User Benchmark = Fake Benchmark

pc-guy · 28 Jul 2019 at 21:34

Brazo said:
Perhaps because it reflects the real world?

that's a pretty terrible statement. do you actual know your CPU usage for sure? I mean multi-core CPU has been around for LONG time. whilst applications nowadays may not have adopted the 8 core 12 core a but definitely quad core is fairly standard stuff even windows and office applications support multicore multithread. I am watching youtube atm on my 4770k and all 4 cores and 8 threads are being used. so i don't know where you believe single core is more important.

as a matter of fact, the weighing of CPU score should be leaning more towards multicore (greater 4 core scores and more towards >4 cores) as opposed to weighing heavier on single thread performance as there will be generally more adoption of this increasing common standard of computing in the future so software developers will be utilising more.

Zernath · 28 Jul 2019 at 21:41

Ross Thomson said:
user benchmark isn't relevant to anything. Does anyone seriously use it to compare cpu's?

Obviously enough that a certain company needed the results changed. :rolleyes:

NinjaCool · 28 Jul 2019 at 21:47

HangTime · 28 Jul 2019 at 22:33

humbug said:
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

I assume the reason is that in some tasks, in particular gaming you get diminishing returns from having more cores, so they don't want the vast benefits you get in certain usage scenarios to skew the overall results too much. You could have a scenario where an 8 core cpu is trouncing a 4 core cpu in some productivity apps but trailing significantly in others such as gaming (e.g. Ryzen 1700 vs 7700k). Ideally they'd have some sort of curve model to handle this but I guess they want to keep it simple.
I haven't looked into it much but if it is genuinely saying 4 core CPUs are better than CPUs with more cores and the same or better performance per core, then obviously that is wrong. I assumed it was more a case of just putting most of the weighting on single/quad core performance.

Brazo · 28 Jul 2019 at 23:20

pc-guy said:
that's a pretty terrible statement. do you actual know your CPU usage for sure? I mean multi-core CPU has been around for LONG time. whilst applications nowadays may not have adopted the 8 core 12 core a but definitely quad core is fairly standard stuff even windows and office applications support multicore multithread. I am watching youtube atm on my 4770k and all 4 cores and 8 threads are being used. so i don't know where you believe single core is more important.

as a matter of fact, the weighing of CPU score should be leaning more towards multicore (greater 4 core scores and more towards >4 cores) as opposed to weighing heavier on single thread performance as there will be generally more adoption of this increasing common standard of computing in the future so software developers will be utilising more.

Try watching YouTube on an i3.

Trust me on two cores it simply works

Try watching it on a vm. Utilising a fraction of one core from a workstation, guess what... it works

PiKe · 29 Jul 2019 at 00:21

Another pretty laughable comparison, 2500k Vs 1700x

That's 4c/4t Vs 8c/16t...

You end up with +80% in the user bench but it's effective speed only ends up at +13% even though it gets +216% on multicore speed.

Used to use this site a lot because their benchmarks actually meant something but not anymore, passmark and cinebench for me from now on.

emodan · 29 Jul 2019 at 07:50

Used to use userbench to compare a customers PC to the minimum requirements of our games. This change has made it harder for me to advise people that their CPU is the limiting factor so I guess I need to switch another site for performance comparisons.

chrcoluk · 29 Jul 2019 at 08:40

ok humbug I will analyse that list later.

chrcoluk · 29 Jul 2019 at 08:44

PiKe said:
Another pretty laughable comparison, 2500k Vs 1700x

That's 4c/4t Vs 8c/16t...

You end up with +80% in the user bench but it's effective speed only ends up at +13% even though it gets +216% on multicore speed.

Used to use this site a lot because their benchmarks actually meant something but not anymore, passmark and cinebench for me from now on.

Why do you need the effective speed tho?

It gives you what you needed right, you can see its +216% on multicore and +8% single core.
You are intelligent enough to work out from those 2 numbers alone which cpu is better for your needs. Just ignore the "effective speed". That seems to be a metric for PC noobs who cant think for themselves.

I dont understand why people are so obsessed with effective speed, it is the least important number on the entire page.

Decide what your most important use case is.

If you a content creator, look at the multi core comparison.
If your priority is legacy games, look at the single core comparison.
If you are mixed, take a note of both multi core and single core and try to get a cpu that balances them both well.

Just ignore the effective speed and think for yourself.

