What is the best way to test ram stability

[MingLee420]

what is the best way to test overclocked ram stability apart from playing my favourite triple a games

[da na]

Memtest86

Prime95 memory test

Windows Memory Diagnostic

[BetteBalterZen]

Mel0nMan pretty much provided the best answer. I just wanna recommend MemTest86 aswell

[freeagent]

These days I have been using TM5, the Anta Extreme profile. I also like Linpack Xtreme, I use v1.15.

[MingLee420]

15 hours ago, Mel0nMan said:

Memtest86

Prime95 memory test

Windows Memory Diagnostic

 

15 hours ago, BetteBalterZen said:

Mel0nMan pretty much provided the best answer. I just wanna recommend MemTest86 aswell

 

14 hours ago, freeagent said:

These days I have been using TM5, the Anta Extreme profile. I also like Linpack Xtreme, I use v1.15.

i did memtest86 over night and passed, 0 errors 

i got 3600mhz c16-21-21-38 but i might tighten it more but if i try 3733mhz it just dosen't boot stock was c18-22-22-39

[DarkSmith2]

memtest86 is not to test stability, its to test defects.

[MingLee420]

13 minutes ago, DarkSmith2 said:

memtest86 is not to test stability, its to test defects.

should i do prime95 and or hci memtest?

[BetteBalterZen]

nvm 

Mod please delete

[BetteBalterZen]

2 hours ago, DarkSmith2 said:

memtest86 is not to test stability, its to test defects.

How is it different? 

RAM OC'ed too hard causing errors would be the same as defect RAM, wouldn't it? In terms of errors produced

[YoungBlade]

3 hours ago, BetteBalterZen said:

How is it different? 

RAM OC'ed too hard causing errors would be the same as defect RAM, wouldn't it? In terms of errors produced

I believe that's correct. Programs like memtest86 are just checking to make sure that the RAM can accurately store and retrieve data. I don't believe there is any meaningful difference between a mistake due to a hardware failure or RAM being pushed too far.

 

I'm interested to see if I'm wrong, though.

[DarkSmith2]

5 hours ago, BetteBalterZen said:

RAM OC'ed too hard causing errors would be the same as defect RAM, wouldn't it? In terms of errors produced

Memtest86 is just unreliable to find errors/instability caused by overclocking. But its very good and fast to find any broken cells at stock settings.

My guess is that the Algorithms being used are just ineffective due to its age and relatively low load for todays Hard- and Software.

 

Or in other words, you can pass memtest86+ while HCI Memtest, Karhu, Testmem5 or Google Stress App will find errors.

So my answer is no, you wont necessarily find the same amount of errors unless a stick is broken.

[BetteBalterZen]

2 minutes ago, DarkSmith2 said:

Memtest86 is just unreliable to find errors/instability caused by overclocking. But its very good and fast to find any broken cells at stock settings.

My guess is that the Algorithms being used are just ineffective due to its age and relatively low load for todays Hard- and Software.

 

Or in other words, you can pass memtest86+ while HCI Memtest, Karhu, Testmem5 or Google Stress App will find errors.

So my answer is no, you wont necessarily find the same amount of errors unless a stick is broken.

Oh alright, gotcha. 

What RAM test would you use out of all the ones you mentioned? 

[DarkSmith2]

Just now, BetteBalterZen said:

Oh alright, gotcha. 

What RAM test would you use out of all the ones you mentioned? 

im using Karhu because its easy to use and "fast" only 2.5hours for 16GB.. but its not free. Some people would advice to use multiple stresstests, some even use Linpack in addition to any other stresstests to look after variations in performance for a glimpse of instability.

But tbh use whichever you like of those options. Just not Memtest86
 

[BetteBalterZen]

7 minutes ago, DarkSmith2 said:

im using Karhu because its easy to use and "fast" only 2.5hours for 16GB.. but its not free. Some people would advice to use multiple stresstests, some even use Linpack in addition to any other stresstests to look after variations in performance for a glimpse of instability.

But tbh use whichever you like of those options. Just not Memtest86
 

Cool, thank you

[MingLee420]

3 hours ago, DarkSmith2 said:

Memtest86 is just unreliable to find errors/instability caused by overclocking. But its very good and fast to find any broken cells at stock settings.

My guess is that the Algorithms being used are just ineffective due to its age and relatively low load for todays Hard- and Software.

 

Or in other words, you can pass memtest86+ while HCI Memtest, Karhu, Testmem5 or Google Stress App will find errors.

So my answer is no, you wont necessarily find the same amount of errors unless a stick is broken.

ok ill try hci memtest

[StDragon]

5 hours ago, YoungBlade said:

I believe that's correct. Programs like memtest86 are just checking to make sure that the RAM can accurately store and retrieve data. I don't believe there is any meaningful difference between a mistake due to a hardware failure or RAM being pushed too far.

 

I'm interested to see if I'm wrong, though.

You are correct. A bit-flip can be caused by both faulty silicon or improper timings and voltages.

 

3 hours ago, DarkSmith2 said:

Memtest86 is just unreliable to find errors/instability caused by overclocking. But its very good and fast to find any broken cells at stock settings.

My guess is that the Algorithms being used are just ineffective due to its age and relatively low load for todays Hard- and Software.

 

Or in other words, you can pass memtest86+ while HCI Memtest, Karhu, Testmem5 or Google Stress App will find errors.

So my answer is no, you wont necessarily find the same amount of errors unless a stick is broken.

Memtest86, and others like it, are better suited for finding specific errors in RAM where there's a specific hardware region at fault. This will manifest itself with a specific address showing up in error over and over.

 

However incorrect timings or voltages will cause errors to occur in random address space (more often than not). The problem here is that for every random error caught in this way, there's a whole lot more errors occurring that can't be identified in real-time as the test isn't checking that specific address at that exact specific time. The algorithm here isn't at fault because you can't test the entire stick of RAM at the exact same time that each and every error occurs from improper timings and voltages. Or to put it another way, testing for these types of errors and finding them is like showing the proverbial 'tip of the iceberg'; for every error found, there's a whole lot more occurring that's not being reported.

[MingLee420]

15 hours ago, DarkSmith2 said:

im using Karhu because its easy to use and "fast" only 2.5hours for 16GB.. but its not free. Some people would advice to use multiple stresstests, some even use Linpack in addition to any other stresstests to look after variations in performance for a glimpse of instability.

But tbh use whichever you like of those options. Just not Memtest86
 

 I have 16 processes at 2048mb overnight some of them are like above 500 percent and some are under 400 and 200 with a few at a 1000 percent do you think tjata good enough