Links to gpu dot userbenchmark dot com should automatically flag posts for deletion.

[Queen Chrysalis]

The CPU individual tests are OK, and the actual fps comparisons are nice WHEN they are available, but the comparisons made across architectures and generations are so mind-bogglingly misleading as compared to just looking up actual in-game tests on youtube that I can't even.  

 

The global general 'speed' comparisons are just downright bad, especially considering that newness and market share are weighed into these calculations without consideration for actual features that accompany newness.

[Queen Chrysalis]

Also worth mentioning that it seems some of the synthetic graphics benchmarks are pretty globally more important than others, even after they redid it a couple years ago.  The MRender seems to almost be the defining one for simulating a lot of stuff actually happening in a game environment, yet for example, my old vega that gets beaten comfortably by my current 6600 in every game i've played, is rated as being like 23 percent faster by the general 'speed' metric.

[Hachi_Roku256563]

Maybe not

Depends on the kind of post

if its a Should i get X gpu vs X gpu and they use Userbenchmark they still have a reasonable question

Explain that userbenchmark is bad and continue on with the question

[Helpful Tech Witch]

Maybe something like a Reddit bot or similar that makes a note saying “The site userbenchmark is known to be inacirate and inconsisten in comparison resaults, consider using a different source for comparisons in the future”

[RONOTHAN##]

50 minutes ago, Queen Chrysalis said:

The CPU individual tests are OK

I give you their 11900k review...

Quote

The i9-11900K is the fastest CPU in Intel’s 11th Gen Rocket Lake-S lineup which brings higher IPC (early samples indicate +19%) and 50% stronger integrated graphics. There are also 500 series chipset improvements including: 20 PCIe4 CPU lanes (up from 16) and USB 3.2 Gen 2x2 (20 Gbps up from 10 Gbps). Rocket Lake’s IPC uplift translates to around a 10% faster Effective Speed than both Intel's 10th Gen and AMD’s 5000 series. Despite Intel’s performance lead, AMD continues to outsell Intel. Given the scale of Intel's operation, it’s inexplicable that their marketing remains so neglected. Little effort is made to counter widespread disinformation such as: “it uses too much electricity” or the classic: “it needs more cores”. Intel’s marketing samples are routinely distributed to reviewers that appear better incentivized to bury Intel's products rather than review them. Not enabling XMP or only testing with BIOS power limits enabled is akin to leaving the handbrake on during a race. Mind-numbing “scientific” and rendering benchmarks are presented as gospel. Different games, mostly unplayed by real users, are cherry picked for each “review”. When it's convenient, canned game benchmarks are chosen such as Ulletical’s CSGO which runs at nearly double the in-game fps. Credible benchmark data, which necessarily includes replicable video footage from popular games, is the exception rather than the rule. At every release, AMD’s marketers coordinate narratives to ensure another feast of blue blubber. Nonetheless, towards the end of 2021, Intel’s Golden Cove is due to offer an additional 20-30% performance increase. At that time, with a net 30-40% performance lead, Intel will probably regain significant market share despite AMD's class-leading marketing. In the meantime, most PC gamers need look no further than the 11400F.

I wouldn't touch their CPU stuff with a 10 foot pole if this is what they end up saying about a CPU that is as bad as the 11900k, then saying "No we're not biased towards Intel, everyone else is just biased towards AMD."

 

Probably a less radical but more effective move would be to just like 2kliksphilips video every time Userbenchmark is even mentioned. It's been my strategy since he released it about a year ago. 

 

[LogicalDrm]

As in this forum? Because that would be limiting opinions that aren't inheritely against any of the major forum guidelines. If you have issues with quality of source, you should explain why and offer better sources. We don't moderate content/opinions, just manner they are represented (aggressive, trollish, attacking, etc.)

Edited March 21, 2022 by LogicalDrm

[manikyath]

yes, userbenchmark's "overall score" and reviews are the absolute ass-est source you can get, but sometimes you need to do a really odd-ball comparison, and then the individual tests (which you can still see on userbenchmark) at least still give an indication, where you cant always rely on comparative game tests exisitng for every combination of two cards.

