Not possible. Firstly nvidia doesn't manufacture the chips, second I doubt anyone has the tools required to manufacture the old VSA100 chips on the specific 250 nm manufacturing process, third finding the specific SDR ram chips would be difficult. Then there is the whole issue of PCI/AGP.

It would cost hundreds of thousands of dollars to retool the labs to manufacture the chips.

If you then went crazy and finished the VSA101 chip for DDR, then that is another process entirely, you may as well just keep selling the 980Ti for serious profit.

Companies are all about making money. If nvidia thought there was any profit in it they'd probably do it. You might sell 5,000 cards tops and then the market would be flooded.

Tetrium wrote:

Yes there would, money! If old ones would sell for $400, then there would be market for manufacturing new ones.

No. Actually quite the opposite. They sell for that amount because they are rare and people think that they are "valuable" collectables. Nobody in his/her right mind will pay that much just to use it in a retro rig (I'm not saying that I'm in my right mind, not always 😊 ). There are better alternatives, already.

When you mass produce them, there won't be any rareness, coolness, and (consequently) "collectability" left, and the price will be plummeting down.

In best case, the original old batch still be collectable (and selling for ridiculous prices, to their collector audience) and the new batch will be selling for a dime to "users" like me and you.

tayyare wrote:

When you mass produce them, there won't be any rareness, coolness, and (consequently) "collectability" left, and the price will be plummeting down.

In best case, the original old batch still be collectable (and selling for ridiculous prices, to their collector audience) and the new batch will be selling for a dime to "users" like me and you.

And don't forget they were never that good to begin with. They were outperformed by competing cards when they were released, and didn't even support hardware T&L when the competition had done so for a while. They are only good for Glide games, and at this day and age I'd take a cheap and faster Geforce4 MX for DX and OGL any day and give my old PSUs some rest.

tayyare wrote:

No. Actually quite the opposite. They sell for that amount because they are rare and people think that they are "valuable" colle […]
Show full quote

Tetrium wrote:

Yes there would, money! If old ones would sell for $400, then there would be market for manufacturing new ones.

No. Actually quite the opposite. They sell for that amount because they are rare and people think that they are "valuable" collectables. Nobody in his/her right mind will pay that much just to use it in a retro rig (I'm not saying that I'm in my right mind, not always 😊 ). There are better alternatives, already.

When you mass produce them, there won't be any rareness, coolness, and (consequently) "collectability" left, and the price will be plummeting down.

In best case, the original old batch still be collectable (and selling for ridiculous prices, to their collector audience) and the new batch will be selling for a dime to "users" like me and you.

~ Yeah ~
That's why the PT Cruiser (a retro 1930's body style) sold so poorly.

mwdmeyer wrote:

Not possible. Firstly nvidia doesn't manufacture the chips,

The point was they could if they wanted to.

Personally I'm thinking along the lines of Voodoo 2 and Voodoo 3 reproductions for retro PCs. Not Voodoo 6.

mwdmeyer wrote:

It would cost hundreds of thousands of dollars to retool the labs to manufacture the chips.

If that were true there would not be brand new video cards for $30-$35, yet there are many.

A 100% markup would be easy to achieve for a niche product like this and my guess is it would be sustainable.
.

Yes it is called economy of scale. Everyone (for video cards) is on tsmc 28nm at the moment therefore it is cheap.

Go read up on chip fabrication, Intel spend billions of dollars ripping out and replacing their fabs every few years.

They wouldn't have the old fabs around for 250nm chips anymore.

A voodoo 2 would cost even more to manufacture as it has multiple chips and isn't integrated like the VSA100.

I agree with everything that's been said since my last reply, including the ones that don't agree with me or correct my statement, as I think this is in fact correct.

But bear in mind I stated if these would sell for $400, and it's obvious the manufacturing price would have to be a whole lot lower in order for such a modern Voodoo card to be economically viable.

I am very aware that with hardware this age, it's mostly demand and supply and that with increased supply, prices will drop (and will certainly drop below the imo outrageous price of $400), but a simple glide card costing maybe $50 or so might get some attention.

And even if there were some kind of Voodoo 2016 card (I'm just throwing in some random name here 😜) that the older original stuff will still hold collector value.

Btw, I would be less surprised if some cheapo PC case manufacturer would consider manufacturing new AT cases. These wouldn't even need to come with all the blingbling and features that modern cases have (perhaps except for a 3-digit LED display 😊 ) and even new ones of cheap quality might become viable in perhaps only a few years (if only because of the absence of any yellowing and because one wouldn't have to seek out second hand ones which would be a pain to find, let alone have shipped to you).

