New build idle ok but small load - gets hot

[SantasGolfClub]

I just built a new gaming pc. A few parts upgraded from my existing build.

Ryzen 7 7700X
Gigabyte B650 Gaming A AX motherboard
Corsair H150i Elite 360mm AIO Cpu cooler
Corsair 32gb 6200mhz c36 ram
Liam li mesh II case with 3 exhaust fans

I built this over the day and I seem to be running hot. Idle, I get 45 degrees. But as soon as I go to over 20% usage on the CPU, I’m hitting the 95 degrees Celsius mark. And I think that’s throttle speed?

i think the cooler is attached to the CPU well. I reset it twice. The first time - I may have over tightened and squished the past out.
I’m new to liquid cooling. Maybe my curves are wrong. Not sure if the motherboard “knows” it’s liquid cooled and maybe should do something different?

Any ideas on what I can be checking? 95 degrees at 20% seems very bad.

 

[ArdWar]

Both temperature numbers don't look right, likely indicating something is physically wrong. This might be silly but are you sure there isn't some peelable protective plastics or similar left between your CPU and waterblock?

 

[SantasGolfClub]

Jeeee. Let me check that! I’ll remove the cooler and make sure. Pretty sure there’s not - but let’s have a look. I may be applying the paste wrong (too much / too little), under/over tightening the cooler to the cpu.

I’m not even convinced the water pump is working. Maybe I wired it wrong.

 

[ArdWar]

Wrong paste application or using even the worst thermal paste out there shouldn't cause temperature that high. As much as I hate using Corsair Link (or whatever the bloatware they're pushing now), you can check the pump status there.

 

[SantasGolfClub]

One thing that worries me, and is cooler related: there’s a fan connector trust comes from the cooler wiring harness that is a 3 pin plug, has only one wire connected, and says it’s the “cpu fan” connector. Not sure how one sure works. But I’ve plugged that into my “Cpu fan” connection on the motherboard. Maybe that’s not right?

Ok. I’ll add the Corsair virus and see what it says.

 

[ArdWar]

Yes, it's only single pin since they only need the fan tachometer signal to let the mobo know that there's a fan for the CPU. If connected properly the wire should connect to the third pin on your CPU fan header.
Considering it boots up without complaining, I assume you already plugged it correctly.

 

[SantasGolfClub]

Yeah. Looking at this, I'm not sure what the colant temp should be. But 34 degrees seems OK?
I'm not sure where that's measured. Assuming at the head? Which would mean the coller is taking away some heat as the CPU temp when I took this screen shot was 44 degrees.

Not helpful, I'm sure, but this is what I have:

So, the radiator is mounted on the front, pulling in the air (One fan is backwards and I will fix it - i.e it's pushing air out - bottom one on the radiator). The two top fans exhaust out the top. The one on the back exhausts. The one at the bottom sucks in air and blows towards the GPU to try to help it. The GPU is a 3080. The cooler has the USB mounted at the bottom (visible USB cable). All the fan headers are that the top above the memory.

And here's some MSI Afterburner history:

Here's a zoom on the processor at around 20% and the CPU heat up at 70.

 

[ArdWar]

Hmm 34C is reasonable if a tad bit on the higher side.

Can anyone with Zen 4 chime in please? lol. Zen 4 clock-thermal profile is supposed to go straight to 95C, but I'm not sure if it's supposed to be like this.
Can you still load the CPU with reasonable performance at 100%?

 

[SantasGolfClub]

And here's me loading DCS. Processor goes up to 83 degrees... CPU stays below 30% (Strange, I thought it would be hammered), coolant stays within a degree of what it was when the CPU was at 40 degrees.

Can you still load the CPU with reasonable performance at 100%?

I'm scared to try! Is there a safe way to stress the CPU, get results, and now blow away half the east coast of Australia with fire ?

And PC reset itself. I left DCS running, CPU was around 52% usage, and last I saw, around 86 degrees c, coolant was up to 38.7 degrees. And it reset itself.
After restart, CPU is at 58 degrees and slowly coming down.

 

[AusWolf]

Idle, I get 45 degrees. But as soon as I go to over 20% usage on the CPU, I’m hitting the 95 degrees Celsius mark. And I think that’s throttle speed?

