Question 5800X3D has high idle temperatures ?

[Jimbob1]

I upgraded to the 5800X3D using a Corsair H115i watercooler. When running R23 benchmark the CPU maxes out at 85 degrees in 10 minutes which is no concern. My idle temps are going from 45-60 degrees now I know zen 4 is a hot chip and I could be worrying over nothing but I'm just wondering if you guys think these are normal idle temps. I've never hit 90 degrees with all the benchmarks I've ran so far. As I'm typing his I'm getting 49 degrees idle.

EDIT: Update, CPU got to 91 Degrees with Prime95.

 

[DavidM012]

It's a Zen 3. Should be pretty easy to find bench tests for it if not on youtube then the web this reddit also says they hit 91c on prime 95.

another with an aio and another with an air cooler (scythe fuma rev b) yes it's a hot chip even with good airflow and plenty of intake & exhaust fans. So there may not be much you can do to chill it more short of a custom loop.


Or lapping the cpu heat spreader flatter which is kinda tough specially with those bendy pins, or de-lidding and applying liquid metal - that could put the cpu die at risk of cracking or shorting something with gallium liquid metal.



crazy idea for a custom mount probably wouldn't recommend it though if you cut a copper plate to size and use a silicon pad of 5mm thickness (and second one of 1mm thickness to cover the back of the cpu socket (clean the back of socket thoroughly with isopropyl alcohol to remove dirt) and then long screws on your mounting brackets to go through the board you could jury rig a mount for a second cooler to pick up residual heat from the back plate.

Well I did it on my not too expensive sabertooth fx & fx cpu and I'm still thinking of ways to improve the design like putting thermal grease on the bolts so heat goes down the screws more, but can only really use long brass screws rather than pure copper because it's too soft so I know it can be done to mount two coolers one on the front and one on the back but the back one isn't so great because you have to use the silicon pads which aren't as excellent as pure metal, but in my experiment my cruddy aerocool power supply blew up and i'm not sure if it's just because it was a crud power supply or because dirt literally contaminated my silicon pad and made it conductive.

Or just conduct heat from the front to the backplate using copper screws without applying a silicon pad directly to the back plate. I mean like literally the capacitors on the back of the mobo where they're soldered maybe they oxidise a little bit over time and the oxides get on the silicone pad so it becomes ever so slightly capacitive which might've caused a short.

Well I don't think you want any experimental innovations or to run without a side panel but it's something I'm considering doing again I just don't want to risk fizzle anything else at the moment.

 

[Karadjgne]

Idle temps with Ryzen are not idle temps with Intel, which has been the 'norm' for what most consider normal.

Intels chop voltages across all cores simultaneously, but all cores retain a certain amount of activity, so at idle every core has only a partial load of the background processes and services. The temp you see is the highest temp out of all the cores, which changes according to which core and when the cores are polled.

Ryzen are somewhat different. At idle, all cores except the preferred core are deactivated, asleep. They share no load, the entire background tasking is dumped on that single preferred core, so is naturally going to register as higher temp than Intel with only a partial load. The preferred core holds that load for about 3 seconds, then wakes and passes the load before being deactivated. That'll roll every core, one at a time, with load so that all cores end up seeing somewhat equal amounts of use.

So if a quad core cpu, Intel would be 30/32/30/30 for example, while a Ryzen would be looking at 24/45/24/24. The polling seeing the 32 as highest for Intel, but the Ryzen registering 45. This is completely normal Ryzen behavior, has nothing to do with whether a 5800x3D is considered 'Hot' running or not.

The job of any cooler is to moderate temps to maintain a working, stable temp within acceptable limits, and does so by changing fan speeds. It's job is Not to maintain the absolute lowest temps possible, that only happens when you manually set the fans to 100%.

Your temps are acceptable, especially concerning Prime95, since I'm going to assume you did not disable AVX technologies, which can hit an approximation of 115% cpu power load, or higher.

 

[Jimbob1]

Idle temps with Ryzen are not idle temps with Intel, which has been the 'norm' for what most consider normal.

Intels chop voltages across all cores simultaneously, but all cores retain a certain amount of activity, so at idle every core has only a partial load of the background processes and services. The temp you see is the highest temp out of all the cores, which changes according to which core and when the cores are polled.

