HW News - AMD Launches 64-Core CPUs, RX 5700 XT THICC, & Intel 10nm Shipping
video description
Date: 2020-05-06
Related videos
Comments and reviews: 10
Kinomora
As a PC Enthusiast I trusted userbenchmark with a general broad idea of CPU performance _because_, just like you elaborated on, the benchmarks are done on people's hardware that isn't tuned. I want to see the full spectrum. If something is wrong in my system after upgrading my hardware, it'll be pretty easy to see how far from the bell curve you fall. Sure having a few hundred or even a few thousand results from people who have no idea what they're doing are going to skew their results, but generally speaking you can see the spread of results by looking at the bell curve formed by all the compiled results. Another reason that UBM's changes bothered me and many other enthusiasts was because people who don't know what they're doing are going to plug in two CPUs into google, like they do, click the first result, which is UBM, and see oh the i3-9350k is faster than the 3900X that means I should get it instead! or similar bull. Of course we know what makes things different, but when they get their shiny new dual core and it runs like shit in their favorite games, they're gonna be mad. The dumb non-enthusiast doesn't know this, so it's up to the people that do to get upset about this kind of stuff, just like I'd expect a car enthusiast to get upset with a company who's making it seem lower quality components are just as good as high-end ones because they adjusted their quality sliders or whatever. We watch out for each other not because we want to get mad about something, but because we want to see the market be the best it can be and help those without the knowledge that we have buy a product without getting screwed over by accidentally looking at the wrong benchmarking site.
reply
As a PC Enthusiast I trusted userbenchmark with a general broad idea of CPU performance _because_, just like you elaborated on, the benchmarks are done on people's hardware that isn't tuned. I want to see the full spectrum. If something is wrong in my system after upgrading my hardware, it'll be pretty easy to see how far from the bell curve you fall. Sure having a few hundred or even a few thousand results from people who have no idea what they're doing are going to skew their results, but generally speaking you can see the spread of results by looking at the bell curve formed by all the compiled results. Another reason that UBM's changes bothered me and many other enthusiasts was because people who don't know what they're doing are going to plug in two CPUs into google, like they do, click the first result, which is UBM, and see oh the i3-9350k is faster than the 3900X that means I should get it instead! or similar bull. Of course we know what makes things different, but when they get their shiny new dual core and it runs like shit in their favorite games, they're gonna be mad. The dumb non-enthusiast doesn't know this, so it's up to the people that do to get upset about this kind of stuff, just like I'd expect a car enthusiast to get upset with a company who's making it seem lower quality components are just as good as high-end ones because they adjusted their quality sliders or whatever. We watch out for each other not because we want to get mad about something, but because we want to see the market be the best it can be and help those without the knowledge that we have buy a product without getting screwed over by accidentally looking at the wrong benchmarking site.
reply
Dawie
Hi guys, love your content. I think you've got the userbench stuff totally wrong on almost every point that you made though, still I do respect your opinion. I don't use it as my only benchmark, but I still do use it and in many cases it is actually very useful especially in the case of looking at overall system health. The actual benchmarks that run are a bit primitive and it's real world usage is debatable, still it's somewhat comparable especially when looking at averages. My last point would be your stance on user reviews which is fair enough, but I could turn your argument on it's head. No benchmarking suite is perfect, my stance would be that both types are NEEDED. I give the likes of you, hardware unboxed etc much more credence, but I look at a bunch of benchmarks, normally userbench is one of them, also it's actually a quick go to when I upgrade or test where my equipment stack up even compared to identical systems. Sites like userbench are indeed necessary, they just need to improve their weighted score to maybe include 6 and 8 core weights independently and drop the single core and quadcore significantly even if they are pandering to the gaming crowd. In my opinion 8 core should be the new quadcore. Multithread can still be low though atleast for gaming scores. Chears guys and thanks for the hardwork you guys put into everything you do. God Bless.
