Discussing UserBenchmark's Dodgy CPU Weighting Changes
ฝัง
- เผยแพร่เมื่อ 5 ต.ค. 2024
- Support us on Patreon: / hardwareunboxed
Intel Core i5-9400F: amzn.to/2LKUDoy
AMD Ryzen 5 3600: amzn.to/2NBu0nT
Intel Core i5-9600K: amzn.to/2ygv9H0
AMD Ryzen 5 3600X: amzn.to/2S2gKap
Intel Core i7-8700K: amzn.to/2LMyDK0
AMD Ryzen 7 3700X: amzn.to/2FYuCh0
Intel Core i7-9700K: amzn.to/2JXcC9c
AMD Ryzen 7 3800X: amzn.to/2FYvja9
Intel Core i9-9900K: amzn.to/32VHBdg
AMD Ryzen 9 3900X: amzn.to/2FYuttY
Discussing UserBenchmark's CPU Speed Index
Disclaimer: Any pricing information shown or mentioned in this video was accurate at the time of video production, and may have since changed
Disclosure: As an Amazon Associate we earn from qualifying purchases. We may also earn a commission on some sales made through other store links
FOLLOW US IN THESE PLACES FOR UPDATES
Twitter - / hardwareunboxed
Facebook - / hardwareunboxed
Instagram - / hardwareunboxed
Music By: / lakeyinspired
This was meant to be a quick 15 mins max discussion… that didn’t happen but hopefully those who wanted us to discuss this issue are satisfied now.
Also the UserBenchmark.com response discussed at 6:30 has been updated, this happened when filming so we read the original statement. While editing I was able to lay in the updated version, so you can see the changes on screen, basically they’re saying the update only downgraded the 3900X by two spots while upgrading the rest of the range quite significantly. I don’t know where the 3900X ranked before the -2 change, it would be interesting to know if it was ahead of the 9900K.
In any case, Tim’s point stands, even before the changes the weighting system didn’t seem to make sense.
Okay
This whole situation is such a fiasco
The 3900X was ahead of the 9900K before the update. The top CPU was the 2990WX, then we had an Intel 18 core part and then one more Intel HEDT part and then the 3900X.
I understand maybe addressing the 2990WX, but while that CPU has its abnormalities, it is a serious product. I would not punish it like that.
@@aurorap0laris430 you know, intel payed them most likely
@@aurorap0laris430 its just started whan AMD is better competition will pay to this Garbage sites and other TH-camrs..
Userbenchmark: "Ryzen 3000 is too fast so we'll adjust our algorithm so that Intel has a chance to compete." 🤣 Yea, seems fair.
bgtubber Ryzen is too good honestly. Price to performance is ridiculous.
Positive discrimination in tech world? :D
@@johnnyggun7141, Affirmative Action.
That's really an opposite case of Nvidia's 3DMark cheating in the past
@John Doe Bro go get those eyes tested.
"MC Mixed average bench - 14%"
"MC Mixed peak OC bench - 8%
"MC Int - 6%"
"MC Float - 7%"
Every benchmark I've found online (from various tech review sources) puts the 3600 10%-30% ahead of the 2600, use-case depending.
Professional software engineer here, I haven't written a pure single thread application since 2005. Their excuse is bs.
Same. who writes a single threaded application. I guess hello world would count lol
@@woodiemarv I find that offensive, my super tiny scripts hardly even use a full thread lol
@@iiiiii7680 lol
@woodiemarv There is a big difference between a single threaded application and a multithreaded application that is nontheless limited by the mainthread.
Amdahls Law is a thing. 100% scaling isnt a thing for multithreaded application in a lot of cases. You still have to wait for locks and signals.
You can offload AI, various checks and so on, but sometimes you just cant split it further, because of dependencies.
Using 8 Threads is fine, but if 7 are at 30% utiliziation and the mainthread at 100%, u´re still bottlenecked by single core performance.
The strategy game stellaris is a prime example for that, or factorio. In Factorios case the developer tried to multithread more heavily, performance got worse.
There wont be a jump in the next 4 years to games using 24 threads effectively, especially when over 80%! are still gaming on a 4 Core or 2 Core CPU (steam hardwaresurvey).
I like HardwareUnboxed but that argument was bs.
Respect for straightforwardness...
UserBenchmark makes people with i3’s happy
Intel don't really care if an i3 is at the top places of the Userbenchmark charts. What's important is for AMD to be lower. 😉
Tech-Justice -Warriors
@@tahiribnmohammad5410 Trans-istor rights?
Nah... Trans rights are human rights!
yeah especially since they're now faster than i9s according to userbenchmark xDD
@@bgtubber a quad core scores higher then one of intels low end xeons. Its clearly no longer a viable way to score cpus. And intel is no longer better then amd its pretty unanimous. The 5 to 10 fps increase is not worth the workload increase. Sorry mate intels only gonna get worse when they switch to the 7nm platform in 2021
Just as additional information:
When Ryzen 1000 was released in 2017 they lowered the multicore weighing from 20% to 10%. With the release of Ryzen 3000 it was again lowered to 2%.
So over the course of 2 years the multicore importance was reduced to a 10th of what it used to be.
20%->10%->2%, each reduction (in)conveniently happening at the launch of a Ryzen generation.
