AM4 mobo's BIOS performance bias?!

[Tataffe]

As long as there's no proof that this feature will be absent on the consumer version of the BIOS or that this version of the BIOS will be unavailable for consumers respectively, I'll jump to the conclusion that this is a feature for OCers. It is obviously not about skewing benchmarks if it's present in such an open way ...

 

Also concerning the NDA end time: It's 15:00 UTC (3 pm). Seriously what's so hard about googling a time zone converter ...

[leadeater]

2 hours ago, Dabombinable said:

Here's what it does: https://www.manualslib.com/manual/949421/Supero-X9srh-7f.html?page=84

Is it actually this though? I mean that is a manual for an Intel motherboard. Those a pretty standard power options on servers and we put all our ESXi hosts to high power/maximum to ensure good performance since we pack in a fair amount of VMs per host.

 

I think those power bias settings for Ryzen might do a bit more than simple power management control, fine control over boost and smart logic were pretty big PR points for AMD in their presentations.

[leadeater]

28 minutes ago, zMeul said:

these settings are tailored specifically for benchmarks, not for regular loads - they even bothered to name them as such

Supply your evidence of that other than the name of the feature.

 

28 minutes ago, zMeul said:

it being available on ASUS' site matters how!?

It matters since you keep bringing up the custom reviewers bios that consumers did not get, if it doesn't matter then why bring it up?

[Urishima]

9 hours ago, MakeAMDGreatAgain said:

Okay, so now I kind of get what it means. This means we should really wait for multiple, and I mean MUTIPLE independent benchmarks.

As always.

 

Seriously people, take a chill pill. Or smoke some weed if that's what it takes to calm your nerves.

[zMeul]

6 minutes ago, leadeater said:

Supply your evidence of that other than the name of the feature.

 

It matters since you keep bringing up the custom reviewers bios that consumers did not get, if it doesn't matter then why bring it up?

1. yeah hey named the options exactly that way by mistake 
2. "you keep bringing up the custom reviewers bios that consumers did not get" - where did I do that!? or you're talking about the parels made to the video card review samples?
3. can you provide guarantee that reviewers. other than Canard PC, are actually aware of the settings and will provide unbiased benchmarks? - so far I haven't seen anyone else even willing to discuss this; surely WCCFTech has jumped on this .. oh wait 

but no, let's wait for biased benchmarks

[AlTech]

10 hours ago, Godlygamer23 said:

I'd advise adding a translated version of the Tweet as many of us cannot speak French. Yes we could translate it manually, but it'd be more convenient to have it ready for us.

Rough Translation:

 

"Get the popcorn, it's going to be fun to test Ryzen. A small before taste with the previous Asus Press BIOS. Auto by default of course"

[Misanthrope]

4 minutes ago, leadeater said:

Supply your evidence of that other than the name of the feature.

Why are you demanding evidence? This is only an image you have no idea what the setting does but the name implies it was done to help the test. The burden of proof is on you since you're basically making the extraordinary claim that the name means nothing without evidence. You can't request evidence to prove a negative your argument is as bad as the people who say "Prove to me that god doesn't exists!"

 

[leadeater]

Just now, zMeul said:

1. yeah hey named the options exactly that way by mistake 
2. "you keep bringing up the custom reviewers bios that consumers did not get" - where did I do that!? or you're talking about the parels made to the video card review samples?

Text is not evidence, supply evidence of it skewing benchmarks or wait for reviews. Name alone means nothing.

9 hours ago, zMeul said:

because that's exactly what happened when GTX10 series AIB cards launched - MSI and ASUS sent reviewers special cards clocked higher than retail ones

let's not pretend it's not an issue

Exactly would mean the bios is not available to everyone...

