User Benchmark = Fake Benchmark

[humbug]

User Benchmark just updated its CPU ranking structure.

The CPU's ranking was based on 30% Single core performance, 60% Quad core performance and 10% multi core performance.

Now it ranks CPU's based on 40% Single core performance, 58% Quad core performance and 2% multi core performance.

This makes CPU's with more than 4 cores almost completely irrelevant in the rankings while heavily favouring low threaded performance.

User Benchmark are going back to 2003 ladies and gentleman.

Why they did this i'll leave you to decide.

I have my own view, User Benchmark lately has been used a lot to measure CPU performance, Intel very recently have been trying to 'encourage' reviewers away from actually real application performance reviewing, such as Blender and Maxcom Cinema 4D to what Intel call "real benchmarks" things just like User Bench.
And of course they cannot have AMD's very high core count CPU's showing them up in benchmarks like this so they removed that problem by 'working with people like User Benchmark' to 're-balance' how they rank performance.

I find Intel hilarious, no this sort of crap doesn't worry me at all because everyone can see whats going on and it's really not a good look for them.

On that note, Intel you-are-pathetic, you slimy ####-weasel, you are 4 or 5 times the size and the only way you can 'even the score' is not by innovating and engendering your way out of your current predicament but by moving the goal posts, Intel you Suck!

-humbug

https://old.reddit.com/r/Amd/comments/chal0r/psa_use_benchmarkcom_have_updated_their_cpu/

https://cpu.userbenchmark.com/Faq/What-is-the-effective-CPU-speed-index/55

 

[humbug]

I don’t see the problem, makes my 3570k relevant again

 

[humbug]

Something tells me you think Intel played a part in this change, I wonder why that could be?

hmmmmm.... ?

 

[humbug]

No that’s your assumptions.

I’m just saying it’s reflecting real world performance.

It doesn't, everything rendering, encoding, compiling... is multi-threaded, why do you think one of the aspects that defines a HEDT CPU is high core counts?

 

[humbug]

gg wp, was nice knowing ya, as relevant as pcper now

lol... so true, Pcper are dead, their live streams get like 32 views

 

[humbug]

Yes I know but they aren’t everyday consumer applications

I think i would mirror what muon said.

You think they did this because CPU's with 4 or more cores are less relevant now that they have been in the past.

I don't believe you think that, i think you're trying some seriously bizarre mental gymnastics to defend it.

 

[humbug]

I’m just suggesting their benchmark is reflecting real world applications/usage. I know that’s an unpopular view and it doesn’t support the ‘moar core’ majority on these boards but these boards are a microcosm.

Goes off to search for 3900x....:

The problem with it is, and i'll use the quote "you only need 4 cores" Intel's defence for a decade of 4 core CPU's, we now know how overpriced those "you only need 4 cores" CPU's were, not that all of us didn't at the time, this is Intel charging high end money for low end CPU's.

And they don't like having to compete again, this is why User Bench now prioritises "4 core CPU's" in the rankings, FFS stop singing to Intel's tunes.

 

[humbug]

humbug your 2 example apps are content creation, and its important to state even tho media reviewers and a fair part of the active OCUK base are content creators, the vast vast majority of global PC users are not.

Now is their new balance a good accurate reflection, no idea. But I do feel that the important of super heavy core counts has been hugely overstated by the media recently vs reality.

The majority of desktop software is still single threaded even in 2019.

Personally I have never used their cpu ranking charts, I just use them to help me look at specific comparisons (without weighting been applied to the tests).

Maybe they should start a serverbench website where its reversed with single core been 2% as that would reflect the server world and appease your desire for more cores.

There are far more content creators than gamers, even then we know games now use 8, some even more.

Let me ask you this, who is this benchmark for?

 

[humbug]

I was wondering why my Ryzen 3600 score went from 109% down to 93% with Userbenchmark.

penalised for having 6 cores, you only need 4.

 

[humbug]

I think the problem is some people worry too much about core numbers rather than how their software is performing. Does it actually matter if software is designed for four cores and we only have four cores or if it’s designed for eight and we have eight? As long as it works and it has been working. Just because there has been a space race in the core count doesn’t mean that what was working fine yesterday suddenly doesn’t work at all.

I suspect the vast majority of office workers would kill for a two core let alone a four core machine rather than some useless vm hub! The speed increase would be phenomenal.

Another thingy I consider which may be more relevant than core count is what platform the software was designed for, for example some adobe products just work better on intel (I’m sur s there’s a vice versa too!)

I think the problem is some people worry too much about core numbers rather than how their software is performing. Does it actually matter if software is designed for four cores and we only have four cores or if it’s designed for eight and we have eight?

This is a completely meaningless and empty statement given that most software will use how many cores the CPU has to offer.

Just because there has been a space race in the core count doesn’t mean that what was working fine yesterday suddenly doesn’t work at all.

Well the point is more cores = higher performance = faster work.

Another thingy I consider which may be more relevant than core count is what platform the software was designed for, for example some adobe products just work better on intel

Not anymore, Ryzen 3000 also does very well in those applications.

