Mining: AMD Opteron or Intel Xeon

[jims2321]

Have you tried bigger number of cores? Like 10-11 or even all 12? I'm getting best results (around 500h/s) on 6276 with 13 of 16 cores enabled. And that with all the same 16MB L3.
And also - do you have Turbo core enabled in BIOS? Without it I'm getting like 425-435 H/s, and with it enabled - all the way to 490-505 H/s. But it draws more electricity then
BTW what is the amount of RAM do you have?

Tim,

Sorry been sidetracked by work lately. I finally got the additional ram for my H8DGU, and wouldn't you know the second cpu socket was bad. After replacing the motherboard with an new H8DGU board, installed the ram and fired it up, allowed xmr-cpu-stak to auto config. I have been getting pretty steady hashrate around 820.7 H/s

[2017-10-12 19:27:11] : New block detected.
[2017-10-12 19:27:42] : Result accepted by the pool.
HASHRATE REPORT
| ID | 2.5s | 60s | 15m | ID | 2.5s | 60s | 15m |
| 0 | 45.4 | 45.8 | 45.5 | 1 | 31.1 | 31.3 | 31.2 |
| 2 | 33.6 | 33.8 | 33.7 | 3 | 33.6 | 33.8 | 33.7 |
| 4 | 58.7 | 58.7 | 58.7 | 5 | 59.1 | 59.1 | 59.2 |
| 6 | 59.1 | 59.1 | 59.1 | 7 | 43.2 | 43.2 | 43.3 |
| 8 | 43.2 | 43.2 | 43.3 | 9 | 43.6 | 43.6 | 43.6 |
| 10 | 43.6 | 43.6 | 43.6 | 11 | 59.2 | 59.2 | 59.2 |
| 12 | 59.2 | 59.2 | 59.2 | 13 | 59.2 | 59.2 | 59.2 |
| 14 | 45.8 | 45.8 | 45.8 | 15 | 31.2 | 31.2 | 31.2 |
| 16 | 33.7 | 33.7 | 33.7 | 17 | 33.7 | 33.7 | 33.7 |
-----------------------------------------------------
Totals: 816.4 817.3 816.6 H/s
Highest: 820.7 H/s
[2017-10-12 19:29:03] : New block detected.

I have not played with the turbo settings in bios, and this is with 2 6238 cpu's.

Tim if you have some recommend bios settings that I might try, by all means let me know.

Jim

 

[Klee]

I have some experience with FX cpu's, they seem to run best with 1 thread per module and for some reason running the threads on the odd numbered cores it hashed at a higher rate.

I have been wanting to play around with opterons but have not come across any screaming deals, as in almost free, yet.

So how about giving this a go?:

Don't do the auto config, you have two 12 core opterons with 6 x 2mb L2 cache each.
Try 12 threads and run them on the odd number cores and see what that does.

Something like this:
"cpu_threads_conf" :
[
{ "low_power_mode" : false, "no_prefetch" : false, "affine_to_cpu" : 1 },
{ "low_power_mode" : false, "no_prefetch" : false, "affine_to_cpu" : 3 },
{ "low_power_mode" : false, "no_prefetch" : false, "affine_to_cpu" : 5},
{ "low_power_mode" : false, "no_prefetch" : false, "affine_to_cpu" : 7},
{ "low_power_mode" : false, "no_prefetch" : false, "affine_to_cpu" : 9 },
{ "low_power_mode" : false, "no_prefetch" : false, "affine_to_cpu" : 11},
{ "low_power_mode" : false, "no_prefetch" : false, "affine_to_cpu" : 13 },
{ "low_power_mode" : false, "no_prefetch" : false, "affine_to_cpu" : 15 },
{ "low_power_mode" : false, "no_prefetch" : false, "affine_to_cpu" : 17},
{ "low_power_mode" : false, "no_prefetch" : false, "affine_to_cpu" : 19 },
{ "low_power_mode" : false, "no_prefetch" : false, "affine_to_cpu" : 21 },
{ "low_power_mode" : false, "no_prefetch" : false, "affine_to_cpu" : 23 },

],

 

[jims2321]

I have some experience with FX cpu's, they seem to run best with 1 thread per module and for some reason running the threads on the odd numbered cores it hashed at a higher rate.

