Ryzen 3700x heat

[Quantum]

Hello. Recently completed my build, I see that the cpu is currently at 63C. I believe this is on the hot side... should I ramp up the CPU fan? Im using the stock cooler that came with the cpu.

[Quantum]

I just did a cinebench cpu test. Here are the results after a "10 min (test throttling)" test on Multicore.
Overall score: 11485
It rendered the image like 10 times, and temps looked well.
Lowest temp recorded was 37C and highest temp recorded was 71C.
Whenever the CPU was under highload i saw that the CPU like to hover around 70-71C.
Does this look normal?

[JohnMars78]

Yes. If you're not bothered by the noise, you don't really have to change the stock cooler.

[Quantum]

The benchmark was also done while being on the AMD balanced power plan. Whenever it returned to idle state CPU temps came back down to 37C and sometimes 36C. voltage likes to just around 0.900-1.240V at idle and on heavy load in benchmark it reached 1.494V max. (according to HWMonitor)

[Quantum]

Should i even bother testing single core? Or should I even do the 30 min test for stability? I think everything checks out...

[Quantum]

JohnMars78 eredeti hozzászólása:

Yes. If you're not bothered by the noise, you don't really have to change the stock cooler.

Thank you!

[r.linder]

Quantum eredeti hozzászólása:

The benchmark was also done while being on the AMD balanced power plan. Whenever it returned to idle state CPU temps came back down to 37C and sometimes 36C. voltage likes to just around 0.900-1.240V at idle and on heavy load in benchmark it reached 1.494V max. (according to HWMonitor)

Should be reversed. FIT won't allow 1.49v at heavy load, ever.

Exactly why I don't use HWMonitor, it gets things wrong with Ryzen systems. Even HWinfo64 does.

The only really accurate tool for monitoring Ryzen is Ryzen Master.

[JohnMars78]

Escorve eredeti hozzászólása:

Quantum eredeti hozzászólása:

The benchmark was also done while being on the AMD balanced power plan. Whenever it returned to idle state CPU temps came back down to 37C and sometimes 36C. voltage likes to just around 0.900-1.240V at idle and on heavy load in benchmark it reached 1.494V max. (according to HWMonitor)

Should be reversed. FIT won't allow 1.49v at heavy load, ever.

Exactly why I don't use HWMonitor, it gets things wrong with Ryzen systems. Even HWinfo64 does.

The only really accurate tool for monitoring Ryzen is Ryzen Master.

It's just the peak. If it allowed that voltage it probably means that the CPU was not all that stressed in that particular moment.

[Quantum]

Hm, well I just checked again and everything seemed even better this time. I think all is well thanks for the help!

[r.linder]

JohnMars78 eredeti hozzászólása:

Escorve eredeti hozzászólása:

Should be reversed. FIT won't allow 1.49v at heavy load, ever.

Exactly why I don't use HWMonitor, it gets things wrong with Ryzen systems. Even HWinfo64 does.

The only really accurate tool for monitoring Ryzen is Ryzen Master.

It's just the peak. If it allowed that voltage it probably means that the CPU was not all that stressed in that particular moment.

As I said earlier, higher voltage is allowed with lower current draw, so ~1.5v is only seen at idle and while just using an internet browser, very light loads. Most gaming loads will range up to around 1.3~1.4v depending on how heavy it is on the CPU, and heavy all-core loads will generally show the safer voltage limits.

Easiest way to find the safest possible voltage is by using Prime95's small fft torture loop. If voltage averages out at 1.18v for example during the loop, that's the safe voltage for an undervolt if you want to have a 100% safe overclock with no risk of damaging the silicon. Earlier 7nm bins will be closer to ~1.2v for most of them while newer bins can get closer to ~1.25v.

And the problem with a lot of programs like HWMonitor is that they misreport for Ryzen based hardware, the most accurate of the bunch was HWinfo64 but it can still misreport discrete clocks and sometimes voltages. People also mistake VID for CPU voltage, but the actual CPU voltage that HWinfo64 uses is SVI2 TFN.

[Illusion of Progress]

Long time PC enthusiast, shorter time PC builder, recent Ryzen 7 3700X owner here.

Upgraded my PC (platform, meaning motherboard/CPU/RAM/cooling) and went with this CPU and actually settled on using the stock cooler for a bit, to space out costs. I usually avoid them due to temperatures and noise, mostly the latter. This one did better than I thought in noise so I actually put off upgrading a month or two. Yes it ran warm, but from everything I've learned, modern CPUs just run warmer, so I figured that was part of it (my results after upgrading supported this), and Escorve explained a lot on how Ryzen especially behaves.

I did end up upgrading to a Dark Rock Pro 4 a while ago. Temperatures under load dropped around a dozen or so degrees, and noise is a bit improved beyond what it was too. I agree with Escorve; if you're going to upgrade beyond the stock cooler (of the Ryzen 7 3700X at least; the lesser Ryzen CPUs indeed have a bit worse stock cooling), it only makes sense if you're going towards higher end air cooling or better, because the stock cooling for this particular CPU isn't too much worse than the cheaper air coolers, so they're not really worth changing to (unless free).

That being said, temperatures barely breaking into the 70C range on stock cooling seem far better than average. I only get that now after having upgraded. Typically near or into the low to mid 80C range is common for these CPUs on stock cooling; warm but common and not dangerous. I'm not sure if that one program is a light one (my testing was done informally just playing a few games) but if you're really barely breaking into 70C territory, that's quite cool and you prrobably don't even need to upgrade unless you want even better noise levels.

Escorve eredeti hozzászólása:

Quantum eredeti hozzászólása:

The benchmark was also done while being on the AMD balanced power plan. Whenever it returned to idle state CPU temps came back down to 37C and sometimes 36C. voltage likes to just around 0.900-1.240V at idle and on heavy load in benchmark it reached 1.494V max. (according to HWMonitor)

Should be reversed. FIT won't allow 1.49v at heavy load, ever.

