3dfx/3dhq Revenge superchip.

[Geeforcer]

aqueel was responsible for them.

I wonder what technical background Aqueel has to produce a design that would theoretically leap so far ahead of the seasoned engineering talent from ATI, Nvidia, etc. Does anyone know if he has an EE degree and work experience related to video cards? He seems to be at the center of some key issues in this thread, but how much is known about his credibility in this area?

Aqueel (aquoess, Syed) has zero cred, a year ago he claimed to have made a new Ogl Icd for 3dfx cards that made Voodoo5 as fast as GF3 and also had software Aniso. The new Icd was to be released last may, but was never seen by anyone... Instead he started this *supersecret* Revenge project...

I remember a discussion Aqueel had with a former 3dfx-hardware engineer (Insider) at the x3dfx-forum. Insider didnt belive the claims of working Aniso for V5, Aqueel responded that he couldnt explain how it worked coz Insider only understood hardware and Aqueel only understood software...

Also Colourless (GlideXP) posted this about Aqueels programming abilities:

Back at the beginning when I first released glidexp, I had some conversations with him and he pretty much didn't really understand much anything in the glide sources. From memory he couldn't even grasp the concept of a pointer properly.

http://pub43.ezboard.com/fx3dfxfrm1.showMessageRange?topicID=14551.topic&start=41&stop=60

This is all good and well until you realize that a total ignorance of even the most basic software and hardware design procedures has never stopped anyone from designing a revolutionary microchip light years ahead of what world's 2 top companies with a thousand engineers between them can come up with.

 

[Ailuros]

The 32-chip setups were a (very) limited run, and had six-digit price tags.

I thought you said specifically that "I can get...", which indicates that they're available.

If the AAlchemy A5 8464 has a 887MPixels/sec fillrate at 22bpp with 4xRGSS, then the 32 chip setup would have the same fillrate with 8xRGSS. Should I cut them in half for 32bpp colour depth?

When cost is no consideration for professional equipment I'd be more intrigued by up to 2.42GPixel/sec at 32bpp with 16xRGMS (I am assuming that Clarity deals with MSAA) on the Independence1 instead.

 

[Tagrineth]

If the AAlchemy A5 8464 has a 887MPixels/sec fillrate at 22bpp with 4xRGSS, then the 32 chip setup would have the same fillrate with 8xRGSS. Should I cut them in half for 32bpp colour depth?

Doesn't quite get cut in half.

But with 8x RGSS, you don't need 32-bit as much. At 4x, 3dfx turns off their "22-bit post filter" altogether (no matter how you fiddle with your settings) because the 4x blend looks better. The only catch is stencil.

Then there's also the obscenely high memory bandwidth on that thing.

 

[Rekrul]

If this information from Aqueel is reported accurately:

(a quote from the first post in this thread)

We had got a board made, from a company that will be left un-named, and it was crazy. Aqueel tested in in Quake 3, and @ 1280x1024 with max gfx @ 8xAA it churned out 254FPS. Now, if you ask me that's faster than anything out there.

If this is accurate (from rashly concerning source for Revenge specs):

Aqueel was responsible for them

If this is true about Aquee l (from Tagrineth):

I think by now we know that Syed isn't the most trustworthy fellow around, anyway

And if there is validity to this:

aqueel was responsible for them.

I wonder what technical background Aqueel has to produce a design that would theoretically leap so far ahead of the seasoned engineering talent from ATI, Nvidia, etc. Does anyone know if he has an EE degree and work experience related to video cards? He seems to be at the center of some key issues in this thread, but how much is known about his credibility in this area?

Aqueel (aquoess, Syed) has zero cred, a year ago he claimed to have made a new Ogl Icd for 3dfx cards that made Voodoo5 as fast as GF3 and also had software Aniso. The new Icd was to be released last may, but was never seen by anyone... Instead he started this *supersecret* Revenge project...

I remember a discussion Aqueel had with a former 3dfx-hardware engineer (Insider) at the x3dfx-forum. Insider didnt belive the claims of working Aniso for V5, Aqueel responded that he couldnt explain how it worked coz Insider only understood hardware and Aqueel only understood software...

Also Colourless (GlideXP) posted this about Aqueels programming abilities:

Back at the beginning when I first released glidexp, I had some conversations with him and he pretty much didn't really understand much anything in the glide sources. From memory he couldn't even grasp the concept of a pointer properly.

http://pub43.ezboard.com/fx3dfxfrm1.showMessageRange?topicID=14551.topic&start=41&stop=60

If all the above is true, we're left with this:

Aqueel testing a custom Rampage board. Was he also the one responsible for getting it built? I find it implausible that a project like that would done as a "work for nothing" arrangement, so I'm assuming that it took a lot of money to get the Rampage card built, including the programming team for the custom drivers (Tag: quite a challenge for a team of only 4, btw. That represents a lot of code to write). Why spend that kind of money for an admittedly illicit project in the first place? What was to be gained?

