Return to Level1Techs.com

(CLOSED) Totally lost, FX 8350 performance abysmally low (er than normal)


#21

Welp...

Neowin.net: If your PC is "abnormally slow" running Windows 10 build 10586.318, here's a possible fix

Chances are you are experiencing the "performance degradation" problem the latest Windows 10 cumulative update caused.


#22

1.47V is insane. My 9590 is running at 1.375V for 4.7GHz. Only when I push it to a sustained 5GHz do I get into the 1.4V range.

Assuming that it's not the great and powerful Winders screwing up everything, I would run passmark and check the test breakdown. If the chip is damaged then you might have a huge hit to your float/int math. The other possibly easier test I would run is to boot a linux iso. If the issue vanishes then it's probably a craptastic Winders install.


#23

Use AIDA64, or that ASUS's AI Suite II, to take a look at the temperature of the VRMs while you are running a stresstest. The CPU may be doing fine, but the VRMs may be overheating and enacting a throttle.


#24

This reminds me of overclocking an Android phone. When pushed too far without overheating or crashing you'll get a lower score than with a lower clock. I mean to say that it's not an uncommon thing for me in general (regarding PCs yeah, it is). Said that why not try to get a lower clock/voltages and do the same tests again to see if that fixes the issue? Maybe the chip has been running with a little too much voltage and it's starting to "bend", so to say. I know I'm not scientific or anything but I think this solution is worth a shot. Good luck!


#25

Thanks for the feedback overnight guys, gonna check out some of these options.

I have also recently checked for CPU core parking, that's definitely not a thing. I agree that 1.47V at that clock speed is heinously high... my old chips would hit at LEAST 4.8 GHz with that kind of juice in the past.

As for VRM temps, I have been checking those as well: the VRMs for this motherboard are designed to hit 150C max temp, and these aren't even hitting the 90's, so I'm fairly confident they're doing fine. I'll get back to you guys a bit later on (possibly Monday, this is my weekend) to see what, if anything, I learn. Again, thank you so much for the great feedback and non-troll-y responses. I have chosen wisely with this forum.


#26

Alright, since I'm a glutton for punishment, I have been stuck om these solutions and one other that I found.

Firstly: ran PassMark. Here's some of my CPU results at a fixed 4.2 GHz clock speed:



Most of the scores seem to be fine here, and there are no failures during testing. Any of the scores that are CLOSE to the 8150 I imagine are involving processes that simply aren't affected much by clock speed at all, or maybe those are a sign of where the problem lies? Haven't ever owned an 8150 so I can't say personally.

Also disabled the Cortana nonsense mentioned in that link that I HAD ALREADY DISABLED IN THE PAST... really need to permanently disable that so-and-so. Tested after in Cinebench, no results changed.

Also went and quadruple checked my Windows Update: there was a USB to Serial Comm port driver that kept failing to install, so I cleared the downloads and stopped the service entirely through the command prompt JUST to make sure it wasn't going to interfere... still no change.

As to the "bending" suggestion, I was thinking that as well at first... until I checked every single multiplier that I validated from highest to lowest, and each performed worse than the last, until I hit the stock, unmodified specs for the chip and re-tested... those results were the WORST, only hit 575 or so with the processor at stone stock, VERY much no bueno!

As for throttling, I mentioned previously that the clock speeds on my chip remain constant throughout all testing, so there's also no throttling occurring...

... which all is leading me to believe that my system is not ONLY AMD, but an even SHITTIER version of AMD than is typical, and that there is very little I can do about it. If there's anything else out there that'd be great... but I'm kind of starting to lose hope here... or maybe I'm just insane... either is valid x-)


#27

1.4 on the V core is not that high, its about what i see when the chip is boosting at stock BIOS settings.
Personally i did not like it and went down to 1.2 at 4.0 GHz.


#28

The chip I presently have undervolts to 1.35 or so at 4.2 which isn't AS bad, but for every increase of 100 MHz I get, I have to apply so much more VCore, it's ra-tarded.


#29

Dude 1.4V is crazy. My 9590 only needs 1.375V to hit 4.7GHz. I can easily hit 5GHz sustained at a little over 1.4V.


#30

Try booting either a linux iso, using a junk HDD and installing a fresh copy of windows on there and running or maybe a windows live CD. Windows live CDs are a bit weird though so I would probably just use a backup HDD.

That should tell you 100% if it's software or hardware. If it is hardware then how old is it? AMD didn't bat and eye when I warrantied my 9590 saying it was overheating and pulling huge amounts of voltage to remain stable at stock clock. One of the only parts I've ever needed to RMA actually. That and a first batch WD Red with bad firmware.

