Discovery on video card.

[MadPistol]

I think you will find this interesting.

Here are a few scores I just tested out on my Geforce FX 5900. First, benchmarks that are tested on normal clock speeds, and round 2 which is tested on the FX 5900 when it is Severely under clocked. take a look at these scores.

Core: 400
Mem: 850

3dMark03: 5557 on 53.03 driver
3dMark2001: 14930
Aquamark3: 46893

Now for the fun stuff

Core: 200
Mem: 300

3dMark03: 2576

I was very surprised at this score. It doesn't really add up because all of the individual tests were still scored higher than my geforce4 Ti 4200, even though the clockspeeds in both memory and core were slower than the Ti 4200. Interesting. Let's see how the results turn out on 3dMark2001.

3dMark2001: 9378

From what I could tell, only the lower ended tests even made that much of a difference from the original. If I'm correct here, the FX 5200 gets a lower score than this at higher clock speeds. Weird huh? Let's look at one more test. Maybe this one is a little more reliable.

Aquamark3: 17421

That is a reliable score there. I will definately believe that Aquamark3's score is accurate. It is the closest score:clock ratio that any of the tests have produced. Hats off to Aquamark3. :t

I guess what I've been trying to prove here is 2 things.

1. Is the difference in architecture in video cards much different in clock speeds?

2. Which benchmark is most accurate?

Question #2 has been answered here. However, Question #1 has yet to have an accurate answer. Any takers? :x

[RamonGTP]

Its really wierd at all, there's a lot more going on than merely clock speeds. For example, the Ti4200 isn't even a DX9 card, so it will naturally do poorly in 3dmark 03.

The 5200 is also vastly different from the 5900, it isn't just clock speed. For starters, the 5900 has a 256bit memory bus, the 5200 is 128 if you're lucky, many are only 64. There are several other differences in the core as well that I've read about but cannot recall.

[Bigjakkstaffa]

As stated, 3mark03 isnt an overall system banchmark, it more of Direct X 9 capability benchmark, if you dotn have direct X 9 hardware your harldy goign to make it past the 1500 Mark no matter hwo good you machine. As such i dont use 3dmark as it doesnt accuratley reflect the systems overall performance

--Jakk:t

[MadPistol]

my point exactly. I guess what I was trying to prove is that even a severely underclocked card will perform better than some of it's predecessors. You are right in saying that 3dMark03 is very dependent on the fact that a card has DX9 capbilities. Since the Ti 4200 doesn't have those capibilities, it severly hurts the score for DX8 cards. However, at those clock speeds, the Ti 4200 performed better on the DX7 test (Wings of Fury). However, since the FX 5900 has a much better architecture than the Ti 4200, it did win in the complex DX8 tests. (Battle of Proxycon and Troll's Lair) I guess what it really depends on is what was the card originally made for?

BTW, when I lowered my cards clock speeds, it didn't lower the overall temps for the card. Keep that one in mind too.