[zMeul]

8 minutes ago, leadeater said:

Exactly would mean the bios is not available to everyone...

that's an example of skewed benchmarking that was discovered only after reviewers published the launch benchmarks

 

the only difference is this was outed before the benchmarks are out

but this does not guarantee the launch benchmarks are not skewed

 

Quote

Text is not evidence

right? so what you'd like for this to be evidence? a signature from ASUS CEO?!

compared to Demerjian's allegations (who didn't provide even a screenshot) picked up by WCCFTech, this is actual evidence 

 

what's next? don't tell me .. Canard PC fabricated the screenshot 

in case that does happen, here's more: https://www.reddit.com/r/Amd/comments/5x04ll/ryzen_1700x_cpuz_more_tests_coming/

 

Edited March 2, 2017 by zMeul

[leadeater]

4 minutes ago, Misanthrope said:

Why are you demanding evidence? This is only an image you have no idea what the setting does but the name implies it was done to help the test. The burden of proof is on you since you're basically making the extraordinary claim that the name means nothing without evidence. You can't request evidence to prove a negative your argument is as bad as the people who say "Prove to me that god doesn't exists!"

 

Actually no that is my point, it is only an image. If you are going to say it's skewing benchmarks then prove it. Very interesting to see it there, would be nice to know actually what it does. Wouldn't it be nice if there were reviews we could look at to find out?

 

You don't get to go round saying someone or a company is guilty of anything without proof. You make a claim you show it, very simple.

[Dabombinable]

4 minutes ago, Misanthrope said:

Why are you demanding evidence? This is only an image you have no idea what the setting does but the name implies it was done to help the test. The burden of proof is on you since you're basically making the extraordinary claim that the name means nothing without evidence. You can't request evidence to prove a negative your argument is as bad as the people who say "Prove to me that god doesn't exists!"

 

I already supplied that for him.

16 minutes ago, leadeater said:

Is it actually this though? I mean that is a manual for an Intel motherboard. Those a pretty standard power options on servers and we put all our ESXi hosts to high power/maximum to ensure good performance since we pack in a fair amount of VMa per host.

 

I think those power bias settings for Ryzen might do a bit more than simple power management control, fine control over boost and smart logic were pretty big PR points for AMD in their presentations.

Name at least 1 other motherboard manufacturer that has a different definition. At least1. And in this scenario its designed to only trigger for certain benchmarks thus skewing results, instead of all benchmarks/programs.

And power management settings can affect performance-there is a noticeable difference in the way things run with my Z97 Sabertooth MKII set to; power saving, normal and performance.

[LAwLz]

Some thoughts after reading this thread...

 

1) We don't know what this feature does. We should remain very skeptical but don't jump to conclusions either.

 

2) I am 99% sure this is NOT the same type of power settings some people in this thread are linking. Power settings are generally put into ~3 different categories. These are settings which will automatically detect which program you are running and change something according to that. Power settings don't do that, at least not to this degree (and only applies them to benchmarks).

 

3) We don't know if this is AMD or Asus doing. I think we should have our pitchforks out, but we should not stab anyone yet because we don't know who is guilty.

 

4) Regardless of what you think this does, having custom tailored settings which changes things based on which benchmark you are running (remember, this will detect when you launch one of the specific benchmarks) is bad. It is deliberately made to enhance the scores of benchmarks while simultaneously not increasing the scores of regular programs (because those won't get automatically detected).

 

5) Wow some people sure are quick to defend AMD...

 

6) Wow some people sure are quick to attack AMD...

 

7) This would not be the first time hardware manufacturers are being dishonest with benchmarks. For example back in 2001 ATI were caught having an "optimization" feature for Quake 3. What happened was that their driver was checking if you were running "Quake 3.exe". As soon as it detected that specific program running, it overwrote the settings you had set in game, and applied things such as lower texture quality. What this meant was that in benchmarks where reviewers were running Nvidia and ATI cards against each other at max settings, it was actually like running the Nvidia card at max and the ATI card at medium (or whatever it was), thus making the ATI card look like it performed better.

Simply renaming the .exe file was enough to make the detection fail, and you got the actual texture quality you had selected in the menus.

 

8) Can we please stop with all the bullshit attacks at zMeul? I get that he is disliked around here, but I have seen several posts in this thread which doesn't even talk about the topic. They are just 100% dedicated to attacking zMeul. The people who make posts like that are, in my eyes, far worse than zMeul is because at least he contributes something to the topic. You on the other hand, is just making the thread shittier without advancing the conversation further. Stop it.

