[SOLVED] Ryzen 3600 long term safe overclock values

[Danymq]

Okay a few days ago I upgraded to a ryzen 3600 and the performance it was well worth it (came from a 1500x), however the weather is really hot (google says 32°C but it fells like 35°C. so the procesor at stock with PBO, boosts between 4 to 4.2ghz, but the temps are not so good, the go straigh up to 76°C and then slowly go up to 81-84°C.

So I started testing a fixed OC and I managed to get it to 4.2ghz with 1.2v (1.18 underload because of vdrop) and the temps when I start Cinebech start at around 60, and slowly rise to around 71°C.

But last nigh I read in an Overclock forum that fixed OC could kill a Ryzen 3000, and that is better to leave it just stock because in the long term it could kill the chip, so now I don't know wheter to leave it stock with higher temps and slightly slower performance, or keep the OC.

My specs are:

1. Asrock B450 HDV R4.0 (I know this is terrible for OC, 4 phases, no heatsink, so I have put a 80mm fan to blow air straight to the VRMs,
2. 16GB Corsair Vengence 3200 Cl16 (with Ryzen safe optimized timings)
3. Ryzen R5 3600 with Arctic Freezer 33 eSports edition.
4. Msi GTX 1650Super Ventus XS OC
5. A non certified Cooler Master 550w (planing to upgrade it soon)
6. Case with 120mm intake fan @1200rpm and 2x 80mm exhaust fans @1600rpm.

So... I'm Happy with my temps, and the performance in single score and multi score is great, PPT, TDC and EDC are all below 100% under full load. So my only question here is: Is this manual overclock going to degrade my chip in a long term?

 

[CountMike]

Okay a few days ago I upgraded to a ryzen 3600 and the performance it was well worth it (came from a 1500x), however the weather is really hot (google says 32°C but it fells like 35°C. so the procesor at stock with PBO, boosts between 4 to 4.2ghz, but the temps are not so good, the go straigh up to 76°C and then slowly go up to 81-84°C.

So I started testing a fixed OC and I managed to get it to 4.2ghz with 1.2v (1.18 underload because of vdrop) and the temps when I start Cinebech start at around 60, and slowly rise to around 71°C.

But last nigh I read in an Overclock forum that fixed OC could kill a Ryzen 3000, and that is better to leave it just stock because in the long term it could kill the chip, so now I don't know wheter to leave it stock with higher temps and slightly slower performance, or keep the OC.

My specs are:

1. Asrock B450 HDV R4.0 (I know this is terrible for OC, 4 phases, no heatsink, so I have put a 80mm fan to blow air straight to the VRMs,
2. 16GB Corsair Vengence 3200 Cl16 (with Ryzen safe optimized timings)
3. Ryzen R5 3600 with Arctic Freezer 33 eSports edition.
4. Msi GTX 1650Super Ventus XS OC
5. A non certified Cooler Master 550w (planing to upgrade it soon)
6. Case with 120mm intake fan @1200rpm and 2x 80mm exhaust fans @1600rpm.

So... I'm Happy with my temps, and the performance in single score and multi score is great, PPT, TDC and EDC are all below 100% under full load. So my only question here is: Is this manual overclock going to degrade my chip in a long term?

Don't believe everything you read. OC you got is better than auto with safer voltage and temperature than most BIOS would give you on auto.

 

[Nightguest]

So my only question here is: Is this manual overclock going to degrade my chip in a long term?

Your values "should" be ok, what other people are referring to are from early on when some people thought that 1.3v was safe no matter what, unfortunately, a few people found out the hard way that it was not.

With Ryzen 3000, each CPU has its own safe limit, to find what your safe limit do this:

1. Deactivate your manual OC and max out PBO. (Make sure LLC is auto).
2. Close all monitoring software except for HWiNFO64
3. Scroll down to CPU Core Voltage SVI2 TFN - this is how much volts your CPU is actually getting (as opposed to Core VID is what it is requesting)
4. Load Prime95 or OCCT and do a small FFT test on all cores.
5. Watch your SVI2 voltage, it will jump around a bit and then it will settle at something, this 'something' is your FIT Voltage and is the maximum safe volts for your CPU.
6. Do note that this voltage is pretty unique per CPU. I've seen some people mention up to 1.3, others have mentioned as low as 1.2.

You should not exceed your FIT voltage, and if you want to be epicly safe, go below it. Also, do keep in mind that heat also plays a role in this, you can be in a 'safe' voltage, but if you're hitting 80C + constantly, you may run into degradation soon.

If you have a relatively new CPU, you should be able to get some really good voltages. The early batches were not great, but since February this year or so, there have been more and more that hit good frequencies on lower voltages.

 

[Danymq]

Your values "should" be ok, what other people are referring to are from early on when some people thought that 1.3v was safe no matter what, unfortunately, a few people found out the hard way that it was not.