@Fellow accelerationist No it looks bad all over.... ;)
@Fellow accelerationist Your CPU that is. Don´t worry I´ve got one myself, just bating a little. :)
@Fellow accelerationist Holy shit you're thick!
Intel to cheat will instead have a 2 core CPU that draws 300W with a ton of blank space and focuses on single core performance
Intel spending more budget on spreading AMD rumors than actually finishing their 10nm cpu
G Kim hahaha agreed
Intel has invested 22 billion into fab upgrades. But that still leaves them with enough in their warchest to spend more on something else.
PC World did a great video on the 3900x and 9900k head to head and it was mentioned how Intel spends millions of dollars to send literally thousands of engineers out to software developers to help them optimize their software for Intel CPUs, so often times the incentive to optimize for AMD hardware is just not there. I was not shocked to hear what was said, but it did catch me off guard a bit.
I think Intel won't be happy people are buying "superior" i3 over the i9 they want to sell you
Jeff Lee they’d still be buying intel tho
Userbenchmark is a homeopatic performance database. The less cores, the better the scores.
It's ridiculous. According to their own score-weighing algorithm, an Intel Core i3-10300 outperforms an i9-10980XE by 3% in terms of effective speed.
And I thought Core i9s were supposed to be much faster than Core i3s?
Beyond the obvious "r5 1600 vs i5 7600k" comparison making their entire argument moot, let's just ignore this for the sake of argument. How could, when CPUs are gaining cores on all sides, "keeping up with the times" result in them bringing multi-thread down? Are they suggesting that multi-core utilisation has regressed?
Yeah no idea how that one works mate ;)
Aren’t we supposed to turn multi-thread off when using Intel?
@@Kathdath That's why they said you only need more than four cores if you have a virus. An excuse so bizarre it makes my head hurt.
They should just delete the "effective speed" tab entirely. One can't measure a complex work of engineering, like a CPU, in a single dimension.
translation = "Buy Intel until they get their chiplet design down then we will raise the multi-core weighting"
AMD : Let's release a bunch of awesome cpus for a good price.
Intel : Lets pay off places to make our stuff look better even if it's obvious.
Precisely
If someone listen to Intel they already said boosting marketing effort which mean deception effort. Userbenchmark is one of them.... yuck! who will Intel go next ?
*Princibled Technolgies* cough
It is apparently cheaper to pay people to say their product is good than it is to actually do their job and make good CPUs.
@Dilet no you will get a(m4) 2500X, it's eXtreme
Userbenchmark is pretty damn silly considering an i3 8100 can beat a 16 core i9 just because of the single core speed
i3 8350K*
AND they even doubled down and said it did offer a similar performance 🤣
@@shernandez31
Just edited and checked with something else, the 9980xe is rated lower than an i5 6600k LOL
Not silly when they are counting cash now.
@@blvk3
I don't see how intel could be paying them off honestly since the new scoring system makes all the intel HEDT CPUs look worse than they actually are
You might actually want to check your facts. They have the 16 core i9 7960x at 26th and the 8100 at 94th. 26 isnt greater than 94 genius. In each of their metrics the i9 wins, by nearly 20% in gaming and well over 110% in workstation.
IF you want to buy a new CPU to play old games, have we got a site for you!
J.D. Martin Like CSGO or dota or LOL?
@Βασίλειος Μπεσλεμές nah
@@chambers5945 DCS World? X-plane 11? S.T.A.L.K.E.R. with tons of mods, which looks better then AAA games in 2020? Why you all people know only the same shit like CSGO or dota?
@@Directx11cool not everyone plays flight sims?
🤣
Time to dust off my 2500k, Apparently it is 15% better than the 3600x...
Teh509 hahahahahahahahahahahahahahahahahaha
Nice 😜🤣😝😉🤪
cpu.userbenchmark.com/Compare/Intel-Core-i5-2500K-vs-AMD-Ryzen-5-3600X/619vs4041 says 90% faster?
I mean the 3600x is 90% faster.
@@larrybishop6622 the 2500k pretty easily doubles it's speed with overclocking, from 3.3ghz it's pretty easy to hit 4.8 to 5ghz.
@@larrybishop6622 Twas a jest.
Damn. I just found userbenchmark today and thought what a great tool this was, until I found this video on userbenchmark's about page, and it is so clear that they really have sold out, and then libellously called Hardware Unboxed 'Incompetent Smearers'. I'm so gutted I've just handed over my machine and user data to this fraudulent website, for anyone else, avoid at all cost.
As a software developer that also happens to game and love PC hardware, I feel the statement saying that it's difficult to scale software to multi-core is BS. Nowadays, it is no longer complicated to utilize all cores, as multi-core programming has had decades to mature. Programming languages have emerged and/or adapted (Go, Scala, even Java) to support and encourage independent processing where programmers no longer need to deal with low-level concepts of multi-core programming (locks, threads, shared memory access, etc), and therefore even beginners can easily write programs that scale across many cores. And for the programs that are single-core only, it's relatively simple to run multiple instances of said programs, say in docker containers or via supervisord in unix.
Bernie Eng I can just use sklearn and set “njobs” to -1. OMFG I just made it multi threaded!
Yes I agree, I'm a developer of over 25 years. Writing Multi-threading applications is a solved problem. Every coder is able to do this easily. So many resources available to new coders to learn the tricks needed. New applications are not single threaded anymore.
