New Ryzen Build Crashing

S

Sovereign

2[H]4U
Joined
Mar 21, 2005
Messages
3,098
RMA-ing the RAM; it BSOD in any slot at stock speeds. It's either the RAM or the motherboard (or possibly both).

I'm sure these are a dime-a-dozen, but I'll try to explain.

So my attempted overclocks weren't stable--that's fine. Nothing's guaranteed beyond stock clocks, after all.

What's bugging me is the following:
- As soon as I get any BSOD for any reason, even resetting to stock won't "save" it--near-instant BSOD on every Windows boot and I have to reinstall the OS. Seems fine in Safe Mode though so I'm really suspecting a driver problem.
- memtest86+ (the v7.x w/UEFI GUI) passed 6+ hours at stock/auto settings
- Prime95 blend @ stock set to use 31500MB RAM near instant-crashes...at stock
- Prime95 torture (that doesn't use much RAM) passed 8 hours with the CPU running at 3.7GHz, 1.35v
- Most BSOD are MEMORY_MANAGEMENT, followed by IRQL_NOT_LESS_OR_EQUAL and DCP_WATCHDOG
- The more drivers I installed, the less stable it became, but it was "better" if I BLOCKED Windows from installing any drivers on its own and only installed drivers directly from ASRock's website

System Parts
- Ryzen 1700
- ASRock X370 Taichi BIOS 1.6 official (I've tried beta 1.55 and 1.94a--both still do this)
- G.Skill 32GB (2x16GB) F4-3200C14D-32GTZ; "auto" goes to DDR4-2133 15-15-15-36 since its XMP is a no-boot; 2933 18-17-17-49 boots but won't survive prime
- PSU is an eVGA G2 1300w (i.e. massively more than enough)

Related to the last point above about drivers, I'm getting the dump file now.

EDIT: One "Unknown" (0xFFFF980010804310--says WhoCrashed) and two nt kernel but it also notes that it could be a "driver that cannot be identified at this time").

If I run prime95 blend in Safe Mode under the same conditions that insta-crash in "regular" Windows, no instant crash.

I think it's a driver problem...
 
Last edited:
Suspect a bad piece of hardware. Can you reset bios to all defaults and test with one stick of ram? Might try different ram slots as well.

Might check voltages on your PSU as well. Just because it's a nice beefy unit doesn't mean it is functioning properly.
 
ewb302 said:
Suspect a bad piece of hardware. Can you reset bios to all defaults and test with one stick of ram? Might try different ram slots as well.

Might check voltages on your PSU as well. Just because it's a nice beefy unit doesn't mean it is functioning properly.
Click to expand...

Unfortunately I lack the ability to check my PSU beyond mobo/software readings. Is that going to be any use?

I can only physically use 3 of 4 slots thanks to super-tall RAM and a big CPU cooler, but I can try moving slots/swapping sticks within that limitation.

I've imaged-back to "before any overclocks" and have booted normal Windows. I am running the test (prime95 blend, 31500MB RAM) that usually causes an instant crash to see if it does so now.

EDIT: Definitely wasn't Ryzen Master--that was a symptom, not a cause.
 
Last edited:
While my error was pretty much the exact same thing every single time, as a person who had issues upon getting Win10 installed, my best advice is: Download the MediaCreationTool.exe and have it download the latest ISO. If you have Windows setup how you like it and don't want to reinstall, then extract the ISO to the hard drive (use a different computer if you need) and then boot back into Win 10. When there, disconnect LAN/WiFi and run Setup, tell it to skip downloading any new updates, and then continue through with Updating Windows using the ISO. (Might be worth trying to start it in Safe Mode, but I don't know if it'll let you).

That solved MY issues, as well as a couple others on [H], but ours was a clock_watchdog_timeout induced BSoD. :\

However, if this is something recent, post-Overclock attempts, it could very well be that some HDD files became corrupted. So an actual reinstall would be smart, in that instance.
 
memory memory memory

either configuration or incompatible with your motherboard

i would try different ram
 
Welcome to the world of building PCs. You'd figure that after 20 plus years of this they would have it figured out.
 