I have been wanting to play around with opterons but have not come across any screaming deals, as in almost free, yet.

So how about giving this a go?:

Don't do the auto config, you have two 12 core opterons with 6 x 2mb L2 cache each.
Try 12 threads and run them on the odd number cores and see what that does.

Something like this:
"cpu_threads_conf" :
[
{ "low_power_mode" : false, "no_prefetch" : false, "affine_to_cpu" : 1 },
{ "low_power_mode" : false, "no_prefetch" : false, "affine_to_cpu" : 3 },
{ "low_power_mode" : false, "no_prefetch" : false, "affine_to_cpu" : 5},
{ "low_power_mode" : false, "no_prefetch" : false, "affine_to_cpu" : 7},
{ "low_power_mode" : false, "no_prefetch" : false, "affine_to_cpu" : 9 },
{ "low_power_mode" : false, "no_prefetch" : false, "affine_to_cpu" : 11},
{ "low_power_mode" : false, "no_prefetch" : false, "affine_to_cpu" : 13 },
{ "low_power_mode" : false, "no_prefetch" : false, "affine_to_cpu" : 15 },
{ "low_power_mode" : false, "no_prefetch" : false, "affine_to_cpu" : 17},
{ "low_power_mode" : false, "no_prefetch" : false, "affine_to_cpu" : 19 },
{ "low_power_mode" : false, "no_prefetch" : false, "affine_to_cpu" : 21 },
{ "low_power_mode" : false, "no_prefetch" : false, "affine_to_cpu" : 23 },

],

Will do, get it changed this evening and fire it up. I will let you know tomorrow, what changes in hashrate I see.

Jim

 

[jims2321]

Okay,

Made the changes, and while individual cpu hashrate increased, overall hashrate dropped.

[2017-10-13 20:47:39] : New block detected.
[2017-10-13 20:47:49] : Result accepted by the pool.
HASHRATE REPORT
| ID | 2.5s | 60s | 15m | ID | 2.5s | 60s | 15m |
| 0 | 59.1 | 59.1 | (na) | 1 | 60.3 | 60.3 | (na) |
| 2 | 60.3 | 60.3 | (na) | 3 | 60.3 | 60.3 | (na) |
| 4 | 60.3 | 60.3 | (na) | 5 | 60.2 | 60.2 | (na) |
| 6 | 60.3 | 60.4 | (na) | 7 | 60.3 | 60.3 | (na) |
| 8 | 60.3 | 60.3 | (na) | 9 | 60.3 | 60.3 | (na) |
| 10 | 60.3 | 60.3 | (na) | 11 | 60.3 | 60.3 | (na) |
-----------------------------------------------------
Totals: 722.7 722.6 (na) H/s
Highest: 722.8 H/s

 

[Klee]

Its also way more consistent.

I bet power use and temps are also noticeably lower.

If you can measure the power usage can you compare H/s to watts?

It might be worth running more efficiently than trying for the absolute highest hash rate.

 

[jims2321]

Its also way more consistent.

I bet power use and temps are also noticeably lower.

If you can measure the power usage can you compare H/s to watts?

It might be worth running more efficiently than trying for the absolute highest hash rate.

I should be able to do that. Let me my killawatt setup and run it both ways for 24 hours.