 

i've never been a fan of using userbenchmark for more than it's individual tests as a vague comparison of the hardware's capabilities (so not necessarily for game performance comparisons), but blocking the platform outright seems like a "nuke from orbit" approach.

 

if we go towards taking action (be that blocking, be that adding them to the community standards, or something else), should we then also be taking action for sellout 'reviewers', and other sources of bad performance data? you cant single out UBM because the fanboy that writes the reviews is a fanboy, and because their overall score calculation was written by a first grader. it's either going for a blanket ban on bad sources, or not banning at all.

[LAwLz]

Bad idea.

The forum itself should not be an arbiter of what is or isn't considered "good sources". If you start doing it for one website, do you also have to do it for other websites that can be misleading? For example LinusTechTips videos are full of misinformation. Should linking to an LTT video also flag the post for deletion, because it could be considered a bad source for information?

 

If you don't like UserBenchmark then explain why when someone brings it up as a source. Don't just go "I don't like this, so it shouldn't be allowed on the forum".

That's the entire point of the forum, to discuss things.

[Arika]

Userbench has it's place. just like all review sites, you should never just use 1 source when determining what is the right product for you anyway.

[Queen Chrysalis]

5 hours ago, Arika S said:

Userbench has it's place. just like all review sites, you should never just use 1 source when determining what is the right product for you anyway.

Right, right, I even used it to just see raw what the single core performance for Alder lake looked like as compared to skylake ++++++++++++++++++++, but that's about it.  There is nothing I'd want to see from them related to a GPU, like ever.  Like, I remember back in 2016 when my roommate was buying a new computer, the 1080 was being called like 15 or so percent faster than the 1070, yet in almost every games it was like a 3 percent fps difference at best (the gap opened up a little bit later with new drivers, but at the time of purchase it would have been really stupid).

[LAwLz]

17 minutes ago, Queen Chrysalis said:

Like, I remember back in 2016 when my roommate was buying a new computer, the 1080 was being called like 15 or so percent faster than the 1070, yet in almost every games it was like a 3 percent fps difference at best (the gap opened up a little bit later with new drivers, but at the time of purchase it would have been really stupid).

Ehm... The difference between the GTX 1080 and GTX 1070 was along the lines of 20%, even at launch. 

Not sure where you got the idea that it was only 3% from.

Spoiler

 

Even in games:

Spoiler

 

If userbenchmark said there was a 15% difference then it was actually pretty close. Much closer than the 3% you thought it was.

[Queen Chrysalis]

We were mainly playing csgo, league and wow.  GTAV may have been more dramatic.

[Queen Chrysalis]

I'll admit the 1070 was overkill for those games, but that was him not me.  I didn't get a new GPU till 2018 for pubg.

[LAwLz]

4 minutes ago, Queen Chrysalis said:

We were mainly playing csgo, league and wow.  GTAV may have been more dramatic.

9 minutes ago, Queen Chrysalis said:

I'll admit the 1070 was overkill for those games, but that was him not me.  I didn't get a new GPU till 2018 for pubg.

But userbenchmark says the 1070 gets an average of 191 FPS in CS:Go, and the 1080 gets an average of 199.

So I am not really sure what your problem with userbench is to be honest. It seems like it is reporting exactly what you claim to be reality (that the 1070 and 1080 performs the same in CS:Go).

 

It seems to me like in this case the issue isn't userbenchmark. It is your way of reading benchmark results that leads you to be confused.

You can't look up results from one benchmark and then assume all workloads will show the same results. Just because techpowerup says the average difference between a 1070 and a 1080 is ~20% does not mean it is "misleading" just because the same does not hold true when playing CS:Go. If you want to know how something performs specifically in CS:Go then you need to look up benchmarks, preferably from multiple sources, of specifically CS:Go.

