce 3 TI-200 at 250mhz GPU was as fast and in a couple of cases
slightly faster than the 8500 at 275. The GeForce 3 TI-500 simply killed
it. The 8500 can't even claim to be in the same league as a GeForce 4
TI-4200.
Click to expand...


But the GF3 Ti500 is as fast, if not faster in some instances,
as the GF4 Ti4200. The speed increase is really not as pronouced as
your trying to make it out to be. You'll see the GF4 Ti4200, or a GF3
Ti500 outshine the 8500 when you raise the resoulutions, and add AA in
the mix, but if you use a screen res of 800x600, or 1024x768, you
really wouldn't see much more than 5 fps, or so.
 
El said:
This is incorrect. Did you view the link I provided? It shows the
8500 and Ti4200 performing about the same in various games.
Click to expand...


I happen to own all 4 mentioned cards. I can even state comparison between
that 8500 and either a 128 meg version of a 4200 or a 64 meg version of a
4200. If you take this HardOCP article as gospel, you should be buying an
ATI 9000 LOL. The only Radeon 9000 worth buying is the mobile version and
you can't be serious implying that the hardocp article is anything like
correct by saying that the 9000 measures up to a TI-4200 or anything else.
Hell, most onboard video beats the 9000 (onboard video from Intel *and*
nVidia). The TI-4200 is a superior card to the 8500 of any flavor.

So, as you revisit articles that are well over a year old and were not
accurate when written, ask yourself where you might find a card or two to
actually compare. Oh, and while you're at it, read Toms article that
compared the 7500 and 8500 to the GeForce 3 TI-200 and TI-500. Tom didn't
agree with hardocp, and neither did anandtech, if I recall.
 
Larry said:
That's retarded. When he runs the "benchmark" part, and it,
for example, shows the fps as 50 with both cards, doesn't that show
he's only getting 50fps with each card. What! If he runs the game
part, then one card will get more fps magicly?
Click to expand...

go to alt.games.unreal.tournament2003 and ask about the benchmark. You will
find common knowledge indicates that the benchmark (especially the fly-by
part of it) is far more CPU intensive than it is GPU intensive.
Furthermore, it is math-processor intensive beyond bandwidth intensive.
 
Li'l ol' me said:
Just bought a ti4200 on the cheap thinking there would be a modest but
noticable speed improvement from a Radeon 8500. But after running
Click to expand...
3DMark03...

I put this together for an earlier post. I just added the 8500 data.


Here's a list of the best scores in 3 benchmarks using 4 different cards. My
scores are included. I don't include any scores that were done with the
51.75 drivers. Here goes:

BTW: These scores are from the FutureMark and AquaMark3 databases. They are
just the best scores posted so overclocking level can vary.

Athlon @ 2.2GHz or slower:

3DMark2001
Best Ti4200 ------ 15534
Best 9600 Pro ---- 14349
My Ti4200 ------- 13967
Best 5600 Ultra -- 13722
Best 8500 -------- 12300

3DMark2003 (A Dx9 test)
Best 9600 Pro ---- 4345 (Dx9)
Best 5600 Ultra -- 3657 (Dx9)
Best Ti4200 ------ 2003 (not Dx9)
Best 8500 -------- 1761 (not Dx9)
My Ti4200 ------- 1720 (not Dx9)

AquaMark3 (CPU @ 2.1GHz or slower for these)
Best 9600 Pro ---- 33119
Best Ti4200 ------ 29028
Best 5600 Ultra -- 25544
My Ti4200 ------- 22118
Best 8500 -------- 18632


My card is OC'd 300/620 for all the tests.

Obviously the Ti4200 is a good performer if you use 3DMark2001 and AquaMark3
benchmarks and not so good if you use 3DMark2003.

If I were upgrading from an 8500 I would get the 9600 Pro. Since I already
have a Ti4200 I don't plan on upgrading any time soon.

