R5 3600 overclocking and strange ram behavior

[ManofGod1000]

Ok, good morning. I have a Asrock B450 Fatality Gaming mITX mainboard with the latest bios installed. It has a Ryzen R5 3600 installed and the reason I am overclocking is to get the temperatures in check. (Reduced voltage in use.) Now, I have a set of GSkill 2 x 8GB 3600 DDR4 ram and a set of Oylo 2 x 16GB 3200 DDR4 ram. I was using the 3600 ram but, when I would overclock and down volt, Red Dead Redemption 2 would randomly crash but would not do so with everything set to default in the bios except for XMP being enabled.

So, I figured I would try to the other ram from my other machine, the 3200 speed ram, since I was getting IRQ errors BSOD when I was not supplying enough voltage to the cpu. However, the exact same issues happened with RDR2 well everything else worked fine. Ok, that is strange, I thought and just to see what would happen, I upped the voltage of the ram from 1.35v to 1.4v. Just like that, BOOM, the problem was gone. This is really strange to me, am I missing a step in the OC process, like maybe not enough SOC voltage or something else? Or does the board and CPU just prefer a bump in memory voltage just to work well? (Oh, and if I would lower the memory speed without upping the voltage, the system would run completely stable then, as well.)

Thanks.

 

[bassnut]

Some boards especially in the older 300 and 400 series have a harder time getting that faster ram speed. AMD and board manufacturers fixed allot of the issues and speeds have gone up. By upping the ram voltage you helped stabilize your ram.

 

[ManofGod1000]

Some boards especially in the older 300 and 400 series have a harder time getting that faster ram speed. AMD and board manufacturers fixed allot of the issues and speeds have gone up. By upping the ram voltage you helped stabilize your ram.

I agree but, it seems odd that both the sets of ram were exhibiting the very same issues. Almost seems like I am missing something key in overclocking an R5 3600 that is different from the previous gen processors.

 

[bassnut]

I'm not the guy to talk ram with. The 450 and 470 boards most were limited to 3200 to 3400 if my memory is right. It wssnt till the new X570 series came out that we have been able to go higher. Some of that's due to better imc but not all.

 

[Kenrou]

https://www.asrock.com/mb/amd/fatal1ty b450 gaming k4/index.asp#Specification

 

[Zerileous]

It could just be down to the power delivery and topology of the board. I wouldn't worry about what works.

Also you're comparing a 32gb kit of lower speed to a 16gb kit of higher speed. Both tax the system in different ways. If you really want a like for like comparison I would just clock down the 16GB kit.

Anyway this is a nice discussion that can really get into the weeds, but what I want to talk about is your temps! Why do you feel the need to undervolt your CPU? What cooler are you using? What temps are you getting (during load, for example a task that stresses all cores like cinebench or a stress test)? What case and fans? Of all CPUs you should not have to undervolt a 3600 due to temps!

FWIW it's not appropriate to compare temps across brands or hardware generations. Every CPU reports temps differently, so it's important to reference the manufacturer spec. 15+ years ago I would set a thermocouple next to the die under the CPU cooler, and if that bad boy got into the 40s C I would start to worry. Now I consider load temps in the 70s to be good and anything under 100C is actually acceptable for the hardware. It's apples to oranges. So if you're seeing temps in the 80s or 90s and worried because of experience with past hardware that would be charcoal at those temps, don't worry. The temps themselves are likely similar if measured in the same way, but measuring has changed with new processes.

 

[trents]

Good questions and good input, Z.

 

[ManofGod1000]

It could just be down to the power delivery and topology of the board. I wouldn't worry about what works.

Also you're comparing a 32gb kit of lower speed to a 16gb kit of higher speed. Both tax the system in different ways. If you really want a like for like comparison I would just clock down the 16GB kit.

Anyway this is a nice discussion that can really get into the weeds, but what I want to talk about is your temps! Why do you feel the need to undervolt your CPU? What cooler are you using? What temps are you getting (during load, for example a task that stresses all cores like cinebench or a stress test)? What case and fans? Of all CPUs you should not have to undervolt a 3600 due to temps!

FWIW it's not appropriate to compare temps across brands or hardware generations. Every CPU reports temps differently, so it's important to reference the manufacturer spec. 15+ years ago I would set a thermocouple next to the die under the CPU cooler, and if that bad boy got into the 40s C I would start to worry. Now I consider load temps in the 70s to be good and anything under 100C is actually acceptable for the hardware. It's apples to oranges. So if you're seeing temps in the 80s or 90s and worried because of experience with past hardware that would be charcoal at those temps, don't worry. The temps themselves are likely similar if measured in the same way, but measuring has changed with new processes.