reply
Hi guys, love your content. I think you've got the userbench stuff totally wrong on almost every point that you made though, still I do respect your opinion. I don't use it as my only benchmark, but I still do use it and in many cases it is actually very useful especially in the case of looking at overall system health. The actual benchmarks that run are a bit primitive and it's real world usage is debatable, still it's somewhat comparable especially when looking at averages. My last point would be your stance on user reviews which is fair enough, but I could turn your argument on it's head. No benchmarking suite is perfect, my stance would be that both types are NEEDED. I give the likes of you, hardware unboxed etc much more credence, but I look at a bunch of benchmarks, normally userbench is one of them, also it's actually a quick go to when I upgrade or test where my equipment stack up even compared to identical systems. Sites like userbench are indeed necessary, they just need to improve their weighted score to maybe include 6 and 8 core weights independently and drop the single core and quadcore significantly even if they are pandering to the gaming crowd. In my opinion 8 core should be the new quadcore. Multithread can still be low though atleast for gaming scores. Chears guys and thanks for the hardwork you guys put into everything you do. God Bless.
reply
EvilMcSheep
UserBenchmark isn't so bad; First of all, it's often the only site to have any data whatsoever on obscure SSDs and flash drives; As unscientific as the tests may be, questionable data is much better than NO data! Guesstimating CPU performance going back 10 years is easy enough, there haven't been enough changes in IPC nor SKU naming to complicate it much at all. For GPUs though, there's just no way that I'd remember the relative performance of every SKU between generations, and compared to modern hardware; Yes, I could look up a bunch of charts, but when they're across a bunch of different sites, where even on the same site, they're testing on different (latest, most popular) software every generation, usually using LAUNCH drivers, not latest/end-of-life drivers, the 'scientifically' collected data becomes no more useful than what UserBenchmark offers. UserBenchmark is certainly not perfect, but how bad do you really expect their data to be? 15%? 25% off? Sure, that's AWFUL by GN standards, but for a quick performance lookup of a 9 year old graphics card, it's certainly 'good enough'.
reply
UserBenchmark isn't so bad; First of all, it's often the only site to have any data whatsoever on obscure SSDs and flash drives; As unscientific as the tests may be, questionable data is much better than NO data! Guesstimating CPU performance going back 10 years is easy enough, there haven't been enough changes in IPC nor SKU naming to complicate it much at all. For GPUs though, there's just no way that I'd remember the relative performance of every SKU between generations, and compared to modern hardware; Yes, I could look up a bunch of charts, but when they're across a bunch of different sites, where even on the same site, they're testing on different (latest, most popular) software every generation, usually using LAUNCH drivers, not latest/end-of-life drivers, the 'scientifically' collected data becomes no more useful than what UserBenchmark offers. UserBenchmark is certainly not perfect, but how bad do you really expect their data to be? 15%? 25% off? Sure, that's AWFUL by GN standards, but for a quick performance lookup of a 9 year old graphics card, it's certainly 'good enough'.
reply
MARTIN0641
The Userbenchmark rant is not accurate imho. Yes, it's not clean data, but thats a good thing from my perspective, its a massive real world aggregate. My system is not a fresh install either, so getting aggregate data from a bunch of messy systems tells me that my 4970K is not faster than the new AMD systems because of the 9 inches of things I have running in the taskbar for work related reasons. Now, if you could add in a messy benchmark with Ninite, and lots of programs in the rear like Spotify, a Youtube video playing on a second monitor in 4k, Google backup, and the kind of REAL stuff people have on their systems, then we could see what performs well in the wild for real hardcore multitaskers. Its like seeing the cars at a local track do runs versus the pristine models that a manufacturer might send to a car review site, I wanna see runs in the real wild world, not a theoretical sterilized lab environment which is the equivalent of a stripped down stock car run with no weight on the chassis. I game with VMs running in the background, like a boss.
reply
The Userbenchmark rant is not accurate imho. Yes, it's not clean data, but thats a good thing from my perspective, its a massive real world aggregate. My system is not a fresh install either, so getting aggregate data from a bunch of messy systems tells me that my 4970K is not faster than the new AMD systems because of the 9 inches of things I have running in the taskbar for work related reasons. Now, if you could add in a messy benchmark with Ninite, and lots of programs in the rear like Spotify, a Youtube video playing on a second monitor in 4k, Google backup, and the kind of REAL stuff people have on their systems, then we could see what performs well in the wild for real hardcore multitaskers. Its like seeing the cars at a local track do runs versus the pristine models that a manufacturer might send to a car review site, I wanna see runs in the real wild world, not a theoretical sterilized lab environment which is the equivalent of a stripped down stock car run with no weight on the chassis. I game with VMs running in the background, like a boss.
