GA-AB350 Gaming 3 default CPU VCORE SOC set too high.

spektrem

Prominent
Nov 17, 2017
12
0
520
Hi everyone,
I have been running my new system since a month now and I haven't experienced crashes, BSODs, freezes or any other kind of performance related issues, however I discovered that the CPU VCORE SOC is set too high and I really wish to know if thats a Bug or if it's correct. I have a Ryzen 7 1700 at stock settings installed on the Gigabyte AB350 Gaming 3 motherboard with 8 GB (2x4) of the Corsair Vengenace LPX 3000 MHz CL15 RAM. The XMP profile is enabled and the memory is perfectly running at 3000 MHz and all other settings are defaulted by the motherboard. The thing is that the CPU VCORE SOC is set by the motherboard itself at 1.250 V and seems to be too high according to AMD that said the following value shouldn't exceed 1.2 V. So my motherboard works by Offset and I am able to set the CPU VCORE SOC to Auto, Normal or values as high as +300 mV. I tried to set it to Normal and the CPU VCORE SOC dropped from 1.250 V to 1.116 V. I ran Prime95 for 30 minutes and I have done 5 consecutive runs on Cinebench R15 with this Voltage and everything was fine. My question is: Should I keep using the Normal setting or should I revert to the Auto setting? It is normal for the motherboard to default itself to 1.250 V?

Thank you.

System:
RX 480 Nitro+ OC
Ryzen 7 1700
Gigabyte AB350 Gaming 3 REV 1.x - BIOS Version F22b
Corsair Vengeance LPX DDR4 3000 MHz CL15 8 GB (2x4)
 
Solution
Auto settings has a tendency to over-volt, and sometimes massively. However, running Cinebench is not a good indicator of load voltages/temps. It's a benchmark after all, not a stress testing app.

You mentioned you are running Prime for stressing. You should run it at Small FFTs only which will stress just the CPU and cache. This will give you the max voltage and heat, to enable you to have a stable CPU OC.

So, at your settings (and with something like HWMon running in the backround) run Prime 95, Small FFts. Monitor your temps with HWMon. If it runs Prime, with no errors (lets say for 30mins as a starting point) and temps don't go above 80c, then you have a good point to start with. I'd expect it to run fine with your CPU at stock speeds.

If it runs fine, you could do a couple of things. Try get a lower Vcore setting, and or OC that CPU, which it is probably begging to do.

I wouldn't worry about that Vcore setting either way. It's pretty normal. An unless you want to OC or undervolt, it's fine.

edit: apologies, I've realised you were referring to SOC voltage rather than CPU. The Vcore SOC is the voltage going to the chipset. On auto, 1.25 is fine. You may be able to reduce that, by tinkering with it. Use the offset - (minus) value. But I would leave it on auto, unless you are really tweaking an OC to get min voltages.

My other points about the CPU and testing are less relative, but something to consider.
 


So, as you said I overclocked the CPU to 3.7 GHz. I did increase the frequency but I haven't increased the CPU VCORE SOC nor the Dynamic VCORE, I did set them both on Normal. The CPU VCORE (Dynamic VCORE) is set to 1.2 V and the CPU VCORE SOC is set to 1.116. I ran Prime95 and all cores did pass the first test, however once the second test started one core crashed and reported "Rounding was 0.5, expected less than 0.4" while all the others kept going. By setting the CPU VCORE SOC on Auto again, so at 1.250 V will it be more stable? Or should I increase CPU VCORE? If so, by how much?

EDIT: I ran the normal benchmark. XMP Profile enabled. Max Temp with stock cooler 71 C°.
 
that failure, means the CPU voltage isn't stable. up it a notch, test again. Once it passes you can work back from there on other voltages. The SOC, 'can' play a part, but leaving it on auto for now is fine. The main things to worry about if OC'ing CPU is vcore, and LLC. Although it's my understanding the AB350 doesn't have LLC settings. LLC can help with higher OC's, like 4.1/4.2 with Ryzen, as the architecture maxes out at that speed.

edit: temps at that OC seem good. If you have a good cooler, and get vcore as low as it can go, you may be able to push beyond 3.7.

sorry, i'm trying to match my mobo settings to yours. I'm running an Asus mobo, and although the settings for the most part to the same thing, they are called different things by the manufacturer. When I say vcore, i'm talking about CPU vcore/VDDRCPU voltage. It may be called something else on yours. But, as you may have guessed it's the CPU voltage that I'm talking about not the SOC.
So by upping it a notch and testing you will get it Prime95 stable, and can work out other components after the CPU is stable. My 1600x is at 3.9ghz with 1.3vcore. I can prob push 4ghz+1, with some extra voltage, but I'm happy with the low temps/voltage and extra performance.
 


