Ryzen 5 3600 turbo speeds

[nguyen]

It's not temps that kill them, its amperage. Thats straight from AMD engineers.

Higher temperature will lead to higher current due to higher leakage. That's how LN2 overclockers get away with insanely high voltages.

Something like this (EM is electron migration)

When you have free time just watch Buildzoid video on chip degradation (it's 1 hour long)

 

[SirPerfluous]

I have been running my 3700x at 1.36 set for almost a year now.

Haven't ran the supposedly shitty 1usmus software, but I would hazard a guess to say my chip is golden or better.
I've managed to run 1910 IF stable on low voltages, run incredibly tight timings and subtimings, and run at *higher* than the average zen 2 all core OC voltage while remaining unscathed so far.

I have lapped the IHS, use an arctic LF 280 to cool it, and seldom does it ever touch 70C in any of the workloads it sees. Temperature plays a big part in degradation.
My chip was made the 6th week of 2020, And since the end of may 2020, It has been running at its maximum possible potential.

4300 all core with 1.36 vcore set, 1900 IF and 1.1v soc.
LLC on medium which allows vdroop to compensate, and I see no greater than 1.32 under any significant load.
I used to run ~4330 and 1910 for a bit before that, but I couldn't ever get it stable in heavy cpu intensive loads. It became stable when I dropped the multiplier.
If I kept it low I could maybe even get the IF higher, but base clock tuning was hard for me on this motherboard.

4300 all core is the peak of this chip. Any more and it doesn't like to be stable below 1.33v under heavy load, which I don't really feel like doing for an extra 30 mhz.
1.325 is the safest maximum I will consider, for a heavy, thread intensive workload. I'm pretty sure you can get away with even more on the 3600 as it should be less heat-dense.
I can sorta achieve the same clocks with less set voltage and more LLC, but it's not 100% stable unless setup like this. so, 1.37v with the proper vrm settings is probably okay on Zen 2.

The way it works now is almost similar to pbo in that it runs higher voltages in low current situations, and under heavy load the voltage drops.
A bit more stupid than pbo, but it has worked great so far.

Also when attempting to setup pbo, I never saw the same performance as what I currently have now, and would often see it run heavy all core loads at 1.38v but only managing 4250mhz or worse.
I did manage to see some impressive single core frequencies up to 4.5 ghz on some cores in hwinfo with the edc bug and some time spent, but the voltages and heat always made me feel worse than this all core oc ever has.
It output far more heat, reaching mid 80's but having worse performance... I never had good luck with pbo, and it wasn't for lack of trying. I spent some time on more than one motherboard fiddling with it.

The 3600 might be more prone to degradation due to it being a lower bin, but I still think its just a lottery.
Perhaps some of the cpu's that are "degrading" were actually just terrible quality silicon that still managed to pass QC?
AMD had record yields for their 7nm node, maybe some of those weren't as good as they thought?
I guess with 7nm being new, It should be expected that the start of it would be a bit unpredictable..


Overall I think Zen 2 is a mixed bag, older chips are probably worse than newer ones, but I also think it takes waaaay more tuning than most people do.

There's not really a set safe voltage for Zen 2, each chip will vary.

Try not to get scared by the reddit posts, experiment and learn for yourself.
Or watch the videos nguyen provided, as those fairly reputable creators spent a number of hours working on that, just to help you be informed. and to maybe make money.

 

[Zach_01]

I'm really happy to see someone else also share my views (which is I firmly believe right). Those zen2 architecture requires excessive voltage already (design error maybe?) and that will certainly degrade the chip sooner or later. I would not push the chip more if I were you.

We are not talking about stock voltages and no one here has said that stock voltage would hurt the CPU (except you). The debate is about what is considered as safe static voltage and speed.

Personal opinion, there is no straight answer.
It depends on the type of load, quantity of load, temperature. In other words = by Current(A). High current with high temp is what causing electromigration and eventually thinned traces

If someone is running a 100% heavy AVX type of loads, voltage should be under 1.3V (1.25~1.3V)
With Cinebench type load it could be 1.3~1.35V
On low or middle loads it could even be 1.4~1.45V.

What would be the specific voltage depends on a lot of things. Chiplets, CCXs, number of active cores per CCX/CCD, type and % of load, current, temperature.