Jim

 

[Joel Wu]

I am running an SuperMicro H8DGU-F with (1) Opteron 6238 active (waiting on some more ram to turn on the second cpu) what sort of hash rate should I be seeing.. the L3 cache is 16 so with 9 threads I am getting around 400+ h/s. I don't see how diklooc could get that rate, maybe 800 h/s. But if you have some insight into tuning the to achieve that rate I am all ears.

config.txt cpu info

"cpu_threads_conf" :
[
{ "low_power_mode" : true, "no_prefetch" : true, "affine_to_cpu" : 0 },
{ "low_power_mode" : false, "no_prefetch" : true, "affine_to_cpu" : 1 },
{ "low_power_mode" : false, "no_prefetch" : true, "affine_to_cpu" : 2 },
{ "low_power_mode" : false, "no_prefetch" : true, "affine_to_cpu" : 3 },
{ "low_power_mode" : false, "no_prefetch" : true, "affine_to_cpu" : 4 },
{ "low_power_mode" : false, "no_prefetch" : true, "affine_to_cpu" : 6 },
{ "low_power_mode" : false, "no_prefetch" : true, "affine_to_cpu" : 8 },
{ "low_power_mode" : true, "no_prefetch" : true, "affine_to_cpu" : 10 },
{ "low_power_mode" : true, "no_prefetch" : true, "affine_to_cpu" : 11 },
],

[2017-09-02 19:07:57] : New block detected.
HASHRATE REPORT
| ID | 2.5s | 60s | 15m | ID | 2.5s | 60s | 15m |
| 0 | 45.4 | 45.3 | 45.3 | 1 | 31.1 | 31.1 | 31.1 |
| 2 | 33.6 | 33.6 | 33.6 | 3 | 33.7 | 33.6 | 33.6 |
| 4 | 58.7 | 58.6 | 58.6 | 5 | 59.1 | 59.0 | 59.2 |
| 6 | 59.2 | 59.1 | 59.1 | 7 | 43.3 | 43.1 | 43.3 |
| 8 | 43.3 | 43.1 | 43.3 |
-----------------------------------------------------
Totals: 407.5 406.6 407.1 H/s
Highest: 411.5 H/s
miner@monerominer2:~$

Hey man what is your powerdraw like. I have a C7000 set up with BL685 G7 blades, with the same processor. You say you only run 9 threads out of 12. Do you think running all 12 threads will make a significant diff in wattage?

I'm trying to keep my wattage down, but still get a pretty high hash rate.

EDIT: Sorry didn't read your last posts properly. I'm a noob. How do you check your power usage and hash rate in XMR-STAK

 

[jims2321]

Hey man what is your powerdraw like. I have a C7000 set up with BL685 G7 blades, with the same processor. You say you only run 9 threads out of 12. Do you think running all 12 threads will make a significant diff in wattage?

I'm trying to keep my wattage down, but still get a pretty high hash rate.

EDIT: Sorry didn't read your last posts properly. I'm a noob. How do you check your power usage and hash rate in XMR-STAK

Joel,

I run both the 6 cores, but with of the hyperthreads running as well, so the recommended config is 18 threads

{ "low_power_mode" : true, "no_prefetch" : true, "affine_to_cpu" : 0 },
{ "low_power_mode" : false, "no_prefetch" : true, "affine_to_cpu" : 1 },
{ "low_power_mode" : false, "no_prefetch" : true, "affine_to_cpu" : 2 },
{ "low_power_mode" : false, "no_prefetch" : true, "affine_to_cpu" : 3 },
{ "low_power_mode" : false, "no_prefetch" : true, "affine_to_cpu" : 4 },
{ "low_power_mode" : false, "no_prefetch" : true, "affine_to_cpu" : 6 },
{ "low_power_mode" : false, "no_prefetch" : true, "affine_to_cpu" : 8 },
{ "low_power_mode" : true, "no_prefetch" : true, "affine_to_cpu" : 10 },
{ "low_power_mode" : true, "no_prefetch" : true, "affine_to_cpu" : 11 },
{ "low_power_mode" : true, "no_prefetch" : true, "affine_to_cpu" : 12 },
{ "low_power_mode" : true, "no_prefetch" : true, "affine_to_cpu" : 13 },
{ "low_power_mode" : false, "no_prefetch" : true, "affine_to_cpu" : 14 },
{ "low_power_mode" : false, "no_prefetch" : true, "affine_to_cpu" : 16 },
{ "low_power_mode" : false, "no_prefetch" : true, "affine_to_cpu" : 18 },
{ "low_power_mode" : true, "no_prefetch" : true, "affine_to_cpu" : 20 },
{ "low_power_mode" : false, "no_prefetch" : true, "affine_to_cpu" : 21 },
{ "low_power_mode" : false, "no_prefetch" : true, "affine_to_cpu" : 22 },
{ "low_power_mode" : false, "no_prefetch" : true, "affine_to_cpu" : 23 },

