Anandtech's Initial remarks about their unusual Ryzen 2 benchmarks

[App4that]

12 minutes ago, Trixanity said:

They've published their results. Results that are reproducible. They're not wrong but they may not be representative either. Anandtech should not withhold benchmarks because they subjectively dislike the results.

 

I'm sure Ian was far too busy to collude with other reviewers to get the "right" results that you desire. I laugh at the notion of Anandtech being biased. They're practically one of the few who does deep dives. You certainly have a preconceived notion of how things should be and what reviewer does what.

 

I'm sure Anandtech has conspired for several years to make Intel look like the only game in town until 2017. That makes sense. 

 

Sarcasm aside. Intel has issues with HPET and they need to address that. Their engineers seem to be unaware since they didn't tell reviewers to disable it - in fact they were indifferent to whatever settings reviewers use.

 

I would have taken minutes to sanity check, through an email or phone call. 

 

If this had been a review placing Coffee Lake over Ryzen that turned out to be false, people would be calling for blood. This is the bias I speak of. Intel bad, AMD good, or face the consiquences...

[Canada EH]

2 hours ago, Lathlaer said:

without as much formal analysis as we typically like to do

interesting, I can understand why as they wanted a lot of website/video hits to cash in.

How is anyone to believe them going forward!

[The Benjamins]

3 minutes ago, Canada EH said:

interesting, I can understand why as they wanted a lot of website/video hits to cash in.

How is anyone to believe them going forward!

The last ryzen and x299 launches review have been complaining with the lack of time to test the chips fully, with many written reviews still being worked on as they release.

[Canada EH]

8 minutes ago, The Benjamins said:

The last ryzen and x299 launches review have been complaining with the lack of time to test the chips fully, with many written reviews still being worked on as they release.

Top Notch Journalism At Work

 

 

[The Benjamins]

3 minutes ago, Canada EH said:

Top Notch Journalism At Work

They just don't have the time needed, and most reviewers know if they miss the NDA lift time they can lose a lot in ad revenue. companies need to give them ample time to test.

[Canada EH]

Yup

 

Fake News is real.

[App4that]

4 minutes ago, The Benjamins said:

They just don't have the time needed, and most reviewers know if they miss the NDA lift time they can lose a lot in ad revenue. companies need to give them ample time to test.

Everyone else made the launch and didn't make this kind of error. Anandtech, and everyone employeed there has a job. Just like we all have jobs. When we can't do our jobs, we're replaced.

 

Anandtech not only made this mistake, they blamed security updates that have been proven not to impact gaming. Anandtech had multiple chances to catch, or fix the issue but chose to stand behind piss poor methodology. Defending Anandtech will only bring worse coverage in the future. 

[The Benjamins]

1 minute ago, App4that said:

Everyone else made the launch and didn't make this kind of error. Anandtech, and everyone employeed there has a job. Just like we all have jobs. When we can't do our jobs, we're replaced.

 

Anandtech not only made this mistake, they blamed security updates that have been proven not to impact gaming. Anandtech had multiple chances to catch, or fix the issue but chose to stand behind piss poor methodology. Defending Anandtech will only bring worse coverage in the future. 

They started investigating it a hour after the review hit, and I would not trust a review that has to call around to compare. also I would not replace a employee of many years of good standing over 1 simple mistake.

 

and again when you play your games do you make sure HPET is on forced? I know I never new it existed.

 

 

[App4that]

Just now, The Benjamins said:

They started investigating it a hour after the review hit, and I would not trust a review that has to call around to compare. also I would not replace a employee of many years of good standing over 1 simple mistake.

 

and again when you play your games do you make sure HPET is on forced? I know I never new it existed.

 

 

Most reviewers sanity check, that's GOOD methodology, not bad. It's why reviewers keep a tight connection to each other, the point is to find the truth, not what you want to see.

 

No other reviewer had an issue with HPET, no one else. You don't need to understand HPET to know their findings were off, that's what's so fustrating about this. 

