Topic: Fury X mining performance - page 2. (Read 22371 times)
I keep seeing people talk about "non-opens source" mining software. Where can we purchase it? I have tried to find it, but all I have found is open source software.
It is opensource now, you don't need to use the binary...
Oh lol, didn't know that
will do some more testing later then 
It is opensource now, you don't need to use the binary...
>As i said before, with those low temps, i know this also could go a lot faster. Energy usage confirmed that.
The bottleneck of CryptoNight PoW algo is that it requires loading small data blocks from random memory addresses, a lot of times. Because addresses are random, memory cache is useless. You see low temps because GPU core does not work all the time, most time it waits data from GPU memory due its latency. And it cannot be optimized because the algo was designed to use memory latency to be asic-resistant. Fury X has 4096 bit memory bandwidth, but it is useless for this algo because big bandwidth is not related to memory latency.
The bottleneck of CryptoNight PoW algo is that it requires loading small data blocks from random memory addresses, a lot of times. Because addresses are random, memory cache is useless. You see low temps because GPU core does not work all the time, most time it waits data from GPU memory due its latency. And it cannot be optimized because the algo was designed to use memory latency to be asic-resistant. Fury X has 4096 bit memory bandwidth, but it is useless for this algo because big bandwidth is not related to memory latency.
same for lyra2re (it looks like fury x performance is worse than 280x), but they are changing the algo (at least for vertcoin) in order to be more gpu-friendly, so that might change.
Did a quick Lyra2re test with stock settings and open source kernel.
Dunno where that cryptoblog got their results.. but they're way off..
here are my results.
Almost 1mhz. That's a serious difference than the 430Kh Cryptoblog wrote...
Energy usage
@Pallas, tried to run your binary but that froze up my rig each time.
it doesn't freeze up when doing that for Quark or any other algo.
Greetings
Thanks for posting! This definitely helps inform my next purchase. I've been debating going the AMD route for a new monero rig. So at current release, it seems that for whatever reason the fury performs as well as the 290x.
You're very welcome!>As i said before, with those low temps, i know this also could go a lot faster. Energy usage confirmed that.
The bottleneck of CryptoNight PoW algo is that it requires loading small data blocks from random memory addresses, a lot of times. Because addresses are random, memory cache is useless. You see low temps because GPU core does not work all the time, most time it waits data from GPU memory due its latency. And it cannot be optimized because the algo was designed to use memory latency to be asic-resistant. Fury X has 4096 bit memory bandwidth, but it is useless for this algo because big bandwidth is not related to memory latency.
It's useless for most algo's.. The bottleneck of CryptoNight PoW algo is that it requires loading small data blocks from random memory addresses, a lot of times. Because addresses are random, memory cache is useless. You see low temps because GPU core does not work all the time, most time it waits data from GPU memory due its latency. And it cannot be optimized because the algo was designed to use memory latency to be asic-resistant. Fury X has 4096 bit memory bandwidth, but it is useless for this algo because big bandwidth is not related to memory latency.
But again, everything was at stock settings, i have been playing around with the core engine and was able to push it to 1190 without a problem. After that it started to fail, temperatures were still very low though, if i was on windows i could play with voltage as it's possible with Afterburner.
I'm not experienced enough to optimize any algo to be Fury X friendly
But i know someone who is and will be playing with it in the next weeks.I'm curious about the results!
>As i said before, with those low temps, i know this also could go a lot faster. Energy usage confirmed that.
The bottleneck of CryptoNight PoW algo is that it requires loading small data blocks from random memory addresses, a lot of times. Because addresses are random, memory cache is useless. You see low temps because GPU core does not work all the time, most time it waits data from GPU memory due its latency. And it cannot be optimized because the algo was designed to use memory latency to be asic-resistant. Fury X has 4096 bit memory bandwidth, but it is useless for this algo because big bandwidth is not related to memory latency.
The bottleneck of CryptoNight PoW algo is that it requires loading small data blocks from random memory addresses, a lot of times. Because addresses are random, memory cache is useless. You see low temps because GPU core does not work all the time, most time it waits data from GPU memory due its latency. And it cannot be optimized because the algo was designed to use memory latency to be asic-resistant. Fury X has 4096 bit memory bandwidth, but it is useless for this algo because big bandwidth is not related to memory latency.
same for lyra2re (it looks like fury x performance is worse than 280x), but they are changing the algo (at least for vertcoin) in order to be more gpu-friendly, so that might change.