I did basically what we're doing now and ran dozens of different tests and then gave them a full page of tests I have preformed and the results. The next day when I checked it it was approved. Just read the rules religiously and write your request accordingly and you should be good. I would also try replacing every part you can just for shits and giggles before hand that way you can confidently say that you did and that it didn't fix anything.

If it's a PIB/w (Processor In Box) fan then you've got 3 years. If it's a PIB/wo fan then you've got 2.
http://support.amd.com/en-us/kb-articles/Pages/AMDCPUWarrantyFAQs.aspx
http://support.amd.com/en-us/warranty


#31

Those scores look about right. Without having a 8350 I couldn't say for sure, but nothing is hugely off which is what I was wondering about. What I was really wondering was if some of the int pipelines or FMACs were damaged and were being switched off by the CPU. I honestly don't know if it can do that but if that was happening then the clock shouldn't change but the amount of actual math that could be accomplished would drop. Since nothing looked like it was taking a hit that doesn't seem likely.


#32

1.47V on Vcore is way too high for 4.5GHz basicly.
But like i said OP should try Northbridge overclocking instead of Multiplier.
I listed some basic settings above, which i found to be just stable on my chip.
Of course every chip is diffrent, but my advice is to give it a shot.

My FX8350 did 4.8GHz Northbridge overclock with a memory OC to 1926mhz at 1.440V.
But i found myself, that the clock increase from 4.5GHz to 4.8GHz did not realy show a significant diffrence realworld.
So to me it was not worth it to run it at 4.8GHz allday.
FX cpu´s are simply abit slow, multi-threaded they are not that bad.
But still they start to show their age.


#33

Have just done multiplier for my FX-8370, but with FX-6100 I did notice that I'm getting same OC, settings & temps no matter which ways I do OC that chip which was curious. Haven't tested that with 8370 though, but I dont see any reason why wouldnt it still be true for 4.5-4.6GHz.

Clarifying, 23 x 200 = 4.6GHz would be same as 20 x 230 = 4.6GHz.

But yes, also noted that 4.6GHz is the sweetspot, 4.3 or 4.4GHz goes with defaults depending from mobo, and also could be done both ways.


#34

Anything above 80C is something to be wary of, if you have any expectation of longevity. A lot of modern motherboards have safeguards to keep your components safe.

I say try putting a fan on the VRM, see if that helps at all.


#35

VRM overheating isnt very likely with an Asus 990FX board.
Never seen a Asus 990FX board to come even close to a dangerous high temp zone.
Even not with world record overclockings.


#36

Yeah, unless the mobo is in a sealed and insulated box with a room heater then the VRMs aren't out of spec. The only way they could be overheating is if they are damaged or if the OPs CPU is pulling over 480W. Unless my math is wrong, thats what the Asus 990FX board should THEORETICALLY be capable of pushing at maximum load according to their bragging about the caps, chokes and MOSFETs they are using.

As I said before. Run a different OS. That should narrow it down to hardware/software. If it's hardware and has been less than 3 years since the OP bought it he can warranty it with AMD.


#37

Just came back to see what other input was offered... as far as loading in to a different OS, that's something that will unfortunately have to wait for the time being since not only am I busy the rest of the weekend, but I lack experience with any other OS, as well as the time to learn another one, even if only to do some testing.

VRMs are definitely not out of spec. Here's the spec sheet for this board's VRMs: On Semiconductors

I have also done BCLK OC as well as Multi OC just see if there is a difference in scores or stability. There is no difference in required voltage for this chip, nor is there a difference in score; they are both within a point or so one way or another.

Based on what I am seeing here, I am likely just crazy and definitely more than a little sad that this is just the way things are going to be for the rest of this system's life. Even worse, I won a Gigabyte Z170 Gaming 3 motherboard not too long ago... and don't have the money to get a system going on it lol.

Thank you for all your help, guys; at the very least, I learned more about how to troubleshoot an issue like this moving forward for others.


#38

Have you tried dialing the voltage down and running it at 4.2? Are the results still comparably low?


#39

http://www.cpu-world.com/CPUs/Bulldozer/AMD-FX-Series%20FX-8350.html
I would link to AMD's page but they don't list them as far as i can see.

Low power P states.
1: 3400 MHz, 1.225V
2: 2800 MHz, 1.125V
3: 2100 MHz, 1V
4: 1400 MHz, 0.875V

Boosted P states
1: 4200 MHz, 1.425V
2: 4100 MHz, 1.4V


#40

Well like i said 1.47V for just 4.5GHz is basicly a bit too high.
You probably just have a dull chip.

Re-installing the OS might be something you could try.
If you have a spare drive laying arround, you could just try to do a new install on a spare drive just for testing.