@Bernie Eng & @Richard Collins
> Writing Multi-threading applications is a solved problem
So Amdahl's Law is no longer valid?
en.wikipedia.org/wiki/Amdahl%27s_law
> to support and encourage independent processing
Independent processing isn't always possible. Often it is, but it doesn't scale.
> low-level concepts of multi-core programming (locks, threads, shared memory access, etc)
But locks, threads... are still used. You may have higher level abstractions for it, but the problem of protected regions or threads waiting for another are still there. The underlying problems of parallel programming remain the same, no matter what your programming language or runtime may offer you to reduce the pain of parallel programming.
@@shawnpitman876 and you are forgetting the one issue with considering EXTREMELY lazy coders that if they don't update with time and new tech especially in the case of the bigger and new emerging companies , they will eventually be replaced by newer generation coders or may be not get promoted for future generation tech development.
@@shawnpitman876 it will happen and consumers will force it, up untill now and a few few years back multicores was not a mainstream hardware , but now corporations and professionals are adopting it, not only on hardware side as well on software. I am an architecture student and have completed my 1 year internship . I will take my own field as a LIVE example. our industry was focused on using Autocad software for drawings which is a very very heavy single threaded software and in 2019 its really pathetic to work with all the time constraints that we have for completing the project. there is other software from same company(Autodesk) i.e. Revit. a intelligent bim software that is heavy both on core clock and multi threads heavy . the efficiency of working on these types of software is soooo miles ahead the a single person can do better work in way less time , that is why more and more architecture firms are moving toward using revit as their platform and replacing their workforce. the returns of these investment is much higher. I know both softwares and prefer revit myself. NOW a software like revit definitely cannot perform on lower cores they need many many cores and threads and heavy core performance. the result is that although revit is much new software and autocad was developed well before still the company is now investing much more on develment of their revit software . SO IT'S DEFINITELY HAPPENING and not slowly because people are moving to newer and better products.
That villainous intent you mentioned sounds way to accurate not to be true. Why would they choose 2% it's such a random number, until you look at the difference between the 9900k and the 3900x and see that at 8% difference in a category could give the 9990k a bump back up to first.
As a newcomer to the PC world I really appreciate the work you guys put into getting information to the consumers.
the MasterRace welcomes with its many many cores.
Agreed, i learned alot from this channel.
Consider Patreon support then. Thanks.
Yes! This is one of my favorite videos they've done recently!
I looked at that website a couple of days ago and I ended up getting a Pentium II. Why isn't my PC going as fast as they said it should?? ☹
F
You should only run one task at a time on your Intel CPU, that way it will be faster than AMD. So, if you are running windows, then don't run anything else, and be happy knowing its faster than an AMD CPU, under such circumstances.
I'll check the scores...
Wait, why are they claiming that a Pentium II 266 is faster than a Core i5-3470?
a pentinum ii is from the late 90's.
Lol according to this new benchmark, i3 9350k will beat any Ryzen.
cpu.userbenchmark.com/Compare/Intel-Core-i3-9350KF-vs-AMD-Ryzen-5-3600/m775825vs4040
dodge af
@@24wherath36 LOL
Someone at UserBenchmark is moving to Intel...
This one's my favourite:
cpu.userbenchmark.com/Compare/Intel-Core-i3-9350KF-vs-Intel-Core-i9-9980XE/m775825vsm652504
You can go to 3dnews_ru CPU reviews and see exactly the same.
Magic )))
Here's a line from their 5800X3D review. I don't understand why anything like this would be put in any professional review of a product? "Also watch out for AMD’s army of Neanderthal social media accounts on reddit, forums and youtube, they will be singing their own praises as usual."
Ryzen releases new CPUs that compete with Intel, then UserBenchmark changes their weighting to be even MORE single-core heavy. Hmmmmmmmm.....
*intel adds HU to their black list... nvidia nods in approval* "eeeeeeeexcellent Smithers"
Why though?
@@olivur_1459 nvidia already has HU on their blacklist of free thinkers
Why the perceived "alliance" between Nvidia and Intel against AMD? Nvidia sees Intel as another potential threat
@@XZagatoX no talk here about an alliance. just approval of certain practises that have nothing to do with amd really.
if you as a tech reviewer dont say - its Great, just get it - giving a honest review of their gpus, nvidia blacklists you. which has happened already. so joke was, nvidia giving approval of such conduct
Intel's dirty tricks won't save them this time
And people said that intel wouldn't resort to dirty tactics again... lolz.
Not sure anyone said that but I know what you mean 😉
@@N1rOx lol...Moores Law is Dead mentioned that Intel would not be able to get away with 'dirty tactics' again. I think you can find the video with the channel name and the exact term. This is pretty dirty though
@@shawnpitman876
sad
Not at all, bud. Not at all.
@@shawnpitman876 wow , how triggered you are by success.
Reccommended to me after watching 2kliksphilip's on the topic.
Wow, they haven't changed at all, in fact they became worse than ever last year in their reviews of Ryzen 5000 series.
Kudos UB, please never change so we can have someone with dependency on one brand only :)
User benchmark should have its financials checked and if they've took money from Intel they should be sued in court
Thare are ways to make it look nice.
