I guess probably as your new update requires using sieve primorial 14 while (at least what I found) 270 and 280X works better on sieve primorial 13.

Can you show miner output on 270 and 280X from versions 8.2 and 9 ?

GPU 0 : 270, GPU 1-2 : 7950, GPU 3-5 : 280X

v8.2 , sieve primorial 13

[GPU 0] T=79C A=99% E=0 primes=0.105580 fermat=366328/sec cpd=5.45/day

[GPU 1] T=69C A=99% E=0 primes=0.105392 fermat=553540/sec cpd=8.09/day

[GPU 2] T=72C A=99% E=0 primes=0.105485 fermat=553552/sec cpd=8.16/day

[GPU 3] T=78C A=99% E=0 primes=0.105531 fermat=665207/sec cpd=9.85/day

[GPU 4] T=82C A=99% E=0 primes=0.105622 fermat=626369/sec cpd=9.35/day

[GPU 5] T=86C A=99% E=0 primes=0.105346 fermat=653325/sec cpd=9.50/day

GPU 0: core=1000MHz mem=1500MHz powertune=20 fanspeed=29

GPU 1: core=1025MHz mem=1500MHz powertune=20 fanspeed=58

GPU 2: core=1025MHz mem=1500MHz powertune=20 fanspeed=58

GPU 3: core=1120MHz mem=1750MHz powertune=20 fanspeed=78

GPU 4: core=1050MHz mem=1600MHz powertune=20 fanspeed=84

GPU 5: core=1100MHz mem=1700MHz powertune=20 fanspeed=91

v9, sieve primorial 14

[GPU 0] T=77C A=99% E=0 primes=0.103699 fermat=402666/sec cpd=5.00/day

[GPU 1] T=69C A=99% E=0 primes=0.104086 fermat=620377/sec cpd=8.00/day

[GPU 2] T=71C A=99% E=0 primes=0.104176 fermat=622481/sec cpd=8.10/day

[GPU 3] T=77C A=98% E=0 primes=0.104209 fermat=744317/sec cpd=9.71/day

[GPU 4] T=83C A=98% E=0 primes=0.104315 fermat=700036/sec cpd=9.23/day

[GPU 5] T=86C A=98% E=0 primes=0.104253 fermat=732571/sec cpd=9.60/day

I have another test rig having 2x 270 which is also showing like 5% slow down.

Basically 270 is affected most. For 7950 and 280x the difference are minimal but v9 is definitely a tiny bit slower (I don't know why the last card looks faster when I took sample this round but a test result is a test result as is.

)