SoC Voltage Increases Stability on Ryzen 2000?

[sl0011]

I've been struggling for a long time with my R7 2700. Random Black Screen crash with PC fans running but no peripherals. It started happening with an As rock basic motherboard that was later changed to a B450 Tomahawk, and the issued persisted (not a motherboard issue). Changing the RAM apparently solved the issue, FOR THE MOST PART. I would like to run a simple OC to 4 GHz at 1.35 – 1.375 V (most users report they can get 4 GHz+ no issues). The chip should be able to do it, it can complete CB R20 several times until the black screen happens again. From what I read online, some users report that increasing SoC Voltage solved the issue for them. So that's my question, should I increase the voltage to help with my CPU OC? 

 

Zen+ IPC is quite slow, so until I can upgrade, I would like to get more performance out of this chip. I have used it at 4 GHz for gaming numerous days without crashing, but the moment I stability test, the issue will appear again. That's why I say I think my CPU should be able to handle 4 GHz.

[RONOTHAN##]

SOC voltage mainly improves the memory stability. If it's crashing in R20, memory is likely not the issue. 

 

Are thermals for the CPU good? If you've got that level of instability id suspect that maybe the CPU is overheating. If they are fine, then sure, try increasing SOC voltage to 1.15V (where most chips sweet spot)  and see if that helps, but I don't think it should help much. It can't really hurt though. 

[sl0011]

29 minutes ago, RONOTHAN## said:

SOC voltage mainly improves the memory stability. If it's crashing in R20, memory is likely not the issue. 

 

Are thermals for the CPU good? If you've got that level of instability id suspect that maybe the CPU is overheating. If they are fine, then sure, try increasing SOC voltage to 1.15V (where most chips sweet spot)  and see if that helps, but I don't think it should help much. It can't really hurt though. 

I don't think temps are the issue here. 79.5 °C seems ok to me. It's just a theory, but I think some CPUs have weak memory controllers, that's why I think SoC Voltage could help with stability.

[BuzzingBee]

You could try increasing SoC voltage to 1.05 -  1.125, I wouldn't go above 1.15v as it can cause instability & degradation. If it didn't help, I guess your CPU is bad luck and you need to decrease clock speed for stability. 

[Guest]

2 hours ago, sl0011 said:

I've been struggling for a long time with my R7 2700. Random Black Screen crash with PC fans running but no peripherals. It started happening with an As rock basic motherboard that was later changed to a B450 Tomahawk, and the issued persisted (not a motherboard issue). Changing the RAM apparently solved the issue, FOR THE MOST PART. I would like to run a simple OC to 4 GHz at 1.35 – 1.375 V (most users report they can get 4 GHz+ no issues). The chip should be able to do it, it can complete CB R20 several times until the black screen happens again. From what I read online, some users report that increasing SoC Voltage solved the issue for them. So that's my question, should I increase the voltage to help with my CPU OC? 

 

Zen+ IPC is quite slow, so until I can upgrade, I would like to get more performance out of this chip. I have used it at 4 GHz for gaming numerous days without crashing, but the moment I stability test, the issue will appear again. That's why I say I think my CPU should be able to handle 4 GHz.

Zen+ and SK Hynix Kits (Corsair Vengeance) where a horrible pair. 

I've run through 5 sets of memory with my 2700X to find resolve to the issue.

 

Was easy, just don't use the SK hynix kits. All my samsung B-die kits where fine.

3733mhz stable and a 3966mhz top frequency because you can't uncouple the IF on Zen+. 

 

When enabling XMP on any kit, the system on auto already increases the SOC to 1.08v. 

You can safely use up to 1.200v for your XMP kit. More than that, you're aiming for gold competitively.

 

If you can't find stability with increased SOC, try and find a 2x8gb set of B-dies and you'll never have the issue again.

In fact, you could say the above about every generation of Ryzen chips. It's just that the 3000 and 5000 chips did in fact have better memory controllers, but you can search the forums for the endless list of "Can't run XMP with my AMD Ryzen Cpu".

[sl0011]

40 minutes ago, Guest 5150 said:

Zen+ and SK Hynix Kits (Corsair Vengeance) where a horrible pair. 

I've run through 5 sets of memory with my 2700X to find resolve to the issue.

 

Was easy, just don't use the SK hynix kits. All my samsung B-die kits where fine.

3733mhz stable and a 3966mhz top frequency because you can't uncouple the IF on Zen+. 

 

When enabling XMP on any kit, the system on auto already increases the SOC to 1.08v. 

You can safely use up to 1.200v for your XMP kit. More than that, you're aiming for gold competitively.

 

If you can't find stability with increased SOC, try and find a 2x8gb set of B-dies and you'll never have the issue again.

In fact, you could say the above about every generation of Ryzen chips. It's just that the 3000 and 5000 chips did in fact have better memory controllers, but you can search the forums for the endless list of "Can't run XMP with my AMD Ryzen Cpu".

Yea, I figured it could still be a RAM issue. Eventhough my current config is fine for the most part. Daily usage on stock seems fine. I'm using custom timings and voltages and tested stabity with MEMTEST86 all night long. Using XMP or running JDEC profile don't seem to change anything regarding CPU OC Stability.

 

[Guest]

12 minutes ago, sl0011 said:

Yea, I figured it could still be a RAM issue. Eventhough my current config is fine for the most part. Daily usage on stock seems fine. I'm using custom timings and voltages and tested stabity with MEMTEST86 all night long. Using XMP or running JDEC profile don't seem to change anything regarding CPU OC Stability.