Spoonie_G said:
Welcome to the world of building PCs. You'd figure that after 20 plus years of this they would have it figured out.
Click to expand...
I sympathize.

If you buy a new car that has a warranty. And you roll it off the lot. If it has problems and doesn't work you get a free rental from the dealer and THEY fix the issue.

With computer parts under warranty it is all on you to spend the hours of your time troubleshooting the problem and identifying it. And sometimes that means you have to go out and buy more parts just to eliminate things from the picture.

Doesn't seem fair. But then cars aren't computers.
 
Riptide_NVN said:
I sympathize.

If you buy a new car that has a warranty. And you roll it off the lot. If it has problems and doesn't work you get a free rental from the dealer and THEY fix the issue.

With computer parts under warranty it is all on you to spend the hours of your time troubleshooting the problem and identifying it. And sometimes that means you have to go out and buy more parts just to eliminate things from the picture.

Doesn't seem fair. But then cars aren't computers.
Click to expand...
I don't know how fair an analogy that is... LOL

It's more like going to the autoparts store and buying a bad ass set of carbon fiber valvecovers to go on those nice new high flow aluminum heads, 2.02/1.60 valves, beehive springs, double-roller rockers, high lift cam, chromeoly pushrods, a SBC 383 Stroker kit with a forged crank, h-beam connecting rods, domed pistons; an aluminum 5.7L block, a sweet fuel injection system, equal length headers, distributorless ignition system, an aluminum oil pan with dry-sump setup, fluid filled harmonic balancer, electric water pump... all of with warranties...

But then when you've put it all together correctly, the thing barely wants to stay running. :cry:

It's not something we can just say "Oh its the fuel injetion system's fault" (lets assume the delivery system is flawless) and send it back to that manufacturer claiming it doesn't work... because it could be any number of things. Could be your spark timing. Could be your rockers are misadjusted. Could be your plugs are mis-gaped and is delivering either too cold or hot of a spark. Could be the fuel injectors are too small for the engine's needs. Could be the air filter is undersized starving the engine or enough air. Heck it could even be you've tightened the connecting rod or crank cap bolts down too tight and there's too much friction.

lol Point is... In a brand new car purchase, THEY are the ones that built the car, so if anything goes wrong it's their responsibility because they warranty their work. In an engine (computer) build, YOU are the one who built everything. Therefore, anything that ends up wrong is first down to you to determine the cause of the issues. Only then can the manufacturer of that warrantied part(s) be held accountable.

On the flip side... that's exactly what the QVL (Qualified Vendor List) that motherboard and DIMM makers create (at least, I know G.Skill does, to an extent), to help ensure that everything IS plug and play. Problem in Ryzen's case, they are new, and so ALL of the kits that were available when we bought our parts, were only certified for Intel. At least we have the G.Skill Flare X and Fortius kits for ensured compatibility :)

[/analogy] :pompous:
lmao
 
Riptide_NVN said:
I sympathize.

If you buy a new car that has a warranty. And you roll it off the lot. If it has problems and doesn't work you get a free rental from the dealer and THEY fix the issue.

With computer parts under warranty it is all on you to spend the hours of your time troubleshooting the problem and identifying it. And sometimes that means you have to go out and buy more parts just to eliminate things from the picture.

Doesn't seem fair. But then cars aren't computers.
Click to expand...


And on top of that. I believe that some of the R&D is passed on the the early adopters.
 
Formula.350 said:
On the flip side... that's exactly what the QVL (Qualified Vendor List) that motherboard and DIMM makers create (at least, I know G.Skill does, to an extent), to help ensure that everything IS plug and play. Problem in Ryzen's case, they are new, and so ALL of the kits that were available when we bought our parts, were only certified for Intel. At least we have the G.Skill Flare X and Fortius kits for ensured compatibility :)

[/analogy] :pompous:
lmao
Click to expand...


Your talking as if adhering to the manufacturer's memory recommendations is a guarantee for success? It isn't. Plenty of folks purchased equipment that the motherboard manufacturers themselves recommended and still had issues. To only be fixed in a later BIOS revision. My point is that they are pushing R&D off to the end user.
 
I have literally the same memory model number except mine are 2x8gb ending in 16GTZ.

