I need some help. I can't get my Ryzen 3900x to mine monero on RandomX any higher than 9200khs. I have the following set up...
MSI X470 Max
Ryzen 3900x
2 8gb Consair Vengence 3600mhz CL18 Ram
570 4gb card.
windows 10
I've tried stock CPU settings.
I've done the DRAM calculator... Anytime I enter in the timings from the calculator, the Memory just locks up and doesn't boot up into Windows 10.
I've done simple adjustments like OC the CPU from 3800mhz to 4100mhz.
Just looking for proper set up of timing to make this CPU mine atleast over 11000 khs.
Thanks in advance
Topic: AMD Ryzen 9 🖥️ 3900x RandomX mining benchmarks (Read 2175 times)
Hitting 15kh/s + on 3900x now. Could prob push it to nearly 16kh/s but 15 keeps it under 200 watts at the wall 
Is it with MSR mod? What's your memory setup and CPU speed?Nice stats, 15 kh/s really nice performance on 3900x
I've prepared a RandomX boost guide for Windows (it's the same MSR mod as the one I had for Linux before):
https://www.reddit.com/r/MoneroMining/comments/e9tuvd/randomx_boost_guide_for_ryzen_on_windows_9100_hs/
It gave me +6% hashrate. Can anyone test it?
https://www.reddit.com/r/MoneroMining/comments/e9tuvd/randomx_boost_guide_for_ryzen_on_windows_9100_hs/
It gave me +6% hashrate. Can anyone test it?
Great. My 3900x can do 14000 H/s now.
Hitting 15kh/s + on 3900x now. Could prob push it to nearly 16kh/s but 15 keeps it under 200 watts at the wall
Is it with MSR mod? What's your memory setup and CPU speed?Yup, 4.1 GHZ
Code:
[2019-12-13 18:51:34.733] net new job from xmr-us-east1.nanopool.org:14444 diff 480045 algo rx/0 height 1988139
[2019-12-13 18:51:45.276] cpu accepted (3002/6) diff 480045 (146 ms)
[2019-12-13 18:51:55.512] speed 10s/60s/15m 15096.1 15094.6 15094.6 H/s max 15104.3 H/s
[2019-12-13 18:52:34.973] net new job from xmr-us-east1.nanopool.org:14444 diff 480045 algo rx/0 height 1988139
[2019-12-13 18:52:55.567] speed 10s/60s/15m 15094.9 15094.6 15094.6 H/s max 15104.3 H/s
[2019-12-13 18:53:31.256] cpu accepted (3003/6) diff 480045 (164 ms)
[2019-12-13 18:53:32.369] cpu accepted (3004/6) diff 480045 (156 ms)
[2019-12-13 18:53:34.975] net new job from xmr-us-east1.nanopool.org:14444 diff 480045 algo rx/0 height 1988139
[2019-12-13 18:53:43.023] cpu accepted (3005/6) diff 480045 (163 ms)
[2019-12-13 18:53:48.818] cpu accepted (3006/6) diff 480045 (155 ms)
[2019-12-13 18:53:55.624] speed 10s/60s/15m 15095.0 15093.6 15094.6 H/s max 15104.3 H/s
[2019-12-13 18:54:26.155] cpu accepted (3007/6) diff 480045 (159 ms)
[2019-12-13 18:54:35.170] net new job from xmr-us-east1.nanopool.org:14444 diff 480045 algo rx/0 height 1988139
[2019-12-13 18:54:50.972] net new job from xmr-us-east1.nanopool.org:14444 diff 480045 algo rx/0 height 1988140
[2019-12-13 18:54:55.684] speed 10s/60s/15m 15092.9 15095.1 15094.6 H/s max 15104.3 H/s
[2019-12-13 18:55:08.841] cpu accepted (3008/6) diff 480045 (162 ms)
[2019-12-13 18:55:41.519] cpu accepted (3009/6) diff 480045 (164 ms)
[2019-12-13 18:55:45.493] cpu accepted (3010/6) diff 480045 (168 ms)
[2019-12-13 18:55:51.156] net new job from xmr-us-east1.nanopool.org:14444 diff 480045 algo rx/0 height 1988140
[2019-12-13 18:55:55.742] speed 10s/60s/15m 15092.7 15085.0 15093.9 H/s max 15104.3 H/s
[2019-12-13 18:56:03.647] cpu accepted (3011/6) diff 480045 (163 ms)
[2019-12-13 18:56:51.194] net new job from xmr-us-east1.nanopool.org:14444 diff 480045 algo rx/0 height 1988140
[2019-12-13 18:56:55.796] speed 10s/60s/15m 15093.4 15095.3 15094.0 H/s max 15104.3 H/s
[2019-12-13 18:57:20.811] cpu accepted (3012/6) diff 480045 (172 ms)
[2019-12-13 18:57:29.038] cpu accepted (3013/6) diff 480045 (166 ms)
[2019-12-13 18:57:51.367] net new job from xmr-us-east1.nanopool.org:14444 diff 480045 algo rx/0 height 1988140
[2019-12-13 18:57:51.965] cpu accepted (3014/6) diff 480045 (175 ms)
[2019-12-13 18:57:55.724] cpu accepted (3015/6) diff 480045 (159 ms)
[2019-12-13 18:57:55.838] speed 10s/60s/15m 15093.6 15095.8 15094.1 H/s max 15104.3 H/s
[2019-12-13 18:57:57.378] cpu accepted (3016/6) diff 480045 (165 ms)
Hitting 15kh/s + on 3900x now. Could prob push it to nearly 16kh/s but 15 keeps it under 200 watts at the wall
Really impressive jstefanop, which cooling solution are you using? and what temperature your 3900x is working.
