New build crashing in games (but not stress tests)

[goshsowitty]

After 10 years it was time to build a new gaming PC on a relatively small budget but something that would last hopefully another 10 years (at a push), so I bought and built this over Christmas and New Year:

PCPartPicker Part List

Type Item Price

CPU	AMD Ryzen 5 5600X 3.7 GHz 6-Core Processor	£160.99 @ Amazon UK
Motherboard	Gigabyte B550 AORUS ELITE AX V2 ATX AM4 Motherboard	£140.99 @ Technextday
Memory	Corsair Vengeance LPX 32 GB (2 x 16 GB) DDR4-3200 CL16 Memory	£87.49 @ Amazon UK
Storage	Crucial P5 Plus 1 TB M.2-2280 PCIe 4.0 X4 NVME Solid State Drive	£86.31 @ CCL Computers
Video Card	PowerColor Fighter Radeon RX 6700 XT 12 GB Video Card	£414.11 @ Amazon UK
Case	Lian Li LANCOOL II-W ATX Mid Tower Case	£109.95 @ Amazon UK
Power Supply	Corsair TX650M Gold 650 W 80+ Gold Certified Semi-modular ATX Power Supply	£79.98 @ Amazon UK
Prices include shipping, taxes, rebates, and discounts
Total	£1079.82
Generated by PCPartPicker 2023-01-05 00:04 GMT+0000

All went smoothly and for the most part the PC runs great for a little while.

In some games, but not all, I can get 10-20 minutes of gameplay before the PC freezes and reboots.

In Fortnite on recommended auto settings I get about 15 minutes if I'm lucky. In Call of Duty Modern Warfare 2 I got about 5 minutes. Fornite can be played for much longer if we turn down the graphics settings a little.

In The Sims 4 my partner was able to play non-stop for hours without issue.

In 3D Mark stress tests it can run 20 passes without issue.

Throughout all of this the system temps are all fine.

Windows (11) logs Event 46, WHEA-Logger which implies RAM:

A fatal hardware error has occurred.

Component: Memory
Error Source: Machine Check Exception

Occasionally, but not consistently, we get Event 18, WHEA-Logger:

A fatal hardware error has occurred.

Reported by component: Processor Core
Error Source: Machine Check Exception
Error Type: Bus/Interconnect Error
Processor APIC ID: 0

Due to the Event 46 I obviously suspected RAM initially. Windows Memory Diagnostic and memtest86 are both clear. I also tried both RAM sticks in the other two dual channel slots, and each RAM stick individually in all slots but we still get the crashes. I experience the issue regardless of the XMP profile of the RAM.

The build details above do not include the original PSU I bought for the build. I realised I bought a pretty bad PSU initially by all accounts - a Gamemax GM 700W. I replaced it with the Corsair TX650M when I realised that Gamemax stuff is not particularly highly thought of. Unfortunately, while I got longer than usual tonight with the Corsair PSU, it eventually crashed after 18 minutes (playing Fornite).

This was the cheapest most suspect part of the build so I don't lose anything by RMAing the original one and ending up with a better PSU but it hasn't significantly helped.

Aside from what I've already mentioned I've done the following:

• Updated the BIOS to the latest available revision (it wasn't too far behind)
• Updated all AMD chipset/graphics drivers

Today I have run stress tests in Prime95, OCTT, HeavyLoad and Furmark and none of them have exhibited any issue.

Some dumps analyzed with Windbg:

MODULE_NAME: AuthenticAMD

IMAGE_NAME:  AuthenticAMD.sys

STACK_COMMAND:  .cxr; .ecxr ; kb

FAILURE_BUCKET_ID:  LKD_0x124_0_AuthenticAMD_PROCESSOR__UNKNOWN_IMAGE_AuthenticAMD.sys