 

 

Well you could try and disable "power down mode" in the timings menu. This should help prevent throttling when idle, I assume that's when the issues most occur? Otherwise in use with heavier loads just fine? 

[BuzzingBee]

56 minutes ago, sl0011 said:

Yea, I figured it could still be a RAM issue. Eventhough my current config is fine for the most part. Daily usage on stock seems fine. I'm using custom timings and voltages and tested stabity with MEMTEST86 all night long. Using XMP or running JDEC profile don't seem to change anything regarding CPU OC Stability.

 

You should use different ram stability test. I use Testmem5 with anta777 extreme config and OCCT with both SSE and AVX2.

[sl0011]

4 hours ago, Guest 5150 said:

Well you could try and disable "power down mode" in the timings menu. This should help prevent throttling when idle, I assume that's when the issues most occur? Otherwise in use with heavier loads just fine? 

Not really, it happens when stress testing with programs such as Prime95, Cinebench. With the old cheap ram, it used to happened on idle.

[sl0011]

4 hours ago, BuzzingBee said:

You should use different ram stability test. I use Testmem5 with anta777 extreme config and OCCT with both SSE and AVX2.

Really good to know that RAM can still be the issue. I wasn't expecting such a weak memory controller on Zen+. I will test stability of RAM with those programs.

[Guest]

Just now, sl0011 said:

Not really, it happens when stress testing with programs such as Prime95, Cinebench. With the old cheap ram, it used to happened on idle.

Who'd thought that you'd try another set of SK Hynix and wind up with completely opposite issued results?!

 

Crazy.

 

Well, you can try and increase SOC. But really if you have to pass 1.20v to get that full stability, the kit is no good for you. It's such an annoying issue. I've got 3 kits of SK hynix, Vengeance 3000mhz 4 sticks and a set of PNY which I didn't even bother putting in my Ryzen system, I overclocked it on my Intel system instead and it was pretty darn stable. But yeah, 2700X is pretty darn picky about memory kits. 

[sl0011]

3 minutes ago, Guest 5150 said:

Who'd thought that you'd try another set of SK Hynix and wind up with completely opposite issued results?!

 

Crazy.

 

Well, you can try and increase SOC. But really if you have to pass 1.20v to get that full stability, the kit is no good for you. It's such an annoying issue. I've got 3 kits of SK hynix, Vengeance 3000mhz 4 sticks and a set of PNY which I didn't even bother putting in my Ryzen system, I overclocked it on my Intel system instead and it was pretty darn stable. But yeah, 2700X is pretty darn picky about memory kits. 

Actually, the first module was micron I believe, and at the time of buying I had no clue about this topic.

[Guest]

2 minutes ago, sl0011 said:

Actually, the first module was micron I believe, and at the time of buying I had no clue about this topic.

Now Micron, is a hit or miss with Zen+. Most kits seemed to be ok.

 

Lest we forget before moving on that frequencies past 2933mhz is no guarantee either. I did notice my 2700X liked a looser tRC and tRFC timings, not sure why that is. I seen a lot of people run much tighter than mine claiming a lot of stability, but perhaps I just didn't have them dialed in well. 

 

Another thing as well, that learning along the way is keeping your memory cool during overclocking. I try and aim for BELOW 40c with the sticks in use during testing. Seems like 50c and up they start loosing stability, but this figure is going to vary widely from system to system depending on frequency and Vdimm used.

 

I dunno. There's quite a bit to memory overclocking and stability, the compatibility, the potential and not to leave out expectations either.... Just takes time and tweaking and trying a bunch of settings. But the issue you have at a load, I'd have to say... try another kit if you feel like you've tried all without positive results.

[sl0011]

7 hours ago, BuzzingBee said:

You should use different ram stability test. I use Testmem5 with anta777 extreme config and OCCT with both SSE and AVX2.

I tested for 1 hour on each SSE and AVX2. Seems ok, should I test for longer?

 

[BuzzingBee]

2 hours ago, sl0011 said:

I tested for 1 hour on each SSE and AVX2. Seems ok, should I test for longer?

 

So far so good, now it's better to try different stability test like Testmem5. 

 

Also do you happen to have WHEA_LOGGER errors on your eventviewer? 

[sl0011]

5 hours ago, BuzzingBee said:

So far so good, now it's better to try different stability test like Testmem5. 

 

Also do you happen to have WHEA_LOGGER errors on your eventviewer? 

It ran for a very short period of time on default config. IDK if that's fine. According to the user, anta777 extreme config is default. Yea, I think I have WHEA_LOGGER errors on my event viewer. I tried to use those logs in the pass without success, they don't tell me much. 

 

 

 

 

 

[Ebony Falcon]

Cba to read all comenets 

soc stabilises memory when ur pushing it 

[BuzzingBee]

3 hours ago, sl0011 said:

It ran for a very short period of time on default config. IDK if that's fine. According to the user, anta777 extreme config is default. Yea, I think I have WHEA_LOGGER errors on my event viewer. I tried to use those logs in the pass without success, they don't tell me much. 

Doesn't seem right, click on Load config &  exit and select Extreme1@anta777.cfg.

 

WHEA_LOGGER shouldn't be ignored unfortunately, likely your SoC voltage was too low. Have you set your SoC voltage to at least 1.1v? Although don't try to set too high yet, in my experience setting SoC voltage too high (e.g 1.2v) also create WHEA_LOGGER errors.