[AncientNerd]

2 minutes ago, App4that said:

Most reviewers sanity check, that's GOOD methodology, not bad. It's why reviewers keep a tight connection to each other, the point is to find the truth, not what you want to see.

 

No other reviewer had an issue with HPET, no one else. You don't need to understand HPET to know their findings were off, that's what's so fustrating about this. 

Let's look at it the other way - frankly for a low level controls guy who depends on HPET all the time what Anandtech discovered is frankly really important information to have. It might not be important to gaming but for controls work it makes me really question Intel as a source for CPU's. If there is this big a difference between Intel and AMD with this enabled then I really want to know - now maybe this specific review wasn't the right place to discover it, but it makes me wonder more about how everyone else missed this.

 

You can look at it from the direction of AnanTech discovered something real that every other reviewer missed - because that difference is real with the HPET forced on, and there are real applications where you run with the HPET forced on. So the performance difference will be real.

[App4that]

Just now, AncientNerd said:

Let's look at it the other way - frankly for a low level controls guy who depends on HPET all the time what Anandtech discovered is frankly really important information to have. It might not be important to gaming but for controls work it makes me really question Intel as a source for CPU's. If there is this big a difference between Intel and AMD with this enabled then I really want to know - now maybe this specific review wasn't the right place to discover it, but it makes me wonder more about how everyone else missed this.

 

You can look at it from the direction of AnanTech discovered something real that every other reviewer missed - because that difference is real with the HPET forced on, and there are real applications where you run with the HPET forced on. So the performance difference will be real.

A quick and dirty google search will show you the performance impact of HPET is well understood. This isn't entirely new information. 

 

Which CPU you buy is subjective, tied to everything from priorities to personal preferences. That's not the point of contention, the point of contention is how Anandtech handled being wrong, and how that impacted the hobby. We're talking about quantitative data here, 30>20. If you can't handle doing what it take to make sure that information is accurate, you shouldn't be trying to provide it. It's bad science, made worse by bad excusses. 

[WMGroomAK]

11 minutes ago, AncientNerd said:

You can look at it from the direction of AnanTech discovered something real that every other reviewer missed - because that difference is real with the HPET forced on, and there are real applications where you run with the HPET forced on. So the performance difference will be real.

If I remember from the article correctly, it is also mentioned that overclocking software (e.g.: Ryzen Master) can force the HPET to be used.  That means a reviewer using software unaware that it is enabling the HPET may actually be reporting lower than optimal values (or the benchmarking software is improperly calculating results)...  Would be nice to know what applications are forcing the use of HPET and what the performance impact is in scenarios as well as whether this is being exasperated via the Meltdown/Spectre patches or if it is specifically exasperated via the HPET timer for Coffee Lake being different from how other HPET timers are set. (Specifically looking at page 2 of the article where most of the HPET timers listed show at 14.32 Mhz, while the 8700k timer shows 24.00 Mhz.)

 

Honestly think that there could be a lot of testing done on this as a specific issue and while the way it was brought up is not ideal, it appears to be an important consideration...

[Trixanity]

2 hours ago, App4that said:

I would have taken minutes to sanity check, through an email or phone call. 

 

If this had been a review placing Coffee Lake over Ryzen that turned out to be false, people would be calling for blood. This is the bias I speak of. Intel bad, AMD good, or face the consiquences...

They investigated after. There was no minute. They had scrapped all data and redone it leading up to deadline whilst making a detailed write-up on Pinnacle Ridge.

 

If there is any bias it may just be because Intel is a giant with a checkered past. Any bad review bounces off them, the same cannot be said for AMD. But frankly, the results and analysis vindicates everyone involved. Intel needs to revise its HPET implementation. It's an important find even if those affected are few. We learn more from mistakes and correcting them even more so. The 2018 bench numbers will all be redone in the coming months and will be with HPET off going forward. That's above and beyond what's required. 

 

If you simply want purchase validation then you should probably look elsewhere. I think most just want the in-depth articles instead of a superficial dump of benchmarks.

