VOGONS


Unconventional 3dfx build

Topic actions

First post, by nolimitsoya

User metadata
Rank Newbie
Rank
Newbie

I always regretted not getting a Voodoo 4/5 back in the day. I went straight from a Voodoo 2 to a GeForce 256 Pro (DDR) that - while at the time positively beastly and all - left me locked out of some of my favourite Glide API games. Eventually Glide wrappers became a thing and I could forego the Voodoo2, with increased resolution and performance as a result. However, the V5 5500 that never was always stuck to my mind. Especially after 3dfx folded.

I've recently got a chance to remedy this, and have been collecting parts for a while now. This will not be period correct. Nor will it be a super high performance time machine build. It might end up being an imagined and evolved side system that a younger me would have dreamt of letting live alongside my main modern system, only to play the occasional Longbow 2, EF2000, Redline Racer or NFS3 session on.

Not all decisions are final yet, and I've accumulated way more hardware than I'll end up using.
Only three components are so far set in stone:

3dfx Voodoo 5500 with Heatkiller copper/black acetal blocks, water cooled!
Dell 2007FP 1600x1200 native LCD
Chieftec Dragon midtower case that I picked up in poor-to-fair condition. I've so far had a most unpleasant experience removing old bitumen mats that were glued on everywhere, and started to mod it to accomodate the water cooling system. It'll need polishing and buffing at some point, as it's been quite badly scratched up by the previous owner, but it's salvageable.

What's not quite decided on yet, but still weighing the options:
Underclocked Athlon on KT333 mainboard, unless VIA chipset gremlings pushes me to a P2/P3 440BX setup that I've acquired as a backup. Either most likely water cooled using a modified copper/black acetal EK Supreme LT block.
Audigy2 or Live! - I have both, but the Audigy2 has a nice front bay I/O like the one I had for my Live! Platinum back in the day. It probably wont be used much, but if there are no downsides to the Audigy2 over the Live! I'd quite like the front bay for nostalgic reasons. I was madly in love with the front bay back then. If both turn up too many compatibility issues I may hunt down an AWE64 to go with the 440BX system mentioned earlier.

This is the current state of modding the Chieftec case:

20231128_213058.jpg
Filename
20231128_213058.jpg
File size
1.6 MiB
Views
1874 views
File license
Public domain
20231128_213108.jpg
Filename
20231128_213108.jpg
File size
756.1 KiB
Views
1874 views
File license
Public domain

I ended up adapting an aquacomputer bracket as a mount as there were to many features on the front face of the case to simply cut a nice radiator mount straight into it.

20231128_213121.jpg
Filename
20231128_213121.jpg
File size
815.25 KiB
Views
1874 views
File license
Public domain

While shopping i found these black nickel shields that I think look quite nice! On the topic of black nickel, I've been debating black nickel or chrome for the water cooling fittings, but I'm actually leaning toward chrome at the moment...

Reply 1 of 23, by nolimitsoya

User metadata
Rank Newbie
Rank
Newbie

The V5500 has had a full recap and is since then outputting excellent video. Pre recap it did have some artifacting in the form of jittery horizontal lines and generally looked quite soft and blurred. That is now all fixed and it's crisp and flawless!

20231128_235515.jpg
Filename
20231128_235515.jpg
File size
1.15 MiB
Views
1838 views
File license
Public domain
20231128_235526.jpg
Filename
20231128_235526.jpg
File size
1.08 MiB
Views
1838 views
File license
Public domain

I just recently fitted the Watercool/Heatkiller blocks together with two Enzotech chipset cooler "backplates" on the back. Completely overkill, but I like how it complements the mounting screws that overwise look a bit lonely and out of place. I'm also a bit partial to the look of big chonky chunks of copper on old hardware. 😁 I plan to tie it into a bit of a theme with Enzotech VRM coolers on the motherboard. Possibly a northbridge cooler as well. There are 2 mm thermal pads under the Enzotech coolers to provide electrical insulation and heat transfer. The springs and thumbscrews will probably have to sit on the back like this as they'd interfere with the plumbing on the front.

I'm still waiting for '90s green heatsinks for the RAM and a batch of retro fittings. Hopefully they'll arrive within the next few weeks.