OSPLATFORM_TYPE:  x64

OSNAME:  Windows 10

FAILURE_ID_HASH:  {f59f17e7-f24e-04f5-3f16-e9425b2acba5}

MODULE_NAME: AuthenticAMD

IMAGE_NAME:  AuthenticAMD.sys

STACK_COMMAND:  .cxr; .ecxr ; kb

FAILURE_BUCKET_ID:  LKD_0x124_0_AuthenticAMD_MEMORY__UNKNOWN_FATAL_IMAGE_AuthenticAMD.sys

OSPLATFORM_TYPE:  x64

OSNAME:  Windows 10

FAILURE_ID_HASH:  {5ce15480-df81-5c3f-8bd4-ee19bb0b5a9a}

At this point, I'm not entirely sure what more I can do to troubleshoot. So I am hoping people have suggestions for what I might be able to do to test, and stress the build quicker and more consistently and to narrow down the issue. Or if anyone else may have experienced anything similar.

Thanks so much if you're still reading to here. Very grateful for any help.

 

[ThrashZone]

Hi,
Use one stick and repeat.
If all is well on one stick take it out and insert the other stick and repeat.

 

[tabascosauz]

@goshsowitty Bus/Interconnect is not RAM, it's unstable Infinity Fabric. Download Zentimings and post up a screenshot.

ZenTimings

ZenTimings is a simple and lightweight app for monitoring memory timings on Ryzen platform.

zentimings.protonrom.com

WHEAs do not report on RAM stability, but best to test that too. Everything you need in the link:

Memory Testing with TestMem5 TM5 with custom configs

Hello everybody I am just making a very light tutorial with a collection of custom config files and a DOWNLOAD LINK for TM5 v0.12 anta777 absolut config Official Intel DDR4 24/7 Memory Stability Thread None of the work is mine but it seems like a pretty good and fast testing app Update...

www.overclock.net

On Ryzen there are three domains to RAM stability:
- RAM itself (test with TM5)
- Infinity Fabric (Bus/Interconnect WHEA)
- Memory controller (in terms of pure speed, usually not a concern at these speeds)

 

[goshsowitty]

Hi,
Use one stick and repeat.
If all is well on one stick take it out and insert the other stick and repeat.

Thanks for the reply!

So I haven't done actual RAM tests with this approach but I did already try RAM in different slots, one at a time etc. and it still crashes. Might be worth doing memory tests with one stick at a time but I'm beginning to think it isn't likely to be RAM at this point.

@goshsowitty Bus/Interconnect is not RAM, it's unstable Infinity Fabric. Download Zentimings and post up a screenshot.

Thanks for the reply!

This unfortunately means nothing to me but I hope something sticks out here

 

[tabascosauz]

@goshsowitty nothing super wrong as it's just XMP, but try some higher SOC Voltage. Maybe 1.1V to start.

You should also be able to find VDDG_CCD and VDDG_IOD on the same BIOS page. IOD can probably come up to 1.0V for now.

It's strange to see Bus/Interconnect at just 1600MHz, even with these settings.

Can you share a picture of the sticker on your RAM sticks? These are obviously 16Gb ICs of some sort.

 

[goshsowitty]

@goshsowitty nothing super wrong as it's just XMP, but try some higher SOC Voltage. Maybe 1.1V to start.

You should also be able to find VDDG_CCD and VDDG_IOD on the same BIOS page. IOD can probably come up to 1.0V for now.

It's strange to see Bus/Interconnect at just 1600MHz, even with these settings.

So everything in my BIOS was set to Auto and I’ve changed SOC as suggested to 1.1V.

There is only one VDDG voltage setting which if I set to Manual I can change the voltage to any whole number, which I found strange but just wanted to make sure you think this looks ok before I save it.

Can you share a picture of the sticker on your RAM sticks? These are obviously 16Gb ICs of some sort.

 

[tabascosauz]

So everything in my BIOS was set to Auto and I’ve changed SOC as suggested to 1.1V.

There is only one VDDG voltage setting which if I set to Manual I can change the voltage to any whole number, which I found strange but just wanted to make sure you think this looks ok before I save it.

