Problems with XMP G.SKILL 16GB (2x8) 3200MHz Ripjaws V Black CL16

[Cinco]

Hello.
So as title says, I've problem with XMP profile, but let me start with PC config:

---

Gigabyte Aorus B450 Elite
Ryzen 5 1600x @3.90
G.SKILL 16GB 3200MHz Ripjaws V Black CL16
GTX 1050 2GB(sic!)
Bitfenix Whisper M 650W
Windows 10 Pro

It runs completly fine without OC, but once i enable XMP, im getting many crashes in games, programs randomly shut down etc. I tried dropping MHz to 3000, 2933... but result was the same, default voltage was set to 1.35 with XMP on, I tried raising it all the way up to 1.4 but that didn't help either. Then after some time in google I found 'DRAM Calculator for Ryzen' and according to it's guide I got some numbers to put manualy in BIOS, sadly without effects. I tested memory on default timings (2133) with MemTest86 and after 5/5 passes it found no errors.
Since I'm clueless about OC I didn't want to do anything on my own as if im not sure what something do I rather not touch it.

About crashes with programs or games, I checked event viewer and most of the times it was 'Exception code: 0xc0000005' - not sure if that info can give you any ideas.

 

[EarthDog]

Try adding SOC voltage for the memory is where I would start.

 

[Cinco]

There's one problem, BIOS of this Aourus MOBO is disaster. I can't find that option, althought there are more options in Ryzen Master

 

[trents]

Does your Bios have a "Try Me!" option in the memory configuration section? If it does, you may find one or more 3200 mhz choices, one with more relaxed timings than the other. If you have that, try one or both of those options.

The other thing that I see is you are running a first gen Ryzen CPU on a second gen AM4 chipset. Nothing wrong with that but the combination of a second gen Ryzen CPU with a second gen AM4 board chipset may have fewer memory compatibility issues. The first gen Ryzen platform was very picky about RAM. To have a good chance of running 3200 mhz on the first gen platform people found they needed to pair it with the memory modules that were made with the Samsung "B" die chips. You could know which those were by selecting the 3200 mhz RAM products that were rated for a CL of 14. They were and are more expensive than the 3200 mhz products with more relaxed timings. When running a second generation CPU on a second generation chipset motherboard there are many more options when running high speed RAM than there were with first gen.

 

[Zerileous]

Dynamic VCORE SOC (second option in first SS). I would just start increasing it gradually until you get more stable. The recommended max is 1.3V, although above 1.15V is not recommended for 24/7 use.

Since you only have an offset available, you'll need to find a monitoring program such as HWiNFO64, which seems to be pretty accurate on these systems. It's important to note that software voltage monitoring is not particularly accurate, so allow some margin of error if you're particularly concerned about longevity.

You'll also need a stress test as well, for memory issues prime95 set to blend works well. You'll want to click blend, then custom and set 12,000MB for memory usage (I've use 16,000 before, but it's not necessary).

Write down the CPU temps (tdie should be correct), vcore and vsoc readings at idle and load. HWiNFO64 can be a little overwhelming if you're not used to having so much data, but take your time and familiarize yourself with the readings.

Once you know the baseline, you can start adjusting the offset to increase the voltage to a stable level. The calculator indicates 1.025V, but your CPU may require more. You can also try to play with vcore a little bit, 1.35V is the recommended 24/7 max.

Don't be surprised to see fairly high temps, it is a stress test and does not reflect normal workloads. That said, you don't want to see the CPU going over 95c during the tests. Are you using the stock cooler?

 

[Cinco]

Okay, so there's no such option allowing me to try something in BIOS, and either only XMP or Manual.
1600x wasn't shipped with stock cooler, but I've Cooler Master Hyper H412R.

CPU temps after OC to 3.90GHz on stock voltage which is 1.4 V (with jumps to 1.412) it seems to be pretty high isn't it?
There's SoC Voltage 0.856 V Max and 0.850 Min in HWiNFO64

Temps:
It's after 5 Tests done in 32 minutes with 0 errors and 0 warnings
CPU (Tctl) 96C (204.8F) Max - 53.8C (127.4F) Min.
CPU (Tdie) 76C (168.8F) Max - 33.8C (92.84F) Min.
I'm not sure which one should I be looking at.

