We need an alternative to userbenchmark.com

[farzher]

userbenchmark is invaluable. i could give countless examples of how useful it is.
but uh yeah, safe to say the tech community is very against the biased people running the site. especially after what just happened with the 7600x userbenchmark

maybe this is a perfect project for the new ltt labs full time software developer to work on?

maybe the community could make an alternative ourselves?

i run userbenchmark on all my computers and all my friends computers. the relative performance percentile is so good.
shows you if your ram is slower than it should be (probably forgot to enable xmp)
shows you how much performance you could gain by overclocking your specific gpu / cpu
shows you if your ssd is slower than it should be. in my case it turned out i had it plugged into a sata 2 port for years!
when your friend says "my laptop is too slow, pls help" i have them run userbenchmark and notice the cpu is running at 0.6GHz. probably the fan is broken or full of lint.
etc.

[da na]

My go-to is Cinema 4D render tests, FurMark, and HWiNFO. But the convenience of having all of that in one with userbenchmark, plus it being a portable application...

Older versions of 3DMark and PCMark are free, for the full version - that's an alternative but they don't have as comprehensive of a comparison database

[SorryBella]

25 minutes ago, farzher said:

i run userbenchmark on all my computers and all my friends computers. the relative performance percentile is so good.
shows you if your ram is slower than it should be (probably forgot to enable xmp)
shows you how much performance you could gain by overclocking your specific gpu / cpu
shows you if your ssd is slower than it should be. in my case it turned out i had it plugged into a sata 2 port for years!
when your friend says "my laptop is too slow, pls help" i have them run userbenchmark and notice the cpu is running at 0.6GHz. probably the fan is broken or full of lint.
etc.

3DMark and Cinebench combo has been my go to for 1 load with either HWiNFO64 or MSI Afterburner to read the sensors, while for all PC load its still combination of P95 and Furmark. But tbh, even with the flawed methodology and benchmark tools their have (daily reminder that they dont have resolution scaling and thus you can "gain" stupid performance by running the benchmark at low res) their database of gullible people are still "grain of salt" good enough for stupid layman. But for me personally? I elect to just look at similarly spec'd system, and compare it to the current system, while keeping my eyes at clock speed and temperature.

[farzher]

26 minutes ago, SorryClaire said:

3DMark and Cinebench combo has been my go to

3dmark score means nothing to me. cinebench score means nothing to me.
"your pc is performing way below expectations, 30% percentile" means a lot.
especially when it shows i have slow RAM with poor performance and only 1 of 4 slots are in use.

[nbrowser]

heh just an aside...UserBenchmark shits the bed testing my not that old gaming laptop, doesn't properly test the GPU and the test falls flat on it's face. No matter what tweaks are done the run always comes up all zeros, quite funny.

[GuiltySpark_]

I don't see what's wrong with Cinebench/CrystalDiskMark(or any disk benchmark)/3DMark. We've been using those tools for years and they make sense. Real easy to compare to independent reviews/benchmark results. 

[LAwLz]

5 hours ago, farzher said:

userbenchmark is invaluable. i could give countless examples of how useful it is.

Is it really though? I have only used it like, once.

 

 

5 hours ago, farzher said:

but uh yeah, safe to say the tech community is very against the biased people running the site. especially after what just happened with the 7600x userbenchmark

Honestly, I think the bigger issue is that people are very biased want want to hate userbenchmark.

It seems like people don't really understand how the site works.

 

It all started when they adjusted their weighting so that single core performance was valued over multiscore performance. I thought that was a completely justified change. A lot of the high scoring CPUs with a lot of cores were not actually better than some of the lower core count parts for everyday tasks. It gave a skewed outlook on things at the time.

It all depends on what the aim of a benchmark is. I think it is important to note that there is no universal truth to which CPU is the "best" when talking about it in general. For some workloads a lot of cores are preferred even if each core is slower. But for other workloads the amount of cores barely matters. It is up to the designer of the scoring system to decide how much or how little weight to put on each task.

 

Secondly, I have noticed at several points on this forum that people don't seem to understand how userbenchmark works. People seem to think that the user submitted writeups are the words of the admins, which they aren't. 

Just because some text appears on their website does not mean it were the admins that wrote it. Likewise, just because this text appears on the website called "LinusTechTips" does not mean the words I have written here are officially endorsed by Linus himself.

 

Now, the user who wrote the review currently only review of the 7600X seems to be an OG of userbenchmark (user ID 8) and they might be an admin. The fact that it is called "Advanced Marketing Devices" and "Zen4 Hype Train" seems to indicate that whoever did this is at the very least a bit of a hater when it comes to AMD. But I am just speaking in general. People have such bad takes when it comes to userbenchmark. It ranges from people not understanding what the site is or how it works, to people deliberately trying to skew facts to make it look bad.