The user compiled list of Corsair rev numbers says Samsung 16Gb M-die? Not a great choice but as long as it'll pass TM5 then no problem, if you're not looking to overclock

As for Fabric stability there's not a whole lot you can do other than continue to game and use your PC as you usually do, and see if anything changes. There are certain stress tests alleged to be good for Fabric testing but all of them are hit or miss due to the complex nature of IF stability.

But get that memtesting down asap, make sure TM5 shows @anta777Extreme1 or @1usmusv3

 

[eidairaman1]

Ugh Corsair, that explains it all

 

[goshsowitty]

But get that memtesting down asap, make sure TM5 shows @anta777Extreme1 or @1usmusv3

Gonna try and play some less intensive games for a bit and run overnight. Thanks for the help so far.

Ugh Corsair, that explains it all

Ugh. Seemed like a good choice at the time. If I manage to prove a RAM fault and do an RMA do you have a suggestion for a better kit?

I went with 2 x 16GB sticks as may use the machine for game development and video editing but happy to pay a little more for a quality set and just go with 2 x 8GB for now if I get a good recommendation.

Not overly concerned with overclocking tbh just want something stable and functional!

 

[eidairaman1]

Gonna try and play some less intensive games for a bit and run overnight. Thanks for the help so far.



Ugh. Seemed like a good choice at the time. If I manage to prove a RAM fault and do an RMA do you have a suggestion for a better kit?

I went with 2 x 16GB sticks as may use the machine for game development and video editing but happy to pay a little more for a quality set and just go with 2 x 8GB for now if I get a good recommendation.

Not overly concerned with overclocking tbh just want something stable and functional!

Yeah no Corsair,

I Suggest Mushkin Black or RedLine, Crucial Ballistix Gaming, Team Group, Geil, Gskill, Or Value Lines from them.

 

[tabascosauz]

Gonna try and play some less intensive games for a bit and run overnight. Thanks for the help so far.

Ugh. Seemed like a good choice at the time. If I manage to prove a RAM fault and do an RMA do you have a suggestion for a better kit?

I went with 2 x 16GB sticks as may use the machine for game development and video editing but happy to pay a little more for a quality set and just go with 2 x 8GB for now if I get a good recommendation.

Not overly concerned with overclocking tbh just want something stable and functional!

Which BIOS exactly are you on?

I'm not sure if it's an AGESA 1207 thing but I ran into a similar problem on one of my MSI boards with a 1207 BIOS, VDDG ended up consolidated into a single option with no way to split CCD from IOD. It's okay though, CCD can also come up to 1.0V, it's not a big deal.

Looks like you have to enter in millivolts, so for 1.0V punch in 1000.

 

[formula383]

After 10 years it was time to build a new gaming PC on a relatively small budget but something that would last hopefully another 10 years (at a push), so I bought and built this over Christmas and New Year:

PCPartPicker Part List

Type Item Price

CPU	AMD Ryzen 5 5600X 3.7 GHz 6-Core Processor	£160.99 @ Amazon UK
Motherboard	Gigabyte B550 AORUS ELITE AX V2 ATX AM4 Motherboard	£140.99 @ Technextday
Memory	Corsair Vengeance LPX 32 GB (2 x 16 GB) DDR4-3200 CL16 Memory	£87.49 @ Amazon UK
Storage	Crucial P5 Plus 1 TB M.2-2280 PCIe 4.0 X4 NVME Solid State Drive	£86.31 @ CCL Computers
Video Card	PowerColor Fighter Radeon RX 6700 XT 12 GB Video Card	£414.11 @ Amazon UK
Case	Lian Li LANCOOL II-W ATX Mid Tower Case	£109.95 @ Amazon UK
Power Supply	Corsair TX650M Gold 650 W 80+ Gold Certified Semi-modular ATX Power Supply	£79.98 @ Amazon UK
Prices include shipping, taxes, rebates, and discounts
Total	£1079.82
Generated by PCPartPicker 2023-01-05 00:04 GMT+0000

All went smoothly and for the most part the PC runs great for a little while.

In some games, but not all, I can get 10-20 minutes of gameplay before the PC freezes and reboots.