Then I ran Blend for about 30 minutes but as above no warnings or errors (this was done on stock memory timings)

In BIOS Dynamic VCORE SOC I can only add voltages but I assume it'll change SoC Voltage that's sitting at 0.850 V right?
And Dynamic VCORE can be dropped only by -0.084 V everything higher than that results with system booting into BIOS with error, at the top of that I can't find any changes so I left it on Auto.

Should I incrase Dynamic VCORE SoC by +0.204 V and try with XMP?


EDIT: Incrased SOC by 0.006 and now it's sitting at 1.100V But shouldn't it be like 0.862?
EDIT2: With SoC Voltage at 1.1V and XMP on (1.35V) after about 15 minutes of Blend test in Prime95 with 12000MB it crashed with Bluescreen:
CPU Running on stock clocks


On Sat 4/27/2019 3:28:29 PM your computer crashed or a problem was reported
crash dump file: C:\Windows\Minidump\042719-6203-01.dmp
This was probably caused by the following module: tcpip.sys (0xFFFFF8061D98BACA)
Bugcheck code: 0xD1 (0x4E21, 0x2, 0x1, 0xFFFFF8061D98BACA)
Error: DRIVER_IRQL_NOT_LESS_OR_EQUAL
file path: C:\Windows\system32\drivers\tcpip.sys
product: Microsoft® Windows® Operating System
company: Microsoft Corporation
description: TCP/IP Driver



EDIT3 And another Blend test with 12000MB running for 45 minutes with no errors or warnings, what i changed was CaS Latency from 16 to 17

 

[wingman99]

If you had to change the CL 17, I would suspect memory trouble.

 

[Zerileous]

CPU OC: 1.4V is not immediately dangerous, especially if your cooling handles it well. It is however beyond the 1.35V recommended by AMD for 24/7 use, so you can expect some degradation in time. This would manifest as an inability to maintain stability at the same settings that were previously stable, either requiring you to increase voltage further (and further exacerbating the degradation) or decreasing the clock. I would try using a negative offset in the bios for DVID and see what your max stable clock around 1.35V is. It may be you only have to down clock by less than 100MHz to run at that voltage. One way to test stability for this is the prime95 small FFT setting, however use caution as it can cause very high temps. I wouldn't worry too much about it, and first focus on getting your memory stable, then you can come back to this if you're concerned about long term longevity. Is your minimum adjustment increment the -0.084V or just that it crashes after that? Since it's an offset to an automatic voltage, you just kind of have to set it and see what it actually does under load.

Temps: You should be looking at Tdie. Tctl is an offset used by AMD so that all of the motherboards can have the same stock fan policy with the stock coolers. You can safely ignore it. Your Tdie of 76C under load is excellent for running prime95 blend. You can expect to see it climb a bit higher under small FFT.

VCORE SOC: 1.1V is fine for the VCORE SOC. For testing/bench marking, I have seen it advised to go as high as 1.4V, however my reference indicates that maximum should be 1.3V. Again that is just for testing/bench marking, but increasing more could help you diagnose the instability as either being SOC related or DRAM related. For long term 24/7 use, it should be 1.15V or less. Right now I would leave it at 1.1V as you got a stable run from that.

Memory timing: Can you please clarify, in Edit3 the memory is still at 3200MHz? I will assume this is the case for my reply. I would generally agree with wingman99 there, however, first gen Ryzen's can be quite picky about timings. If you're concerned about it you can try increasing the DRAM voltage to 1.4V and decreasing the timings to 16 or even 15. However real world it's not going to make a significant difference. Also G.Skill would be likely to simply tell you that every CPU is different and that's the silicon lottery, especially since the Ripjaws V is "optimized for Intel." You can always contact them and see what they as if it bothers you. If it was me however, I would just be happy to run 3200MHz stable.

In summary, I wanted to provide a lot of details / options for the sake of knowledge and depending on what you want to get out of the system. Assuming this is a daily driver OC for gaming / general computing etc, my advice is: 1. Leave SOC voltage and memory unchanged. Do one more stress test of prime95 blend with 12,000MB memory for 1 hour to confirm stability. Then I would use prime95 small FFT to test CPU stability, aiming for the highest clock possible at 1.35V VCORE. A little over it should be fine. Alternatively, if you're not super concerned about using this system for 2+ years at this overclock, leave it at 1.4V and see how it goes.

