There are different flavors of DX4, at both Intel and AMD, so there is no obvious answer without knowing what exactly you have there.
Both had early models that didn't do WriteBack caching, these are noticeably slower than the later ones that do (on boards that support this feature, of course). The later ones are recognized by them also supporting the CPUID instruction and having some rudimentary power saving features ("SL enhancements"
.
Furthermore, AMD's DX4 initially had 8 KBytes of cache, while Intel's always had 16. Much later after both introduced WB caching, AMD brought us the superior DX5-133 with 16 KBytes of cache, and along with it a new DX4 with the same amount to finally match. AMD also had a DX4-120 for 40 MHz local bus boards.
AMD marked their "old" DX4 NV8T and the "new" ones SV8B and SV16B (for SL enhanced 8/16 KB cache writeBack capable). Intel's must be put into a system and inquired through software, since their manufacturing S-spec information for these parts seems long lost, or at least isn't available to the public.
Intel's DX4-100 had 2x and 3x multiplier support, for running it on 33 and 50 MHz boards. There also was a DX4-75 for 25 MHz boards, I'm not sure whether they ever actually sold that one.
AMD's NV8T and SV8B models also have 2x and 3x multipliers, while the late SV16B ones have 4x and 3x.
All DX4 and DX5 are 3.3 or 3.45V by the way, with the exception of the Intel DX4 overdrive, which is a 3.3V chip with an on-chip power converter to make it work in a 5V-only socket.
Finally, Intel and AMD 486-class processors had exactly the same instruction and data throughput in their processor core. Performance is exactly the same if you have models with the same feature set and cache size. Only Cyrix's 5x86 "M1SC" was noticeably faster per clock, already having the superscalar core that became most famous in the Pentium-class 6x86 and is still alive in the MII.
Regards, Peter
Facebook
Twitter