These AT cases might even be made compatible with mATX boards (I know, I'm stretching things a bit more here), but since PC cases are mostly just metal, it wouldn't require a multi-million dollar fab to manufacture those.

Interesting thoughts, hehe

PCBONEZ wrote:

The point was they could if they wanted to.

Personally I'm thinking along the lines of Voodoo 2 and Voodoo 3 reproductions for retro PCs. Not Voodoo 6.

Personally I was thinking more a Voodoo 4.
There were ones capable of Universal AGP slotting while the rest wasn't, the Voodoo 4 is a single GPU card and the lack of a second GPU (or the lack of 3 more, irrelevant really) could make AGP implementation simpler and thus cheaper, the Voodoo 4 did support 32 bit and there was the option to increase the total amount of memory on Voodoo 4s, right? V3 was stuck with 16MB iirc

Voodoo 2 would also be interesting for retro gamers, but I'm not sure how practical it would be to manufacture a newer version since the V2 seems like a more complicated card than the V4

tayyare wrote:

Tetrium wrote:

Yes there would, money! If old ones would sell for $400, then there would be market for manufacturing new ones.

No. Actually quite the opposite. They sell for that amount because they are rare and people think that they are "valuable" collectables.

This particular case is a little more complex. I live in Argentina. There are many idiots and/or ignorants and/or oportunists who thinks they can sell any crap for ridiculous prices and live an entire month with it using the "collectable" speech. In most of this cases, the items are far from what you can see as mint condition in ebay or some sites like that. And our culture and economy is so far away from yours, so nobody would spend such amount for old things. At the end, these vendors never sell anything.

mwdmeyer wrote:

They wouldn't have the old fabs around for 250nm chips anymore.

Well, would it be that hard to create a die-shrink instead of just producing the old chips as is?

Sure it's technically possible but you'd not make your money back on 5,000 cards. The only possible way todo it cheaply is to find a bunch of unused VSA100 chips and just manufacture the physical board.

mwdmeyer wrote:

They wouldn't have the old fabs around for 250nm chips anymore.

A voodoo 2 would cost even more to manufacture as it has multiple chips and isn't integrated like the VSA100.

No.
Why on earth would they do it that way. The only functions need reproduced, not the original chip.
They could build the functions on the present die size without having to retool anything AND by doing so they could do it one chip.
It would in fact be inexpensive.

I'm familiar with Intel fabs.
I was offered a QA job in one and I got 'the big tour' when they were trying to recruit me.
I have several former coworker friends that took jobs there and that still work there.
Hillsboro Oregon was not working for me or I'd probably be there right now.
The idea of working in a clean room and wearing a f*cking poopie suit all day didn't thrill me either.
.

Right so you'd have a whole engineering team just to create a new design for 5,000 chips!

mwdmeyer wrote:

Right so you'd have a whole engineering team just to create a new design for 5,000 chips!

5,000 is arbitrary and your own creation.

It would be nothing more than a mod to an already existing design.
Not even a big deal.
.

I disagree, if it really was that easy we'd have seen it by now.

Maybe if you could program an existing FPGA to emulate it, but then that isn't the real thing now anyway.

mwdmeyer wrote:

I disagree, if it really was that easy we'd have seen it by now.

Maybe if you could program an existing FPGA to emulate it, but then that isn't the real thing now anyway.

We agree that we disagree then.

You haven't seen it because corporate suits are generally single minded morons.

I've worked in the high-tech engineering field (including manufacturing) and it wouldn't even be difficult to pull off.
.

Initial costs for spinning the ASIC would be substantial.

mwdmeyer wrote:

Sure it's technically possible but you'd not make your money back on 5,000 cards. The only possible way todo it cheaply is to find a bunch of unused VSA100 chips and just manufacture the physical board.

Well, apparently you can get the VSA100 chips from here:
http://www.ebay.com/itm/3DFX-VOODOO-VSA100-CH … 4AAAOSwPe1Tz-PL
"Price:US $6.80"

Dunno how many this seller actually has though

Yep that would do it. I have a couple of trays of new Voodoo 2 texture unit chips somewhere I purchased a while ago. Got them for less than $1 each.

Still you'd need to design the PCB and source the other components (RAM, DAC if for Voodoo 2, glue logic etc), once you've done that you'd need to get a company in China to manufacture them, and for small runs it isn't always that easy/simple. Not exactly a cheap or quick project! Although much more possible than designing the chip.

Going back on topic, if the S3 card is still available you could go for that still. Not fantastic, but it would at least be an improvement over the Trident. Or if you want 3D capabilities and you find one for cheap/free a Trident Blade 3D would not be terrible, at least compared to what you currently have 😀 and should be fairly common. EDIT: Not SUGGESTING a Trident Blade, just saying that they are not as terrible as I expected them to be 😊