The 24 hour numbers are:

idle wattage is 144 watts
18 threads running:

mining 281 watts
24 hour run total wattage 6.95Kwh
hashrate average 817.7 h/s
hashrate/watt 2.94

Just started with running only the physical cores per Klee suggestion

12 threads

{ "low_power_mode" : false, "no_prefetch" : false, "affine_to_cpu" : 1 },
{ "low_power_mode" : false, "no_prefetch" : false, "affine_to_cpu" : 3 },
{ "low_power_mode" : false, "no_prefetch" : false, "affine_to_cpu" : 5},
{ "low_power_mode" : false, "no_prefetch" : false, "affine_to_cpu" : 7},
{ "low_power_mode" : false, "no_prefetch" : false, "affine_to_cpu" : 9 },
{ "low_power_mode" : false, "no_prefetch" : false, "affine_to_cpu" : 11},
{ "low_power_mode" : false, "no_prefetch" : false, "affine_to_cpu" : 13 },
{ "low_power_mode" : false, "no_prefetch" : false, "affine_to_cpu" : 15 },
{ "low_power_mode" : false, "no_prefetch" : false, "affine_to_cpu" : 17},
{ "low_power_mode" : false, "no_prefetch" : false, "affine_to_cpu" : 19 },
{ "low_power_mode" : false, "no_prefetch" : false, "affine_to_cpu" : 21 },
{ "low_power_mode" : false, "no_prefetch" : false, "affine_to_cpu" : 23 },

Initial numbers:

mining 263 watts
hashrate average 722.7 h/s
hashrate/watt 2.74

Estimate electric cost difference per month of $1.55

Just a note, I don't expect the hashrate/sec to change when I originally ran the 12 threads last week, it was pretty rock solid at that rate. So the 18 threads consumes about 18 more watts, for a .2 increase of hash/watt.

I am still playing around with the thread configuration a few hours on the weekend, but these older AMD opterons don't have a lot of wiggle room for tweaking.

I sat with my system hardware people last week looking at AMD Epyc versus Intel scalable processors for some upgrade projects, and got to talking to them about the my crypto mining rig during a break. They were pretty much in agreement, what I am seeing is the upper limits of practical configuration, without overclocking and sending power consumption thru the roof.

Jim

 

[jims2321]

Just some money numbers to go along with the hash numbers. Running this thru nicehash calculator for cryptonight algorithm. Monthly numbers. This is not earth shattering numbers I will never retire from running my rig. It's an interesting experiment and fun to play with.

18 threads:

Income $50.81
Power Cost $24.28
Profit $26.53

12 threads:

Income $45.21
Power Cost $22.72
Profit $22.49

Electric cost diff $1.55
Profit difference $4.04
Net difference between 18 versus 12 threads $2.49/m

The profit numbers should go up as my electric rate drops to .08 during the Oct-March time frame, but the gap will not increase between 18 and 12 threads.

 

[Joel Wu]

Just some money numbers to go along with the hash numbers. Running this thru nicehash calculator for cryptonight algorithm. Monthly numbers. This is not earth shattering numbers I will never retire from running my rig. It's an interesting experiment and fun to play with.