And my favorite:

Revenge designed by a person who apparently does not understand hardware or software engineering. People may snicker at BitBoys, but at least they use qualified engineers to design their chips.

There's no evidence that convinces me of the validity of Revenge, the existence of the mysterious Rampage card or it's tested results because the credibility and competency of the primary source of this information is at issue. Much of this thread therefore falls into a "how many angels can dance on the head of a pin" type of discussion (i.e., the miraculous Rampage vs. 500 chip R350 vs. Godzilla...etc.).

 

[Geeforcer]

While we are talking about all of this, perhaps some could clear up another long-standing bit of 3dfxU lore:

Has anyone here (Tag, I am looking at you) finally wound out anything about the mysterious Hank Semenec, the supposed owner of SLI patent? Alternatively, does anyone know the patent number itself?

 

[Ailuros]

But with 8x RGSS, you don't need 32-bit as much. At 4x, 3dfx turns off their "22-bit post filter" altogether (no matter how you fiddle with your settings) because the 4x blend looks better. The only catch is stencil.

For a 3dfx fanatic, 32bpp will never be actually needed ROFL

Doesn't quite get cut in half.

So what? It's still low for today's standards.

Then there's also the obscenely high memory bandwidth on that thing.

<3GB/sec per chip. Peanuts compared to today's 11, 16 or even 19GB/sec bandwidth per chip. Besides I'd rather prefer say 15-20Gb/sec bandwidth with highly effective bandwidth saving techniques, then >80GB/sec raw bandwidth with no bandwidth saving features at all. May I remind you that each chip has also to process each triangle individually too?

anything about the mysterious Hank Semenec, the supposed owner of SLI patent? Alternatively, does anyone know the patent number itself?

Uh oh....hehehehe

 

[sonix666]

For a 3dfx fanatic, 32bpp will never be actually needed ROFL

And the same people, while obviously proven wrong, will start their crusade against floating point modes in modern video cards. And we'll not be able to convince them, while in the mean time more and more people will start getting disgusted with the amount of banding that more and more 32 bpp games show with their enormous amounts of texture layers and transparencies, especially in the dark.

 

[Tagrineth]

I didn't say 32bpp is unnecessary.

I said it's less necessary.

Basically what I meant was that 16-bit with 8x AA wouldn't look anywhere remotely near as bad as standard 16-bit, and would easily come close to 32-bit quality most of the time (of course there will be exceptions though)... UNLESS it has the semi-rare V5 6k's 4x/8x severe colour component loss. Colourless should know what I'm talking about...

But of course, 16-bit < 32-bit < 128-bit FP. You can come close through some trickery, but the standings shall always remain that way in all cases.

 

[Fafner]

While we are talking about all of this, perhaps some could clear up another long-standing bit of 3dfxU lore:

Has anyone here (Tag, I am looking at you) finally wound out anything about the mysterious Hank Semenec, the supposed owner of SLI patent? Alternatively, does anyone know the patent number itself?

If you wanna talk to Hank he posts at x3dfx now and then, for example this thread two days ago: http://pub43.ezboard.com/fx3dfxfrm2.showMessage?topicID=2909.topic

Faf

 

[OpenGL guy]

I didn't say 32bpp is unnecessary.

I said it's less necessary.

Basically what I meant was that 16-bit with 8x AA wouldn't look anywhere remotely near as bad as standard 16-bit, and would easily come close to 32-bit quality most of the time (of course there will be exceptions though)... UNLESS it has the semi-rare V5 6k's 4x/8x severe colour component loss. Colourless should know what I'm talking about...

You seem to be implying that AA can improve the results 16-bit color can give. If this is true for 16-bit color, then it should also be true for 32-bit color.

Please give us the math that shows that 16-bit color with 8x AA gives results close to 32-bit color with 8x AA... I think you'll have a tough time with this.

 

[parhelia]

I wonder what technical background Aqueel has

Here's Aqueel alledged "diploma" :

 

[rashly]

heh, i like the construction set thing on that diploma.

