Glaze3D - A Bit of History

[Dave]

I have been sitting on an extra Glaze3D engineering sample for several years now from my days at Bitboys. As this was never released there is a bit of history surrounding it and I personally consider it to be a bit of a collectors item. I was considering putting one on Ebay if anyone is interested in that sorta thing, but before doing so I wanted to see if there was any interest.

Thanks,
Dave

 

[XMAN26]

I have been sitting on an extra Glaze3D engineering sample for several years now from my days at Bitboys. As this was never released there is a bit of history surrounding it and I personally consider it to be a bit of a collectors item. I was considering putting one on Ebay if anyone is interested in that sorta thing, but before doing so I wanted to see if there was any interest.

Thanks,
Dave

I'm interested in knowing more about what bitboys was trying to build but never seemed to get beyond drawing board.

 

[Dave]

I'm interested in knowing more about what bitboys was trying to build but never seemed to get beyond drawing board.

Well I think the point here is that it was well beyond the drawing board. What would you like to know?

 

[XMAN26]

Well I think the point here is that it was well beyond the drawing board. What would you like to know?

What levels of performance were you getting against which hardware? What size of edram were in use? what type of interconnect? 64/128/256/512/1024 and was it 1/4, 1/3, 1/2, 2/3, 3/4 or full speed compared to core clock? transistor counts? memory sizes? PCB size? Why it never made it to market as it seemed like they had something that could have been very fast under most conditions compared to cards out at the same time talk of the boards were?

 

[Dave]

What levels of performance were you getting against which hardware? What size of edram were in use? what type of interconnect? 64/128/256/512/1024 and was it 1/4, 1/3, 1/2, 2/3, 3/4 or full speed compared to core clock? transistor counts? memory sizes? PCB size? Why it never made it to market as it seemed like they had something that could have been very fast under most conditions compared to cards out at the same time talk of the boards were?

Most of this information was public information. If you pull up any old articles on Glaze you should be able to find the answers.

It never made it to market because it was the first chip Bitboys completely made on their own and a number of unexpected issues came up along the way. By the time it was completed it still had a number of advanced features but other areas had it lagging behind. So we moved on to a next-gen product that was actually progressing very well and would have been out in a reasonable time frame and been competitive. However because of various reasons the focus was changed to the mobile market.

 

[JoshMST]

Heh, do working drivers come with it?

 

[_xxx_]

Yeah and how about Vista drivers?

 

[Rys]

Surely it needs to find a home in the Beyond3D museum? So if you're interested in selling, I'm certainly interested in buying.

 

[Geo]

I can think of at least one Editor who started salivating when he saw this thread. . .

 

[AlNom]

Vote for BitsBoy article

 

[Dave]

Heh, do working drivers come with it?

Ha, no. Keep in mind it is just the GPU itself, not a whole board. They didn't give me one of those.

 

[Dave]

Vote for BitsBoy article

I wrote one a long time ago on FiringSquad about Bitboys history and what had taken place up to that point. I'm not sure if it is still there anymore, but you should at least be able to find it in an archive on google or something.

 

[AlNom]

Thanks for the hint, found it. (The images still work too)

 

[swaaye]

The story is pretty sad. So many disappointments and so many close calls that fell through... And the worst part is that they designed a few generations of very interesting stuff that could've seriously influenced the market if it had worked out.

 

[Gunhead]

Dave, what can you tell about the Thor geometry/master chip for the original* DX6 Glaze3D?

There were hints at checkerboard-style tiled load-balancing (for two or fours rasteriser chips), but not much beyond that. How far was it designed -- did it even reach the drawing board or was it just a bunch of concepts and sketches?

And did it initially include TnL or was that added to the concept/design later? (I remember some talk of schedule moved back due to addition of TnL to Glaze3D chip itself, don't know how this relates to Thor though...)

How would it have compared to say a Geforce 256 in geometry acceleration? (I have no idea how useful it was in the Geforce actually -- saw conflicting reports and reviews, opinions ranging from "kicks ass in games" to "too weak to be useful", so really don't know.)

(* Actually I hazily recall the very first version was an eDRAMless "Glaze3D 400" design with Rambus memory, but I suppose this got abandoned before the design progressed anywhere.)

Thanks for any info or just anecdotes you can share -- I remember how it was all a big mystery then and Thor was the most mysterious piece of it all, heh

 

[Pressure]

I remember how we used to ponder about Glaze3D at Rage3D and the later chips. Hopefully ATI, ehm AMD, got some cool technology by the take-over in May 2006.

Would really have liked a third competitor in the high-end market of graphic cards and it sounded like they were close in several aspects.

 

[Tahir2]

Anyone remember this interview?

A few years later I did ask them for another interview but at that time the companies direction had changed and the Bitboys in Finland kindly declined.

From my experience, they were/are a top bunch of lads.

 

[Dave]

Gunhead,

I really didn't know much of anything about Thor because by the time I arrived that had come and gone. I did ask though and because it has been so long it seems a bit hard for others to remember the details as well! A hardware model was never written, only simulation scenarios. Sorry for the lack of detail here, it has just been a long time!

As for Rambus, it is correct that the original plan was to go in that direction. When the use of EDRAM became a viable option that was dropped.

 

[Dave]

Tahir2,

They were definitely top notch. They were very intelligent and knew what they were doing, but beyond that they were just great to work with and hang out with.

 

[Dave]

Pressure,

The purpose of the acquisition was to do with the mobile designs that had taken place, which were EXTREMELY effective. It was really amazing what we were doing with how many transistors it was being done. They did pickup stuff from the next-gen design on the desktop end too, but only a couple of things would likely have been of interest at that point (Matrix AA and a patented means of addressing occlusion culling).