AMD chipset driver revision 1.9.27.1033 is out

Anyone's idle clocks change? Mine appear to be much higher now.

i'll try them shortly but could be that the drivers are just being early released for the new windows update that's suppose to fix the priority core usage problem. that being said if you're using the ryzen power plan clocks are always going to be higher since it's trying to keep the cores unparked.

just putting these here to check once my drivers update.
1/2/3/5 sleep 500/403 .995-1.005v

ok installed and tested.. cores are definitely staying parked for less time now and cycling clocks but my voltage is down from what i posted above to .75-.89v at idle even though cores are staying asleep less but are now all staying below 400mhz. this is with chrome open as well. also seems to attempt to park up to 5 of the 6 cores instead of previously where it would only park at most 3-4 cores.. PPT is down from 28% idle to 24% at idle so it's definitely using less power.

this is using windows power plan not the ryzen PP btw.
 
Last edited:
i seem to be running at 3.7-ish idle on my 3700x stock speeds now it runs up to 4.22 at times......i guess thats good. Enermax AIO 120 mm push pull config
 
i seem to be running at 3.7-ish idle on my 3700x stock speeds now it runs up to 4.22 at times......i guess thats good. Enermax AIO 120 mm push pull config

use ryzen master to check it, don't use hwinfo or windows performance to check the idle clocks.. for some reason hwinfo still can't properly detect idle clock speeds through AMD's api..
 
my 2600 runs at its rated boost speed of 3.9ghz instead of 3.85ghz. this machine should kick some serious ass now. i'm so happy
 
A friendly reminder that if you are not using project mercury on your ryzen cpu you might lose up to 20% of your FPS in games
 
A friendly reminder that if you are not using project mercury on your ryzen cpu you might lose up to 20% of your FPS in games


You need to give that project mercury to someone to do a review if it actually works that well. It needs more exposure.
 
  • Like
Reactions: c3k
like this
Sven, nobody replied to your post. It was an excellent improvement. Have you shared this with AMD?
 
Sven, nobody replied to your post. It was an excellent improvement. Have you shared this with AMD?

I have not but other have.
Part of it is going to be fixed in upcomming windows update for win 10 . if its not arledy in there ( to load up one CCX core first instead of thread jumping back and forth)


The issue is the windows thread scheduler treats alle logical cores as equale. which they are on a unloaed system. but once you have one thread loaded to a logicla cores. all other logical cores for the same physical core is slower. due to the sharing of the real physical ressources.
This is Where SMT hurts performance
related you have that cache data for a thread jumping baxk and forth between the 2 ccx units has to go over the slow infintyi fabric which is much slower than the l3 cache. this is the CCX performance penalty issue


With chiplets coming up the thread scheduler for windows really needs improvemnt and understand not all cores are equal for a given thread even though they are equal on their own
 
Giving it a shot on my 1700 @ 3.8.
Seems to have speed up reaction times? I think is the best way to explain it. Nothing computationally seems to be faster according to benchmarks, but there's a little less lag when doing basic stuff like throwing windows across my multiple screens and such. Hard to quantify.
 
I just download it on x470 with 3700x .. I have it in eco mode testing it .. does cut it a bit in power as almost what a 3700 may look like .
 
Giving it a shot on my 1700 @ 3.8.
Seems to have speed up reaction times? I think is the best way to explain it. Nothing computationally seems to be faster according to benchmarks, but there's a little less lag when doing basic stuff like throwing windows across my multiple screens and such. Hard to quantify.
Snappier? That's the word all Intel users use when they describe how their CPU is better than AMD. Intel is just snappier. Does this software help make AMD snappier than it was before?
 
A quick note that idle clock speed doesn't seem to affect power draw with these CPU's being able to enter sleep states quickly.

Using a watt meter I saw no higher idle power draw using the Ryzen power plan (higher idle clocks, 95% power at idle in power plan) vs the windows power plan (lower idle clocks, 5% power at idle in power plan).

However, the Ryzen plan had higher benchmarks in things like Kraken web benchmark.
 
A quick note that idle clock speed doesn't seem to affect power draw with these CPU's being able to enter sleep states quickly.

Using a watt meter I saw no higher idle power draw using the Ryzen power plan (higher idle clocks, 95% power at idle in power plan) vs the windows power plan (lower idle clocks, 5% power at idle in power plan).

However, the Ryzen plan had higher benchmarks in things like Kraken web benchmark.

yeah agree, i went back and forth between the plans as well and seems like the drivers now completely ignore the power state setting in the power plan which i'm glad to see. there is something else that was added to the ryzen one that's mentioned in the notes on one of the chipset driver versions but can't remember what it is. now just waiting for windows to release that damn update that fixes the scheduler.
 
with project mercury enable things do feel more responsive. If i turn settings down to 720 there is a noticable increase in frame rates.Running a RX580 at 1080 or 1440 can't tell the differnce card is probably the limiting factor.
I thought that the scheduling problem had been taken care of and the new updates were aimed at intel processor's.


btw if you have nothing better to do you can head over to tom's hardware and read their cooking review.It's amd vs intel. These pepole really need to get a life.
 
Last edited:
with project mercury enable things do feel more responsive. If i turn settings down to 720 there is a noticable increase in frame rates.Running a RX580 at 1080 or 1440 can't tell the differnce card is probably the limiting factor.
I thought that the scheduling problem had been taken care of and the new updates were aimed at intel processor's.


btw if you have nothing better to do you can head over to tom's hardware and read their cooking review.It's amd vs intel. These pepole really need to get a life.

no the new update is aimed all processors that have a priority core(aka amd's "golden" core) so that it'll always be the first core loaded before moving to a different core. not sure if intel uses something like that as well.
 
Back
Top