Has anyone here (Tag, I am looking at you) finally wound out anything about the mysterious Hank Semenec, the supposed owner of SLI patent? Alternatively, does anyone know the patent number itself?

hank posts over at x3dfx every once in a while. he knows ALOT. he got some data sheets on the vsa100 and we were discussing how to modify teh v5 to have 64 megs of ram per chip.

as for aqueel testing a spectre, what he told me was that he tested one left over from 3dfx which a former employees had who works at "some" company now. this was in summer 2002 and he told me 170fps 1600x1200x32 with 8xfsaa and 16 tap af in quake 3. he also said he took pics of the card (2 rampage, 2 sage) and was getting them developed. i heard nothing about teh pics since then.

as for the 3dfx fanboy argument, it really doesnt hold water anymore. it did at a time, but most 3dfxers moved on and DO realize other tech is alto better. no one would rather have a v5 than a r300. the things 3dfx users will say is that the v5 is nice to have because of old glide games, 22 bit looks alot better than 16 bit and its a good performance midpoint for the v5, and the v5 fsaa still is of the best quality (quality at removing jaggies, not performance). also, we do claim that rampage wouldve been awsome if it came out when it was supopsed to.

 

[Tagrineth]

You seem to be implying that AA can improve the results 16-bit color can give. If this is true for 16-bit color, then it should also be true for 32-bit color.

Please give us the math that shows that 16-bit color with 8x AA gives results close to 32-bit color with 8x AA... I think you'll have a tough time with this.

No, no, no, not compared to 32-bit with 8x. Just vanilla 32-bit versus 8x AA in 16-bit.

I can't provide the maths; in fact when there are threads talking about just what ATi's FP24 spec entails my head starts spinning ^_^;

I just remember that 3dfx, when asked how to turn on the 5500's '22-bit post filter', would answer how, then say if you are using 4x FSAA, the post-filter would be disabled because the 4x blend provided better IQ.

I said that as a response to people accusing me of using the old 3dfx 'WE DONT NEED 32 BIT COLOUR' argument for V3. It was true, for the most part, at the time, but since I got my Kyro II, I haven't ever used 16-bit in anything that gave me the option for 32-bit.

 

[OpenGL guy]

You seem to be implying that AA can improve the results 16-bit color can give. If this is true for 16-bit color, then it should also be true for 32-bit color.

Please give us the math that shows that 16-bit color with 8x AA gives results close to 32-bit color with 8x AA... I think you'll have a tough time with this.

No, no, no, not compared to 32-bit with 8x. Just vanilla 32-bit versus 8x AA in 16-bit.

Ok then. Show how 16-bit w/ 8x AA comes close to 32-bit color.

I can't provide the maths; in fact when there are threads talking about just what ATi's FP24 spec entails my head starts spinning ^_^;

C'mon, we're talking about 8 16-bit numbers and one 32-bit number.

 

[Tagrineth]

Ok then. Show how 16-bit w/ 8x AA comes close to 32-bit color.

I would show you a comparison of 16-bit, 16-bit with '22-bit post filter', 16-bit with 4x FSAA, and 32-bit... but my Voodoo5 broke two years ago. Care to fix it?

 

[OpenGL guy]

I would show you a comparison of 16-bit, 16-bit with '22-bit post filter', 16-bit with 4x FSAA, and 32-bit... but my Voodoo5 broke two years ago. Care to fix it?

You'd probably have better luck if you asked someone who knew something about the Voodoo 5

 

[Ailuros]

It was true, for the most part, at the time, but since I got my Kyro II, I haven't ever used 16-bit in anything that gave me the option for 32-bit.

16bpp colour depth quality on a TBDR is a chapter of it's own and would make a bad comparison anyway.

rashly,

....we do claim that rampage wouldve been awsome if it came out when it was supopsed to.

No one ever denied that. You know where there seems to be something rotten in the House of Denmark? See your own sentence here, picturing Aqueels claims:

....this was in summer 2002 and he told me 170fps 1600x1200x32 with 8xfsaa and 16 tap af in quake 3.

In conjuction with the absolutely ridiculous speclist of what "you" or "they" call Revenge.

If you pay close attention I haven't so far involved any individuals and their claims, from when I personally had discussions with them. I've got a few stories to tell on my own, but I don't because prementioned individuals are not present and I'm not really fond of digging up past arguments. Claims in them had the same flair of wild exaggerations and not one shred of credibility; and that is exactly the point where I disagree here.

 

[Ailuros]

Ok then. Show how 16-bit w/ 8x AA comes close to 32-bit color.

I would show you a comparison of 16-bit, 16-bit with '22-bit post filter', 16-bit with 4x FSAA, and 32-bit... but my Voodoo5 broke two years ago. Care to fix it?

There are more than a few people around that still have V5's either in their main or secondary rig.

Here you go:

http://www.pvrdev.com/eg/h/ITC.htm

65.5kb's download.

 

[Althornin]

so the R300/R350's have internal truecolor, eh?
I get none fo the hideous dithering artifacts my GF3 gets with my 9700

 

[rashly]

Ailuros,

i never said i believed those numbers. i dont believe them. i just said thats what he told me. i do believe that he saw a working rampage card though, but i think he exaggerated the numbers.