Is userbenchmark being payed by intel? WTF

[rx590overclocker]

Where is our AMD at the top?

[SpookyCitrus]

Userbenchmark is highly unreliable, just forget about it. It's all user results, most are going to be overclocked or not at stock settings, could even be tainted results. If you're looking for a reliable benchmarking utility go with 3Dmark. 

[rx590overclocker]

Just now, SpookyCitrus said:

Userbenchmark is highly unreliable, just forget about it. It's all user results, most are going to be overclocked or not at stock settings, could even be tainted results. If you're looking for a reliable benchmarking utility go with 3Dmark.

I know that...

I am trying to figure out why, it is coming out like this.

[SpookyCitrus]

1 minute ago, rx590overclocker said:

I know that...

I am trying to figure out why, it is coming out like this.

You are filtering with AVG% which is the overall system(all components together) and it's ranking, not the effectiveness or power of the CPU's themselves that is what the core points columns are for.

[rx590overclocker]

1 minute ago, SpookyCitrus said:

You are filtering with AVG% which is the overall system(all components together) and it's ranking, not the effectiveness or power of the CPU's themselves that is what the core points columns are for.

oh...

[SpookyCitrus]

2 minutes ago, rx590overclocker said:

oh...

Also again, just to reiterate, the results will more than likely still be skewed, Userbenchmark is highly unreliable and you shouldn't even consider using it to compare anything.

[Atmos]

1 hour ago, rx590overclocker said:

 

Where is our AMD at the top?

Formatting tools are broken for me because I use vivaldi, so please dont mind the mess im bouta paste.

 

"The UBM effective speed measures performance for typical consumers. For example, we de-emphasize deep queue depth data transfer and heavily multi-threaded CPU workloads as these metrics are not generally consumer orientated."

The intel line of cpus is still technically faster in the single thread, and since the overwhelming majority of cpus tested on UBM are at stock speed, not overclocked, and many, many AMD systems are running sub-par ram speeds, it taints the scores even more. If you actually compare cpus like the 3700x to the 9700k you still will find that the 3700x dominates it in workload metrics even on UBM, and only loses marginally in gaming performance, as well it would.

 

For example, actually comparing the "highest avg score" cpu, the 9900ks, and the 3900x, a cpu more than 200$ cheaper, still shows that in workload benchmarks and metrics the 3900x absolute dominates the much more expensive 9900ks

https://cpu.userbenchmark.com/Compare/Intel-Core-i9-9900KS-vs-AMD-Ryzen-9-3900X/m929964vs4044

again, as well it should being a 12c/24t with only marginally lower single-core performance, compared to the 8c/16t hot garbage that the 9900ks is.

 

 If you want accurate down to the percentage for specific workloads, then you need to find good reviews on the individual processors or components. UBM is basically just to get a jist, but for the love of god, look at the detailed comparisons of components, not just the average scores.

 

 

53 minutes ago, SpookyCitrus said:

Also again, just to reiterate, the results will more than likely still be skewed, Userbenchmark is highly unreliable and you shouldn't even consider using it to compare anything.

userbenchmark is a perfectly fine metric for examining particularly average workloads, not specific ones. I would never recommend using it to decide on a specific component upgrade, its not a guide, just a broad sweeping tool for system checking and general reassurance. Its perfectly fine for comparing in certain metrics, but once again as posted above if you want real detailed information then you need to retrieve more indepth reviews for each component and weigh them against each other.

 

UBM isnt even remotely gpuboss, let's not put them in the same classification shall we?

[seon123]

18 minutes ago, Atmos said:

userbenchmark is a perfectly fine metric for examining particularly average workloads, not specific ones. I would never recommend using it to decide on a specific component upgrade, its not a guide, just a broad sweeping tool for system checking and general reassurance. Its perfectly fine for comparing in certain metrics, but once again as posted above if you want real detailed information then you need to retrieve more indepth reviews for each component and weigh them against each other.

 

UBM isnt even remotely gpuboss, let's not put them in the same classification shall we?

I would not consider them to be serious, and would not really bother using them for anything. Surely, no even remotely professional site would be this unserious?

Quote

Why we do it

Marketers make it tough to choose hardware. An army of anonymous paid social media accounts spread hype and disinformation to drive sales. Incompetent (moar core) smearers would sell ice to Elsa.

From https://www.userbenchmark.com/page/about

The hyperlink in "Incompetent (moar core) smearers" goes to a Hardware Unboxed video, which called them out on some interesting changes.

[Atmos]

19 minutes ago, seon123 said:

I would not consider them to be serious, and would not really bother using them for anything. Surely, no even remotely professional site would be this unserious?

 

From https://www.userbenchmark.com/page/about

The hyperlink in "Incompetent (moar core) smearers" goes to a Hardware Unboxed video, which called them out on some interesting changes.

