I wrote a CPU Benchmark

[RyzenX770]

I made Nidam Benchmark to measure how fast your CPU can authenticate users using the OAuth 2.0 standard. It’s a full-stack Java + OAuth 2.0 microservice benchmark (reverse-proxy, auth server, BFF, protected API and in-memory DB) that simulates real users logging in and calling protected APIs. The authentication flow uses scrypt, a password hashing algorithm designed specifically not to be hardware accelerated for security reasons. Modern production systems favor scrypt and similar memory-hard algorithms instead of hardware-accelerated primitives like AES or SHA for password storage, which means the workload stresses raw CPU cores, sustained frequency, memory bandwidth, and latency in a very real-world way. Results are produced as a neat HTML report with throughput, latency, thread-scaling charts and an “optimal thread” score.

 

If you’re a reviewer or bench nerd, give it a spin (Windows and Linux; GUI + CLI; x64 + ARM) and compare how your chip behaves under end-to-end backend load. I’ve linked the project page below — please post your system specs and report HTML here, and open issues if you see anything odd (ARM builds are untested on my side, so extra feedback there is especially welcome).

 

https://nidam.derbyware.com/benchmark/

https://github.com/Mehdi-HAFID/Nidam-Benchmark-Application/releases/tag/2.0

 

 

[MadAnt250]

Would you recommend your benchmark for overclocking stress tests?

[danalog]

I will -TRY- to get this running on a hardware cryptography accelerator. I am curious to see the results. 

[notabitail]

I'm going to have to give this a try. 

 

Does it have system requirements or will this run on just about any x64 device?

[RyzenX770]

13 hours ago, MadAnt250 said:

Would you recommend your benchmark for overclocking stress tests?

it not a stress test, for each threat count the benchmark runs for 50 seconds, and before moving to the next thread count it does some light stuff before starting again for 50s. so it does NOT keep the cpu at 100% all of the time, that was not the goal. but give a try and let me know.

 

12 hours ago, danalog said:

I will -TRY- to get this running on a hardware cryptography accelerator. I am curious to see the results. 

first time i heard of such a thing, not surprised though. i asked gpt, and it said it won't work, scrypt is intentionally designed to resist hardware acceleration. but let know how it goes.

 

11 hours ago, notabitail said:

I'm going to have to give this a try. 

 

Does it have system requirements or will this run on just about any x64 device?

I tested on an old 8GB ram laptop, and it run. but the whole thing requires about 5/5.5GB Ram, you see it has a database of 1 million users. and they're loaded in memory.

 

so did anyone run it? I would love to see you scores. just when the html page opens, there is a colorful 'Download PNG', no need to take a screenshot.

 

also if anyone with an arm pc I would love if you can confirm if it works. as I say in the benchmark page I don't have an arm pc so I don't know if they work, they are supposed to work but until someone confirm I will never know.

[notabitail]

12 hours ago, RyzenX770 said:

it not a stress test, for each threat count the benchmark runs for 50 seconds, and before moving to the next thread count it does some light stuff before starting again for 50s. so it does NOT keep the cpu at 100% all of the time, that was not the goal. but give a try and let me know.

 

first time i heard of such a thing, not surprised though. i asked gpt, and it said it won't work, scrypt is intentionally designed to resist hardware acceleration. but let know how it goes.

 

I tested on an old 8GB ram laptop, and it run. but the whole thing requires about 5/5.5GB Ram, you see it has a database of 1 million users. and they're loaded in memory.

 

so did anyone run it? I would love to see you scores. just when the html page opens, there is a colorful 'Download PNG', no need to take a screenshot.

 

also if anyone with an arm pc I would love if you can confirm if it works. as I say in the benchmark page I don't have an arm pc so I don't know if they work, they are supposed to work but until someone confirm I will never know.

I'm going to give it a try.

 

I have a PineTab 2 on ARM, going to dig it out and give this a try.

[notabitail]

@RyzenX770

This is on Kubuntu 25.10, Intel Core i7-1185G7

 

It's still loading itself on the PineTab 2, that thing is slow.

[notabitail]

And on ARM, PineTab 2

[RyzenX770]

12 hours ago, notabitail said:

I'm going to give it a try.

 

I have a PineTab 2 on ARM, going to dig it out and give this a try.

Man You're a life safer, it runs on arm linux, hell yeah. I really appreciate you doing that, I really do.

 

for the 1185G7 since it 4C/8T, the value you want to enter is: 1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16 this will give you a score.

the reasons is you want to test to 2 times the the threads count your cpu has, and since it is 4 cores, increment the value by 1, if you have an 8 core or more cpu, then increment by the thread count per core for the cpu which is 2 if it has SMT, for example 8C/16T: 1,2,4,6,8,10,12,14,16,18,20,22,24,26,28,30,32 it.

I just took a quick look and seen your comment and i said i had to write a comment now, in an hour or so. I will write anther one they are some other things i want to say.

 

THANK YOU, it is really rewarding that someone downlaod the benchmark and run it. thanks a lot.

[RyzenX770]

@notabitail

The thing i want to address are the error logs are the end:

I just want to let you know that the error at the end of logs has nothing to do with the benchmark but with the OS, gpt put it better than i can:

Quote

Those messages are harmless GUI warnings that appear after the benchmark has already finished.

On Kubuntu 25.10 (Wayland), the embedded Chromium UI logs a few shutdown warnings related to Wayland/Vulkan and thread cleanup. They do not affect the benchmark, results, or CPU measurements in any way.

 

Chromium is trying to use Vulkan GPU acceleration, which isn’t fully supported with Wayland in this setup.

 

you can safely ignore the messages — the run was successful and the results are valid.

 

as for 1185G7, if you happen to run it again with the value mentioned in the previous comment, let me know the score, you can even embed the html file in your comment.

[notabitail]

6 hours ago, RyzenX770 said:

Man You're a life safer, it runs on arm linux, hell yeah. I really appreciate you doing that, I really do.

 

for the 1185G7 since it 4C/8T, the value you want to enter is: 1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16 this will give you a score.

the reasons is you want to test to 2 times the the threads count your cpu has, and since it is 4 cores, increment the value by 1, if you have an 8 core or more cpu, then increment by the thread count per core for the cpu which is 2 if it has SMT, for example 8C/16T: 1,2,4,6,8,10,12,14,16,18,20,22,24,26,28,30,32 it.

I just took a quick look and seen your comment and i said i had to write a comment now, in an hour or so. I will write anther one they are some other things i want to say.

 

THANK YOU, it is really rewarding that someone downlaod the benchmark and run it. thanks a lot.

Thanks, I'll give this a try. 

 

I have a stupid amount of devices I can run this on for the fun of it, I'm going to probably spend a few days playing around with it.