The internal unit appears to be an Adaptec Ultra-Wide to the system. (We use Linux on one of these boxes as a production WWW server in an Enterprise Class operation that's about to be opened up to the world shortly...)
After all, Eddie was designed for doing HA/HP clustering of web servers. Use the bigger iron to handle the MySQL database engine and web serving and then have the old box handle just web serving. I'd bet that would handle a hell of a lot more than it would by itself....
Mystique/Millenium 3D "suppoprt" is painfully weak
on
Friday Quickies
·
· Score: 1
(Wiping egg off face...)
Hrm... Wonder where it could have been in the specs doc that I could have missed that one- I must have read the thing time and time again and didn't see it in the docs. Guess I need to look it over again (since Z-buffering would help out nicely in speeding things up...) and find it.
It won't do for me to continue making mistakes like this if I want to continue development of 3D support (and I do!)- thanks for the correction!
I plan on initially using that one...
on
Friday Quickies
·
· Score: 1
But, since it's proprietary (written under an un-needed NDA- the stuff's no longer a trade secret even if 3DLabs is still acting like it is!) I plan on extending the thing with non-closed-source code at some point.
Actually, this is all a moot point when XFree86 4.0 comes out, we're likely to be graced with a new architechture that is more geared for 2D/3D acceleration than what we have- and I'll be working in that playground as soon as it comes out (Maybe sooner- I'm debating on joining XFree86 to get earlier access to it to jumpstart the extended acceleration support...)
Bugs I can live with. Bugs I can help fix. I understand the code that is in the driver quite well- it was the interaction between Thomas' code and Mesa that was opaque to me (it appears that there was enough changes in the internals of Mesa from when the driver was written to the 3.1 release that it's obfuscated to anyone not intimately familiar with both version's internals). I don't mind spending a little time working with something to provide for the current installed user base while we wait for 4.0 to come out.
Does anyone happen to have working code?
on
Friday Quickies
·
· Score: 1
In the case of the Millenium driver, I happen to have the original code and I've been trying to make it happen with the 3.1 beta code. Do you know if anyone has managed to succeed where I've been failing (I have a few ideas to make it work "better" under the current way Mesa does things, but I've got to integrate it in and I've not had a lot of time to succeed in that regard!)?
About 3-6 months, I'd say...
on
Friday Quickies
·
· Score: 1
The anticipated release timeline for XFree86 4.0 is July of this year. At this time, PI plans on releasing the GLX/Multipipe Rendering Sample Implementation (SI, as they refer to it) at the same time as part of the 4.0 package. The sample implementation will be primitive and probably only support simple acceleration on a couple of adapters but provide all the pieces to make it way better. In 2-3 months after that, expect to see the full thing on XFree86. It's happening- and it couldn't have been at a better time (Games are coming to Linux and we NEED the acceleration NOW!)
Where can you get PCI models??
on
Friday Quickies
·
· Score: 1
I've looked, and couldn't find. As it stands, I happen to have the specs to the Permedia 2 (courtesy of an "oops" by TI) and I plan on using it since the performance of the Permedia is adequate and it's relatively cheap to obtain a card nowadays...
As for the Mystique register level info, it's not under NDA (I happen to HAVE them along with the Millenium I and II specs)- but as for drivers, it's like I said, the lack of interest is due to the fact that they should have never called these things 3D accelerators.
Mystique/Millenium 3D "suppoprt" is painfully weak
on
Friday Quickies
·
· Score: 1
I have the specs and I've been trying to reconcile what I have with Mesa and I'm afraid that it would fall under the dreaded category of "3d decellerators" if used with Mesa as it is.
The chips used in the Mystique and Millenium (I AND II) has support for Gouraud shaded and texture-mapped polys (and that's ALL it supports!), but you've got to do all the geometry pre-proc for it yourself (no z-buffering, no engine to make your meshes be "nice" for the rendering engine (almost every 3D card needs the triangles to be broken up so that part of their verticies land on a scanline so that it is easier for the engine to render), etc.) which makes the CPU have to do the work. In order for me to wring decent performance out of the Mystique/Millenium line of adaptors, I would have to re-write Mesa's entire architecture so that I'd be able to optimize the operations of GL to the card.