Those numbers are normal on a 7700X. The way it works is, the CPU will target 95 °C, or the highest attainable spot on the voltage-frequency curve under load.

Idle is fine, too, I have 42 °C with a 280 mm AIO liquid cooler.

What CPU cooler do you have? What CPU frequency do you see under load after reaching 95 °C?

 

[ArdWar]

I'm scared to try! Is there a safe way to stress the CPU, get results, and now blow away half the east coast of Australia with fire ?

It should protect itself. Assuming the cooler don't suddenly fall off or something.

And PC reset itself.

Okay, maybe after you fix whatever caused this.

 

[AusWolf]

Can anyone with Zen 4 chime in please? lol. Zen 4 clock-thermal profile is supposed to go straight to 95C, but I'm not sure if it's supposed to be like this.
Can you still load the CPU with reasonable performance at 100%?

Yes and yes.

You will reach 95 °C with a 7700X at stock unless you have liquid cooling with a 280 mm rad, or something bigger. What matters is the CPU frequency you see when you reach 95 °C.

 

[SantasGolfClub]

I’m running a Corsair H150i Elite 360mm AIO Cpu cooler. Installed today. And my first attempt with liquid cooling / AIO.
Something I just realised I’m failing to mention is - I am having an issue which I believed was related. But if your post about the temps being OK are right - then maybe I have a separate issue.
Sometimes - my PC just gets a black screen. It just happened now, under load from DCS. But happens randomly. I put it down to overheating.
The screen goes black. Monitor goes to “No signal” mode and then goes into standby. But PC still running. Fans spinning. Internal RGB all on. I see that the little “post” lights come on. And my eyes let me down, but I’m pretty sure it’s the top right red led that stays illuminated. Looking at the manual - I think it’s the DRAM led. When you boot up - the top left one comes on and goes off, then the DRaM one comes on and after a while, goes off. Then the lower two do the same. But when I get the black screen - the DRAM ones stays on and nothing happens until I hold the power button. Even then a reboot may happen but screen remains black. It never gets past that LED. After maybe 5 minutes - it all works again. I know it’s working because some external USB devices come to life - and then the monitor starts the startup process.
I thought this was due to heat.

 

[AusWolf]

I’m running a Corsair H150i Elite 360mm AIO Cpu cooler. Installed today. And my first attempt with liquid cooling / AIO.
Something I just realised I’m failing to mention is - I am having an issue which I believed was related. But if your post about the temps being OK are right - then maybe I have a separate issue.
Sometimes - my PC just gets a black screen. It just happened now, under load from DCS. But happens randomly. I put it down to overheating.
The screen goes black. Monitor goes to “No signal” mode and then goes into standby. But PC still running. Fans spinning. Internal RGB all on. I see that the little “post” lights come on. And my eyes let me down, but I’m pretty sure it’s the top right red led that stays illuminated. Looking at the manual - I think it’s the DRAM led. When you boot up - the top left one comes on and goes off, then the DRaM one comes on and after a while, goes off. Then the lower two do the same. But when I get the black screen - the DRAM ones stays on and nothing happens until I hold the power button. Even then a reboot may happen but screen remains black. It never gets past that LED. After maybe 5 minutes - it all works again. I know it’s working because some external USB devices come to life - and then the monitor starts the startup process.
I thought this was due to heat.

That's not overheating - the CPU can't overheat, that's why it has the 95 °C max. temperature. Overheating would be something like 110-115 °C, but it can never reach that thanks to the built-in protection.

Edit: Your motherboard's VRM circuitry can overheat, though, but I don't think that's the case here.

To decide if your 95 °C is OK or not, you need to look at the CPU's clock frequency when it runs at 95 °C. If it doesn't drop or fluctuate too much, you're fine. I consider it a bit too high for a 360 mm AIO, though.

If the DRAM led lights up, then you have a RAM or motherboard, or memory controller issue. Is your BIOS up-to-date? My B650 motherboard was a near unusable piece of [insert something here] and wouldn't work together with my RAM no matter what I did until an update came a couple months hater.

Speaking of RAM, I see you have a 6200 MHz kit, which is a bit over the recommended max. OC of Ryzen 7000 series. Can you gradually clock it back with 200 MHz increments, and see if the problem persists?