Stock cooler and 3000 CL14 memory. Running around 78C
Hitting 15kh/s + on 3900x now. Could prob push it to nearly 16kh/s but 15 keeps it under 200 watts at the wall
Really impressive jstefanop, which cooling solution are you using? and what temperature your 3900x is working.
Hitting 15kh/s + on 3900x now. Could prob push it to nearly 16kh/s but 15 keeps it under 200 watts at the wall
Is it with MSR mod? What's your memory setup and CPU speed?
Hitting 15kh/s + on 3900x now. Could prob push it to nearly 16kh/s but 15 keeps it under 200 watts at the wall
I am a fan of AMD procie, and if we talk about the price AMD processors are much cheaper compare to Intel Processors, glad to see that they launched the ryzen 9 I hope that it will available soon in my country.
guys who where telling me that 2 memory sticks will be cap, and there wont be any speed increase if i add more memory where wrong. Ading 2 more sticks to my current system increased my performance by 10% without any over-clocks. Video : https://www.youtube.com/watch?v=xeVhjAfgMqw&t
That is an interesting find, I would not have thought that anything other than dual channel memory or quad channel memory(in a system that supports it) would have an impact the hash rate. It makes me wonder if populating all the memory channels in a threadripper system would have similar results. I do not think the gains are enough to warrant the extra memory, but it would be a fun experiment just for the sake of science.
I like your build in the video, the MSI cards look pretty sweet
Yes its all about caclulation best price per hash. Like my RAM was quite cheap 3600 mhz 8gb CL19 was 50euro for a pair. So it makes sense to pay extra 5% to get 10% speed increase. I wonder if i had better quality ram, would it affect my performance.
P.S man these MSI RX5700 are so sexy they really feel premium cards in metal case.
I've prepared a RandomX boost guide for Windows (it's the same MSR mod as the one I had for Linux before):
https://www.reddit.com/r/MoneroMining/comments/e9tuvd/randomx_boost_guide_for_ryzen_on_windows_9100_hs/
It gave me +6% hashrate. Can anyone test it?
https://www.reddit.com/r/MoneroMining/comments/e9tuvd/randomx_boost_guide_for_ryzen_on_windows_9100_hs/
It gave me +6% hashrate. Can anyone test it?
guys who where telling me that 2 memory sticks will be cap, and there wont be any speed increase if i add more memory where wrong. Ading 2 more sticks to my current system increased my performance by 10% without any over-clocks. Video : https://www.youtube.com/watch?v=xeVhjAfgMqw&t
That is an interesting find, I would not have thought that anything other than dual channel memory or quad channel memory(in a system that supports it) would have an impact the hash rate. It makes me wonder if populating all the memory channels in a threadripper system would have similar results. I do not think the gains are enough to warrant the extra memory, but it would be a fun experiment just for the sake of science.
I like your build in the video, the MSI cards look pretty sweet
guys who where telling me that 2 memory sticks will be cap, and there wont be any speed increase if i add more memory where wrong. Ading 2 more sticks to my current system increased my performance by 10% without any over-clocks. Video : https://www.youtube.com/watch?v=xeVhjAfgMqw&t
Yes, the theory is that having more independent memory ranks is better for RandomX. Two dual-rank sticks are better than two single-rank sticks, 4 same-rank sticks are better than 2 same-rank sticks. But more sticks and dual-rank sticks usually mean worse timings, so speedup is not guaranteed.
