Slashdot Mirror
The Return of S3
flynn_nrg writes "Just saw this article on ExtremeTech about S3's new graphics card. S3 is back on the scene with its first new GPU architecture in five years. Rather than take aim at the high-end, S3 has set its sights on the midrange price/performance category, which is currently dominated by ATI's Radeon 9600 XT and nVidia's GeForce FX 5700, both of which are under $200. Today S3 unveils the DeltaChrome S8 GPU, which represents the midrange of its upcoming line of DeltaChrome GPUs."
Without some razzle-dazzle high end cards to "wow" people with, they probably won't get the publicity needed to sell these midrange cards.
I have been using out S3 supply (outrageously large) of these cards for servers for a long time. And it doesn't get any better then that.
Basically, we have tons of these things and they were used back in the day when we didn't spend all of our money on expensive computer peripherals.
I would recommend using these for anyone that does not use the computer as a workstation - such as a file server or in my case, a home machine that I ssh into. Heck, I don't ever turn on the monitor quite so often for that thing.
Go S3!
Wow.
There 3d cards sucked back in 96 when I bought my S3 virge. I figured it was going to be the defacto standard since Vodoo was new and never heard of. Just upgrading to NT4 and Linux from DOS, I assumed it was up to the game makers to provide the drivers and not up to directx and opengl to provide support.
But I have upgraded to 2 newer pc's since. I forgot all about them and assumed they went under. I doubt they will support FreeBSD/Linux and X as they did in the past with their own Xserver.
http://saveie6.com/
But for only $150, nothing should hold this card back aside from name recognition. The $150 print point almost seals the deal for me, only that I'm holding out for better offerings from ATi and NVidia before moving up from my GeForce2 MX (I'm not much of a gamer).
.13 micron manufacturing process, the same as ATi and NVidia, which should allow them to crank out higher-speed cards within the next few months, at least allowing S3 to remain competitive.
Overall, I have to agree with the concensus that S3 is back, and may be primed to stay in the market for some time. The article mentions that they are using a
Either way, the video card market may just be heating up for 2004.
The Political Programmer
Mid-range graphics cards seem a slightly pointless purchase, given that you can buy top-of-the-range cards from 6 months ago for a fraction of their original prices (not to mention the second hand prices).
Why buy something mediocre but brand new, when you could buy something that absolutely kicked ass six months ago for a similar amount of money?
Organic free-range music... yum!
Remember the Kyro II? The chip used a unique tile-based rendering system that produced performance similar to the then-current Geforce 2s (although some synthetic benchmarks indicated otherwise) while being priced more in line with the MX line of cards. After much reading and research, a buddy of mine decided to pick one up for his machine, his reasoning being that he wasn't a super hardcore gamer, but wanted to be able to throw down with us every once in a while.
Flash forward a couple of years, and while NVidia and ATI are still willing to release updated drivers for their cards of that era, the Kyro lingers unsupported, even though NEC (the chip designer) and Guillemot/Hercules (the card manufacturer) are still going strong. My friend wanted to play Halo, and even though the card should've been able to support the game (albeit at a lower resolution/framerate), he can't because his card is basically ignored and unsupported by the game manufacturers and the source comapnies for the card itself.
The moral of the story: S3 is a reasonably well-known name. So is Hercules/Guillemot/NEC. It's gonna take a hell of a price/performance ratio to get me to recommend a video card not based on Ati or NVidia after the Kyro debacle.
I want a new video card, but no one makes the card I want. I do not give a crap about playing games, I want the modern equivalent of a Matrox card, one that is cheap yet renders beautiful color for 2D apps. Hell I still prefer the ATI Rage card with all 8mb of ram to the crappy Geforce with 64mb. I want clean bright color, I do not care how fast the card is. I am willing to pay about 50 bucks for it. So where is it? Maybe I need to find some new old stock matrox cards? Any pointers towards those?
