Monero Mining Performance

[MiniKnight]

We All know the 2mb/thread cache requirement. What impact does the clock freq have on hash rate? I'm trying to see if the increased power and heat for the higher MHZ and the sweetspot of hashrate, wattage, cost. Aren't we all.

Sent from my Nexus 6 using Tapatalk

There's a site that did this
Monero Mining Benchmarks - CPU Mining With AMD Ryzen 7 Systems

 

[Marsh]

'm trying to see if the increased power and heat for the higher MHZ and the sweetspot of hashrate, wattage, cost.

It really depends on cost of electricity KW/H.
For me in CA, $0.28 to $0.40 KW/H.
If I am paying $0.0.7 KW/H, I'll run the fastest and power hungry CPU like E5-2670 v1, the cost differential is so minimal.

In fall and winter, I'll fire up a bank of 20 x E5 V1 nodes. Instead of paying for natural gas for heat, I'll use the miner.

 

[funkywizard]

seeing around 970 h/s with stak miner on dual e5-2660v2, with 8 of 10 cores enabled and 24 threads (12 per cpu). Power use of 154w according to ipmitool. Haven't measured with a killawatt yet.

 

[Marsh]

Looks like E5-2660 v2 is a good candidate for mining.
Hope E5-2660v2 price would drop more before winter.

 

[funkywizard]

the two servers I tested, e5-2660v1 vs v2, were different motherboard, chassis, etc. Will do an apples to apples comparison when I get a chance.

 

[Patrick]

Added Xeon Gold 6138
Added AMD Threadripper 1950X

 

[Marsh]

Wow, nice CPU
I am seeing my future miner if I wait for few years.

 

[Patrick]

Here is an interesting tidbit, going DDR4-2666 to DDR4-2800 increased the mining hashrate to 1207H/s on the 1950X. You would think that there would be no impact yet there is a very tiny one. Measured over 1000 shares accepted so it is not just a one off result.

 

[funkywizard]

Dual E5-2680v2 with stak miner: 1250h/s

Power use seems high (no surprise I guess), 220w at first try. This is with 8x8gb ram and all cores -- trying to get an apples to apples comparison so will post that when I've got it.

 

[Bingda733]

Dual E5-2660v2 with xmr-stak-cpu running on Windows Server 2008

1030H/s


Tested on quanta openrack :
2x(2x E5-2660v2+2x8G REGECC)
12 threads on each CPU
Power usage:297W

 

[Marsh]

I like the E5-2660 v2 1030 H/s .
The down side is high 297w power usage.

Could you please verify the power usage?
What other hardware?
Thanks

 

[Bingda733]

I like the E5-2660 v2 1030 H/s .
The down side is high 297w power usage.

Could you please verify the power usage?
What other hardware?
Thanks

Hello
This server comes with 2 node(2 cpu each node) in 1 server,so 297W is actually 4x 2660v2's usage,I think it is not really high compare to E5v1.
no other hardware ,headless and diskless bulid only for compute

 

[Patrick]

Added dual Xeon Gold 6152 results.

Two interesting notes (at least):

1. EPYC is still awesome at Monero mining
2. It is >10% performance gain while also dropping price by over 10% generation on generation.

 

[nthu9280]

seeing around 970 h/s with stak miner on dual e5-2660v2, with 8 of 10 cores enabled and 24 threads (12 per cpu). Power use of 154w according to ipmitool. Haven't measured with a killawatt yet.

That's interesting. I'm getting ~ 875 H/s (that's almost 100 h/s less) and using 225W (+75W) IPMI. S2600Cp2J with 2x E5-2660V2, 7 cores enabled. Using the STH docker image. I can try xmr-stak on CentOS to see if the H/s goes up. I know I have additional DIMMs, LSI 2308 card and a dual 10GB card but they are idling probably adding 20W not sure what else is using the rest of 50W.

 

[Marsh]

@Bingda733

Sorry, Let me understand the hashrate and power usage

You are saying 2 x 1030 H/s = 2060 H/s consumes 297 watt.

 

[funkywizard]

That's interesting. I'm getting ~ 875 H/s (that's almost 100 h/s less) and using 225W (+75W) IPMI. S2600Cp2J with 2x E5-2660V2, 7 cores enabled. Using the STH docker image. I can try xmr-stak on CentOS to see if the H/s goes up. I know I have additional DIMMs, LSI 2308 card and a dual 10GB card but they are idling probably adding 20W not sure what else is using the rest of 50W.

Stak is definitely faster. Your result is only a little low for the wolf docker image. As you've already disabled cores in bios, not sure why you're using so much more power. Usual suspects might be a low-efficiency power supply, or a chassis that has an excessive number of high speed fans, which are set to run at much faster than necessary. Also, all my tests use a single ssd, which should add 1w or less typically. If you have any regular hard drives, that will substantially increase power use.

 

[Klee]

Stak is definitely faster. Your result is only a little low for the wolf docker image. As you've already disabled cores in bios, not sure why you're using so much more power. Usual suspects might be a low-efficiency power supply, or a chassis that has an excessive number of high speed fans, which are set to run at much faster than necessary. Also, all my tests use a single ssd, which should add 1w or less typically. If you have any regular hard drives, that will substantially increase power use.

Yep I prefer Stak.

I have not even tried to tweak for the absolute minimum power usage at all. Still have all cores enabled and am running some WD black 1TB hard drives, I already had them so I used them and they give excellent performance for hard drives.

I plan to ad some more ram to one of the open compute servers and use it to mine and run a full Monero node as soon as the next fork hits in a week and a half or so.

 

[funkywizard]

The wd black is going to add some power no doubt. ballpark 10w each. I've got plenty of those drives spare as well, but don't want to spend the power on them.

 

[nthu9280]

Stak is definitely faster. Your result is only a little low for the wolf docker image. As you've already disabled cores in bios, not sure why you're using so much more power. Usual suspects might be a low-efficiency power supply, or a chassis that has an excessive number of high speed fans, which are set to run at much faster than necessary. Also, all my tests use a single ssd, which should add 1w or less typically. If you have any regular hard drives, that will substantially increase power use.

Got the stak working on CentOS and getting 975+ H/s which is inline and is good.
The power draw is still 220W is little perplexing. Fans are running @ ~1300RPM. Intel 750W Platinum PS. 1 SSD. The only variables I can think of are extra RAM & SAS card.

 

[funkywizard]

Got the stak working on CentOS and getting 975+ H/s which is inline and is good.
The power draw is still 220W is little perplexing. Fans are running @ ~1300RPM. Intel 750W Platinum PS. 1 SSD. The only variables I can think of are extra RAM & SAS card.

1300 rpm is very low. Are you sure this is not 13,000? What size fans / server is this? Shouldn't see 1300rpm on a fan smaller than 120mm, which would require a 3u or bigger server.