So now that the dust has settled, it doesn't seem to beat the 2070s on average but gets consistently close enough vs a 2070s stock to justify it's price tag. However, considering the crappy drivers from AMD, coupled with the fact that 2070s does seem to have about 200mhz headroom OC putting it closer to an 2080, the extra 100$ isn't completely unjustified either. The 5700xt no longer seems to be the cemented choice in this price segment.
Many lines i dissagree with in this post.
1."2070s stock to justify it's price tag" Niether 5700/5700xt/2060/2060 super/2070/2070super/2080/2080 super/2080ti have the right msrp for the cards they replace. So 5700xt should be in 300-350 usd range with 2070 super. Only buyer is loser in this gen.
2. "considering the crappy drivers from AMD" To be honest i have had more problems with Nvidia drivers to date(both maxwell and pascal cards). This is an old statement mostly found in wccftech troll department.
3."200mhz headroom OC putting it closer to an 2080" This is true for both 2070 super and 5700xt when oced the proper way.
4. " choice in this price segment" 2060 super and 5700xt is in the same price segment.
Right now at current drivers RDNA and Turing have about same ipc when compairing cores and mhz or flops. 2560 (5700xt) and 2560 (2070 super) perform about the same at same clocks(some games favor nvidia some amd). Think even RDNA has some 1% lead in ipc avg over some games tested in one review site. 5700xt and super has about same max oc on air(2050-2100) and (2150-2200) watercooled depending on silecon lottery. Big problem is that many reviewers get strange results many all over the place. Many cases likly thermal trottling or max 50 % powerlimet set with 1.2v witch result in bad avg clocks for the radeon cards not to mention an great power draw. Only reason to buy 2070 super is RTX or if you play Nvdia sponsor games to justify 100 dollar more. The stock cooler on 5700xt is fine if you undervolt when pushing for higher core clocks.
Some testing posted earlier in this thread. Some Firestrike extreme gpu score testing 5700xt.
voltage scaling stock cooler ( set at 100% at :
core/mem volt watts avg real clock avg Powertarget junktion temp(highest) Fsx Gfx score
1860/910 mhz 913mv 110 w 1810mhz 0% 66 degrees 12110
1920/910 mhz 940mv 125w 1870mhz 0% 66 degrees 12520
1960/910 mhz 970mv 135w 1910mhz 0% 66 degrees 12756
2000/910 mhz 1000mv 145w 1945mhz 0% 72 degrees 12932
2070/910 mhz 1050mv 170w 2015mhz 0% 76 degrees 13220
2110/910 mhz 1080mv 180w 2045mhz 0% 81 degrees 13410
2150/910 mhz 1100mv 205w 2100mhz 0% 87 degrees Fail. card fail in scond test when junktion hit 87 degress.
Watercooling. Soft power table mod. water 24 degrees. 31 degrees idle gpu.
1860/910 mhz 908mv 105w 1810mhz 0% 52 degrees 12122 (210-215w of powerdraw at wall :O)
2151/910 mhz 1091mv 190w 2100mhz 0% 62 degrees 13700
2200/910 mhz 1141mv 210w 2150mhz 0% 65 degrees 13920
2250/910 mhz 1220mv 240w 2200mhz 10% 71 degrees 14050
2270/910 mhz 1250mv 270w 2220mhz 20% 78 degrees Fail. card fail in test 2 when junction hit 78 degrees.
Would love if some 2070 super owner did some scaling charts for voltage / clocks/ power draw temps cooling etc.