Sure, that's a good answer, but I doubt they're likely to just fire all their driver staff ( even if they do deserve it ) and turn the whole thing out in the open, right? At the very least, I can't see the windows driver being replaced with an open effort ( call it cultural resistance ), and Windows is where the money is. They also need a good driver at ship time - they can't wait around while some volunteers put one together. And lord help them if there's a bad bunch of open drivers due to inadequate testing / mistakes ( hell, we all make them ), and S3 couldn't help out their customers because they don't support the driver (!!) that would be a PR blow not soon recovered from.
What I'm trying to say is that out here in Linux/BSD land, our market share is so vanishingly small that I'd be surprised if they offer a binary driver + wrapper, ala NVidia. They're going to be busy enough copping a hell of a ride from the established players.
YLFIOne god, one market, one truth, one consumer.
If you want to make it a political issue, that's fine -- more power to you. But recognize that you're among a minority. I'm not sure I believe that there are "hundreds of thousands" of *desktop* Linux users, I refuse to believe that any preponderance of them "are more interested in non-proprietary kernel modules than raw performance."
I'm a Linux user, and I believe in/contribute to "the open source movement". When it comes down to it, however, I care a lot more about things working right than whether or not I have the source code.
NVidia's drivers work (relatively) well, and they've been writing Linux drivers longer than most hardware vendors. For this reason, they'll continue getting my money until somebody beats them in Linux support. Maybe it could be S3.
A $75 card won't last you too much longer now... I find my Radeon 8500 struggling with Halo at 800x600, and DeusEx 2 at 640x480 (with shadows completely off!). All these games now with their realtime physics and shadows...
Back in my day we were happy to have textures...
Who doesn't like free music?
Sometimes I'm reminded of why RMS draws a hard line between Open Source and Free Software. :-)
NVidia's drivers work (relatively) well
For some applications, maybe. For others, the closed drivers are clearly inferior to XFree's "nv" module. For example, if you're running Linux on non-Intel hardware, or running a non-Linux Unix on Intel, then you're pretty much out in the cold. Sure, they release a FreeBSD module every now and then, but that's no help for NetBSD or OpenBSD folks. Do they offer binaries for PowerPC Linux? I'm not sure, and not interested enough to look it up at the moment.
I'm a good programmer. I have some experience debugging hardware drivers and submitting source patches. However, if the "NVidia" module crashes, there's nothing I can do except send in a half-informed bug report and hope that enough other people gripe about the same problem to motivate someone to fix it. Remember, the FSF started as a consequence of RMS not being allowed to fix a broken printer driver. :-
So, if "work[s...] well" means "usually executes without crashing and offers decent performance", then I won't argue. However, that's not the standard of "works well" that I use for myself and my employer.
Dewey, what part of this looks like authorities should be involved?
I reciently bought a ATI AIW 9600 pro card, and It is one of the biggest computer let downs I have ever seen. The drivers were crap, and it took me the better part of two weeks to get all of the 'features' on the card to work.
Maybe now, with more competition in this segment of the market, the card makers will start putting out a good final product, and not make the buyers be the the BETA testers!
Diamond is back. Best Data picked them up and relaunched the company(sans the audio division, which someone else owns).
I recently quit one of the big 2 GPU companies to pursue other opportunities...which one is irrelevant, but this is an AC post none-the-less. This is a brief look at the business end...I'll leave the "it's great" or "it's garbage" discussions to others.
To use an overused buzzword, lets assume that the S3 chip has the best "price/performance ratio" of any chip. S3 still has little chance to gain any real market share, mostly because they have little chance to get in OEM systems.
Let me explain. The retail market (where you go to BestBuy or newegg.com) makes up a very small percentage of the overall market. I can't give real numbers (I don't know if they're NDA'd or copywritten by the research company, so better safe than sorry), but lets just say, it's the OEM sales that pay everyone's salaries and keep the investors happy.
Since OEM sales are so important, lets jump into the mind of the OEM. There are 3 major things that the OEMs care about when choosing the chip to put in their computers.
1)Does this chip perform SIGNIFICANTLY better than what we're already using?
2)Is there any benefit with using company X over company Y?
3)Are we getting a better deal from the new company?
So, what does this mean for S3 (lets throw in XGI also). To put it simply, change is difficult and expensive. Assembly lines need to be retooled, software needs to be changed and re-validated. There needs to be a good reason for an OEM to change.