If you wanted to turn AA and/or AF on, then the 9600 Pro and 5600 Ultra
would move to the top of all the lists. Even for "real" games. :)


Too_Much_Coffee ®
 
i'm_tired said:
go to alt.games.unreal.tournament2003 and ask about the benchmark. You
will find common knowledge indicates that the benchmark (especially the
fly-by part of it) is far more CPU intensive than it is GPU intensive.
Furthermore, it is math-processor intensive beyond bandwidth intensive.
Click to expand...

So are you saying that it is not typical of the game?

Or that it doesn't test the graphics independatly of the CPU?

My point earlier about knowing what you are testing is everything. If you
are using hte benchmark to measure the performance of the game on a given
system, and the benchmark is typical of the game, then thats fine. If
you're using it to measure different GPUs on systems with different CPUs
then you don't know how to perform comparisons anyway.

Ben
 
The gf3 was a bit faster when the 8500 first arrived, before ATI got their
drivers straightened out, then it was slower. I had a 275/275 R8500 in a
via kt266a board and a buddy had the gf3ti200 in an ECS K7s5a board with
sdram. We tried swapping cards (clean installs from ghost images), and
swapping his sdram with ddr. CPU's were xp1700, t-bird1.2G or so (can't
remember exactly, it was his). Then we benched for comparisons. I don't
have a lot of hard data on this but the basics are as below:
K7S5A was about 7% faster with ddr than sdr (I always wondered how big an
improvement the ddr was).
The 8500 was a bit faster than the gf3ti200 in benches but gaming
performance was similar, except maybe in slow areas of maps where the 8500
held up a bit better (mohaa, 1024, no AA). Image quality was definately
better with the ATI even without AF, and the ATI was ok with AF while we
didn't use it on the gf3.
Buddy then bought a gf4200 and while we didn't get together to do another
benching comparison, he didn't get a huge improvement over the 8500 (he had
borrowed it for a while). He then ditched the ECS board and got a kt333 and
xp1800. Card took off but we never did bench the difference or try the 8500
in his new board/cpu. But the difference between our systems was then
minimal, and the 8500 played well at the same game settings as the 4200,
though with frame rates a bit slower. In the end he turned eye candy lower
than I did with my 8500 but he was a fps freak while I wanted godd (above
50fps balanced with nice image). There was a difference, but I do remember
that it wasn't worth paying for.

Mike
 
I didn't expect this much response, but in answer to some queries:

My system:
Abit NF7-S
XP2100 (tbred)
768mb DDR 3200 (Twinmos)
Win XP SP1

I didn't buy the ti4200 as an upgrade as such, but I saw it on ebay for just
30 of our quaint little english pounds and couldn't resist. I can sell my
8500 for more than that, so it's win-win!

To be honest, I was quite happy with my 8500. I don't undestand people on
these groups that pay hundreds for cards just for high benchmark scores and
kudos, especially when you'll have to wait 12 months for most games to catch
up. If a card can do a minumum sustained rate of 50 FPS at my desired res,
then it's good enough. If your desired res is 1600x1200 for some reason,
then maybe you do need an expensive card, although my 8500 can run TOCA car
racer at 1600x1200 (but not much point on a 17" monitor). I think the 8500
is a great entry-level card, and deserves to still be on sale rather than
crappy sub 9500 cards like the 9000.

I was considering looking for a soft-moddable 9500, but with Doom 3 and
Half-life 2 and other DX9 games being put back, there seems no rush. I'm
looking to get a 17" TFT (or bigger) when they get a little better, and with
a native res of 1280x1024, I'll need a card that can do a minumum 50 FPS at
this res. Don't think a ti4200 or 8500 will suffice here for the very
latest games.

I have not tried AA and all that stuff as I don't really notice the effect.
The only game I have with jaggies that bother me is the first Serious Sam
game, and I only notice them in fly-bys.

