why not download the benchmark suite here: Phil's Ultimate VGA Benchmark Database Project and run it with both cards? Then you'll have an accurate answer for your system 😉

I would guess S3 is faster. However, you should benchmark to confirm. Use the link provided by keropi. We would appreciate if you added your data to our database after testing. 😀

Wrong forum. This should be in Marvin.

The Trio should be faster unless the Cirrus has 2MB of RAM, then it should be pretty even.

RAM does not matter, RAM is for framebuffer only.

RacoonRider wrote:

RAM does not matter, RAM is for framebuffer only.

That's not entirely correct...

Many graphics cards have sockets for RAM chips. If there are empty ones, some cards perform slower. I had this with an on-board ATI Mach 32 card. It has soldered memory and slots. I filled the slots and my benchmark scores improved 😀

RacoonRider wrote:

RAM does not matter, RAM is for framebuffer only.

normally that's true, but the Cirrus chipset only utilizes it's 64bit memory bus if it's equipped with 2MB of memory. With only 1MB, it's stuck at 32 bit.

The Trio has no such limitation.

Still it might be that the additional theoretical bandwidth is not utilized because the chip itself can not keep up. You can see this for example with a TSENG ET6000 where plugging additional 2.25 MB MDRAM just increases speed marginally. It is always better to run some benchmark than to trust some theoretical performance gains.

Still it might be that the additional theoretical bandwidth is not utilized because the chip itself can not keep up. You can see this for example with a TSENG ET6000 where plugging additional 2.25 MB MDRAM just increases speed marginally.

Technically you shouldn't be seeing any gain by putting the extra memory on an ET6000, because the ET6000 actually only has a 32-bit memory bus. (Compensated by the fact that MDRAM runs at about twice the clock speed of normal DRAM/VRAM of the time.) With 2 or 2.25MB, it's already using the full memory bus width... adding two more chips won't change that.

Just reread some info and you are right, it just applies if you go from 1 chip to 2 chips that the bus width increases from 16 to 32 bit.

Old Thrashbarg wrote:

Still it might be that the additional theoretical bandwidth is not utilized because the chip itself can not keep up. You can see this for example with a TSENG ET6000 where plugging additional 2.25 MB MDRAM just increases speed marginally.

Technically you shouldn't be seeing any gain by putting the extra memory on an ET6000, because the ET6000 actually only has a 32-bit memory bus. (Compensated by the fact that MDRAM runs at about twice the clock speed of normal DRAM/VRAM of the time.) With 2 or 2.25MB, it's already using the full memory bus width... adding two more chips won't change that.

But even if you have maxed out the speed on the bus, wouldn't the extra memory still be useful for increasing the maximum resolution and color depth? I know with my two Mach64 ISA cards, the one with 4mb is capable of displaying more pixels and more colors than the 2mb card.

Well yeah, you're absolutely right on that point. I didn't intend to imply the extra memory wouldn't be useful, I was simply pointing out that it doesn't affect the speed on that particular card.

I've tried a Speedstar Pro VLB (CL5426 IIRC) once that had palette corruption and so I fell back to Trio64v+ PCI. Not sure if 5436 exhibits similar issues.

and that was 1997.