I have a couple 9370s at home, the newer one has a lower PState than the older one and the older one id on an AIO as my gaming machine. What do you want to know S_B. I can get the P_states tonight or tomorrow. I know the older one is buried somewhere in the just putzen thread but it'll be easier to just get them again.
Page 27 of 318
Well the P-States aren't quite what I expected on the FX-9370.
P-State FID 0x1A - VID 0x04 - IDD 18 (21.00x - 1.500 V)
P-State FID 0x16 - VID 0x0D - IDD 14 (19.00x - 1.387 V)
Yeah it is that jump in Vcore requested by cpu of the mobo in jumping from 3.8GHz to 4.2GHz that is a shocker. I would have to do some looking but 4.2GHz for most FX-8350s is not that high of a P-State voltage request as my memory serves me.
I would believe the 9590 could pull off 22.0x at 1.500v and perhaps a touch less.
On average with green on at load, the 9590 is USUALLY running 1.4xx and higher.
Let her do some F@H last night, almost completing a WU around 6.5 hours or so, HW monitor shows a max of 1.644v. This can't be accurate imo. All green on stock, this set up has never shown me a voltage that high while I was actually attending to it.
Cpu temp (socket) 30c current 26c min and a max of 55c while the package or core temp reached a max of 54c. If this cpu hit 1.644v (which I doubt) 55c is a pretty good temp.
Is there any software we can call reliable? I had no other temp monitoring program open during all of this. So I'm curious if any one else has seen a v spike like this?
I also noticed that HWmonitor claims a MAX clock speed on all cores at 5114mhz (lowest core 2) and highest at 5142 (core 0)
Seems possible HWmonitor gets a little flakey with W10 after some period of unknown time..... (within 6 hours?)
S_B not talking like this to argue a point but to allow those that have a different perspective to see another side as it were.
I don't doubt the FX-9590, left mostly to its' own devices, that HWMonitor 'might' have had a minor portion of time at 1.64V to cpu. It takes a lot of juggling for me to get my FX-8350 to not have a pretty fair overshoot of Vcore when stressful apps load and unload and time has been going on for a while and heat is up. I used to sit and watch HWMonitor do the overshoot thing but I finally mastered it at about 4.8GHz all cores on.
I have not heard of CPUID doing anything to HWMonitor to make it better fitted to Win10. So you maybe spot on that over time HWMonitor and loses focus. Hehehe. I have gotten bored before myself and lost focus.
I wish we had a wider base of FX-9xxx to draw from but my other side of the brain is glad we don't have so many to have to suggest that X system is just not up to an FX-9xxx. Split decision man.
I think this was a brilliant idea to beat around the two biggies from AMD since they really act totally bonkers compared to the FX-83xx in my little pea brain. Thanks man.
It's almost as if these 9 series chips are made to go heavy on voltage and very extreme cooling. At least the Board I'm using is totally and completely built around LN2 and High frequency everything EXCEPT HT. If I OC the Ram manually, HT drops down to match NB at 2200mhz. This is curious to me because either way there's plenty of bandwidth at either speed. But the stability is there, so no complaints. Next step would be to pop it back to what we might call stock of 2.6ghz leave ht voltage at auto with Cpu/Nb and see how it goes.
A note on F@H really quick.
I haven't folded for a long time. It used to be cpu cores done individual WUs while the GPU could really knock down small ones. I just DLed a 10+ day WU for my GTX 760... wow that's a long time. I don't know if I can choose WU units any more, I pretty much DLed it set up and let it do it's thing. Now the Cpu works ALL 8 cores to complete one WU. I gotta say I like that better. F@H really uses a multi core cpu as a program should.
While running the GPU client, my cpu temps have raised 3c so far while Running Cpu client with it.
Well both chip pstates for comparison would be great. The 1.500v for the top 3 pstates says enough. Realistic temps while running AIO with and without additional vrm socket cooling would be a plus. No testing needed, off your head temps would be close enough.
Many thanks. Trying tobturn this more into a 9 series thread more over just the 9590 by itself, sometimes wondering the differences between these two chips with average daily stuff if you follow.
Many thanks
- - - Updated - - -
Oh and any info scraped up on batch 1429 chips would be great as well. I am collecting as much data as possible.
Here' an older shot of the "old" 9370 on my AIO with no additional coolin I don't usually put fans on the VRM unless I'm really going to put the boots to it or having a heat problem. This is also on the sabertooth IIRC which is what I have it on currently but location forced me to slow it down a bit. Also I'm almost certain that "new" 9370 has it's highest bin set at 1.475 and I do know it's a 1429 batch.
Now as far as doing testing with the "new" one on the AIO, we'll see. I can say one thing it won't be anytime soon. I'm using it for benching ATM and the AIO is stuck into the HTPC so it's all boxed up. If I get to it it'll likely be a couple of months.
P-State FID 0x1A - VID 0x04 - IDD 18 (21.00x - 1.500 V)
P-State FID 0x16 - VID 0x0D - IDD 14 (19.00x - 1.387 V)
Yeah it is that jump in Vcore requested by cpu of the mobo in jumping from 3.8GHz to 4.2GHz that is a shocker. I would have to do some looking but 4.2GHz for most FX-8350s is not that high of a P-State voltage request as my memory serves me.
Either my loop needs a good over haul (it does theres no denying it..) OR this gash dang chip just steadily makes heat.
