OC Ryzen 5 1400 3.7Ghz

[Tiamat]

Hello guys! I recently got a Ryzen 5 1400 and i've been trying some OC with it. Can u help me?

So, i'm not trying to break any records, but i'm trying to see how much can i OC ryzen SAFELY, for a long time (years) OC. I'm making some tests and trying to understand the behavior of my Gigabyte GA-B350-DS2 and the Ryzen 5 1400.

In the attach i'm showing 4 tests i've made. The first one with the stock settings after 15 min of prime95 stress test. The 2th with a pre-made OC from the Gigabyte motherboard at 3.4 Ghz after 15 min of stress test. 3th with a 3.6Ghz OC after 15 min of prime95 stress test. And the last one is a 3.7Ghz test after 1 hour stress test.

As far as i know (and that is not much), cpu temperature (~55ºC), Vcore (~1.248V), chipset temp (55ºC) are pretty good, but i'm worried with the VRM MOS temperature, that was around 80ºC even with stock settings (reaching 84ºC in OC). I don't know if i should take care of other sensors (CPU core power maybe?). So i would like your experts opinion

Beyond that i tried 3.8ghz, but the system stopped working as soon as opened Prime95. I didn't try any change in the Dynamic Vcore (DVID) or Dynamic Vcore SOC (DVID) settings. What should i do?
I'm using easytune from Gigabyte for this first tries.

Thank you

 

[GoldenX]

Up to 1,35v should be good for a long life CPU overclock. I would avoid using the auto overclock of the motherboard.
Mmm, I think 84º on the VRMs is good, bout could be improved with a direct cooler.
Only other relevant voltage is SoC, it helps with RAM overclocking past 2666MHz. I would not use over 1,15v for a 24/7 oc.

 

[Tiamat]

. I would avoid using the auto overclock of the motherboard.

Hello @GoldenX ! Ty. To avoid use the auto settings, i just need to set Dynamic Vcore (DVID) = 0?

RAM overclocking past 2666MHz

I don't intend o OC ram, cause i just have some simple generic Kingston memory... no heat dissipation and other things...

I would not use over 1,15v for a 24/7 oc.

Do u mean 1,15V over the Dynamic Vcore (DVID) or the SoC voltage?

 

[HD64G]

 

[Tiamat]

Hey @HD64G ! Man, i was worried when i saw the tittle of that vid. I just thought i did something really bad haha.

So, let me try to understand. Do u mean i shouldn't mess up with DVID and DVID SoC at all or just be carefull. I 'm really new in these OC things, and i'm trying to be as carefull as i can, so ANY info matters. I did some tests now and could get the Ryzen 5 1400 pretty stable at 3.8 Ghz and 1.34V Vcore. But, as i can assume from the video, i shouldnt trust that info and just avoid getting closer to that? If so, that 3.7ghz that i did with some stock voltages (as far as i know) are safe?

Please, i'm really trying to learn with u guys!

 

[GoldenX]

"Dynamic" Vcore. Sounds like auto tuning from the motherboard.
I meant SoC, 1,25v can degrade the controller according to AMD, and more than 1,0v seems necessary for going over 3200MHz on RAM.

If it is your first overclock, don't touch any voltage, only set the CPU multiplier, and set RAM to their stock frequencies and timings. I recommend overclocking from the BIOS, if not, with Ryzen Master, I always had bad results with the motherboard's software. Personally I run stock most of the time and use 1,25v-3800MHz (constantly, when my 1200's stock 3100MHz is not enough) or 1,45v-4000MHz (when really needed).

According to AMD on vCore, 1,35v is the max recommended for a 24/7 overclock (that means full frequency and voltage, all day, all week), and 1,45v can degrade the lifespan of the CPU (maybe one or two years from the 10 it has?). So I would stick to 1,35v if temp. allows, maybe 1,4v.

Regarding RAM, mine are also normal 2400 ones, and they can reach 3200MHz with loose timings. A heatsink is not needed on them for a simple overclock, they are more of a marketing lie. Heatsinks are not needed in normal RAM since DDR2. RAM frequency is very important on Ryzen, so after learning the limits of the CPU freq., take some time with RAM tunning. Also, I would have preferred two sticks for dual channel, either 4GB or 8GB, but I can see RAM prices and having two slots being a problem.

 

[Tiamat]

. I recommend overclocking from the BIOS, if not, with Ryzen Master,

I tested both now... BIOS settings were usable to reach the 3.8 Ghz with 1.34V Vcore. But, after @HD64G vid, i decided to dont get too close of that 1.35V limit. So i took a step back and used the Ryzen Master. Now i'm running at 3.7Ghz (37x mult) and 1.25V locked in Ryzen master config... I guess 1.25V is the stock Voltage isn't? CPUid, actually shows vcore around 1.236 and 1.248 volts under load. Besides that, i'm using the balanced ryzen energy plan...

Do the stock setting would be any safer than these 3.7Ghz settings considering both are at the same vcore? I'll attache a fast comparison of both later, so u can judge it.

@GoldenX What are your stock voltages showed on CPUID (ouHWGmonitor)?