Same motherboard, must be the dual rank sticks or something just wrong with the memory.

Got any other DDR4 systems you can test the memory in?
 
Spoonie_G said:
Your talking as if adhering to the manufacturer's memory recommendations is a guarantee for success? It isn't. Plenty of folks purchased equipment that the motherboard manufacturers themselves recommended and still had issues. To only be fixed in a later BIOS revision. My point is that they are pushing R&D off to the end user.
Click to expand...
I think the main problem in most of these cases is that, like Corsair for example, tags their RAM modules with a "version" and that interprets into being something that sometimes can mean a shitload of a difference. Unfortunately for us end users, we're not able to determine said version during purchase since all the seller websites state are the manufacturer's model number and stated timings. Granted, the RAM is indeed specced and binned to operate at those published settings, but in a situation like Ryzen, that version can mean the world of difference when it comes down to the modules having totally different memory ICs :(

I haven't looked that closely to the QVLs on Ryzen boards given mine, with shipping BIOS (dated Feb 22), worked at DDR4-3200 right out of the box. That being said, using Corsair for the example still, do the QVLs state which specific version of DIMM they tested with? If not, then it's definitely the motherboard maker's fault for holding that back. However, I blame Corsair just as much since they obviously bounce around manufacturers so much as to require versioning as broad as being up to (and likely higher than) v4.31 on the 16-18-18-36 DDR3-3600 kits. At the end of the day it makes the motherboard makers liars for putting X-Kit by Y-Maker on their QVL, and then it ends up not working because Y-Maker has switched IC manufacturers and that now results in a different memory product that can't be compared against anymore. Not to say AMD is squeaky clean in all this, but they're working on it at least, so that's a positive. Indeed though, it'd have obviously been better to had gotten with the motherboard makers earlier to get this all ironed out. These motherboards have been designed for quite a few months now. Look at the PDF for the ASUS Prime B350 Plus and they include their FCC certificate, which was dated the end of last year. :confused:

There was definitely time to sort issues out had the communication efforts been made, just AMD was concerned word would get to Intel and they'd make an effort to get ahead of Ryzen however they could, which I can't really Blame AMD for given all the shit Intel has pulled in the past *shrug*
 
Sounds like typical early-adopter issues.

I'd pull a full swap and grab a MSI Titanium along with some QVL-approved RAM.

(My old-ass i5 build was a total nightmare from Nov '09 - April 2010. That's when the BIOS finally came into its own etc.)
 
horrorshow said:
Sounds like typical early-adopter issues.

I'd pull a full swap and grab a MSI Titanium along with some QVL-approved RAM.

(My old-ass i5 build was a total nightmare from Nov '09 - April 2010. That's when the BIOS finally came into its own etc.)
Click to expand...
Exactly, but I don't think many folk are factoring that in *shrug* Companies can only test stuff so thoroughly w/o it becoming a massive expense. It's once the product makes it into the wild, into the hands of users who have a wide range of varying system components, that the real work begins.

Also on the MSI Titanium QVL note... For whatever reason they're not bothering to test the CL15 kits, at least by G.Skill. So I'd like to officially for [H] declare G.Skill Trident Z F4-3200C15D-16GTZ 2x8GB 15-15-15-35 1.35V kit working at DDR4-3200 on Ryzen :p All you need to do is enable A-XMP in the OC menu.

Mind you, Ryzen has a peculiarity when it comes to odd number CAS latencies, so these default to 16-15-15-35 when using A-XMP; HOWEVER, manually adjusting it to 14-15-15-35 works just fine as well.
Additionally worth noting: the DRAM Voltage sets itself a little high, at 1.375V instead of 1.35V, so worth knocking that down and manually inputting it (which will run at 1.36V BTW).

Can't say with certainty but currently I'm running at 14-15-15-35 with 1.32V (1.328V) stable thus far. Last night I even managed 1.29V (1.30V) and while it booted into Windows and passed my testing, it did initially hit a windows Bootloader "Kernel error" that gave me the option of "Continue" or "F8 for Boot Options". Continuing worked fine *shrug*
 
You must log in or register to reply here.
Back
Top