Lyra2re performance is worse than a 280x you mean? i haven't tested that algo "yet".. so i don't know. I ignored Cryptomining blog results because most of them aren't correct compared to my results.
Regarding Cryptonote, seeing the results with other cards (stock settings) the fury beats them all, not by much, but it's still the fastest it seems
Greetings
>As i said before, with those low temps, i know this also could go a lot faster. Energy usage confirmed that.
The bottleneck of CryptoNight PoW algo is that it requires loading small data blocks from random memory addresses, a lot of times. Because addresses are random, memory cache is useless. You see low temps because GPU core does not work all the time, most time it waits data from GPU memory due its latency. And it cannot be optimized because the algo was designed to use memory latency to be asic-resistant. Fury X has 4096 bit memory bandwidth, but it is useless for this algo because big bandwidth is not related to memory latency.
The bottleneck of CryptoNight PoW algo is that it requires loading small data blocks from random memory addresses, a lot of times. Because addresses are random, memory cache is useless. You see low temps because GPU core does not work all the time, most time it waits data from GPU memory due its latency. And it cannot be optimized because the algo was designed to use memory latency to be asic-resistant. Fury X has 4096 bit memory bandwidth, but it is useless for this algo because big bandwidth is not related to memory latency.
same for lyra2re (it looks like fury x performance is worse than 280x), but they are changing the algo (at least for vertcoin) in order to be more gpu-friendly, so that might change.
>As i said before, with those low temps, i know this also could go a lot faster. Energy usage confirmed that.
The bottleneck of CryptoNight PoW algo is that it requires loading small data blocks from random memory addresses, a lot of times. Because addresses are random, memory cache is useless. You see low temps because GPU core does not work all the time, most time it waits data from GPU memory due its latency. And it cannot be optimized because the algo was designed to use memory latency to be asic-resistant. Fury X has 4096 bit memory bandwidth, but it is useless for this algo because big bandwidth is not related to memory latency.
The bottleneck of CryptoNight PoW algo is that it requires loading small data blocks from random memory addresses, a lot of times. Because addresses are random, memory cache is useless. You see low temps because GPU core does not work all the time, most time it waits data from GPU memory due its latency. And it cannot be optimized because the algo was designed to use memory latency to be asic-resistant. Fury X has 4096 bit memory bandwidth, but it is useless for this algo because big bandwidth is not related to memory latency.
Thanks for posting! This definitely helps inform my next purchase. I've been debating going the AMD route for a new monero rig. So at current release, it seems that for whatever reason the fury performs as well as the 290x.
Ok i have some new tests regarding Cryptonote.
As seen in the screenshot. 1 Fury at stock settings runs @ around 850+h/s
Cards were almost at idle temperatures. So in other words, they were hardly doing their best
Screenshot of hash rate
As i said before, with those low temps, i know this also could go a lot faster. Energy usage confirmed that.
Not even 500 Watt with 3 Cards.
Screenshot
I have only one algo that drives my fury's over 40 degrees. And that is Siacoin. This is because i'm pretty sure that they're almost working at full power on that algo. Also the power usage when running sia is more then double than what it was pulling on Cryptonote. Not only Cryptonote but also X11, Quark etc..
Again, if someone wants to test something, Just ask
Greetings!
As seen in the screenshot. 1 Fury at stock settings runs @ around 850+h/s
Cards were almost at idle temperatures. So in other words, they were hardly doing their best
Screenshot of hash rate
As i said before, with those low temps, i know this also could go a lot faster. Energy usage confirmed that.
Not even 500 Watt with 3 Cards.
Screenshot
I have only one algo that drives my fury's over 40 degrees. And that is Siacoin. This is because i'm pretty sure that they're almost working at full power on that algo. Also the power usage when running sia is more then double than what it was pulling on Cryptonote. Not only Cryptonote but also X11, Quark etc..
Again, if someone wants to test something, Just ask
Greetings!
anyone tried to mine XMR with the Fury X?
Ok here are some details.
