Googling around it looks like this problem was associated with the Adlib Gold and a few clone cards? There's a thread about it that suggests it's not so much CPU speed that's the problem as bus timing issues. (IE, there's a person who claimed they got it to work fine on a 440BX motherboard, which of course would be a Pentium II or better CPU.)
Some 486 motherboards had BIOS-adjustable ISA speed dividers. (IE, 486 CPUs came in speeds anywhere from the rare 16mhz 486SX all the way up to the 50mhz non-doubled 486DX.) The "standard" for ISA is usually specified as 8.33mhz, so a 33mhz bus 486 typically has a 1/4th divider. It's probably something you'd be more likely to find on a generic clone than your gateway, but it's *possible* you might have the option of selecting a 1/5th or 1/6th divider for ~6mhz or ~5mhz respectively. Of course this will slow *everything* down, including video and hard disk performance, but it in theory might let the card work if it doesn't like the default setting.
(On the flip side, on my 486 that had this option I overclocked the bus to 11mhz using the 1/3rd divider. It's a great way to make an IDE hard disk with a cache run faster, or at least look a lot faster on dumb benchmarks.)