PKT Performance Testing Results

We’ve asked several miners to validate our results. After further testing, we found that miners on Akash produce about 2x the amount of PKT as a similar server on AWS. The AWS (c4.2xlarge) EC2 instance with 8 cores yielded 66 PKT per day, while a 10 CPU container on Akash yields 140 PKT per day.

Based on our data the ideal configuration for PKT miners was 10 CPU, 1 GB RAM, 1 GB storage. Let us know if you find a better configuration.

AWS Performance

AWS Cores Daily Yield (PKT)
8 66
2 12

Akash Performance

CPUs Mem HDD Copies Threads Akash Price
(uakt / block)
PKT / hour PKT / day1 AKT / month PKT / day
PKT / month
Cost of PKT
(USD / PKT)
Profit
Margin
10 1 1 1 10.00 53 6.8 143 21.1 139 4180 $0.0140 53.65%
10 1 1 1 10.00 53 6.7 142 21.1 140 4195 $0.0140 53.81%
10 1 1 1 10.00 53 6.6 140 21.1 139 4158 $0.0141 53.41%
10 1 1 1 10.00 53 6.0 128 21.1 127 3813 $0.0154 49.19%
1 1 1 1 1.00 8 0.7 15 3.2 15 442 $0.0200 33.87%
2 1 1 1 2.00 13 1.1 24 5.2 23 703 $0.0204 32.37%
1 1 1 1 1.00 8 0.6 15 3.2 14 422 $0.0209 30.74%
1 0.5 0.5 10 10.00 62 6.7 - 24.7 108 3229 $0.0212 29.80%
1 0.5 1 10 10.00 63 6.1 - 25.1 98 2927 $0.0238 21.33%
1 1 0.5 10 10.00 72 6.6 - 28.7 104 3130 $0.0254 15.92%

As you can see in the updated table above, we tested several different CPU configurations to identify the ideal configuration on Akash. We did performance testing with 10 CPU, 2, 1 CPU and tried lower settings for memory and storage.

It’s important to understand that with Akash you can run 10 miners with 1 CPU as easily as running 1 miner with 10 CPU. In our initial testing we found a better price-performance ratio by running 10 miners each with 1 CPU, but after publishing the blog post we found that 10 CPUs on 1 miner gave a better price for the same performance as 10 miners with 1 CPU.

To change the resources section of the SDL file:

      resources:
        cpu:
          units: 10

How is it possible that we can beat AWS?

Running software on “bare metal” provides better performance than virtual machine. The Akash providers we used run Kubernetes containers on bare metal. Not every provider on Akash has the same performance.

Because we use Containers not VMs and not virtualization, the miner shares the same system resources as the host OS, but it is managed by Kubernetes. As long as you request whole numbers of CPU cores (e.g. 1, 2, 4, 8, 10) you can have a “dedicated” CPU assigned to your container.

Most modern scalable web apps run microservices in containers now, and there has been a move away from VMs and AWS style Cloud. Large enterprise companies are now building “private clouds” on bare metal to get better performance.

You can learn more about how CPU is managed in kubernetes here: Feature Highlight: CPU Manager | Kubernetes

Don’t agree with our data?

Just try it! Run a PKT miner on Akash, and share your results with the community here.

6 Likes

That’s not accurate at all as per the latest difficulty. I ran three servers from Akash with 1core, 1GB RAM, and 1GB storage. The total earning was just 1.75PKT per hour from three servers (combined) — 42 PKT per day, 14 PKT/server. The total cost per server is $15 per month that is not economical at all. When did you test it? It seems a year old result.

1 Like

Update: I ran 10 cores, 1GB RAM, 1GB storage. Each hour it produces 6 PKT, meaning 4230 PKT every month. You will earn 126 dollars if you sell at a rate of 0.03$ with an investment of 23AKT (2.7*23=$62/month). In other words, you’ll earn 100% from your investment.

Here is the script I used:

version: "2.0"
services:
  miner:
    image: chandrastation/pkt_on_akash:v1.0.0
    env:
      - WALLET_ADDR=pkt1qs78fejrr68hs7c73tfquzj4nyup3vck5w60fmx
      - POOL1=http://pool.srizbi.com
      - POOL2=http://pool.pktpool.io
      - POOL3=http://pool.pkt.world
      - POOL4=http://pool.pkteer.com
    expose:
      - port: 3000
        as: 80
        to:
          - global: true

profiles:
  compute:
    miner:
      resources:
        cpu:
          units: 10
        memory:
          size: 1Gi
        storage:
          size: 1Gi
  placement:
    miner:
      attributes:
        host: akash
      signedBy:
        anyOf:
          - "akash1365yvmc4s7awdyj3n2sav7xfx76adc6dnmlx63"
      pricing:
        miner:
          denom: uakt
          amount: 1000

deployment:
  miner:
    miner:
      profile: miner
      count: 1
3 Likes

I was able to mine at $0.03/AKT with 5 vCPU. But then, something happened overnight yesterday and it stopped bringing in rewards, could be because of some pool going down or something. Anyways, here’s the script, it’s just like the one above, but with half the vCPU. Add your address after WALLET_ADDR= if you want to mine for yourself using this config.

