cancel
Showing results for 
Search instead for 
Did you mean: 

9590 overclock core speed droop.

TheOnlyDoor
Level 7
I'm hoping some one with some knowledge can help me with this.
I have some basic experience overclocking. I was able to get my 8350 to a P95 stable 4.6 by tweaking DIGI settings, voltage, etc.
I now have a 9590 and was hoping to get it to at least 5ghz stable. So I went through the normal tweaking process that I had done with my 8350. (I was using the M5a99x evo R2.0 for the 8350, am using the Formula-Z for the 9590).
I started by setting core speed to a solid 4.7 (23.5), turned off turbo core, bumped my memory freq. to 1866, but did not touch anything else, left everything else on Auto.
Prime 95'd just fine...so 4.7 is perfectly stable. Got my first P95 error at 4.8 with all the same settings listed above. So I decided to start tweaking voltage, LLC, etc just like I had with my 8350.
However now, even if I don't get any Illegal sumouts on P95 and even if my temps are below 55c I notice in CPUZ that my core speed will droop from 4800 down to 1400mhz during P95. It will sit there for 10-15 seconds then go back up to 4800. It does this off and on the whole time I'm Priming.
Yes Apm,C6, turbo core, etc are ALL disabled. Again, so no P95 errors when this happens but I'm sure the core speed should not be drooping intermintently like this. Must have somthing ro do with voltage being taken off auto or the DIGI settings.
Any thoughts?
Thanks in advance.
2,664 Views
3 REPLIES 3

MeanMachine
Level 13
Hi TheOnlyDoor
Yes the 9590 is a different beast in terms of temperature. I have a slightly larger Cooler which copes but yours should be OK.
First you need to establish if it is heat related, As you are stable at 4.7GHz and seems to be throttling at 4.8GHZ.
I originally had the same issue and found P95 which is too harsh IMO for extended periods, and I could only pass two cycles on Small FFT.

Use RealBench for a realworld stress test and benchmark. Constantly monitoring temps in the background for voltage droops and temperature spikes.
Other changes like CPU Current Capability set to 130% and Digi+ Power Control> LLC set to (Ultra High) helped me achieve 5.2GHz. This was where the VRMs were getting too hot so I backed off to 5Ghz, opened up my case and had a fan blowing on the VRMs. This made a difference so I decided to fit a fan on the back plate of the CPU which dropped temps by 4-5c and achieved stability at 5.2GHz.

44591
We owe our existence to the scum of the earth, Cyanobacteria

My System Specs:

MB:ASUS ROG Crosshair VII Hero/WiFi GPU:EVGA GTX 1080 sc PSU:Corsair AX-1200i
CPU:
AMD R7 2700X Cooler: Corsair Hydro H115i Case: Corsair Carbide 780t

Memory:G.Skill TridentZ F4-3200C14D-16GTZR SSD:Samsung 500GB 960 EVO M.2


[/HR]

Well I ran 5 passes with Realbench, most hung around 69200, 1 pass did hit 71000 (which I uploaded) . Firestrike I ran 3 times and got around 10,500.
So benchmarks seem to be stable.
Im going to run an 8 hr realbench stress test. I'm a little confused however, unlike P95 which will show sumout errors, what am I looking for in the realbench stress results?
Is every pass saying " Hash Match" mean the OC is stable?
Thanks for your help!

MeanMachine
Level 13
Hi TheOnlyDoor For a good insight as to how RealBech works and what you can do with it.

http://rog.asus.com/275272013/overclocking/realbench-benchmarking-stress-test-insights/
http://rog.asus.com/tag/realbench/

Have fun and your posted results are great for your first attempt.

Compare your results with those above you to gain an insight on how to improve scores.
There is a batch file to close unnecessary services during your benchtest, which can also improve scores.
We owe our existence to the scum of the earth, Cyanobacteria

My System Specs:

MB:ASUS ROG Crosshair VII Hero/WiFi GPU:EVGA GTX 1080 sc PSU:Corsair AX-1200i
CPU:
AMD R7 2700X Cooler: Corsair Hydro H115i Case: Corsair Carbide 780t

Memory:G.Skill TridentZ F4-3200C14D-16GTZR SSD:Samsung 500GB 960 EVO M.2


[/HR]