PC Tech PC won’t load Windows after reboot.

[AbsolutSurgen]

Does your MB allow you to run Memtest86 from Bios?

[Spork3245]

3 minutes ago, AbsolutSurgen said:

Does your MB allow you to run Memtest86 from Bios?

Not from BIOS, but I’m currently running it off USB. It loads, unlike the Windows installer. Currently testing the new Corsair RAM.

[Spork3245]

With the new RAM received yesterday giving me the same issue as my 4x 3000mhz Corsair RAM, I’m not sure if it’s an issue with the actual RAM, though. Not to mention 4 sticks of RAM going bad at once is improbable on it’s own, but for 4 sticks that have been fine for over 4 years to go bad + 2 sticks of brand new RAM to also be bad is kind of insane. None of it 100% impossible, though!

 

I’m leaning towards something with the memory controller, maybe? I seriously have no clue.

[AbsolutSurgen]

1 minute ago, Spork3245 said:

With the new RAM received yesterday giving me the same issue as my 4x 3000mhz Corsair RAM, I’m not sure if it’s an issue with the actual RAM, though. Not to mention 4 sticks of RAM going bad at once is improbable on it’s own, but for 4 sticks that have been fine for over 4 years to go bad + 2 sticks of brand new RAM to also be bad is kind of insane. None of it 100% impossible, though!

 

I’m leaning towards something with the memory controller, maybe? I seriously have no clue.

I would try running Memtest86 on the same stick in each of the four slots, to determine if the MB has a problem with one of the slots.

[Spork3245]

3 minutes ago, AbsolutSurgen said:

I would try running Memtest86 on the same stick in each of the four slots, to determine if the MB has a problem with one of the slots.

Well Windows won’t load with a Corsair stick in any of the slots, but the GSkill RAM works in any. AFAIK they’re all dual-rank.
It’s perplexing! 

[cusideabelincoln]

I've been doing some memory overclocking lately and it can be very finnicky. Changing a value by 1 out of the 50 or so different timings and voltages can cause the RAM to go from perfectly stable to not bootable.

 

Now as to why your system is not setting the proper default values, which should work, is a head scratcher.

 

Matching sticks of RAM is definitely the best, because there are about 5-6 different memory chips that are prevalent and they all require different secondary timings to be stable. Even Corsair will use chips from different vendors under the same "brand" name. It seems unlikely that your system is setting a proper value for the G.skill but not for any of the Corsairs... but I guess it could be possible. Dual rank sticks do require different subtimings than single rank sticks, and if you run all 4 sticks that also activates two more subtimings that need to be configured properly.

 

Or if your testing is just a result of random instability:

 

If it's a failing memory controller, you could try setting your Corsair sticks to default timings and speed. Then set your SoC voltage to 1.1V and your VDDG-IOD to 1.2V and see what happens, as these should be the highest safe values for 24/7 usage while the default values should be 1V and 0.9V. I think this is a place to start, and if these voltages don't work then somehow your board is setting incorrect timings.... or you CPU is going bad.

 

@Spork3245

 

Oh and did you clear CMOS before installing the brand new Corsair sticks?

 

[Spork3245]

30 minutes ago, cusideabelincoln said:

I've been doing some memory overclocking lately and it can be very finnicky. Changing a value by 1 out of the 50 or so different timings and voltages can cause the RAM to go from perfectly stable to not bootable.

 

Now as to why your system is not setting the proper default values, which should work, is a head scratcher.

 

Matching sticks of RAM is definitely the best, because there are about 5-6 different memory chips that are prevalent and they all require different secondary timings to be stable. Even Corsair will use chips from different vendors under the same "brand" name. It seems unlikely that your system is setting a proper value for the G.skill but not for any of the Corsairs... but I guess it could be possible. Dual rank sticks do require different subtimings than single rank sticks, and if you run all 4 sticks that also activates two more subtimings that need to be configured properly.

 

Or if your testing is just a result of random instability:

 

If it's a failing memory controller, you could try setting your Corsair sticks to default timings and speed. Then set your SoC voltage to 1.1V and your VDDG-IOD to 1.2V and see what happens, as these should be the highest safe values for 24/7 usage while the default values should be 1V and 0.9V. I think this is a place to start, and if these voltages don't work then somehow your board is setting incorrect timings.... or you CPU is going bad.

 

@Spork3245

 

Oh and did you clear CMOS before installing the brand new Corsair sticks?

 

 

The sticks in question were working for 3+ years in this system, no changes were made. The new RAM was purchased when I saw that the GSkill RAM worked as I assumed the old Corsair RAM must be bad, CMOS was cleared and various timings + voltages (even dropping speed down to 1833mhz) was tried on the new 3600mhz RAM. Just tried the old Corsair RAM in my Dad's Intel system, it worked fine.

[stepee]

Have you tried throwing in the towel and building a brand new 7800x3D build?

