Reading your results¶
A finished run answers one question in five dimensions: what changes if you move from the estate you have to the hardware you are considering.
The five measures¶
Energy reduction¶
Annual kWh saved, and the same figure as a percentage. This is the primary result — everything else follows from it.
Carbon reduction¶
Scope 2 emissions avoided, in CO₂e, derived from the energy reduction and the carbon intensity of your grid.
Server consolidation¶
How many servers the target hardware needs to do the same work, and the floor space that releases. Modern hardware is usually denser, so the count falls.
Energy rating improvement¶
Efficiency grades for your current estate and for each target. This is the fastest way to see whether a given piece of hardware is worth considering at all.
Cost savings¶
Annual operational cost reduction, built from the energy reduction and whatever cost parameters were in force for the run.
Servers left out of a run¶
A run analyses the lines of your specification that are matched to our catalogue. Lines that are still unmatched when you submit are left out, and every figure (servers, cores, memory, energy, the estate's rating) covers the matched servers only. When that happens, the results page and the report say how many specifications and servers were left out. To include them, match the lines and run the simulation again.
What the colours mean¶
Every figure that shows a change is coloured by whether the change is good for you: green when it is better, red when it is worse, and uncoloured when nothing changed. Better depends on the figure. Less energy, less carbon, fewer servers, fewer cores, a lower running cost and a lower power draw are better; more performance is better; a change in memory is neither, so it is not coloured.
That means a negative reduction is red. A carbon reduction of −7% is a carbon increase, and shows red. The sign and the wording ("Increase", "Cost Increase") always stay beside the colour, so a report printed in black and white still reads correctly. The results page and the PDF report always colour a figure the same way.
Power and energy¶
Each target also shows its power and energy against your current estate:
- Annual energy, in kWh a year, including cooling and power distribution.
- Power draw, in watts at peak and at idle, for all the servers together: every server in your estate today, and as many of the target as the simulation says you would need. This is the servers alone.
Power draw needs a wattage for every validated server and for the target. If any is missing, that figure shows as not available rather than as a partial total. GPU simulations do not show power draw. These figures are worked out when a run finishes, so a run completed before they were added does not show them; run the simulation again to see them.
Efficiency grades¶
Every configuration in our catalogue is graded from A+ down to F, based on its ITEE score — a measure of how much useful compute a server returns for the electricity it draws. Higher is better.
| Grade | ITEE score | What it means |
|---|---|---|
| A+ | 14,000 and above | Substantially above average. Current, high-performance hardware |
| A | 12,000 – 14,000 | Well above average. Recent generation |
| B | 9,000 – 12,000 | Above average. Reasonably current |
| C | 7,200 – 9,000 | Moderate. Likely a few generations behind |
| D | 6,000 – 7,200 | Below average. Ageing, worth reviewing |
| E | 4,700 – 6,000 | Poor. A strong candidate for replacement |
| F | Below 4,700 | Very low. Replace these first |
| U | — | Ungraded. We could not determine a rating |
Grades appear as coloured badges throughout: on specification lines, on targets, on server detail pages, and in the dashboard charts. They are the quickest way to see the shape of an estate and where the opportunity sits.
Where the grades come from¶
Grades are not our opinion. Each configuration is rated against industry benchmark data — peak power draw, idle power draw, and benchmark throughput — and the ITEE score falls out of that. Ratings are fetched when a configuration first appears, and the catalogue is refreshed daily so new hardware and revised ratings arrive without anyone asking.
What a U grade means¶
U means we could not rate that configuration. Usually one of:
- The hardware is new enough that it is not yet in the ratings data.
- The vendor, model or CPU does not match anything we know — most often a typo or a non-standard name in the source inventory.
- The rating service was briefly unavailable when we looked.
The fix is nearly always the second one. Check the details on the line, correct anything wrong, and it will be rated on the next attempt. See Specifications.
Ungraded lines still appear in your specification, but contribute nothing to the energy calculation — so a simulation with many U grades understates both the current estate and the saving.
How far to trust it¶
The analysis is only as good as three things:
- How well your estate was matched. Lines flagged and left unconfirmed, and lines left at U, are the main source of error. See Specifications.
- Whether you supplied your own energy price. The default is a platform assumption and will not be your tariff.
- How complete your import was. Optional fields you did not map still sharpen the result.
The report states which figures were assumed rather than supplied. Read that section before circulating the result.
Comparing targets¶
Where a simulation has several targets, the results place them side by side. The best choice is rarely the one with the largest headline energy saving — consolidation and licensing often move the cost result further than energy does.
When part of a run fails¶
If some targets fail and others succeed, the run still completes. The failed targets carry their error message, and their credits are refunded automatically. You are never charged for analysis that produced nothing.
Exporting¶
Every result can be exported as a PDF report. See Reports.