>double the die size>double the power consumption>only 50% faster raster at BEST
>>107883389>170 rops RTX 5090 vs 112 rops RTX 5080>152% faster at 4K per TPU sharts>170/112 = 1.518wow... if only there was an explanation
>>1078834051.518 is 51%thank you for confirming my thesis
>>107883452yeah. it turns out they have a SM tensor heavy configuration for GB202 and a ROP heavy configuration for GB203 almost like one of them is designed for gaming and the other is ai/cuda slop master
>>107883467Why do they need ROPs if raster is the past and RT is the future?
>>107883389Cope you poor piece of shit. I bought a 5090 right before prices started going crazy. Now I’m set for at least half a decade.
>>107883492>poor piece of shit.Linus torvalds is infinitely more rich than you and doesn't buy bloated hardware that will need to be upgraded in 4 years time anyways
>>107883492well you'll have to ask Nvidiabut it's a contributing factor why the massive gb202 isn't that useful for gamers
>>107883389The 90 series is meant to be performance over efficiency what do you expect?