Hot on the heels of repairing the Cirrus Logic VLB card, I acquired a dead Trident TGUI9440 based VLB graphics card to see if I could repair it.
This is the card I received. It is a 1MB card which is expandable to 2MB, missing the bracket and has some physical damage. I was pretty sure I knew what was wrong before I purchased it. So, I went ahead and bought it. You can see some jumper pins are damaged. That isn’t the main problem, but something I will repair.
Now, for the problem. Experts in VLB cards are probably screaming the answer right now. It comes down to the jumper pins on the far left of the board.
The cool thing about cards of this era, and even many cards today, is that they are all based on a reference design. This means that the jumper numbers are usually the same from board to board, even if the physical layout is a little different. On Trident TGUI94xx VLB cards (and some others), this is the layout:
Those jumper pins are J1. There is nothing on them. So, the card doesn’t know what timing to use. Therefore, it is unlikely to boot. I immediately popped a jumper across pins 2-3.
Now to test the card. I popped it only my test VLB motherboard and connected it to my VGA monitor, and…
Wow! OK, this is going to be easier than I thought then.
As with the two Cirrus Logic cards that take DIP RAM, this requires 8x256KB 4bit DRAM chips to upgrade. I had some spare, so I whacked them in. I also removed the bent jumper pins and added some new ones.
Excellent! A full 2MB of RAM detected.
The card came without a bracket, and it is a good idea to have one to make the card nice and secure. To resolve this, I took this design and modified it slightly to make it stronger. This is what the end result looked like before printing.
When mounted to the card, this is what it looks like.
Let’s try side-by-side with my Cirrus Logic GD-5429 based VLB card. The tests are on my AMD 5×86 133MHz based PC. The first image is 3DBench 1.0c (Faster PCs), the second is Doom max. details, both from Phil’s Computer Lab benchmark pack. This is the results of the Cirrus Logic:
As for the Trident in this blog post (Update 2025-02-13: this is an unfair test, see next section below):
Both results are showing that the Cirrus Logic GD-5429 is about 5% faster than the Trident TGUI9440. I’m not sure which I’ll keep in my 486 PC at this stage, but I’m happy that this turned out to be a quick and easy repair.
I was doing some further testing early this morning and realised this was an unfair test. VLB cards have two speed options “≤33MHz” and “>33MHz”. The general idea is if your card is unstable you should set the jumper to “>33MHz” to add additional wait-states to slow the card down.
Now, the Cirrus Logic had this set to “≤33MHz”, whereas the Trident had it set to “>33MHz” (jumper JX7 in the table near the beginning of this post). When I set the Trident to the same as the Cirrus Logic, we get:
So, in reality, the Trident is about 5% faster than Cirrus Logic.
I’ve decided to keep the Trident in my machine for a couple of reasons. The speed difference is negligible, but the Trident has a cleaner DAC, which makes the image quality a lot better on my LCD screen. In addition, it has a faster pixel clock, which means it can output 1024×768 at a refresh rate that my monitor understands. But I would say that both are excellent DOS gaming cards.
Trident often gets a bad reputation for being a “budget” choice. Sure, they had some dud cards / chips. But the 94xx range and its siblings were budget because they did things such as removing the need for the expensive 74F-type interface logic.
It seems that every time I document how to do Xilinx JTAG programming from a…
Fellow vintage computer collector, Gammitin, challenged me to fix four VLB graphics cards, as well…
You may remember my recent 486 PC build, I enjoyed doing it, but I had…
One of the projects I wanted to do with my new 486 PC was to…
As many of you are probably aware, I am one of the developers of wolfSSL.…