Elder III
|
|
July 24, 2019, 03:38:52 PM |
|
there was a video of few + rx5700 doing ~1600-1700 and you say the same for one without screens not info, and just troll
thats my 5700 Looks pretty good. How much power? My Vega 56 do 1800 each but are fairly power hungry. My kilowatt wall meter said around 150w or 170w stock. I'm not too good at undervolting so I used the wattman auto undervolt template to get to 150w. doesn't seem to effect the hashrate too bad. the memory clock can be increased to raise the hashrate to 1750h/s but the watts gets around to 230w at that point What drivers and settings are you using in your config? I have tried using xmr stak a few times in the past week and have not had any success getting it to mine at all on a 5700 XT. *hmmm I noticed GPUz has you on 19.7.1 and I'm on 19.7.2
|
|
|
|
|
|
|
In order to achieve higher forum ranks, you need both activity points and merit points.
|
|
|
Advertised sites are not endorsed by the Bitcoin Forum. They may be unsafe, untrustworthy, or illegal in your jurisdiction.
|
|
|
alexforneus
Jr. Member
Offline
Activity: 133
Merit: 2
|
|
July 24, 2019, 04:31:32 PM |
|
after new lower prices this is not a bad variant i guess
|
|
|
|
ronnieb
Jr. Member
Offline
Activity: 198
Merit: 1
|
|
July 24, 2019, 04:48:43 PM Last edit: September 18, 2019, 10:14:44 PM by mprep |
|
My kilowatt wall meter said around 150w or 170w stock. I'm not too good at undervolting so I used the wattman auto undervolt template to get to 150w. doesn't seem to effect the hashrate too bad. the memory clock can be increased to raise the hashrate to 1750h/s but the watts gets around to 230w at that point
Use the Afterburner instead, enable voltage control and you can start to play with undervolting and OC. Are you sure that you adjusted only memory clocks? Higher memory clocks will need some extra power, but definitely not 80W more... Ill take another look when I get home today. I believe I took the mem clock from 1750 to 1850 and thats when the jump happened. Ill do another test with afterburner instead and see if i get different results.
there was a video of few + rx5700 doing ~1600-1700 and you say the same for one without screens not info, and just troll
thats my 5700 Looks pretty good. How much power? My Vega 56 do 1800 each but are fairly power hungry. My kilowatt wall meter said around 150w or 170w stock. I'm not too good at undervolting so I used the wattman auto undervolt template to get to 150w. doesn't seem to effect the hashrate too bad. the memory clock can be increased to raise the hashrate to 1750h/s but the watts gets around to 230w at that point What drivers and settings are you using in your config? I have tried using xmr stak a few times in the past week and have not had any success getting it to mine at all on a 5700 XT. *hmmm I noticed GPUz has you on 19.7.1 and I'm on 19.7.2 I downloaded the latest drivers from AMD for the rx 5700 card. I used XMR STAK auto config for crypto gpu and it started to hash around 1300 h/s I then went into the AMD config file and doubled the thread to 2. which increased the hashrate to around 1600h/s. Im surprised the XT is not functioning the exact same way or better.