version: "2.0"
services:
  miner:
    image: chandrastation/pkt_on_akash:v1.0.0
    env:
      - WALLET_ADDR=
      - POOL1=http://pool.srizbi.com
      - POOL2=http://pool.pktpool.io
      - POOL3=http://pool.pkt.world
      - POOL4=http://pool.pkteer.com
    expose:
      - port: 3000
        as: 80
        to:
          - global: true

profiles:
  compute:
    miner:
      resources:
        cpu:
          units: 5
        memory:
          size: 1Gi
        storage:
          size: 1Gi
  placement:
    miner:
      attributes:
        host: akash
      signedBy:
        anyOf:
          - "akash1365yvmc4s7awdyj3n2sav7xfx76adc6dnmlx63"
      pricing:
        miner:
          denom: uakt
          amount: 1000

deployment:
  miner:
    miner:
      profile: miner
      count: 1

Here’s an updated breakdown with the profitability based on AKT at $2.75 and PKT at $0.03.

CPUs Mem HDD Copies Threads Akash Price
(uakt / block)
PKT / hour PKT / day 1 AKT / month PKT / day 2
PKT / month
Cost of PKT
(USD / PKT)
Profit
Margin
10 1 1 1 10.00 53 6.8 143 21.1 139 4180 $0.0140 53.65%
10 1 1 1 10.00 53 6.7 142 21.1 140 4195 $0.0140 53.81%
10 1 1 1 10.00 53 6.6 140 21.1 139 4158 $0.0141 53.41%
10 1 1 1 10.00 53 6.0 128 21.1 127 3813 $0.0154 49.19%
1 1 1 1 1.00 8 0.7 15 3.2 15 442 $0.0200 33.87%
2 1 1 1 2.00 13 1.1 24 5.2 23 703 $0.0204 32.37%
1 1 1 1 1.00 8 0.6 15 3.2 14 422 $0.0209 30.74%
1 0.5 0.5 10 10.00 62 6.7 - 24.7 108 3229 $0.0212 29.80%
1 0.5 1 10 10.00 63 6.1 - 25.1 98 2927 $0.0238 21.33%
1 1 0.5 10 10.00 72 6.6 - 28.7 104 3130 $0.0254 15.92%

Happy to provide the wallet addresses and the spreadsheet if you want to double check my work!

Can confirm similar numbers. Had some fun and ran 10 x 10cpu 1+1 Mem and HDD and it threw off 1446pkt in 24 hrs. Need a way to scale on the deployment tool and then auto sell pkt.

1 Like

Sell?! I thought everyone is a HODLer! Big thank you for confirming our results. I am going to keep mine running until… forever. How does this compare to other cloud hosts?

Only to buy more AKT!!!

1 Like

I stopped my miners so I could get an exact amount of AKT spent and PKT mined. Here are my results after about 64 hours:

Parameter Value
Spot (USD/AKT) $2.85
vCPUs 10
Cost (AKT) 2
Cost (USD) $5.77
Cost (USD/PKT) $0.01533
Spot (USD/PKT) $0.03000
Income (PKT) 377
Income (USD) $11.30
Profit (USD) $5.52
Margin 49%
Profit 196%
2 Likes

Hi, from 10 vcpu deployed - pkt miner - i generate not more than 5 pkt per instance. Can you confirm that other instances generate more than 6 pkt hourly ?

Hi @yeszie sorry I didn’t understand your question. What’s your akash address?

sorry, i edited - so i think pkt miner deployed with 10vcpu cant mine more than 6 pkt hourly

Yes that’s what I’m getting too. 6 PKT / hour is common, but I got 6.8 / hour on some of mine.

Thank you, from few hours i run 18 instances (10 vcpu) - and it is mining total 80 PKT / hourly (4.5 PKT / instance per hour). Maybe tomorrow will be better.

I’m getting around 4.8/hr running 1 instance with the 10 CPU 1 Mem 1 HDD configuration, too.
It’ll be interesting to see if there’s any variation on that as the day progresses.

This is from PKT telegram:
We are going to make a change to our ann handler addresses. This might cause certain miners breaking in case they are using an old version of the mining code. If your miners are using the latest version there’s no action required.

To fix the old miners you can upgrade the code by recompiling it. If you only want a temporary fix you can restart the miners and they will start working again.

Yea that’s what I suspect. I am also getting 4.8pkt/h for a 10c server

Maybe we need new deployment image. Tomorrow will see if miners with 6.8 PKT/hour will not be below 5.

1 Like

Update:
I’m still getting a pretty consistent 4.8pkt/h on average across the miners I deployed.
It ain’t bad, but I’m curious: is anyone still able to reach the ~6pkt/h mentioned in posts above?
If so, it’d be really cool to hear how you optimized it and what your pool configurations are!

I am also curious to know about others’ results. I am currently only getting about 3.7 PKT / Hr using 10 vCPUs. Anyone having better luck?