Hot Temps for 5800X3D

[tabascosauz]

90/60/90 @ -30mv and +0.05vcore
View attachment 269421

So it didn't really help, only made things worse. Without the vcore offset it can maintain 4.1-4.125ghz.

I'll check tomorrow. I would imagine like 30-35c, atm its sitting at 38c.

Windows 11 uses a lot of memory. Just idling with nothing running it uses 8-9gb. But yes, I have a few apps like steam or spotify in the background.

I did try testing on a windows 10 bootable OS, and my cinebench scores on there were like 1000 points higher at the same clock speeds. Just wish it improved thermal/clock behaviour too.

Hmmm. How's your ST results? Around 1450-1500?

What background stuff do you have running? I've been using 11 basically since release - at idle I'm around 4GB used physical RAM, on 22H2. I don't really slim down my 11 installs either, except a few tweaks from here (mostly settings not removing stuff):

Windows 11 Tweaks for GPU Benchmark

Updated for 22H2 - Install without Internet - When it prompts you to go online, press Shift+F10 and type "OOBE\BYPASSNRO" (that's an o not a zero at the end). After the automatic reboot you can install without network - Install on systems without TPM, UEFI or other requirements...

www.techpowerup.com

For a 21C ambient, I'm at 22C mobo temp in the morning. Unless your case has really bad airflow, that's a big difference for supposedly similar ambient temps

Easy way to tell if your windows install is exceptionally dirty is just to run R23 in safemode. I get the same scores within margin of error. I don't expect it to be the cause of the low clocks though

 

[LordsnLadies]

Hmmm. How's your ST results? Around 1450-1500?

I ran it twice now (that was a painful wait), got 1388 and 1362pts.

For a 21C ambient, I'm at 22C mobo temp in the morning. Unless your case has really bad airflow, that's a big difference for supposedly similar ambient temps

That is pretty cool temps. I don't think mine ever hits such low temps, ill have to confirm tomorrow.

What background stuff do you have running? I've been using 11 basically since release - at idle I'm around 4GB used physical RAM, on 22H2.

4GB on windows 11, didn't even think that was possible.. Currently I've got steam, spotify, windows mail, one drive (can disable this because i dont use), icue, razer synapse, steelseries software, free download manager. There is one thing that has been plaguing me since windows 10 and its this super long wait after you enter your credentials to log in. I think its creating a system restore or something.

Thanks for that link though, will check it out.

Easy way to tell if your windows install is exceptionally dirty is just to run R23 in safemode. I get the same scores within margin of error. I don't expect it to be the cause of the low clocks though

I'm certain my OS is bloated and interfering with results. Which is why I see such a big difference going to a windows 10 bootable which has basically no extra programs running.

 

[tabascosauz]

I ran it twice now (that was a painful wait), got 1388 and 1362pts.

Even in safemode?

That's either really low ST clocks, or a metric ton of background processes. Would be interested to see how you're clocking during the ST test.

I didn't think much of background programs at first, because in your screenshots Global Frequency Limit looked fine (is basically the clock ceiling at any given moment, if you have a very busy idle then you will probably never see Global Limit >4450 at idle). But that clearly isn't the case.

4GB on windows 11, didn't even think that was possible.. Currently I've got steam, spotify, windows mail, one drive (can disable this because i dont use), icue, razer synapse, steelseries software, free download manager. There is one thing that has been plaguing me since windows 10 and its this super long wait after you enter your credentials to log in. I think its creating a system restore or something.

I'm certain my OS is bloated and interfering with results. Which is why I see such a big difference going to a windows 10 bootable which has basically no extra programs running.

In daily use I'm not far off, at about 10GB. With nothing active open, that comes down to about 5GB without going out of my way to close background stuff.

But when benching I don't keep anything open that I don't absolutely need. I disable all necessary startup stuff, disable internet, stop and disable the Windows Search service, and then reboot (important after disabling internet for a cleaner boot). Why reduce perf for no reason, right.

The likes of CAM, iCUE or Synapse are not generally things viewed as favourable to performance.

That is pretty cool temps. I don't think mine ever hits such low temps, ill have to confirm tomorrow.