reply
mark
i don't agree with your point about user benchmarks, Yes there scores are generated in a un-controlled environment but the facts is that every gpu is tested by at least 1000 systems what makes it fair again. And because the hardware is tested by real users for x around of times in environments that match the casual pc user there environments the benchmarks are actually trustworthy. Not in what you see as trustworthy as in i need to know the difference just between the gpu's and not other hardware but trustworthy as in a 2080ti owner probably also has a top of the line i9 and a budget gamer buys an i3 with a rx580. Trust me they are probebly smart enoth to removed the tests using a celeron bind with rtx titan. And yes there benchmark itself is questionable but as long as the results on there website are trustworthy why not right? I also think everyone need to take the results on there website results with a big grain of salt but that is the case with all benchmarks even on youtube.
reply
i don't agree with your point about user benchmarks, Yes there scores are generated in a un-controlled environment but the facts is that every gpu is tested by at least 1000 systems what makes it fair again. And because the hardware is tested by real users for x around of times in environments that match the casual pc user there environments the benchmarks are actually trustworthy. Not in what you see as trustworthy as in i need to know the difference just between the gpu's and not other hardware but trustworthy as in a 2080ti owner probably also has a top of the line i9 and a budget gamer buys an i3 with a rx580. Trust me they are probebly smart enoth to removed the tests using a celeron bind with rtx titan. And yes there benchmark itself is questionable but as long as the results on there website are trustworthy why not right? I also think everyone need to take the results on there website results with a big grain of salt but that is the case with all benchmarks even on youtube.
reply
v1asec
I've been using UserBenchmark for quite some time. It is hard to find comparisons of parts from different time. It is hard to find benchmarks of less known SSDs. I consider UserBenchmark an estimate rather than a real benchmark. But that can still be useful for quick orientation. That said, the effective weighted averages probably suck in every category (just in the GPUs, it's harder for me to understand the metrics. So, I never really cared about effective score, but the people who are led there by Google may care. And that is probably the reason for the whole outrage, that these people are misled, where it's intentional or not. By overrating i3-9350K, a gamer might get the idea that it is a nice alternative to i7-9700K, being just a little bit slower. It also seems better than Ryzen 5 3600. Adding octa-core metric is probably useful and they should start collecting 16-core data so that they don't have to look like complete idiots again, in a few years from now.
reply
I've been using UserBenchmark for quite some time. It is hard to find comparisons of parts from different time. It is hard to find benchmarks of less known SSDs. I consider UserBenchmark an estimate rather than a real benchmark. But that can still be useful for quick orientation. That said, the effective weighted averages probably suck in every category (just in the GPUs, it's harder for me to understand the metrics. So, I never really cared about effective score, but the people who are led there by Google may care. And that is probably the reason for the whole outrage, that these people are misled, where it's intentional or not. By overrating i3-9350K, a gamer might get the idea that it is a nice alternative to i7-9700K, being just a little bit slower. It also seems better than Ryzen 5 3600. Adding octa-core metric is probably useful and they should start collecting 16-core data so that they don't have to look like complete idiots again, in a few years from now.
reply
Lucas
I understand Steve when he says we should rely on proper methodology and such for performance, and not on UserBenchmark. However, for comparing lower spec parts it's still very much valid, as it's not thst easy to find quick comparisons for those. When was the last time Steve or basically every famous tech reviewer focused on lower spec parts, like i3's, for example? Recently had to assist a family member on choosing a lower-budget laptop for basic use, and although the user might not notice much difference in performance between an i3 and an i5, they also want to make sure they're getting a good cost-effective machine, or even knowing if it's worth choosing an upgrade or not for the money. So in the real world, far from the high-performance top-tier parts, UserBenchmark is very much useful. Maybe Steve, Linus & Co could start benchmarking lower-spec parts. Wouldn't be as nice to see, but would certainly be useful for us mortals.
reply
I understand Steve when he says we should rely on proper methodology and such for performance, and not on UserBenchmark. However, for comparing lower spec parts it's still very much valid, as it's not thst easy to find quick comparisons for those. When was the last time Steve or basically every famous tech reviewer focused on lower spec parts, like i3's, for example? Recently had to assist a family member on choosing a lower-budget laptop for basic use, and although the user might not notice much difference in performance between an i3 and an i5, they also want to make sure they're getting a good cost-effective machine, or even knowing if it's worth choosing an upgrade or not for the money. So in the real world, far from the high-performance top-tier parts, UserBenchmark is very much useful. Maybe Steve, Linus & Co could start benchmarking lower-spec parts. Wouldn't be as nice to see, but would certainly be useful for us mortals.