Simply put, there's an awfully good reason why they had a weak OpenGL MCD for NT and nothing else but DirectX. Like everyone else out there, Matrox was guilty of Marketing Sickness- they had to have a "3D offering" out there since everybody and his dog was doing it. The G200 is the first serious attempt at doing things the right way- for that alone, they have to be applauded. The fact that they're releasing the specs is another, indeed.
They have released the specs on their current chip (Used in Diamond's OEM low-end card, the SpeedStar A50. Since I don't have AGP (yet!), I can't say what this card will do quality and performance wise- but as soon as I do get an AGP motherboard, I'll announce what it can/can't do to the world!).
I hope they release the specs to the next generation version of their chip- it's supposed to be better than the previous design and supposedly supports DVD decoding right on the chip.
Would rather have the quality...
on
Friday Quickies
·
· Score: 1
It's a matter of quality over quantity.
- At 15 fps, the framerate is usable, but jumpy. - At 30 fps, the framerate smooths up and the game is very playable at that point. - At about 50-60 fps, most people will notice no difference with the 30 fps- those that do will like it and will see some benefit to it. - Past 60 fps, you will not see much of any difference, no matter what they say (It's past your body's ability to normally process the information...). It's down to "my card's better than yours" games at that point.
Simply put, if I see 30-60fps, and it's at least twice as good visual quality-wise, I'll take that over any 70+ fps any day. I do believe that quailifies it over the Voodoo- on top of them releasing specs, it's a all around winner in my book.
Just as often as the fsckups with stuff that genuinely should have been classified "ts" is the situations where something that should have been classified at a lower level got classified "ts" or "secret"- even though it really was confidential or unclassified in nature. It's a fscking nightmare to straighten out something incorrectly classified. (Which is part of the reason why things like the "ts" stuff getting out in the open by accident happen)
Then you aren't on a classified network.
on
Pentagon Cyber Wars
·
· Score: 1
Data classified secret or top secret cannot be accessed on a machine with outside access- no exceptions. Anything less is considered by the Federal Government to be a breach of security, punnishable by a minimum of $20,000 (possibly a LOT more) and possibly a 20 year stint in a Federal Pen.
Went and interviewed with their portables line (didn't work out- HR jinxed the deal) they had plaques with all the patent covers etched into them on their walls, something like a 50-100 of them- all for the notebook line. They do a hell of a lot of R&D for the portables line.
In order for this to work, they'd have to proprietarize Linux and WINE. Since both seem to be GPLed items, distribution would REQUIRE distribution of the changes to the source code that make it "MS Linux". And if you don't think that someone would come up with either a normal lawsuit or a class-action over the violation of the GPL that would be needed by MS to make it happen- you're sadly mistaken. And it would be ugly for MS to attempt such a thing- especially at this point in time.
You see, MS signed a few agreements as part of their spinoff of SCO that were required for SEC reasons- one of them is that they don't go into the Unix market- ever.
I was afraid it was a mistake, not as severe as the Linux as a "competitor" defense- but a mistake just the same.
Now, I know better- it was as gigantic a mistake on MS' part bringing Gassee's company into the picture as a "competitor". Whomever said Gassee was from the "scorched earth" school of business wasn't kidding.
They consider us to be worth making a production environment but not the player for the same? I would have thought if they had "got it" that we'd be seeing the player first. But hey, what do I know?
But FP performance is really only an issue if you're doing "Pentium" tricks- issuing an FP instruction and being able to out of order execute one or more integer ops while the FP unit was chugging on the request. More often than not, you only need these sorts of tricks on the rendering engine code- where 3DNow! has already been proven to shine over a stock PII at the same clock speed.
You see, MacOS cripples PPC performance. People have done off the cuff performance comparisons between PII's and G3's using Red Hat and Linux PPC. Integer performance is about on a par with the PII, floating point is faster on the G3.
You really can't do side by sides using Windows and MacOS on these processors- it's comparing apples and oranges (no pun intended).