Temps are still critical, even with my 3600. On my Asrock B450 mITX mainboard, no matter the processor installed, when it hits 82C on the cpu, the computer will always spontaneously reboot. CPU does not matter, ram does not matter, power supply does not matter, nothing else seems to matter. My guess is there is a setting somewhere on the board that causes that behavior but, I am not certain where it is at.

I did some mainboard swapping around and now use my Asrock AB350 mITX Fatality Gaming mainboard in that system, instead. It just seems to have fewer quirks, short of the fact that post times are longer when using the 3600 and the latest bios. My Corsair H100i V2 needs to be RMAed so I am trying a Hyper 212 Evo I have sitting around and a Wraith Spire Cooler than came with my 1600.

 

[trents]

Temps are still critical, even with my 3600. On my Asrock B450 mITX mainboard, no matter the processor installed, when it hits 82C on the cpu, the computer will always spontaneously reboot. CPU does not matter, ram does not matter, power supply does not matter, nothing else seems to matter. My guess is there is a setting somewhere on the board that causes that behavior but, I am not certain where it is at.

I did some mainboard swapping around and now use my Asrock AB350 mITX Fatality Gaming mainboard in that system, instead. It just seems to have fewer quirks, short of the fact that post times are longer when using the 3600 and the latest bios. My Corsair H100i V2 needs to be RMAed so I am trying a Hyper 212 Evo I have sitting around and a Wraith Spire Cooler than came with my 1600.

Did you ever check the VRM temp on that mATX?

 

[Zerileous]

The hyper 212 EVO should be plenty for a 3600 stock. Also 82C should be normal. Can you post your full system specs. A good way to test if it's possibly VRM related though would be to use the Wraith Spire cooler, since it is a down facing fan it will flow more air over the VRM. The two boards appear to be very similar in terms of VRM. I think the major concern is the mosfets that do not have heatsinks located just above the socket. One would hope that these are for vSOC and not vCore, and do not really have a high load or need a heatsink, but I really can't seem to tell. Using a liquid cooler would limit airflow to the VRM and potentially exacerbate a VRM cooling problem.

The VRM consists of 3 parts, the mosfet, the choke or inductor, and the capacitor. The mosfet is an integrated circuit that can require cooling. I've attached a picture below. The mosfets of concern are circled in red. There is also a blue arrow pointing to the heatsink which is covering some, but not all of the mosfets.

The solution, if this is in fact the problem (which will be diagnosed by using the Spire cooler or using another fan to blow air directly onto that part of the motherboard), would be to add heatsinks to this area. It is relatively simple to do, but important to ensure that they do not touch any other components on the motherboard when installed. They can be a little pricey but I would suggest the Enzotech MOS-C1. They can also be difficult to find but there are some listings on a popular auction site.

 

[bassnut]

So.ething never touched on is case airflow .... what kind of case and what and how many fans are in the case + orientation of those fans to bring in cool air and exhaust hot air.

 

[ManofGod1000]

The hyper 212 EVO should be plenty for a 3600 stock. Also 82C should be normal. Can you post your full system specs. A good way to test if it's possibly VRM related though would be to use the Wraith Spire cooler, since it is a down facing fan it will flow more air over the VRM. The two boards appear to be very similar in terms of VRM. I think the major concern is the mosfets that do not have heatsinks located just above the socket. One would hope that these are for vSOC and not vCore, and do not really have a high load or need a heatsink, but I really can't seem to tell. Using a liquid cooler would limit airflow to the VRM and potentially exacerbate a VRM cooling problem.

The VRM consists of 3 parts, the mosfet, the choke or inductor, and the capacitor. The mosfet is an integrated circuit that can require cooling. I've attached a picture below. The mosfets of concern are circled in red. There is also a blue arrow pointing to the heatsink which is covering some, but not all of the mosfets.
View attachment 209348
The solution, if this is in fact the problem (which will be diagnosed by using the Spire cooler or using another fan to blow air directly onto that part of the motherboard), would be to add heatsinks to this area. It is relatively simple to do, but important to ensure that they do not touch any other components on the motherboard when installed. They can be a little pricey but I would suggest the Enzotech MOS-C1. They can also be difficult to find but there are some listings on a popular auction site.

Thanks for the reply. I think what I am running into on the B450 board is a bios cpu temp limit but I do not know where to adjust that. The reason I say that is because I have already installed my Asrock AB350 mItx board instead and it does not have that problem. Even when the cpu would get up to 85C, the machine still kept running. (This was in the last couple of days.) Wraith Spire, Noctua NH-u14s, Corsair H100i v2, it did not matter, it would reboot at exactly the same temp on my B450. (I have the Noctua NH-u14s in my NZXT H200 case.) The case I have my 3600 installed in will really only fit 120mm size coolers or less.