In Fortnite on recommended auto settings I get about 15 minutes if I'm lucky. In Call of Duty Modern Warfare 2 I got about 5 minutes. Fornite can be played for much longer if we turn down the graphics settings a little.

In The Sims 4 my partner was able to play non-stop for hours without issue.

In 3D Mark stress tests it can run 20 passes without issue.

Throughout all of this the system temps are all fine.

Windows (11) logs Event 46, WHEA-Logger which implies RAM:

A fatal hardware error has occurred.

Component: Memory
Error Source: Machine Check Exception

Occasionally, but not consistently, we get Event 18, WHEA-Logger:

A fatal hardware error has occurred.

Reported by component: Processor Core
Error Source: Machine Check Exception
Error Type: Bus/Interconnect Error
Processor APIC ID: 0

Due to the Event 46 I obviously suspected RAM initially. Windows Memory Diagnostic and memtest86 are both clear. I also tried both RAM sticks in the other two dual channel slots, and each RAM stick individually in all slots but we still get the crashes. I experience the issue regardless of the XMP profile of the RAM.

The build details above do not include the original PSU I bought for the build. I realised I bought a pretty bad PSU initially by all accounts - a Gamemax GM 700W. I replaced it with the Corsair TX650M when I realised that Gamemax stuff is not particularly highly thought of. Unfortunately, while I got longer than usual tonight with the Corsair PSU, it eventually crashed after 18 minutes (playing Fornite).

This was the cheapest most suspect part of the build so I don't lose anything by RMAing the original one and ending up with a better PSU but it hasn't significantly helped.

Aside from what I've already mentioned I've done the following:

• Updated the BIOS to the latest available revision (it wasn't too far behind)
• Updated all AMD chipset/graphics drivers

Today I have run stress tests in Prime95, OCTT, HeavyLoad and Furmark and none of them have exhibited any issue.

Some dumps analyzed with Windbg:

MODULE_NAME: AuthenticAMD

IMAGE_NAME:  AuthenticAMD.sys

STACK_COMMAND:  .cxr; .ecxr ; kb

FAILURE_BUCKET_ID:  LKD_0x124_0_AuthenticAMD_PROCESSOR__UNKNOWN_IMAGE_AuthenticAMD.sys

OSPLATFORM_TYPE:  x64

OSNAME:  Windows 10

FAILURE_ID_HASH:  {f59f17e7-f24e-04f5-3f16-e9425b2acba5}

MODULE_NAME: AuthenticAMD

IMAGE_NAME:  AuthenticAMD.sys

STACK_COMMAND:  .cxr; .ecxr ; kb

FAILURE_BUCKET_ID:  LKD_0x124_0_AuthenticAMD_MEMORY__UNKNOWN_FATAL_IMAGE_AuthenticAMD.sys

OSPLATFORM_TYPE:  x64

OSNAME:  Windows 10

FAILURE_ID_HASH:  {5ce15480-df81-5c3f-8bd4-ee19bb0b5a9a}

At this point, I'm not entirely sure what more I can do to troubleshoot. So I am hoping people have suggestions for what I might be able to do to test, and stress the build quicker and more consistently and to narrow down the issue. Or if anyone else may have experienced anything similar.

Thanks so much if you're still reading to here. Very grateful for any help.

A few things come to mind here. What cpu cooler do you have and is it working correctly? How is the case flow is it possible to remove the front panel and dust screen and possibly ramp up the case fan speeds for testing?

The perhaps more likely cause is a defective PSU. this can happen with any psu but honesly the TX series from corsair is pretty mediocre at best imho. And its corsair, great warranty support. But pretty half assed products unless you know exactly what your looking for and its almost always well over priced for what you are getting. RM series from corsair is said to be good if they still make them and i think the HX series might be ok as well. IMO look up SuperFlower leadex iii or gold units, or even some seasonic focus are solid units as well.

as for wattages you want around double the power that you can use max so your psu is never put under heavy stress and causing it to go past its optimal power efficiency curve most units hit peak around 50~70% usage.

