GCN 1.0 Async Compute issue fixed

[Prysin]

@Kwee I called out your sensational bullshit when you were making your original thread and subsequently decided to fill it with paranoid conspiracy theories, but here we are. It was a bug.

 

Quote

like last drivers (17.2.1) re-enabled async-compute for GCN1 GPUs. Not sure how much will impact games (Currently my 280 is used as slave GPUs for non-game applications), but at least This should cause less problem to developers.

Here's a shot by GPU View, as you can see, the compute queue is back and runs in concurrency with the default queue. Please note that some flip and presentation issues are caused since the GCN1 GPU does not directly output to the monitor (Which is controlled by a R9 380X, to GCN3 GPU) and all is handled by WDDM magic.

 

 

NOTE: there is no mention of this fix in AMDs official patch notes. I have checked, twice.

 

 

Not much to say here. There was a bug, it's importance and impact was blown out of proportion by fanboys, and it was fixed.

As we get more sources, ill try update the OP.

 

Source: http://www.bitsandchips.it/52-english-news/8053-welcome-back-asynchronous-compute-queues-gcn-1-0

[Morgan MLGman]

Good to know, people tend to overhype things like that without any further analysis.

[Kinda Bottlenecked]

5 minutes ago, Morgan MLGman said:

Good to know, people tend to overhype things like that without any further analysis.

 

heh Ryzen

[zMeul]

5 minutes ago, Morgan MLGman said:

Good to know, people tend to overhype things like that without any further analysis.

really?!

AMD said nothing when they removed Async Compute from GCN 1.0 and again said nothing when they put it back

in the mean time you had a lot of people on AMD's forums asking what was going on .. with some AMD rep blaming game devs 

 

analyze that

[Prysin]

4 minutes ago, zMeul said:

really?!

AMD said nothing when they removed Async Compute from GCN 1.0 and again said nothing when they put it back

in the mean time you had a lot of people on AMD's forums asking what was going on .. with some AMD rep blaming game devs 

 

analyze that

null-hypothesis. You cannot prove or disprove your proposition, so your entire premise is irrelevant.

[Morgan MLGman]

1 minute ago, Pohernori said:

heh Ryzen

That too, and many more... I personally just gather all of the "leaks" and "rumours" and when people ask about Ryzen or Vega I'm just telling them what info floats around the web, adding that this isn't confirmed info.

 

2 minutes ago, zMeul said:

really?!

AMD said nothing when they removed Async Compute from GCN 1.0 and again said nothing when they put it back

in the mean time you had a lot of people on AMD's forums asking what was going on .. with some AMD rep blaming game devs 

 

analyze that

Another sensationalistic comment from you, see the entire point of this post is to inform people that it was a bug and it got fixed. Right under it you say that AMD "removed" Async Compute and put it back again. Come on, not everything has to have a hidden conspiracy theory behind it.

[zMeul]

6 minutes ago, Morgan MLGman said:

Another sensationalistic comment from you, see the entire point of this post is to inform people that it was a bug and it got fixed. Right under it you say that AMD "removed" Async Compute and put it back again. Come on, not everything has to have a hidden conspiracy theory behind it.

"bug" what the fuck does that mean?

does this software walks around on it's own legs? does it delete itself? truly amazing if it managed to do that

 

"software bugs" do not exist! it's the hipster way to say the software devs fucked up / didn't do their job properly

 

conspiracy?! what conspiracy 

what happened, most likely, was that they discovered issues with GCN 1.0 and removed Async Compute without telling anyone; they fixed it and added it back .. again without telling anyone

[Morgan MLGman]

7 minutes ago, zMeul said:

"bug" what the fuck does that mean?

does this software walks around on it's own legs? does it delete itself? truly amazing if it managed to do that

 

"software bugs" do not exist! it's the hipster way to say the software devs fucked up / didn't do their job properly

It means that either yeah, someone, somewhere screwed up or for instance when they added another feature to the drivers it was conflicting with the Async on GCN 1.0 and it stopped working. Stuff like this happens, you have no idea how VMware (which is the most-commonly used virtualization software in enterprise environments) is fucked up in both GUIs, one feature works in the web vSphere client and the same feature doesn't work in the regular vSphere client, there is another feature that works only in the regular client and isn't shown in the web client at all for some reason. Won't tell you which ones exactly, I can only hint that they're related to configuring storage under the iSCSI protocol.

 

Stuff like that happens. Those people are still only people and do make mistakes.

[PlayStation 2]

insert conspiracy ecks dee

Hmm, I really am confused on how they accidentally crippled asynchronous compute performance. Oh well, they fixed it.

[zMeul]

12 minutes ago, Morgan MLGman said:

It means that either yeah, someone, somewhere screwed up or for instance when they added another feature to the drivers it was conflicting with the Async on GCN 1.0 and it stopped working. Stuff like this happens, you have no idea how VMware (which is the most-commonly used virtualization software in enterprise environments) is fucked up in both GUIs, one feature works in the web vSphere client and the same feature doesn't work in the regular vSphere client, there is another feature that works only in the regular client and isn't shown in the web client at all for some reason. Won't tell you which ones exactly, I can only hint that they're related to configuring storage under the iSCSI protocol.

 

Stuff like that happens. Those people are still only people and do make mistakes.

that's what incompetence looks like, from the senior programmer to the individual programmers

changes made to the code needs to be cataloged and revised and not added with the shovel 

[Centurius]

23 minutes ago, Prysin said:

null-hypothesis. You cannot prove or disprove your proposition, so your entire premise is irrelevant.

You can actually pretty easily prove whether or not AMD has said something about its removal and subsequent reintroduction as well as the statement that AMD reps blamed game developers. At least try to apply the scientific method properly.

[Prysin]

3 minutes ago, XenosTech said:

Think I remember that thread... God damn that xeon comment from a while that that he got flamed for while for

Everyone knows that the reason AMD disabled Async Compute for GCN 1.0, is beacuse its not GCN 1.0, but actually a Maxwell GPU. Duh

[XenosTech]

9 hours ago, zMeul said:

that's what incompetence looks like, from the senior programmer to the individual programmers

changes made to the code needs to be cataloged and revised and not added with the shovel 

Sigh you do know nothing made by humans is perfect.. not even the bloody tech we use is perfect but it works

9 hours ago, Prysin said:

Everyone knows that the reason AMD disabled Async Compute for GCN 1.0, is beacuse its not GCN 1.0, but actually a Maxwell GPU. Duh

Dang I should of gotten me one of those maxwell titans eh ?

[Prysin]

5 minutes ago, Centurius said:

You can actually pretty easily prove whether or not AMD has said something about its removal and subsequent reintroduction as well as the statement that AMD reps blamed game developers. At least try to apply the scientific method properly.

Can you prove or disprove if AMD is right in their allegations?

No. Because none of us has the information or insight on the topic from AMDs perspective.

 

Can you thus claim AMDs claim is false? 

No. Because you cannot prove or disprove your hypothesis simply by the fact that we will most likely never get the truth to the story about AMD blaming devs.

 

There is no logic in blaming AMD without knowing the counter argument/proof to their claim. 

 

If you are to criticize my critic, atleast make sure your argument stand up to rigor.

[zMeul]

19 minutes ago, XenosTech said:

Sigh you do know nothing made by humans is perfect.. not even the bloody tech we use is perfect but it works

did anyone asked to be perfect?

when you have a issue that interests both developers and consumers, you inform people about it - why do they even have release notes with each driver pack then?!

 

I propose that each driver update is like a RPG mini-game, you have to discover on your own what's new, what's fixed and what's missing 

[XenosTech]

2 minutes ago, zMeul said:

did anyone asked to be perfect?

when you have a issue that interests both developers and consumers, you inform people about it - why do they even have release notes with each driver pack then?!

You know somethings don't show up in a test environment especially when that problem isn't even related to what you're testing for... If they're testing for stability and comparability an a-sync test probably wasn't even in order since it didn't really affect neither of those things.

[Ekin]

6 minutes ago, zMeul said:

when you have a issue that interests both developers and consumers, you inform people about it

 

Because you weren't informed does not mean they did not respond. Get your facts right and put some sources behind your claims.

 

They even talked about it at pcper: (@1:00:07)

 

 

[zMeul]

1 minute ago, XenosTech said:

You know somethings don't show up in a test environment especially when that problem isn't even related to what you're testing for... If they're testing for stability and comparability an a-sync test probably wasn't even in order since it didn't really affect neither of those things.

they're testing for shit

software doesn't modify itself - you just don't add some game compatibility and boom, all Async Compute for GCN1.0 goes missing - that's not how software works

[XenosTech]

1 minute ago, zMeul said:

they're testing for shit

software doesn't modify itself - you just don't add some game compatibility and boom, all Async Compute for GCN1.0 goes missing - that's not how software works

I'll show myself the door before I say something that gets me banned. @Prysin you care to explain to this one ?

[zMeul]

2 minutes ago, XenosTech said:

I'll show myself the door before I say something that gets me banned. @Prysin you care to explain to this one ?

you're quoting ame and passing the ball to him?! have some balls

 

do you even have the slightest idea of how software development works?

[Prysin]

6 minutes ago, zMeul said:

they're testing for shit

software doesn't modify itself - you just don't add some game compatibility and boom, all Async Compute for GCN1.0 goes missing - that's not how software works

you say that on what basis? your extremely extended but undocumented and unproven knowledge about software and software development?

6 minutes ago, XenosTech said:

I'll show myself the door before I say something that gets me banned. @Prysin you care to explain to this one ?

He cannot see my posts, he has blocked me a long time ago

[XenosTech]

2 minutes ago, Prysin said:

you say that on what basis? your extremely extended but undocumented and unproven knowledge about software and software development?

He cannot see my posts, he has blocked me a long time ago

Seems he likes to block ppl who prove him wrong ? I mean it's a simple concept that we've seen time and time again... Dev make a change to their code and something else breaks, totally normal. Some times you pick it up in testing some times you don't, just a matter of how what you're testing for at that specific time. Async being disabled didn't render the cards unstable, dead nor caused BSODs 

[Mira Yurizaki]

Wow this thread derailed fast.

[Prysin]

5 minutes ago, M.Yurizaki said:

Wow this thread derailed fast.

Really? you think so?

 

Well, all in all, this news post is a "statement". There isnt much to discuss here, other then @zMeuls conspiracy theory.

[Mira Yurizaki]

1 minute ago, Prysin said:

Really? you think so?

 

Well, all in all, this news post is a "statement". There isnt much to discuss here, other then @zMeuls conspiracy theory.

Over half the posts are you guys bickering about it so...