News Controversial benchmarking website goes behind paywall — Userbenchmark now requires a £10 monthly subscription

[rluker5]

It's actually worse than claiming a Raptor Lake i9 will use 330 W, because there are at least some situations where the latter is factually accurate.

[/QUOTE]
I think the more obvious and prevalent one is worse. As an example:

View: https://youtu.be/2MvvCr-thM8

Vs finding some snarky trolling on a page you have to look for. And yes, I still find Gamers Nexus credible, just sometimes misleading. But by all means call them both out. Neither some guy at userbenchmark trolling in his own personal comment section, nor youtubers misleading for clicks are appropriate.

You'd be surprised how much ads can bog down a web page. Not only that, but anti-malware makes it even worse. I have a 16-core/24-thread Alder Lake CPU in my work laptop, and Chrome bogs down even worse on that machine than when I use firefox on my Sandybridge w/ anti-spyware. Same web sites, same internet connection!


What the heck are you even talking about?

There is a benchmark that shows website performance: PCMark10. I couldn't find the 9654 so I went with what I could find, the Epyc 7742. It's clocks are 300mhz lower than a 9654 so I also went with a 300 mhz lower 12100f for comparison:
https://www.3dmark.com/pcm10b/1387910

I scored 5 507 in PCMark 10

AMD EPYC 7742, NVIDIA GeForce RTX 3090 x 1, 229238 MB, 64-bit Windows 10}

www.3dmark.com

And the 12100f had a +25% web browsing score. There are probably more but the ones I've seen get more synthetic and use different operating systems, etc. You don't see anything load in geekbench, for example. Who knows what else they are skipping in the steps.

The reason your mobile chip is lagging is that it is in a mobile setup. I've got an i3 Haswell that is more responsive than my 7700hq laptop chip as well.

 

[bit_user]

There is a benchmark that shows website performance: PCMark10.

That's a composite score. If you look at the web benchmarks that Toms uses in its CPU reviews, you can see that some are more thread-friendly than others.

I couldn't find the 9654 so I went with what I could find, the Epyc 7742.

Why??? This is a nonsensical comparison.

The reason your mobile chip is lagging is that it is in a mobile setup.

This is absurd. I looked it up and it's a i7-12850HX, with 55 W base power, 157 W turbo power, 4.8 GHz turbo clock, DDR5 memory, and PCIe 4.0 NVMe SSD.

The Sandybridge I'm talking about is a i7-2600K with 95 W TDP, 3.8 GHz turbo, DDR3 memory, and a SATA SSD.

And yet, I'm to believe that the mere fact that the former is a mobile CPU, in spite of being HX-series mobile workstation and using the same die as desktop Alder lake i9, somehow makes it worse than a desktop CPU from 10 years earlier? Dude, I know you're smarter than that, so you must be joking. I ain't laughing.

Furthermore, the HX laptop uses a Nvidia dGPU, while my Sandybridge is using its iGPU, further reducing its memory bandwidth. Both are connected to the same 1440p monitor. Also, both use gigabit Ethernet to connect to the same switch.

 

[rluker5]

This is absurd. I looked it up and it's a i7-12850HX, with 55 W base power, 157 W turbo power, 4.8 GHz turbo clock, DDR5 memory, and PCIe 4.0 NVMe SSD.

The Sandybridge I'm talking about is a i7-2600K with 95 W TDP, 3.8 GHz turbo, DDR3 memory, and a SATA SSD.

And yet, I'm to believe that the mere fact that the former is a mobile CPU, in spite of being HX-series mobile workstation, somehow makes it worse than a desktop CPU from 10 years earlier? Dude, I know you're smarter than that, so you must be joking. I ain't laughing.

It isn't the CPU, it is the power plan, ram timings, etc. The setup. If you were to take that same CPU and put it in an lga1700 adapter it would likely be as responsive as it's desktop equivalent.

But back to the topic of some benchmarks not being as good for a typical PC owner at providing understanding as to how their PC should perform, the Epyc 7742 absolutely destroys the 12100f in a great many things. And many benchmarks back this up. They just aren't the same things that a typical PC owner would find themselves doing with their PC. Userbenchmark if focused more on the needs of some low information user and is better suited for them. It would be better still if they ditched the biased CPU recommendations, but hopefully they are obvious enough for anybody normal to see through.

 

[bit_user]

It isn't the CPU, it is the power plan, ram timings, etc. The setup. If you were to take that same CPU and put it in an lga1700 adapter it would likely be as responsive as it's desktop equivalent.

This is even with it plugged-in, on max performance settings. And I make sure the air ducts are free of dust. I even have the back end of it propped up slightly, for better airflow.

But back to the topic of some benchmarks not being as good for a typical PC owner at providing understanding as to how their PC should perform, the Epyc 7742 absolutely destroys the 12100f in a great many things. And many benchmarks back this up.

Your own example of PCMark 10 contradicts you.

But, I'll give you this: I've been going through some CPU reviews, on 3 different review sites, and I can't find a recent review with a web browsing benchmark that clearly indicates a benefit from threading (beyond the baseline of 8 threads or whatever minimum spec CPU was included in the lineup). I still maintain that web browser are well-threaded, but perhaps the point of diminishing returns is already less than the baseline number of threads for modern CPUs.

