GTX 1080
> GDDR5X: 484 GB/s
RX Vega 64
> HBM2: 484 GB/s
not a video card trolling thread. memory question: can somebody please explain the point of HBM? i thought it's advantage was higher bandwidth?
>>61667556
to look cool and edgy while underperforming
>>61667556
>gtx 1080
>484 GB/s memory
Are you kidding?
>>61667556
bro just wait
>>61667681
sorry, i should've said 1080 ti
both do use GDDR5X
https://en.wikipedia.org/wiki/GeForce_10_series
but that's besides the point. where is the bandwidth promised by hbm?
>>61667880
Not sure about bandwidth, but HBM doesn't take up as much room. It's tiny compared to GDDR
>>61667556
It's advantage is size.
>>61667880
Up rajas ass.
I'm saddened by the fact the 16gb version doesn't have anything extra either
>>61667556
HBM2 has absurdly lower latency that allows it to be used as an L4 cache
>>61667556
It's a marketing gimmick. We're not bandwidth starved.
>>61668791
link or gtfo
>>61667556
Because Vega is a pro/enterprise GPU first and a gayming GPU second. Plus AMD doesn't have the money and resources to do a separate consumer platform and Polaris doesn't really scale past the 580 tier GPUs so we get cut down Vega stuff with HBM.
>>61669881
implying that HBM is better for "enterprise" than GDDR5X
kek
>>61667556
the HBM memory controller is smaller, probably also consumes less power along with memory chips themselves
>>61670151
i have no doubt that there are advantages to HBM over DDR, otherwise the industry would not have made such a huge investment in it. but BANDWIDTH was supposed to be THE advantage, and it ain't there. so WTF happened to Vega's HBM b/w?
>>61667556
Ones in bits and ones in bytes you gigantic retard.
that retarded gpu runs so hot, they probably couldn't afford the watts to clock the hbm any higher
>>61670197
try again shit-for-brains ;)
>>61669853
hbcc
>>61667556
Give back the core memory Ubuntu
>>61667880
It's the reason Vega has less frame drops, we only saw amd's charts but it's not difficult to deduce this if you read about hbm memory.
>>61670195
>so WTF happened to Vega's HBM b/w?
2 stacks of 4gb
Instead if the furyX 4 stacks if 1gb
>>61670235
You are an absolute retard, g is full of fucking morons who cant read.
>>61667556
HBM as the advantage of:
>lower power
>smaller (see Fury nano and Vega nano)
>lower latency
>Higher bandwidth (theoretically)
My only guess to the low bandwidth is only using 2 stacks instead of four like on the Fury
>>61670137
Say this on my face, you fucking donkey!
manufacturing/supply problems with hbm. that's why its bandwidth is shit and vega was delayed from spring 2017 to mid-august 2017
get your shit together AMD!
>>61670415
>Supply
>Bandwidth
Pick one
>>61670137
Even Nvidia uses HBM.
>>61667556
>expecting a normal conversation about videocard technology without fanboy manchild screeching
>>61667556
to raise cost and power usage for no reason
>>61674431
>power usage because of hbm2
haha no, kill yourself