Going down the checklist:
1) They do not, and never will, have a part that performs that much better than nVidia's or ATI's midrange part (if they keep the "we only want the midrange" strategy). This is because the big 2 can generate a better midrange part by either lowering the price on a higher-end part, or by tweaking the binning of the higher-end parts (a high-end part that fails may be able to run as a mid-range part). Obviously, the low and mid-range parts make up the bulk of sales (and therefore contribute most to market share), so there's no way ATI or nVidia would give up any market share without a fight...and both companies have much more ammo (graphics IP) than S3 or XGI.
2)Positive mindshare in the IT world is a HUGE thing. Most of the time it is more important than the quality of the product. Though, a good product usually generates a greater mindshare, it's not always the case (read: Microsoft...to the uneducated masses). In graphics, it's been shown that the easiest way to generate a positive mindshare is to have the fastest & most stable product. nVidia built it's reputation on it's Riva and GeForce lines. ATI got back in the game with it's 9700. For S3 or XGI to gain mindshare, it can't elicit a "ooh, it's competitive" remark. It needs a "holy shit, that's fast" remark...that or some kick-ass marketing.
3)This would have to be one hell of a deal. Switching involves a risk that they will not sell as many PCs (and make as much money) as they already are. If money alone is driving the deal, the OEM would have to feel that there is a good chance of them making more money while selling fewer PCs...it doesn't take an economics major to see what that would mean for S3's or XGI's profit margins.
So, how could S3 or XGI really take market share from ATI and nVidia? Simple, make the fastest part out there at a price that rivals what nVidia and ATI sell their high-end parts for. Can one/both of them do that? Maybe, but it won't be easy. If they can do that, then they will have a solid foundation for deriving the mid-range parts, and the mid-range parts will practically sell themselves.
Don't get me wrong -- I'd love to see a completely open driver from NVidia, but because of patent issues and licenses they have with other companies, it simply will not happen. Ever. But I need to do actual work on my computer that requires the use of a working graphics card, and the status of the source code is secondary to the performance of the binary.
Tech Report also has DeltaChrome preview with screen shots of just how messed up S3's drivers are in some applications.
Wow...they even brought back the same logo. Brings a tear to the eye. Not that I have entirely fond memories of Diamond products...the Stealth II was nice, but I was always annoyed at the complete lack of support for the original proprietary Monster Sound cards (never even wrote a driver for Windows NT/2K/XP, much less released specs to the linux community -- but I wouldn't have cared at all if it didn't have pretty decent analog output quality, and more power than almost any other PCI card I've used). But the circumstances of their demise left a rather nasty taste in my mouth. The story involves S3 to a large extent, although like Diamond, S3 then was not S3 now.
Diamond was one of the more prominent aftermarket expansion card marketer of the nineties. They were very successful selling mostly video cards, based first on S3's chipsets, which were very competitive until 3D acceleration became popular, and later nvidia and 3dfx. They branched out into a wide array of products, including SCSI controllers, motherboards (after acquiring Micronics), modems (after acquiring Supra), and audio cards. They invented the portable MP3 player, with the original Rio, and developed some of the first telephone-line and power-line home networking products. But, largely because of acquisition and competition, they were constantly losing money.
S3 was probably in a much worse bind. They were also losing money, but had none of the innovation that characterized Diamond's last years. They had been surpassed by new competition in graphics chipsets, and had no real other business. But through a lucky investment in TMSC fabrication plant, they had some cash on hand, and decided to buy out Diamond. At the time everyone assumed they were going to follow 3dfx's lead and produce sell graphics cards based on their own chipsets directly. But the truth is, they were looking for an exit both from Diamond's core computer component business, and their own graphics chipset line. After the rushed-to-market, broken, Savage 2000 was a market failure, they abandoned expansion cards entirely, throwing away the legacy of two PC hardware pioneers in favor of the Rio MP3 players, and another technology they had acquired, ReplayTV's personal video recorders. At the same time, the graphics chipset operation was spun off as a joint venture with VIA. This is what is now known as S3. The rest of the company was renamed SonicBlue. Completing the trajectory set by S3 management since the days of the Virge, they went bankrupt recently, and the Rio and ReplayTV units changed hands yet again, hopefully to more competent management. Best Data apparently picked up the old Diamond brand at the same time.