Like Intel pays editor and his family for coming to Barcelona for some IT event.
So kind of them.
There is nothing illegal about being paid to lie
Nothing wrong with taking money from someone.
And to be honest that would'nt help.
Instead they could be forced to correct the 'errors'.
Pay back whatever money they got.
And then thru media apologise.
That would correct things a bit i think. (If not give them crap so they closed the crap down)
@@DanielFrost79 You guys ever realize the people who are extreme to one side or another can't type properly? Just a thought.
If Intel actually payed cash to userbenchmark, then it's to late, Intel just got done being fined 1.5 billion dollars, if they did more sketchy stuff like this, they would do anything to hide it.
No doubt if Intel payed anything, they have NDA agreements out the back end, and they wouldn't hesitate to shutdown a little website like userbenchmark, long before the beans where spilled, and the site reversed.
So no you're wrong, but in all likelihood no actual money changed hands.
userbenchmark went the way of CPU-Z and became unreliable trash as a benchmark.
Again and again, Hardware Unboxed continues to blow me away with their videos showing their insights and of course bringing me to Benchmark Valhalla. Good job guys!
@Hardware Unboxed
Wow, did you see the About page of Userbenchmark. They are calling you "Brazen shills" and "Incompetent smearers", linking to this video. 🥴
Seems like Farcry is now Intel's favourite gaming benchmark :D
What's funny is that it's an AMD sponsored title. I guess they only focused on optimizing for AMD GPU's, not CPU's.
gary tallowin I think far cry is sponsored by amd
Far cry 4 was sponsored by nvidia tho
I got that joke. For once, something didn't go over my head.
@@champnotchicken4318 insert *that's the joke meme* here
When you see that an i3 is better (albeit by 2%) in the score than a 2700X, you know there is something really wrong with the way they score...
I'm guessing they will say that i3 is still 2% faster than a 64 core Epyc Rome! Sheesh.
Datacenters should be powered by i3s according to UB
@@Wahinies, UB meaning Useless Benchmarks?
@@vsevolodalipov4375 a.k.a. Uber Bullshit ;)
@@Wahinies too good a word for ubm.
They do have the separate workstation scoring, seems decent.
AMD is starting to make gains? Our algorithm must be weighted wrong...
What I don't get is why people neglect the fact that even if an app is single or low core count, at this time of age almost NO ONE runs one app at a time all the time. Even gamers have apps/utilities and os running concurrently in the background while gaming (maybe not the hardcore ones but we are talking about the average here).
Some examples:
No software developer has only their IDE/Editor running, but concurrently have a couple of VMs for testing (different OS) as well as compiling, web browser, utilities, collaboration tools, pdf, etc and same goes for web developers that also run multiple server/browsers for extra testing.
Graphic designers/3d designers run a 3d and/or 2d design program along with photo editing program as well as graphic libraries, multiple browser tabs, printing utilities, and other software based on the task, etc.
Musicians have sound editors open along with their DAW as well as software for libraries too and even broadcast software. Daw itself is (or can be) multithreaded especially when you involve a lot of plugin effects.
Office workers have multiple office software apps running with multiple documents/database/spreadsheets open at the same time as well as their company's dedicated software, browser (for chat, if anything else).
Even casual users have multiple browser tabs, as well as communications software for voice/chat and a lot of the above cases (as well as others) also have an antivirus/antimalware running as well as other OS tasks like downloading (movies, software... anything).
The notion that people ONLY run one app (not even the OS?) and that app is single core/thread is really not representative of the current state of computers.
This is not a true representative of real life cases too. Chances are you never really hit 100% cpu usage for casual users
@@Centrioless It's not about hitting 100% CPU usage or not. It's about that in the current state of computers, even tho few single core/thread apps may still exist, very few people run ONLY one such app or doing one task at any given time. So if people run few (or several) tasks at the same time, even if those tasks are single-core/thread tasks, multicore CPUs makes the situation better.
When one compress files, for example, one just doesn't look at the screen waiting for the task to finish (specially if the files we compress are several GBs) but rather lets compression run in the background while continuing to work, or do whatever else like browse the internet which uses different cores on our CPUs.
In reality, we often execute multiple, single tasks at the same time which makes the multicore part of the CPUs valuable and as such I think that multicore should "weight" more than the single-core performance.
From my experience, almost no one is executing a single task only, even tho they seem they do. Even multiple tabs on a single browser are executed in different threads and so benefit from multicore/multithread CPUs even tho they seem like a single task.
Now occasionally someone somewhere might do that, (run only one task) but those I believe are exceptions to the general rule.
@@Centrioless Unless you have a quad core and try to play modern games :)
I hope the tech community can create enough bad publicity that their business suffers because of this.
As it should be. What they're doing es extremely shady.
Yeah... wont ever happen
The site is too big
@@iiiiii7680 ok UBM shill
4 years later and their reputation is in tatters. No one trust them any more.
This isn't the first time a benchmarking algorithm was adjusted to be more "correct". CPU-Z also adjusted their benchmarking tool shortly after 1st gen Ryzen released in 2017 because Intel CPUs were losing to Ryzen. They also quickly changed the algorithm to be "more correct". 🧐
Cpu-z was adjusted?