I can set the mobo to anything between 133 and 213 FSB. The higher FSB is
what enables me to run TOCA at max res, but has little effect on most other
games/benchmarks (at 213, the ti4200 gives a massive 3% increase in UT2003
over the 8500 at 1024x768). I often set the CPU muliplier to 213 x 8 to get
the same CPU clock speed as 133x13, just to see what effect the extra FSB
bandwidth gives. TOCA responds to this if nothing else.

I have to confess, I've just tried 3dMark 2001SE and found a 10% increase
for the ti4200 at 1024x768 (over 11000 compared to over 10000 for the 8500).
With stock FSB (133) I got only 9000-ish with the 8500.

I'm not a benchmark expert, but I'm well aware that UT2003 fly-by is gfx
dependent and the bot-test is CPU dependent. I used 3dmark 2003 as I heard
it's a pure gfx test. I even read that I'd get almost identical scores with
the same card on a P2 board! I know it's a DX9 test, but as both 8500 and
ti4200 are DX8, neither has the advantage or disadvantage.

I've tried overclocking this card and have got nowhere. Default speeds are
250/513. Best (stable) I can get are about 266/532.

I tried a GF3 ti200 a while ago, and the difference was noticably worse than
an 8500. Not hugely though, so I'd like to try a ti500.
 
Your approach is sensible, and I agree with almost everything you say.
There's no point in spending a ton of money on hardware until the software
is there to use it, at which time it'll be cheaper anyway.

I couldn't hold out any longer, however, when I was able to pick up a
moddable Sapphire Radeon 9500 for cheap and turn it into a 9700 Pro -- the
mouse that roared.

However, there is one feature I think you have yet to appreciate. Once
you're really seen what AA and AF can do on a current Radeon with little
performance hit, there's no turning back.

Now that I've figured out how to configure them properly, I can play system
busters like Morrowind with smooth framerates at 1024x768 with 4X AA and 8X
AF, and it looks and plays better than 1280x1024 without them.
 
Ben said:
So are you saying that it is not typical of the game?
Click to expand...

Yes. That is what I'm saying. The UT2K3 benchmark is not representative of
the performance you'll see in the game. While playing, you'll find that a
greater benefit can be had from a 533 or 800 mhz FSB pentium (the faster and
more bandwidth the better) but the benchmark will give you far better
results (again especially in the flyby portion) with a nice AMD XP processor
(more instructions per clock and a much much stronger math processor). Also
at resolutions in the 1024X768 or 1280X720 range, generally an nVidia
equipped AMD machine will score somewhat higher in the benchmark. Well,
after reading the actual results others get while playing the game, I think
if I built a machine to only play UT2K3, I'd have to build one with an
800mhz P4 and an ATI 9700 or better.
Or that it doesn't test the graphics independatly of the CPU?
Click to expand...

The benchmark certainly does not represent a method of testing graphics
independantly of the CPU (I'm not even sure there is any game engine or
benchmark anywhere that can really do that)
My point earlier about knowing what you are testing is everything.
If you are using hte benchmark to measure the performance of the game
on a given system, and the benchmark is typical of the game, then
thats fine. If you're using it to measure different GPUs on systems
with different CPUs then you don't know how to perform comparisons
anyway.
Click to expand...


Again, the UT2K3 benchmark is not a good representation of what you can
expect for any system's performance while playing UT2K3. I won't discard it
as absolutely worthless as a system bench, though. It may serve some purpose
for a system bench, but it just isn't going to accurately tell you how well
your system will actually play that particular game.
 
He's not... thats why he's replacing it... unfortunately with a card of a
similar generation and has gone sideways rather than up.
Click to expand...

That was my point -- he expected a big change, but he did not
(understandably) see one.

jw
 
ELVIS2000 said:
That was my point -- he expected a big change, but he did not
(understandably) see one.

jw
Click to expand...

I did not expect a "big" change, I expected a "modest" change.

Please read my posts properly in future.
 
Back
Top