|
|
|
|
andrucrypt
Member
Offline
Activity: 719
Merit: 49
|
|
July 24, 2019, 07:31:25 PM |
|
Got my own 5700 XT and what can I say... drivers are f*cking disaster %) They have finally implemented volatile keyword correctly for OpenCL, this is some kind of "plus", but this removed possibility to control register allocation a bit while compiling OpenCL kernels. reqd_work_group_size broken now, in some cases it can just crash compiler with error "ran out of registers". Memory operations... 10x times slower than on RX series ) AMD once again made dumb as hell OpenCL compiler, so developers should switch to GCN/RDNA assembler(or Vulkan at least), and showed how bad their driver team no matter what for so many years xD
|
|
|
|
WhyMe
|
|
July 24, 2019, 09:16:32 PM |
|
Well, that's a bad news
|
|
|
|
doktor83
|
|
July 25, 2019, 06:19:41 AM |
|
Got my own 5700 XT and what can I say... drivers are f*cking disaster %) They have finally implemented volatile keyword correctly for OpenCL, this is some kind of "plus", but this removed possibility to control register allocation a bit while compiling OpenCL kernels. reqd_work_group_size broken now, in some cases it can just crash compiler with error "ran out of registers". Memory operations... 10x times slower than on RX series ) AMD once again made dumb as hell OpenCL compiler, so developers should switch to GCN/RDNA assembler(or Vulkan at least), and showed how bad their driver team no matter what for so many years xD
Can you compile any of your CL kernels for 5700 ? so that they at least work
|
|
|
|
andrucrypt
Member
Offline
Activity: 719
Merit: 49
|
|
July 25, 2019, 06:40:35 AM Last edit: July 25, 2019, 08:26:40 AM by andrucrypt |
|
Can you compile any of your CL kernels for 5700 ? so that they at least work Yep, and compute power of this card is pretty good, close to Vega56/Vega64 with lower power consumption(can say about simple blake2b and bmw512 speeds) But... I don't understand this random nature of how it works xD Here some mind blowers: - first time I see loop not unrolled: the optimizer was unable to perform the requested transformation... for some kernels - in some cases compiler just throwing error when structure data types used and stop compile kernel; upd: found a problem, it's "tricky" pointer logic, not structure data types - simple sha256 kernel fails with AMD HSA Code Object loading failed... wtf %)
|
|
|
|
doktor83
|
|
July 25, 2019, 07:10:37 AM |
|
Can you compile any of your CL kernels for 5700 ? so that they at least work Yep, and compute power of this card is pretty good, close to Vega56/Vega64 with lower power consumption(can say about simple blake2b and bmw512 speeds) But... I don't understand this random nature of how it works xD Here some mind blowers: - first time I see loop not unrolled: the optimizer was unable to perform the requested transformation... for some kernels - in some cases compiler just throwing error when structure data types used and stop compile kernel - simple sha256 kernel fails with AMD HSA Code Object loading failed... wtf %) Yes, computing part should be even faster than on Vega64, or at least this is what the numbers say RDNA looks good on 'paper', time will show is this the case in practice. I don't have a 5700 card but when i tried to fix some of my kernels blindly, and compile them offline, the compiler was so friendly that it just said 3 errors found, but it won't tell me what are those
|
|
|
|
cikusa
Jr. Member
Offline
Activity: 39
Merit: 2
|
|
July 25, 2019, 08:06:57 AM |
|
Can you compile any of your CL kernels for 5700 ? so that they at least work Yep, and compute power of this card is pretty good, close to Vega56/Vega64 with lower power consumption(can say about simple blake2b and bmw512 speeds) But... I don't understand this random nature of how it works xD Here some mind blowers: - first time I see loop not unrolled: the optimizer was unable to perform the requested transformation... for some kernels - in some cases compiler just throwing error when structure data types used and stop compile kernel - simple sha256 kernel fails with AMD HSA Code Object loading failed... wtf %) Yes, computing part should be even faster than on Vega64, or at least this is what the numbers say RDNA looks good on 'paper', time will show is this the case in practice. I don't have a 5700 card but when i tried to fix some of my kernels blindly, and compile them offline, the compiler was so friendly that it just said 3 errors found, but it won't tell me what are those i could maybe provide u with tw access to my test rig with 5700xt that i got yesterday when i get back from buisness trip tommorow, if u are interested?
|
|
|
|
andrucrypt
Member
Offline
Activity: 719
Merit: 49
|
|
July 25, 2019, 09:14:44 AM |
|
Compiled all my main kernels... well, need to say bye-bye to writing any code using OpenCL from now xD Results are bad for complex kernels and playing in game "compiler, please, don't spill so many registers" is not fun anymore.
|
|
|
|
karim_hassan777
Newbie
Offline
Activity: 58
Merit: 0
|
|
July 25, 2019, 01:22:22 PM |
|
Compiled all my main kernels... well, need to say bye-bye to writing any code using OpenCL from now xD Results are bad for complex kernels and playing in game "compiler, please, don't spill so many registers" is not fun anymore.
what do you mean
|
|
|
|
jstefanop
Legendary
Offline
Activity: 2097
Merit: 1396
|
|
July 25, 2019, 03:41:33 PM |
|
Compiled all my main kernels... well, need to say bye-bye to writing any code using OpenCL from now xD Results are bad for complex kernels and playing in game "compiler, please, don't spill so many registers" is not fun anymore.