18 threads:

Income $50.81
Power Cost $24.28
Profit $26.53

12 threads:

Income $45.21
Power Cost $22.72
Profit $22.49

Electric cost diff $1.55
Profit difference $4.04
Net difference between 18 versus 12 threads $2.49/m

The profit numbers should go up as my electric rate drops to .08 during the Oct-March time frame, but the gap will not increase between 18 and 12 threads.

I finally managed to set my C7000 running with 1 BL685C G7 Blade. Those have 4 AMD Opteron 6238s in them, I'm using 47 threads on CPUMiner-Multi but the screen says I'm only getting round 1160h/s. That doesn't seem right for 4 CPUs...

 

[Klee]

Joel,

I run both the 6 cores, but with of the hyperthreads running as well, so the recommended config is 18 threads

{ "low_power_mode" : true, "no_prefetch" : true, "affine_to_cpu" : 0 },
{ "low_power_mode" : false, "no_prefetch" : true, "affine_to_cpu" : 1 },
{ "low_power_mode" : false, "no_prefetch" : true, "affine_to_cpu" : 2 },
{ "low_power_mode" : false, "no_prefetch" : true, "affine_to_cpu" : 3 },
{ "low_power_mode" : false, "no_prefetch" : true, "affine_to_cpu" : 4 },
{ "low_power_mode" : false, "no_prefetch" : true, "affine_to_cpu" : 6 },
{ "low_power_mode" : false, "no_prefetch" : true, "affine_to_cpu" : 8 },
{ "low_power_mode" : true, "no_prefetch" : true, "affine_to_cpu" : 10 },
{ "low_power_mode" : true, "no_prefetch" : true, "affine_to_cpu" : 11 },
{ "low_power_mode" : true, "no_prefetch" : true, "affine_to_cpu" : 12 },
{ "low_power_mode" : true, "no_prefetch" : true, "affine_to_cpu" : 13 },
{ "low_power_mode" : false, "no_prefetch" : true, "affine_to_cpu" : 14 },
{ "low_power_mode" : false, "no_prefetch" : true, "affine_to_cpu" : 16 },
{ "low_power_mode" : false, "no_prefetch" : true, "affine_to_cpu" : 18 },
{ "low_power_mode" : true, "no_prefetch" : true, "affine_to_cpu" : 20 },
{ "low_power_mode" : false, "no_prefetch" : true, "affine_to_cpu" : 21 },
{ "low_power_mode" : false, "no_prefetch" : true, "affine_to_cpu" : 22 },
{ "low_power_mode" : false, "no_prefetch" : true, "affine_to_cpu" : 23 },

The 24 hour numbers are:

idle wattage is 144 watts
18 threads running:

mining 281 watts
24 hour run total wattage 6.95Kwh
hashrate average 817.7 h/s
hashrate/watt 2.94

Just started with running only the physical cores per Klee suggestion

12 threads

{ "low_power_mode" : false, "no_prefetch" : false, "affine_to_cpu" : 1 },
{ "low_power_mode" : false, "no_prefetch" : false, "affine_to_cpu" : 3 },
{ "low_power_mode" : false, "no_prefetch" : false, "affine_to_cpu" : 5},
{ "low_power_mode" : false, "no_prefetch" : false, "affine_to_cpu" : 7},
{ "low_power_mode" : false, "no_prefetch" : false, "affine_to_cpu" : 9 },
{ "low_power_mode" : false, "no_prefetch" : false, "affine_to_cpu" : 11},
{ "low_power_mode" : false, "no_prefetch" : false, "affine_to_cpu" : 13 },
{ "low_power_mode" : false, "no_prefetch" : false, "affine_to_cpu" : 15 },
{ "low_power_mode" : false, "no_prefetch" : false, "affine_to_cpu" : 17},
{ "low_power_mode" : false, "no_prefetch" : false, "affine_to_cpu" : 19 },
{ "low_power_mode" : false, "no_prefetch" : false, "affine_to_cpu" : 21 },
{ "low_power_mode" : false, "no_prefetch" : false, "affine_to_cpu" : 23 },

Initial numbers:

mining 263 watts
hashrate average 722.7 h/s
hashrate/watt 2.74

Estimate electric cost difference per month of $1.55

Just a note, I don't expect the hashrate/sec to change when I originally ran the 12 threads last week, it was pretty rock solid at that rate. So the 18 threads consumes about 18 more watts, for a .2 increase of hash/watt.