 

9) When reviews comes out, people should keep this in mind. If the reviewer doesn't specify that they had this setting turned off (at least on this motherboard), then the scores might be slightly higher than what they actually should be.

 

10) It being available in the consumer version of the BIOS should not matter. It's not that the setting exists that's the problem, it's that it applies special settings to artificially increase benchmark scores that's the problem. Benchmarks exists so that you can compare different products. If one company is trying to artificially increase their scores then it is no longer a fair comparison.

[Misanthrope]

12 minutes ago, leadeater said:

Actually no that is my point, it is only an image.

 

OP had a link to wheere some versions of the BIOS could be previously found. They're not there any more but it's at least 2 images.

 

EDIT: Not a link an image.

 

Quote

If you are going to say it's skewing benchmarks then prove it. Very interesting to see it there, would be nice to know actually what it does. Wouldn't it be nice if there were reviews we could look at to find out?

 

That's a straw man: The BIOS setting doesn't has to be nefarious or unfairly skewing the benchmark. Some people are claiming it's suspect but nobody is claiming they know exactly what it does. What we know is that the name, if it turns out to be more than a fabrication, very clearly implies it addresses some kind of issue with the benchmark. The legitimacy of the setting or whenever or not it was fair to include is not into question, the only concern I have is that the name implies that the leaked benchmarks (and possibly even "review" benchmarks after reviewers reveal which BIOS was used) might not be indicative of the final product performance tests.

 

That doesn't means that the setting unfairly skews in favor of a better benchmark, it doesn't means it actually won't have those settings by default on the final product at launch, etc. What you're saying is that you definitively need to show evidence that the name means nothing, which is a fallacious request for Burden of Proof that's on you: The name definitively means it's doing something to aid the test, nothing more.

 

You should have just stated that you wanted evidence that the setting skews negatively or unfairly but instead you're using a blanket term to discredit all concerns unfairly. You're not being skeptical enough of AMD.

[zMeul]

@leadeater

[leadeater]

Just now, Dabombinable said:

And in this scenario its designed to only trigger for certain benchmarks thus skewing results, instead of all benchmarks/programs.

That's what the name of the setting implies yes, but do we actually know if it ONLY triggers on that one benchmark. That is rather unlikely since the bios does not have that kind of access in to the OS and CPU logic is not designed to identify specific applications and boost only for that singular one.

 

Bios's have been isolated from OS's for a very long time for virus and malware protection. If you think this has changed I'd be more worried about system compromises than skewed benchmarks.

 

They can look at usage patterns, cache utilization and other factors to guess that there is a benchmark running and react to it. But then of course you are having to pre-define the testing being run so there can't be too much smarts around it which comes back to my point that those same setting may be useful outside of the named value of that text field in the bios.

 

Those settings obviously help the named benchmarks other wise why name them as such, that does not mean it only helps those and nothing else and that is what I am saying. We don't know that this is the case, all we can be reasonably confident of is those settings do help the named benchmark and we don't even know by how much.

 

What if the Aida/Geekbench setting helps games that only use 4 cores? I'd say that setting would be rather useful outside of the named benchmarks if that was the case.

 

4 minutes ago, Dabombinable said:

And power management settings can affect performance-there is a noticeable difference in the way things run with my Z97 Sabertooth MKII set to; power saving, normal and performance.

Yep never said anything counter to this and even gave an example of how I use it every day in a business environment to increase performance.

 

13 minutes ago, Dabombinable said:

I already supplied that for him.

You gave me a link to a Supermicro motherboard with an Intel chipset for an Intel CPU, what exactly is that evidence of for this AMD system?