It does feel like people want a website or video presenter to tell them what cpu to buy.

chrcoluk · 29 Jul 2019 at 08:47

HangTime said:
I assume the reason is that in some tasks, in particular gaming you get diminishing returns from having more cores, so they don't want the vast benefits you get in certain usage scenarios to skew the overall results too much. You could have a scenario where an 8 core cpu is trouncing a 4 core cpu in some productivity apps but trailing significantly in others such as gaming (e.g. Ryzen 1700 vs 7700k). Ideally they'd have some sort of curve model to handle this but I guess they want to keep it simple.
I haven't looked into it much but if it is genuinely saying 4 core CPUs are better than CPUs with more cores and the same or better performance per core, then obviously that is wrong. I assumed it was more a case of just putting most of the weighting on single/quad core performance.

The only time I see cpus with less cores ranked higher than ones with more cores is when there is a noticeable gain on SC performance. If the gain is small, the cpu with more cores comes out on top. What they have done is your last sentence.

pc-guy · 29 Jul 2019 at 21:29

chrcoluk said:
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.

you are all sort of wrong here. some of those apps are single thread but VLC is definitely multi-threaded, office is definitely multi-thread since like office XP, AVs are multi-threaded these days. i don't know if you still live in the millennium bug era.

but the point is that lots of these software can run at the background like AV so with multi-thread it enables the work load to be shared out so you can get on with other stuff on the computer like spreadsheeting or typing report while watching champions league on the second screen or stream spotify or watch the Opens.

I meant even bother to list Notepad and MS Paint is pathetic - do them actually take up any CPU resources???! that is a very narrow and very small hole you have dug there.

Think about normal average users - web browser, streaming, youtubing, ituning, gaming, office work (multi-thread), multi-tasking, online shopping lots of these are all content rich and are all multi thread workloads. if you got a single core CPU i can guarantee you that you will only be able to do one of those things at the same time as opposed to simultaneously. most of the people who buys a PC nowadays are probably going to be using it for some kind of gaming or content creation, otherwise they would be using a phone or tablet.

it is simply insane to suggest technology that is effectively Jurassic in computing world is relevant nowadays and given more priority over newer tech - for the sake of propping up a particular brand and company. Do you honestly believe in what you said in the above...seriously.

pc-guy · 29 Jul 2019 at 21:34

Brazo said:
Try watching YouTube on an i3.

Trust me on two cores it simply works

Try watching it on a vm. Utilising a fraction of one core from a workstation, guess what... it works

it works but it doesn't work very well. I had a i3 on my streaming box, and a core2duo laptop and a i5 laptop guess which one performs the best at doing the same thing whilst also having headroom for me to remote dial in to setup as a download box also...yes the i5. the i3 and i5 were of the same gen similar clocks but one just had more headroom to do more due to the cores and threads available...simples.

Panos · 30 Jul 2019 at 02:35

humbug said:
There is a contradiction in what User Bench are saying.

They are saying that applications and games above 4 cores are irrelevant because those applications don't really exist, or that they are a tiny fraction of applications / games, and yet they had to adjust the more than 4 core weighting from 10% to 2% because 'as they put it' there were getting a lot of highly threaded performance results skewing their results in favour of the highly threaded Ryzen CPU's.

Well if those application don't use all those threads on Ryzen then why are they skewing the results to such an extent that you feel you need to reduce how much that makes up the overall result by such massive margins?

Are these guys stupid or do they just think we are?

Don't forget also. Until Ryzen 1000 came out, multi core weight was 20%. The moment the 1800X came out, Userbenchmarks dropped the multicore weight to 10%

https://web.archive.org/web/2017011.../Faq/What-is-the-effective-CPU-speed-index/55

Now the 12 & 16 cores are mainstream, they dropped multi core to 2%. The plot thickens.

So with the current weights, I won't be surprised if the 9980XE is beater by a 6600K

4K8KW10 · 30 Jul 2019 at 06:37

Panos said:
Don't forget also. Until Ryzen 1000 came out, multi core weight was 20%. The moment the 1800X came out, Userbenchmarks dropped the multicore weight to 10%
https://web.archive.org/web/2017011.../Faq/What-is-the-effective-CPU-speed-index/55

Now the 12 & 16 cores are mainstream, they dropped multi core to 2%. The plot thickens.

