How Accurate Is UserBenchmark?

[Kraigolas]

I found this forum post which was specifically created to answer the question in my title. Comments linked to this other post, and the end result is that both posts discuss how User Benchmark is bad and inaccurate without sourcing any of their information or justifying it whatsoever. 

 

I am, however, skeptical of UserBenchmark simply because my AMD Radeon software recommends I use an AMD Ryzen 9 3950X processor (a whopping $1000) over my Intel Core i5-9600K (valued at $300) and according to UserBenchmark, at least for gaming FPS, the Core i5 is still the winner. Based on my current knowledge, AMD generally makes reasonably priced processors with a lot of cores, which is exceptional for multiprocessing, so by no means am I insinuating that Intel is better than AMD, but I just haven't been able to find an AMD processor on UserBenchmark which can beat my Core i5, especially not for the price. Call of Duty Warzone sends all 6 of my cores to 100% according to Resource Monitor, from which I assume "gaming is difficult to execute in parallel" is actually coming to an end, because the developers must have done so in this game (or I am missing something) so I'd expect AMDs high core counts to become more relevant moving forward. 

 

Regardless, I was wondering what the basis for the benchmarks are: after a benchmark, UserBenchmark asks you to supply the FPS for the games that you play, so it's possible that AMD users are under reporting, Intel users are over reporting, or something nefarious is going on. I don't expect anyone to be knowledgeable of UserBenchmark's algorithms for filtering out incorrect FPS reports, perhaps they construct their own rigs to test in game FPS but I haven't a clue, but what has YOUR experience been with UserBenchmark, have you taken their advice and felt that the performance of your processor / graphics card reasonably represents what you were told to expect? 

[gloop]

1 minute ago, Kraigolas said:

my AMD Radeon software recommends I use an AMD Ryzen 9 3950X processor (a whopping $1000) over my Intel Core i5-9600K

They are doing that because they want to sell you their flagship product, what do you expect them to do? Even with my 2700x it recommends a 3950x.

[Prodigy_Smit]

2 minutes ago, Kraigolas said:

I use an AMD Ryzen 9 3950X processor (a whopping $1000) over my Intel Core i5-9600K (valued at $300)

There is a known bug with the Radeon software, it does not care what the hell hell your system is and tells you to upgrade to a Ryzen 9 3950x and a 5700xt. There are people with 5700xt and the software tells them that your gpu is underpowered for your games upgrade to a 5700xt.

[Kraigolas]

1 minute ago, gloop said:

They are doing that because they want to sell you their flagship product, what do you expect them to do? Even with my 2700x it recommends a 3950x.

This was a separate question that I was going to ask: does AMD just recommend a reasonable upgrade, or do they recommend their "best product", so I appreciate you answering it. It seems silly of them to recommend a $1000 "upgrade" to my $300 CPU. 

[Chris Pratt]

Just now, gloop said:

They are doing that because they want to sell you their flagship product, what do you expect them to do? Even with my 2700x it recommends a 3950x.

Yeah, the Radeon Software recommendations are stupid. That's been a know issue for a while.

 

I think you pretty much answered your own question with UserBenchmark, though. The reports can be manipulated too much, resulting in highly inaccurate data overall. Take 3DMark, for example. There's very strict control over when you can report your benchmark. For example, if you simply don't run it full screen or interact with another app while it's running, it will not let you report.

 

[You_found_dom]

You tell me how accurate it is

[Kraigolas]

Just now, BlueScope819 said:

Again, see my post. The numbers speak for themselves. (in the screenshots)

The screenshots show only the very top of the comparison, without any in depth information, like where UserBenchmark actually compares in game FPS, compares memory latency, single, duo, quad, octa core speeds, etc etc etc. To be fully honest, I don't really see what you mean by the numbers speak for themselves. They rate the PC separately for gaming, desktop use, and workstation use, so even if you just looked at the top of the page you'd see there's different applications for each use, and I could appreciate that it's difficult to rate a processor for gaming because depending on the game, you might use less cores or more cores, and from there it's unclear which will be better. But on a direct comparison, user benchmark sees the cores of the i3 are faster and this should in general translate to a better gaming experience assuming the game can't use more cores than you have to offer. 

