Question Post your MaxxMem² score

[igor_kavinski]

Thanks to Det0x for using this utility in his benchmarking and thanks to AdamK47 for posting his score, thus making me aware of this utility.

Det0x's score: https://forums.anandtech.com/threads/zen-4-builders-thread.2605910/post-40890534

AdamK47's score: https://forums.anandtech.com/threads/zen-4-builders-thread.2605910/post-40890893

This is what I got on my Dell Optiplex work PC:



This is the best of three runs.

No other software should be running as it will result in lower scores.



Compare feature for DDR3 shows bandwidth is acceptable but latency could be better.

Download link: http://maxxpi2.de/pages/downloads/maxxmemsup2---download.php

Click the name next to the tiny Winzip icon:

 

[tcsenter]

Anyone else getting security/malware detections for MaxxMem2?

 

[igor_kavinski]

Anyone else getting security/malware detections for MaxxMem2?

It is using CPU-Z.exe in the background in the Windows Temp folder to query the information about RAM timings and CPU model etc. Maybe that's why.

I think it has the CPU-Z.exe embedded in it.

 

[igor_kavinski]

Anyone else getting security/malware detections for MaxxMem2?

Curious which antivirus you have because yes, this is suspicious behavior that Maxxmem tries to do by creating an EXE in the temp folder. The author probably found some easy hack instead of having to pay for the actual CPU-Z library meant for developers.

 

[tcsenter]

Curious which antivirus you have because yes, this is suspicious behavior that Maxxmem tries to do by creating an EXE in the temp folder. The author probably found some easy hack instead of having to pay for the actual CPU-Z library meant for developers.

Both Malwarebytes and BitDefender.

Malware.AI.4286915064, C:\USERS\TESTPC\DOWNLOADS\MAXXMEM2\MAXXMEM2.EXE

 

[igor_kavinski]

Both Malwarebytes and BitDefender.

Since it's AI, most likely false positive.

 

[igor_kavinski]

9950X3D Stock (no PBO anything), DDR5-7200C34 FCLK 2133 UCLK 1800 Windows Ultimate Power Plan

 

[Super Spartan]

Here are my results on my Razer Blade 16 (2025), which has AMD Ryzen AI 9 HX 370/ 64 GB LPDDR5X 8000 MHz RAM

 

[igor_kavinski]

Here are my results on my Razer Blade 16 (2025), which has AMD Ryzen AI 9 HX 370/ 64 GB LPDDR5X 8000 MHz RAM

View attachment 122016

This can be better if you use 7-max.

 

[Super Spartan]

This can be better if you use 7-max.

 

[aigomorla]

DDR5 ECC Reg. Gskill 6400. 48GB Sticks x 4.



Not bad for Server Ram eh?

 

[igor_kavinski]

Not bad for Server Ram eh?

You will really love it in LM Studio (if watching your computer "think" is something you enjoy) because LLMs want all the bandwidth they can get.

Also, do you have NPS settings under AMD CBS settings?

If I recall correctly, NPS1 gives the best performance. But you should try all settings from NPS0 to 4 to see which ones has the lowest latency or highest bandwidth.

Also, using 7-Max, you will get the "true" performance. 7-Max uses larger memory pages that optimized applications use by default.

6400C32 quad channel is pretty impressive!

 

[igor_kavinski]

My best so far (weirdly, my RAM is slower than my previous score config yet more performant. I must be at least an intermediate level Ryzen Master Master by now )

 

[igor_kavinski]

Not bad for Server Ram eh?

MaxxMem is actually a pretty bad benchmark for testing server RAM bandwidth. It can't use the extra channels effectively. OCCT gives a better bandwidth score with extra channels.

 

[igor_kavinski]

@leoneazzurro's 9955HX3D



With 7-Max:

 

[igor_kavinski]

I got this laptop in 2024:



And the OOB score I got:



That latency looked so bad that my heart sank and I ignored the laptop till yesterday. Booted it up, went into BIOS and turned off power saving related settings. And I got:



Latency halved!

With 7-Max:



Got the best performance with HT off:



And an interesting test to see how multitasking workloads affect bandwidth and latency:



Those four instances are continuously accessing their allocated pool of 500MB RAM and calculating the latency in the background.



Notice anything different? The correct maximum speed of LPDDR4-3200 is showing with the correct timings because an application (iggy RAM latency) is keeping the RAM working at full speed. In previous attempts, when MaxxMem would start, RAM would be chilling so querying speed and timings resulted in its "chill speed" being reported.