With Ryzen 3000, each CPU has its own safe limit, to find what your safe limit do this:

1. Deactivate your manual OC and max out PBO. (Make sure LLC is auto).
2. Close all monitoring software except for HWiNFO64
3. Scroll down to CPU Core Voltage SVI2 TFN - this is how much volts your CPU is actually getting (as opposed to Core VID is what it is requesting)
4. Load Prime95 or OCCT and do a small FFT test on all cores.
5. Watch your SVI2 voltage, it will jump around a bit and then it will settle at something, this 'something' is your FIT Voltage and is the maximum safe volts for your CPU.
6. Do note that this voltage is pretty unique per CPU. I've seen some people mention up to 1.3, others have mentioned as low as 1.2.

You should not exceed your FIT voltage, and if you want to be epicly safe, go below it. Also, do keep in mind that heat also plays a role in this, you can be in a 'safe' voltage, but if you're hitting 80C + constantly, you may run into degradation soon.

If you have a relatively new CPU, you should be able to get some really good voltages. The early batches were not great, but since February this year or so, there have been more and more that hit good frequencies on lower voltages.

Thanks for your answer I purchased this chip in "preorder" in amazon the past month so my guess is that it's a recent chip, my SV12 settled at around 1.35v after 6 minutes, however it also was still jumping between 1.3 and 1.38 sometimes, here is a screenshot:

Everything was stock except for the XMP profile. Also, I don't like the Power reporting deviation feels a bit too high.

 

[Nightguest]

Thanks for your answer I purchased this chip in "preorder" in amazon the past month so my guess is that it's a recent chip, my SV12 settled at around 1.35v after 6 minutes, however it also was still jumping between 1.3 and 1.38 sometimes, here is a screenshot:

No problem at all, it seems that you have a very nice CPU.

That means that your OC is perfectly safe, if you find a way to get your temperatures under control a bit, you can even bring up your voltage to see if you can hit 4.3 (or even 4.4) all-core.

If you do get the heat controlled, I'd still recommend that you do not go over 1.28v just to be safe.
(That's what I'd do if I had your CPU, and I won't advise things to people that I'd not do myself).

For reference, I've got a 3700x that has a fit of 1.25v
I run 4.2 all core at 1.218v (purely for temperature reasons, it's much more chilled this way and I don't notice any performance drops).

If you want to squeeze our more performance, try and tighten your RAM timings a bit.
You said that you have the Corsair Vengence, (you may have the same as I do), if that's the case then tighten your timings to CL14.
I hit 14-17-17-17-34 at 1.4v and I've not had the time to really see what else I can do (work takes up all my time), but I do have a feeling that 3600CL16 should be possible.

The one caveat to that is that the Corsair Vengence 8gb sticks come in 2 dies, Samsung C-die and a low-grade B-die.

If you have the B-die variant, then 1.4v is perfectly fine, but if you have the C-die, do you exceed 1.35v. I've even heard that lowering the voltage on C-die sometimes helps stability for the OCs.

Unfortunately, the only real way to check what you have is by opening your case and looking on the RAM stick itself. What you're looking for is a version number (an amazing thing that corsair does to make identifying easier). If your 16Gb is 2x8gb sticks (I hope that it is) then the version numbers mean:

4.31: B-die
4.32: C-die

If you have something else, then the full list is here.

Note that this is important as sometimes thaiphoon burner gets lost between c-die and b-die.
(If you have b-die, don't jump too high just yet, officially this is 'down binned' b-die' and that means that it is sometimes not the good stuff that gets people hyped about - you'll just have to test what you have.)
If you have c-die, then this place will be of great help to you.

Another way to get more performance is with FCLK overclocking.
If you cannot get 3600 on your RAM then see if you can increase your FCLK to 1900.
Generally speaking, you want FCLK and RAM speeds to be 1:1 (so 3200:1600) as having it otherwise will force a latency penalty. However, the exception to that rule is if you can get your FCLK to be over 200mhz faster than your RAM speeds.

A note on FCLK
First, not all processors will hit 1900, bad ones will be stuck at 1700, most go to 1800, the better ones go to 1900. This is because you're playing the silicon lottery all over again, this time with your IO die.

That means that you can have an excellent all core OC, but since the IO die is a different chip entirely, that needs to be good as well.

 

[prophet51]

Thanks for your answer I purchased this chip in "preorder" in amazon the past month so my guess is that it's a recent chip, my SV12 settled at around 1.35v after 6 minutes, however it also was still jumping between 1.3 and 1.38 sometimes, here is a screenshot:

Everything was stock except for the XMP profile. Also, I don't like the Power reporting deviation feels a bit too high.

My 3600 pulls 120w in the same situation so I think your board is under reporting PPT. 1.25-1.275v should be a very safe OC voltage, you will easily get 4.3ghz or higher with that. I would definitely try to get the memory to 3600mhz like the other poster mentioned.

edit - looks like your EDC is limited to 70A but the stock limit is 90A and it runs at 90A for me even with EDC set to 250A.

 

[Danymq]

No problem at all, it seems that you have a very nice CPU.

That means that your OC is perfectly safe, if you find a way to get your temperatures under control a bit, you can even bring up your voltage to see if you can hit 4.3 (or even 4.4) all-core.

