Originally Posted by ZealotKi11er
Look at how much faster VII is vs Vega 64. Not much. 1TB/s does not help it in games. We will have to wait and see but I still think AMD needs to beat 2080 Ti.
I know with my V64 at 1050mhz HBM2 (537.6GB/s bandwidth), any gains past that point were minimal at 3200x1800 or higher and far less than noticeable lower than that. Mind you my gpu sits at 1630mhz so it's a bit slower than a VII clock speed wise but I've got the full fat 64CU too which from research was never really a benefit compared to the V56 as scores/fps are near identical clock for clock.
Any idea what kind of bandwidth speeds GDDR6 would typically produce assuming a 256bit bus and average clocks (I know, this is not exactly an easy request)?
I did see quite nice gains, atleast in BFV going from 945mhz mem to 1050 so i'd imagine faster clocks, the full 64CU's but being limited to less bandwidth would really kill some performance, at least at higher res. Also, despite the move to 7nm it's not like VII is a low power gpu. GDDR6 is only going to increase this so I'd imagine Navi would have memory clocks on the reserved side.
Forgot the 2080 used GDDR6 w/ a 256bit bus. So assuming the same 14ghz we are talking 448gb/s when a stock clocked V64 is at 483.8gb/s. Considering the gains I saw going up 53.8gb/s I can't imagine Navi being much faster than a V64 with less bandwidth despite a fairly large jump in clock speed.
VII is faster than a half decent V64 but in all reality, it's certainly not earth shattering. At stock it has double the bandwidth (still blows my mind to see over 1,000gb/s on a GPU) and a boost clock barely over 100mhz faster than a stock liquid cooled V64 (which face it, those speeds are typically easy enough to achieve on any V64).
Navi is going to have to be priced really well to sell or it's going to have to pull a high clock trick out of it's hat.