 

[GoldenX]

The moment you set something in Ryzen Master, EasyTune or in BIOS, the motherboard disables the stock voltage variations.
Stock is way lower on normal clocks and way higher on XFR clocks, for example in my case with the 1200 at stock, its 1,05v for the 3100MHz under load, and somewhere between 1,4v and 1,5v for the 3450MHz XFR, since it's not used on long runs, it's completely safe.
Test stability of that 3700MHz overclock, maybe you can try for 3800, or a lower voltage for 3700.

 

[Tiamat]

The moment you set something in Ryzen Master, EasyTune or in BIOS, the motherboard disables the stock voltage variations.
Stock is way lower on normal clocks and way higher on XFR clocks, for example in my case with the 1200 at stock, its 1,05v for the 3100MHz under load, and somewhere between 1,4v and 1,5v for the 3450MHz XFR, since it's not used on long runs, it's completely safe.
Test stability of that 3700MHz overclock, maybe you can try for 3800, or a lower voltage for 3700.

Ty @GoldenX I'll do that. Do u know if undervoltage could cause any cpu damage (even without causing system crashs ---- only temporary stops for example)

 

[HD64G]

My not huge experience leads me to not ever trust motherboard bios auto voltage control and set the lowest possible for the frequency I target to have the best combo of efficiency-performance-temps-durability of my pc parts. So, I would go to the frequency I want and then try to find the lowest possible Vcore. If that Vcore was over or very close to the limit I would prefer to lower the frequency 5% to keep it safe. My 5 cents.

And undervoltage is always good and helps efficiency and temps when you find a stable setting.

 

[Deleted member 67555]

Just to let you know 3.8Ghz is good enough of an overclock on these... It took 1.55v on the vcore to hit 4.0Ghz but it only took 1.35v to get 3.8Ghz.
At least on my R5 1400 temps stay good up to 1.375v on the vcore.

 

[GoldenX]

No, undervoltage doesn't cause damage. But to be considered an undervoltage you would have to go bellow 1v on vCore.

 

[Tiamat]

z but it only took 1.35v to get 3.8Ghz.

@jmcslob Actually, i was abble to do the same with 1.34V on Vcore... but according to @HD64G vid, those sensor values can be much higher in reallity... I've checked those reviews affirming that 1.35V is safe for a long term OC on ryzen 5, but none of them mention if 1.35 would be the value showed on software (like cpuz for example), or the real voltage passing through the system. So, considering those sensor values could be underestimated, an 1.35V could actually mean something around 1.4V+... very near to the 1.45V limit. As i could get 3.7Ghz whitout forcing the vcore voltage (actually the max values are lower than stock settings), i guess 3.7Ghz would me much safer for a small 100hz sacrifice compared to 3.8ghz. What do u think?

@GoldenX I've made that comparison i told u earlier. As u can see in the attach, Vcore average was a little bit smaller in stock settings, but Vcore max was signiffcantly higher when compared with 3.7Ghz OC. That certainly corresponds to what u said:

Stock is way lower on normal clocks and way higher on XFR clocks

But, wich config would imply in less stress for the cpu in a long term system? I mean, comparing those data, Vcore on OC never passed 1.25V (at least at the sensor), CPU and VRM MOS temp were only 5ºC and 3ºC higher than stock, respectively -- at the cost of higher Hyper TX3 rpm cooling), while SOC and DRAM voltage remained the same, and max CPU core voltage and VSOC MOS temp were lower than stock.

 

[GoldenX]

The only way to be sure of the sensor data is to measure it manually. The sensor is supposed to report the right value.
Only the voltage is responsible for degradation (on the long run), so it doesn't matter if it's 3,8 or 3,7, vCore is what you should be looking at.

Also, don't look at averages in voltage, look at the present value when stressing all cores, and when stressing a single one, then you can see the right values. The CPU-Z bench or Cinebench can help with that, those programs test both all cores o a single one.
Anyway it looks like your CPU uses 1,25v at stock. A tad high, but that would explain why it needs so much vCore for reaching 4GHz.

 

[Tiamat]

So, here i am again

I did some more tests and faced some small problems. I checked on some foruns, and most topics are about reaching 1.35V-1.45V wich is something i'm trying to avoid for now. So i decided to keep posting here as i try to get the 3.7Ghz stable OC.

I did some tests trying to reduce voltage at 3.7Ghz and i was not successful. So i just kept the 1.25V and tried a more long stress test with prime95. My tests were OK, but after more than 2:30 hours of test, system freezed. I dont know if i could consider that stable, since some reviewers says that 1 hour should be more than enough... What do u think?

My next step will be increase voltage till 1.275 and check stability... I'm trying to reach at least 4 hours of prime95 test.

 

[HD64G]

Keep it at 1.25V and use it for whatever you want it. If it frozes again, give it the smallest bump possible in BIOS. Check stability and you should be ok.

 

[Tiamat]

@HD64G is there any risk that this small instability could slowly damage the cpu?

 

[infrared]

Instability will only cause crashes and potential data corruption, it won't do any damage. Also IMO you're leaving a huuuuge safety margin... I've spent a few hours benchmarking at 1.525v, I regularly run 1.44v while gaming, and 1.34v at full load the rest of the time.. I've had that chip since June '17 and I've not noticed any degredation at all.