 

[humbug]

I have never ran the benchmark so I dont know how the app itself is affected I am talking about the website, so I am guessing the benchmark itself gives some kind of rating when its done and this rating is affected?

Its in the OP.

User Benchmark just updated its CPU ranking structure.

The CPU's ranking was based on 30% Single core performance, 60% Quad core performance and 10% multi core performance.

Now it ranks CPU's based on 40% Single core performance, 58% Quad core performance and 2% multi core performance.

So CPU's with more than 4 cores, those extra cores only account for 2% of the overall score.

 

[humbug]

Yes that statement to me wasnt obvious its about the app, as they use that metric to calculate the "effective speed" on their website and cpu ranking.

I agree that is just weird, there should be no quad core scoring, and all cores count equally on multi core testing.

In my experience with software you tend to get a ton of software that is single threaded, some software that is dual threaded, and then after that if more are supported its usually at least 8 but often more. I dont know of much software that is limited to specifically 4 threads.

The heaviest non gaming software on my devices is web browsers. Chrome the leader in the market will thread heavily if enough tabs are running. On the flip side nearly all my other desktop software is single threaded. Which for most of them is fine as multi threading is inefficient, and should only be used when a single thread would bottleneck as is clearly the case in a multimedia web browser.

So I think quad core shouldnt even be a metric, in terms of single vs multi threaded, I think that should be user customisable for the weighting. On my main rig I would probably weight it at 20-30% multi and rest single if I had the choice. On my ryzen it would be 100% multi.

I would question what your software experience is if its not beyond Dual core, with respect my computer is stuffed with software designed to use as many threads that are available, i have a hobby in game development which involves multiple disciplines in 2D / 3D rendering, Ray traced shading / light baking, various forms of compiling, just Game development i have 10's of applications at my disposal all of which would push HEDT CPU's for all they are worth.

Even User Bench with their old method of awarding CPU's with more than 4 cores just 10% of the overall score was outdated and "very strange" 'updating it to awarding above 4 core performance just 2% of the overall score is not just strange its extremely suspect.

Why have it fractionalise its ranking in this strange way in the first place?, if you look at real world applications that have a performance benchmarking tools built in, Like Blender, 7Zip, Maxcom Cinema 4D, Handbreak.... they don't distinguish their rankings to however many cores the CPU tested has, it just ranks them based on the performance they have, because that's the reality.

User Bench have been influenced by "four cores are all you need" Intel for a long time.

 

[humbug]

As I said pretend you a typical user not a content creator, exclude all your apps related to content creation, so typically apps that encode or decode. Then look at how many threads get used.

So e.g.

MS paint single threaded
VLC player single threaded (the app itself supports multi threaded but it isnt default operating mode, and most decoders for it are single threaded)
Microsoft office single threaded
Notepad single threaded
Windows media player single threaded.
Most a/v single threaded. This can be very frustrating doing a virus scan and seeing its clearly bottlenecked by one core.
Most small apps like ftp clients, ssh clients, hwinfo, afterburner etc. single threaded.

The exceptions tend to be content creation and archive software.

Now there is nothing wrong with needing all those threads, in your case you clearly do. However you seem to be assuming that your use case is relevant the PC userbase as a whole.

Humbug you do seem emotionally attached to AMD given your posts on them, its good they released these cpus that are good for your workload, just be happy with that

All your listed applications can be run equally well on Windows Tablet, or any Tablet given most of that stuff is available for Android and Apple.

Intel also want reviewers to stop reviewing CPU's by benchmarking them in high stress real world applications and instead benchmark web browsers and media players, would VLC and FireFox run any better on an i9 9980XE vs a Celeron G4900?

No!

So what is the Bloody point? Intel have lost the performance crown, so now it wants to change the performance testing landscape, you know what i read in that? Intel think they can't compete with AMD going forward.

 

[humbug]

All your listed applications can be run equally well on Windows Tablet, or any Tablet given most of that stuff is available for Android and Apple.

Intel also want reviewers to stop reviewing CPU's by benchmarking them in high stress real world applications and instead benchmark web browsers and media players, would VLC and FireFox run any better on an i9 9980XE vs a Celeron G4900?

No!

So what is the Bloody point? Intel have lost the performance crown, so now it wants to change the performance testing landscape, you know what i read in that? Intel think they can't compete with AMD going forward.

This is the real world.

 

[humbug]

I think I will end my part in here now, humbug you do seem obsessed with intel man, all the best

I've had multiple Intel CPU's, Pentium 2, 3 and 4 HT, i7 930, 4690K, multiple AMD CPU's, Athlon 3800+, Athlon 64 X2 5800+, Phenom II X6 1090T, FX 8350, FX 9590, Ryzen 1600, Ryzen 3600.

I don't care any more for Intel than i do AMD, i just care what i get for my money, and sometimes curiosity.

The irony is AMD are ahead of Intel in IPC.

Indeed.