Slashdot Mirror
The internal unit appears to be an Adaptec Ultra-Wide to the system. (We use Linux on one of these boxes as a production WWW server in an Enterprise Class operation that's about to be opened up to the world shortly...)
After all, Eddie was designed for doing HA/HP clustering of web servers. Use the bigger iron to handle the MySQL database engine and web serving and then have the old box handle just web serving.
I'd bet that would handle a hell of a lot more than it would by itself....
(Wiping egg off face...)
Hrm... Wonder where it could have been in the specs doc that I could have missed that one- I must have read the thing time and time again and didn't see it in the docs. Guess I need to look it over again (since Z-buffering would help out nicely in speeding things up...) and find it.
It won't do for me to continue making mistakes like this if I want to continue development of 3D support (and I do!)- thanks for the correction!
But, since it's proprietary (written under an un-needed NDA- the stuff's no longer a trade secret even if 3DLabs is still acting like it is!) I plan on extending the thing with non-closed-source code at some point.
Actually, this is all a moot point when XFree86 4.0 comes out, we're likely to be graced with a new architechture that is more geared for 2D/3D acceleration than what we have- and I'll be working in that playground as soon as it comes out (Maybe sooner- I'm debating on joining XFree86 to get earlier access to it to jumpstart the extended acceleration support...)
Bugs I can live with. Bugs I can help fix. I understand the code that is in the driver quite well- it was the interaction between Thomas' code and Mesa that was opaque to me (it appears that there was enough changes in the internals of Mesa from when the driver was written to the 3.1 release that it's obfuscated to anyone not intimately familiar with both version's internals). I don't mind spending a little time working with something to provide for the current installed user base while we wait for 4.0 to come out.
In the case of the Millenium driver, I happen to have the original code and I've been trying to make it happen with the 3.1 beta code. Do you know if anyone has managed to succeed where I've been failing (I have a few ideas to make it work "better" under the current way Mesa does things, but I've got to integrate it in and I've not had a lot of time to succeed in that regard!)?
The anticipated release timeline for XFree86 4.0 is July of this year. At this time, PI plans on releasing the GLX/Multipipe Rendering Sample Implementation (SI, as they refer to it) at the same time as part of the 4.0 package. The sample implementation will be primitive and probably only support simple acceleration on a couple of adapters but provide all the pieces to make it way better. In 2-3 months after that, expect to see the full thing on XFree86. It's happening- and it couldn't have been at a better time (Games are coming to Linux and we NEED the acceleration NOW!)
I've looked, and couldn't find. As it stands, I happen to have the specs to the Permedia 2 (courtesy of an "oops" by TI) and I plan on using it since the performance of the Permedia is adequate and it's relatively cheap to obtain a card nowadays...
As for the Mystique register level info, it's not under NDA (I happen to HAVE them along with the Millenium I and II specs)- but as for drivers, it's like I said, the lack of interest is due to the fact that they should have never called these things 3D accelerators.
I have the specs and I've been trying to reconcile what I have with Mesa and I'm afraid that it would fall under the dreaded category of "3d decellerators" if used with Mesa as it is.
The chips used in the Mystique and Millenium (I AND II) has support for Gouraud shaded and texture-mapped polys (and that's ALL it supports!), but you've got to do all the geometry pre-proc for it yourself (no z-buffering, no engine to make your meshes be "nice" for the rendering engine (almost every 3D card needs the triangles to be broken up so that part of their verticies land on a scanline so that it is easier for the engine to render), etc.) which makes the CPU have to do the work. In order for me to wring decent performance out of the Mystique/Millenium line of adaptors, I would have to re-write Mesa's entire architecture so that I'd be able to optimize the operations of GL to the card.
Simply put, there's an awfully good reason why they had a weak OpenGL MCD for NT and nothing else but DirectX. Like everyone else out there, Matrox was guilty of Marketing Sickness- they had to have a "3D offering" out there since everybody and his dog was doing it. The G200 is the first serious attempt at doing things the right way- for that alone, they have to be applauded. The fact that they're releasing the specs is another, indeed.
