AMD FX-8350 underclocks under load

[tddk25]

Hey everyone! So today I noticed that my CPU was underclocking itself when I would run either a CPU-Z stress test or benchmark. My CPU is an AMD FX-8350, Mobo is an ASRock 970A-G/3.1. I also have a 650w PSU so I know it's not suffering from insufficient power. When it idles it stays at 4.08GHz, but when I run the stress test it goes down and sits around 3.45GHz. I opened NZXT CAM and noticed my CPU won't touch 70 degrees if that helps. I am very confused and would appreciate any help.

[Sjaakie]

Could be power limited. Does it max out at a certain wattage?

[tddk25]

Sorry I don't really know much about how to monitor computers. Is there any software I could use that would be able to monitor that?

[Fasauceome]

8 minutes ago, tddk25 said:

my CPU won't touch 70 degrees if that helps

These FX series CPUs don't have the same thermal limits as modern CPUs. While you might expect an i5-10400 to reach 85 Celsius and be just fine, these older FX CPUs needed to be kept in their 60s. What cooler are you using? 

 

Also, make sure to use the quote button so we can see your responses 

Edited August 13, 2022 by Fasauceome

[tddk25]

I'm currently using its stock cooler with the red led ring.

[tddk25]

I was using a cooler master cooler (don't know exact model) but replaced it with that because it had a very loud whine and was caked in hard dust.

[tddk25]

11 minutes ago, Sjaakie said:

Does it max out at a certain wattage?

I downloaded Open Hardware Monitor, and at idle the CPU is at 1.380 volts and when I stress test it the CPU will steadily drop to 1.200 volts.

[Guest]

1 hour ago, Fasauceome said:

These FX series CPUs don't have the same thermal limits as modern CPUs.

Yes they did....

 

The 65c (roughly) core temp, socket temp was a little higher, but was a "recommended temp" when on a full load. 

If memory serves me correctly today, 90c it should throttle hard.

 

He's witnessing the normal P-states in action. 

 

Better to OC this one manually in my opinion.

[tddk25]

23 minutes ago, Guest 5150 said:

Better to OC this one manually in my opinion.

I would OC, but the previous times I've tried to, even by just 0.1GHz or more my PC would bluescreen right after boot. And OCing by 50MHz doesn't sound worth it to me. 

[Guest]

35 minutes ago, tddk25 said:

I would OC, but the previous times I've tried to, even by just 0.1GHz or more my PC would bluescreen right after boot. And OCing by 50MHz doesn't sound worth it to me. 

Where you using the bus clocks (200mhz) and increasing that? 

 

You should be able to set the multiplier manually, leave the 200 bus clocks alone.

 

I'd try like 3.6ghz which would be a 200x 18 multiplier. 

Set the V-core around say 1.32v manually. 

 

 

 

[tddk25]

So when I OCed my CPU I used my old mobo's automatic overclocking feature (ASUS M5A97 R2.0) and it would bluescreen. I just got this ASRock mobo and haven't been able to find the manual OC settings anywhere in the bios. When I try to use ASRock's automatic overclock, the overclock works until I restart and then it goes back to stock speeds.

[SkilledRebuilds]

Overheating VRM temps from the board could be dropping your clocks under load, a big deal when these were popular years ago.

Talking about a whole range of boards here...

Not specifically this one though.

 

Even if VRMs were rated for say 100+*c the board still would downclock on some boards programmed to do so nearing 70*c or so.

Adding a small fan blowing over the VRM area can help maintain FX CPU load clocks if your board suffers the same issues as differing boards using FX8350 / OC, but are fine with FX4/FX6.

[tddk25]

On 8/13/2022 at 2:56 PM, Fasauceome said:

Also, make sure to use the quote button so we can see your responses 

Sorry Fasauceome I'll be sure to quote from now on.

[tddk25]

18 hours ago, SkilledRebuilds said:

Overheating VRM temps from the board could be dropping your clocks under load, a big deal when these were popular years ago.

Talking about a whole range of boards here...

Not specifically this one though.

 

Even if VRMs were rated for say 100+*c the board still would downclock on some boards programmed to do so nearing 70*c or so.

Adding a small fan blowing over the VRM area can help maintain FX CPU load clocks if your board suffers the same issues as differing boards using FX8350 / OC, but are fine with FX4/FX6.