So 4 cores 10 minutes 5ghz 110 bus and stock multi. Package temps hit 73c Cpu temp was only 62c. Temps seemed to be stable by 10 minutes, but Im not one to push my luck if its not needed.
With nearly ALL oc settings on auto mutiplier only 25X oc temps are much much lower no green activated. I dont have screen shots on my phone, but heres voltage settings I DO NOT RECOMMEND to any over clocker on ambient.
I have found that setting LLC lower tends to get lower temps on most of the FX. The other side of the coin is higher voltage setting to compensate. It seems to me like you say there are "spikes" where the chip need extra volts for whatever reason but most of the time will hum along with much less.
Here's an example of what I mean . It's not a 9xxx chip but shows the volts.
Its not stable all cores or even four cores under 1.5250v, Ive tried it.
Ok to get the point across, this chip is not going to be stable manually ocing it while running a set speed with anything less than its equipped pstate or more. Perhaps with the right temps, but not as is. Had a hang at 4.4ghz 1.440v wont even boot at 1.35v....
Under volting isnt an option with this set up. Or at least my lucky chip dont like it. 1.5V wont sustain any oc over 4.9ghz for any amount of time.
Cause if it where to be that easy, every one else with a 9590 would be running 5ghz all cores, but it just doesnt happen.
I understand what you're saying S_B and you're right you're not the first or to be the last with one of those "finicy" chips. Just for arguments sake try what I have put up in he post above. Try setting a V-core of 1.53 ish and IIRC the LLC setting was regular. See if that helps with the high volt requirement and also the heat.
Alright I quickly ran through this thread. Tl;Dr? This chip is power hungry and air/wc is not enough to offset the heat generated? It also sounds like it sucks a lot of amps causing Vdroop?
The first post caught my eye and I am now curious. May allow for some interesting moding if my assumptions are true.
Alright I quickly ran through this thread. Tl;Dr? This chip is power hungry and air/wc is not enough to offset the heat generated? It also sounds like it sucks a lot of amps causing Vdroop?
The first post caught my eye and I am now curious. May allow for some interesting moding if my assumptions are true.
I plan on a delid. Already know what it can do with chilling, thats a given.
Basically, Im trying to stay within the relm of reality with ambient temps. It is very likely with a better loop than mine, temps could be more manageable, but not for thw intesive purposes of trying to beat the turbo frequency of 5ghz.
Any one can use chilling... Thats easy. Simple as a bucket of ice water. Or move in with Johan during the winter and borrow some rads out the window....
I'm not moving up there, I'm just buying about 1000 miles of 1/2" x 3/4" tubing and running it up to his house. He has spare windows I can use..I'm sure of it. Anyone know if my magdrive pump will push the water that far?
OK I have given this a try just now. 11:25pm central time.
As requested 1.521250v was the closet I could come to request. 25x Multipler. Reference clock (200) auto. LLC on regular had horrible v-droopage. Insta hang while trying to run Cinebench. Had issues all the way up to High setting. High core temps, cpu temps seem better, but the "HANG" issue still resides here. Will not pass less heavy WPrime 1024m but does pass WPrime 32m. So it makes headway with more more more voltage. totally power hungry chip. Seems to thrive on it. Lovesd it like new born baby.
Now being not having tons of time.... I'm letting the loop cool down to try the heat soaking down. And it would help if I added 3 fans, running only 5 of 8.... I figured most people are only on a 120.2 rad at best and trying to keep this "real" as possible for you guys.
This thread is pretty much intended for the new guy in town looking at OCing the hottest most voltage hungry cpu AMD has made to date.
ALSO - Would like to take note on hear thoughts on L2 and L3 cache and the role they play in actual heat production. I see this a lot, big difference, between WPrime and Cinebench. Prime95 is a heat monster too. So is OCCT which I like to use all 4 of these for testing. Will and have run Prime95 and OCCT back to back for maximum super load and even run F@H on the Gpu to make it proper. But for our readers, the Gpu will remain silent as we concentrate solely on what we call an overclock.
So far ANYTHING solid clock from 4.4GHZ to 5GHZ this cpu wants 1.5250v with Ultra High LLC and not a drop less. However you can get away with some lower volts leaving Cpu auto v-core setting and let the Cpu decide when to nab that 1.5250v, but will wind up in a hang.
TURBO = 1 core. I have not seen this cpu use more than ONE CORE during standard stock settings. Reason? It will just create way too much heat ESPECIALLY with an AIO cooler boxed with chip. Buying THIS is way out of the question, just buy the CPU itself and get as many rads as possible on a home build custom loop. ONLY way to go!! FX-9590
Here's be a couple of screen shots sir. regular and high. Medium hung as I was trying to get a screen capture during a 1024m run.... Just a simple low volt hang. Same as this X86 cores ever where..... No BS, no errors, just stop dead in tracks.
And believe me..... it drives me completely bonkers.
First pick idle no benching btw on regular. second pick is High setting LLC with at least a 32m. Instant hang as soon as I hit run on Cinebench.
Oh any questions related to the Memory, I simply hit "load PC chip settings" and let it fly. Perfectly stable. Did a couple of F@H packets for team 11314. Did fine. Notice HT is 2600mhz everything else is on auto.
This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
By continuing to use this site, you are consenting to our use of cookies.
Anyone know if my magdrive pump will push the water that far? 
