5900x temperatures high on water

[DotComBubblegum]

Hi, 

I have a 3090 and 5900x with alphacool GPU AIO Eiswolf and a cpu AIO Eisbaer connected together in a full loop, with 2 360mm rads in a O-11 dynamic case. 

 

I have some temperature issues with the CPU, the chip runs at 65 degrees idle, opening chrome bumps the temp to 75 or 80, while gaming it usually stays around 85 degrees. With prime 95 running  it thermal-throttles at 90 degrees and frequency drops. 

 

All this is happening while in the same loop 3090 is idling at 40 degrees with max 70 degrees in furmark even if ran at the same time as prime 95. 

 

No idea what is happening with my cpu are these Temps normal for 5900x on water, would you have any ideas? 

[Jorban]

That sounds like PBO. 100% normal.

 

Edit:

Sorry, I was looking @ GPU temps and thinking they were CPU temps.

 

Check the usual:

Pump speed @ max

CPU mounting pressure

 

also Prime95 is an unrealistic test and ultimately shows synthetic workloads. Nothing like that will happen in real world scenarios.

Try OCCT using AVX2.

 

On a 3700x with PBO enabled to my liking I see around 40c Idle and 70c under full load.

Mess around with CTR from 1usmus, or perhaps look into using a different power plan in Windows if you have it set to MaxPerformance.

 

Make sure that you are using a notable sensor reader like HWINFO64 and not HWMonitor/Speccy/TaskManager.

[krakek]

1 hour ago, DotComBubblegum said:

Hi, 

I have a 3090 and 5900x with alphacool GPU AIO Eiswolf and a cpu AIO Eisbaer connected together in a full loop, with 2 360mm rads in a O-11 dynamic case. 

 

I have some temperature issues with the CPU, the chip runs at 65 degrees idle, opening chrome bumps the temp to 75 or 80, while gaming it usually stays around 85 degrees. With prime 95 running  it thermal-throttles at 90 degrees and frequency drops. 

 

All this is happening while in the same loop 3090 is idling at 40 degrees with max 70 degrees in furmark even if ran at the same time as prime 95. 

 

No idea what is happening with my cpu are these Temps normal for 5900x on water, would you have any ideas? 

If the CPU waterblock is new there is always the possibility of forgetting to remove the plastic sticker from the cold plate.

 

Happened to me.

[DotComBubblegum]

7 minutes ago, Jorban said:

That sounds like PBO. 100% normal.

 

Edit:

Sorry, I was looking @ GPU temps and thinking they were CPU temps.

 

Check the usual:

Pump speed @ max

CPU mounting pressure

 

also Prime95 is an unrealistic test and ultimately shows synthetic workloads. Nothing like that will happen in real world scenarios.

Try OCCT using AVX2.

 

On a 3700x with PBO enabled to my liking I see around 40c Idle and 70c under full load.

Mess around with CTR from 1usmus, or perhaps look into using a different power plan in Windows if you have it set to MaxPerformance.

 

Make sure that you are using a notable sensor reader like HWINFO64 and not HWMonitor/Speccy/TaskManager.

Thanks I will try undwrwolting using PBO curve optimizer after work. 

Both pumps are at 100%

What is occt? 

I did use HW monitor but I also compared it in ryzen master it was a bit off but just a few degrees. 

[DotComBubblegum]

5 minutes ago, krakek said:

If the CPU waterblock is new there is always the possibility of forgetting to remove the plastic sticker from the cold plate.

 

Happened to me.

Hahah if only it would be that easy. No I have remounted and repasted it 3 times now with varying paste amounts and mounting pressures. 

[Jorban]

7 minutes ago, DotComBubblegum said:

Thanks I will try undwrwolting using PBO curve optimizer after work. 

Both pumps are at 100%

What is occt? 

I did use HW monitor but I also compared it in ryzen master it was a bit off but just a few degrees. 

OCCT is a benchmark / stressing utility.

Ryzen Master and HWMonitor are both awful sensor tools.

I would uninstall Ryzen Master and do all of your PBO tweaks in BIOS.

[freeagent]

2 hours ago, Jorban said:

That sounds like PBO. 100% normal.

Nope not normal. My 5900X is running PBO +200 -30 and she sits in the mid 20s. With an air cooler.. Loads still get up into the 70s low 80s for all core stuff.

[DotComBubblegum]

2 hours ago, freeagent said:

Nope not normal. My 5900X is running PBO +200 -30 and she sits in the mid 20s. With an air cooler.. Loads still get up into the 70s low 80s for all core stuff.

Currently it's running on auto OC setting without any PBO adjustment all defaults. The voltage goes up to 1.45v. I think this might be the issue. Or I got the worst trash tier silicon. 

[freeagent]

1 hour ago, DotComBubblegum said:

Currently it's running on auto OC setting without any PBO adjustment all defaults. The voltage goes up to 1.45v. I think this might be the issue. Or I got the worst trash tier silicon. 

 Not sure.. I see 1.513v fairly consistently.

[Demonic Donut]

Get HWInfo and not HW monitor. You can't trust any of the info with HW monitor.

 

It seems like you have background tasks or you have a bad mount. I know you've redone it, but the high idle makes me think one of those two things is happening. Your load temps are completely normal. The 5900x is a hot chip. The normal load temps make me think you have background processes causing high "idle" temps.

 

I'm not sure how the 5000 series compares, but I know the stock 3000 series chips push more voltage than necessary. Again, you need HWInfo and look at the svi2/tfn voltage to see what voltage you are actually getting on the cpu.

 

My 3600 will get 1.45V on stock settings during minimal tasks like web browsing etc at runs hotter than necessary, and drops voltage and clocks during all core. I static OC 4.3ghz all core at 1.3V set voltage, I don't remember what llc setting I use, and in everything except p95 I'm running cooler while achieving higher frequency.

