Major Update

CoreBench 1.6.2 emphasizes the effect that 1.6 was supposed to achieve, accommodated by several dramatic scoring system changes and algorithm rewrites.

Change Summary

• All score calculations in the full CPU test (excluding the GFLOPs and Full Load test) have been fully rewritten to be balanced around modern midrange processors.
• The multithread test runs many times in order to achieve a stable result of high enough quality to submit.
• The multicore test runs many times in order to achieve a stable result of high enough quality to submit.
• Other improvements to better user experience.

• Bypassed Python GIL using numba nogil for the multithread test
• Multithread score rebalanced (very unstable in 1.6.0)

• Mass rebalancing of scores in the multi-core and multi-thread tests.
• Multi-thread test uses Python's threading instead of multiprocessing again.
• The full load test has significantly less Python overhead now.
• Fixed small visual bugs.
• You are now asked to re-enter your API key if it is invalid in an attempt to resubmit your score.
• Multi-core and multi-thread tests now work with process queues and workers instead of forced batches of individual tests, giving a better representation of true CPU performance.

Fixed a crash when running test commands.

Fixed issues when using normal test command syntax would break.

Fixed a bug when not all cores would be fully utilized in the full load test. Added progress indicator.

Fixed a bug when calling "colours.bold()" during the CPU full load test would cause the program to crash due to it not existing.

Added new metric recorded to the database: "Full Load"

This tests the entire capability of your CPU (by utilizing all logical cores) to determine its performance in a situation when the CPU is able to run at full capacity.

This metric is mostly useful for workstation owners.

Signature fixes

Updated version, visual fixes, small improvements.
(Hotfix 2)