So for your use case i would look for 660w at the very absolute minimum (do note a 450w may power the system just fine its more of a matter of for how long will it do it with out crashing) psu with a good quality psu that is. 850w would be the "sweet spot" for long term use. and even 1000w+ would still be OK. even if fairly overkill ish for your current use. This all depends on pricing dont feel like you cant use a 1300w if the price is right and its a good unit. the SuperFlower leadex gold 1300w is 229 with a 50 instant rebait so 179$ for a very solid psu thats not a bad price and even smaller units can be priced similar. But if your on a tight budget i would aim for around the 850w mark for your build. Again if quality of the unit is good and price is also good. So it just depends what is available in your area.

If you have another psu to test try it. make sure the cooler is making proper contact with the cpu and temps are NORMAL not just OK ish. If cooling is good and psu is good then its likely a bad mother board.

 

[freeagent]

Corsair has a bad rep with Zen 3..

I have some Adata sticks with those same timings.. They run at XMP just fine, but if I cut the power to the PSU, I have to re-enter all settings, but fine with my B-Die

Honestly, I really don't have too much good to say about Corsair, I have been bitten by them a few times, now I am bitter

as for wattages

He is fine.

 

[Norton]

Corsair has a bad rep with Zen 3..

Running Corsair Vengeance 3600 ram on my 5950X rig (4×8GB timing 16-19-19-36 ( XMP). No issues/errors running full load 24/7 while crunching.

OP- try dropping ram speed to 2800 with rest of the timings, etc at XMP. I have had some success with odd memory issues/crashes by just cutting down the speed a bit

 

[freeagent]

Running Corsair Vengeance 3600 ram on my 5950X rig (4×8GB timing 16-16-16-39 ( XMP). No issues/errors

Is that Sammy B-Die?

If so could be why..

 

[Norton]

Is that Sammy B-Die?

If so could be why..

It's not Samsung B... micron iirc. Will check and update this post when I get home.

EDIT- timing is 16-19-19-36 and it's Micron E die ram
CORSAIR Vengeance LPX 32GB (4 x 8GB) 288-Pin PC RAM DDR4 3600 (PC4 28800) Desktop Memory Model CMK32GX4M4D3600C16 - Newegg.com

 

[eidairaman1]

It's not Samsung B... micron iirc. Will check and update this post when I get home.

EDIT- timing is 16-19-19-36 and it's Micron E die ram
CORSAIR Vengeance LPX 32GB (4 x 8GB) 288-Pin PC RAM DDR4 3600 (PC4 28800) Desktop Memory Model CMK32GX4M4D3600C16 - Newegg.com

Crucial Ballistix Gaming 3600 (NO RGB) on Ry7 5800, 32GB no issues

 

[goshsowitty]

OP- try dropping ram speed to 2800 with rest of the timings, etc at XMP. I have had some success with odd memory issues/crashes by just cutting down the speed a bit

No joy unfortunately.

Which BIOS exactly are you on?

It’s the latest version of the B550 Aorus Elite AX V2 BIOS, F16d.

I’m going to set those voltages now to see if this makes a difference. (Everything is back to how it was).

 

[eidairaman1]

A few things come to mind here. What cpu cooler do you have and is it working correctly? How is the case flow is it possible to remove the front panel and dust screen and possibly ramp up the case fan speeds for testing?

The perhaps more likely cause is a defective PSU. this can happen with any psu but honesly the TX series from corsair is pretty mediocre at best imho. And its corsair, great warranty support. But pretty half assed products unless you know exactly what your looking for and its almost always well over priced for what you are getting. RM series from corsair is said to be good if they still make them and i think the HX series might be ok as well. IMO look up SuperFlower leadex iii or gold units, or even some seasonic focus are solid units as well.

as for wattages you want around double the power that you can use max so your psu is never put under heavy stress and causing it to go past its optimal power efficiency curve most units hit peak around 50~70% usage.