guys who where telling me that 2 memory sticks will be cap, and there wont be any speed increase if i add more memory where wrong. Ading 2 more sticks to my current system increased my performance by 10% without any over-clocks. Video : https://www.youtube.com/watch?v=xeVhjAfgMqw&t
Can anyone proficient in Linux try this RandomX boost guide?
https://www.reddit.com/r/MoneroMining/comments/e962fu/9526_hs_on_ryzen_7_3700x_xmrig_520_1gb_pages_msr/
I was able to get +5% on my R7 3700X, I think 9526 h/s is a record for this CPU now.
https://www.reddit.com/r/MoneroMining/comments/e962fu/9526_hs_on_ryzen_7_3700x_xmrig_520_1gb_pages_msr/
I was able to get +5% on my R7 3700X, I think 9526 h/s is a record for this CPU now.
What kind of power draw are you getting with this setup, I imagine its something north of 150+ watts for the cpu give or take 10 watts.
This setup got me to 8880 h/s from my original 8444 h/s on a 3ghz threadripper and dropped the power draw by about 9-10 watts going from 185 to 175-176 at the wall. I am sure I could get it to 10000+ h/s but that would take water and a much larger power draw. Pretty happy with this gain and drop in power consumption
Can anyone proficient in Linux try this RandomX boost guide?
https://www.reddit.com/r/MoneroMining/comments/e962fu/9526_hs_on_ryzen_7_3700x_xmrig_520_1gb_pages_msr/
I was able to get +5% on my R7 3700X, I think 9526 h/s is a record for this CPU now.
https://www.reddit.com/r/MoneroMining/comments/e962fu/9526_hs_on_ryzen_7_3700x_xmrig_520_1gb_pages_msr/
I was able to get +5% on my R7 3700X, I think 9526 h/s is a record for this CPU now.
What kind of power draw are you getting with this setup, I imagine its something north of 150+ watts for the cpu give or take 10 watts.
Can anyone proficient in Linux try this RandomX boost guide?
https://www.reddit.com/r/MoneroMining/comments/e962fu/9526_hs_on_ryzen_7_3700x_xmrig_520_1gb_pages_msr/
I was able to get +5% on my R7 3700X, I think 9526 h/s is a record for this CPU now.
https://www.reddit.com/r/MoneroMining/comments/e962fu/9526_hs_on_ryzen_7_3700x_xmrig_520_1gb_pages_msr/
I was able to get +5% on my R7 3700X, I think 9526 h/s is a record for this CPU now.
What kind of power draw are you getting with this setup, I imagine its something north of 150+ watts for the cpu give or take 10 watts.
Can anyone proficient in Linux try this RandomX boost guide?
https://www.reddit.com/r/MoneroMining/comments/e962fu/9526_hs_on_ryzen_7_3700x_xmrig_520_1gb_pages_msr/
I was able to get +5% on my R7 3700X, I think 9526 h/s is a record for this CPU now.
https://www.reddit.com/r/MoneroMining/comments/e962fu/9526_hs_on_ryzen_7_3700x_xmrig_520_1gb_pages_msr/
I was able to get +5% on my R7 3700X, I think 9526 h/s is a record for this CPU now.
I tested this RagerX yesterday and it was slower than XMRig on my R7 3700X. 8200 h/s RagerX vs 8450 h/s XMRig (all stock 3700X with tuned memory).
well as your the xmrrig dev and are prob optimising your miner for your own equipment its not really a fair comparison is it?
My equipment is pretty standard, I just make sure everything is tested in perfect conditions - properly tuned, no background programs eating CPU, updates and unneeded Windows services turned off etc. I need this to easily reproduce the same hashrate at every test - it's hard to tell if some change in the code improves hashrate +0.1% or makes it worse -0.1% without this kind of stability."According to the devs..."
Why does CMB publish this kind of crap without any attempt to verify the outrageous claims.
Any miner that locks you into their ecosystem is to be avoided. There's no telling what kind of malware
might be buried in their OS.
No one is going to magically come up with a new faster implementation. There is strong competition between
xmrig & xmr-stak, not to mention Tevador's reference implementation. I doubt there is much room for improvement.
I tested this RagerX yesterday and it was slower than XMRig on my R7 3700X. 8200 h/s RagerX vs 8450 h/s XMRig (all stock 3700X with tuned memory).
well as your the xmrrig dev and are prob optimising your miner for your own equipment its not really a fair comparison is it?
My equipment is pretty standard, I just make sure everything is tested in perfect conditions - properly tuned, no background programs eating CPU, updates and unneeded Windows services turned off etc. I need this to easily reproduce the same hashrate at every test - it's hard to tell if some change in the code improves hashrate +0.1% or makes it worse -0.1% without this kind of stability. Jump to: