Hercules Thriller Conspiracy (early T&L card)

G

GinX

Newcomer
Hi all,

I'm relatively new to this forum but I've been lurking for years.

I thought I'd open a thread for a recent addition to my VGA collection.
It's one of those talked about cards that didn't make it into mass production so only a handful of prototypes were made.
One of them is this card:




Some marketing materials.

T-shirt that was handed out to people working on this project at that time:


The card is functional but it's picky about the mainboard. I got it to POST in an Intel R440FX UP server board and here as picture of the BIOS information:


Unfortunately I haven't found drivers for the Pinolite chip in order to fully test under Windows so if anyone has them or know where to source them please let me know.

Cheers,
GinX
 
You could try the latest Hercules beta drivers found here (I believe those are the latest drivers ever coded for an Hercules Rendition Rxxx board):

http://vintage3d.org/driver.php#sthash.2RqhtGfq.dpbs

Who knows, maybe the beta support Fujitsu's additional DSP. I remember seeing the card at a trade show but it's nevertheless interesting to know there is at least *one* out in the wild ;-)
 
I have mostly suppressed how primitive old PC electronics looks, compared to the incredibly densely packed devices we have today. Ugh, thanks Ginx, for reminding me! :)
 
Thanks for sharing this GinX.
That card really is among the rarest I've (not) seen yet. :) If you ever grow tired of collecting graphics cards, let me know. :)
 
Curious to see the Cyrix CPU in the promotional materials. Looks like by 1998 the 6x86 arch family was really hitting the bottom-end of the performance rankings and obviously was the perfect showcase for the benefits of HW TnL acceleration.
 
fellix said:
Curious to see the Cyrix CPU in the promotional materials. Looks like by 1998 the 6x86 arch family was really hitting the bottom-end of the performance rankings and obviously was the perfect showcase for the benefits of HW TnL acceleration.
Click to expand...

But the 6x86 probably had exceptional price/performance ratio, just like Verite. I can see them becoming good combination for value gaming computers, asuming the TnL would be cheap. Bigger problem would be game support.
 
msxyz said:
You could try the latest Hercules beta drivers found here (I believe those are the latest drivers ever coded for an Hercules Rendition Rxxx board):

http://vintage3d.org/driver.php#sthash.2RqhtGfq.dpbs

Who knows, maybe the beta support Fujitsu's additional DSP. I remember seeing the card at a trade show but it's nevertheless interesting to know there is at least *one* out in the wild ;)
Click to expand...

Thanks for the link although I think I had those drivers stored somewhere for my other Rendition prototypes. I'll give these a try but I don't expect much.
Fujitsu was probably providing the drivers for the Pinolite since they had separate boards with it internally.

CarstenS said:
Thanks for sharing this GinX.
That card really is among the rarest I've (not) seen yet. :smile: If you ever grow tired of collecting graphics cards, let me know. :smile:
Click to expand...

My pleasure. I also enjoy when people share their rare finds with the rest of the world.
I don't see that anytime in the near future I'll stop collecting. Over the years the passion has grown and I've extended a bit beyond just video cards :)

Putas said:
But the 6x86 probably had exceptional price/performance ratio, just like Verite. I can see them becoming good combination for value gaming computers, asuming the TnL would be cheap. Bigger problem would be game support.
Click to expand...

The marketing for this card addressed consumers with lower budgets who have slower non-MMX CPUs. These would benefit most from the T&L chip.
You can see in the link bellow the Hercules press release and targeted price of $149
http://web.archive.org/web/19981202023032/http://www.hercules.com/presale/release/conspiracy.htm
 
fellix said:
Curious to see the Cyrix CPU in the promotional materials. Looks like by 1998 the 6x86 arch family was really hitting the bottom-end of the performance rankings and obviously was the perfect showcase for the benefits of HW TnL acceleration.
Click to expand...

The Cyrix had quite ok integer performance but the fpu couldn't keep up per clock at all with the pentium, much less p6 core. And that PR233 cpu used in the marketing material had only some clock around ~180Mhz, so the fpu is at best 1/3 as fast as the one used in the compared cpus (PII-300)... And of course FPU is what was needed for cpu TNL.
 
If this card could've somehow launched in 1997, then it would've been amazing. But by mid-98 the window was closing for V2200 and slow CPUs like a 6x86. I think 3DNow is superior to Pinolite's throughput and K6-2 was quite cheap. Celeron was also faster and didn't need special programming.

The pricing for Conspiracy was supposed to be around $150. I remember Diamond started liquidating their Verite inventory in September 98 and I bought a Diamond Stealth II S220 for $50 direct from them.
 
They were looking for developers who had custom TnL to port to it, to show it off. As it had a full x86 processor the idea was that it was easy to take all our nasty custom TnL routines and offload them. A lot of us back then had integer TnL so it was a good fit (classic FPU divide whilst int ops stuff).

I don't think we ever got past the talking about it stage but vaguely recall talking about the interface and how we would communicate across the bus.
Essentially the idea was that we would just pass up state variables and run as much as we could of the render engine on the card itself.