 

[Cinco]

Hello and thank you for your reply!
I'm not using this PC for any kind of work, just for games and general use.
Yes after changing VCORE SOC to 1.1V and changing CL to 17 so far I can say it's stable while running @3200MHz, also I've done Prime95 Blend test for 1hour and it returned no errors or warnings.
I'm more concerned about CPU tho. This BIOS is either ignoring my settings or HWiNFO64 is lying to me, VCORE reduction gives almost no results it's running all around 1.294V to 1.497V with MAX of 1.512 V.

Also, in HWiNFO to check CPU Voltage where should I look at? Until now I've been taking a look at "CPU Core Voltage (SVI2 TFN)"

EDIT: I tried changing values under Dynamic VCORE and when it was on Auto then it's been running at 1.4V (max) when i swapped to 'Normal' it dropped to 1.375 and when I manually changed to - values like -0.0016... It was upping voltage by that number, and another weird thing, using + voltages like +0.0012 V (or more) then it lowered maximum Voltage to 1.375 V however any further changes like +0.0016 V or 20 did not changed anything but it's still running at 1.375 V same result with +0.000 it stays at 1.375 V no matter what. I'm confused, should I leave it as it is or play with it more? Prime95 during FFT small test @3.90GHz was going all the way up to 95C (Tdie), then i dropped to 3.80GHz and it's not going over 87C. I think it's maximum I can squeeze with current cooling for now.

 

[Zerileous]

Yeah we need to get that voltage setting figured out. You can look at the motherbaord section where it says vcore and see what that reads. You can also look at the VR loop sections. VR VOUT will match vcore for one and SOC voltage for the other. Sounds like some kind of bug for the vcore. You really don't want it over 1.45V. You might try flashing to the latest bios, if it's not already. I think you'd be pretty safe at 1.375V to be honest. Maybe a little high but better than where you were before.

It's the voltage that's giving you those high temps. What is the vcore when set to 3.8GHz?

Also running at 3200MHz is actually overclocking part of the CPU, not just the memory, so it will increase the temps too. It's called infinity fabric, if you want to do more research. This is part of the benefits of the memory OC.

 

[Cinco]

That's how it looks like. I'm out of ideas, changing Dynamic Vcore in BIOS doesn't do anything but changes what's mentioned in my previous comment. Maybe I should return mobo and ask for refund, as it seems to be broken?

When I changed voltage in Ryzen Master then it was showing 1.35 as set in RM, but is using software for such things is safe, and also it doesn't apply manually every restart.

I updated BIOS to F31 recently

 

[Zerileous]

Depending on where you purchased it, you may be able to return it for those reasons. Other users have had good luck using Vcore under the motherboard section, in this case labeled "GIGABYTE B450 AORUS ELITE." For my motherboard they are all within a tenth of each other. According to that reading, you're in a good place right now @ 1.356V. You might want to check what it reads at load though, to make sure it's similar. It's also important to remember that these software monitors are not especially accurate, and inaccuracies of more than a 10th of a volt are common. In truth we're not pushing out systems hard enough for that to matter, however extreme OC people use hardware monitors for this reason.

You could try to corroborate the voltage read in Ryzen Master with the readings in HWiNFO64, to see which sensor matches it most closely. Also there is nothing wrong with using Ryzen Master for your overclock. Of course we prefer BIOS overclocks, but Ryzen Master is fine if it works better for your board. Only downside is having to re-enable the profile each reboot.

I wouldn't worry too much about it if you're stable and everything is running well. You could even try for 3.9GHz now that you got the voltage figured out.

