UserBenchmark -> Read this

[ctank]

Hello.

 

I used Userbenchmark named well, benchmarking software couple of times, and started to wonder why my whole disk score didn't show. Apparently, it requires 100 Gigabytes free space on hard drive. Obviously on especially SSD:s all people wont have that much free space, which basically no reason warps the results of benchmark score.

 

Now, the Benchmark does CPU, memory and GPU tests in 45 seconds. The HD tests on 3 drives lasts 3 minutes. I was starting to wonder, why benchmark like this puts so much weird effort (writing 100gigs) in disk benchmark, when it just seems to casually test the most important parts, CPU, GPU and memory. 

 

There is no reason to do that. It will do "Sustained writes" for  60 seconds per hard drive, and that time CPU usage is on highish level. Also you can't cancel the benchmark, except obviously from task manager which is weird because its directed to non power user (cancel and X button on corner greyed out). Also, even there is checkboxes on start to select only specific tests, they greyed out and you cant unselect anything. 

 

I will suggest that the program might do something we don't want it to do, maybe mine virtual currencies or something while it does these superficial tests.

 

 

[Jurrunio]

run it in a clean windows install with internet access disabled (until right before it finishes) then. That's the way to get the most score anyway

[ctank]

1 minute ago, Jurrunio said:

run it in a clean windows install with internet access disabled (until right before it finishes) then. That's the way to get the most score anyway

You missed the point by far.

[Oalei]

He/she/neither just want to win, being competitive is healthy 

[ctank]

You guys might want to read the whole post. And think for a minute.

 

EDIT: I made the point little more obvious

[TheBean]

yea no. definatly not doing that. i dont understand why they dont let you cancel or edit tests but it is not for nafarious reasons. 

it wants free space because ssd speeds can significantly decrease when space runs out. that will give you a really low score for your ssd even if it could be fast. 

[.Apex.]

8 minutes ago, ctank said:

There is no reason to do that. It will do "Sustained writes" for  60 seconds per hard drive

Actually it does sustained write for 30 seconds, I really don't see the problem, the whole point of sustained write is to check if it can sustain write speed, it's in the name, i don't know why i'm explaining this...

Also It doesn't require 100GB because it's gonna write 100GB, do you really think the program can write 100GB in 30 seconds? or even 60 seconds as you say

[ctank]

4 minutes ago, Saksham said:

yea no. definatly not doing that. i dont understand why they dont let you cancel or edit tests but it is not for nafarious reasons. 

it wants free space because ssd speeds can significantly decrease when space runs out. that will give you a really low score for your ssd even if it could be fast. Also It doesn't require 100GB because it's gonna write 100GB, do you really think the program can write 100GB in 30 seconds? or even 60 seconds as you say

"Definatly not doing that" how do you know? Have you reverse engineered the software?

 "De facto" standard in disk speeds that is Crystaldiskmark uses 5 gigs for example. 

 

And thats one of the reasons its weird, requiring actually 132Gb for me, so my guess is that it just stalls the test to do something else. My SSD has 2200Mb/sec writ, so it requires 60 times max write speed which is just idiotic way to do disk speed test. 

[TheBean]

2 minutes ago, ctank said:

Its completely unnecessary for this kind of superficial test. "De facto" standard in disk speeds that is Crystaldiskmark uses 5 gigs for example.

i dont know how much space it wants, i doubt it is 100gb (doubt it is higher than 2gb). just saying that free space matters

[.Apex.]

1 minute ago, ctank said:

Its completely unnecessary for this kind of superficial test. "De facto" standard in disk speeds that is Crystaldiskmark uses 5 gigs for example.

I just said it doesn't write 100GB, where did you even get this 100GB thing from? I just ran the test and my SSD has 45GB left, and it didn't warn me about anything.

[Fasauceome]

27 minutes ago, ctank said:

people wont have that much free space, which basically no reason warps the results of benchmark score.

Having very little free space on an SSD greatly reduces drive performance, it's not nefarious.

[Jurrunio]

29 minutes ago, ctank said:

Apparently, it requires 100 Gigabytes free space on hard drive.

Just tried it, it doesnt have that requirement.

 

https://www.userbenchmark.com/UserRun/16392445

 

12GB free on the boot SSD, 27GB on the bulk storage 1TB HDD.

 

Yes both perform shit because how I filled them up to the brim, but it still runs

[ctank]

9 minutes ago, _Syn_ said:

Actually it does sustained write for 30 seconds, I really don't see the problem, the whole point of sustained write is to check if it can sustain write speed, it's in the name, i don't know why i'm explaining this...
 