[AresKrieger]

I thought that review was nonsense glad I didn't consider it legit from the getgo, regardless AMD's fans are starting to irritate me at this point with their nonsense. At least the BS on this particular case has been clear, also I see no reason to look at most sites besides GM it would seem

[App4that]

17 minutes ago, Trixanity said:

They investigated after. There was no minute. They had scrapped all data and redone it leading up to deadline whilst making a detailed write-up on Pinnacle Ridge.

 

If there is any bias it may just be because Intel is a giant with a checkered past. Any bad review bounces off them, the same cannot be said for AMD. But frankly, the results and analysis vindicates everyone involved. Intel needs to revise its HPET implementation. It's an important find even if those affected are few. We learn more from mistakes and correcting them even more so. The 2018 bench numbers will all be redone in the coming months and will be with HPET off going forward. That's above and beyond what's required. 

 

If you simply want purchase validation then you should probably look elsewhere. I think most just want the in-depth articles instead of a superficial dump of benchmarks.

Purchase validation? You know I have a Ryzen system right? LOL

 

Anandtech got it wrong, when no one else did. Anandtech then made the statement it was a security patch that multiple reviewers and testers had proven had no impact on gaming, long before Anandtechs review of Ryzen 2.

 

I make the same fuss when anyone gets it this wrong and tries to cover it up about ANY manufacturer, I'm well within my right to question you and anyone else defending Anandtech if they would say the same if the review had wrongfully placed Intel ahead. Since the whole Con Lake and Steve vs Jim issues literally just happened.

 

The majority of the press are bias against Intel, which on a personal level ios fine. Intel have been dicks, so has AMD, since AMD says I have to run the stock cooler on my 1700X or I lose my warranty. Frantic searching in my 1700X's box is yet to provide such a cooler LOL.

 

Anandtech let their personal bias, influence quanitative data, not cool. 

 

If Anandtech had NOT tried to say it was security updates already proven not to have a performance impact, then it's a case of shit happens. But.

[SpaceGhostC2C]

3 hours ago, WMGroomAK said:

Basically goes back to Ian's overclocking roots and trying to ensure that individual motherboards are not creating an issue...  Definitely encourage reading the whole article as there are a lot of details discussed that may not be considered.  

No one has time for reading! Just tell us where to take our pitchforks already!

[Trixanity]

2 minutes ago, App4that said:

Purchase validation? You know I have a Ryzen system right? LOL

 

Anandtech got it wrong, when no one else did. Anandtech then made the statement it was a security patch that multiple reviewers and testers had proven had no impact on gaming, long before Anandtechs review of Ryzen 2.

 

I make the same fuss when anyone gets it this wrong and tries to cover it up about ANY manufacturer, I'm well within my right to question you and anyone else defending Anandtech if they would say the same if the review had wrongfully placed Intel ahead. Since the whole Con Lake and Steve vs Jim issues literally just happened.

 

The majority of the press are bias against Intel, which on a personal level ios fine. Intel have been dicks, so has AMD, since AMD says I have to run the stock cooler on my 1700X or I lose my warranty. Frantic searching in my 1700X's box is yet to provide such a cooler LOL.

 

Anandtech let their personal bias, influence quanitative data, not cool. 

 

If Anandtech had NOT tried to say it was security updates already proven not to have a performance impact, then it's a case of shit happens. But.

Tinfoil engaged?

 

There is no cover up. Anandtech never made any official explanation prior to today. Everything else have been speculation on their part. Public speculation but that'd obvious considering the amount of feedback and questions they've received.

 

You may question as much as you want if there is basis for it. There isn't. You're basically saying Anandtech is straying from their well-worn path yet there is no evidence to support it except your baseless accusations pulled out of nowhere. 

 

From reading your post it's clear to me why you're spouting nonsense since you're parroting other conspiracies. There is no bias against Intel though. Intel has the most power in the industry and you don't bite the hand that feeds you. They may not give them any favors but they're not plotting against them either. Either everyone is in on it or no one is since it's relatively easy to spot as it'd stick out like a sore thumb. And given we're discussing this, someone diverged and later corrected it when the work was complete. Conspiracy averted.