I'll check on VRM temps, thanks.

[tddk25]

On 8/14/2022 at 5:02 PM, tddk25 said:

I'll check on VRM temps, thanks.

Which of these sensors is for the VRM?

[-rascal-]

1 hour ago, tddk25 said:

Which of these sensors is for the VRM?

 

Do you see more temperature readings if you scroll down?

HWMonitor might show it as a "TMPIN#", or it will actually read "VRM".

 

Otherwise, you can try using HWiNFO instead, since it is able to read / show A LOT more information than HWMonitor.

AsRock says the motherboard uses a 8+2 phase VRM, but my guess it is not a TRUE 8-phase, but a 4-phases and using a doubler.

The VRM could be thermal throttling.

EDIT: Yes, it is a 4+1 VRM, running a doubler to get so-called "8+2"

 

[tddk25]

On 8/13/2022 at 2:50 PM, Sjaakie said:

Could be power limited. Does it max out at a certain wattage?

HWiNFO is saying that my CPU isn't even using 50 watts. It is using about 40 watts on a stress test with a max of 47 watts.

[tddk25]

2 hours ago, -rascal- said:

 

Do you see more temperature readings if you scroll down?

HWMonitor might show it as a "TMPIN#", or it will actually read "VRM".

 

Otherwise, you can try using HWiNFO instead, since it is able to read / show A LOT more information than HWMonitor.

AsRock says the motherboard uses a 8+2 phase VRM, but my guess it is not a TRUE 8-phase, but a 4-phases and using a doubler.

The VRM could be thermal throttling.

EDIT: Yes, it is a 4+1 VRM, running a doubler to get so-called "8+2"

 

I'm not seeing any temp readings for the VRMs.

[Sjaakie]

9 hours ago, tddk25 said:

HWiNFO is saying that my CPU isn't even using 50 watts. It is using about 40 watts on a stress test with a max of 47 watts.

That's pretty low. Check in the bios if it's power limited. Eco-modes often limit it at 45 watt so could be that.

[tddk25]

15 hours ago, Sjaakie said:

That's pretty low. Check in the bios if it's power limited. Eco-modes often limit it at 45 watt so could be that.

I found that Cool 'n' Quiet was enabled, so I disabled it. The CPU still has the same symptoms under the stress test.

I did notice the thermal throttle is on auto. Should I try a different option?

 

(Edit: When I disabled Cool 'n' Quiet the CPU would go up to 72°C and the downclock was more gradual, but it was still only using 47 watts.)

Edited August 18, 2022 by tddk25

[Beerzerker]

On 8/13/2022 at 3:47 PM, tddk25 said:

Hey everyone! So today I noticed that my CPU was underclocking itself when I would run either a CPU-Z stress test or benchmark. My CPU is an AMD FX-8350, Mobo is an ASRock 970A-G/3.1. I also have a 650w PSU so I know it's not suffering from insufficient power. When it idles it stays at 4.08GHz, but when I run the stress test it goes down and sits around 3.45GHz. I opened NZXT CAM and noticed my CPU won't touch 70 degrees if that helps. I am very confused and would appreciate any help.

How do you know the above is accurate?
And there is this too:

 

On 8/13/2022 at 3:57 PM, tddk25 said:

I'm currently using its stock cooler with the red led ring.

I'll go ahead and sum up what's up - It's throttling like a MOFO while running at least as hot as Hell's basement under load....
Let me repeat that - UNDER LOAD, not at idle.

You're trying to OC a chip with a cooler only made to run it at stock speed in the first place.
All FX coolers I've ever seen (And I've had quite a few of them) never had a LED ring on them, red or whatever so you must be talking about something that looks like a stock cooler but it's not, or maybe one of the AM4 stock Wraith coolers.

Then there's the fact 8 core FX chips run warm(er) to hot and now that factors in along with the above, equalling 2 strikes vs your OC.

I have to say the following too:
Power draw of an FX CAN go well above 250W just for the chip alone, that's a big chunk of 650W's being used by the CPU alone. You're not drawing that much I know but any FX, once OC'ed can and does go above 125W's of power draw which is why I say the following.

If basing power needs by the PSU's wattage rating along that's not "Quite"  accurate... I mean it is yet it isn't. 