- - - Auto-Merged Double Post - - -

So.ething never touched on is case airflow .... what kind of case and what and how many fans are in the case + orientation of those fans to bring in cool air and exhaust hot air.

Airflow itself is fine for the cases themselves. It is just that for whatever reason, that B450 board will always cut off at the exact same temp, no matter the cpu, power supply or case. I am honestly thinking I am hitting a bios set ceiling that is by default but, I have no idea what. (Could be wrong but, the AB350 board does not exhibit this issue, which is why I am running the 3600 in it, instead.)

 

[bassnut]

Of course the board will shut down or throttle when it hits a certain temp …… Your job is to figure out why ….. poor cooling on CPU or MB or bad airflow where you cant get the heat out are your most likely culprits ….. figure out which and do something about it so your system will run much better.

EDIT ……

Here's an old post where I did just that and what I had to do to achieve what I wanted from my system. I know different issues but the same idea behind each problem.

https://www.overclockers.com/forums...FX-Tamer-Water-Cooled-Upgrade?highlight=tamer

 

[ManofGod1000]

Of course the board will shut down or throttle when it hits a certain temp …… Your job is to figure out why ….. poor cooling on CPU or MB or bad airflow where you cant get the heat out are your most likely culprits ….. figure out which and do something about it so your system will run much better.

EDIT ……

Here's an old post where I did just that and what I had to do to achieve what I wanted from my system. I know different issues but the same idea behind each problem.

https://www.overclockers.com/forums...FX-Tamer-Water-Cooled-Upgrade?highlight=tamer

The board shuts down at the exact same temp whether a 1600, 1700, 2600 or 3600 is being used. It does not matter whether it is in the NZXT case, the case I am using now for the 3600 and does not matter what cpu cooler I am using, literally. (The ones I am actually using, as I have already listed in this thread.) The AB350 Fatality Gaming mITX board does not exhibit this problem, at all and it is pretty much the same board as the B450 one.

Unless there is a faulty component on the board, I have to imagine that the problem is caused by a default setting in the bios. The question I have for that is, what setting could that possibly be?

Edit: Oh, and I am sharing the CPU temp thing now since it was asked. I simply gave up on trying to run the 3600 on the B450 with the ram speed flakiness and temp issues. I now have the 1600 in it with a NH-u14s and it is running at 4.0 GHz at 1.4v. (Does not make it to that hard 82C ceiling so it never has an issue.)

 

[Zerileous]

Ahh I thought it was having the problem with both boards. In that case it's probably a faulty board or a bios problem, but I'm thinking more like a corrupt bios than a setting.

 

[ManofGod1000]

Ahh I thought it was having the problem with both boards. In that case it's probably a faulty board or a bios problem, but I'm thinking more like a corrupt bios than a setting.

Well, if it is a corrupt bios, that would suck. I have pretty much had that going on since I bought it new back about 1 year ago. I did not realize it was an actual problem so I just stuck a better cooler on it and forgot it. I have updated the bios a few times over the last year.

 

[Woomack]

I was testing B350, Z370, B450 and X470 from these ITX ASRock series. B450/X470 ITX boards are overclocking worse with any memory kit. For example, X370 ITX could run at 3733 ... X470 could run at 3466 with the same memory. It's normal and at first, AMD was the one to blame as they failed 3 AGESA releases just after 400 chipset series and were making problems to mobo vendors. Later I have no idea if they fixed these issues or not ... looks like not but I wasn't following topics related to these motherboards.

I have no idea how these motherboards are working with Ryzen 3000 series but with 1000/2000, 3200 memory clock was max guaranteed on every memory kit. 3466 was max which I could set stable some memory kits, typically lower capacity Samsung based.

Regarding Ryzen CPUs and their temps. It's normal that on an average cooler like Hyper 212 there are over 80°C temps. Actually I'm surprised that not more as fully loaded 3600/3600X/3700X on Noctua U12A can pass ~87°C and it's a much better cooler. Corsair H100 is actually worse than the Noctua U12A.
75+ degrees is a general limit for 1000/2000 Ryzens at which mobo will show overheating. 95+ is a limit for Ryzen 3000 above which the CPU is throttling. You can probably set a higher TJ limit in BIOS but I don't remember exactly if it was available in that ITX ASRock.

 

[ManofGod1000]

I was testing B350, Z370, B450 and X470 from these ITX ASRock series. B450/X470 ITX boards are overclocking worse with any memory kit. For example, X370 ITX could run at 3733 ... X470 could run at 3466 with the same memory. It's normal and at first, AMD was the one to blame as they failed 3 AGESA releases just after 400 chipset series and were making problems to mobo vendors. Later I have no idea if they fixed these issues or not ... looks like not but I wasn't following topics related to these motherboards.

I have no idea how these motherboards are working with Ryzen 3000 series but with 1000/2000, 3200 memory clock was max guaranteed on every memory kit. 3466 was max which I could set stable some memory kits, typically lower capacity Samsung based.

