Ryzen 3700x "Overclocking" for simpletons...(and mainly for thermal reasons)...

[SheldonCooper]

Hi there,

I recently swapped a 2700x for a 3700x into the same mobo and with the same cooling...and I was surprised to see that the 3700x ran way hotter than the 2700x...even with the huge Noctua NH-D15 with 2 fans. Like at idle, it would sit around 50-60 degrees and under load would be around 70-75.

Obviously higher temps = more fan noise, and regardless of whether it's safe for the chip to run hotter, it's a pain in the ass hearing the fans whirring up and down all the time when i'm just browsing the interwebs - and even with days of fan curve messing...i've not found anything to tame the chip.

Until...I found a video on Youtube of some dude that just used Ryzen Master to set a all-core frequency of 4.3Ghz and a V-core of 1.25 and it lowered his idle and load temps. I'm not really someone who has the patients to deep dive into eeking every last bit of performance out of it...and this solution seemed like a really nice set it and forget it option (providing it was stable).

So I have just spent a quick hour doing some preliminary tests, starting at 4.1 and working up to 4.3Ghz @ 1.25v, using Cinebench as a quick "does it it immediately crash" test - and I've so far run 3 successful tests on each set of settings, and my score went up by between 400-500 points and thermals are a nice 40ish on idle and 65 under load.

It all looks sweet...but I can't help but have that feeling of it being too good to be true (or too easy to be true). Just wondering if anyone has any thoughts on whether this is legit? I have looked through reddit and everyone is doing different things and have differing opinions...could do with some reassurance from non-reddit people

Thanks!

 

[merlin6r]

I also have a Ryzen 3700x and have experimented in the same way. An overclock does indeed lower the temps.
It gives good all core scores (cinebench etc) but you lose the 4.4ghz for things like games. My Timespy score in 3dMark went down.

I've settled on using a -0.030v offset and leaving the CPU alone. This has lowered the temps and my boost still works.

I could go lower, but at -0.050v my cinebench scores start to go down. I'll try -0.040v at some point.

I've seen videos on youtube of people running their Ryzen 7/9 at 1.0v and saying its a good idea. Its not. Because of way these chips work it will all APPEAR to be fine, no crashes etc, but your performance will be crap.

Try a small offset (-0.010v) and work down....

PS I'm only using an Arctic Freezer 33 eSports - nothing as good as your Noctua and my temps are 38-44c right now. They do jump around a little but I'd rather have the boost...

 

[PieEater]

I would suggest that you make sure you are on the latest BIOS for your board and have downloaded the latest chipset drivers from AMD then experiment with the AMD Ryzen Power Plans in Windows. You may find out that your idle temps improve as the newer power plans (particularly "balanced") are supposed to allow the idle voltages to scale down better.

With regards to setting an all core overclock I would echo what merlin6r said that this may negatively impact on performance in less intensive workloads (gaming) where your CPU would normally allow one or more cores to boost above these speeds. Unless you have an exceptional chip that will sustain an all core overclock of 4.4Ghz (max boost speed) at voltages of 1.35v or less then I would tend to leave it to it's own devices.

It's not really helpful to suggest that all chips are equal and all 3700Xs will be able to acheive 4.3GHz on 1.25v and I would tend to discount anyone's advice that says this is true, there are so many variables involved and no single chip is exactly the same.

Unless you do a lot of rendering Cinebench is not a realistic benchmark to use for performance testing you would be better off with an in-game benchmark if your primary use is gaming. Cinebench is also not the best for stability testing, I use OCCT which normally will highlight any instabilities within 10 minutes.

In summary your load temps seem fine, your idle temps need adressing but I would first look towards BIOS and chipset updates to resolve this before messing around with things that may negatively affect your overall performance.

 

[SheldonCooper]

Thanks for the responses. Have had a week or so now with the CPU. I don't think the temps are as wild at idle as they were. Not sure why this was as my BIOS was on the latest one from when first swapping the CPU. I tried again messing with seeing what I could push the CPU too, and it crashes at anything other than 4.3GHz all cores @ 1.25v. I can't even set the strongest core to 4.4 and have the others at 4.3 - crashes.

When I let it just do its own thing and just game as normal, only 1 core turbos to 4.4GHz while all the others boost to only around 4.35Ghz, obviously it's not doing that all the time. I am very much all for leaving it as it is to do it's own thing - but I dunno, in my head (and please correct me if I am wrong) - other than the odd little burst of an extra 100mhz on a core here and there, surely the benefits of having all cores guarantee 4.3Ghz solid 100% of the time when required? That make sense?

I would say I predominately game on the computer - but I do also use it for Adobe products too. This is where the 4.3GHz all cores all the time thing for rendering Premiere and After Effects stuff would be ideal? If I all I am losing is 100mhz bursts here and there?

I hope all of that makes sense.

I will be honest, I've been on Ryzen for nearly 2 years now and I still don't understand how the CPU's boost and manage themselves power wise and I'm not really interested in learning cause I'm not really chasing every last ounce of performance. I just want something that will feed my 2080ti with frames and that I can get like..what...90% of the raw power out the chip?

I waffled there. Sorry. Any advice welcomed and again, really appreciate you taking the time to read and reply.

 

[PieEater]

Given you have arguably the best air cooler on the market you would be perfectly safe upping the voltage to 1.35v to see what if any gains that brings in terms of stability at over 4.3Ghz. Personally I had to use 1.375 to be fully stable at that speed and was happy to do so given my temps never exceeded 75c but we are moving into "at your own risk" territory with those voltages. If you can get a stable overclock over 4.3Ghz with voltages and temperatures that you are happy with then definitely stick with it. At 4.3Ghz you really are in swings and roundabouts territory, yes an all core overclock will be beneficial for all core workloads but equally games will benefit from single core performance that you will be missing out on. Like you say though we are only talking of 100Mhz and whether the gains in either scenario would be noticeable outside of benchmarks is debatable, which is why the consensus is it's not worth bothering with overclocking these chips unless you have a really good one.

Also the consensus with these chips is you are better off tweaking the memory and infinity fabric if you want performance gains, so have you done anything along those lines?

 

[Joxeon]

If you can get your all core OC to run 4.4 then you wouldn't be losing any single core speed either but you may need a bit more voltage at say 1.3 unless you have a great sample.