The side of the PSU (Rail) that supplies power to the CPU via the 8 pin or 4/4 combo CPU power plug MUST be able to deliver the power it needs or it's starving for power.
Luckily most PSU's these days are made to have good support on that part of it, but FX can go well beyond the norm and that's where you'd run into trouble with it.

So, to sum all this up:
Ditch the stock-type coolers and forget about any stock cooler types period with FX if OC'ing one, NONE will work sufficiently to cool an OC'ed FX chip and keep it going that way for very long (If you get that far) with it, esp an 8 core.

The board you have can run it, not the best but it's capable.

As for coolers, there are plenty that can let you go further but even the best aircooler you can get for one won't let you get much beyond 4.5 continuous (If that), on good water 4.8-5.0 is doable with careful tweaking but I must assume you're on air and that's it.
Clock it to somewhere around 4.3-4.5GHz, check temps UNDER LOAD, not while it's sitting doing nothing (Idle) and be sure to try and keep temps below 63c if at all possible, again under load.

Good luck.



 

[tddk25]

33 minutes ago, Beerzerker said:

You're trying to OC a chip with a cooler only made to run it at stock speed in the first place.
All FX coolers I've ever seen (And I've had quite a few of them) never had a LED ring on them, red or whatever so you must be talking about something that looks like a stock cooler but it's not, or maybe one of the AM4 stock Wraith coolers.

Then there's the fact 8 core FX chips run warm(er) to hot and now that factors in along with the above, equalling 2 strikes vs your OC.

I have to say the following too:
Power draw of an FX CAN go well above 250W just for the chip alone, that's a big chunk of 650W's being used by the CPU alone. You're not drawing that much I know but any FX, once OC'ed can and does go above 125W's of power draw which is why I say the following.

If basing power needs by the PSU's wattage rating along that's not "Quite"  accurate... I mean it is yet it isn't. 

The side of the PSU (Rail) that supplies power to the CPU via the 8 pin or 4/4 combo CPU power plug MUST be able to deliver the power it needs or it's starving for power.
Luckily most PSU's these days are made to have good support on that part of it, but FX can go well beyond the norm and that's where you'd run into trouble with it.

I'm not trying to OC my chip, I've just ran benchmarks and they have said my PC is underperforming so I ran the CPU-Z stress test. Also, AMD's website states that my CPU comes with a Wraith stock cooler for AM3.

I also just bought this 650 watt EVGA PSU just a few months ago to replace my previous sketchy 525 watt Raidmax PSU.

 

Now that I think about it. I need to say something I forgot to before.

When I had the 525 watt Raidmax PSU I also had an Asus M5A97 R2.0 mobo, which was fried by the raidmax. So I replaced it with the ASRock 970A-G/3.1 and the PSU to ensure this mobo wouldn't get fried by it too.

Is there a possibility that the raidmax PSU screwed up my CPU when it fried my mobo?

[Beerzerker]

42 minutes ago, tddk25 said:

I'm not trying to OC my chip, I've just ran benchmarks and they have said my PC is underperforming so I ran the CPU-Z stress test. Also, AMD's website states that my CPU comes with a Wraith stock cooler for AM3.

I also just bought this 650 watt EVGA PSU just a few months ago to replace my previous sketchy 525 watt Raidmax PSU.

 

Now that I think about it. I need to say something I forgot to before.

When I had the 525 watt Raidmax PSU I also had an Asus M5A97 R2.0 mobo, which was fried by the raidmax. So I replaced it with the ASRock 970A-G/3.1 and the PSU to ensure this mobo wouldn't get fried by it too.

Is there a possibility that the raidmax PSU screwed up my CPU when it fried my mobo?

I don't believe that chart is quite accurate, I'm seeing alot of AM4 reference in it. 
I bought a 8350 new (In 2019 no less) and a 8370 before that new too, neither had a cooler like that. 

In fact as said, I don't recall any FX stock coolers with an LED in it but regardless that's a mute point, the stock cooler just won't cut it in some situations - Some being such as stress testing for example. 
I mean it should but that's not aways the case.