@@BudgetGamerz Yup. Do a web search with the terms "CPU-Z: new benchmark, new scores CPUID". It's an article from May 1, 2017 in their website. Their explanation is something along the lines of "Our benchmark up until Ryzen was perfectly fine, but then Ryzen came out and we realized the benchmark was actually sh*t and unrealistic so we changed the algorithm so that Ryzen can perform equally as worse as Intel. But that's fine because our benchmark is more realistic now that Ryzen is slower in it, we promise!".
userbenchmark before ryzen 1000 had multycore at 20% one month after first ryzen launch its went down to 10% and now 2%
@@razoo911 Really? Wow, I didn't know that. At this rate, next Ryzen release it's going to be -10% 😂Meaning, if your CPU has more cores/threads, you'll actually be penalized for it. Craziness. 🤡
@@bgtubber the problem is also they give too much score at clock speed and overclock headroom instead ipc and single thread then in my test seems for single thread they chose everytime less fast core clock on ryzen then viceversa
HOW ABOUT
25% singgle core, 25% quad core, 25% octa core, 25% multi core
@@Tw1stTech well if they actually implement our ideas, it would be a great site
I'd suggest something along the lines of 25% Single core, 45% quad core, 20% octacore, 10% multicore. It'd be more realistic for real life.
I really don't like being THAT guy, but...
INTEL is getting hammered even harder with the ZEN 2 Mjolnir, so what do "independent" outlets do? Change the rules so that INTEL looks better.
There...
EDIT: Yes, saying that there's even a chance in hell that they changed the rules NOT trying to make INTEL look better was supposed to be ironic, because OF COURSE THEY DID!
... it hurt intel more than amd. It makes intel's 100 dollar i3 look better than their several thousand dollar cpu. Amd does t sell shit that expensive lel
Something that bothers me is how they talk about "cores".
It's not just about cores anime, it's also about "threads", right? Even when their system compares CPUs with the same amount of "cores" the Ryzen CPU is punished because it has more "threads", when in the reality is the absolute contrary. This even affects Intel CPUs, making i7 get lower scores than i5 from the same generation!
@@shawnpitman876 So where is the AMD consumer chip that price matches a 9990XE?
@@shawnpitman876 And does AMD have any auction only chip? Or are there anything AMD makes that price matches a 9980XE even? Or if we're doing all chips, where is that Xeon Platinum 9282 price match equivalent?
@@shawnpitman876 It must feel good to be blindly aggressive without reason like that. What does "Dumb Intel cuck" even mean? For a given core count with SMT features, Intel options are almost always more expensive, thus this UserBenchmark change effects cheaper AMD processors the same as it effect more important Intel ones, thus it harms Intel more. Literally the entire stack segmented by core count sees AMD cheaper. How about you name one core count where Intel has it cheaper?
Suddenly my i7 2600k is awesome once again!
Nothing like regress
The score on my 2600 non k went down, lol. It's at 65% down from 73% effective speed.
You know that it's a special video when Tim and Steve are in the same place.
28:58 "does nothing"
I never investigated their algorithms previously, however, I no longer am going to use their site. Although now, I won't even treat it like the wikipedia-esque source that I was treating it as for much the reasons pointed out here.
Good expose djents, got my sub.
i would rather believe Wccftech instead of UserBenchmark
I'll take that as a compliment I guess, lol. I did do a write up on the situation several days ago :P
@@WccftechTv I don't know how linked you are to the Wccftech website, but the website is a joke. I don't remember the last time they got one of their "credible source leaked" rumours correct. They just make stuff up based on what sounds good. Their "sources" said the gtx 1080ti was coming out on 4 different occasions before it actually did. No one should trust anything written on that website.
@@TheHalfGlassFullGuy There's a difference between reporting on a leak or rumor and an exclusive leak but our latest was regarding Frank Azor joining AMD since we reported on that about a week before it was announced :)
@@WccftechTv You've had plenty of "exclusive" leaks that have been flat out false. As I said, if you bothered to actually read my comment, you've had many articles where you have leaks from "reliable insider sources" that have been off by a long shot. This is an extremely common theme on your website, and many other similar websites to yours. Take a leaf out of Hardware Unboxed's page, and do some research before publishing.
@@TheHalfGlassFullGuy 👍
At the very least this gives the appearance intentionally favoring Intel. At worst they actually are.
And I'm s some one with three 9900Ks, though an AMD build is in the works. This move does not encourage my goodwill towards the site.
More to the point it's clearly driven by Intel so your ill will would be better directed to that dodgy mob and how much they are paying this site to make the change. Filling someone's back pockets for sure
+632% value, -2% speed. KEK
🤡🤡👌
@@A.Froster It's important to be patient for good days catch...
@@m.streicher8286 especially into this word
hello fren.
@Fellow accelerationist Based.
An i3-9350k beats almost all of Intel's and amd's products, it beats the 2990wx by like 11%.
You tell me if you think this is good.
The only Intel CPUs the i3 doesn't beat are the more expensive 14nm CPUs of the same class. (Duh)
Userbenchmark? More like use-LESS-benchmark.
ObViOuSlY iT's GoOd, We NeEd To GiVe InTeL a ChAnCe!
Their new "review" of the 11900k is just pure comedy
Userbenchmark is the modern day equivalent of Sysmark, where they repeated the tests that favored Intel CPUs multiple times to show them in a better light (Intel paid Bapco to do this, which may or may not be the case with Userbenchmark)...