See, that's the thing, if I roughly go off your mobo temps (which are not precisely indicative of ambient, but provide some reference), CPU temps don't seem unreasonable considering that you are pushing a greater proportion of Core power than most people.

But for an ambient between 20-25C, temps are not great.

 

[damric]

Sorry didn't read the whole thread.

What's your water temp?
What's your ambient temp?

If they are close within 5 degrees that's good.

If not you need more radiator. I bet that RTX is just heating up the water.

I prefer overkill on radiators.

 

[LordsnLadies]

Even in safemode?
That's either really low ST clocks, or a metric ton of background processes. Would be interested to see how you're clocking during the ST test.

I tried using safe mode on win11 but the cpu reading wouldn't show more than 3.4ghz, so here's ST and MT runs from win10 hirens bootable. It was hitting 4.4-4.5ghz on ST.

142/80/120 @ -30mv. Although for ST i used 90/60/90

Sorry didn't read the whole thread.

What's your water temp?
What's your ambient temp?

If they are close within 5 degrees that's good.

If not you need more radiator. I bet that RTX is just heating up the water.

I prefer overkill on radiators.

Ambient temp: 28c~
Coolant temp on balanced profile: 35c

More radiator... I'd need a bigger case then .

@tabascosauz @Mussels

This is the friends setup with the golden 5800x3d just for comparison. You'd think my setup would win over a mid-tower that's cramped.

 

[tabascosauz]

Sorry didn't read the whole thread.

What's your water temp?
What's your ambient temp?

If they are close within 5 degrees that's good.

If not you need more radiator. I bet that RTX is just heating up the water.

I prefer overkill on radiators.

AIO, not copper loop, GPU is not in it

not sure I would apply the ol loop temps wisdom here

I tried using safe mode on win11 but the cpu reading wouldn't show more than 3.4ghz, so here's ST and MT runs from win10 hirens bootable. It was hitting 4.4-4.5ghz on ST.

142/80/120 @ -30mv. Although for ST i used 90/60/90

Ambient temp: 28c~
Coolant temp on balanced profile: 35c

More radiator... I'd need a bigger case then .

@tabascosauz @Mussels

This is the friends setup with the golden 5800x3d just for comparison. You'd think my setup would win over a mid-tower that's cramped.

3.4GHz sounds like something's up with your BIOS. I ran in safemode just fine yesterday. I don't use monitoring software because a lot of them are borked in safemode (Benchmate also doesn't run in safemode), only there for the scores

3.4Ghz is what happens if you try to use a 5800X3D on any BIOS earlier than 1206. Locked to base clock.

At 142W ST score looks good. MT score also looks good for 4.3GHz.

28C ambient sounds much more likely.

 

[freeagent]

28c is pretty warm, but don't tell @Mussels that

My mobo temp is 23c right now

 

[tabascosauz]

28c is pretty warm, but don't tell @Mussels that

My mobo temp is 23c right now

Like always, wait for you to accomplish the impossible all the 15200+ results I've seen use BCLK

I ain't catching up to you with temps like these

I tried using safe mode on win11 but the cpu reading wouldn't show more than 3.4ghz, so here's ST and MT runs from win10 hirens bootable. It was hitting 4.4-4.5ghz on ST.

This is the friends setup with the golden 5800x3d just for comparison. You'd think my setup would win over a mid-tower that's cramped.

Just reminded me of something, a while ago another feller had some appreciable clocks/temps issues, albeit living in a hotter climate with an ambient of around 30-35C

It turned out to be the AIO in the end for whatever reason, and on a D15 they managed to get temps down to 78C @ 100W with -30.

If the TC12DX is getting the exact same temps, might want to consider testing the AIO elsewhere or thinking about an RMA.

help with undervolting 5800x3d

If you heard cracking then something isn’t right, take it apart and start over. i got my hands on a d15 chromax today rented it from a shop to test with it and see in a few hours after work i will try again ok got the cooler and here is the test result please tell me it's good

www.techpowerup.com

 

[QuietBob]

I tried using safe mode on win11 but the cpu reading wouldn't show more than 3.4ghz, so here's ST and MT runs from win10 hirens bootable. It was hitting 4.4-4.5ghz on ST.