So with the current weights, I won't be surprised if the 9980XE is beater by a 6600K

i9-9980XE is beaten by an i3 :eek:

humbug · 30 Jul 2019 at 09:42

Panos said:
Don't forget also. Until Ryzen 1000 came out, multi core weight was 20%. The moment the 1800X came out, Userbenchmarks dropped the multicore weight to 10%
https://web.archive.org/web/2017011.../Faq/What-is-the-effective-CPU-speed-index/55

Now the 12 & 16 cores are mainstream, they dropped multi core to 2%. The plot thickens.

So with the current weights, I won't be surprised if the 9980XE is beater by a 6600K

oh wow.... what are they going to do when soon 16 cores is mainstream and they get a lot of them bubbling to the top of the charts? Change it to 0.5%?

4K8KW10 · 30 Jul 2019 at 09:55

humbug said:
oh wow.... what are they going to do when soon 16 cores is mainstream and they get a lot of them bubbling to the top of the charts? Change it to 0.5%?

No, they must change the Multi-core weight to at least 50%, better 70%, giving the rest 30-50% equally divided between Single-core weight and 4-core weight.

Why is there 4-core weight, at all?

humbug · 30 Jul 2019 at 10:07

4K8KW10 said:
No, they must change the Multi-core weight to at least 50%, better 70%, giving the rest 30-50% equally divided between Single-core weight and 4-core weight.

Why is there 4-core weight, at all?

A legacy from Intel's CPU's being only 4 cores mainstream, i swear this is just a shill site for Intel, they are always at the top of Googles search results when you search CPU benchmarks or comparisons, it costs huge amounts of money to stay at the top of Googles search results.

DragonQ · 30 Jul 2019 at 10:45

chrcoluk said:
I dont understand why people are so obsessed with effective speed, it is the least important number on the entire page.

And yet it is presented as the most important. Everyone on this forum understands that the number is worthless but the entire point here is that Joe Bloggs doesn't know that and is being misled by this stupid weighting.

4K8KW10 · 30 Jul 2019 at 11:30

DragonQ said:
And yet it is presented as the most important. Everyone on this forum understands that the number is worthless but the entire point here is that Joe Bloggs doesn't know that and is being misled by this stupid weighting.

Yup, imagine all those 12-14-year-olds who will see this and jump on the 4-core intel bandwagon. It's disgusting :rolleyes:

chrcoluk · 30 Jul 2019 at 11:48

pc-guy said:
you are all sort of wrong here. some of those apps are single thread but VLC is definitely multi-threaded, office is definitely multi-thread since like office XP, AVs are multi-threaded these days. i don't know if you still live in the millennium bug era.

but the point is that lots of these software can run at the background like AV so with multi-thread it enables the work load to be shared out so you can get on with other stuff on the computer like spreadsheeting or typing report while watching champions league on the second screen or stream spotify or watch the Opens.

I meant even bother to list Notepad and MS Paint is pathetic - do them actually take up any CPU resources???! that is a very narrow and very small hole you have dug there.

Think about normal average users - web browser, streaming, youtubing, ituning, gaming, office work (multi-thread), multi-tasking, online shopping lots of these are all content rich and are all multi thread workloads. if you got a single core CPU i can guarantee you that you will only be able to do one of those things at the same time as opposed to simultaneously. most of the people who buys a PC nowadays are probably going to be using it for some kind of gaming or content creation, otherwise they would be using a phone or tablet.

it is simply insane to suggest technology that is effectively Jurassic in computing world is relevant nowadays and given more priority over newer tech - for the sake of propping up a particular brand and company. Do you honestly believe in what you said in the above...seriously.

The way to test a/v is easy enough, run a virus scan, then watch the cpu load whilst you doing it. I have not tested every single a/v but I have tested a lot of the mainstream ones and every one was scanning using only a single core.

VLC note my comments, the app supports multi threading, but most of the decoders do not, especially the most used one's since the bottleneck is typically the decoder than thats whats relevant.

The apps I listed specifically were designed in mind to represent the typical PC user, most people use whats freely available to them and are not professionals. So wont use high end professional tools. now the tasks you listed are nearly all online stuff which go via a web browser. The web browser is the one mainstream app that of course stands out as been multi threaded, I never said otherwise tho. However one thing to bear in mind with online activities via a web browser, they are heavily multi threaded, but its one thread per tab. So if you visit a website that needs a lot of cpu grunt to do its job, you will be bottlenecked by what one core can do. Chrome dev's are working to improve this tho so it will get better in the future, but is a work in progress. So far they have put as in testing features thread offloading for network traffic, and any draw call stuff for the GPU. There is also a testing feature that does this for html5 videos. Firefox is no where near as threaded as chrome and are quite far behind in that respect.