This is a strange time to up the emphasis of single core performance over quad and "multiple core" performance, so I would say that it smells like intel money (obviously I may very well be wrong, but you know, the company has a pattern of trying to manipulate benchmarks at the source, they do/did the same with their compiler, ensuring AMD's CPUs used the slow path by not using extended instructions when they could on their products).
Times don't change.
On a side note, the way they mitigate the delta of results on a given CPU is that they show a nice graph of the results for your CPU below the overall number, this way you can see how you stand against all other 9900k (it's a nice way to check if your overclock is good or great, or if you throttle/have some performance issue, which is a great use for the site).
USerbenchmark is actually good if you view the individual results. The Ranking is terrible and I did not know people actually unironically use it, when 3 centimeters under it you have the breakdowns of the CPUs in a more logical, easy to understand manner. I will still use it and will just continue ignoring their shitty ranking as its retarded.
We came to the same conclusion after about 30 mins ;)
@@Hardwareunboxed We agree. I am still watching it. Even at 2x speed it takes some time.
@@CharcharoExplorer you have to remember that people that arent as much of a tech enthusiast, they may just search their cpu vs (insert cpu) then go to userbenchmark and look at the ranking.
@@calebjames7330 I know. That is why I want it changed, but my point is it isnt useless even for an enthusiast, as long as they keep a few things in mind.
10% for Single-Core (niche productivity and emulation), 30% for dual-core (general desktop and windows/linux use, old games), 40% for 6-core (2019 gaming, to be changed to 8 core in 2020), and 20% for multi-core would actually be a better representation of modern CPUs, overall.
@@CharcharoExplorer They would be better off with simply two parallel columns ranking gaming and workstation separately.
lmao the i3-9350K is ranked above the i9-9980XE with their current ranking system.
@@rickylonghaul682 Which is okay to them, because the type of person using their site isn't going o be comparing a i3-9350k against a 9980XE. They will be comparing it to something on a similar price point, such as a R5 2600. That there is collateral damage in their slewed data is meaningless if it's not really going to effect the audience that would come to the site thinking it is legitimate.
AMD will overcome even single/quad core performance at Ryzen 4th Gen. What will userbenchmark do at that time?
They already did. Ryzen 3000 has higher IPC than intel.
Destination @#$/
Nerf Ryzen is some other bullshit way. What else can they do ?
inb4 more cores = worse
Introduce am penalty for core performance if not run on intel procesor.
UFD tech did a test 1 fast core vs multi slow ones. Just booting win 10 was painfully slow on one core, multi slow cores was still slower than stock but it worked ok. Its was last year sometime if anyone wants to look it up. So yes single core is becoming less important tgese days vs more cores and threads.
Wow another long discussion and analysis video from HUB. I love you guys.
It’s confirmed user benchmarks gets paid by intel
Sources?
@@SconVideos common sense?
common sense != confirmed.
@@MephistoDerPudel I didnt say that, but its bloody obvious. occams razor is unforgiving lol
@@SconVideos It was a joke
Should be
25% single 50% quad 25% multi
Or even
25% single 40% quad 35% multi
Yeah multi should be higher percentaged, gaming is now becoming more multi core and anything other than gaming also uses more cores...
Just get rid of "quad"
@@paul2609 exactly, and more multithreaded %
I'm on the side of userbenchmark, because:
-Most of the users are going there to compare gaming performance.
-In the wast majority of cases, you will not notice any major difference in gaming performance, as long as you have a 4 core sandy bridge or newer CPU.
-If they don't want to mislead the average dilettante gamer to buy a 12 core CPU, they have to weight in heavily the single core and four core performance in the overall speed.
Those, who have a modicum of knowledge, can just look at the benchmark's breakdown, and evaluate the data themselves.
@@TheGyuuula Did you even watch the video? A 2700x is significantly better in most modern games than the lowly quad core that it is being compared to yet Userbenchmark assigns it a higher score. Even in games where the quad core performs better, it is only by a hair. That seems incredibly disingenuous as Userbenchmark is objectively wrong here.
People buying a 12 core CPU probably know what they're doing and won't be using Userbenchmark anyway. Meanwhile, some poor kid is going to buy a quad core piece of crap instead of a similarly priced 6 or 8 core CPU. Even if you only cared about gaming, quality of life is just so much better with >4 cores.
me, who saw the title: 'what's wrong with userbenchmark? it's quite ok..'
me, while watching the video: 'ohh.. they changed the rating system.. wow.. indeed.. that looks like intel metrics..'
It's good you guys pushed this out. I was almost reliant on this site because I was 5 years out from my last build. But I found the site was off feeling. But to someone brand new they wouldn't have a clue.
So to them, multitasking isn't a thing? I listen to videos on youtube, play games and do other things that caused my 7600k choke and I had to upgrade!
I love these open discussions.Especially the hardware related ones!
Tbh I’m just finishing up my first pc build in 10 years and I stumbled upon you guys in a monitor review, and then I kept coming back for reviews on CPU’s and GPU’s as I put together my build. Much appreciated!
World: Having 6 cores instead of 4 affect 1% performance A LOT in more and more modern games
Userbenchmark: Great! Since it's only 1% let's measure its importance as 2% of a total performance!