142/80/120 @ -30mv. Although for ST i used 90/60/90

Your run in this post looks like stock CPU, and the R23 scores are good. But you stated otherwise. How did you undervolt the CPU? By -30mv do you mean -30 CO in PBO2 Tuner? It seems that the settings you chose did not get applied.

Could you post a screenshot showing HWinfo during R23 at full stock settings?

 

[Chicken Patty]

Sorry guys, been testing couple different things without having to completely re-do my loop. I did re-paste and re-seat the CPU block, felt it helped a bit but not much, maybe 2-3 degrees max, which is something. I removed by 7" LCD display from the rear case position that I used to monitor the PC, and added a rear case fan. I tried both way, pulling air in and pushing air out. Pulling air out definitely worked best in terms of bringing down water temps. This case by default is not the greatest for airflow so I've read and with rads mounted in front and up top, I would suppose air flow is not the greatest. My fans are Lian Li SL Infinity's which are very pretty and the daisy chain design is awesome, but don't flow the most air. However, lately temps have been pretty good since I added the rear fan exhausting out. What I did notice is that the overall temp inside the case was very hot, components were really warm to the touch, ever since it has dropped my water temp by about 2-3 degrees and that's really helped.

Room is 27-28ºc right now as I just got home from work, left it running a flight in flight simulator 2020, and water temp is 35-36ºc, and CPU hottest was about 70ºc with default values. Not using PBO2 Tuner or anything.

So I definitely think between the re-paste/re-seat, and the extra fan blowing that hot air out of the case, it's improved tons.

This is the build, I was never able to upload a picture when I created this thread.

Average core temps. Much better than before.

 

[Mussels]

Something does seem off with the AIO performance-wise if it can't even beat a TC12DX

The review i linked showed that unless its at high pump speeds, it's simply not great. They did testing with all sorts of combinations and it basically didnt handle things well, on a much easier to cool CPU unless it was on extremely loud settings - but they tried to make it sound good since it could be quiet and good performing, just not at the same time.

 

[LordsnLadies]

Your run in this post looks like stock CPU, and the R23 scores are good. But you stated otherwise. How did you undervolt the CPU? By -30mv do you mean -30 CO in PBO2 Tuner? It seems that the settings you chose did not get applied.

Could you post a screenshot showing HWinfo during R23 at full stock settings?

Just to be clear, this is on agesa 1206. On 1207 my chip gets throttled to 3.8ghz at stock. Ambient temps in the room is 29c.

The review i linked showed that unless its at high pump speeds, it's simply not great. They did testing with all sorts of combinations and it basically didnt handle things well, on a much easier to cool CPU unless it was on extremely loud settings - but they tried to make it sound good since it could be quiet and good performing, just not at the same time.

I only get a 2-3c reduction if I put it on extreme profile for pumps and fans, so clearly it doesn't help much. I don't actually have the capellix, here's one of the few reviews of mine: review

I feel like I wasted a whole lot of money for nothing on this cooler. I was just naive in thinking an AIO would provide significant changes. R3.3k (190$~) vs R750 cooler (43$~), yeah you'd think there would be some improvement...

 

[QuietBob]

Just to be clear, this is on agesa 1206. On 1207 my chip gets throttled to 3.8ghz at stock. Ambient temps in the room is 29c.

There is nothing wrong with your CPU. At stock, it draws a little more power than mine in R23. This could be due to the difference in silicon quality, your mobo's default voltages, the VRMs or indeed CPU temperature. Your room temperature in the last run is clearly lower than 29c. It seems to be in the 23-24c range, as evidenced by the SSD temps. Why don't you get a cheap digital thermometer to confirm it?

That said, it looks like your tweaks have little effect on the CPU. The first shot is supposed to be "142/80/120 @ -30mv", while the second full stock:

There is barely any difference between the power drawn. In fact, max CPU core voltage is higher in the first run, which makes no sense for "-30mV". When applied properly, -30 CO should bring the max core voltage below 1.2v and package power close to 100w. Again, to me that points to the changes not being in effect.

