[SOLVED] Importance of Samsung B-Die for Ryzen

[Sanegency]

I've had my new PC up and running for a week or two now, and I've been encountering consistent hard freezes during use, entirely out-of-game, usually on Youtube. My system crashed about once a day at 3200MHz, once in six days at 3000, once a day at 2666 and once a day at 2133 under equivalent load. I'm led to believe (and I was told by a tech help line) that the problem is my RAM, which isn't listed on the QVL for my motherboard (ASUS ROG Strix b450-f Gaming). I have to buy a new RAM set in order to make the system work properly.

The problem is that while I'm trying to select a new set from the QVL to order in, few of the sets listed are the same 16GB size as the set i have now, even fewer are equipped with Samsung B-Die which is supposed to help with stability for Ryzen (I run an R5 2600X at stock speed), and the few that do are mostly in the $250-300 range which is far more than I'm able to spend on RAM right now.

The only way I'm realistically going to be able to order a new set of RAM in the 3000/3200MHz range is if I go with an offbrand kit or a kit with no chip or with a HYNIX chipset (either one of which I can get for about $115), which I've heard very mixed reports on. How important is it that I have BDie on my RAM?

EDIT: As I was running my last quality checks before waiting for a reply, I realized that I'd never enabled XMP in my BIOS, the reason for that being that ASUS decided to call it DOCP instead and I didn't think it was the same thing. I have it enabled now, and at DOCP 3200 my timings changed from about 16:27:27:54 to 16:18:18:36. I'm not exactly hopeful that that will fix the issue. If you don't see another edit below this, assume I'm still experiencing crashes, and that I still need to buy RAM.

 

[CountMike]

On 2nd gen Ryzen and on 400 series MBs, pretty well all RAM would work at at least their rated XMP. Samsung b-die usually brings best overclock with slower memory but not much OC with 3200MHz or faster.
More important part is to have single rank memory and have it in right slots, usually A2 and B2. Also make sure you have latest BIOS installed, new ones came out just few days ago and new AGESA code helps with RAM stability.

 

[Sanegency]

Update: the system continues to freeze randomly even with DOCP enabled. Though some games such a Battlefield V have stuttered, I haven't seen any freezes in-game so far. All but two of my dozen-plus freezes have occurred on YouTube, in which case the video's audio continues to play while the screen freezes and the mouse and keyboard stop responding. The other two crashes occurred while attempting to close Battlefield V and while attempting to run Minecraft in a half-screen window alongside a YouTube video, both with similar symptoms.

The Windows memory diagnostic tool returned no errors with my ram, and the windows event log shows no errors or failures besides the fairly benign Event 10016 before or after any of my freezes. At this point I'm seriously unsure if my ram or my motherboard is causing the issue here.

 

[gskill support]

Have you tried one at a time to see if one performs differently than the other?

 

[Karadjgne]

QVL is qualified Vendor list, not qualified ram list. There's only a few actual ram OEMs, like Samsung, SkHynix, Micron etc. They manufacture all the ram for every vendor. The serial number on the ram covers a bunch of different scenarios. F4-3000C16D-16GISB for instance is G-skill Aegis, F4 (ddr4)-3000MHz-Cas16-D(?) 16Gb kit-I(?) S(single rank) - B(black). You can have the exact same ram F4-3000C16D-16GISR except its red. Just in g-skill Trident-Z 3200 alone is 5 different models, 8 colors, +rgb, in 4 different kits, 3 different Cas. If you add it all up, there's over 200 different individual serial numbers just in the Trident-Z alone. Then add the value, Aegis, sniper, RipJaws IV, RipJaws V etc and g-skill alone has thousands of different serial numbers. Multiply that by all the different speeds and sizes. Then add Adata, Patriot, Crucial, Kingston and all the rest and you'd have a little old man spending thousands of man hours to make a QVL thousands of pages long, testing the same 10 ram over and over, since they all come from the same 10 places, using the same 10 OEM IC chips.

Kinda absolutely pointless. So mobo vendors will grab a few sticks from a few different kits in each size, maybe a few donated prototypes and test them. That's all.

In almost 40 years of dropping ram into pc's, I don't think I've ever once had a single stick that was actually exactly QVL. Same speed, same vendor, but it was red, or blue or Cas10 and the QVL had black Cas 9.

Don't rely on QVL as gospel, you'll be disappointed. It's just a tool really that says that a variety of different sticks and kits will work as promised. If your exact number isn't on it, it really is, you just have a different color. Different speed. Different kit. But it's all the same ram, just under a different named heatsink from a different vendor.

DOCP is an AMD thing, XMP is Intel eXtreme Memory Profile. It amounts to the same thing.

Screen freezes, almost always have something to do with the gpu and/or drivers. Can also be caused by out of date motherboard chipset drivers for USB, Sata, audio and Lan. So make sure a) latest bios, b) any and all mobo chipset drivers are current (get from same place as bios) c) clean install of gpu drivers d) it's a new build, with lots of temp stuff in, erased, deleted, left behind etc, so use something like ccleaner and it's reg tool (use defaults) to run through and clean out all the crap that can cause conflicts with dead end addressing and orphans etc.

 

[Sanegency]

Have you tried one at a time to see if one performs differently than the other?

What do you mean by this?

 

[Sanegency]

Screen freezes, almost always have something to do with the gpu and/or drivers. Can also be caused by out of date motherboard chipset drivers for USB, Sata, audio and Lan. So make sure a) latest bios, b) any and all mobo chipset drivers are current (get from same place as bios) c) clean install of gpu drivers d) it's a new build, with lots of temp stuff in, erased, deleted, left behind etc, so use something like ccleaner and it's reg tool (use defaults) to run through and clean out all the crap that can cause conflicts with dead end addressing and orphans etc.