Well I'll try to run Prime95 again with Auto SOC Voltage, since I have done the test with 1.116 V, so on Normal. If it does the same I'll bump the VCORE up by +0.012 V or should I begin with higher voltage?

I did read somewhere that the SOC voltage shouldn't go past 1.2 V anyways, thats why I asked.

EDIT:
So the CPU VCORE is called Dynamic VCORE
The SOC VCORE is called CPU VCORE SOC
I currently have the CPU at 3.7 GHz with a vcore of 1.2 V and and SOC voltage of 1.116 V. As I said before one core crashed with these settings. So I wanted to know if I might fix it by setting the SOC on AUTO so at 1.25 V
 
when set on auto, the voltages can go beyond normal parameters. For example, some mobo's offer an OC setting. This may get your CPU to 3.9ghz, but 'occasionally' the voltage required to main stable operation (on auto) may put the voltage up around the 1.49/1.50 vcore mark ( for Ryzen CPU's). That's pretty scary. But, it's only for a split second. As long as it's not sustained voltage that high, the CPU will be okay.

So by setting things manually, you can reduce the vcore/soc/mem voltage to it's min, to reduce heat and push the CPU/Mem to it'e limits and maintain your OC. As I've said, focusing too much on the SOC voltage right now, is of little benefit. It will yield no performance impact, but rather only may help with tweaking your settings down the line.

Focus on the CPU/Mem, and when you have it 100% rock solid, then look at reducing other voltages, or tighter timings on ram, or whatever.
 


No, to fix the prime crash (cpu instability) increase the vcore (dynamic vcore), the SOC is not important at this stage.

BTW 1.2 Dynamic core is good. But obviously not stable. Try 1.225 or something and test prime again. AMD has said going up to about 1.37v is fine for the CPU, but ideally you want it as low as possible. SO start low, and work up on the voltage until your CPU passes prime at 3.7 for an hour. You'll know it's pretty stable then. Although, personally I run it for 8hrs overnight. That's my opinion, because I wanna know my OC is 100%
 


Okey thank you. I am pretty new to this so sorry if I am asking way too many questions. To clear things out one last time: Should I leave the SOC on Normal (1.116 V) or set it on Auto (1.250 V)? Then how much should I increase the cpu vcore to begin with?

edit: just saw your last post, thank you man :) will let you know how it goes. If you could answer my question about the SOC voltage anyways, I would appreciate it that.
 


good. no probs. ask as many questions as you need to, to understand. I'm happy to help. report back and let me know how you get on, or PM me if you need any more help :)

Just to be clear, if your Vcore (Dynamic Vcore) is at 1.2, raise it up in small increments, and test each time. Once you pass Prime95 for 1hr with no errors, you are good to go. Just keep in mind for absolute 100% stability it's advised to test for longer, as the temps still creep up even after 1hr testing. You don't want your temps to go above 80c, and your vcore (Dynamic) going above 1.375. Apart from that, good luck :)
 
Solution


on the SOC voltage, set it to normal. It won't do any harm. Look at it as separate thing to consider though. If you find any difference between normal and auto (for which i'd be very surprised) then switch it back to auto, or bump it to the next level above normal. Even that will be better than the Auto settings, which as I've said will overvolt.

 


So I did increase the voltage to 1.225 V. It did crash after 25 minutes of Prime95 (75 C° max temp). Everything is perfect while gaming, the CPU is a beast, I went from 1410 CB to 1650 CB. I will give it more voltage but not now, I want to make sure temp isn't too high. One thing I don't really understand is why when I am gaming the CPU is at 50 C° and sometimes it spikes up to 60 to quickly go back to 50 again. CPU utilisation is not increasing, but the temp is spiking sometimes.
 
you have to consider here, that you are talking about an 8 core CPU. so even a small change in processor speeds will impact the temps (even just for that split second). I see the same with my 1600x. Idling along, and if I open a chrome tab, the temps spike on the package temp up 10c and then back down. It sseems to be normal.