F5F Stay Refreshed Power Users Overclocking Do you require running 2048K+ in Prime95 to stress test your XMP profile?

Do you require running 2048K+ in Prime95 to stress test your XMP profile?

Do you require running 2048K+ in Prime95 to stress test your XMP profile?

C
Coollegoguy380
Member
Find
51
11-28-2020, 10:12 PM
#1
I've assembled a PC for a friend and am currently conducting tests to verify proper functionality.
Typically, after overclocking RAM, you should run Memtest86 followed by Prime95 in custom mode with a range of 512K-4096K. The 512K-1024K range primarily tests the IMC and I/O-lanes, while 2048K+ targets the actual DIMMs.
Now, if the goal is to check stability and performance of XMP/DOCP settings, do you really need to stress-test the DIMMs with 2048K+ FFTs? Would it suffice to only test the 512K-1024K range?
They are already binned and should perform well if the rest of the system meets the specifications.
To be honest, he's setting up the system tonight, so I'm running low on time and can't let Prime95 run for more than a few hours without asking him to visit tomorrow instead.
Reply
Reply
C
Coollegoguy380
11-28-2020, 10:12 PM #1

I've assembled a PC for a friend and am currently conducting tests to verify proper functionality.
Typically, after overclocking RAM, you should run Memtest86 followed by Prime95 in custom mode with a range of 512K-4096K. The 512K-1024K range primarily tests the IMC and I/O-lanes, while 2048K+ targets the actual DIMMs.
Now, if the goal is to check stability and performance of XMP/DOCP settings, do you really need to stress-test the DIMMs with 2048K+ FFTs? Would it suffice to only test the 512K-1024K range?
They are already binned and should perform well if the rest of the system meets the specifications.
To be honest, he's setting up the system tonight, so I'm running low on time and can't let Prime95 run for more than a few hours without asking him to visit tomorrow instead.

Reply
Reply
M
monkeyboo2004
Member
Find
117
12-11-2020, 08:16 PM
#2
Prime95/ blended (default) puts extra strain on RAM beyond what’s needed at the standard settings.
A short period of 10 to 15 minutes should be sufficient, according to my view...
If you’re concerned about unstable RAM performance, you can perform a full nightly Memtest86 test...
Reply
Reply
M
monkeyboo2004
12-11-2020, 08:16 PM #2

Prime95/ blended (default) puts extra strain on RAM beyond what’s needed at the standard settings.
A short period of 10 to 15 minutes should be sufficient, according to my view...
If you’re concerned about unstable RAM performance, you can perform a full nightly Memtest86 test...

Reply
Reply
I
Its_Brian2002
Member
Find
74
12-17-2020, 10:03 PM
#3
Prime95 isn't the best choice for memory testing. It has some unusual access patterns that most utilities don’t use, and it takes a long time to spot errors. Memtest86 is meant to check for defective memory, so it can also be used for OCs, but it only covers specific bad memory cases and misses many normal access patterns. This is why people often pair Prime95 with Memtest86.

I discovered something useful when testing my RAM: using OC mode with TM5 0.12 and the 1usmus preset on this page produced a lot of errors in about half an hour, while Memtest86 ran smoothly for hours. I don’t have enough data to confirm it catches everything, but it does find many issues quickly. Based on my experience, this test alone has often been sufficient to determine system stability.

I can’t promise it works for everyone, but since it’s just an XMP profile, I think it’s probably effective.
Reply
Reply
I
Its_Brian2002
12-17-2020, 10:03 PM #3

Prime95 isn't the best choice for memory testing. It has some unusual access patterns that most utilities don’t use, and it takes a long time to spot errors. Memtest86 is meant to check for defective memory, so it can also be used for OCs, but it only covers specific bad memory cases and misses many normal access patterns. This is why people often pair Prime95 with Memtest86.

I discovered something useful when testing my RAM: using OC mode with TM5 0.12 and the 1usmus preset on this page produced a lot of errors in about half an hour, while Memtest86 ran smoothly for hours. I don’t have enough data to confirm it catches everything, but it does find many issues quickly. Based on my experience, this test alone has often been sufficient to determine system stability.

I can’t promise it works for everyone, but since it’s just an XMP profile, I think it’s probably effective.

Reply
Reply