what do you mean The only way your going to get any good performance out of this architecture is with ASM kernels...which people are already running privately of course
|
|
|
|
ronnieb
Jr. Member
Offline
Activity: 198
Merit: 1
|
|
July 25, 2019, 05:23:32 PM |
|
Seems like whattomine is stating that ryo is the most profitable coin to mine at the moment for most Nvidia cards.
the hashrate for cryptonightgpu on a nvidia 2060 is comparable to amd 5700 at the moment
|
|
|
|
karim_hassan777
Newbie
Offline
Activity: 58
Merit: 0
|
|
July 25, 2019, 06:26:00 PM |
|
new drivers guys 19.7.3 has any one test it
|
|
|
|
ronnieb
Jr. Member
Offline
Activity: 198
Merit: 1
|
|
July 26, 2019, 12:38:58 AM |
|
Can anyone mine cryptonightgpu with the 5700XT version?
|
|
|
|
andrucrypt
Member
Offline
Activity: 719
Merit: 49
|
|
July 26, 2019, 05:52:33 AM Last edit: July 26, 2019, 11:59:07 AM by andrucrypt |
|
Released WildRig 0.19.0 preview with RX 5700 series support. Here some results for stock Sapphire RX 5700 XT(better results can be achieved with undervolt) blake2b - ~2.1GH/s lyra2v3 - ~72MH/s lyra2vc0ban - ~65MH/s Download and Details
|
|
|
|
Old_Timer
Member
Offline
Activity: 131
Merit: 13
In the fray since 2013.
|
|
July 26, 2019, 09:26:32 AM |
|
Released WildRig 0.19.0 preview with RX 5700 series support. Here some results for stock Sapphire RX 5700 XT(better results can be achieved with undervolt) blake2b - 2.1GH/s lyra2v3 - 72MH/s lyra2vc0ban - 65MH/s Download and DetailsCan anyone provide numbers for vega 64 / radeon vii in order to compare?
|
|
|
|
andrucrypt
Member
Offline
Activity: 719
Merit: 49
|
|
July 26, 2019, 09:33:12 AM |
|
Released WildRig 0.19.0 preview with RX 5700 series support. Here some results for stock Sapphire RX 5700 XT(better results can be achieved with undervolt) blake2b - 2.1GH/s lyra2v3 - 72MH/s lyra2vc0ban - 65MH/s Download and DetailsCan anyone provide numbers for vega 64 / radeon vii in order to compare? Vega 64 gives up to 90-100MH/s on Lyra2v3 as I know(with proper undervolt of course). And 5700 XT should lift more than current 72-78Mh/s, probably we will see performance near to Vega56/Vega64.
|
|
|
|
cikusa
Jr. Member
Offline
Activity: 39
Merit: 2
|
|
July 26, 2019, 10:55:14 AM |
|
Released WildRig 0.19.0 preview with RX 5700 series support. Here some results for stock Sapphire RX 5700 XT(better results can be achieved with undervolt) blake2b - 2.1GH/s lyra2v3 - 72MH/s lyra2vc0ban - 65MH/s Download and DetailsSo let's put it this way, i think navi got alot of potentional, i just got around 65mh on lyra2v3 with undervolt to 850mv and 1650mhz on core, power consumtion is around 110w in gpuz, i can't measure from wall atm, but i will post as soon i can do it
|
|
|
|
Elder III
|
|
July 26, 2019, 06:09:13 PM |
|
Released WildRig 0.19.0 preview with RX 5700 series support. Here some results for stock Sapphire RX 5700 XT(better results can be achieved with undervolt) blake2b - ~2.1GH/s lyra2v3 - ~72MH/s lyra2vc0ban - ~65MH/s Download and DetailsI've been mining to your VTC address for a little under 2 hours so far with a rx 5700 XT. So far the average hashrate has been ~72 mh/s on stock settings, using 180 watts and Temp of ~75c (92c Hotspot) at 50% fan. I have not tried adjusting any clocks or voltages.
|
|
|
|
|