Ryzen are somewhat different. At idle, all cores except the preferred core are deactivated, asleep. They share no load, the entire background tasking is dumped on that single preferred core, so is naturally going to register as higher temp than Intel with only a partial load. The preferred core holds that load for about 3 seconds, then wakes and passes the load before being deactivated. That'll roll every core, one at a time, with load so that all cores end up seeing somewhat equal amounts of use.

So if a quad core cpu, Intel would be 30/32/30/30 for example, while a Ryzen would be looking at 24/45/24/24. The polling seeing the 32 as highest for Intel, but the Ryzen registering 45. This is completely normal Ryzen behavior, has nothing to do with whether a 5800x3D is considered 'Hot' running or not.

The job of any cooler is to moderate temps to maintain a working, stable temp within acceptable limits, and does so by changing fan speeds. It's job is Not to maintain the absolute lowest temps possible, that only happens when you manually set the fans to 100%.

Your temps are acceptable, especially concerning Prime95, since I'm going to assume you did not disable AVX technologies, which can hit an approximation of 115% cpu power load, or higher.

Thanks for the reply. What do you mean by the AVX technologies?

 

[Karadjgne]

AVX is one of the instruction sets the cpu recognises and allows use of. It's an advanced vector extension used for partical analysis, imagine a wall blowing up, AVX is one of the possible instructions used to keep track of every single little chunk of brick flying around, at different speeds and directions. It's faster and more effective to use AVX rather than some other instruction, but most games only make limited use of it because it's very brutal on the cpu.

Consequently, it's going to push a cpu harder than what is considered a maximum normal load, which for games is somewhere around 70% power use. It would temporarily be closer to 85% power use. When you run a 100% load like Prime95, with AVX, you get a virtual 115% power use.

If using Prime95, at the bottom, before actually starting it, there's check boxes to disable AVX, AVX2, and possibly AVX-512 on some cpus. Prime95 should be run as Small (not smallest) fft with AVX disabled. That simulates a 100% constant load, a worst possible case for 'normal' cpu gaming use.

 

[Jimbob1]

AVX is one of the instruction sets the cpu recognises and allows use of. It's an advanced vector extension used for partial analysis, imagine a wall blowing up, AVX is one of the possible instructions used to keep track of every single little chunk of brick flying around, at different speeds and directions. It's faster and more effective to use AVX rather than some other instruction, but most games only make limited use of it because it's very brutal on the cpu.

Consequently, it's going to push a cpu harder than what is considered a maximum normal load, which for games is somewhere around 70% power use. It would temporarily be closer to 85% power use. When you run a 100% load like Prime95, with AVX, you get a virtual 115% power use.

If using Prime95, at the bottom, before actually starting it, there's check boxes to disable AVX, AVX2, and possibly AVX-512 on some cpus. Prime95 should be run as Small (not smallest) fft with AVX disabled. That simulates a 100% constant load, a worst possible case for 'normal' cpu gaming use.

Okay thank you I didn't know that. Would you say I should just leave everything the way it is as it's normal temperatures. I'm seeing people get a lot lower temperatures and they claim they haven't undervolted or modified the CPU at all.

 

[Karadjgne]

You could, or you could change the fan curves. I'm doubting you fans are hitting 100%, or even over 90% fan speed, so others saying they get lower temps is going to be dependent on what fan curves they use, like changing from silent to standard or performance etc.

I get 60-62 gaming on a custom loop. With fan speeds on the rads at 35%. I could easily drop that to 50-52 by jacking up the fan speeds, but there's no point, the temps are perfectly good for a Ryzen at 60° and I don't feel like listening to the jump in noise for no gains.

 

[Jimbob1]

You could, or you could change the fan curves. I'm doubting you fans are hitting 100%, or even over 90% fan speed, so others saying they get lower temps is going to be dependent on what fan curves they use, like changing from silent to standard or performance etc.

I get 60-62 gaming on a custom loop. With fan speeds on the rads at 35%. I could easily drop that to 50-52 by jacking up the fan speeds, but there's no point, the temps are perfectly good for a Ryzen at 60° and I don't feel like listening to the jump in noise for no gains.

Could you possibly show me a fan curve you would use?