reply
Rodrigor
Well, I updated to the lastest BIOS, which as today I still not sure if its maybe a beta one, but it is AGESA 1003 ABB, and the PCIE gen 4 option still there. MOBO: Gigabyte B450 Gaming X, Bios F42b (maybr the b is for Beta. i don't know. As for userbenchmark, even though I wouln't recommend a purchase based only on what they ranks shows, is not a horrible site to get another look at some products. Specially components that may not be reviewed by the big boys, lets be honest you guys (GN, HU, LTT, J2C, Biwit, Paul's Hardware, Derbaur, etc) are awesome, but the day has only 24 hours and you cannot review everything the market has to offer. Im not defending the site, or how the test samples/results are taked, by whom, and how they are uploaded to it, but since the results are avergage among thousands or hundreds (in some cases) of samples, they may not be trusted, but they are so so. (Just my humble opinion)
reply
Well, I updated to the lastest BIOS, which as today I still not sure if its maybe a beta one, but it is AGESA 1003 ABB, and the PCIE gen 4 option still there. MOBO: Gigabyte B450 Gaming X, Bios F42b (maybr the b is for Beta. i don't know. As for userbenchmark, even though I wouln't recommend a purchase based only on what they ranks shows, is not a horrible site to get another look at some products. Specially components that may not be reviewed by the big boys, lets be honest you guys (GN, HU, LTT, J2C, Biwit, Paul's Hardware, Derbaur, etc) are awesome, but the day has only 24 hours and you cannot review everything the market has to offer. Im not defending the site, or how the test samples/results are taked, by whom, and how they are uploaded to it, but since the results are avergage among thousands or hundreds (in some cases) of samples, they may not be trusted, but they are so so. (Just my humble opinion)
reply
sputnik13
I'm a fan of GN and love the in depth coverage, however the rant on userbenchmark is rather narrow minded and elitist. I don't use the site myself, but comparing average scores across a large population of samples has value. If you throw that idea out then any and all statistics based inferences (of which there are a LOT) become meaningless. You could argue about the population not being representative but the scores are classified per hardware/component model, so I'm not sure how you could argue that. Looking at their cpu benchmark page, it shows the lowest sample size in the top 10 to be 16k samples. When's the last time GN took 16k samples for a single model of anything? Userbenchmark's business model (SEO) aside, the availability of a large population of measurements that you can draw statistical (average value) data from has value.
reply
I'm a fan of GN and love the in depth coverage, however the rant on userbenchmark is rather narrow minded and elitist. I don't use the site myself, but comparing average scores across a large population of samples has value. If you throw that idea out then any and all statistics based inferences (of which there are a LOT) become meaningless. You could argue about the population not being representative but the scores are classified per hardware/component model, so I'm not sure how you could argue that. Looking at their cpu benchmark page, it shows the lowest sample size in the top 10 to be 16k samples. When's the last time GN took 16k samples for a single model of anything? Userbenchmark's business model (SEO) aside, the availability of a large population of measurements that you can draw statistical (average value) data from has value.
reply
Aljaž
I upgraded one of my computers to SSD a year ago and run UserBenchmark on it. It performed bellow expectation and it asked if I have last gen cable. I didn't, computer was around 10 years old and I havent thought that cable would change, it looked the same and could be used the same. Then I also figured that motherboard doesnt have new gen port, so eventually I upgraded everything, but I wouldnt notice all of this without UserBenchmark. So it has to have some value. I use it when upgrading one component or meddling with BIOS to see which components get hit by a change in the same system. I know there are other benchmarks that are better, but they are also WAY harder to read when just trying to figure out if everything just works as expected.
reply
I upgraded one of my computers to SSD a year ago and run UserBenchmark on it. It performed bellow expectation and it asked if I have last gen cable. I didn't, computer was around 10 years old and I havent thought that cable would change, it looked the same and could be used the same. Then I also figured that motherboard doesnt have new gen port, so eventually I upgraded everything, but I wouldnt notice all of this without UserBenchmark. So it has to have some value. I use it when upgrading one component or meddling with BIOS to see which components get hit by a change in the same system. I know there are other benchmarks that are better, but they are also WAY harder to read when just trying to figure out if everything just works as expected.
reply
Add a review, comment
Other channel videos