 

But please stick to the facts otherwise my toes may just curl up so much that I'll never be able to walk right again.

[WMGroomAK]

20 minutes ago, App4that said:

AMD, since AMD says I have to run the stock cooler on my 1700X or I lose my warranty. Frantic searching in my 1700X's box is yet to provide such a cooler LOL

From what I've read on this, that was on an old FAQ that had not been updated in 5+ years...

 

http://www.guru3d.com/news-story/your-ryzen-processor-warranty-states-you-may-only-use-the-stock-cooler.html

Quote

We got word back from AMD, and much of the staff is as surprised to find out what is written there. AMD mentioned that the FAQ and info distributed is old (5 years) and is being updated soon to clarify things. 

 

They have since updated this to basically if the Heatsink/Fan is not designed to support the operation of AMD processors to be within their conformance.

https://www.pcper.com/news/Processors/AMD-Clarifies-Warranty-Terms-Ryzen-CPU-use-Third-Party-Coolers

Quote

As it turns out, this was an older support page that does not accurately reflect the warranty of modern AMD processors. AMD has since updated the warranty page to provide clarification.

 

Now, the page reads that the warranty shall be null and void if the processor "is used with any heatsink/fan (HSF) that does not support operation of the AMD processor in conformance with AMD’s publicly available specifications."

https://support.amd.com/en-us/warranty/rma/terms/pib

Quote

This Limited Warranty shall be null and void if the AMD microprocessor which is the subject of this Limited Warranty is used with any heatsink/fan (HSF) that does not support operation of the AMD processor in conformance with AMD’s publicly available specifications.  Use of HSF solutions determined by AMD as incapable of such performance or which are determined to have contributed to the failure of the processor shall invalidate the warranty.

 

[App4that]

9 minutes ago, Trixanity said:

Tinfoil engaged?

 

There is no cover up. Anandtech never made any official explanation prior to today. Everything else have been speculation on their part. Public speculation but that'd obvious considering the amount of feedback and questions they've received.

 

You may question as much as you want if there is basis for it. There isn't. You're basically saying Anandtech is straying from their well-worn path yet there is no evidence to support it except your baseless accusations pulled out of nowhere. 

 

From reading your post it's clear to me why you're spouting nonsense since you're parroting other conspiracies. There is no bias against Intel though. Intel has the most power in the industry and you don't bite the hand that feeds you. They may not give them any favors but they're not plotting against them either. Either everyone is in on it or no one is since it's relatively easy to spot as it'd stick out like a sore thumb. And given we're discussing this, someone diverged and later corrected it when the work was complete. Conspiracy averted.

 

But please stick to the facts otherwise my toes may just curl up so much that I'll never be able to walk right again.

The facts are there, I've repeatedly posted them. Anandtech make a major mistake, one no other reviewer made. Anandtec failed to sanity check. When cornered with their benchmarks not matching any other, Anandtech went the route of blaming security patches already proven to have no impact on performance.

 

So, as the outside observer, I have two choices. Find Anandtech incompetent, or bias. 

 

Which is worse?

[App4that]

6 minutes ago, WMGroomAK said:

From what I've read on this, that was on an old FAQ that had not been updated in 5+ years...

 

http://www.guru3d.com/news-story/your-ryzen-processor-warranty-states-you-may-only-use-the-stock-cooler.html

 

They have since updated this to basically if the Heatsink/Fan is not designed to support the operation of AMD processors to be within their conformance.

https://www.pcper.com/news/Processors/AMD-Clarifies-Warranty-Terms-Ryzen-CPU-use-Third-Party-Coolers

https://support.amd.com/en-us/warranty/rma/terms/pib

 

You, kinda proved my point. I was being facetious about the cooler, but if AMD is accused of anything people work to remove the threat.

 

If Intel is accused of anything, people rush in pitchforks and torches at the ready.

 

I'm not defending Intel, I'm calling out the bias. 