Regarding Ryzen CPUs and their temps. It's normal that on an average cooler like Hyper 212 there are over 80°C temps. Actually I'm surprised that not more as fully loaded 3600/3600X/3700X on Noctua U12A can pass ~87°C and it's a much better cooler. Corsair H100 is actually worse than the Noctua U12A.
75+ degrees is a general limit for 1000/2000 Ryzens at which mobo will show overheating. 95+ is a limit for Ryzen 3000 above which the CPU is throttling. You can probably set a higher TJ limit in BIOS but I don't remember exactly if it was available in that ITX ASRock.

I think I found the setting but, I have not bothered changing it, since the NH-u14s cools really well, anyways. I will let you know in a bit when I look again in the bios.

Edit: Advanced\AMD CBS\NBIO Common Options - Processor Temperature Control.

 

[trents]

I was testing B350, Z370, B450 and X470 from these ITX ASRock series. B450/X470 ITX boards are overclocking worse with any memory kit. For example, X370 ITX could run at 3733 ... X470 could run at 3466 with the same memory. It's normal and at first, AMD was the one to blame as they failed 3 AGESA releases just after 400 chipset series and were making problems to mobo vendors. Later I have no idea if they fixed these issues or not ... looks like not but I wasn't following topics related to these motherboards.

I have no idea how these motherboards are working with Ryzen 3000 series but with 1000/2000, 3200 memory clock was max guaranteed on every memory kit. 3466 was max which I could set stable some memory kits, typically lower capacity Samsung based.

Regarding Ryzen CPUs and their temps. It's normal that on an average cooler like Hyper 212 there are over 80°C temps. Actually I'm surprised that not more as fully loaded 3600/3600X/3700X on Noctua U12A can pass ~87°C and it's a much better cooler. Corsair H100 is actually worse than the Noctua U12A.
75+ degrees is a general limit for 1000/2000 Ryzens at which mobo will show overheating. 95+ is a limit for Ryzen 3000 above which the CPU is throttling. You can probably set a higher TJ limit in BIOS but I don't remember exactly if it was available in that ITX ASRock.

Would you elaborate on that Woomack? What do you mean by "show signs of overheating"?

 

[Woomack]

Ryzen 1/2k CPUs run at much lower temps than 3k series. In theory, max temps for all are 95°C and that we can read in specs. In reality, there is a 20°C offset on older chips and also older chips run at much lower temps in general. For example, a typical max temp for 1/2k series is ~80°C ... but it includes 20°C offset where 3k has about 90°C on above-average coolers without any offset.
2k is like 1k after some improvements, 3k is a totally different chip. Simply can't compare them.

3k chips are designed to run up to 95°C and throttle down or at least limit their power so a chip will stay below that 95°C. In a typical situation, AMD stock coolers are too weak and are causing new CPUs to drop voltage and clock under full CPU load. In real they stay within a specified range but because of the heat, boost clock is lower during work. You can expect up to 200MHz lower boost clock on all cores because of low performing cooler. On the other hand, the difference between top coolers and these let's say average (but still better than stock AMD) is about 10°C. The difference in 1 and 2 fans on tower coolers is 1-2°C (something like U9/U12S/U12A coolers with 1 and 2 fans).

Once Tj is unlocked then you can see 115°C on a stock AMD cooler running stock R5 3600 in something like Prime95. At least when I was testing R5 3600/3600X then the CPU was still 100% stable at over 110°C.

The same because of a different memory controller, various chips act in a different way on new and old motherboards. I don't believe that older motherboards couldn't support higher memory frequency but then what would be the point to buy the most expensive motherboards with X570 chipset. No one sees any point in supporting old products which are bringing only losses and no one will invest in work on new BIOS for so many available motherboards when there are new products on the market and soon will be a new chipset.
There is no mobo manufacturer who was prepared on Ryzen 3000 when 400 series chipsets were released. Believe me, AMD is making problems with their CPU samples even to their partners. Later it looks like end-users get new motherboards with "beta" BIOS which is named as the official one. Updates and tweaking usually last the next 2-3 months and after that, we can say there is an official and stable release (sometimes earlier).
Excluding the top motherboard series, you can see that most motherboards get no BIOS updates after about half a year after release. The exception are top overclocking series like ASUS ROG or EVGA Dark which have a different kind of support and we actually pay a premium for these series so we expect more.
With 400 series chipsets and Ryzen 3000 was like all mobo manufacturers released only required updates so CPUs could work. There are no additional patches and tweaking to make them run like on X570 mobos. On the other hand, it meets general specification so the CPU declared frequency and memory at DDR4-3200. This is what 95% of users expect so it's fine from the sales point of view.