I seriously suggest a better cooler for it, like these for example:
Cooler Master Hyper 212 Black Edition CPU Air Cooler, Silencio FP120 Fan, Anodized Gun-Metal Black, Brushed Nickel Fins, 4 Copper Direct Contact Heat Pipes for AMD Ryzen/Intel LGA1700/1200/1151 - Newegg.com
Or:
Scythe Mugen 5 Rev. C CPU Air Cooler, 120mm Single Tower, Intel LGA1700, LGA1151, AMD AM4/Ryzen - Newegg.com

Nice thing is once you have one, the cooler itself will last for years.
All you have to do is blow it out/clean it maybe once or twice a year and that's about it. I will admit though the case you have your system in matters for room to have a cooler like this in it - Small, cramped cases are not much good for anything except making the system run hot because there is so little volume to allow for good airflow through the case and that affects everything.

You also have to remember, the CPU cooler is drawing air to use from what's inside the case and if that's already warm, guess what?

Here's what you can do first:
Check and make sure your case has good airflow through it, not just a fan cramming air in but not alot coming out. Check for slow running case fans and replace those that aren't running at speed if you have to with higher CFM units - Don't need to go for hurricane force fans but you get my meaning. I'd have to say about 10-20 CFM above what's already there is fine. May make a bit more noise but that's the tradeoff of it.
If there are obstructions in the case such as cables hanging everywhere, try to tuck and neaten all that up, you'd be suprised how all that can affect airflow through a case.
Try that first, see if temps improve and go from there.

One more thing, for an 8350's stock speed you shoudn't need anymore than 1.35v's and alot of them can run there with less such as 1.32v's or even lower. More voltage means more heat.
In fact why not list what tweaks you are using such as CPU, CPU-NB, chipset and other voltages.
I may spot something that can help.
 

[tddk25]

1 hour ago, Beerzerker said:

Nice thing is once you have one, the cooler itself will last for years.
All you have to do is blow it out/clean it maybe once or twice a year and that's about it. I will admit though the case you have your system in matters for room to have a cooler like this in it - Small, cramped cases are not much good for anything except making the system run hot because there is so little volume to allow for good airflow through the case and that affects everything.

You also have to remember, the CPU cooler is drawing air to use from what's inside the case and if that's already warm, guess what?

Here's what you can do first:
Check and make sure your case has good airflow through it, not just a fan cramming air in but not alot coming out. Check for slow running case fans and replace those that aren't running at speed if you have to with higher CFM units - Don't need to go for hurricane force fans but you get my meaning. I'd have to say about 10-20 CFM above what's already there is fine. May make a bit more noise but that's the tradeoff of it.
If there are obstructions in the case such as cables hanging everywhere, try to tuck and neaten all that up, you'd be suprised how all that can affect airflow through a case.
Try that first, see if temps improve and go from there.

One more thing, for an 8350's stock speed you shoudn't need anymore than 1.35v's and alot of them can run there with less such as 1.32v's or even lower. More voltage means more heat.
In fact why not list what tweaks you are using such as CPU, CPU-NB, chipset and other voltages.
I may spot something that can help.
 

My case could be the problem. It doesn't have any exhaust on the top, just 1 fan in the rear. I also only have 1 intake fan on the bottom and passive intake in the front. I don't think PWM is working on my CPU fan because I can't see its rpm on NZXT CAM or HWiNFO.

I currently am not using any tweaks besides XMP 1.3 for my ram.

Thank you for the cooler recommendations.

[Beerzerker]

28 minutes ago, tddk25 said:

My case could be the problem. It doesn't have any exhaust on the top, just 1 fan in the rear. I also only have 1 intake fan on the bottom and passive intake in the front. I don't think PWM is working on my CPU fan because I can't see its rpm on NZXT CAM or HWiNFO.

I currently am not using any tweaks besides XMP 1.3 for my ram.

Thank you for the cooler recommendations.

Small cases are known for that, sounds like you just don't have enough airflow going through it.
I don't know what size case you have but if wanting a different case, go for one with plenty of room inside if you can. That does mean at least a midtower case or better but the trade off is more room internally to let air flow, plus those tend to have room for more fans if you need them.
However I'd see about dealing with obstructions to airflow first as described earlier, any cables sticking out or just in a wad where the air would normally flow will impede airflow so look at that first and see if it makes any difference for you once done. Zipties are really handy for tidying up things in that regard.