2
\$\begingroup\$

I have watched many benchmarks on RAM memories and I was shocked to find out that sometimes lowering the memory timings can result in lower performance in BOTH synthetic benchmarks general use of programs.

In my opinion if everything works correct in the RAM modules, lower timings should always result in better performance. The only case that I can think of where the performance would be impacted negatively is if the memory would actually have transmission errors and the data would have to be re-transmitted. I think this is why ECC memory was invented.

The last video I watched which is a very detailed analisys of the results after changing the memory timings is this one. https://www.youtube.com/watch?v=zrsuP48qsJE

The results seem mathematically impossible to me. Are there errors that cause this loss in performance or is it something else? How can I start to find out more about this subject?

\$\endgroup\$
  • 1
    \$\begingroup\$ No, ECC memory has nothing to do with improving performance, but detects memory errors in stored data (e.g. from background radiation). \$\endgroup\$ – Brian Drummond Aug 15 '17 at 21:14
  • 5
    \$\begingroup\$ Also, sometimes you find as you lower the clock period, you need to increase a delay to meet a memory's timing requirements (published in its datasheet) - e.g. if Trc must be >= 40ns that's a CAS delay of 1 clock cycle at a 40ns (25MHz) clock, but 2 cycles ( 60ns) at a 30ns (33MHz) clock, so the faster clock reduces that aspect of performance. So study the memory's datasheet to find its limitations. \$\endgroup\$ – Brian Drummond Aug 15 '17 at 21:43

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.