Originally Posted by steadly2004
My VEGA was holding 1800 or so, didn't write down or screen shot.... While my son was playing prey on my VEGA via steam link. I had to adjust wattman when he was done to go custom (instead of balanced) to go back to mining when not in use. Maybe I was seeing something. But I'll definitely have to re-test in the future. I'm pretty sure is was 1802 or something like that on the core. Perhaps it was a fluke, or perhaps I am tripping. I don't know.
What was doing the reporting? Be sure to update whatever it is to the latest and or greatest version. There were some issues with false reports and "amazing" clocks which are simply not true. Now to be fair, and to clarify, that a Vega hit some high frequency... once... and didn't hose is noteworthy, but doesn't really mean all that much.
By maximal clock, I mean the "average" sustained clock rate over a 10 minute time span averaged in a repeating bench, including maxima and minima of the frequency curve performance. HWinfo64 is very handy for this. This number is going to vary from application to application, but ultimately, the goal for me is to realize the best possible clock for the most performance while minimizing the driver / application failure. Now, I am fairly tolerant of a dodgy setup if it gets me at least 2 fps at my native res in my minimums. (For freesync purposes @ 3440 @ 100 hz, so my minimum acceptable low is 48, ideally 55+).
The reason I call BS is just from my own experience and hours of tweakin' n' benching on the sig rig with a Vega 64 Ref on the LC Bios, EKWB; 600 mm of rad, 3/4 gal fluid, 18w pump; in a positive pressure room with a dedicated LG inverter and 2 air circulation tower fans.Warning: Spoiler! (Click to show)
1737 held for 22 minsWarning: Spoiler! (Click to show)
1796ish application crash
1710-1720s has been (in my experience, your mileage may vary) a physically realizable frequency that is perfectly "game-able". Not sure I would trust it mining or doing significant work for $$$... drop another 10mhz for reliability. To go further is to get into reg. tweaking the core voltage... which looks dodgy.
Originally Posted by Flaxen Hegemony
Your explanation of the clocks and benchmarks makes sense to me, which is why the statement struck me as odd. If the air cooled and water cooled were equivalent at chip level except for a cooler, then AMD was literally just selling an AIO for $500. There has to be a performance difference. I mean, even that may not be worth $500, but still.
Well even at that, there are not all that many AIOs which can be set @ stock settings and still hold up with a +50 power... most that I have seen... crash. Vega is only going to clock up if there is suitable power and volts in reserve. The stock AIO setting... never going to see that P7 except once in a blue moon... its near pointless. P6 can help out there, once you find your Vegas P7 limit, or... it finds your patience threshold with all the rebooting.
Frame rates are all!