I’ve watched all BZ’s videos related to this degradation issue. And he admitted that degrading his 3700X has involve a lot of heavy type loads with 1.37V, when he shouldn’t cross 1.3V for those.

It’s very complicated to find what should be suitable for each type of load. And it’s far easier to find it when the load is 100%.
Gaming for example is for sure a very mixed bag of load type and % quantity even on 1 game. Let alone on the dozens of different games out there.

I usually not suggesting static OC because of the complexity it requires from the user to find the proper voltage and speed for his/her type or variety of usage scenarios.

Especially on 3000 static OC is not worth it unless someone is under clocking for better thermals.
For gaming it’s better to be left on auto boosting.

And my friendly advice is to stop running stress/burn tests all the time and try to find every last point on benchmarks and just enjoy your systems on whatever you do.
Unless someone’s hobby is to run benchmarks all day. By all means feel free to do it but be aware of the dangers.

 

[Muaadib]

Recently got a 3600 as a temp cpu for my B550 so i thought i might chip in. I used ClockTuner which said i have a silver sample and set the following (PBO,Voltages set to auto in BIOS):
4.225@1.245v MT
4.375@1.345v ST

It automatically switches between the two depending on the load. The MT setting in particular dropped my temps by almost 10c compared to stock while simultaneously getting higher scores in R20.

Just my 2 cents.

 

[brandon7171]

I have been running my 3700x at 1.36 set for almost a year now.

Haven't ran the supposedly shitty 1usmus software, but I would hazard a guess to say my chip is golden or better.
I've managed to run 1910 IF stable on low voltages, run incredibly tight timings and subtimings, and run at *higher* than the average zen 2 all core OC voltage while remaining unscathed so far.

I have lapped the IHS, use an arctic LF 280 to cool it, and seldom does it ever touch 70C in any of the workloads it sees. Temperature plays a big part in degradation.
My chip was made the 6th week of 2020, And since the end of may 2020, It has been running at its maximum possible potential.

4300 all core with 1.36 vcore set, 1900 IF and 1.1v soc.
LLC on medium which allows vdroop to compensate, and I see no greater than 1.32 under any significant load.
I used to run ~4330 and 1910 for a bit before that, but I couldn't ever get it stable in heavy cpu intensive loads. It became stable when I dropped the multiplier.
If I kept it low I could maybe even get the IF higher, but base clock tuning was hard for me on this motherboard.

4300 all core is the peak of this chip. Any more and it doesn't like to be stable below 1.33v under heavy load, which I don't really feel like doing for an extra 30 mhz.
1.325 is the safest maximum I will consider, for a heavy, thread intensive workload. I'm pretty sure you can get away with even more on the 3600 as it should be less heat-dense.
I can sorta achieve the same clocks with less set voltage and more LLC, but it's not 100% stable unless setup like this. so, 1.37v with the proper vrm settings is probably okay on Zen 2.

The way it works now is almost similar to pbo in that it runs higher voltages in low current situations, and under heavy load the voltage drops.
A bit more stupid than pbo, but it has worked great so far.

Also when attempting to setup pbo, I never saw the same performance as what I currently have now, and would often see it run heavy all core loads at 1.38v but only managing 4250mhz or worse.
I did manage to see some impressive single core frequencies up to 4.5 ghz on some cores in hwinfo with the edc bug and some time spent, but the voltages and heat always made me feel worse than this all core oc ever has.
It output far more heat, reaching mid 80's but having worse performance... I never had good luck with pbo, and it wasn't for lack of trying. I spent some time on more than one motherboard fiddling with it.

The 3600 might be more prone to degradation due to it being a lower bin, but I still think its just a lottery.
Perhaps some of the cpu's that are "degrading" were actually just terrible quality silicon that still managed to pass QC?
AMD had record yields for their 7nm node, maybe some of those weren't as good as they thought?
I guess with 7nm being new, It should be expected that the start of it would be a bit unpredictable..


Overall I think Zen 2 is a mixed bag, older chips are probably worse than newer ones, but I also think it takes waaaay more tuning than most people do.

There's not really a set safe voltage for Zen 2, each chip will vary.

Try not to get scared by the reddit posts, experiment and learn for yourself.
Or watch the videos nguyen provided, as those fairly reputable creators spent a number of hours working on that, just to help you be informed. and to maybe make money.

Thats really nice numbers, i have the same processor but with the X570 board, with a 4X8 kit G.Skill Ram