Now letīs see what we have with real life
benchmarks. I made a mixture of older and newer games. But as with the 3DMark99 results:
The drivers are not final now. The newest ASUS drivers for the V3800 are v1.25! But the
v1.20 drivers are OK for a direct comparsion of the cards. I made some tests with the ASUS
V3800 drivers and the results were nearly the same.
Testsystem: Intel Pentium II 450MHz, ASUS P2-B
Mainboard, 128MB SDRAM (7ns)
ASUS V3800 and STB Velocity running with NVIDIA Detonator reference drivers v1.20
UPDATE 27.04.99: New
values added. ASUS AGP-V3800TVR overclocked to 150MHz core TNT2 clock and 183MHz memclock
(yellow)
Iīve overclocked 'the beast' to
150/183MHz. The card ran stable and without faults. You can get an impression of the power
of the TNT2 Ultra version of the card. But these Ultra cards will have 5ns or 5,5ns RAM.
The memory used on this card is 7ns and has higher timings (RAS/CAS). So weīll see that
an Ultra is faster than this..... apart from higher values through overclocking the Ultra.
| GL-Quake |
|
| Timedemo 1 |
|
| Bigass |
|
| Timedemo |
|
| Bigass |
|
| Comment |
|
The
TNT2 is a lot faster in GLQuake than the TNT. The falloff at
higher resolutions is not as huge as on TNT. A first impression is
that the TNT2 has a much higher throughput. You can see that
Quake is a bit older now. The CPU has enough time to feed the
cards. The TNT canīt follow the TNT2 here. |
| Quake2 |
|
| Massive |
|
| Crusher |
|
| Massive |
|
| Crusher |
|
| Comment |
|
Quake
2 is a bit different. Watch the crusher results. Crusher needs
more CPU power. The cards canīt show their real performance.
Only the higher clockspeeds of the TNT2 are responsible for the
advantage. But I must say: 6fps are a lot in crusher! And TNT2
will perform a lot better with faster clockspeeds. |
| Turok 2 |
|
Settings:
V-Sync ON
Tripple -
Buffering |
|
|
|
|
| Comment |
|
Itīs
hard to say something about Turok2. The engine can render
with 32Bit colors but you have heavy harddisk access during the
benchmarks. This may be responsible for the results here. In
addition, you can see delays during the benchmark, but the values
are not effected of that.... But I show the results to keep it
complete.ge ich die Ergebnisse aber trotzdem. |
| Incoming |
|
Settings:
V-Sync OFF
|
|
|
|
|
| Comment |
|
Incoming
is simmilar to GLQuake. The CPU has enough time to feed
the cards. The V3800 is running away from the TNT. |
| Kingpin |
|
Settings:
Big
|
|
|
|
|
| Kommentar |
|
Kingpin
is based on the Quake2 engine. You can say that it is like
a heavy Crusher benchmark. The CPU canīt feed the cards with
data. So you can see only small differences here. I think it is a good
benchmark to compare different types of graphic chips. But the TNT2
is an evolution of the TNT - so you canīt see gig differences here.
You will see larger values at higher clockspeeds because this is
one of the two factors that produce the performance. |
You can
see a huge increase with the new clockrates in Incoming and GLQuake. Incoming shows the
great 32Bit power of the TNT2. As expected you canīt see an increase in games like
Kingpin or Q2 Crusher Demo because of the high CPU load.
RIVA Station 1999 - Lars Weinand |