Exactly why I don't use HWMonitor, it gets things wrong with Ryzen systems. Even HWinfo64 does.

The only really accurate tool for monitoring Ryzen is Ryzen Master.

My thinking was that 1.484V was simply the maximum value it logged, not that it reached that under load, which makes sense. That's about what I see on mine as well.

I've actually returned to HWMonitor on my system, because I love the simplicity over HwInfo64 (but there's a few QoL changes it could really use). But, I mostly use it to keep an eye on CPU/GPU temperatures. Ryzen Master probably is the more accurate thing to use for it. I initially abandoned HWMonitor thinking it couldn't read CPU core temperatures, since my older Intel platforms reported those, to find out these Ryzen CPUs don't report them the same way regardless of software.

[r.linder]

Illusion of Progress eredeti hozzászólása:

Long time PC enthusiast, shorter time PC builder, recent Ryzen 7 3700X owner here.

Upgraded my PC (platform, meaning motherboard/CPU/RAM/cooling) and went with this CPU and actually settled on using the stock cooler for a bit, to space out costs. I usually avoid them due to temperatures and noise, mostly the latter. This one did better than I thought in noise so I actually put off upgrading a month or two. Yes it ran warm, but from everything I've learned, modern CPUs just run warmer, so I figured that was part of it (my results after upgrading supported this), and Escorve explained a lot on how Ryzen especially behaves.

I did end up upgrading to a Dark Rock Pro 4 a while ago. Temperatures under load dropped around a dozen or so degrees, and noise is a bit improved beyond what it was too. I agree with Escorve; if you're going to upgrade beyond the stock cooler (of the Ryzen 7 3700X at least; the lesser Ryzen CPUs indeed have a bit worse stock cooling), it only makes sense if you're going towards higher end air cooling or better, because the stock cooling for this particular CPU isn't too much worse than the cheaper air coolers, so they're not really worth changing to (unless free).

That being said, temperatures barely breaking into the 70C range on stock cooling seem far better than average. I only get that now after having upgraded. Typically near or into the low to mid 80C range is common for these CPUs on stock cooling; warm but common and not dangerous. I'm not sure if that one program is a light one (my testing was done informally just playing a few games) but if you're really barely breaking into 70C territory, that's quite cool and you prrobably don't even need to upgrade unless you want even better noise levels.

Escorve eredeti hozzászólása:

Should be reversed. FIT won't allow 1.49v at heavy load, ever.

Exactly why I don't use HWMonitor, it gets things wrong with Ryzen systems. Even HWinfo64 does.

The only really accurate tool for monitoring Ryzen is Ryzen Master.

My thinking was that 1.484V was simply the maximum value it logged, not that it reached that under load, which makes sense. That's about what I see on mine as well.

I've actually returned to HWMonitor on my system, because I love the simplicity over HwInfo64 (but there's a few QoL changes it could really use). But, I mostly use it to keep an eye on CPU/GPU temperatures. Ryzen Master probably is the more accurate thing to use for it. I initially abandoned HWMonitor thinking it couldn't read CPU core temperatures, since my older Intel platforms reported those, to find out these Ryzen CPUs don't report them the same way regardless of software.

Ryzen Master also only reports effective clocks and not both effective and discrete clocks, unfortunately.

Reporting ~1.5v is always normal, it's typical idle voltage because FIT allows it, it's that simple. Only way to stop it is to cap voltage, but a manual voltage setting also stops the FIT from doing its job. So if you keep voltage at 1.3v and run P95 small FFT torture loops, you're pretty much guaranteed to degrade the silicon. Gradually, sure, but some people run the test for a whole day when testing OC stability, and that'll really mess it up. When I didn't know any better I caused minor damage to my 3900X with a few hours worth of that torture loop on ~1.387v and a number of other Zen2 users have noted degradation was noticeable even as low as 1.3v and below, it's silicon lottery that ultimately determines what the FIT allows as safe based on a given current.

[Illusion of Progress]

Yeah, I used to overclock (between the Core 2 to older Core i series days especially, there was so much extra performance/frequency to be had NOT to), but I'm 100% happy letting it do its own thing now, since it's not only gotten a lot more complicated, but it pretty much pushes itself as far as it can by default, at least with modern Ryzen systems. Intel systems may not do this as much by default and leave more headroom, so it's not totally out of the question on them, but even that headroom is pretty small compared to what it used to be.

[r.linder]

Illusion of Progress eredeti hozzászólása:

Yeah, I used to overclock (between the Core 2 to older Core i series days especially, there was so much extra performance/frequency to be had NOT to), but I'm 100% happy letting it do its own thing now, since it's not only gotten a lot more complicated, but it pretty much pushes itself as far as it can by default, at least with modern Ryzen systems. Intel systems may not do this as much by default and leave more headroom, so it's not totally out of the question on them, but even that headroom is pretty small compared to what it used to be.

Yeah we're reaching the physical limits of silicon it seems. Eventually it'll need to be phased out in favour of a superior material, but to my knowledge the only suitable replacement we have so far is graphene but it's both difficult and expensive to mass produce.

[Illusion of Progress]

Yep, with heat again becoming such a huge hurdle, like during the Netburst days, they're going to really have to overcome it in some big way or progress will be stifled more by it.

I remember hearing something about DNA even being possible but I don't recall the specifics.

Decades ago, quantum computers were supposed to be the next big breakthrough.

[Bad 💀 Motha]

Illusion of Progress eredeti hozzászólása:

Decades ago, quantum computers were supposed to be the next big breakthrough.

Those are coming; after 2025