 

[DavidM012]

They may have a different mobo, different case, different power supply, different thermal paste, different fans, you can guess more quality is more expensive, they might have a slightly flatter cpu heat spreader due to variations in manufacturing, could be many variables so replicating other builds may not yield better results. They could be in a different climate, or have air con. Lower ambient temps means your cpu idles lower.

If you had a shed in the arctic you could run at 0c probably. If you were in a desert with 50c ambient temps your cpu might throttle immediately if you open a browser or something. Since it has a delta of 20-25c above ambient or something idle.


Certainly radiant heat from the power supply can increase ambient temps inside the case and warm up nearby components while a super dooper expensive platinum excellent power supply of a higher capacity than you need will run probably run cooler.

It would be very difficult. You'd have to try to stack up gains of a few degrees C here and there but if the heat spreader is concave, convex or slightly uneven only lapping it flat would improve the contact of the heatsink and lapping the chip is hard because you have to contrive to hold it flat without bending the pins and sand it flat and not sand it unevenly. You don't want one side higher than the other. *And lapping voids your warranty.

I lapped my fx cpu using a glass panel with sandpaper taped to it and graduated from 36 grit to 100, 400, 800, 1000, 2000, 3000, 5000 to 7000 grit to polish it off, took about half an hour and I used the cpu plastic packing to hold the cpu with the center cut out to avoid bending the pins and you hold it by the edge and I kind of just scraped it ten times in each direction rather than rubbing it up and down like sanding a door frame and then went to the next grit and then coated it with a silver polish 'cos I had some so that's the amount of faffin about you have to do otherwise.

End of the day if the chip won't hit thermal limits it's fine to run at 85c gaming and not worth the otherwise exorbitant costs to get premium components that only provide marginal gains.

Extra quiet fans or 140mm fans or larger start from $10 apiece and still your system is only as quiet as the loudest fan so that doesn't help you much. A premium case is over $100 .. a premium power supply over $100, custom cooling isn't for the faint of heart, all sorts of nik naks fixtures fittings and tubing you need for that, as well as a premium water block to chase 10ths of a Degree C gains $60-$100, all phooey.

you could throw money at it

revving up your fans is the only free thing you can do and if they are the fans that ship in the box with the various cooler or case, probably get loud.

Don't try and tighten the screws on your cooler probably overtighten and break something. You could just reseat it to be sure it's as level as you can get it

The chances are the cooler examples paid more for their cooler coolers. Or cooler solution.

 

[sizzling]

I am assuming your Corsair H115i has the same options as my H150i Elite in iCue. What is the pump set to iCue, quiet, balanced or high performance? If not already try setting to high performance. I would have thought increasing the water flow will have more of an effect at reducing idle temperature than adjusting the fans on the radiator assuming you already have good coolant temperature.

 

[Jimbob1]

I am assuming your Corsair H115i has the same options as my H150i Elite in iCue. What is the pump set to iCue, quiet, balanced or high performance? If not already try setting to high performance. I would have thought increasing the water flow will have more of an effect at reducing idle temperature than adjusting the fans on the radiator assuming you already have good coolant temperature.

I had a look in iCue and the pump was on balanced, so I put it on extreme. Idle temps have improved to a steady 45-48. I ran Cinebench R23 for 10 minutes, CPU maxed to 86 degrees. I'm happy with the result.

 

[Jimbob1]

They may have a different mobo, different case, different power supply, different thermal paste, different fans, you can guess more quality is more expensive, they might have a slightly flatter cpu heat spreader due to variations in manufacturing, could be many variables so replicating other builds may not yield better results. They could be in a different climate, or have air con. Lower ambient temps means your cpu idles lower.

If you had a shed in the arctic you could run at 0c probably. If you were in a desert with 50c ambient temps your cpu might throttle immediately if you open a browser or something. Since it has a delta of 20-25c above ambient or something idle.


Certainly radiant heat from the power supply can increase ambient temps inside the case and warm up nearby components while a super dooper expensive platinum excellent power supply of a higher capacity than you need will run probably run cooler.

It would be very difficult. You'd have to try to stack up gains of a few degrees C here and there but if the heat spreader is concave, convex or slightly uneven only lapping it flat would improve the contact of the heatsink and lapping the chip is hard because you have to contrive to hold it flat without bending the pins and sand it flat and not sand it unevenly. You don't want one side higher than the other. *And lapping voids your warranty.