I did have a weird experience where the CPU fan got stuck and went to 0 RPM. The CPU immediately throttled down to running a single core at 800 MHz. Before I figured out the problem was the CPU fan, I first disabled some Intel Power Management service, which enabled all cores to run at 800 MHz. The difference between 1 core and all cores was night-and-day. The web browser, on the same page, went from completely unusable, to sluggish-but-usable (barely). There's something in that...

 

[TerryLaze]

I still maintain that web browser are well-threaded, but perhaps the point of diminishing returns is already less than the baseline number of threads for modern CPUs.

They are but they are so for stability, so that your whole browser doesn't freeze when one page freezes.
You can open task manager and see all the separate tasks running but you will see them running at 1-2% each, a web browser is not one task running mutithreaded, it's a lot of small tasks, some of them running very small mutithreaded some of them running mostly singlethreaded, you will notice the singlethreaded ones way more since the tasks in general are too small to have any wait time in multithreaded.
You have to go waaaaay out of your way to find something webbased that will run one big thing multithreaded, some accountant thing or whatever.


Process hacker shows an even more detailed view of the thread that each task runs for anybody that is interested.

 

[TerryLaze]

I did have a weird experience where the CPU fan got stuck and went to 0 RPM. The CPU immediately throttled down to running a single core at 800 MHz. Before I figured out the problem was the CPU fan, I first disabled some Intel Power Management service, which enabled all cores to run at 800 MHz. The difference between 1 core and all cores was night-and-day. The web browser, on the same page, went from completely unusable, to sluggish-but-usable (barely). There's something in that...

Yes there is that you aren't running the browser in limbo, windows will run terribly on a single core on its own. let alone trying to run anything on top of that.

 

[ivan_vy]

Yes, of course, but it comes down to both Tom's and UBM earning money, and possibly so for their survival (maybe not Tom's so much any more). This is the only way UBM can make money. Can we blame them for trying?

I think the problem here is not the intention (everyone needs money) but the execution (annoying captcha, expensive subscription model)

 

[Order 66]

This is the most "meh" thing I have read all day. Good riddance as I am sure it will be a tremendous success...not.

I remember looking at fps for a game and seeing 1080p low getting more fps on the same hardware than 720p low. Didn't make any sense, and at that point (this was before I was even into computers) I knew not to trust it.

 

[Order 66]

You'd be surprised how much ads can bog down a web page. Not only that, but anti-malware makes it even worse. I have a 16-core/24-thread Alder Lake CPU in my work laptop, and Chrome bogs down even worse on that machine than when I use firefox on my Sandybridge w/ anti-spyware. Same web sites, same internet connection!

Believe me, I know. Used an old school laptop (Intel Celeron n3350 and 4GB of RAM with chrome) it was awful. I actually believed the userbenchmark score for my laptop, because it was that bad. I think it was like a 3% or something comically low.

 

[ODuffer]

Userbenchmark now requires a £10 monthly subscription​

Good luck with that🤣

 

[AgentBirdnest]

Not sure if it is $10 per year or month. The title "$10 monthly subscription" but the article says "subscribers to the $10-per-year Pro plan can test "

It's per-year. The Tweet/X (Twix? ... I want a Twix bar now) says it's per year, UBM says per year. TomsHardware magically pulled the word "month" out of their digital butt for the title, either out of incompetence or to get clicks.

Get your s*** together, TH. Honestly... if it weren't for Jarred and Paul's reviews, and the forums, I would never use this site again.

 

[Order 66]

It's per-year. The Tweet/X (Twix? ... I want a Twix bar now) says it's per year, UBM says per year. TomsHardware magically pulled the word "month" out of their digital butt for the title, either out of incompetence or to get clicks.

Sure, but even at $10 a year it is not worth it.

 

[AgentBirdnest]

Sure, but even at $10 a year it is not worth it.

Agreed!

 

[bit_user]

Get your s*** together, TH. Honestly... if it weren't for Jarred and Paul's reviews, and the forums, I would never use this site again.

Look at who wrote it. I rarely see one of this author's articles that I don't have at least one nit-pick with. He's a freelancer who used to write for this site, then stepped away for several years. He's been back since the end of last year.

https://www.tomshardware.com/author/matthew-connatser
​

At least you can say his articles are reasonably well-written, even if sometimes getting facts wrong or simply not doing all of the appropriate research. TBH, I never cared as much about grammar as some people seem to. If I have to sacrifice grammar or accuracy, I'd much rather give up the former.

To be fair, a few of his articles have gone a bit deeper or explained a bit more than I expected. So, I think it's not as if he never does any diligence. It's just not often enough.

 

[madaraosenpai]

It's like they WANT to die.

 

[Nine Layer Nige]

Thanks for posting this article.
I had used UBM many times and often recommended it to Toms forum users who had problems with their rig and often exposed obvious things to fix or clear weak components. Its almost impossible to get on the server now - maybe it was/is being spammed by so many users or bots that it was necessary. I still want to recommend it to help Tom users, but no longer can as it's no longer click and run and cant be bothered to explain the hoops to jump through to succeed. Its a great disappointment.

 

[rgd1101]

an update https://forums.tomshardware.com/threads/userbenchmark-benchmark-is-now-malware.3850488/