Slashdot Mirror
S3 DeltaChrome S4 Graphics Chip Reviewed
EconolineCrush writes "The Tech Report has a preview of S3's budget DeltaChrome S4 graphics chip for PC graphics cards. While not the fastest option for games, the S4 looks like a credible alternative to ATI and NVIDIA's dominance of the graphics market - there are some handy analysis graphs comparing performance in Wolfenstein: Enemy Territory, Unreal Tournament 2004 and Far Cry. Better still, the S4 has component HDTV output built right into the chip, making it an intriguing option for home theater systems."
Being in the market for a new graphics card (Doom 3 anybody?) I have to admit this wouldn't even show up on the radar. I have enough concerns about ATI stability, or the fact that I need to buy a separate minitower and nuclear power supply to power the nVidia cards.
The component out is a major selling point however, for home theatre people anyway who might want to play the occasional game.
This is more interesting for being the graphics technology that will be incorporated in upcoming VIA integrated chipsets however.
I'd still get a low-end ATI or nVidia card above this however. What will S3's support be like for Linux?
Not a top score, but an alternative more credible than XGI, IMHO.
Bye!
SeqBox
I thought S3 was bought out or disbanded or something quite a while ago. Is my memory playing tricks on me? Since when did they start making chips again?
And making decent graphics chips, no less. As someone who used a S3 ViRGE for much more time than anyone should have to, this is a certainly a surprise to me....
I concur with this statement. Closed-source drivers are a PITA to deal with. I'd happily dump my ATI card and get an S3.. even if it was somewhat inferior in terms of performance, just so that I could not have to deal with installing yet another program every time I recompile my kernel. Plus being open source and all, a lot of performance could probably be gotten through various optimizations over time.
100% agreed. I would (and do) buy damn near any old vid card, as long as I have a reasonable belief that open source programmers have the docs they need from the manufacturer to produce a good driver. Cards like the Rage128 and Millenium II are good (old) examples. The driver ATI puts out is not a useful product and while nVidia produces a fairly high quality driver, they don't cover all the platforms I might care to use.
So I second that. S3: steal this market!
Normally I'd disregard this as the usual slashbot knee-jerk, but in this case opening the driver source is actually plausible.
NV and (to a lesser extent) ATI have invested a huge amount of effort in their drivers. A good GL driver was never trivial, and if anything is becoming more complicated as drivers take on responsibilities like compiling and optimizing shader code. Even without the oft-rumoured third-party IP issues, I don't see much chance of the big players releasing their source anytime soon.
S3, on the other hand, may be starting with a pretty clean slate. Their drivers are probably still pretty shaky once you step off the usual Quake rendering paths, and tightening them up could take years if they only have in-house dev resource. They're positioning this as a budget part, and are presumably very keen to keep costs down. They're an outsider at the moment and might happily grab a niche like Linux as a toehold from which to make a play for the wider market.
Fingers crossed.
Hercules CGA? I thought that Herclues cards used the HERCULES standard, and to run a CGA app, you had to use an emulator.
Hmm... Well, then why not release all of the chip specs? Is the DESIGN of an nVidia chip not totally nV?
Why I can get PDFs or books full of info about AMD, Intel, ARM or TI processors so I can program them, avoid the "errata" problems, target new CPUs better or whatever, but ATI or NVidia can not provide any basic info now? Do they have anything to fear from others programming their chips?
I doubt the validity of these benchmark:
unless the radeon 9550 is radically different than 9600 pro (which I own), the 9550 should destroys in any benchmark test the nvidia 5200 fx(which I also own). 5200 is in fact just a little bit faster than a gf4mx440.They are two very low-end by today standard. 9600 (and so is 9500) is a mid-range card. So why in most test the 5200 got better result than 9550? Even more,I'm not even sure than 9550 exist. I know for sure 9500 and regular 9600, but these two are two close in performance for worthing a half '50 version.
Does anyone know more about the component outputs?
from the pictures it looked like it was an adapter that went to the svideo port, however from the small picture they had it was hard to tell.
I really don't know all that much about the video standards and wiring capacities, but I thought svideo couldn't cary hdtv signals.
These cards are programmed extensively using VLIW microcode, which contains the implementation of cross-licensed technologies. Since, NDA's/patents are only valid if the technology has not been released into the public domain, it's not possible to release this code. That's why you have third party extensions (SGI, HP, SUN, 3Dlabs etc..) in consumer OpenGL drivers. These extensions are protected by patents. As the device drivers install this code upon startup, releasing the source code to the device drivers would allow users access to the implementation of these patents.
Even if someone did write a pure inhouse architecture, the hardware register set can change so rapidly that it would immediately break anything hardcoded to the metal.
Vintage computer adverts: http://www.vintageadbrowser.com/computers-and-software-ads
The problem with the video card market can be seen right here. Look at the Slashdot section this is in: Games.
Video card manufacturers have stopped marketing their products to normal people, and have focused on gamers. Your MeshBlitter 99900 FireCore+ selling for 599 dollars and 99 cents isn't going to do a damned thing to improve my word processing. Heck, it will probably make it worse by driving me nuts with the attached Hoovermatic cooling system.
Yeah, all you gamers living in your parent's basement are going to mod this down for heresy, but the truth cannot be ignored, and that truth is that most people don't need more RAM for their GPU than their CPU.
Don't blame me, I didn't vote for either of them!
Like what 3dfx used to do, except that they had their own API that they could actually convince game developers to write to.
The reason NVidia destroyed 3dfx was their decision to implement Microsoft's reference rasterizer as fast as they could in hardware. S3, on the other hand, tried to design their way around cost-cutting decisions in ways which were arguably incompatible with D3D and then pitch the flaming turd over the fence to the driver guys, who were left looking like crap when the driver either went really slow or really buggy.