[Dabombinable]

2 minutes ago, zMeul said:

@leadeater

Yep. Just like my Z97 Sabertooth MKII (although I have to manually set the power mode-especially since I'm using the AIsuite version that goes with my H87M Pro)

[zMeul]

33 minutes ago, leadeater said:

That's what the name of the setting implies yes, but do we actually know if it ONLY triggers on that one benchmark. That is rather unlikely since the bios does not have that kind of access in to the OS and CPU logic is not designed to identify specific applications and boost only for that singular one.

BIOS setting on my mobo

from my mobo's manual: http://www.gigabyte.com/Motherboard/GA-H170-D3H-rev-10#support-manual

Quote

3DMark01 Enhancement
Allows you to determine whether to enhance some legacy benchmark performance. (Default: Disabled)

 

Quote

Bios's have been isolated from OS's for a very long time for virus and malware protection. If you think this has changed I'd be more worried about system compromises than skewed benchmarks.

except it's UEFI and UEFI communicates with the OS and vice-versa 

Edited March 2, 2017 by zMeul

[Dabombinable]

1 minute ago, leadeater said:

That's what the name of the setting implies yes, but do we actually know if it ONLY triggers on that one benchmark. That is rather unlikely since the bios does not have that kind of access in to the OS and CPU logic is not designed to identify specific applications and boost only for that singular one.

 

Bios's have been isolated from OS's for a very long time for virus and malware protection. If you think this has changed I'd be more worried about system compromises than skewed benchmarks.

 

They can look at usage patterns, cache utilization and other factors to guess that there is a benchmark running and react to it. But then of course you are having to pre-define the testing being run so there can't be too much smarts around it which comes back to my point that those same setting may be useful outside of the named value of that text field in the bios.

 

Those settings obviously help the named benchmarks other wise why name them as such, that does not mean it only helps those and nothing else and that is what I am saying. We don't know that this is the case, all we can be reasonably confident of is those settings do help the named benchmark and we don't even know by how much.

 

What if the Aida/Geekbench setting helps games that only use 4 cores? I'd say that setting would be rather useful outside of the named benchmarks if that was the case.

 

Yep never said anything counter to this and even gave an example of how I use it every day in a business environment to increase performance.

 

You gave me a link to a Supermicro motherboard with an Intel chipset for an Intel CPU, what exactly is that evidence of for this AMD system?

Still didn't answer " Name at least 1 other motherboard manufacturer that has a different definition. At least1. " There is no contrary evidence at all that says "performance bias" means anything other than the definition in Supermicro's manual.

[Dabombinable]

4 minutes ago, leadeater said:

That's what the name of the setting implies yes, but do we actually know if it ONLY triggers on that one benchmark. That is rather unlikely since the bios does not have that kind of access in to the OS and CPU logic is not designed to identify specific applications and boost only for that singular one.

 

Bios's have been isolated from OS's for a very long time for virus and malware protection. If you think this has changed I'd be more worried about system compromises than skewed benchmarks.

Pst. I can overclock my i7 4790K using Ai suite or Intel XTU and change a lot of settings to do so-including power related settings, and Windows communicates with the BIOS to manage power BTW.

[LAwLz]

3 minutes ago, leadeater said:

That is rather unlikely since the bios does not have that kind of access in to the OS and CPU logic is not designed to identify specific applications and boost only for that singular one.

Correct me if I am wrong, but I am pretty sure Asus actually has several programs which allows you to change BIOS settings from within Windows. If that is the case, which I am fairly sure it is, then the OS and BIOS can communicate.

The fact that there is an auto setting, which will detect which benchmark is running, is very worrying if you ask me.

 

4 minutes ago, leadeater said:

They can look at usage patterns, cache utilization and other factors to guess that there is a benchmark running and react to it. But then of course you are having to pre-define the testing being run so there can't be too much smarts around it which comes back to my point that those same setting may be useful outside of the named value of that text field in the bios.

The benchmarks listed seems way too specific to actually be "it's just general settings being changed". It even lists two different versions of Cinebench.

 

6 minutes ago, Dabombinable said:

Yep. Just like my Z97 Sabertooth MKII (although I have to manually set the power mode-especially since I'm using the AIsuite version that goes with my H87M Pro)

Again, this is not like the "power mode" and "performance mode" you got on your motherboard. Those settings do not automatically change, nor do they detect which programs you are running. Those are static and won't try and differentiate between Cinebench 11.5 and Geekbench or Cinebench 15.