91c in R23 with a 360mm AIO in a 23c room isn't a great result, TBH. If you've checked the mounting and applied TIM properly, it would suggest that this AIO simply isn't the most effective cooling solution for the 5800X3D. I get the following in a 22c room:

This was achieved with a stock Assassin III, a single stock 140x25 intake fan @ 1150rpm, and a single Arctic F12 120x25 exhaust fan @ 1450 rpm. No OS tricks were used, normal Windows boot with an active internet connection and third-party AV protection running in the background. In fact, this was the third consecutive run -- I had to disable HWinfo to score over 15,200. To get maximum MT boost you need to stay below 75c.

EDIT: Naturally, with -30 CO on all cores.

 

[LordsnLadies]

Your room temperature in the last run is clearly lower than 29c. It seems to be in the 23-24c range, as evidenced by the SSD temps. Why don't you get a cheap digital thermometer to confirm it?

I used an infrared thermometer pointed at the ceiling to get an ambient temp. I don't know how accurate that is.

That said, it looks like your tweaks have little effect on the CPU. The first shot is supposed to be "142/80/120 @ -30mv", while the second full stock:

It definitely does have an effect making it -30mv. My clocks go up to 4.3ghz over stock and that's the max it can achieve on cinebench because of thermals and/or silicon lottery. The best I can get it is just over 4.3ghz with the 80tdc and 120tdc limit.

91c in R23 with a 360mm AIO in a 23c room isn't a great result, TBH. If you've checked the mounting and applied TIM properly, it would suggest that this AIO simply isn't the most effective cooling solution for the 5800X3D. I get the following in a 22c room:

Here's what hardware unboxed got with the h150i capellix at 21c ambient: link

Granted mine isnt a capellix, it shouldn't be too far off from this. Don't know if their result is any good either.

 

[tabascosauz]

@QuietBob to be fair, the screenshots are being taken during R23, so it appears from the current Vcore values that CO is making a difference. But I have to say, the crazy peak spikes in Vcore I haven't ever seen from a 5800X3D, CO or no CO. Just one of a number of bizarre behaviours.

@LordsnLadies honestly at this point I would just take the AIO off and reduce the list of suspects by 1. Assuming a warmish ambient, you already know that your old TC12DX roughly does the job. And if you can buy locally/Amazon a better cooler that you can return within a month, could use that to test and make sure hardware is the culprit. Or if you can find a Fuma 2 to level the playing field with the friend's X3D, even better.

I don't really believe these issues are confined to the AIO. Too many strange things going on around your clocks and voltages on this board.

 

[QuietBob]

I used an infrared thermometer pointed at the ceiling to get an ambient temp. I don't know how accurate that is.

These may show 5 degree variation measurement to measurement, even the more expensive semi-pro models. Your room temperature cannot possibly be 29c when your SSD recorded 25c minimum and your case 26c. Your current SSD temp is 27c.

It definitely does have an effect making it -30mv. My clocks go up to 4.3ghz over stock and that's the max it can achieve on cinebench because of thermals and/or silicon lottery. The best I can get it is just over 4.3ghz with the 80tdc and 120tdc limit.

Are you using PBO2 Tool? Undervolting your CPU by 30mV is not the same as applying -30 CO. The latter will have a much more pronounced effect on power consumption and temps.

Here's what hardware unboxed got with the h150i capellix at 21c ambient: link Granted mine isnt a capellix, it shouldn't be too far off from this. Don't know if their result is any good either.

Steve's 84c in R23 with 21c ambient isn't particularly low. My rig was 85c full stock in a 25c room. This leads me to believe that the h150i isn't a great performer given its size, and most probably not best suited to the 5800X3D. You may get better results with air cooling. What is your case and fan config?

 

[LordsnLadies]

Are you using PBO2 Tool? Undervolting your CPU by 30mV is not the same as applying -30 CO. The latter will have a much more pronounced effect on power consumption and temps.

Yeah sorry, when i say -30mv i mean through pbo tuner.

You may get better results with air cooling. What is your case and fan config?

