mirror of
https://github.com/Brandon-Rozek/website.git
synced 2024-11-09 18:50:34 -05:00
Compare commits
No commits in common. "75f46da99b8c05f21f380dbdec5529bdc8aa4ad0" and "01a28e2a4a54b66ecba994742434ea29737dfaa4" have entirely different histories.
75f46da99b
...
01a28e2a4a
3 changed files with 0 additions and 83 deletions
|
@ -1,45 +0,0 @@
|
|||
---
|
||||
title: "Disabling CPUs to Save Power"
|
||||
date: 2024-04-06T20:48:52-04:00
|
||||
draft: false
|
||||
tags: []
|
||||
math: false
|
||||
medium_enabled: false
|
||||
---
|
||||
|
||||
Looking for ways to reduce the power usage of my home server? This post shows a not-so-scientific look at how disabling CPUs can help save some power.
|
||||
|
||||
I run a Dell PowerEdge R430 with an Intel Xeon E5-2643 v3 CPU. This gives me 12 physical cores, and with hyperthreading this presents itself as 24 logical cores.
|
||||
|
||||
Given that I mostly use my server as media storage, most of those CPUs sit idle most of the time. My thought is, how much power can I save if I disable some of these unused cores?
|
||||
|
||||
For idle workloads disabling the CPU does not result in any noticeable power savings. The power savings is significant, however, if you analyze the system under load.
|
||||
|
||||
To conduct this experiment, I used a [Kill a Watt](https://en.wikipedia.org/wiki/Kill_A_Watt) measuring device which monitors the power usage of whatever is plugged into it.
|
||||
|
||||
To disable a CPU in Linux, use the following command:
|
||||
|
||||
```bash
|
||||
# Disable CPU 23
|
||||
echo 0 | sudo tee /sys/devices/system/cpu/cpu23/online > /dev/null
|
||||
```
|
||||
|
||||
Repeat for any CPUs you want to disable. You can see which CPUs are available using `ls /sys/devices/system/cpu`. The `htop` tool will display in addition to the CPU utilization, which CPUs are offline.
|
||||
|
||||
To get the system under load, I used the `stress` tool.
|
||||
|
||||
```bash
|
||||
# Spin off 24 dummy tasks that max out each CPU
|
||||
stress -c 24
|
||||
```
|
||||
|
||||
I ran the stress tool with the corresponding number of logical CPUs I had online to come up with the following table:
|
||||
|
||||
| # Online Logical CPU | Power (Watts) |
|
||||
| -------------------- | ------------- |
|
||||
| 24 | 360 |
|
||||
| 12 | 295 |
|
||||
| 6 | 194 |
|
||||
| 4 | 173 |
|
||||
|
||||
Cutting down to 4 logical CPUs can cut the power usage under load in half! Do note though, that this exchanges performance for power savings. If you are running a low amount of services on your home server or can wait a bit of extra time for a computation to finish, consider disabling some of your CPUs.
|
|
@ -30,7 +30,6 @@ The following menu contains meals that are in rotation at my home. These are mad
|
|||
|
||||
## Soups
|
||||
|
||||
- [Beef Taco](beef-taco-soup/)
|
||||
- [Pork Pozole](pork-pozole/)
|
||||
- Chicken & Gnocchi
|
||||
- {{< vegetarian >}}Broccoli Cheddar
|
||||
|
|
|
@ -1,37 +0,0 @@
|
|||
---
|
||||
title: "Beef Taco Soup"
|
||||
date: 2024-04-06
|
||||
hideDate: true
|
||||
draft: false
|
||||
---
|
||||
|
||||
## Ingredients
|
||||
|
||||
- Beef
|
||||
- 1/2 yellow onion
|
||||
- 1/2 green bell pepper
|
||||
- 1/2 red bell pepper
|
||||
- 24 oz beef broth
|
||||
- 1 can diced tomatoes
|
||||
- 2 ounces cream cheese
|
||||
- 1/2 cup of heavy whipping cream
|
||||
- 1 cup cheddar cheese
|
||||
- 1 tablespoon chili powder
|
||||
- 1 tablespoon cumin
|
||||
- 2 teaspoons onion powder
|
||||
- 3 teaspoons oregano
|
||||
- 1 teaspoon garlic powder
|
||||
- 1 teapoon paprika
|
||||
- 1/2 teaspoon cayenne pepper
|
||||
|
||||
## Recipe
|
||||
1. Cook beef
|
||||
2. Blend diced tomatoes and beef broth together.
|
||||
3. Put tomato and broth blend in pot and bring to a boil,
|
||||
then reduce heat to a simmer.
|
||||
4. Meanwhile start cooking the onion and bell peppers.
|
||||
5. Add beef, onion, bell peppers, and spices to the pot.
|
||||
6. Cook for 30 minutes
|
||||
7. Add cream cheese, heavy whipping cream, and cheddar cheese
|
||||
to the pot.
|
||||
8. Wait until everything is all melted and smooth then serve.
|
Loading…
Reference in a new issue