I lapped my fx cpu using a glass panel with sandpaper taped to it and graduated from 36 grit to 100, 400, 800, 1000, 2000, 3000, 5000 to 7000 grit to polish it off, took about half an hour and I used the cpu plastic packing to hold the cpu with the center cut out to avoid bending the pins and you hold it by the edge and I kind of just scraped it ten times in each direction rather than rubbing it up and down like sanding a door frame and then went to the next grit and then coated it with a silver polish 'cos I had some so that's the amount of faffin about you have to do otherwise.

End of the day if the chip won't hit thermal limits it's fine to run at 85c gaming and not worth the otherwise exorbitant costs to get premium components that only provide marginal gains.

Extra quiet fans or 140mm fans or larger start from $10 apiece and still your system is only as quiet as the loudest fan so that doesn't help you much. A premium case is over $100 .. a premium power supply over $100, custom cooling isn't for the faint of heart, all sorts of nik naks fixtures fittings and tubing you need for that, as well as a premium water block to chase 10ths of a Degree C gains $60-$100, all phooey.

you could throw money at it

revving up your fans is the only free thing you can do and if they are the fans that ship in the box with the various cooler or case, probably get loud.

Don't try and tighten the screws on your cooler probably overtighten and break something. You could just reseat it to be sure it's as level as you can get it

The chances are the cooler examples paid more for their cooler coolers. Or cooler solution.

I totally agree. I know the chip is a hot chip anyway with the extra layer of 3D cache. I'm only maxing it to 86 degrees when on full load which isn't ever going to happen in gaming. I have a 4090 too which as you can imagine, with the shear size of it, the inside of the case is going to run warm. I updated the chipset too. I've done everything on my end to try and get the best possible result, I'm not going to fiddle around anymore. Of course I'd like the idle temperatures to be lower but baring that, I'm just going to leave it

 

[Karadjgne]

With a full custom loop with 2x 240mm rads, my Ryzen 3700x idles between 32-40 normally with a 22° ambient.

Liquid cooling, whether aio or loop has multiple different properties, involving the rad, the tubing, the pump itself, the cold plate etc. Increasing flow rate may or may not have any affect.

Coolant has a very large capacity to absorb the heat given off by the cpu. Think of it like a line of ppl carrying buckets and you have a hose. The water comes out of the hose into the bucket of the first person starts filling it and moves on, then empties the bucket into a tank. It's not going to matter if the bucket is ¼ full or ¾ full, the tank only gets filled so fast. The flow being the length of time they stood with you waiting for water. Making the line move faster, the buckets are less full. End result is same amount of water is dumped into the tank, no matter how fast the flow.

Moving to performance from balanced changes not only the pump speed, increasing flow, but also changes the fan curves. The fans spin slightly faster at idle and will get faster, sooner. That pushes more air, harder through the radiator, making it more effective at removing the transfered heat. That in turn keeps the coolant at a lower temp, which makes it easier and faster to transfer the heat from the cpu to coolant. Cpu temps go down.

That's the basics, there's a lot more involved, such as ambient temps, case temps, location of rad, push or pull fans, fan speeds, fan types, size of tubing, loop resistance, head pressure etc.

As I said, the job of the cooler is to maintain the temp of cpu within acceptable limits. I should clarify, those limits are what's acceptable to the cpu, not you personally. The cpu decides if it's too warm or not and will increase fans or flow or lower clock speeds and/or voltages accordingly. Your input to the cpu determines by how much that happens, by settings like silent or balanced or performance which are presets, or custom which is manual manipulation.

Fan curves are personal, and different to everyone, unless using a preset. You can set whatever starting temp you choose, with whatever fan speed you choose, with whatever ramp up speed according to temp you choose. So you could set the fans to spin at 80% from 0-60° and an instant climb to 100% at 61°+. Or set fans at 40% at 20° with a long slow ramp up to 70% at 90° only going to 100% at 91°+. Or anything else. Totally your choice.

I generally start with Silent mode, see how it performs, and then modify that to my taste. Asus, and maybe other software, has fan recognition in its software, so using that it can discover the properties of the fans and help with tailoring the fan curve. Presets are nice and simple, but I find custom settings are generally better all around.