So, did you read my post? Doesnt seem like you did...

39 minutes ago, Atmos said:

userbenchmark is a perfectly fine metric for examining particularly average workloads, not specific ones.

39 minutes ago, Atmos said:

For example, actually comparing the "highest avg score" cpu, the 9900ks, and the 3900x, a cpu more than 200$ cheaper, still shows that in workload benchmarks and metrics the 3900x absolute dominates the much more expensive 9900ks

39 minutes ago, Atmos said:

"...heavily multi-threaded CPU workloads as these metrics are not generally consumer orientated." -UBM

I dont care about drama. I dont care what the guys maintaining the site do because they have 0 weight over the actual benchmark results, they only control the average arbitrary scores given for each category. i dont care what fake "ranking" they have. I showed how it can be used in a fairly accurate and useful way, and that they are clear and open in telling people that the score isnt an all round score. All it took was clicking a single link right next to the score that tells you exactly what the score means and how its determined.

 

I explained why their overall avg scores are what they are. Explained that you should use it to compare specific benchmark tests and results only, and that it only serves as a broad sweeping tool when examining specific components, against specific components, not just by looking at a bunch of completely arbitrary numbers and scores handed out for what best suits the median consumer.

 

This is a segment of the market online that is filled with sites that actively falsify benchmarks for brands, and suppress general information on testing methodology and benchmarks. UBM is by far and wide the least offensive of the bunch, and also by far and wide the most broadly accessible. I would rather they lose the global rankings and average scores all together, or at least include more than 8 cores into their "avg score" benchmark totals, instead of giving it its own category unrepresented on the global ranking page, but it is what it is, and they are upfront about it.

 

Until some tech reviewer with reliable benchmarks or a conglomeration of them compile all their results both current and historic into an online, user-friendly database, then this segment of the market isnt going to change, or get any more accurate than what UBM is offering.

Edited March 12, 2020 by Atmos

[rx590overclocker]

15 minutes ago, Atmos said:

So, did you read my post? Doesnt seem like you did...

I dont care about drama. I dont care what the guys maintaining the site do because they have 0 weight over the actual benchmark results, they only control the average arbitrary scores given for each category. i dont care what fake "ranking" they have. I showed how it can be used in a fairly accurate and useful way, and that they are clear and open in telling people that the score isnt an all round score. All it took was clicking a single link right next to the score that tells you exactly what the score means and how its determined.

 

I explained why their overall avg scores are what they are. Explained that you should use it to compare specific benchmark tests and results only, and that it only serves as a broad sweeping tool when examining specific components, against specific components, not just by looking at a bunch of completely arbitrary numbers and scores handed out for what best suits the median consumer.

 

This is a segment of the market online that is filled with sites that actively falsify benchmarks for brands, and suppress general information on testing methodology and benchmarks. UBM is by far and wide the least offensive of the bunch, and also by far and wide the most broadly accessible. I would rather they lose the global rankings and average scores all together, or at least include more than 8 cores into their "avg score" benchmark totals, instead of giving it its own category unrepresented on the global ranking page, but it is what it is, and they are upfront about it.

 

Until some tech reviewer with reliable benchmarks or a conglomeration of them compile all their results both current and historic into an online, user-friendly database, then this segment of the market isnt going to change, or get any more accurate than what UBM is offering.

this dude can type effing fast

[GDRRiley]

30 minutes ago, Atmos said:

UBM is by far and wide the least offensive of the bunch, and also by far and wide the most broadly accessible. I would rather they lose the global rankings and average scores all together, or at least include more than 8 cores into their "avg score" benchmark totals, instead of giving it its own category unrepresented on the global ranking page, but it is what it is, and they are upfront about it.

given their search result place, they have a lot more influence and should be punished for not showing using actuate performance comparisons.

they aren't upfront, they have mostly hidden how the create the score. It is in a few blog posts.

they went backwards not forwards in how they balanced muticore scores.

it looks like a 2008 workload not a 2020.

 

single core should make up maybe 10-20% of it.

12-16 threaded workload. 60%

highly threaded 20-30%

That would be a much more modern and actuate representation of processors.

[Atmos]

1 hour ago, GDRRiley said:

given their search result place, they have a lot more influence and should be punished for not showing using actuate performance comparisons.

they aren't upfront, they have mostly hidden how the create the score. It is in a few blog posts.

they went backwards not forwards in how they balanced muticore scores.

it looks like a 2008 workload not a 2020.

 

single core should make up maybe 10-20% of it.

12-16 threaded workload. 60%

highly threaded 20-30%

That would be a much more modern and actuate representation of processors.

How they weigh scores is literally a single link away, directly under the score. It actually could not be more upfront unless they wrote the entire article in each score box.