Edit 2: Also, what is your FCLK and UCLK (Infinity Fabric and Memory Controller clocks)?

 

[SantasGolfClub]

Thanks. The 6200 was an oversight by me when I got all the parts. I didn’t realise I was over the max. Would it be a good idea to somehow - the the bios, “underclock” the memory down to the max? Is that even possible? Is that what you mean by clicking back? Would I be doing that in the bios?

I just watched a YouTube video of a chap explaining what you just said. The 7700 runs hot. That’s a relief. Would MSI afterburner be a suitable tool to try and chart the temp, cpu usage and cpu frequency?
I’m. It sure what FCLK or UCLK is. If you tell me where to check (bios, I assume?). I can get that.

mad for firmware - I updated today. It seemed
More stable. But in the last hour - two resets.

 

[AusWolf]

Thanks. The 6200 was an oversight by me when I got all the parts. I didn’t realise I was over the max. Would it be a good idea to somehow - the the bios, “underclock” the memory down to the max? Is that even possible? Is that what you mean by clicking back? Would I be doing that in the bios?

What I mean is, Zen 4's official supported memory speed is 5200 MHz, but AMD says overclocking should be usually fine up to 6000 (not always, but usually).

So instead of running at 6200 MHz, you could try lowering it to 6000, and see if it's stable. If it's not, lower it to 5800 and check again, and so on.

I just watched a YouTube video of a chap explaining what you just said. The 7700 runs hot. That’s a relief. Would MSI afterburner be a suitable tool to try and chart the temp, cpu usage and cpu frequency?
I’m. It sure what FCLK or UCLK is. If you tell me where to check (bios, I assume?). I can get that.

Every 7000X CPU runs hot, to be fair. The non-X versions are much easier animals to tame.

Afterburner is generally a GPU overclocking tool, but you can use its monitoring tool to chart temps and frequency. HWinfo would be even better.

FCLK is the Infinity Fabric's clock (the part responsible for chiplet communication), UCLK is the memory controller's clock. You can see them in the BIOS, or in HWinfo. In an ideal memory overclock of 6000 MHz, the FCLK is at 2000 MHz, and the UCLK is at 3000 (that is 1:1 with RAM speed). Sometimes, when it is not stable (the mem controller can't handle it), or when the motherboard decides to do so, UCLK sometimes drops back down to 1:2 compared to RAM speed. It's more stable, but slower this way.

 

[QuietBob]

Sometimes - my PC just gets a black screen. It just happened now, under load from DCS. But happens randomly.

What is your specific PSU model?

 

[ArdWar]

I'd rather go straight to 5200 to quickly eliminate the possibility of the stick being bad and then working down from 6000 if it's fine at 5200, but that's just preference.

Use ZenTiming to check the clocks

 

[AusWolf]

I'd rather go straight to 5200 to quickly eliminate the possibility of the stick being bad, but that's just preference.

Interestingly enough, my PC hung up during boot with EXPO 6000 MHz, and gave me blue screens and memory errors in TestMem5 with JEDEC 4800 MHz with the factory BIOS. Only increasing the DRAM voltage by +0.05 V at JEDEC made it stable. Both issues disappeared with BIOS v15.

Other than that - I see your point. If it's rock solid at 5200 MHz, why not.

 

[SantasGolfClub]

Ok thanks. So step one is to work out in the bios, how to get the memory speed down. Because it’s hard trying to test stuff and I get the black screen and dram led error. Hopefully that solves that.

Then grab HWInfo. Does that allow for on screen temp displays like afterburner ? Or can I run both? Does it graph? So that I can upload results?

thr thing I want to watch thrn would be CPU temp corollated with CPU frequency. Hoping it stays stable. Is there a preferred (free?) app to push the CPU? I guess tue hope is to see it go to 100%, the temp, we expect will go to 95 degrees and the frequency will stay stable. Probably drop off a bit?

And then also report that FCLK and UCLK.
On it.

What is your specific PSU model?

I’m running a Silverstone 750W Gold Plus modular PSU.

 

[mama]

Jeeee. Let me check that! I’ll remove the cooler and make sure. Pretty sure there’s not - but let’s have a look. I may be applying the paste wrong (too much / too little), under/over tightening the cooler to the cpu.

