[an error occurred while processing this directive]
September'2001 3Digest
S3 Savage4
This chip is presented by Diamond Ste alth III S540 32 MBytes, AGP videocard.
[an error occurred while processing this directive]
Features:
Tweaks up to 155/160 MHz (see summary diagrams).
For the 20th of June, 2001 the latest drivers version is JCG 3.05.
[an error occurred while processing this directive]
[an error occurred while processing this directive]
[an error occurred while processing this directive]
[an error occurred while processing this directive]
[an error occurred while processing this directive]
[an error occurred while processing this directive]
[an error occurred while processing this directive]
[an error occurred while processing this directive]
[an error occurred while processing this directive]
[an error occurred while processing this directive]
[an error occurred while processing this directive]
A lot was written about the quality of this videocard, a lot of drivers FAQs are out to help set them, there are also some tweaked drivers, supporting popular games, written by enthusiasts. By the way, I'd advise to use these tweaked drivers instead of S3 references that are usually shipped along with cards. Tweaked drivers give a stability effect in the work of these videocards. Before getting good results in games you'd have to go through various lock ups. Therefore, as an exclusion Savage4 was tested with tweaked drivers that look like a cocktail made from different parts of the references, and also there's an S3Tweak utility in addition. I'd advise to every owner of this videocard to get this program and tune your settings with it.
And speaking about comparatives I should mark that versions don't differ much, but the JSG 3.05 latest drivers showed good speed results, thus it can be recommended. And namely this version took part in usability coefficients calculating.
And, in general, cards based on Savage4/Savage2000 are in the decline nowadays as no one releases new drivers for them, and thus the technical support is getting worse and worse. We'll continue to review new 'cocktails' from old drivers, but it's clear that the reliability of this cards on new platforms has to be supported by completely new drivers, that is unlikely. Thus we are forced to put these articles into archives.
I advise you to read about S3TC influence on speed and perfomance, because S3TC realization in OpenGL is important nowadays, and it shows good results.
In addition, S3 Metal API worked bad. For example, Unreal (S3 Metal) locked up immediately in spite of the S3Metal driver version (and everything worked on KT133 platform). The same version worked stable on both platforms in Unreal Tournament (though having some artefacts). For normal play with API S3 Metal Vsync should be turned on, because if not, there would be a huge lag.
We should say that a number of opinions about the speed of this chip can be equal to the number of the cards based on it. Somebody after a week of tuning was able to get better results in some specified games and somebody wouldn't. Don't blame people that bought this card due to lack of the experience and are not able to set it right. Not everybody can continuously get the card working as it should do. And you should pay attention to the fact that due to this chipset's low price many Chinese manufacturers produced many boards of different quality, their stability can vary a lot.
All problems mentioned above might occur due to S3's bad attitude to writing drivers. Since newly organized S3 Graphics, created on the base of VIA's S3 department, has refused to support Savage4/2000, the only our hope is continuously reducing number of enthusiasts. Because of bad drivers there are problems with video quality. Here are some of them.
Unreal Tournament is the best in working and stability, especially with S3TC (S3 Metal) support. Still there are some bugs:
[an error occurred while processing this directive]
And No One Lives Forever has some textures missing:
[an error occurred while processing this directive]
Other screenshots are below, you can compare them to the reference ones taken on NVIDIA GeForce3.
| Games / Cards | 1 | 2 | 3 | 4 | 5 | 6 |
| S3 Savage4 |
|
|
|
|
|
|
| NVIDIA GeForce3 |
|
|
|
|
|
|
| Games / Cards | 7 | 8 | 9 | 10 | 11 | 12 |
| S3 Savage4 |
|
|
-
|
|
|
|
| NVIDIA GeForce3 |
|
|
|
|
|
|
| Games / Cards | 13 | 14 | 15 | 16 | 17 | 18 |
| S3 Savage4 | 
|
|
|
|
|
-
|
| NVIDIA GeForce3 | 
|
|
|
|
|
|
| Games / Cards | 19 | 20 | 21 | 22 | 23 | 24 |
| S3 Savage4 |
-
|
-
|
|
-
|
|
|
| NVIDIA GeForce3 |
|
|
|
|
|
|
Game list:
OpenGL trilinear filtering doesn't work in spite of multitexturing settings. But there's an effective trilinear approximation that "cheats" Quake3. E.g. with "r_colorMIPLevel 1" the game shows the absense of trilinear filtering any way, but the borders between MIP-levels are being slightly smoothed:
To the left is a shot with bilinear filtering on, to the right - the one with trilinear approximation:
[an error occurred while processing this directive]
[an error occurred while processing this directive]
[an error occurred while processing this directive]
But still this is not complete trilinear filtering thus there are some artefacts in game dynamics.
2D quality is fair enough. At 1024x768 resolution you can notice as scene "soaps" so to say. Pay attention to the fact that 2D quality of such cards strongly depends on the manufacturer and the exact card.