That would be the largest variance (by a mile) on a batch of chips I've ever seen. Are you seriously suggesting the B grade chips can do less than 1/3rd of the clock speed of the A chips? Pull the other one.

If the chips are functional (as in, not defective) the error rate will directly linked to the clock speed they operate at. In this case, clocking them to produce 350 GH is clearly too much already as a near-10% error rate is ridiculous. Maybe they can do 250 GH reliably. So the A chips now need 4 times higher clockspeed. Hilarious.