I am still playing around with the thread configuration a few hours on the weekend, but these older AMD opterons don't have a lot of wiggle room for tweaking.

I sat with my system hardware people last week looking at AMD Epyc versus Intel scalable processors for some upgrade projects, and got to talking to them about the my crypto mining rig during a break. They were pretty much in agreement, what I am seeing is the upper limits of practical configuration, without overclocking and sending power consumption thru the roof.

Jim

I guess the FX cpu's with their much higher clocks and higher power usage using less cores is more efficient vs the Opterons lower speed and power requirements.

 

[dwright1542]

I finally managed to set my C7000 running with 1 BL685C G7 Blade. Those have 4 AMD Opteron 6238s in them, I'm using 47 threads on CPUMiner-Multi but the screen says I'm only getting round 1160h/s. That doesn't seem right for 4 CPUs...

What OS? Did you set 4 different processes for the 4 differenet NUMA nodes? That makes a HUGE difference.

 

[Joel Wu]

What OS? Did you set 4 different processes for the 4 differenet NUMA nodes? That makes a HUGE difference.

I'm installed ubuntu server on a blade. I'm a noob though, so I don't know if its different processes for 4 NUMA nodes. I don't know how to set them. Mind pointing me in the right direction?

 

[dwright1542]

Check out
How to tell CPU topology on Linux - Fibrevillage

You'll need 4 different mining setups, each pointed to the virtual CPU's on each socket.

TURN OFF HYPERTHREADING

On xmrig, for example, on a dual 6 core, I have
"cpu-affinity":"0x3F" for one thread which specifies CPU's 0-5 and
cpu-affinity": "0xFC0 for the second which specifies CPU's 6-10

 

[Greg C]

Guys I have several options regarding this. I currently have a R815 Dell Poweredge Server, with 4 AMD Opteron 6174 processors. Those are 12 cores and I thought that they would be slightly worse than the 6274 in terms of hashing power, but so far, it was dismal. Bloody 450h/s on all 48 cores. What's worse is that my power draw was around 730watts.

So right now I am trying a C7000 blade server with BL685 blades. The G7 ones I've got have 4 AMD Opteron 6238 processors. Those have the AES instruction set that the 6174 does not. But I am still unsure as to the profitability of XMR mining using that.

Another option would be to buy RX460 GPUs and affix them to the Dell servers but yet again I am yet to test the hash rate and the power draw.

If any of you could enlighten me please do let me know.

Joel, the 61xx performance is almost tragicomic. I can envision the screaming fans and wildly spinning power meter thrashing wildly to poop out ....450 H/S.

The AES crypto performance probably explains the puzzling price of the Bulldozer / Pilredriver based Opteron CPUs. I'm another Noob, piecing together a 2 x G34 box for chess analysis and just "to fiddle with." (Unlike Monero, the 61xx chips go quite well running chess engines. If you're looking for another benchmark, fire up Ipman Chess) After buying a mobo and PS, I went shopping for CPUs. The 61xx Phenom-based chips are under $10 a piece, while the Dozer-based chips go for $50~100.

Profitability? Maybe...
Assuming decent scaling, Jim's numbers suggest you'll get about 1500 H/S.
On the off chance you haven't see these sites:
Here's a handy calculator. Mining Calculator Bitcoin, Ethereum, Litecoin, Dash and Monero
and assorted benchmarks
GPU & CPU BENCKMARKS FOR MONERO MINING!

 

[Joel Wu]

Joel, the 61xx performance is almost tragicomic. I can envision the screaming fans and wildly spinning power meter thrashing wildly to poop out ....450 H/S.

