Anandtech's Initial remarks about their unusual Ryzen 2 benchmarks

[mr moose]

3 minutes ago, App4that said:

Following scientific method would have saved Anandtech in this case, they would have contacted other reviewers and compared methodology, this would have made the posibility of finding the cause of the disparity in results happen before any drama could have happened much higher.

They did.  They all post their results at the same time.  No scientists call other scientists and say hey, I've just a donme a test and the results are weird, they publish the results and wait for correction.  That is how science works.

 

3 minutes ago, App4that said:

I've said that LOL. Lost count in how many times I've said that. If you find something that goes against expectations, you sanity check. Also why we have peer review. 

It's a benchmark review not an article for nature. Who is going to offer their time to peer review a review before deadline?  please that is just taking a an almost moot situation to way off the chart.

 

3 minutes ago, App4that said:

Yeah, in hindsight it was a misstep to bring up the AMD bias. Puts people in an emoptional state, something I'm not immune to so should expect anything different from others. 

Because biases and fanboyism is on every side and it doesn't help the conversation. 

[App4that]

7 minutes ago, mr moose said:

They did.  They all post their results at the same time.  No scientists call other scientists and say hey, I've just a donme a test and the results are weird, they publish the results and wait for correction.  That is how science works.

 

It's a benchmark review not an article for nature. Who is going to offer their time to peer review a review before deadline?  please that is just taking a an almost moot situation to way off the chart.

 

Because biases and fanboyism is on every side and it doesn't help the conversation. 

It's how science works, on every level I've been exposed to. Do you know how many conferences we go to each year for that very reason? Let alone trying to get published without being supported. 

 

Is Anandtech bound to the same rules I and most the people I know live by? Hell no. They are paid to write articles people want to read. Doesn't mean you can't use the scientific method to show what went wrong with how Anandtech handled this launch. By showing if Anandtech had followed the scientific method, they wouldn't have made these mistakes, you show how the scientific method could be use to improve how the media and reviewers handle these launches. 

[mr moose]

1 minute ago, App4that said:

It's how science works, on every level I've been exposed to. Do you know how many conferences we go to each year for that very reason? Let alone trying to get published without being supported. 

 

Is Anandtech bound to the same rules I and most the people I know live by? Hell no. They are paid to write articles people want to read. Doesn't mean you can't use the scientific method to show what went wrong with how Anandtech handled this launch. By showing if Anandtech had followed the scientific method, they would have made these mistakes, you show how the scientific method could be use to improve how the media and reviewers handle these launches. 

You are now trying to argue that they can use the scientific method to mitigate issues in the publication of reviews that are timeline dependent.  Journal publications take as long as they need,  but reviews have to be out on the hour regardless.  The scientific method did work.  The consensus highlighted and issue, the author investigated and corrected the issue.  End of story.

 

 

[Razor01]

5 minutes ago, App4that said:

It's how science works, on every level I've been exposed to. Do you know how many conferences we go to each year for that very reason? Let alone trying to get published without being supported. 

 

Is Anandtech bound to the same rules I and most the people I know live by? Hell no. They are paid to write articles people want to read. Doesn't mean you can't use the scientific method to show what went wrong with how Anandtech handled this launch. By showing if Anandtech had followed the scientific method, they wouldn't have made these mistakes, you show how the scientific method could be use to improve how the media and reviewers handle these launches. 

Did they know about the error prior to testing or when creating their testing methodology?  If not they did just fine, if they did then you have a point.  I don't think they knew about the issue.  Results may have came out unexpected but they do retest to verify those results.  After retesting and getting the same results, they thought they were correct but still not knowing there was a fault in the testing methodology.

[App4that]

Just now, mr moose said:

You are now trying to argue that they can use the scientific method to mitigate issues in the publication of reviews that are timeline dependent.  Journal publications take as long as they need,  but reviews have to be out on the hour regardless.  The scientific method did work.  The consensus highlighted and issue, the author investigated and corrected the issue.  End of story.

 

 

You miss that Steve at Gamers Nexus didn't have this issue, and the reason is Steve at Gamers Nexus sanity checks. He follows the scientific method and stays on schedule. He had a leg up on this launch, but in past launches he's checked with other reviewers before publishing. Because he has a background in the sciences. 

 

I get it, I picked on Anandtech. I promise you, they have zero fucks I did and for good reason lol.

[mr moose]

10 hours ago, App4that said:

You miss that Steve at Gamers Nexus didn't have this issue, and the reason is Steve at Gamers Nexus sanity checks. He follows the scientific method and stays on schedule. He had a leg up on this launch, but in past launches he's checked with other reviewers before publishing. Because he has a background in the sciences. 

 

I get it, I picked on Anandtech. I promise you, they have zero fucks I did and for good reason lol.

I did not miss anything. Steve was not the only one who didn't have issue because  again:  the consensus highlighted an issue and the testing was revised to find and correct the error, said correction was published. it is quite evident to people there is nothing going on here so this is the last post I am making on the topic. 

[asus killer]

back in February GN had already called this, so it's stupid to be falling for this in April.

[App4that]

7 hours ago, mr moose said:

I did not miss anything. Steve was not the only one who didn't have issue because  again:  the consensus highlighted an issue and the testing was revised to find and correct the error, said correction was published. it is quite evident to people there is nothing going on here so this is the last post I am making on the topic. 

By people, you mean anyone who agree's with you. Interesting. Since Steve covered the issue before. 

[WMGroomAK]

8 hours ago, asus killer said:

back in February GN had already called this, so it's stupid to be falling for this in April.

 

55 minutes ago, App4that said:

Since Steve covered the issue before

I'm fairly sure that the Windows Sleep bug and this are separate issues (that may have a tangential relation) since the Sleep Bug was showing increased performance on AMD systems whereas this bug is showing both AMD and Intel having decreased performance with a markedly larger performance decrease on Intel vs AMD.  

[App4that]

1 minute ago, WMGroomAK said:

 

I'm fairly sure that the Windows Sleep bug and this are separate issues (that may have a tangential relation) since the Sleep Bug was showing increased performance on AMD systems whereas this bug is showing both AMD and Intel having decreased performance with a markedly larger performance decrease on Intel vs AMD.  

It's an issue of not trusting odd results, not the specific bug. The point of contention is that Anandtech not only published the results that went against our knowledge of both companies and gaming performance, but then blamed security patches rather than considering it was their methodology. 

[n4da]

Anybody with a shred of brainmatter left would have already known that the 8700K cannot lose so much performance overnight. Just like anyone that actually has a 8700k/Coffee Lake CPU can test for themselves how disastrous the HPET is for framerates and frametimes. It's going from silky smooth to a stuttering mess, it's not just some performance loss.

 

I am glad however to see the matter straightened out. Those benchmarks looked ridiculously dumb. I would have never published them before understanding WHY it is just my website that has different results.

 

Also Intel are barking mad with their recommendations for HPET. I hope they change that crap for their own good, if they care about good press.