A Christmas Gift For Phoronix Readers - Improving Graphs

[phoronix]

Phoronix: A Christmas Gift For Phoronix Readers - Improving Graphs

First of all, Merry Christmas for those of you celebrating today or Happy Holidays regardless...

http://www.phoronix.com/scan.php?pag...h-Improvements

[tchiwam]

Thank you for your contribution.

I'd like to see more of operations vs size dependent graphs.

One example from my mini bench mark:
https://github.com/tchiwam/ptrbench/...s/MaxFLOPS.png

I don't know if this would be feasible... But just an idea.

[bug77]

I think the graphs are pretty fine (red for AMD, blue for Intel, green for Nvidia, with the tested piece properly highlighted).

There are only two suggestions I have:
1. Make it more obvious which graphs are "higher is better" and which are "lower is better". The small font is atrocious on the eye, maybe use a slightly different background for one category.
2. When testing more than one thing using the same test (e.g. min/avg/max fps+fps/W for one game), maybe group all graphs together, add tabs and make them clickable or changeable on mouse hover.

Happy holidays everyone.

[atomsymbol]

Hi. I noticed that on an Android phone in the Google app the charts/graphs have distorted character positions.

For example, "Operation: Sharpen" renders as "Ope ra tion: Sha rpe n".

In the Chrome browser on the same phone the charts look fine.

Below is a screenshot from the https://www.phoronix.com/scan.php?pa...d-athlon-3000g article.

Maybe you can take a quick look at this issue to determine whether the cause is in the SVG files (which might be fixable) or in the Google app (which cannot be fixed by modifying the SVGs).

[atomsymbol]

The small font is atrocious on the eye

The small text in the graphs is harder to read on a 1080p desktop display (or on a 768p notebook display). However, the price of a 4K desktop monitor is (slowly) approaching 200 EUR/USD, and so the problem of unreadable fonts in the graphs will become less severe in the near future as people upgrade to newer displays over time.

In my opinion, if deciding between a 1080p/1440p 144Hz monitor and a 4K 60Hz FreeSync monitor, I would choose the latter (4K) option.

[xxmitsu]

When drawing graphs, Min FPS is an irrelevant statistic (like Max FPS), it's a single frame that could screw the whole statistics. But a 95 percentile, if you could implement that, it would give a much better impression with regards to its performance.

[Michael]

When drawing graphs, Min FPS is an irrelevant statistic (like Max FPS), it's a single frame that could screw the whole statistics. But a 95 percentile, if you could implement that, it would give a much better impression with regards to its performance.

For engines/games exposing per-frame data, that is where you see the box plots displayed as secondary metrics to the FPS graphs... Unfortunately, most games don't expose the per-frame information so can't report percentile data, so just the average (and min / max when exposed) are reported.

[Zyklon]

Better scaling for mobile would be good, currently the graphs are cut off in vertical mode.

(And there are also some other issues on mobile forum usage...)

[boxie]

For engines/games exposing per-frame data, that is where you see the box plots displayed as secondary metrics to the FPS graphs... Unfortunately, most games don't expose the per-frame information so can't report percentile data, so just the average (and min / max when exposed) are reported.

I vaguely remember something about there being a Vulkan layer that could expose frame timing info? was there ever any movement on this?

[boxie]

never mind, I spoke without thinking it through