They have released the specs on their current chip (Used in Diamond's OEM low-end card, the SpeedStar A50. Since I don't have AGP (yet!), I can't say what this card will do quality and performance wise- but as soon as I do get an AGP motherboard, I'll announce what it can/can't do to the world!).
I hope they release the specs to the next generation version of their chip- it's supposed to be better than the previous design and supposedly supports DVD decoding right on the chip.
It's a matter of quality over quantity.
- At 15 fps, the framerate is usable, but jumpy.
- At 30 fps, the framerate smooths up and the game is very playable at that point.
- At about 50-60 fps, most people will notice no difference with the 30 fps- those that do will like it and will see some benefit to it.
- Past 60 fps, you will not see much of any difference, no matter what they say (It's past your body's ability to normally process the information...). It's down to "my card's better than yours" games at that point.
Simply put, if I see 30-60fps, and it's at least twice as good visual quality-wise, I'll take that over any 70+ fps any day. I do believe that quailifies it over the Voodoo- on top of them releasing specs, it's a all around winner in my book.
Just as often as the fsckups with stuff that genuinely should have been classified "ts" is the situations where something that should have been classified at a lower level got classified "ts" or "secret"- even though it really was confidential or unclassified in nature. It's a fscking nightmare to straighten out something incorrectly classified. (Which is part of the reason why things like the "ts" stuff getting out in the open by accident happen)
Data classified secret or top secret cannot be accessed on a machine with outside access- no exceptions. Anything less is considered by the Federal Government to be a breach of security, punnishable by a minimum of $20,000 (possibly a LOT more) and possibly a 20 year stint in a Federal Pen.
Went and interviewed with their portables line (didn't work out- HR jinxed the deal) they had plaques with all the patent covers etched into them on their walls, something like a 50-100 of them- all for the notebook line. They do a hell of a lot of R&D for the portables line.
Thank you for re-iterating what I've been saying all along- both to the hackers (me and my peers' crowd) and the suits (my boss' crowd).
I'd think more like at least 10 times that amount- 20-25's not enough to really wring things out for something this big.
;->
By the way, best of luck to everyone that applied- hope you all get lucky (and I'm planning on being lucky!
In order for this to work, they'd have to proprietarize Linux and WINE. Since both seem to be GPLed items, distribution would REQUIRE distribution of the changes to the source code that make it "MS Linux". And if you don't think that someone would come up with either a normal lawsuit or a class-action over the violation of the GPL that would be needed by MS to make it happen- you're sadly mistaken. And it would be ugly for MS to attempt such a thing- especially at this point in time.
You see, MS signed a few agreements as part of their spinoff of SCO that were required for SEC reasons- one of them is that they don't go into the Unix market- ever.
What else would they be doing? They sure as hell wasn't going to be supporting the site to benefit Linux.
I've got to be dreaming.
I was afraid it was a mistake, not as severe as the Linux as a "competitor" defense- but a mistake just the same.
Now, I know better- it was as gigantic a mistake on MS' part bringing Gassee's company into the picture as a "competitor". Whomever said Gassee was from the "scorched earth" school of business wasn't kidding.
Boies got him to out and out admit that they wanted Netscape to not compete with them- this alone is a violation of the Sherman Antitrust Act.
Rosen's the last nail in the coffin, I think...
Microsoft is so toast it's not even funny- and with the court antics, they're going to not win any appeals either.
They consider us to be worth making a production environment but not the player for the same? I would have thought if they had "got it" that we'd be seeing the player first. But hey, what do I know?
But FP performance is really only an issue if you're doing "Pentium" tricks- issuing an FP instruction and being able to out of order execute one or more integer ops while the FP unit was chugging on the request. More often than not, you only need these sorts of tricks on the rendering engine code- where 3DNow! has already been proven to shine over a stock PII at the same clock speed.
You see, MacOS cripples PPC performance. People have done off the cuff performance comparisons between PII's and G3's using Red Hat and Linux PPC. Integer performance is about on a par with the PII, floating point is faster on the G3.
You really can't do side by sides using Windows and MacOS on these processors- it's comparing apples and oranges (no pun intended).