They are showing accurate scores for their testing, they simply ignore anything over 8 cores in their avg score ranking, and are upfront and directly state that.

They were getting hate from normal consumers who dont care if a 24c/48t xeon is massively faster for rendering, if its slower for browsing facebook and playing call of duty. Thats an extreme example, but just a reason why they changed. I dont like it, because it only devalues my experience, but I am able to understand why. I still want them to change it however.

They weigh their tests on the benchmark performance, which measures single core, dual, 4 and 8 core performance. Thats it. Thats why xeons and threadripper cpus do not absolutely dominate the rankings. That is perfectly fine, because they clearly state so in the scoring methodology available DIRECTLY next to each score.

 

Once again, I want them to change, and at the very least add in a column that shows complete cpu performance as opposed to ranking 8 core performance as "avg score". They should change "avg score" to "Daily-Use Score" and create a new column called something like "Workstation Score" in the global ranking that shows fully multi-threaded benchmark results. That way those of us interested can see those cpus more easily and how they stack up with user-benchmarks.

 

We need to keep in mind that once again, they are not being deceptive. They clearly, and plainly state that the "avg score" ranking is NOT based on actual cpu performance, but what most consumers will experience with the CPU. They have to cater to a majority, and in the computer world enthusiasts who care about 128c/256t cpus are in the absolute minority.

 

Quote

this dude can type effing fast

lmao, im not the fastest, and i re-wrote that many times over before posting. I just use computers a lot, and have for many years now. I know people who double my WPM and have less than a tenth of the probably 65,000 hours i have put in at a computer in the last 15 years.

[GDRRiley]

1 minute ago, Atmos said:

How their weigh scores is literally a single link away, directly under the score. It actually could not be more upfront unless they wrote the entire article in each score box.

They are showing accurate scores for their testing, they simply ignore anything over 8 cores in their avg score ranking, and are upfront and directly state that.

They were getting hate from normal consumers who dont care if a 24c/48t xeon is massively faster for rendering, if its slower for browsing facebook and playing call of duty. Thats an extreme example, but just a reason why they changed. I dont like it, because it only devalues my experience, but I am able to understand why. I still want them to change it however.

They weigh their tests on the benchmark performance, which measures single core, dual, 4 and 8 core performance. Thats it. Thats why xeons and threadripper cpus do not absolutely dominate the rankings. That is perfectly fine, because they clearly state so in the scoring methodology available DIRECTLY next to each score.

 

Once again, I want them to change, and at the very least add in a column that shows complete cpu performance as opposed to ranking 8 core performance as "avg score". They should change "avg score" to "Daily-Use Score" and create a new column called something like "Workstation Score" in the global ranking that shows fully multi-threaded benchmark results. That way those of us interested can see those cpus more easily and how they stack up with user-benchmarks.

 

We need to keep in mind that once again, they are not being deceptive. They clearly, and plainly state that the "avg score" ranking is NOT based on actual cpu performance, but what most consumers will experience with the CPU. They have to cater to a majority, and in the computer world enthusiasts who care about 128c/256t cpus are in the absolute minority.

no its 3-4 away and even then the breakdown isn't clear.

they are, 99% of people are going to buy based on rating and nothing else.

if we are taking most consumers, they can do facebook on a core 2 duo with 4gb of ram and an SSD just fine.

I'm not saying threadripper 3990x should be at the top but right now the 3950x should be. if they want to move server and workstations chips out onto their own space fine.

[Delicious Cake]

Userbenchmark is about as trustworthy and accurate as a targeted political advert on Facebook.

[Pickles von Brine]

I only use it as a VERY rough guide for video cards. Generally to see if one is faster. Even then i take it with a grain of salt and also look at reviews. Though it can be tough to find a good answer. 

[LogicalDrm]

This was covered in length back in last fall. The sum of it, they changed algorithm to make single-core performance have bigger deal than overall or multi-core performance. Which essentially dropped any last trustworthiness they might have had before it.

[rx590overclocker]

49 minutes ago, LogicalDrm said:

This was covered in length back in last fall. The sum of it, they changed algorithm to make single-core performance have bigger deal than overall or multi-core performance. Which essentially dropped any last trustworthiness they might have had before it.

That is what I thought.  I thought about it, and did some comparisons, and almost every time, the CPU with the higher clock speed did WAY better.

[Atmos]

8 hours ago, GDRRiley said:

no its 3-4 away and even then the breakdown isn't clear.

they are, 99% of people are going to buy based on rating and nothing else.

if we are taking most consumers, they can do facebook on a core 2 duo with 4gb of ram and an SSD just fine.

I'm not saying threadripper 3990x should be at the top but right now the 3950x should be. if they want to move server and workstations chips out onto their own space fine.

