CPU reporting - Linode/Longview vs. New Relic

I've noticed that Linode & Longview consistently report CPU usage at 10x higher than New Relic. If New Relic says 20%, Linode says 200%. If New Relic says 3%, Linode says 30%. Is this normal? Which is "right"?

top and htop both usually give numbers in line with New Relic (smaller order of magnitude) - although this may not be accurate since top / htop are looking moment-to-moment and Longview / New Relic are presenting averages over time.

Thanks for any insight.