Interesting concept but was killed by custom HW. In someways the fore-runner of larrabee and MIC ;)
 
That's news to me and I have done a metric ton of Fujitsu MB862xx research over the years... Maybe it's in the whitepapers and I've forgotten? Been a while since I looked at the MB86242/"Pinolite" documentation I've scratched together.

As I said over at VOGONS, epic acquisition! :)
 
The Cyrix was great, had it as a young kid, a P150+ at 120MHz. Revisiting the old DOS games and some I hadn't played before was fun. Everything 320x200 running at butter smooth 60fps or more, no more slideshows, no more loading times. And the Windows 95 stuff was working adequately.

Still, when we upgraded the box to an Intel P166 (which was already outdated, but we got it for very cheap and the board didn't take the damn split voltages for better chips) it was embarassingly faster at e.g. Quake.

We had a Virge btw. Would have loved a full version of the Virge Descent 2 and Terminal Velocity that came with it.. Else Virge was a great value when considering the 2D performance (just a couple years before that was the territory of esoteric high end 2D cards or Matrox Millenium)
 
Last edited by a moderator:
I had a couple of those 6x86 back then for less than a year, so not much memories about it. Looking now at some archived reviews from Anand and Tom's Hardware, the thing was really slow in anything that required intensive math computations. The FPU was essentially ripped off from a generic 486 design -- non-pipelined and high instruction latency. AMD's K6 at least offered low-latency for its x87 implementation and generally could keep with the clock-rates, while 6x86 had to use the stupid PR rating for the entire product life. The thing was cheap though, but at least here in my country all the rage then was K6 and K6-2 -- those were sold by truck loads in the mainstream channels.
 
6x86 were available before the K6. 1996 vintage, woo! The K5 would have competed with it, but it was so late and so low clocked (except the very last ones) that its existence is barely known at all.

Then the poor Cyrix follow-ups ended up as bottom of the barrel, as was said. The final humiliation, after Cyrix was bought, was the "Cyrix III" processor that got released, which was in fact a Centaur Winchip variant later known as VIA C3.

In 96 or 97 you could still find yourself running some 3D game or engine that only made use of integers (or fixed point), even many new DOS games were still published. The first Quake was very recent and was the first game really built around a fast FPU (leaving good old DX/2 66 toiling at 5 fps or so)
 
Last edited by a moderator:
Stiletto said:
That's news to me and I have done a metric ton of Fujitsu MB862xx research over the years... Maybe it's in the whitepapers and I've forgotten? Been a while since I looked at the MB86242/"Pinolite" documentation I've scratched together.

As I said over at VOGONS, epic acquisition! :)
Click to expand...

I might have the wrong codename etc. (hell even the wrong vendor) but putting a cheap 6x86 onto the card was talked about with us in the late 90's. I remember because they were particularly looking for Custom TnL games.
 
For those who don't remember the old discussions on the Hercules Thriller Conspiracy or the Fujitsu "FGX-1"/"Pinolite" MB86242:

“Fujitsu Develops World’s First Three Dimensional Geometry Processor,” press release, Fujitsu
Fujitsu press release 1997-0139, July 2, 1997.
https://pr.fujitsu.com/jp/news/1997/Jul/2e.html
https://pr.fujitsu.com/jp/news/1997/Jul/2.html

“Geometry Processor for PC 3D Graphics : Pinolite”, Fujitsu, 49, 2, pp.103-106 (03 1998) (Japanese)
http://img.jp.fujitsu.com/downloads/jp/jmag/vol49-2/paper03.pdf

There is also an English translation of the above Japanese paper in IEICE Trans. Electron., Vol. E81-C, No. 5 May 1998.
"3D Graphics Geometry Processor for PC", Makoto Awaga, p.733-736
http://search.ieice.org/bin/summary.php?id=e81-c_5_733&category=C&year=1998&lang=E&abst=
I have a copy of this but it's easy enough to obtain.

There is also a document reference mentioned in US patent US6854003 (regarding a Hyundai processor) as “MB86242 <PinoliteTM> Product Specifications,” Fujitsu, Version 1.0, (1997) - which while the patent is easily accessible, I do not believe I have the document referenced at the moment.

That's all the info I still have AFAIK.

According to the Fujitsu MB86291 "Scarlet" datasheet, two Pinolite processing cores with enhanced rendering cores were integrated into the MB86291, along with an enhanced Cremson core (x5 previous performance), dedicated video capture/scaler and embedded wide-bus band SDRAM.
http://www.datasheetarchive.com/MB86291/Datasheets-SW10/DSASW00188342.html
http://www.datasheetarchive.com/mb86291-datasheet.html
 
Last edited by a moderator:
Stiletto said:
For those who don't remember the old discussions on the Hercules Thriller Conspiracy or the Fujitsu "FGX-1"/"Pinolite" MB86242:
Click to expand...

Thank you for all the links.

It's unfortunate these cards never made it to mass production but it's interesting reading about the technology behind them.

From what I've been able to gather these boards were handed out to select partners for the purpose of developing/testing applications that would take advantage of it but I still couldn't find the drivers they used.
 
You must log in or register to reply here.
Back
Top