Around 1300 h/s per card
Greetings
Thank you for testing.
Is the 490w for all 3 cards combined?
would be nice if the images was a tad smaller
Would be even nicer if your resolution was a tad bigger :-)
I'll remove the image tags, specially for you :-)
would be nice if the images was a tad smaller
anyone tried to mine XMR with the Fury X?
Ok here are some details.
Around 1300 h/s per card
screenshot
http://s24.postimg.org/fve5l9zhx/Schermafbeelding_2015_07_24_om_21_34_24.png
Power usage at that time
Greetings
@Eliovp Thanks for your detailed power consumption and hashrate charts.Fury X really interesisting card, pls keep updating informations,Still can't deceide to buy 970 or fury X ))
))
anyone tried to mine XMR with the Fury X?
Ok, got some new X13 results!
Also own Mod
9.7Mhz per card @ 1050 Engine clock pulling 756Watt with 3 cards.
Watt
http://s29.postimg.org/ksobyeh93/IMG_7069.png
Greetings!
Also own Mod
9.7Mhz per card @ 1050 Engine clock pulling 756Watt with 3 cards.
Watt
http://s29.postimg.org/ksobyeh93/IMG_7069.png
Greetings!
thanks for the tests.
in the end the performance is 0-20% higher than a 290x depending on the algo:
sia - about the same
groestl - about 10% more
x11 and quark - about 20% more
let's hope they make a better opencl compiler to unleash its full power!
finally a curiosity: did you get valid blocks with that ultra fast sia miner?
No problem in the end the performance is 0-20% higher than a 290x depending on the algo:
sia - about the same
groestl - about 10% more
x11 and quark - about 20% more
let's hope they make a better opencl compiler to unleash its full power!
finally a curiosity: did you get valid blocks with that ultra fast sia miner?
Mm, Sia about the same? on my 290x i hit around 1.25 Ghz (oc'ed), on my Fury's i hit 1.9 Ghz (stock clocks)
That's a difference of 40%, non open source though..
As you said, a better opencl compiler and optimized kernels would be seriously interesting. Very curious what they are capable of.
Yes i have found plenty of blocks with that miner
Greets
thanks for the tests.
in the end the performance is 0-20% higher than a 290x depending on the algo:
sia - about the same
groestl - about 10% more
x11 and quark - about 20% more
let's hope they make a better opencl compiler to unleash its full power!
finally a curiosity: did you get valid blocks with that ultra fast sia miner?
in the end the performance is 0-20% higher than a 290x depending on the algo:
sia - about the same
groestl - about 10% more
x11 and quark - about 20% more
let's hope they make a better opencl compiler to unleash its full power!
finally a curiosity: did you get valid blocks with that ultra fast sia miner?
Siacoin hashrate is impressive! Are you using the public miner?
I'm interested in knowing hashrate and power usage for the groestlcoin/diamond algo.
Please use both the opensource v1 kernel and the experimental v2 binary in my thread.
Many thanks!
I'm interested in knowing hashrate and power usage for the groestlcoin/diamond algo.
Please use both the opensource v1 kernel and the experimental v2 binary in my thread.
Many thanks!
Hey Pallas.
no i'm not using the public miner, hashrate on the public miner is just above 1Ghz
As for your Groestl/diamond interest.
I installed this version of sgminer
https://github.com/prettyhatemachine/sph-sgminer
when running Groestl/diamond on stock settings "1050 engine" i hit around @ 778 Watt
Watt
http://postimg.org/image/v14qbrulp/
When i use your Hawaii binary, rename it, run it, i run around 40Mhz+ but with HW errors
Hashrate was still rising and this is with stock clocks. So with 1100 (which those cards can easily handle it should go to 50Mhz)
Oh and a screenshot of Siacoin speed
Greetings
Siacoin hashrate is impressive! Are you using the public miner?
I'm interested in knowing hashrate and power usage for the groestlcoin/diamond algo.
Please use both the opensource v1 kernel and the experimental v2 binary in my thread.
Many thanks!
I'm interested in knowing hashrate and power usage for the groestlcoin/diamond algo.
Please use both the opensource v1 kernel and the experimental v2 binary in my thread.
Many thanks!
Jump to: