Windows 10 Scheduler bug causes poor RyZen performance

[Mira Yurizaki]

5 minutes ago, SaladFingers said:

So, If Windows goes to sleep, I will be able to run Crysis afterwards?

 

meh, wrong thread sorry

Yes. :3

[SaladFingers]

1 minute ago, M.Yurizaki said:

Yes. :3

It's funny how there was news about a bug making Ryzen faster and now more news about another bug that makes it slower, lol. I mixed up the threads.

[juri-han]

3 hours ago, LAwLz said:

Read the next paragraph in my post. It addresses that point specifically.

The high latency can not be fixed. What can be fixed is that Windows can try to avoid needing to data between the CCXs, but that's no easy thing to do.

 

I just watched the video and I did not hear them say the 6 core would not have this issue. Do you have a timestamp?

As for how the 6 core will work, my guess is that they will just take a defect CCX, disable two cores and then use that along with a fully functioning one. So you have a chip with some broken cores, but 6 fully working ones.

yeah sorry that's what i ment not a fix but rather a work around witch would not be easy but it still will improve performance, and when they talked about 6 cores it was more of a off-hand comment saying in this design they would of been better off with 6 cores rather then 8.. its a rather long video ill try and find a time code lol 

[LAwLz]

1 hour ago, Prysin said:

Ryzen is a dual core. Obviously. 2 big clusters with "AMD cores" inside.... come on man, get the facts straight!!!

Honestly, that is not any worse than what AMD did before with "compute cores".

You can't list a quad core with 8 GPU core iGPU as a "12 compute core" processor AMD! Shame on you!

At least they had some radical videos to market it with:

[Prysin]

8 hours ago, LAwLz said:

Honestly, that is not any worse than what AMD did before with "compute cores".

You can't list a quad core with 8 GPU core iGPU as a "12 compute core" processor AMD! Shame on you!

At least they had some radical videos to market it with:

atleast that one they had a whole API built around the idea of using the APUs like that. And AMD DID, for once, clarify and demand that all retailers clarify it using AMDs official clarification. Which was understandable for most....

[xAcid9]

On 3/13/2017 at 2:09 PM, leadeater said:

I would do 1 CCX since it should be cheaper and no cross CCX issues but then it could all depend on how many partial faulty CCX's there are and how well 6 core CPUs sell to use those faulty CCXs with disabled cores.

On 3/14/2017 at 1:43 AM, LAwLz said:

As for how the 6 core will work, my guess is that they will just take a defect CCX, disable two cores and then use that along with a fully functioning one. So you have a chip with some broken cores, but 6 fully working ones.

From Anandtech.

Quote

We have confirmation from AMD that there are no silly games going to be played with Ryzen 5. The six-core parts will be a strict 3+3 combination, while the four-core parts will use 2+2. This will be true across all CPUs, ensuring a consistent performance throughout.

Sauce: http://www.anandtech.com/show/11202/amd-announces-ryzen-5-april-11th

Yup, Data Fabric is still there even for the 4C/8T. Gonna need an expensive fast RAM to squeeze out most performance out of them. Not sure how it will perform with some core disabled though.

[leadeater]

26 minutes ago, xAcid9 said:

From Anandtech.

Sauce: http://www.anandtech.com/show/11202/amd-announces-ryzen-5-april-11th

Yup, Data Fabric is still there even for the 4C/8T. Gonna need an expensive fast RAM to squeeze out most performance out of them. Not sure how it will perform with some core disabled though.

Awesome thanks for the info, kinda disappointed but not surprised.