I ran some initial tests with an IBM 5x86c-100HF at 150 MHz on the Biostar MB-8433UUD motherboard while using a few different peltier pads. I started with conservative DRAM and SRAM timings. The 51 W peltier at 12 V was insufficient. I upped it to 75 W at 12 V and was able to complete DOS Quake with the Cyrix register features disabled. Unfortunately, when I enabled all the features, Quake crashed half-way through the timedemo. Thus, it would seem that one or more of the Cyrix features isn't stable at 150 MHz. It would take quite a bit of time to narrow down which feature(s), but I decided that I did not want to risk frying the CPU and have ceased testing.
Vcc to the CPU was 3.75 V. I tried it up to 3.82 V, but there is still some Cyrix feature which doesn't cooperate with that operating frequency. For the record, using conservative DRAM/SRAM settings with with all of the Cyrix features disabled, the Quake score was 16.8 fps at 150 MHz. PCI was run at 2/3*50=33MHz.
By contrast, I was able to use the 51 W peltier to run an AMD DX2-66-WB at 180 MHz with just 3.45 V to the CPU. DOS Quake scores of the Am5x86-180 and the Cyrix 5x86-133/2x are the same at 19.2 fps w/256K.
EDIT: There seems to be some variation in the power at 12V for these peltiers. Some quote the max wattage of heat dissipated, while others are quoting the electrical current needed to run the peltier at max current. For clarity, I used a TEC1-12708 for the IBM 5x86c and a TEC1-12704 for the AMD at 180 MHz.
Plan your life wisely, you'll be dead before you know it.