I think you got the order right already...

to paraphrase the old ad - "who wants Trident?" (crickets)

True, Trident was probably the slowest graphics card that you could put into a PC.
However, I believe that the TVGA8900D was a later model that was slightly faster.

I just performed the same benchmark tests again on a 386SX-25 and, although understandably slower, all 3 cards have exactly the same benchmark results.
The same goes for the performance of the 3 cards on the Cyrix 486DLC-40.

But, I think the order I have it in is probably correct.

I kind of like the signal quality of trident 8900's and 9000's. And they are cheaper than CL or ET cards.
So... When the speed are aprox the same on DLC's and 386sx25.
Then that must mean something like Trident are the way to go on this era of hardware.

Personally, I can only recommend Trident's on a 286-8/10mhz as I have on in my machine.
Stuff feels like it is running slightly faster on a Trident than on a Ega card.
And the Tridents can do all the Ega titles that I need to run from that era. Win-Win situation.

Doom realticks are pretty granular. If you want to break the tie (and have the patience), run the Doom benchmark in Phil's VGA package.

Decided to try out clueless1's suggestion (using the Doom benchmark):

• ET4000AX: 7097 realticks which equates to 10.52 fps
• Cirrus Logic CL-GD5422: 7048 realticks which equates to 10.60 fps
• Trident TVGA8900D: 7099 realticks which equates to 10.52 fps

Basically, all three cards has more or less the same performance, but I was a bit surprised that the Cirrus Logic CL-GD5422 was able to pull slightly ahead of the Tseng Labs ET4000AX.

Test system was a Cyrix 486DLC-40 with 16 MB of RAM on an ECS FA 386 motherboard with 128KB of external cache (UMC chipset).
However, I previously ran the benchmark with the ET4000AX and with only 5 MB of onboard RAM and got the same results.

Yeah, that's surprising! I fully expected the ET4000AX to be the winner. Cirrus Logic is actually one of the better DOS accelerators, at least with external graphics. I have an onboard CL on my Pentium and it lags pretty bad compared to Voodoo3 and TNT2 M64 PCI.

FYI, I just came across a Wolf3d timedemo that you might want to run too. It's in this RAR archive:
Performance comparison of CPU: 286-25 vs 386DX-25 vs 386SX-25

jesolo wrote:

Decided to try out clueless1's suggestion (using the Doom benchmark): […]
Show full quote

Decided to try out clueless1's suggestion (using the Doom benchmark):

• ET4000AX: 7097 realticks which equates to 10.52 fps
• Cirrus Logic CL-GD5422: 7048 realticks which equates to 10.60 fps
• Trident TVGA8900D: 7099 realticks which equates to 10.52 fps

Basically, all three cards has more or less the same performance, but I was a bit surprised that the Cirrus Logic CL-GD5422 was able to pull slightly ahead of the Tseng Labs ET4000AX.

Test system was a Cyrix 486DLC-40 with 16 MB of RAM on an ECS FA 386 motherboard with 128KB of external cache (UMC chipset).
However, I previously ran the benchmark with the ET4000AX and with only 5 MB of onboard RAM and got the same results.

Looking at prices makes an ET-4000 less sought after.
Or, in other words, is an ET4000 worth the price?

These are some comparative results I recorded when I benched my ISA graphics cards on a high-end VLSI-based 386 motherboard w/DOOM.

feipoa wrote:

These are some comparative results I recorded when I benched my ISA graphics cards on a high-end VLSI-based 386 motherboard w/DOOM.

ISA_Bench_VLSI.png

That confuses me a bit. Is the reason for better result's, the motherboard, cpu or both?

I think it's more to do with the motherboard, but the CPU could also play a role, since the Ti486SXL-40 has 8 KB of L1 cache, as opposed to 1 KB of L1 cache on the Cyrix 486DLC-40 (would be interesting to see how these two CPU's stack up against each other in benchmark tests like NSSI 0.60, Norton Sysinfo 8.0, Landmark 2.0, etc.)

I recently benchmarked my Cyrix 486DLC-40 on another hybrid 3/486 motherboard that has an OPTI 495XLC chipset (using similar BIOS settings).
On that motherboard, I only managed to get 7924 realticks (equates to 9.43 fps) with my ET4000AX, which is quite a significant diferrence and proves that, in this case, the motherboard/chipset can play a large role.

jesolo wrote:

I think it's more to do with the motherboard, but the CPU could also play a role, since the Ti486SXL-40 has 8 KB of L1 cache, as opposed to 1 KB of L1 cache on the Cyrix 486DLC-40 (would be interesting to see how these two CPU's stack up against each other in benchmark tests like NSSI 0.60, Norton Sysinfo 8.0, Landmark 2.0, etc.)

I recently benchmarked my Cyrix 486DLC-40 on another hybrid 3/486 motherboard that has an OPTI 495XLC chipset (using similar BIOS settings).
On that motherboard, I only managed to get 7924 realticks (equates to 9.43 fps) with my ET4000AX, which is quite a significant diferrence and proves that, in this case, the motherboard/chipset can play a large role.

From what I can find here on this forum that SXL is a good deal faster than the DLC by far.

Hmmm. So the extra cache is what makes the ET run faster? No cache = no need for anything else than trident's then.

Only the 8900D is a fast Trident, but the output quality at certain resolutions are not nearly as clean as with the higher-end cards.

There were many different brands of Tseng ET4000. Some might use slower memory timings. Some have jumper for 0-Wait states operation. But overall the ET4000 is around the top. Fastest is CL 5434. ATi Mach64 might be close if configured with 0-WS. That's for DOS usage. In Windows Mach64 is on the top by large margin.

I re-tested all of the above cards on a VLB based 486 motherboard using an Intel 486DX-33.

I'll post my results a bit later on, but the Cirrus Logic CL-GD5422 still came out on top (albeit by a slight margin).

I've also noticed that my Cirrus Logic has much less "interference" or "noise" on my widescreen LCD monitor when I just have a black screen (DOS prompt) in front of me (with the other two cards, I notice a slight distortion on a black background). It's almost like the signal is "stronger" coming from graphics card, giving a clear (no "interference/noisy" background.