[SpaceGhostC2C]

26 minutes ago, Trixanity said:

There is no cover up. Anandtech never made any official explanation prior to today. Everything else have been speculation on their part. Public speculation but that'd obvious considering the amount of feedback and questions they've received.

Indeed. They were cautious from the beginning. They immediately set out to find the source of discrepancies. They found them. They reported them, they made a decision about which methodology to use, and they're adjusting the results accordingly.

And, in case people forget, they didn't even make a mistake. All of their results are accurate. The simply were obtained under conditions which, with what they know now, they don't consider to be the most useful, hence why they are updating their numbers to reflect what they believe is more relevant.

To me, it is flawless.

 

Other things to consider:

• Other reviewers didn't do differently "because they knew better". They were as unaware as Anandtech about the HPET issues. It wasn't a conscious decisions, it's just that they run their tests differently.
• Other reviewers have reported biased results as well. In particular, any reviewer who used Ryzen Master before running their benchmarks is reporting Ryzen 2 numbers with HPET on. Luckily, the effect on Ryzen 2 wasn't as pronounced as in Intel. Still, in light of these results, everyone would want to double-check their testing conditions. The reason why they didn't have the same problem with Intel is that there is no obvious way to accidentally force HPET in that case, as with Ryzen Master.
• What happened to Anandtech is the same as what happened at Coffee Lake launch with MCE. Same as now, there were discrepancies across reviewers. Serious reviewers investigated those results (of course, after publishing them), narrowed it down to MCE, and barely anyone updated any review. Not too long ago, Gamers Nexus did a breakdown on the effect of MCE in testing (speaking of "almost a week" being "too long"...). Again, that's what serious reviewers do: run tests, report performance to the best of their knowledge, inspect oddities, correct what needs correction and educate their audience in the process.
• People thinking that rejecting data when it contradicts your expectations is the right way to go about things because "this time it would have made me right" is like people saying it's best to disregard traffic lights because "last time I crossed the street that way nothing happened and I arrived more quickly". Going with their data, whatever their preconceptions could be, was the most honest thing they, or anyone, could do. Just like reporting the results of their investigation now.
• Some people are also going hyperbolic about the "huge misleading impact" of their review. Everyone was saying CL is the leader for high framerate gaming on April 18, and everyone was saying CL is the leader for high framerate gaming on April 20. Stop looking for excuses to be outraged.

[Brooksie359]

5 hours ago, App4that said:

Everyone is out for those advertising dollar, but Steve at Gamers Nexus understands his responsibity, and does a sanity check if results are outside expectation.

 

You're saying Anandtech has no one they can talk to, compare results? That's silly. They chose to run with the results. 

They did but thought it was due to the meltdown patch. Obviously they now know the culprit. It's not like their results were inaccurate. They were just done with a different setting than others. The main thing to note here is they did investigate and figured it out. 

[App4that]

5 minutes ago, Brooksie359 said:

They did but thought it was due to the meltdown patch. Obviously they now know the culprit. It's not like their results were inaccurate. They were just done with a different setting than others. The main thing to note here is they did investigate and figured it out. 

They couldn't have thought it was the meltdown patch, if they did it shows a level of incompetence I'm not sure should be possible.

 

Them saying it could be the meltdown patch, it the reason it's obvious they are biased. 

 

 

[mr moose]

This is just a classic example of how monitoring different test results can prove individual test efficacy.   Well done to Anandtech for looking into it and finding the fault.  If the entire industry keeps doing this the only outliers will be the shills.  And I like it when they are easy to spot.

[Brooksie359]

4 minutes ago, App4that said:

They couldn't have thought it was the meltdown patch, if they did it shows a level of incompetence I'm not sure should be possible.

 

Them saying it could be the meltdown patch, it the reason it's obvious they are biased. 

 

 

Hey if you had results that were different than other had been up for long periods of time double checking and couldn't figure out what it is you might just end up thinking the only possible explanation is that it was because of the meltdown patch. I mean you are picking them apart for making a mistake that in all honesty most of us wouldn't have been able to diagnose and if we did it would take a long time. You act like a they aren't people but benchmarking machines.