If you do get the heat controlled, I'd still recommend that you do not go over 1.28v just to be safe.
(That's what I'd do if I had your CPU, and I won't advise things to people that I'd not do myself).

For reference, I've got a 3700x that has a fit of 1.25v
I run 4.2 all core at 1.218v (purely for temperature reasons, it's much more chilled this way and I don't notice any performance drops).

If you want to squeeze our more performance, try and tighten your RAM timings a bit.
You said that you have the Corsair Vengence, (you may have the same as I do), if that's the case then tighten your timings to CL14.
I hit 14-17-17-17-34 at 1.4v and I've not had the time to really see what else I can do (work takes up all my time), but I do have a feeling that 3600CL16 should be possible.

The one caveat to that is that the Corsair Vengence 8gb sticks come in 2 dies, Samsung C-die and a low-grade B-die.

If you have the B-die variant, then 1.4v is perfectly fine, but if you have the C-die, do you exceed 1.35v. I've even heard that lowering the voltage on C-die sometimes helps stability for the OCs.

Unfortunately, the only real way to check what you have is by opening your case and looking on the RAM stick itself. What you're looking for is a version number (an amazing thing that corsair does to make identifying easier). If your 16Gb is 2x8gb sticks (I hope that it is) then the version numbers mean:

4.31: B-die
4.32: C-die

If you have something else, then the full list is here.

Note that this is important as sometimes thaiphoon burner gets lost between c-die and b-die.
(If you have b-die, don't jump too high just yet, officially this is 'down binned' b-die' and that means that it is sometimes not the good stuff that gets people hyped about - you'll just have to test what you have.)
If you have c-die, then this place will be of great help to you.

Another way to get more performance is with FCLK overclocking.
If you cannot get 3600 on your RAM then see if you can increase your FCLK to 1900.
Generally speaking, you want FCLK and RAM speeds to be 1:1 (so 3200:1600) as having it otherwise will force a latency penalty. However, the exception to that rule is if you can get your FCLK to be over 200mhz faster than your RAM speeds.

A note on FCLK
First, not all processors will hit 1900, bad ones will be stuck at 1700, most go to 1800, the better ones go to 1900. This is because you're playing the silicon lottery all over again, this time with your IO die.

That means that you can have an excellent all core OC, but since the IO die is a different chip entirely, that needs to be good as well.

Thanks for all this info, according thaiphoon burner, my Dram is a Hynix A-die, however I think I will leave it at 4.2ghz 1.2v for the moment, because my motherboard is not the best for OC, and my PSU seems to be at the limit of tis life and gets really hot, so I won't push my luck any further. I really appreciate the info, it will definitely help me after I upgrade my case and PSU (one that at least is 80+ bronze).

Also my monitor is 75hz, and my graphics is a GTX1650S so at the moment the CPU performance is good enough for the rest of my components.


Prepared by Thaiphoon Burner Super Blaster
-------------------------------------------------------------
MEMORY MODULE
-------------------------------------------------------------
Manufacturer : Corsair
Series : Vengeance LPX
Part Number : CMK16GX4M2B3200C16
Serial Number : Undefined
JEDEC DIMM Label : 8GB 1Rx8 PC4-2133P-UA0-10
Architecture : DDR4 SDRAM UDIMM
Speed Grade : DDR4-2133P downbin
Capacity : 8 GB (8 components)
Organization : 1024M x64 (1 rank)
Register Manufacturer : N/A
Register Model : N/A
Manufacturing Date : Undefined
Manufacturing Location : Taiwan
Revision / Raw Card : 0000h / A0 (8 layers)
-------------------------------------------------------------
DRAM COMPONENTS
-------------------------------------------------------------
Manufacturer : Hynix
Part Number : H5AN8G8NAFR-TFC
Package : Standard Monolithic 78-ball FBGA
Die Density / Count : 8 Gb A-die (Deneb / 21 nm) / 1 die
Composition : 1024Mb x8 (64Mb x8 x 16 banks)
Input Clock Frequency : 1067 MHz (0.938 ns)
Minimum Timing Delays : 15-15-15-36-50
Read Latencies Supported : 16T, 15T, 14T, 13T, 12T, 11T, 10T...
Supply Voltage : 1.20 V
XMP Certified : 1600 MHz / 16-18-18-36-54 / 1.35 V
XMP Extreme : Not programmed
SPD Revision : 1.0 / January 2014
XMP Revision : 2.0 / December 2013

 

[Danymq]

My 3600 pulls 120w in the same situation so I think your board is under reporting PPT. 1.25-1.275v should be a very safe OC voltage, you will easily get 4.3ghz or higher with that. I would definitely try to get the memory to 3600mhz like the other poster mentioned.

edit - looks like your EDC is limited to 70A but the stock limit is 90A and it runs at 90A for me even with EDC set to 250A.

Yes, power deviation confirms that, when I manually OC, the power deviation disappears and reports even or higher tempos and power draws than in aunto OC, so in auto OC the motherboard is undereporting power consumption.