[TetraSky]

14 minutes ago, Kraigolas said:

AMD Radeon software recommends I use an AMD Ryzen 9 3950X processor

AMD's software always recommend the highest tier of hardware they have available.

Even for me with me 3600, they say it's not good enough and they recommend the 3950x, because more cores = more better.

I disabled recommendations after that and suggest you do the same. 

Click the cog in the top right > General > Preferences > Upgrade Advisor (While you're there, disable Advertisements, too.)

[Prodigy_Smit]

7 minutes ago, Kraigolas said:

$1000 "upgrade" to my $300 CPU. 

It is kinda retarded, like telling a guy with a Toyota Camry to buy a maxed out Tesla if he wants to go green or move to electric.

[Kraigolas]

1 minute ago, YT_DomDaBomb20 said:

You tell me how accurate it is

Just to add a clarification to this post, I'm asking about the accuracy once you scroll down. Yes their algorithm for summarizing information at the top is clearly garbage now that I look at it, but every screenshot I've seen now is just showing that very top. It reports the Ryzen 7 1800X to be much faster in almost every category once you scroll down. I'm not asking about cherry picked data, yes it could be misleading to users, but in all honesty "Gaming 77%" means absolutely nothing to me, I look at the head to head scores when I scroll down the page, and that is what I'm asking about. 

[You_found_dom]

1 minute ago, Kraigolas said:

Just to add a clarification to this post, I'm asking about the accuracy once you scroll down. Yes their algorithm for summarizing information at the top is clearly garbage now that I look at it, but every screenshot I've seen now is just showing that very top. It reports the Ryzen 7 1800X to be much faster in almost every category once you scroll down. I'm not asking about cherry picked data, yes it could be misleading to users, but in all honesty "Gaming 77%" means absolutely nothing to me, I look at the head to head scores when I scroll down the page, and that is what I'm asking about. 

Oh gotcha

[Kraigolas]

Just now, YT_DomDaBomb20 said:

Oh gotcha

I do appreciate your response though, because that clearly displays how bad their summary algorithm is. The other examples were ambiguous. 

[Kraigolas]

2 minutes ago, BlueScope819 said:

If you look at this comparison:

https://cpu.userbenchmark.com/Compare/Intel-Core-i5-9600K-vs-AMD-Ryzen-5-3600/4031vs4040

You can see that the 3600 is better in every way except 9% quad core speed (which again is pointless as many games use 6 cores or even more) yet UB says that the i5 9600k is 13% better because of "memory latency" which as I said earlier is defined by the memory, not the processor.

I'm not sure what you mean, I clicked the link you sent and the Core i5 wins in everything EXCEPT the octascore speed. Memory latency, single, dual, quad core speeds are better for the Core i5, overclocked even more so. 

[Kraigolas]

6 minutes ago, BlueScope819 said:

You can't rate a processor for Gaming, Workstation, and Desktop without actually knowing what the workload is.

That's very true. I wanted this post to discuss more the accuracy of things like in game FPS (which isn't available for all head to head comparisons). 

[Kraigolas]

1 hour ago, comander said:

For the record, the i3 has a serious risk of generating lag spikes and the i9 can be expected to be silky smooth.

Unfortunately, PCPartPicker does not have a reference to this CPU, and Amazon only has about 10 (fake looking) reviews for it. However, Newegg has over 70 reviews, many of which are in-depth (and thus less likely to be bot spam) and they are all 3 eggs and up. Tom's Hardware benchmarks also show it putting up very solid FPS in several games. So 

Quote

It's grossly misleading 

Seems misleading in itself. Just by looking at the provided graphics on FPS and the article summary, the author does not mention your concern, and I found no references to it at all. 

 

Surely there are better processors, and the Tom's Hardware author even recommends against the Core i3-9350KF if you read through the article (although for gaming, it appears to be fine), but if you could please reference considerable evidence to back up your claim it would be greatly appreciated.