Why you write even you don't know? Can you read, read this text from their website 

[TheBean]

what ssd do you have? the faster it is, the more free space it wants. 

to do a real sustained performance test, it will need to continously write stuff. if you dont have space, it cant do that and you will get a improper result.

[.Apex.]

1 minute ago, ctank said:

Why you write even you don't know? Can you read, read this text from their website 

But you can clearly see it's counting down from 34 seconds, and mine counted down from 32 seconds.....

[Jurrunio]

31 minutes ago, ctank said:

The HD tests on 3 drives lasts 3 minutes.

If you ask me, I think it's scaling up the test time to compensate for cache and some storage acceleration things like AMD StoreMI and Samsung Magician's rapid mode. 

 

33 minutes ago, ctank said:

Also you can't cancel the benchmark, except obviously from task manager which is weird because its directed to non power user (cancel and X button on corner greyed out). Also, even there is checkboxes on start to select only specific tests, they greyed out and you cant unselect anything. 

taking the highest priority is the best way to make sure background apps dont take away resources and make the score inaccurate

[ctank]

2 minutes ago, _Syn_ said:

But you can clearly see it's counting down from 34 seconds, and mine counted down from 32 seconds.....

I dont know, other drives it counts from 60 second, like my system drive

 

 

3 minutes ago, Saksham said:

what ssd do you have? the faster it is, the more free space it wants. 

to do a real sustained performance test, it will need to continously write stuff. if you dont have space, it cant do that and you will get a improper result.

I have EVO 970, obviously more space it wants but that is not the point, but that I suspect it does something else while reserving system mostly with disk tests.

 

I suggest that someone with time and tech savvy to investigate this more, would do so. 

[Firewrath9]

Becuase, say a drive that is full will perform much worse than a drive that is half full, or even 80% full. Thats why I have 15% of empty space. Besides, lets say they mine a cryptocurrency, for what, 1 or 2 cents?

[ctank]

4 minutes ago, Jurrunio said:

taking the highest priority is the best way to make sure background apps dont take away resources and make the score inaccurate

You can take highest process priority while enabling ability to cancel test. Never seen other benchmark which does that. And its not like this test is serious and accurate anyway, the results are named "UFO, Nuclear Submarine" and so on lol. 

[TheBean]

1 minute ago, ctank said:

You can take highest process priority while enabling ability to cancel test. Never seen other benchmark which does that. And its not like this test is serious and accurate anyway, the results are named "UFO, Nuclear Submarine" and so on lol. 

those are just names. it can be anything. doesnt matter. IMO those are good names. you can easily identify which is which.

 

it reserves space so that no other program can take it while doing the test. it writes a blank file. you can try it yourself as there is a cmd command to make one yourself. 

[.Apex.]

2 minutes ago, ctank said:

dont know, other drives it counts from 60 second, like my system drive

Okay just ran some tests and it seems to start doing the "sustained write" test during the GPU benchmark, and when it finishes the GPU benchmark it starts counting down how much left for the sustained writes, that's why it's not consistent, but still I don't see the problem, how much it writes depends on the speed of your drive, for an HDD it will most likely be little, let's say 100mb/s? 60*100mb = 6GB, for a SATA SSD it's 30GB as stated

Sustained write is an important test in my opinion, some hard drives or SSD's do great in the normal sequential read and sequential write tests, which only last a few seconds, but once you stress it for longer than that then it will show its true colors, which is very realistic

[Jurrunio]

1 minute ago, _Syn_ said:

which only last a few seconds, but once you stress it for longer than that then it will show its true colors,

exactly. SLC caching on SSDs for example, is super fast but unrealistic as they hardly last in real world use when speed matters.

[ctank]

46 minutes ago, Saksham said:

it reserves space so that no other program can take it while doing the test. it writes a blank file. you can try it yourself as there is a cmd command to make one yourself. 

How do you know? Have you opened the test file and looked what it writes? Of course it reserves space, completely not related to my point. 

[.Apex.]

https://ssd.userbenchmark.com/SpeedTest/494033/Samsung-SSD-970-EVO-250GB

 

Look at this 970 EVO for example, the sequential write is 1400mb/s, but the sustained write is 400mb/s, which is what you would see in a real world scenario when transferring files.

[.Apex.]

https://ssd.userbenchmark.com/SpeedTest/498971/Samsung-SSD-970-PRO-512GB

 

But look at the 970 PRO, the sequential write and sustained write are almost the same, see why it's important?