No man, it isnt. Let's not push a narrative.

 

https://www.userbenchmark.com/Faq/What-is-UBM-Effective-Speed/95

 

This is precisely the reasoning that they changed it further and "de-emphasized" ignore* massively multithreaded workloads. For most consumers, that doesnt matter. And again, I dont like that, and at the very least feel they should change the category and add another that would show complete multithreaded performance.

 

2 minutes ago, rx590overclocker said:

That is what I thought.  I thought about it, and did some comparisons, and almost every time, the CPU with the higher clock speed did WAY better.

What were you comparing? I'd like to see for myself, because in my own usage I've yet to find issue except with their seemingly completely arbitrary rankings.

[rx590overclocker]

5 minutes ago, Atmos said:

No man, it isnt. Let's not push a narrative.

 

https://www.userbenchmark.com/Faq/What-is-UBM-Effective-Speed/95

 

This is precisely the reasoning that they changed it further and "de-emphasized" ignore* massively multithreaded workloads. For most consumers, that doesnt matter. And again, I dont like that, and at the very least feel they should change the category and add another that would show complete multithreaded performance.

 

What were you comparing? I'd like to see for myself, because in my own usage I've yet to find issue except with their seemingly completely arbitrary rankings.

[nox_]

Here's a nice example also :

[Atmos]

Just now, rx590overclocker said:

-snip-

Once again... I'm starting to sound like a broken record here.

Their arbitrary "effective speed" and ranking is irrelevant. Look at the other actual categories.

Average benchmark on the 9800x for 8 cores is higher.

Average overclocked benchmark for 8 cores is higher.

Value and sentiment is pointless, and nice to haves are also irrelevant.

 

[Atmos]

7 minutes ago, noxdeouroboros said:

Here's a nice example also :

 

6 minutes ago, Atmos said:

Once again... I'm starting to sound like a broken record here.

Their arbitrary "effective speed" and ranking is irrelevant. Look at the other actual categories.

12% higher weighted average user benchmark on up to 8 cores for the threadripper

3% higher weighted oc benchmark scores on up to 8 cores on the 8350k

value and sentiment is irrelevant.

Nice to haves this time, because you're comparing a cpu with more than 8 cores shows up to 64 core performance, where the threadripper obviously takes off into orbit compared to an i3.

 

And if you actually look into the benchmark scores you see that yes the i3 has marginally higher 1-4 core performance, but then the 8 core performance is handed back to the threadripper. Its all there,

 

 

[nox_]

2 minutes ago, Atmos said:

 

Spoiler

 

12% higher weighted average user benchmark on up to 8 cores for the threadripper

3% higher weighted oc benchmark scores on up to 8 cores on the 8350k

value and sentiment is irrelevant.

Nice to haves this time, because you're comparing a cpu with more than 8 cores shows up to 64 core performance, where the threadripper obviously takes off into orbit compared to an i3.

 

 

You should consider the fact that most people will only look at the top that states that threadripper is overall weaker cpu than i3.

[Atmos]

4 minutes ago, noxdeouroboros said:

You should consider the fact that most people will only look at the top that states that threadripper is overall weaker cpu than i3.

Because for most people. it. is.

Its almost like UBM has literally millions of data points to pull from... or something... and has literally no bias or obligation to try and push an agenda when they aren't promoted, sponsored, or paid by intel, nvidia, amd, or any of them. I dont like it, but thats what it comes down to

Edited March 12, 2020 by Atmos

[LAwLz]

11 hours ago, GDRRiley said:

single core should make up maybe 10-20% of it.

12-16 threaded workload. 60%

highly threaded 20-30%

That would be a much more modern and actuate representation of processors.

Completely disagree.

 

Most programs are still heavily reliant on single threaded scores. Even those who programs who can "use more cores" are often still very reliant on single core scores, or maybe 2-4 in some better cases. By your logic, something like an AMD FX-8150 would probably be fairly competitive with a Ryzen CPU, even though something like quad core Ryzen will run circles around it in 90% of all benchmarks.

 

 

 

I am not sure how the weighted scores work for Userbenchmark but if I were in charge I'd probably put like 60% of the score on 1-4 threaded workloads. 30% on 4-8 threaded workloads and the remaining 10% on 8+ threaded workloads. I think that accurately represent how most applications actually work and scale. But then again, these super high core count CPUs aren't really meant for consumer applications either. So you have to take a stance on who the target audience is. Is it the average consumer? Then I think my weight system is accurate. Is it professionals who use specialized software? Then the weight system should look differently.

 

 

 

 

Also, OP, just because Intel does decently in some tests doesn't mean there is a conspiracy and something illegal about it. I think you need to cool down a bit with the fanboy attitude. No need to foam at the mouth and throw around accusations as soon as AMD isn't at the top of every chart.