[Queen Chrysalis]

The complaint is focused on the primary 'speed' metrics less informed consumers will be greeted with at the splash of the page.  The frustration with the 1070/1080 thing just case from my memory,  but the larger issue still persists with broad comparisons users are funneled to immediately that leads to bad decision making in general, especially at a time when a lot of people are buying older cards. 

[Helpful Tech Witch]

18 minutes ago, Caroline said:

Not that I'm a cop but were you under the influence of alcoholic beverages while writing this?

 

wait a minute.... didn't your username used to be helpful tech wizard? now it's just wiard, huh

Dude I wish

but your about 7 years too early

also yea it’s a typo when I changed it. I have to wait to June to fix

have you seen the site? They say a 5800x is better than a 5950x….

[IkeaGnome]

Just going to leave this here.

On 9/30/2021 at 10:59 AM, IkeaGnome said:

Depends on how you interpret this. 8350k is 10% better than a 2990x, 8350k and 7700k are 2% faster than 5600x, but 7700k is 6% faster than 8350k?

  Reveal hidden contents

 

Should add some comedy to this thread.

[LAwLz]

19 hours ago, Helpful Tech Wiard said:

have you seen the site? They say a 5800x is better than a 5950x….

Ehm, no?

 

It says that the 5950X is very, very slightly ahead of the 5800X for gaming and general tasks, but the 5950X is way ahead for workstation tasks. That is pretty accurate if you ask me.

[Helpful Tech Witch]

1 hour ago, LAwLz said:

Ehm, no?

 

It says that the 5950X is very, very slightly ahead of the 5800X for gaming and general tasks, but the 5950X is way ahead for workstation tasks. That is pretty accurate if you ask me.

Sure, but the average person will look at the 87 vs 65 scores and assume that the 5800x is faster…

[Avocado Diaboli]

20 minutes ago, Helpful Tech Wiard said:

Sure, but the average person will look at the 87 vs 65 scores and assume that the 5800x is faster…

Those are user ratings:

 

Is it a confusing layout? Maybe, but that's what happens when you let users rate their stuff. You don't have to prove that you own one to rate one, the rating is basically arbitrary and has nothing to do with any benchmarks whatsoever.

[LAwLz]

3 hours ago, Helpful Tech Wiard said:

Sure, but the average person will look at the 87 vs 65 scores and assume that the 5800x is faster…

As Avocado pointed out, you're looking at user ratings.

It might not be that obvious at first, but the up and down arrows was, in my opinion, a pretty clear indicator that it was some kind of user rating and not the benchmark scores.

 

 

3 hours ago, Avocado Diaboli said:

Those are user ratings:

 

Is it a confusing layout? Maybe, but that's what happens when you let users rate their stuff. You don't have to prove that you own one to rate one, the rating is basically arbitrary and has nothing to do with any benchmarks whatsoever.

Honestly, I have never payed much attention to userbenchmark but after reading this thread I am fairly sure the people criticise it are just confused. The handful of scores I have looked at on userbenchmark during this thread have been fairly spot on.

 

People being upset that an aggerate "gaming" score does not perfectly match the performance in a particular game, even though it matches other aggregate scores. This is perfectly normal and applies to all benchmarks, not just userbenchmark. If you want to know the performance in a particular application you have to look up benchmarks for that application, not generalized aggregate scores. That's not to say aggregate scores are bad by the way. I think they are usually more useful for the average consumer who uses their computers for a wider variety of tasks.

 

People thinking that user ratings are the benchmark scores, rather than look at the actual scores titled "benchmarks". Maybe this is an issue with the design of the website that makes people confused, but I thought it was pretty straight forward. 

 

Earlier I saw someone (@RONOTHAN##) post a user comment and implying that it is an official statement from UserBenchmark, which it isn't. It's just a comment from a user. You can't attribute a user comment to the site owner. If we start doing that then are we going to start claiming that whenever someone posts something dumb on this forum, Linus should be criticised for the content of that comment? If someone on this websites says an old i3 is faster than the 5600X, should we say "Linus thinks an old i3 is faster than the 5600X, it says so on his website"? That would be ridiculous.