20231128_234358.jpg
Filename
20231128_234358.jpg
File size
943.37 KiB
Views
1838 views
File license
Public domain

As free breathing as a Dragon chassis is going to get.

20231128_235310.jpg
Filename
20231128_235310.jpg
File size
1.01 MiB
Views
1838 views
File license
Public domain

Front panel still fits!
The 240x60 mm Bykski radiator takes up quite a bit of room in the front, but still less than the drive cages used to. The brackets for the FDD may only look hot glued from this angle, but they are double riveted to the 5.1/4" cage with 4.7 mm blind rivets and very sturdy. The hot clue was for finding the positioning and keeping them in place for drilling and riveting.
I'll only be able to use one 3.5" bay as the radiator is blocking the other, and even then it's a tight fit for the radiator. There is only a single millimeter of space between it and the bottom, and about three towards the FDD.

Reply 6 of 23, by nolimitsoya

User metadata
Rank Newbie
Rank
Newbie

Fittings and sinks have arrived!

20231219_124827.jpg
Filename
20231219_124827.jpg
File size
1.22 MiB
Views
1181 views
File license
Public domain

I did a quick mockup just to see how it will turn out. I'm not sure I'm sold on how the red motherboard goes with the V5500, visually. It worked out better in my head. The case fan, however, is a Yate-Loon with quite strong blue LED lighting that I'm still hoping will take the edge out of the contrasting colours once the system is turned on. What do you guys think?

20231219_123039.jpg
Filename
20231219_123039.jpg
File size
1.3 MiB
Views
1181 views
File license
Public domain
20231219_124424.jpg
Filename
20231219_124424.jpg
File size
1.76 MiB
Views
1181 views
File license
Public domain

I'm more and more warming up to plan B...

20231219_124456.jpg
Filename
20231219_124456.jpg
File size
1.81 MiB
Views
1181 views
File license
Public domain

But, in the mean time!

20231205_115254.jpg
Filename
20231205_115254.jpg
File size
540.12 KiB
Views
1180 views
File license
Public domain

Reply 8 of 23, by nolimitsoya

User metadata
Rank Newbie
Rank
Newbie

Not as such, but I have enough blocks laying around that I could surely find something that can be adapted to fit Slot1 with a bit of work. The block pictured for the s462 system is actually a s775 block, originally.

Reply 9 of 23, by H3nrik V!

User metadata
Rank Oldbie
Rank
Oldbie

I wouldn't trust that water cooler setup - you have absolutely no control whatsoever over the flow - you may only have flow in one block with the other totally stale. I would definetly have the flow serial, to ensure equal and adequate flow in both.

Please use the "quote" option if asking questions to what I write - it will really up the chances of me noticing 😀

Reply 10 of 23, by nolimitsoya

User metadata
Rank Newbie
Rank
Newbie

Parallel flow is no problem at all and even used to be the most common way of connecting SLI setups. The water is flowing under pressure and will always slow down/lower pressure if possible. Meaning, that if there are multiple paths of equal flow resistance the water flow (and pressure) will be divided equally between the two paths.
You can even use this to your advantage if, for some reason, you want lower flow through part of your loop, adjusting flow rate by matching tubing size to the block flow resistance in each branch to get the desired ratio. It can be very helpful in loops including small, restrictive blocks like VRM/VRAM/SSD blocks that you definitely do not want limiting the flow rate of the entire loop and where cooling requirements are lower.

Think of the blocks as resistors in an electrical circuit if that makes more sense. You wouldn't ever have current through only one of two resistors connected in parallel.

Reply 11 of 23, by H3nrik V!

User metadata
Rank Oldbie
Rank
Oldbie
nolimitsoya wrote on 2023-12-20, 10:41:

Parallel flow is no problem at all and even used to be the most common way of connecting SLI setups. The water is flowing under pressure and will always slow down/lower pressure if possible. Meaning, that if there are multiple paths of equal flow resistance the water flow (and pressure) will be divided equally between the two paths.
You can even use this to your advantage if, for some reason, you want lower flow through part of your loop, adjusting flow rate by matching tubing size to the block flow resistance in each branch to get the desired ratio. It can be very helpful in loops including small, restrictive blocks like VRM/VRAM/SSD blocks that you definitely do not want limiting the flow rate of the entire loop and where cooling requirements are lower.

