Something else you can do is finding another card, doesn't have to be the same brand or model, with the sockets filled and use those chips 😀

They might be the same as on this card
http://www.ebay.com/itm/111720622017

KM416C2568J-7
or
S416257DJ-06

but I can't find a source for those.

I recently bought a pack of 10 from Aliexpress, worked perfectly in my S3 Virge. 35ns chips.

$5.70 US plus any shipping or might be free shipping.

https://www.aliexpress.com/item/10PCS-lot-V53 … 39-bc2adf8494de

Imperious wrote:

I recently bought a pack of 10 from Aliexpress, worked perfectly in my S3 Virge. 35ns chips.

Awesome! Thanks for that link 😀

If you have a card that can be sacrificed where the RAM is soldered on, warm the entire board with the gun, tilt it upside down and give it a sharp knock onto a flat surface. Pretty much all the surface mount parts will have molten solder and they'll fall off cleanly.

Slight derail of this thread.

I recently got this card that uses 40-pin ZIP memory for upgrade.

It works fine at baseline, and reports 1MB as it should.

But I wanted to upgrade it to 2MB. So I got 40-pin ZIP chips, 256kx16, so 0.5MB each, two would give 1MB.

They're 70ns chips, same as the original ones on the card. I plugged them in, made sure that the orientation is correct (pin1 to pin1). The card still reports 1MB at POST. I can't seem to find references to this type of memory chips in any other configuration than what I've got - so why didn't it work?

Maybe you have to set the three jumpers near the 26 pin header to a specific configuration? It's the only thing I can think of.

Provided that those ZIP chips are not dead, I see no reason why they wouldn't work... very strange indeed.

I have a dead PCI TGUI9440-3 you can have for a couple bucks + shipping. Just grab the ram off it? Idk offer is there

TheMobRules wrote:

Maybe you have to set the three jumpers near the 26 pin header to a specific configuration? It's the only thing I can think of.

Provided that those ZIP chips are not dead, I see no reason why they wouldn't work... very strange indeed.

I have no information about the jumpers. I tried 2 different sets of ZIP chips, same result. The tube they came in was original from the factory (got a tube with 10 chips) and it looks like it wasn't opened for years or decades - the rubber end caps of the tube had become fused with the tube walls, and I actually had to break one down with pliers to open the tube. I have no other means of testing the chips, but have no reason to believe they are dead.

It's almost certainly the jumpers, my VLB card (S3) has a jumper to set 1MB/2MB. I'm struggling to find documentation on what the jumpers on your card are for however

It's not worth adding extra ram to that card, it would be useless. It was a pretty good card back in its day, pretty fast for a 32 bit card and very compatible but like many other ancient cards had a crappy ramdac. With a crt you'll be limited to 640 and 16 bit or 1024 with 256 colors, anything above that will use 56 hz.
According to the manual for the tvga9440agi @vgamuseum 2 mb only would be useful for getting acceleration @800 and 16 bit but my card (tgui9440) would only do 56 hz at that resolution, you could get 1280 @60 hz as well at least in theory. And yes, there are jumpers to set the ram size.

TheAbandonwareGuy wrote:

I have a dead PCI TGUI9440-3 you can have for a couple bucks + shipping. Just grab the ram off it? Idk offer is there

Might take you up on that 😉

Pics?

Imperious wrote:

I recently bought a pack of 10 from Aliexpress, worked perfectly in my S3 Virge. 35ns chips.

$5.70 US plus any shipping or might be free shipping.

https://www.aliexpress.com/item/10PCS-lot-V53 … 39-bc2adf8494de

Has anyone tested whether faster add-on memory gives any type of measurable performance boost? Let's say my S3 has 45ns ram soldered on, and the sockets have 60ns chips in them. If I replaced those with 35ns chips from the link above, I wonder if there would be a measurable improvement. I don't imagine it would be significant, but even something in the 5-10% range of improvement might be worth the trouble.

Not gonna happen 😉

They put fast enough memory chips. Faster won't do anything unless you overclock the card itself.

clueless1 wrote:

Has anyone tested whether faster add-on memory gives any type of measurable performance boost?

Not unless there is a way to tell the card that you have faster memory. This generally requires you to patch and reflash the BIOS, but on old cards like this, I doubt that's even possible.
The card will just use the memory timings that it is programmed to use. It can't tell how fast or slow memory is (this is a feature that is only present on mainboard memory afaik, and was a relatively recent development, I believe in the DDR3 era, where they add a simple ID chip to each DIMM so the motherboard can query the modules for their timings, like SPD and XMP: http://www.intel.com/content/www/us/en/gaming … rofile-xmp.html).

There are some programs that you can tweak some hardware via DOS. I've used it to change timings on ET6000 and some Award motherboards, but I don't know if it supports Trident.

Read my example again. In my example, the soldered ram is 45ns while the socketed ram is 60ns. Wouldn't the speed of the soldered ram be what the VGA BIOS is expecting? Therefore, wouldn't replacing the 60ns socketed ram with 35ns allow the VBIOS to run all is ram at the rated 45ns? Or is it doing that anyway by overclocking the 60ns socketed ram?

Most probably factory installed overrated chips. If everything is working fine with 60ns then they are fast enough. Of course they could be overclocked, but you don't know at what speed they are actually running now.

clueless1 wrote:

Therefore, wouldn't replacing the 60ns socketed ram with 35ns allow the VBIOS to run all is ram at the rated 45ns?

It has no way to tell what speed the RAM is. So no.
As I said, if it even has configurable timings at all, they would be configured in the BIOS.