I’m not even convinced the water pump is working. Maybe I wired it wrong.

You may have the radiator set up wrong.

 

[SantasGolfClub]

Thanks. Just watched that and from what I could understand - the worst thing about my mount is possible noise. But not a performance issue. I have mine mounted on the front panel - sucking in air. The tubes are top. They’re about 7cm above the pump. Meaning the tubes enter the radiator 7cm above the position if the pump. I could switch it around I think. But it seems this is more an audio issue (trickling water sounds - which I don’t hear) more than a performance fix? Maybe king term I should do it. But not sure it’s “required” right now?

UCLK = MEMCLK

can’t find the other setting.

Even in the bios settings while getting those images, PC freezes and the DRAM light lights up.

Also. Unable to get latest bios version.

This morning this seems worse. I brought this fabric thing from 2000mhz to 1733mhz. But it seems I’m getting a freeze up while in bios settings. Maybe a minute after boot. It just freezes. And the red led comes on. If I power off and on - sometimes the red light goes to dram and nothing happens. It just sits there. Other times the dram led lights up for maybe 20 seconds and then my usb devices light up and I know it’s going to work. The dram led goes out and the other two below it flash for a second each and the system starts.

The led that stays lit is the top right one of that group of 4. Labelled dram.

Could it be my memory is just not good on this board/processor?


Corsair 32GB (2x16GB) CMH32GX5M2B6200C36W Vengeance RGB C36 6200MHz DDR5 DRAM

Here’s my complete parts list for this build.

https://au.pcpartpicker.com/list/XKxw4s

I could go switch my memory for this. It will cost me a restocking fee. But would maybe eliminate an incompatibility issue. Would that be a good idea?

Corsair Vengeance RGB 32GB (2x16GB) C36 5600MHz DDR5 RAM (CMH32GX5M2B5600Z36K) - msy.com.au

Buy Corsair Vengeance RGB 32GB (2x16GB) C36 5600MHz DDR5 RAM (CMH32GX5M2B5600Z36K) - Buy Now, Pay Later + FREE pick-up & same day dispatch!

www.msy.com.au

OK, I've removed my EVGA 3080 video card, and .... no crashes. I'm running off the onboard video. This isn't good. How can my video card, be causing a DRAM error light on the motherboard!? Unless the next light is VGA, and it can't get to that one?

This is what I can now capture without the RTX 3080 installed.

Still running without the RTX 3080 in. I have a 3060 in another pc. I’ll put that in and see what happens. And out the 3080 into that pc. If the other pc starts failing - I’m in a spot of bother. But closer to working out the issue. If both PCs run fine. Then I’m confused. I’ll return to my PC the 3080 and see what happens. This could be expensive.

Ok. So I put the 3080 in my older pc which has the 3060. (Ryzen 7 3700x) and - it’s running fine. No crashes.
I put the 3060 into my pc. And ….
Running fine.

so it seems the 3080 mixed with this board/processor/memory is the issue?

Edit: PC still crashing with 3060. Old pc running smooth with 3080. So video card eliminated. PCIe issue?

 

[AusWolf]

Now, this is getting more interesting, because apparently, you're not even running a 6200 MHz RAM. EXPO/XMP is disabled, so you're only at 4800 MHz.

You don't need to touch the Infinity Fabric clock, an Auto setting and UCLK=FCLK is fine.

Technically, the system should be stable with 4800 MHz RAM, but considering that even mine wasn't with the factory BIOS, we might tweak some stuff on that route.

Can you enable XMP/EXPO and see if things are stable? If they're not, what you'll need to decrease is the "system memory multiplier", which should show "62" with XMP on. Then, when you decrease it to 60, 58, etc. you should see a more stable system. If you don't, and it's still unstable regardless of speed used, then we can go from there.

 

[SantasGolfClub]

OK, and keep the video card out? The system is so unstable with it in, I have difficulty getting into Windows. I'll quickly get all the stuff I am running at the moment, as I worry all the stuff above was changed over time, and it's a skewed picture. so right now:

And stand by... Just rebooting to get images of my current bios settings.

 

[AusWolf]

Wait a minute... is the system fully stable with no video card in it?