[leadeater]

34 minutes ago, zMeul said:

what's next? don't tell me .. Canard PC fabricated the screenshot

Your missing the point, show me it effecting the benchmark and only the benchmark and no other application.

 

7 minutes ago, Dabombinable said:

Still didn't answer " Name at least 1 other motherboard manufacturer that has a different definition. At least1. " There is no contrary evidence at all that says "performance bias" means anything other than the definition in Supermicro's manual.

See your still not getting my point, it's not about the meaning of performance bias it's about the claim that those setting only effect the named benchmarks in an attempt to skew reviews and make Ryzen look better than it actual is in real work application.

 

Those setting will influence those benchmarks, the else have them. It is rather unlikely it will only have an effect on the set benchmark and nothing else while it is in effect.

[Bouzoo]

15 minutes ago, zMeul said:

@leadeater

Interesting. Can't wait for TPU, Guru3D and Anandtech to strip them open. 

[Dabombinable]

Just now, LAwLz said:

Correct me if I am wrong, but I am pretty sure Asus actually has several programs which allows you to change BIOS settings from within Windows. If that is the case, which I am fairly sure it is, then the OS and BIOS can communicate.

The fact that there is an auto setting, which will detect which benchmark is running, is very worrying if you ask me.

 

The benchmarks listed seems way too specific to actually be "it's just general settings being changed". It even lists two different versions of Cinebench.

 

Again, this is not like the "power mode" and "performance mode" you got on your motherboard. Those settings do not automatically change, nor do they detect which programs you are running. Those are static and won't try and differentiate between Cinebench 11.5 and Geekbench or Cinebench 15.

Performance bias individually adjust settings on the fly (when set to automatic that is) to optimise performance. It has a similar to the individual modes on my motherboard-except those are all manual and work for all programs.
 

 

1 minute ago, leadeater said:

Your missing the point, show me it effecting the benchmark and only the benchmark and no other application.

 

See your still not getting my point, it's not about the meaning of performance bias it's about the claim that those setting only effect the named benchmarks in an attempt to skew reviews and make Ryzen look better than it actual is in real work application.

 

Those setting will influence those benchmarks, the else have them. It is rather unlikely it will only have an effect on the set benchmark and nothing else while it is in effect.

Those 3 benchmarks are the only things mentioned other than "auto" under performance bias. That along with every other menu on any other motherboard listing all possible options alongside "auto" and "disabled" within the same menu. Also taking into consideration that reviewers are starting to notice as well....

[leadeater]

2 minutes ago, LAwLz said:

Correct me if I am wrong, but I am pretty sure Asus actually has several programs which allows you to change BIOS settings from within Windows. If that is the case, which I am fairly sure it is, then the OS and BIOS can communicate.

The fact that there is an auto setting, which will detect which benchmark is running, is very worrying if you ask me.

Yea the god awful Ai Suite, horrible applications.

 

That's sort of a tricky area as the protections are around the bios not being able to access the CPU and it's cache rather than stopping the ability to change fan controllers, multipliers, voltages etc. Was a little broad in my statement.

 

Basically the bios isn't able to see what applications are running in the OS or see what is in the CPU cache, you could however have an OS application installed together with that profile set in the bios detect what application is running and change the multiples and CPU voltage on all or specific cores.

[Dabombinable]

3 minutes ago, leadeater said:

Yea the god awful Ai Suite, horrible applications.

 

That's sort of a tricky area as the protections are around the bios not being able to access the CPU and it's cache rather than stopping the ability to change fan controllers, multipliers, voltages etc. Was a little broad in my statement.

 

Basically the bios isn't able to see what applications are running in the OS or see what is in the CPU cache, you could however have an OS application installed together with that profile set in the bios detect what application is running and change the multiples and CPU voltage on all or specific cores.

Hint hint: chipset drivers.