 

[Cinco]

I made a ticket on Gigabyte website, now I'll have to wait 3-7 days for response.
So I did 1h Prime95 test at @3.80GHz (3.90 is going over 90C) with stable 85.4C. Under load HWiNFO average CPU Core Voltage was 1.369 V, Vcore similar to that 1.367 V. I checked with CPU-Z it's been jumping around 1.356 to 1.380 V, Ryzen Master tells me 1.38125 V. And shortly Prime95 ended I experienced BSOD:

On Sun 4/28/2019 2:28:30 PM your computer crashed or a problem was reported
crash dump file: C:\Windows\Minidump\042819-6531-01.dmp
This was probably caused by the following module: ntoskrnl.exe (nt+0x1B3730)
Bugcheck code: 0x4E (0x2, 0x346028, 0x3FFFFF, 0x8)
Error: PFN_LIST_CORRUPT
file path: C:\Windows\system32\ntoskrnl.exe
product: Microsoft® Windows® Operating System
company: Microsoft Corporation
description: NT Kernel & System
Bug check description: This indicates that the page frame number (PFN) list is corrupted.

 

[trents]

What are you cooling the CPU with? Are you using the stock cooler that the CPU came boxed with? What is the make and model of your case and what about case ventilation? Your CPU temps seem on the high side. Please run a stress test and report back with another pic of HWinfo64 showing the section were VR MOS is located.

I think you may be able to get away with less vcore at 3.8 ghz. Try lowering that.

Can you pass the stress tests if the all bios settings are on default except the memory set to XMP 2?

 

[Cinco]

Cooler Master HYPER H412R. Case "SilentiumPC Regnum RG1W" - some cheap one that I used for previous rig, I'm currently upgrading old PC and for now it must be as it is.

Thing is whatever I do it's impossible to modify vcore, only option I have is vcore offset which gives no results whatever it's changed to.
Here are some screenshots from HWiNFO after Dynamic VCORE changes I made.

Vcore -0.084


Vcore -0.096


Vcore -0.120


Vcore set to AUTO - No changes made, just default settings in BIOS to see if something changed*


*I removed CMOS battery for 10 minutes to reset BIOS thinking it could help, but it did not.

Every screenshot was taken right before I stopped Prime95 test.

I googled some and found a guy with same problem as mine, he's been told to change some Pstate but I'm not sure where and what to change.

 

[wingman99]

Looks like you were having the desired effect, the Vcore was lowering with negative dynamic core voltage.

 

[Cinco]

Hmm, so I think reseting BIOS by removing CMOS battery helped?

-0.052


Here's result from Small FFT, after OCing to 3.80GHz and lowering offset to -0.052V, that's lowest I can go, -0.056V is booting to BIOS with error. Is that safe to run for 24/7 usage or should I downclock?
@3.80GHz Ryzen master always show 1.41875GHz, default @3.60 it's jumping around 1.2 to 1.38 and results as you can see comment before.

 

[trents]

I would say that is safe temps for sure and probably safe vcore for 24/7 as well. AMD recommends not going any higher than 1.35 for vcore for 24/7 to avoid CPU degradation but you are so close to that I wouldn't worry about it. Plenty of folks are running 1.4 or better vcore 24/7 to squeeze every last ounce out of their overclock and I have not heard of any chip failures from this as of yet.

Your CPU (Tctrl) is quite high at 99c but my understanding is that is positively offset temp adjustment to introduce fan speed consistency across the Ryzen product line. I think CPU (Tdie) is the actual reading and your are safe there at under 80c. Also, the CPU+SoC Power seems too high at 170.343c. Not sure how accurate that is. For a six core/12 thread Ryzen I would have thought it would be lower at your vcore.

 

[Cinco]

Allright, so when I'm looking at voltage, should I look at Vcore or CPU Core Voltage (SVI2 TFN)? Because those values confuse me.

 

[Zerileous]

Assume you're using prime95 to load during these shots?

Use "Vcore" under the section "GIGABYTE B450..." not the CPU Core Voltage under (SVI2 TFN), as that does not appear to be accurate on this board. If you look in all you're previous pics, the "Vcore" did seem to change with each step, just not the "CPU Core Voltage."

If your last pic was taken running prime95 and it will run small FFT for 1 hr, then you're in great shape and should be good to go for the 24/7 OC. What trents said about the temp is correct, tdie is the actual temp. tctl is offset so the same default fan curve that works for a 2200G can be used on a Threadripper. Not really sure why, since they're different chipsets, but that's what they did.