Intel doing their shady business again. The same as Nvidia.
Same as AMD, too
@@XZagatoX between amd nvidia and intel we know who paid billion of dollars to the companys to force to not sell amd cpu
I used to use UserBenchmark all the time however wont any longer they have clearly taken a back hander from Intel according to them the Ryzen 9-3900x is only 4% faster than the i3-9350KF despite simple maths revealing the 3900x should be over 300% faster. They also have the i9-9900k as the fastest cpu despite the 3900x being more powerful by around 15-20% due to the extra cores and threads.
Effective speed is explicitly stated for gaming performance
I just went and compared the dual-core i3 7350K to 32 core Threadripper 2990WX and it said Threadripper was only 15% faster lol
So now my 6600k is on top and I don't need to upgrade for any reason whatsoever. Thanks user bench you saved me money and confused me, your the best.
Hey guys. It would be interesting to compare the 4x1 vs 1x4 Core/GHz scenario. In my opinion the performance of those two setups would be roughly the same. The single core would have to do 4 times more context switches (which are expensive) for multi program workload while 4 slower cores would have probably content for memory access as the memory bus is usually shared between cores. Of course cache might come into play heavily in this scenario as 4 core setup would probably have 4 times more L1 and L2 available. This is all theoretical and I think it might go beyond the level of depth this channel usually goes into.
All in all really a good talk, given that it was not scripted I think it was surprisingly on point and informative for 40+ minutes.
Also, one last note, it might be worth exploring for many percent of tested games and apps in a given comparison utilize how many cores (i.e. 20% is siglnglethreaded, 30% uses four cores, 20% uses up to 8 and 30% scales infinitely). This would then help to create the weights for the interested buyer (not necessarily for the web you criticized).
Cheers!
Love the format of you two discussing a topic in detail, listened as a podcast in the car. more more more :-) Please.
If you don't do parallel threading then most companies won't hire you at this point. They really want parallel threading coding because the market has change in the most common hardware. AMD cpus are really commonly used any high core count cpus and multi threaded cpus are even in laptops.
They should just have weighted 6 and 8 threads aswell if they didn't want to intentionally skew the results.
They can give multithread a low percentage, but 6 & 8 should be much higher, atleast for games.
I mean they can easily do some analysis regarding the cpu usage of modern games.
The new way of calculating the speed of a CPU might give a slightly better visualisation of what to expect from a particular processor in terms of gaming, but it makes it much harder for people like me to choose a CPU for productivity.
Productivity software like all of the Adobe CC products, Vegas Pro, Autodesk Maya, Autodesk 3Ds Max and others always use 100% of the available CPU resources, which means that the multi-core speed is essential there.
vegas pro doesn't use 100% of CPU always
The benchmark is described as a gaming/desktop one though. It'd be unwise to look at that anyways.
The website always gave the options to look at multi-core scores and "workstation score" - which if you used the website only and nothing else, you would look at for productivity unless you're negligent
But the fact that it has to be debated what the benchmark represents is such a silly thing anyway.
They should have either done nothing, or made a separate category with these new weightings, like "production" weighting and "gaming" weighting.
E: Or much better, just remove all weightings, all "quad core" benchmarks, and just have two lists for single core and multi-core respectively
Also with this new system an i3 9350kf BEATS a 9980xe. Ridiculous.
I don't think that represents any real-world workload.
*Not sure if they discuss this, still watching*
You've done a few videos on "old i7 quad cores in 2019", but please would you consider doing a video of how the old 6 core HEDT platforms hold up? x79/3930k/4930k @ 4.4Ghz, x99/5820k etc - Those cores weren't that useful in pure gaming back in 2011-2014 but seem to be providing more longevity nowadays...
I'm having flashbacks of 2000-2004 working at Fry's & having customers (& even fellow employees in the build-your-own "Components" section!) argue that if AMD was better than Intel, then you'd find them in OEM systems.
...few years pass & the truth comes out, unfortunately too late for AMD by that time.
intel paid a lot to keep amd out from oem more then 15 billion of dollars to the companys
Regarding the "single core CPU would choke to death on today's OS and games" comment, I just responded to someone in a post and decided to post this in open forum, someone might actually be interested in my findings. TLDR "You were absolutely right....CHOKED TO DEATH"
I have a 1st Gen Threadripper. I also have my OS installed on a RAID-0 NVME array (2 x 1TB 960 EVOs). It takes 3 seconds to go from BIOS to login screen. I went into my BIOS and disabled all but one core of the CPU. I also disabled multi-threading. When I rebooted it took over 7 mins to get to the login screen. I made the stupid decision to try to run Red-Dead Redemption 2 as my first game test. It took 35 mins to load the game and start my savegame. When I got there, it would get 35 FPS (normal is 65-80 for my system) but then every 2-3 seconds would just stop operating for 2-3 seconds....zero FPS. Then pick up again at 35 FPS. When I alt-tab out and open my browser, you could actually watch it draw the lines of boarder of the window like it was an Amiga 500 from 1987. And Alt-Tabbing back into the game took nearly 2 mins. (This is the part where people message me back and tell me I'm wrong, I misunderstood, I tested improperly, and call me a n00b or some other millennial meme. Don't bother)
Suspicious timing
Ryzen 3000 launch
And ice lake launch in few day ... VERY SUSPICIOUS
now guess what: back when Ryzen 1000 launched they decreased the multicore weight from 20% to 10%!