I reinstalled my gpu and chipset drivers. Everything else is up to date. I got ccleaner and ran the default space cleaning and registry cleaning tools. I'll let you know if the problem persists. Almost everyone I've talked to irl about this has recommended changing ram.

 

[Karadjgne]

Get memtest64 or memtest86+ and run it overnight. The longer the better. If there's issues with the ram more in depth than what the WMDT can find, memtest will find it.

 

[Sanegency]

Get memtest64 or memtest86+ and run it overnight. The longer the better. If there's issues with the ram more in depth than what the WMDT can find, memtest will find it.

I've just finished doing this. MemTest64 ran for about 9.5 hours with 0 errors.

 

[gskill support]

With more than one module you can test one module at a time to see if all modules have consistent results. If not, it can give some insight on what may be going on.

 

[Karadjgne]

Ooh, actual g-skill tech! Cool! Nice to see you.

That said, with 0 errors after over 9hrs looping, I'm doubting it's the ram, I'd be more inclined to believe it's a defective gpu or possibly overheating vram on the gpu causing issues. Have you tried underclocking the gpu and running the game yet?

To hopefully cure that event 10016.

1. Type or paste ‘regedit’ into the Search Windows box. Hit file and save a copy of the registry.
2. Navigate to ‘HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Ole’.
3. Delete these four keys if they appear, DefaultAccessPermission, DefaultLaunchPermission, MachineAccessRestriction, MachineLaunchRestriction.
4. Reboot your computer.

 

[Sanegency]

That said, with 0 errors after over 9hrs looping, I'm doubting it's the ram, I'd be more inclined to believe it's a defective gpu or possibly overheating vram on the gpu causing issues. Have you tried underclocking the gpu and running the game yet?

I really doubt it's overheating vram, I've been monitoring the temps and it's way in the safe zone. As for defective gpu, I had a benchmark tell me that my gpu is underperforming but that same benchmark told me my gpu benched at 105% expected performace, and I figure it can't be all that defective if I can still get the high framerates I do. As for underclocking it, I don't have any problems with the card other than a stutter in one specific game (BFV) that sometimes happens and sometimes doesn't.

The PassMark 3DMark 9.0 GPU test shows performance in the 95th percentile for RTX 2060s even after being penalized score for not having a 1440p monitor

 

[Karadjgne]

Unless you have one of a very few cards, I doubt that you can monitor vram/vrm temps. Simple reason being there's no thermal sensor at all in that area. The only thermal sensor is inside the gpu itself, so any temp reading is from the cpu on the video card. It's not uncommon for vrms/vram to well exceed 90°C and the gpu be in the 60's. All depends on usage, cooling, how well the thermal adhesive was applied, if it was built at 4:59pm on a Friday etc.

Everything points at the gpu in one way or another, be it freezes, stuttering etc. That could be the gpu itself, it could be driver conflicts or errors, could be storage or even the psu. Could even be a bad solder on the pcie x16 socket itself and when the fans spin up it breaks loose. Games locking up on exit is almost always a driver conflict with the gpu drivers, and event 10016 usually happens with unclean windows 10, especially on 10 that got upgraded from 8. It's pretty much where the game is telling the game server its logging off, and when the server replies ok, your pc has no idea who the server is talking to and denies permission to turn stuff on/off. Which makes the gpu drivers loop and eventually freeze. (eventually to a pc is usually 3-4 seconds). Maybe the above fix for 10016 will cure just about everything, or maybe not, I dunno, but until you can rule out stuff, everything is suspect.

It may take a full, clean install of windows to fix, hopefully not, but never know, it's cured many such issues.

Passmark is good for ability tests, but not so good for torture tests. For that I'd use Unigine Heaven. Even msi kombuster has a series of tests that'll hurt a gpu, and not with just the fuzzy donut.

 

[Sanegency]

Games locking up on exit is almost always a driver conflict with the gpu drivers, and event 10016 usually happens with unclean windows 10, especially on 10 that got upgraded from 8.

Passmark is good for ability tests, but not so good for torture tests. For that I'd use Unigine Heaven. Even msi kombuster has a series of tests that'll hurt a gpu, and not with just the fuzzy donut.

Hm. Well I'll keep the gpu thing in mind. I've only encountered a freeze on exit once.
As for Windows, I set up this computer with an unactivated copy and only activated after about a week and a half of consistent use.
As for Heaven, I've used it before. I only ran it for about 45 minutes due to having to leave home that day, but it ran entirely solid with no stutters, frames never dropping below 70 on the highest graphics setting, and CPU temps stable in the safe zone. As I've said before, it's almost entirely web browsing that is causing my freezes, and not games except for my sole freeze on exit and my sole freeze in window mode. I turned off hardware acceleration in my browser just to make sure.

 

[Sanegency]

To hopefully cure that event 10016.

1. Type or paste ‘regedit’ into the Search Windows box. Hit file and save a copy of the registry.
2. Navigate to ‘HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Ole’.
3. Delete these four keys if they appear, DefaultAccessPermission, DefaultLaunchPermission, MachineAccessRestriction, MachineLaunchRestriction.
4. Reboot your computer.

None of these four keys appeared under the path.

 

[Sanegency]

Update to include all the information I have.

The crashing program seems to have stopped. The only three things I did that could possibly have solved the issue are reinstalling my gpu driver, reinstalling my chipset driver, and turning off hardware acceleration in my browsers. Since all but one of my crashes were on or related to YouTube, I would have to guess it was hardware acceleration that was causing the issue. I don't expect to see the problem again, so I'll mark the thread solved. Thanks for your help.