[Jorban]

1 hour ago, Demonic Donut said:

Get HWInfo and not HW monitor. You can't trust any of the info with HW monitor.

 

It seems like you have background tasks or you have a bad mount. I know you've redone it, but the high idle makes me think one of those two things is happening. Your load temps are completely normal. The 5900x is a hot chip. The normal load temps make me think you have background processes causing high "idle" temps.

 

I'm not sure how the 5000 series compares, but I know the stock 3000 series chips push more voltage than necessary. Again, you need HWInfo and look at the svi2/tfn voltage to see what voltage you are actually getting on the cpu.

 

My 3600 will get 1.45V on stock settings during minimal tasks like web browsing etc at runs hotter than necessary, and drops voltage and clocks during all core. I static OC 4.3ghz all core at 1.3V set voltage, I don't remember what llc setting I use, and in everything except p95 I'm running cooler while achieving higher frequency.

Coming from personal experience, I have seen my 3600x degrade over 6 months by using 1.3v static on the core.
I am not saying that yours will, mine was most likely just a really poor quality chip.
However, I would keep an eye on your core qualities by using OCCT's CPU testing, and it will let you know if there are any errors on the core.
 

8 hours ago, DotComBubblegum said:

Currently it's running on auto OC setting without any PBO adjustment all defaults. The voltage goes up to 1.45v. I think this might be the issue. Or I got the worst trash tier silicon. 

Personally, I would remove Ryzen Master, and then run PBO settings default from your motherboard.
The easiest way to do this is to uninstall Ryzen Master, turn the system off, reset the CMOS, and run it as-is.

Stock PBO & Default PBO set by the motherboard are two very very different things.
Generally if you leave the motherboard set to AUTO on PBO, it will put in it's own settings.
Coming from a Asus B450-F that's where I see the most performance out of my 3700x - however, your results may varry.
It's often that I will see idle loads in the 40c range and all core loads under the higher 70's with the balanced Windows Power Profile.

6 hours ago, freeagent said:

 Not sure.. I see 1.513v fairly consistently.

Is this single core or all core performance voltages?
That could be normal for really really light workloads, but in an all core situation, you shouldn't be seeing anything higher than 1.39 spikes.

[freeagent]

Try giveing it 1.5v all core and see what happens

 

For self regulated all core stuff usually 1.4-1.45v

[Demonic Donut]

2 hours ago, Jorban said:

Coming from personal experience, I have seen my 3600x degrade over 6 months by using 1.3v static on the core.
I am not saying that yours will, mine was most likely just a really poor quality chip.
However, I would keep an eye on your core qualities by using OCCT's CPU testing, and it will let you know if there are any errors on the core.

I'll check out that tool. Is it the one that recommends OC settings? Or is it just another stress test?

 

I've been running this OC for 2 years now and haven't had any problems, but all I really do is game. My temps are great and I don't pull enough current for me to be all that worried. If I was running a lot of CPU work I'd go lower. I have always lost worker 3 and 4 on p95 with small FFT and everyone else behaves, but I'm stable with the use case I have and I got a measurable increase in performance with no crashing.

 

[Jorban]

Just now, Demonic Donut said:

I'll check out that tool. Is it the one that recommends OC settings? Or is it just another stress test?

 

I've been running this OC for 2 years now and haven't had any problems, but all I really do is game. My temps are great and I don't pull enough current for me to be all that worried. If I was running a lot of CPU work I'd go lower. I have always lost worker 3 and 4 on p95 with small FFT and everyone else behaves, but I'm stable with the use case I have and I got a measurable increase in performance with no crashing.

 

OCCT is a stressing tool.
1usmus' CTR is the tool that recommends overclocks based off of short burst tests that it does.

[DotComBubblegum]

16 hours ago, Demonic Donut said:

Get HWInfo and not HW monitor. You can't trust any of the info with HW monitor.

 

It seems like you have background tasks or you have a bad mount. I know you've redone it, but the high idle makes me think one of those two things is happening. Your load temps are completely normal. The 5900x is a hot chip. The normal load temps make me think you have background processes causing high "idle" temps.

 

I'm not sure how the 5000 series compares, but I know the stock 3000 series chips push more voltage than necessary. Again, you need HWInfo and look at the svi2/tfn voltage to see what voltage you are actually getting on the cpu.

 

My 3600 will get 1.45V on stock settings during minimal tasks like web browsing etc at runs hotter than necessary, and drops voltage and clocks during all core. I static OC 4.3ghz all core at 1.3V set voltage, I don't remember what llc setting I use, and in everything except p95 I'm running cooler while achieving higher frequency.

yea I got HWinfo now it looks much better based on those temps, I indeed had a parasitic background process - Logitech G hub bugged out and for some reason was using 10% of my CPU. 

I did some PBO curve optimization with a 17 step all core negative offset. Now I the CPU runs about high 40s low 50s degrees idle and pretty much same load temps. 

 

Attached the snip of HWinfo after 10 min of Aida64 stress test for CPU + GPU.

[Jeppes]

On 9/22/2021 at 3:50 PM, DotComBubblegum said:

Currently it's running on auto OC setting without any PBO adjustment all defaults. The voltage goes up to 1.45v. I think this might be the issue. Or I got the worst trash tier silicon. 

Never use auto-oc. It just does not work well ever and you are much better off using the curve optimizer.

[DotComBubblegum]

23 minutes ago, Jeppes said:

Never use auto-oc. It just does not work well ever and you are much better off using the curve optimizer.

I  think all it does is adding 100mhz to boost clock overrides