The AES crypto performance probably explains the puzzling price of the Bulldozer / Pilredriver based Opteron CPUs. I'm another Noob, piecing together a 2 x G34 box for chess analysis and just "to fiddle with." (Unlike Monero, the 61xx chips go quite well running chess engines. If you're looking for another benchmark, fire up Ipman Chess) After buying a mobo and PS, I went shopping for CPUs. The 61xx Phenom-based chips are under $10 a piece, while the Dozer-based chips go for $50~100.

Profitability? Maybe...
Assuming decent scaling, Jim's numbers suggest you'll get about 1500 H/S.
On the off chance you haven't see these sites:
Here's a handy calculator. Mining Calculator Bitcoin, Ethereum, Litecoin, Dash and Monero
and assorted benchmarks
GPU & CPU BENCKMARKS FOR MONERO MINING!

Hahaha no worries. I've since written off the 61xx series for Monero, but as for now, I have 4 of the processors that Jim has. What's puzzling to me is that I get ~1100 H/S as opposed to the 1500+ h/s I'm supposed to be getting. I haven't a clue as to why. I'm running cpuminer-multi with default settings (which are 47 threads for the 4 CPUs). If anyone has any idea please let me know.

I don't think Chess requires AES, which the 6238 Opteron does. That explains why the 61xx isn't too popular.

 

[dwright1542]

Hahaha no worries. I've since written off the 61xx series for Monero, but as for now, I have 4 of the processors that Jim has. What's puzzling to me is that I get ~1100 H/S as opposed to the 1500+ h/s I'm supposed to be getting. I haven't a clue as to why. I'm running cpuminer-multi with default settings (which are 47 threads for the 4 CPUs). If anyone has any idea please let me know.

I don't think Chess requires AES, which the 6238 Opteron does. That explains why the 61xx isn't too popular.

Already told you why. You need 4 workers, 1 per processor. Read the NUMA guide I linked above.

 

[spfoo]

Hahaha no worries. I've since written off the 61xx series for Monero, but as for now, I have 4 of the processors that Jim has. What's puzzling to me is that I get ~1100 H/S as opposed to the 1500+ h/s I'm supposed to be getting. I haven't a clue as to why. I'm running cpuminer-multi with default settings (which are 47 threads for the 4 CPUs). If anyone has any idea please let me know.

I don't think Chess requires AES, which the 6238 Opteron does. That explains why the 61xx isn't too popular.

You are running too many threads. 6238 is a 12-core CPU. On AMD Opteron you can't run 1 thread per core. You will probably max out with about 32 threads. More power efficient would be 24 threads - one on every second core. You can do this with taskset for example:
taskset -c 1,3,5,7,9,11,13,15,17,19, ...,47 minerd...

I have a 4 CPU board with one dead CPU socket so 3x6234. It gives 1100 h/s without Turbo core.

I haven't seen any benefit of using the NUMA isolation per CPU on Opteron 62xx's on dual or quad socket boards.

 

[jims2321]

To quickly identify the numa regions for your cpu's run "lscpu" at the prompt. with 4 opterons, I would expect there are 8 numa regions, 2 per cpu. So something like this:

NUMA node0 CPU(s): 0-5
NUMA node1 CPU(s): 6-11
.
.
.
NUMA node 7 CPU(s):41-47

Jim

 

[Joel Wu]

You are running too many threads. 6238 is a 12-core CPU. On AMD Opteron you can't run 1 thread per core. You will probably max out with about 32 threads. More power efficient would be 24 threads - one on every second core. You can do this with taskset for example:
taskset -c 1,3,5,7,9,11,13,15,17,19, ...,47 minerd...

I have a 4 CPU board with one dead CPU socket so 3x6234. It gives 1100 h/s without Turbo core.

I haven't seen any benefit of using the NUMA isolation per CPU on Opteron 62xx's on dual or quad socket boards.

Thanks for your input. I was just wondering if this was what I was supposed to do.

Sent from my LG-H815 using Tapatalk