Out of the box the reference 5700 XT has high burst speeds but under load it overheats and therefore drops frames to the extent that it is more or less unusable for demanding games like PUBG. In order to settle the card it was under volted by 120 mV and the maximum clock was lowered to 1,980 MHz (the stock BIOS and 19.9.1 driver defaulted the card to 2,030 MHz). The fan curve also had to be turned up to around 50% power at 75 degrees (which is a lot noisier than stock). After making these changes the card delivered far more consistent performance albeit with a reduced top speed and unacceptable (hair dryer) levels of noise. There were also incompatibilities with GTAV: enabling reflection MSAA resulted in very poor, almost matt, reflection fidelity (the same bug appeared on several Navi and Vega cards). The reference 5700 XT is great for beating cherry picked benchmarks, but it is not so great for playing games. Thousands of people, duped by sponsored marketing material, purchased the reference card expecting flagship performance, instead they got a shopping trolley with a V6 engine. It appears that the same marketing tactics were employed for the reference Vega 56 and 64 series of graphics cards which we will purchase for our gaming lab and generate effective Fps gaming metrics as soon as possible (results here). AMD appear to have very short term marketing strategists at the helm, they seem more concerned with this years bonuses than the longevity of the brand. ^{[Jul '19 GPUPro]}

MORE DETAILS