 

I think this particular case in indefensible, but in my eyes this is their first offense. All other controversies have just been bullshit that quite frankly aligns with the venomous comments written by that user on userbenchmark.

 

 

Here is the previous thread where I debunk a lot of misconceptions regarding userbenchmark.

[LogicalDrm]

-> Moved to Programs, Apps and Websites

[X-System]

13 hours ago, farzher said:

userbenchmark is invaluable. i could give countless examples of how useful it is.
but uh yeah, safe to say the tech community is very against the biased people running the site. especially after what just happened with the 7600x userbenchmark

maybe this is a perfect project for the new ltt labs full time software developer to work on?

maybe the community could make an alternative ourselves?

i run userbenchmark on all my computers and all my friends computers. the relative performance percentile is so good.
shows you if your ram is slower than it should be (probably forgot to enable xmp)

 

I use AIDA64 to run the benchmark to see how much GB/s between theory and real for the frequency.

 

MHz * 8 = MB/s

 

And multiply by 2 for dual channel or multiply by 4 for quad channel or multiply by 8 for octal channel, etc...

 

For example : 2 sticks DDR4-3200 in dual channel mode. In theory, it does 51.2GB/s and if you set 2133MHz by mistake, it gives 34.128GB/s. So you would know there is a wrong in the settings.

 

Quote

shows you how much performance you could gain by overclocking your specific gpu / cpu

 

3DMark and Cinebench.

 

But you said : "3dmark score means nothing to me. cinebench score means nothing to me"

 

First, you run the tests in stock and get the scores. Then overclock you CPU / GPU and run the tests to get others scores.

 

100 * (Score in overclock - Score in stock) / Score in stock = gain in percent

 

Quote

shows you if your ssd is slower than it should be. in my case it turned out i had it plugged into a sata 2 port for years!

 

I use CrystalDiskMark and ATTO to see the performance of SSD in MB/s.

 

Quote

when your friend says "my laptop is too slow, pls help" i have them run userbenchmark and notice the cpu is running at 0.6GHz. probably the fan is broken or full of lint.

 

For CPU, check with HWinfo sensors for the frequency and the temperature. And if the laptop uses HDD 5400rpm and fragmented, it also does the laptop slow. And if the laptop has a lot of bloatwares, etc...

[farzher]

4 hours ago, X-System said:

 

I use AIDA64 to run the benchmark to see how much GB/s between theory and real for the frequency.

 

MHz * 8 = MB/s

 

And multiply by 2 for dual channel or multiply by 4 for quad channel or multiply by 8 for octal channel, etc...

 

For example : 2 sticks DDR4-3200 in dual channel mode. In theory, it does 51.2GB/s and if you set 2133MHz by mistake, it gives 34.128GB/s. So you would know there is a wrong in the settings.

 

 

3DMark and Cinebench.

 

But you said : "3dmark score means nothing to me. cinebench score means nothing to me"

 

First, you run the tests in stock and get the scores. Then overclock you CPU / GPU and run the tests to get others scores.

 

100 * (Score in overclock - Score in stock) / Score in stock = gain in percent

 

 

I use CrystalDiskMark and ATTO to see the performance of SSD in MB/s.

 

 

For CPU, check with HWinfo sensors for the frequency and the temperature. And if the laptop uses HDD 5400rpm and fragmented, it also does the laptop slow. And if the laptop has a lot of bloatwares, etc...

..... or simply run userbenchmark. wtf lmao.

and the thing is, i don't always know what i'm looking for. if i'm just doing a checkup on a friend's laptop. i'm not even sure what they have in there. i have no clue how many mb/s read/writes the ram/drive should be getting. i need that database of relative performance.

[X-System]

13 hours ago, farzher said:

..... or simply run userbenchmark. wtf lmao.

and the thing is, i don't always know what i'm looking for. if i'm just doing a checkup on a friend's laptop. i'm not even sure what they have in there. i have no clue how many mb/s read/writes the ram/drive should be getting. i need that database of relative performance.

Then don't check on a friend's laptop and leave that laptop to a computer expert who doesn't need Userbenchmark for checkup because Userbenchmark isn't a checkup tool.

[IkeaGnome]

On 7/31/2022 at 5:03 PM, farzher said:

the relative performance percentile is so good.

I'm not against using it for a quick diagnosis for apples to apples comparisons. 7700k to 7700k, 5600x to 5600x, 3070 to 3070 etc.

I won't use it for apples to oranges of any kind. 7700k to 5600x etc.

13 hours ago, X-System said:

checkup because Userbenchmark isn't a checkup tool.

I'd liken it to your usual checking of the dipstick on an engine or transmission. Is oil level good? Is oil looking muddy? That's about it. I wouldn't use it to try to figure out why the oil is getting water in it.