My setup

I shared this before and it seems the biggest sticking point is the hdd bays in front. I'm going to remove the hdd bays and put a spare 140mm fan on top to get more exhaust from the rad.

 

[3x0]

Yeah sorry, when i say -30mv i mean through pbo tuner.

Only PBO2 Tool can set the CO and custom PPT/TDC/EDC limits, the options in the BIOS are not functional despite being there. Only modified BIOSes have those options enabled and functional.

 

[LordsnLadies]

Only PBO2 Tool can set the CO and custom PPT/TDC/EDC limits, the options in the BIOS are not functional despite being there. Only modified BIOSes have those options enabled and functional.

The official asus rog strix bios does allow you to change ppt, tdc and edc limits. Just not CO.

 

[3x0]

The official asus rog strix bios does allow you to change ppt, tdc and edc limits.

But only to lower values than official AFAIK. Anyway, -30 CO value is equal to about -90mV since every increment of CO is 2-3mV

 

[tabascosauz]

Yeah sorry, when i say -30mv i mean through pbo tuner.


My setup

I shared this before and it seems the biggest sticking point is the hdd bays in front. I'm going to remove the hdd bays and put a spare 140mm fan on top to get more exhaust from the rad.

This dispels a lot of the uncertainties......

I can tell you right now that 0 air is getting through those drive cages unless you run all case fans on full blast.

Even if it did, none of it is useful to CPU cooling.

Having seen the photos, your friend's setup is much more preferable. You won't get any optimal results out of either the AIO or air cooler like this, regardless of how fast you run your fans.

 

[LordsnLadies]

@tabascosauz So i removed the hdd bays and replaced one of the intake fans.

Would you say its worth mounting fans on top of the rad for more exhaust too?

Edit:

Alright so i installed 3x 120mm top exhaust fans and it does seem to have made a difference.

Just as an experiment, I put the aircon on at 18c and waited for the coolant temp to drop down significantly (23.9c in this case). Put AIO fans at 100% and pump on extreme.

142/80/120 @ -30 co

 

[LordsnLadies]

@QuietBob

I did some more tests this evening with the AC in the room set to 16c and maxed out fans on radiator.

90/60/80 @ -30 co:

90-60-80.png

drive.google.com

142/80/120 @ -30 co:

142-80-120.png

drive.google.com

no limits @ -30 co:

no limits -30.png

drive.google.com

stock:

stock.png

drive.google.com

What do you think?

Sorry didn't read the whole thread.

What's your water temp?
What's your ambient temp?

If they are close within 5 degrees that's good.

If not you need more radiator. I bet that RTX is just heating up the water.

I prefer overkill on radiators.

Sorry for only replying now. At idle the coolant temp is about 5c higher than ambient if i max out the fans, otherwise an extra 2c on top of that.

 

[Zach_01]

88/60/90 @ -20mv

88/60/90 @ -30mv

Stock limits @ -30mv

Don’t know if anyone said so far but I saw you keep putting mV at the end or the undervolting.
I assume you are referring to CurveOptimizer negative magnitude.

FYI these are negative (or positive) 1-30 steps and it’s not mVolts.
Each step is a range of 3-5mV
3mV on high/highest clock and 5mV on low/lowest clock.

So when you set CurveOptimizer to -30 it means you are shifting the v/f down by 90-150mV.

 

[QuietBob]

I did some more tests this evening with the AC in the room set to 16c and maxed out fans on radiator.

I'm confused. Do you have snapshot CPU polling enabled in HWinfo?

Stock results look right, but your clock is lower than it should be -- only 4180 with max temp at 80c. Unless that was the moment when a new cycle in R23 starts? I got 4300-4325 at 85c (25c room). Also, your PPT/TDC/EDC limits were maxed out at some point, while mine were at 83%/80%/93% tops. Perhaps the BIOS is messing up with some settings?

With -30 CO the clocks look much better at 4415, as expected at this temperature. But your voltage remains largely unaffected by the CO, while power usage is the same as stock and the limits are still being maxed out.

I'd clear CMOS and stick with AGESA 1206. If that doesn't help, I'm afraid a clean OS install is in order.