So for your use case i would look for 660w at the very absolute minimum (do note a 450w may power the system just fine its more of a matter of for how long will it do it with out crashing) psu with a good quality psu that is. 850w would be the "sweet spot" for long term use. and even 1000w+ would still be OK. even if fairly overkill ish for your current use. This all depends on pricing dont feel like you cant use a 1300w if the price is right and its a good unit. the SuperFlower leadex gold 1300w is 229 with a 50 instant rebait so 179$ for a very solid psu thats not a bad price and even smaller units can be priced similar. But if your on a tight budget i would aim for around the 850w mark for your build. Again if quality of the unit is good and price is also good. So it just depends what is available in your area.

If you have another psu to test try it. make sure the cooler is making proper contact with the cpu and temps are NORMAL not just OK ish. If cooling is good and psu is good then its likely a bad mother board.

Only way to know this is to try different ram

 

[tabascosauz]

No joy unfortunately.


It’s the latest version of the B550 Aorus Elite AX V2 BIOS, F16d.

I’m going to set those voltages now to see if this makes a difference. (Everything is back to how it was).

Keep us posted. Hopefully it's TM5 stable and the bump in volts solves the WHEAs.

If not, I would honestly RMA the CPU. We like to complain about having dud CPUs running into Bus/Interconnect at 3600/3733/3800 but 3200 is literally AMD official spec. 1.1V SOC is already plenty high for 3200, and should honestly be unnecessary to sustain 1600MHz Fabric with reasonable VDDGs.

 

[goshsowitty]

I’m going to set those voltages now to see if this makes a difference. (Everything is back to how it was).

Weirdly I noticed that it looked like the settings I found before for VDDG didn’t take effect.

These were in the Tweaker tab.

But theres another page of settings under AMD Overclocking and it lets me change CCD and IOD independently.

I’ve set both to 1000 now and that seems to have taken effect according to ZenTimings.

Will game for 20 minutes and see what happens.

 

[tabascosauz]

@goshsowitty you can try it instead of the VDDG control under Tweaker, but generally we don't touch most of AMD OC, at least as a first choice, because the menu there is designed to be changed by Ryzen Master.

If you have independent VDDG control, CCD can generally live around the 0.9-1.0V mark. It usually isn't the lynchpin of Fabric stability, unlike VSOC and IOD

 

[goshsowitty]

Unfortunately didn’t make a difference.

It’s late so I’m going to go back to my stock config with the XMP profile and run TM5.

I guess we can safely say it isn’t RAM at that point if this is successful.

We suspect CPU at that point? It bothers me that the various stress tests I’ve done haven’t tripped over.

It’s also frustrating having to reproduce the issue each time by playing games. I wish there was a quicker way to get it to fall over.

 

[tabascosauz]

Unfortunately didn’t make a difference.

It’s late so I’m going to go back to my stock config with the XMP profile and run TM5.

I guess we can safely say it isn’t RAM at that point if this is successful.

We suspect CPU at that point? It bothers me that the various stress tests I’ve done haven’t tripped over.

It’s also frustrating having to reproduce the issue each time by playing games. I wish there was a quicker way to get it to fall over.

The WHEAs are pretty clear. If massaging the VDDGs and VSOC can't fix the problem then the CPU should go back. Not getting lucky enough to hit 4000 1:1 is one thing; having to dip below official DDR4 spec for this platform is something completely different and not acceptable.

Though, between the board/RAM/CPU things can definitely defy logic and expectations.

I returned a B550M Steel Legend at B550 launch because the BIOSes available at the time were all incredibly unstable and crashed under the slightest system load even with known stable RAM profiles.

I last year RMA'd my original 5700G, B550I Aorus AX, and Patriot 4400CL19 kit because apparently they all went bad.


One last thing you can try is disabling Global Cstates. I don't remember where it is in Gigabyte BIOS. It was the last hope for my 3700X that was not only bad IF like yours (Bus/Interconnect) but bad cores as well (Cache Hierarchy).

 

[phanbuey]

it might be the motherboard not feeding enough volts at idle. Is there a low load setting that bumps the voltage to the CPU a bit?

System unstable on idle | Overclock.net

^ this thread might help