Known Issues & Resources
Share
Explore
Known Issues & Resources

AWS Hash Rate Testing

This article contains hash rate testing on AWS machine tiers for an Iron Fish node. It was written and submitted to Iron Fish by a community member named Ryan through email.
Hey, our AWS team has been doing some node performance testing recently and wanted to share our results with you, for whatever they may be worth.
Search
Type
CPU Cores
CPU GHZ
CPU Type
RAM GB
Spot Rate
Hash rate
1
c6gd.metal
64
3.3
ARM
128
$0.64
3900000
2
c6i.metal
128
3.5
Xeon
256
$1.28
3100000
3
c5a.24xl
96
3.3
EPYC
192
$0.91
3100000
4
z1d.12xl
48
4
Xeon
384
$1.34
1600000
5
z1d.metal
48
4
Xeon
384
$1.34
2700000
6
t3-2xl
8
3.1
Xeon
32
$0.10
680000
7
t3a-2xl
8
2.5
EPYC
32
$0.09
680000
There are no rows in this table
⁠
This was the mean hashrate over the course of a couple hours when running 1 node and 1 miner process with 1 thread per core.
All ran at the same time except the c6gd -- we had some issues getting it installed from source and it was tested the next day (the rest ran docker). We ran the c6i.metal at the same time though to confirm network conditions weren't significantly different. (it performed the same)
We tried altering thread counts over CPU count and under CPU count. In general we found 1=1 worked best. EXCEPT we found that running more ARM threads than CPUs provided higher hashrate. We went up to about 4,900,000 with 128 threads. However, going up to 192 threads brought us back to 3,500,000 or so. That gave us an idea though and we're currently running 1 node process and 3 miner processes set to 64 threads each and getting throughput of about 2,700,000 on each process for a total of 8,100,000 H/s.
tldr; ARM outperforms EPYC outperforms Xeon. metal outperforms virtual.
⁠
image.png
⁠
⁠
Gallery
Share
 
Want to print your doc?
This is not the way.
Try clicking the ⋯ next to your doc name or using a keyboard shortcut (
CtrlP
) instead.