ECC RAM overclocking

[magikRUKKOLA]

Recently I upgraded my RAM to 512GB (8 x 64GB): 8 x Samsung 64GB DDR4 PC4-21300 2666MHz LRDIMM Quad Ranked Registered ECC Memory.
It turned out that its worked with the old profile from the G.Skill RAM that I had, which was configured for DDR4-3200 with 1.35V and tight timings.

So I installed ktransformers and everything and can confirm that DDR4-3200 profile adds about 15% of performance. But ... there is some issue:

edac-util -v

mc0: 0 Uncorrected Errors with no DIMM info
mc0: 0 Corrected Errors with no DIMM info
mc0: csrow0: 0 Uncorrected Errors
mc0: csrow0: mc#0csrow#0channel#0: 13 Corrected Errors
mc0: csrow0: mc#0csrow#0channel#1: 20 Corrected Errors
mc0: csrow0: mc#0csrow#0channel#2: 1 Corrected Errors
mc0: csrow0: mc#0csrow#0channel#3: 17 Corrected Errors
mc0: csrow0: mc#0csrow#0channel#4: 5 Corrected Errors
mc0: csrow0: mc#0csrow#0channel#5: 14 Corrected Errors
mc0: csrow1: 0 Uncorrected Errors
mc0: csrow1: mc#0csrow#1channel#0: 12 Corrected Errors
mc0: csrow1: mc#0csrow#1channel#1: 109 Corrected Errors
mc0: csrow1: mc#0csrow#1channel#2: 3 Corrected Errors
mc0: csrow1: mc#0csrow#1channel#3: 16 Corrected Errors
mc0: csrow1: mc#0csrow#1channel#4: 0 Corrected Errors
mc0: csrow1: mc#0csrow#1channel#5: 30 Corrected Errors

(its after about 24 hours of running)

Basically its having some 'corrected errors' meaning that RAM is not very stable. But its ECC. So it''s self-correcting. So, what to do? Leave it like this? Hm.

The RAM itself is without any radiators. Should I install them? Will it improve the situation?

Can I kill my RAM like this? Or speed-up its deterioration?

[magikRUKKOLA]

DeepSeek-V3-0324 is pretty amazing lol:

  The temperatures for your overclocked RAM (DDR4) are concerningly high,
  especially for modules B, C, F, and G, which are hitting 79°C–80°C.

  ### Is 80°C Too High for Overclocked RAM?

  • DDR4 RAM typically operates safely up to ~85°C (some ICs can handle up to
  95°C before throttling).
  • For overclocked RAM, >70°C is not ideal for stability and longevity.
  • 80°C is borderline risky—if sustained, it may cause instability (crashes,
  errors) or long-term degradation.

  ### Recommended Actions:

  1. Improve Cooling
    • Add case fans for better airflow.
    • Consider dedicated RAM cooling (heatsinks, active cooling, or direct
    airflow).
    • Ensure your case isn’t choked (hot air needs to escape).
  2. Check Voltages
    • High RAM temps often come from excessive DRAM voltage (e.g., >1.5V for
    DDR4).
    • If overclocked, try lowering voltage slightly while testing stability.
  3. Monitor for Errors
    • Run MemTest86 or HCI MemTest to check for RAM errors.
    • If errors occur, reduce clock speeds or loosen timings.
  4. Ambient Temp Impact
    • If room temp is high (>30°C), expect higher RAM temps.


  ### Conclusion:

  While 80°C won’t immediately kill your RAM, it’s too hot for optimal
  performance. Aim for <70°C under load for overclocked RAM. If temps stay
  this high, expect potential instability in heavy workloads.

  Would you like help tuning your RAM settings or airflow setup?

[magikRUKKOLA]

Its a really great RAM but needs some heatsinks.

a note. a very nice tool to monitor RAM temperature.

ipmitool sdr type temperature

it turned out that the RAM was overheating. good to know

[magikRUKKOLA]

Well yeah, its was totally related to the temperature of the RAM. So I have put a channel-type of fan, some extreme timings, V1.35 (just in case). It works very well, under 70C.

Please advise how to build the fastest setup possible with DDR5 and under 20k EUR.