Think of the blocks as resistors in an electrical circuit if that makes more sense. You wouldn't ever have current through only one of two resistors connected in parallel.

Agreed, but in the electrical circuit, I would know the two resistances, I would never trust those t-junctions TBH 🤣

Please use the "quote" option if asking questions to what I write - it will really up the chances of me noticing 😀

Reply 12 of 23, by CwF

User metadata
Rank Member
Rank
Member

Some bogus assumptions here-
Temp along the way will drastically alter flow. The parallel circuits assumed equal will never be. A single hot spot could totally stall flow in a branch.

I used to know what I was doing...

Reply 13 of 23, by chrismeyer6

User metadata
Rank l33t
Rank
l33t

Jay from Jayztwocents has proven multiple times over the years that there basically no difference between serial or parallel flow in a water loop. He's also shown that loop order doesn't matter as well. My experience with custom loops over the years mirrors his

Reply 14 of 23, by nolimitsoya

User metadata
Rank Newbie
Rank
Newbie
CwF wrote on 2023-12-20, 16:00:

Some bogus assumptions here-
Temp along the way will drastically alter flow. The parallel circuits assumed equal will never be. A single hot spot could totally stall flow in a branch.

Respectfully, no.

Reply 15 of 23, by CwF

User metadata
Rank Member
Rank
Member
nolimitsoya wrote on 2023-12-20, 17:53:
CwF wrote on 2023-12-20, 16:00:

Some bogus assumptions here-
Temp along the way will drastically alter flow. The parallel circuits assumed equal will never be. A single hot spot could totally stall flow in a branch.

Respectfully, no.

Practically, you're right. Overwhelming flow volume means you may never hit the condition. But that doesn't mean your not leaving the possibility open to see 'vapor lock'. Technically, 'never equal' doesn't mean 'not good enough'.

I used to know what I was doing...

Reply 16 of 23, by nolimitsoya

User metadata
Rank Newbie
Rank
Newbie

If a V5, or any other components for that matter, is ever close to boiling the water and creating a "vapor lock" its long gone. This is not a condition relevant to computer cooling and I have no interest in debating hypothetical bro-science. Thank you.

Reply 17 of 23, by CwF

User metadata
Rank Member
Rank
Member
nolimitsoya wrote on 2023-12-20, 19:20:

If a V5, or any other components for that matter,

You mean on a low wattage V5 it won't matter. Plenty of chips hit 100C, so generally...

I used to know what I was doing...

Reply 18 of 23, by Trashbytes

User metadata
Rank Member
Rank
Member

IIRC the coolant in a custom loop wont boil at 100C in any case, pretty sure modern coolants have a much higher boiling point than plain distilled water. Anyone using just distilled water is either mad or has enough knowledge of their components and setup to never have a situation where vapour lock can occur. This aside if the VSA100 ever hits 100c its pretty much toast well before that and the OP has much bigger issues than a Vapour lock.

-Edit Yes I know most AIOs use distilled water but we are not discussing a CLC where its one component being cooled.

The setup OP is using looks perfectly fine for a Voodoo 5 and would have been a ballin setup back in the day.

Reply 19 of 23, by nolimitsoya

User metadata
Rank Newbie
Rank
Newbie
Trashbytes wrote on 2023-12-21, 03:39:

IIRC the coolant in a custom loop wont boil at 100C in any case, pretty sure modern coolants have a much higher boiling point than plain distilled water.

Marginally at best. The additives are for reducing corrosion and growth, not to do with temperature. The main mistake i think is made here, is that just because some minor part of a modern chip is above a 100°C it does not mean that the coolant will be subjected to anything near that. Your single gram, or cubic millimeter, or whatever small unit of silicon transfers that heat (J, not C!) to copper that is not only tens or hundreds the mass and volume, but has a higher molar heat capacity as well. Again, energy is transferred - not temperature!
The coolant in a reasonably well circulated and cooled modern loop is not above 30-35°C and the side of the blocks in contact with the coolant is not much hotter than this. Then it's a gradient through the layers of copper, interfaces, heatspreaders and silicon towards that chipset hotspot.