The big deal about that video is specifically the SoC voltage, which is much more sensitive to overvoltage. Since you're on a Gigabyte board it might be worth setting that low to 1.0v and see if you're still stable (if it's not stable that means it's probably setting/reading the voltage fairly close). Judging by your 3.8ghz clock at 1.25v I'd say vcore is in the ball park, and if it is overvolting, it's not likely to be more than 0.05v so you still have at least 0.1v of safe (long term) headroom.

tl;dr You could push that cpu a lot harder and it'll still be obsolete before it dies.

 

[Tiamat]

Ty @infrared !
Nice explanation =D. Really helped.

But , when u said:

Judging by your 3.8ghz clock at 1.25v I'd say vcore is in the ball park, and if it is overvolting, it's not likely to be more than 0.05v so you still have at least 0.1v of safe (long term) headroom.

I was in doubt, cause i got an almost stable 3.7Ghz at 1.25V.
3.8Ghz was only stable after i got at 1.33 V- 134V... and i didnt run a long stress test...

So i didnt understant if you were referring to the 1.25V or the 1.35V. If the overvoltage is around 0.05V as u said, so 1.25V would actually means 1.3V, only 0.05V under the safe 1.35V limit. On the other side, 1.35V would mean 1.4V in true, 0.05V above the 1.35V limit...

The big deal about that video is specifically the SoC voltage

About SoC voltage, i'm a totally analphabet. I understood that 1.2V would be the limit, but i should stay under 1V because of that difference between sensor and real voltage, right? Beyond that, i just know that SoC voltage could help RAM overclock (as @GoldenX said), wich i dont know a lot either...

Thank you for your help guys!

 

[infrared]

3.8Ghz was only stable after i got at 1.33 V- 134V... and i didnt run a long stress test...

So i didnt understant if you were referring to the 1.25V or the 1.35V. If the overvoltage is around 0.05V as u said, so 1.25V would actually means 1.3V, only 0.05V under the safe 1.35V limit. On the other side, 1.35V would mean 1.4V in true, 0.05V above the 1.35V limit...

Ah okay, sorry I misunderstood your post, I thought you were stable at 3.8-1.25v.

Check out this voltage/frequency curve for ryzen.. your voltages line up pretty well with 1.25v and 1.34v having roughly a 1ghz jump, so your vcore is probably pretty close to what you set/read.

(don't worry about frequencies matching, the shape of the curve will apply to all ryzen's, just at different frequencies depending on how lucky you were)

About SoC voltage, i'm a totally analphabet. I understood that 1.2V would be the limit, but i should stay under 1V because of that difference between sensor and real voltage, right?

Exactly If there's a potential inaccuracy you'd want to err on the safe side like you're doing, unless you verify the voltage with a multimeter. It might not be as far out as gamers nexus' particular board. 200mV is a huge discrepancy!

 

[NdMk2o1o]

Thats a pretty accurate graph though for 4ghz I think it should go up slightly steeper, I have a 1600 that hits 3.9ghz with 1.335v though it took 1.45v to hit 4ghz that was cinebench stable (it was bootable into windows just wouldnt finish a cinebench run at anything below 1.45) I'm not clocking my ram as I have 2800mhz and they run XMP with no issues since raven ridge BIOS update.

OP why are you so hesitant to go above 1.25 vcore? AMD recommends anything up to 1.35 is a safe value so I think you'll hit 3.8 very easily with about 1.3v and that won't cause any degradation to your CPU either.

 

[GoldenX]

I think the same, but efficiency goes to hell if you can do 3700 with 1,25v, but for 3800 you need +0,1v.
Set profiles on Ryzen Master, one for 1,25v, one for 1,35v, and one for all out 4GHz at whatever voltage it needs. That's what I do.

 

[NdMk2o1o]

I think the same, but efficiency goes to hell if you can do 3700 with 1,25v, but for 3800 you need +0,1v.
Set profiles on Ryzen Master, one for 1,25v, one for 1,35v, and one for all out 4GHz at whatever voltage it needs. That's what I do.

vCore shouldnt be constant anyway without any load, mine regularly hovers arounbd 0.6-0.7 under light use and browsing etc, albeit the clocks remain the same, my only gripe with overclocking Ryzens really, would be nice to have the CPU throttle down when idle, I know you can with P-States and I almost got the hang of using them once but not quite and havent ventured back into using them since

 

[GoldenX]

If you leave settings at stock in bios (except SoC or DRAM voltage if you need) and overclock with Ryzen Master, you get the overclock you set on it under load, and the frequency and voltage will throttle when idle.

I think my english is failing here, please mark me any errors.

 

[NdMk2o1o]

If you leave settings at stock in bios (except SoC or DRAM voltage if you need) and overclock with Ryzen Master, you get the overclock you set on it under load, and the frequency and voltage will throttle when idle.

I think my english is failing here, please mark me any errors.

Nope, no problems with your English I know Ryzen Master works this way, I just prefer bios overclocking, not sure why they couldn't implement C-States when overclocking with a bios update tbh