 

[humbug]

I wouldn't blame intel individually for this status quo. This is an industry-wide issue and it's above intel - it's probably coming from the world government.
AMD is in the game, too. At the very least, because they have agreedments with intel.
AMD is the underdog which behaves very strangely - when their market share approaches 50%, they behave as if they have 80 or 90%.
Intel, on the other hand, when have 50% market share, behave like crazy and start releasing very major performance upgrades ala Core 2 Duo for desktop.

What about the industry-wide lack of adoption of Ultra HD monitors?

The world government wants from you to have office computers with dual-core processors and Full HD monitors.
And stay online using your smartphone, which can reach 800 pixels per inch pixel density, which is insane overkill and not needed.

I know my post essentially amounts to an accusation of conspiracy, but its far from without merit, Intel has been found guilty of exactly this type of anti trust conspiracy all over the world, many times, including in the US.

IMO your conspiracies have no reconcilable history to validate, they are just tin foil, a fist full of conspiracy too far.

 

[humbug]

It might not be that it matters more before, just that their view on it has changed (i.e. they recognised that they were previously underestimating the importance of single-core performance). So maybe they thought **** we should have previously put 50% on single core instead of 30%, but now due to changing software demands and increasing importance of multi-core, 40% is more appropriate.

Why have 4 core and 4 core+ separate at all?

You would only do that if in your results table you wanted to put more enfaces on 4 core or lower CPU's by giving them an artificial boost up the rankings, this seems like a very strange thing to do and what would be the reason for it?

Why wouldn't you, as everyone else does, just say your per core performance is X, you CPU has multiple cores and your performance is Y.

As it currently stands what its saying is 4 core CPU's are better than CPU's with more than that. "you only need 4 cores" -Intel

 

[humbug]

God already punishes them. They can't make anything 10nm, nevermind 7nm and lower
In the end, only bankruptcy awaits them.

But it won't be bad - there are the VIA, ARM, AMD, nvidia to join efforts for our future wellness.

I like this idea unfortunately there would have to be a complete shift in software development for that to happen and we all know how lazy and reluctant to change software developers are, everything is X86_64, Intel owns the X86 part of that, AMD the X64 part and they ain't licensing those to nVidia or ARM, VIA who do have access to X86 don't have access to X64, so that's windows out of the question... and Linux is slowly going AMD64 exclusively too.

 

[humbug]

On your Intel system navigate to C:\Windows\WinSxS

 

[humbug]

Their explanation to be fair is reasonable (yes I responded back but not to humbug directly with his anti intel crusade).

So the reasoning is that cpu's which have very bad per core performance but massive core counts were been rated higher than cpu's with good per core performance and much less cores, and they felt this was misleading for people looking for cpu's that are good for gaming, I think that as a motivation is reasonable. (per core performance across the entire PC gaming library is still king, some of us forget that) Also to be fair when discounting the bad heat/power issues, a 9900k is a better gaming cpu than a ryzen 3000 chip.

However they should point out that they are rating cpus for gaming (Which they currently do not). Most of the ryzen 3000 chips seemingly got promoted when they changed the weighting system. The cpu's that suffered from AMD were the threadripper chips. Also intel chips with high core counts and low clock speeds (server chips) got demoted as well.

They also seem to have acknowledged they perhaps went too extreme and seem open to another adjustment later down the line.

The site is excellent as a whole in my view, a database supplied with user ran benchmarks. Providing data for different metrics of performance. I consider the data from here vastly better than the likes of unboxed hardware who massively over emphasise multi core and content creation. They also have a habit of not testing low end hardware enough and is lots of low end hardware on userbench.

But there is always the problem that there is different "types" of users. So I think the next logical step is either the customisable weighting system or add a use type category e.g. "server" "content creator" "gamer" "streamer" "office productivity" "casual user". Then weight based on what is selected. There is also the AMD/intel fanboys which sadly nothing can be done for those people, they will only be happy when their favourite brand is on top.

I suppose you could argue userbench is fighting the corner for people like me who think the importance of per core performance is grossly under stated by most of the media, its clearly an unpopular move as people have been manipulated so heavily by their favourite youtubers and co, but there is nothing wrong with a website not following like a sheep. I also stand by what I said earlier I use my own opinion and concentrate on using these sites for data, I just happen to accept this explanation as rationale.

And a 3600 is a better gaming CPU than a 9600K, or an 8600K, it has 12 threads vs 6 to better cope with highly threaded modern games, even if the 6 core Intel gets higher averages in these games the 3600 is smoother because with 12 threads its under far less stress.

The "you only need 4 cores" trope didn't wash when Intel made that excuse when selling low end CPU's for high end prices, it's even worse when you repeat it now and the same goes for User Bench now using Intel's ridiculous marketing excuses.

You don't believe that crap any more than i do @chrcoluk and simply dismissing people who point out the flaw in that reasoning as "AMD fanboys" is typical of Intel shills who know they have lost the argument, you know you're wrong, typically its the thing people like you always reach for, these days everyone is an AMD fanboi, because these days Intel have no responce for the arse kicking they are getting.