look here this is from Nov 2016 pre Ryzen lanuch (Ryzen launched Feb 2017)
web.archive.org/web/20161118235647/cpu.userbenchmark.com/Faq/What-is-the-effective-CPU-speed-index/55
quote:
"single-core mixed (30%), quad-core mixed (50%) and multi-core mixed (20%)"
this is post Ryzen launch from June 2017
web.archive.org/web/20170606033519/cpu.userbenchmark.com/Faq/What-is-the-effective-CPU-speed-index/55
quote:
single-core mixed (30%), quad-core mixed (60%) and multi-core mixed (10%)
that's interesting timing I guess.....100% suspicious!
Steve, Tim, even though you question the validity of user generated bench marks it is a decent way to have a quick comparison on hardware performance and no other site gives good quick look data on hardware in this way. The only other way to get performance numbers for comparison is from in depth analysis like you and GN do. Your issue with outliers is moot point since since it generally averages out over all the test runs. The weight of multi core performance in the over all score being skewed is a BIG issue though.
The other websites that offer data like this are measurably worse, usually extremely so.
04.04.2021. Userbenchmark still links here from their "about" page.
Still the only worthwhile information on the entire website.
Watching from 2020, it is weird seeing you both in the same room.
Video encoding, streaming etc not mainstream. Is that website living in 2002-2003?
Ive done some experiements out of curiosity using my R5 2600 and both VMs and turning on/or settings in the bios and I found that for most modern games a decently powerful 6 core CPU should be enough but its the bare minimum for enjoyable 60 fps experience.
This is one of my favorite videos you guys have done recently.
The *obvious* shill-be-gone aura surrounds the both of you here lol.
You two are clearly driven by data, and real world tests.
Tim nails it at 28:00:
"...It's not like the website was ever good."
the website also his some of the worse prices like the r9 3900x is almost 700 on there site like wtf
Came here again today from their "About" section, lol that was one *yuuge* MELTDOWN (pardon ye pun :D) on UB's end xD
These guys at UserBenchmark are the Ghost of 2004!!!
How much money must Intel have spent to resurrect these dead people! :)
for the point at 10:10 with the 1 core at 4GHz and 4 core at 1GHz .. the UFD tech channel did somehting similar on launch of the 9900k .. they took 1 core and maxed it out (hit like 5,6 or 5,5 stable) and then chose the closes frequency they could set to 5,5 divided by 8.
and the results of 1 super fast core vs. 8 cores below 1 GHz where amazing ... even if in gaming the average fps was around the same, the 1% low result was for the 8 cores something around double of the single core at 5,5 GHz ... even running windows was a real load of crap and laggyness on that 1 core. i mean it wasnt particualrly superb for the 8 cores below 1ghz either .. but MUCH better than the single core variant.
It really shows that even today multi core is a central part of our software and this effect will only increase with time
Why they simply do not give "buy Intel" as a result. We only lose time to enter the processor model, because the results are fake anyway.
While most of my PC use is single-core (like watching this video), its performance doesn't actually matter once I get to a certain point. It's the performance when I'm hammering my system hard (like encoding) that really matters to me or even mild use like you mention in the video (opening Chrome windows or general Windows usage).
I used the site but never really considered the overall rating a good one. I used the breakdown data to see how each cpu performed tasks. Then would look at reviews such as yours to then see their controlled performance comparison.
A new user buying purely from their data could easily be misguided. Its the same as looking at one games benchmark graph and basing all performance from that without understanding how the graph was made or test conditions.
The site is best for comparing older stuff when searching the used market to get a quick idea if its worth the effort to investigate further.
Thanks for the detailed feedback and stats guys. Appreciate you doing this.
Lol userbenchmark links to this video calling you incompetent “moar cores” marketers. Its hilarious. I dont even think they are paid by intel, but they are just so dug in their position and seem like they want to contrarians.
Haha thats just petty 🤣
Really appreciate you guys going into detail on this. I look at that site all the time and had no idea of the issues with their scoring. Consider me educated!Thankyou!
Well didnt cpu-z do sort of the same when the 1gen Ryzen came out
Yup, they did. 🙂
I'm cracking up, they rank same gen K sku i5 over an i7 even though its frequency is lower and it doesn't have hyper-threading.
I think it's good that you and other youtubers cover this shennanigans from Userbenchmark.com for those that are not so informed.
i was a noob and use to use that site, then I found HUB and GN and Tech Deals among others, now I don't look back and get wayyy better value for my money. Thank you guys for all your hard work and showing us the way. You are very much appreciated.
They still say you are "Incompetent 'moar core' marketers" on their website. As of Jan 18th 2022
0:24 - 0:27 am i the only one that finds the expression changing from cheeky to dead serious on the guy on the right so hilarious?!
First video ive seen from you on my newly built 3700x x570 aorus elite build!
*"What kind of math is that? That's fools math playa" - Jizzy B*
Userbenchmark lost its credibility for me atleast.
they never had any, their testing methodology was always flawed, just like all the GPU sites out there too
You guys are great keep up the good work honest and fair comparisons across the board are a great benefit for all of us Thank you