i was thinking about one of those datacenters and workstations GPUs , anyone tried this monsters before ? i think one of them should go over 100 MH at least
are you people stupid no the first card will hash slower than the R9 290x the second one is going to do 26MH/s and the last one can do up to like 58 MH/s
If you knew anything about GPUs you'd know that the professional cards have the exact same core as the consumer cards they just get higher bins, and ECC memory and sometimes a lot more memory the first card uses a Hawaii 28nm core (290, 290x) the second card uses dual Tahiti 28nm cores (7950, 7970, 280, 280x) the third card uses dual Fiji 28nm cores (Fury, Fury X, Nano)
they do not have any faster memory. More memory will not do crap for hashrate. You want tighter timings and higher clocks, which the professional cards DO NOT OFFER. And if they do, its definitely not worth the massive price gap. Forget about it.
@midoprince the R9 Fury series (Fury X, Fury, Nano) all have 4096 bit memory buses. It's comprised of 4 1024-bit 1GB HBM stacks. heres the problem though the Fury series does not hash at a million MH/s since its the first time HBM has ever been used, the memory controller on these cards is kinda garbage. Theoretically, with a Fury series card, you shouldn't even need to overclock HBM to gain performance (gaming load). However, overclocking the HBM on the Fury cards is like the best way to gain performance on these cards. Something's wrong here. (The entire memory subsystem is primitive and not fully developed yet) That's why I think Vega will be huge If its an RX 480 core with more Stream Processors and HMB2 with a new controller that thing could maybe do 60+ MH/s like everyone is saying I dont know this for sure but with Vega, AMD has had time to refine the power management and memory controller... should help Tighter timings is always better (well of course if the memory can handle it) Boost clocks, lower timings, lower voltages Thats how you get an efficient card The RX 480 makes me really sad its SOOOO heavily memory bottlenecked the core is capable of so much more but is bound in performance by the shitty memory setup The hashrate difference of a card that does 2250 memory and a card that does 2045 memory is like 2 MH/s!! and we usually underclock the 480 to 1175MHz core to save on power and voltage...
So, sorry to necro this post. I actually just got some of the S100000's in. Everyone's been scalping the crap out of the Vega's and 1080 series cards and reselling at double and triple the retail price. Anyway, I found a few of the S10000's dirt cheap. I'll start benchmarking later tonight, and I'll let you know how it goes. I'm really new to the boards, so I haven't looked through all the link posting rules just yet, but if it's allowed, I'll leave a link to the supplier if anyone's interested after I establish the benchmarks, get a feel for how close they get to max tdp, etc.
So how did the S10000's perform? I have an S9150 and Win-10 won't recognize it even with latest AMD Drivers specific to the GPU. Any thoughts on how to see the S9150 in Dev Manager?
Comments
While I do not have the expertise to say exactly why, I do not believe for a second these are worth the price.
we need someone already tested it
first one have memory bus 4096 bit while 256 bit on r 480 thats mean 4 times higher and ethereum is memory hungry
12kVbrmN6aB1o3yX6vVwxDq8h6hrnmaBYY
no
the first card will hash slower than the R9 290x
the second one is going to do 26MH/s
and the last one can do up to like 58 MH/s
If you knew anything about GPUs you'd know that the professional cards have the exact same core as the consumer cards
they just get higher bins, and ECC memory
and sometimes a lot more memory
the first card uses a Hawaii 28nm core (290, 290x)
the second card uses dual Tahiti 28nm cores (7950, 7970, 280, 280x)
the third card uses dual Fiji 28nm cores (Fury, Fury X, Nano)
they do not have any faster memory. More memory will not do crap for hashrate. You want tighter timings and higher clocks, which the professional cards DO NOT OFFER. And if they do, its definitely not worth the massive price gap. Forget about it.
but what you think about the memory bus difference from 256 bit to 4096 ?
can you explain the timing section if you have any more info please ?
the R9 Fury series (Fury X, Fury, Nano) all have 4096 bit memory buses. It's comprised of 4 1024-bit 1GB HBM stacks. heres the problem though
the Fury series does not hash at a million MH/s
since its the first time HBM has ever been used, the memory controller on these cards is kinda garbage. Theoretically, with a Fury series card, you shouldn't even need to overclock HBM to gain performance (gaming load). However, overclocking the HBM on the Fury cards is like the best way to gain performance on these cards. Something's wrong here. (The entire memory subsystem is primitive and not fully developed yet)
That's why I think Vega will be huge
If its an RX 480 core with more Stream Processors and HMB2 with a new controller
that thing could maybe do 60+ MH/s like everyone is saying
I dont know this for sure but with Vega, AMD has had time to refine the power management and memory controller...
should help
Tighter timings is always better (well of course if the memory can handle it)
Boost clocks, lower timings, lower voltages
Thats how you get an efficient card
The RX 480 makes me really sad
its SOOOO heavily memory bottlenecked
the core is capable of so much more but is bound in performance by the shitty memory setup
The hashrate difference of a card that does 2250 memory and a card that does 2045 memory is like 2 MH/s!!
and we usually underclock the 480 to 1175MHz core to save on power and voltage...