Skip to content

FAQ.md

Latest commit

 

History

History
47 lines (36 loc) · 3.17 KB

FAQ.md

File metadata and controls

47 lines (36 loc) · 3.17 KB

Frequently asked questions (FAQ)

??? faq "Could not find CPU model" In our plugin, we are trying to match your CPU model name to entries in our TDP data table. Unfortunately things are not as smooth sailing as we would like it to be.

**Main issues**

1. Manufacturers have not given us access to their CPU data. While the manufacturers provide information to more than 4000 models
    on their websites, redistribution and automatic parsing of this info needs their permission. We currently use
    [WikiChip](https://en.wikichip.org/wiki/WikiChip) as our main source, which encompasses approximately 1400 models.
    Our current hope is that the new [MIT processor database](https://processordb.mit.edu) will solve this question in the next few months.
2. Naming of CPU models is inconsistent. Sometimes the true model name ("Model X 1000") is appended with
    extra information ("Model X 1000 @10MHz"), slightly changed ("Manufacturer - Model X 1000"), or hidden ("M1"). While
    we are working on accounting for all these cases, it's not a trivial problem to solve.
3. Cloud providers sometimes use custom processors that are not publicly listed.

<br>
**💡 Solution**

So you encountered a warning like this:
```
[WARN] Could not find CPU model "Model X 1000" in given TDP data table. Using default CPU power draw value (100.0 W).
```
As previously mentioned, all information we need should be out there somewhere — the TDP of your model should be
easy to find with a quick online search. Once you know the TDP and the number of logical cores you can either:

- Create a [small table](./parameters.md#custom-tdp-table) with your CPU model names, as they are presented in the warnings, and
  supply the path to this table via the `customCpuTdpFile = <path>` parameter.
- Set the TDP via `powerdrawCpuDefault = <TDP per core>` and then ignore the warning with `ignoreCpuModel = true`.

For more information see our documentation on [power draw parameters](./parameters.md#hardware-power-draw). You can additionally
report the model with the ["Missing chip" GitHub issue](https://github.com/nextflow-io/nf-co2footprint/issues/new?template=missing_chip.yaml).

??? faq "Differing memory values"

The conversion of `Byte`-type units, such as `memory` is done in accordance with [SI standard, decimal-based naming](https://en.wikipedia.org/wiki/Metric_prefix#List_of_SI_prefixes), rather than [Nextflows binary-based system](https://en.wikipedia.org/wiki/Byte#Multiple-byte_units).
In essence, we use: 1 GB = 1 000 000 000 bytes (1 * 1000^3), while Nextflow uses: 1 GB = 1 073 741 824 bytes (1 * 1024^3), which should be 1 GiB (Gibibyte).

Example: <br>
&emsp;    -> You request `process.memory = 12 GB` <br>
&emsp;    -> Nextflow converts it to 12884901888 bytes and passes the restriction onto the executor <br>
&emsp;&emsp;        => 12.88 GB = 12 GiB are allocated for the process <br>
&emsp;    -> The plugin uses the the former to proceed with calculations and reporting, while Nextflow (incorrectly) reports using the latter <br>