As to this new graphics chipset...I wouldn't take it seriously unless it is proven to perform decently (well, actually I wouldn't take it seriously unless it also had Linux support on par with the old Matrox card I use now, but I digress...). As far as I can see VIA is just looking for some paying beta testers to work out the bugs in the core before they embed it in their next-generation southbridge chips, so don't look for a renewed commitment to serious graphics hardware from "S3".
"(Man) tries to live his own life as if he were telling a story. But you have to choose: live or tell." --Sartre
Ok, I just bought this card and it seriously took over a week to configure to get things stablize. I jumped around from the catalyst 3.7, 3.8, 3.9, 3.10 before things would work decently.
I think if S3 can build a card with drivers stable on the first install... they'll have my money. From what I know the latest geforce FX5900 has the same problems. It's just mind boggling having to pay so much and still dealing with such a bad out-of-box-experience.
I am playing some of the most common games (RTCWET, battlefield 1942, call of duty) and they all took a massive amount of driver tweaks and install sequence to work right. The market is flooded with premature products if you ask me.
If you haven't given ATI a shot lately, give 'em a try. I was nVidia only until my 128M Radeon 9700 Pro. Prior to it I'd had a TNT, a TNT2 Ultra, a GF2, and a GF3. Given the "optimizations" nVidia chose on my behalf (sacrifice image quality for speed) I decided to give the 9700 Pro a shot right after it came out. I've been very happy, and have had no issues to date. I'm running WinXP though, so if you are running Linux, I can't really speak for the driver support there.
My Tech Posts on Twitter
What's the point to not releasing documentation, when your card is not "high speed"? What you have to hide?
By opening source of drivers and releasing documentation - company could gain:
And it means money, because better drivers and better karma means bigger sell.
Hahaha, that's great. After the TNT I bought a TNT2Ultra made by Guillemot (which was a great card) and a few years later donated it to my brother. He had been using that card to play every single new game (including Max Payne) up until a few months ago when he built a new PC from the ground up.
One nice thing about Nvidia's driver upgrades over the years is that each release has improved the performance of damn near every card they make. My assumption is that the drivers are 50% of the card's performance..which would make sense in the context of them being unable to fully open-source the driver.
I've seen POST failures with an AGP card not pushed all the way in, but never with no card at all. Remember, the 8-but MDA / MDPA card was actually undetectable to the BIOS selftest routine {remember, even on your spankiest new 80686-class processor, all the BIOS stuff has to run in "traditional" 8086 emulation mode}, or at least indistinguible from a more powerful card. As long as there is a faint glimmer of a possibility that someone could actually wire up some old 16-bit slots to a motherboard and plug in one of these old cards, then motherboards will have to pass POST with the graphics card missing presumed MDA.
Je fume. Tu fumes. Nous fûmes!
Overall, I have to agree with the concensus that S3 is back, and may be primed to stay in the market for some time.
Indeed.
I find this card interesting for home theatre applicatons, where 3d capabilities (while nice and IMHO necessary for a complete entertainment system, including xmame and 3d simution support) don't have to be cutting-edge fast. Of particular note is this card's component output capabilities and ability to do 1080p, 1080i, 720p, etc. Right now my home theatre PC has an ATI card connected to my TV's DVI input. However, the TV only has one DVI input, while it has three component inputs, so being able to connect the computer to a component input (and free up the DVI interface for an HDTV tuner) would be nice.
Of course, until and unless there are decent Linux drivers for the hardware it will be of absolutely no interest for those of us building truly DRM-free home entertainment systems. Which is where S3, like so many others, may shoot itself in the foot (more's the pity). Here's hoping I'm wrong, and we do so solid 3d, X11, and linux driver support for the mpeg2/mpeg4 chip, tv and component outputs.
The Future of Human Evolution: Autonomy