[Spork3245]

8 minutes ago, stepee said:

Have you tried throwing in the towel and building a brand new 7800x3D build?

 

 

 

 

Seriously, though, and honestly, I can't afford that at all right now. The vet bills for my dog that passed in June were nearly $7k, on top of the $2k for the puppy we reserved a few weeks before she got sick and passed (the puppy was supposed to be "hers"), then in October my wife and I had to put out $2.5k to get her car fixed. Just stretched beyond stretched at the moment.

[cusideabelincoln]

14 minutes ago, Spork3245 said:

 

The sticks in question were working for 3+ years in this system, no changes were made. The new RAM was purchased when I saw that the GSkill RAM worked as I assumed the old Corsair RAM must be bad, CMOS was cleared and various timings + voltages (even dropping speed down to 1833mhz) was tried on the new 3600mhz RAM. Just tried the old Corsair RAM in my Dad's Intel system, it worked fine.

 

Starting to sound like a degraded memory controller since the lower-stress G.skill works, but then the dual-rank and quad-rank setups are harder on the memory controller and don't work.

 

Again SoC and VDDG-IOD voltages are specifically for the memory controller so those should be the only voltages you need to change to help the CPU out, if possible.

 

Command Rate, Gear Down Mode, ProcODT and the various DrvStr values can also have an effect on stability and could be worth changing to attempt to get the system running.

[Spork3245]

3 minutes ago, cusideabelincoln said:

 

Starting to sound like a degraded memory controller since the lower-stress G.skill works, but then the dual-rank and quad-rank setups are harder on the memory controller and don't work.

 

Again SoC and VDDG-IOD voltages are specifically for the memory controller so those should be the only voltages you need to change to help the CPU out, if possible.

 

Command Rate, Gear Down Mode, ProcODT and the various DrvStr values can also have an effect on stability and could be worth changing to attempt to get the system running.

 

I just got this motherboard, I RMA'd my previous and ASUS sent me this one (September iirc). Unless you think it's the IMC on the CPU? Is the GSkill Single Rank RAM (I have no idea)? It's Ripjaws series.

[stepee]

Glad @cusideabelincoln is here, I’m horrible at hardware troubleshooting and want our boy Spork to get settled

[cusideabelincoln]

44 minutes ago, Spork3245 said:

 

I just got this motherboard, I RMA'd my previous and ASUS sent me this one (September iirc). Unless you think it's the IMC on the CPU? Is the GSkill Single Rank RAM (I have no idea)? It's Ripjaws series.

 

 

 

IMC is only on the CPU. The motherboard is just there to provide a physical connection between the CPU and RAM and control voltages. With higher speed memory a better quality motherboard means better quality physical traces to the CPU which helps prevent signal loss and electromagnetic interference.

 

I have no idea if the G.skill is single or dual rank, but I just assume single rank since you said it was older. But that also depends on the capacity, as it could easily be dual-rank. However maybe some of the other subtimings are more relaxed.

 

I would try these settings first at the JEDEC standard 2400 MHz with default timings on the new Corsair set:

VDIMM (RAM voltage): 1.35V

2T command rate, much easier on the memory controller.

Gear Down Mode: Enabled (should auto to enabled anyway)

SoC voltage: 1.05 to 1.15V

VDDG-IOD: 0.95 to 1.05V (make sure this is always at least .04V less than SoC)

 

See how those go, if a no go then try...

ProcODT: Start at 40 Ohm, and work your way up to 80 Ohm if it fails

ClkDrvStr: Start at 20 Ohm and work your way up to 120 Ohm. The three other DrvStr you can leave on auto.

 

[cusideabelincoln]

Also if you get the G.skill back in and booted up, download a program called ZenTimings and we can see what your motherboard is setting the default values to for a baseline to compare against the Corsair sticks.

 

You could also run a test on the IMC/Infinity fabric. To do so, you need a program called OCCT and the latest version of Prime95. You run OCCT's VRAM stability benchmark and then run Prime95's LargeFFT at the same time. OCCT will report WHEA errors fairly quickly if there's an issue.

[cusideabelincoln]

Also reading over this again... could still be motherboard training (training is when it autodetermines the timings) issue.

 

7 hours ago, Spork3245 said:

So the RMA replacement motherboard started doing the same issue on Tuesday, however, flashing the BIOS no longer fixes the issue.

I remembered that I had an (extremely) old pair of GSkill Ripjaw memory (2400mhz C15) from an X99 build. Tried that, Windows booted. Went back to the 3000mhz C15 Corsair RAM but only with two of the sticks in DIMMs 2 and 4 (same I tried the GSkill in), Windows booted, tried reinstalling the other two Corsair sticks, same issue, removed all 4, put the “working” two in DIMMs 1 and 3, Windows would not boot. “Okay, bad DIMM?” I thought, so I moved the two back to 2 and 4, same thing. “Did I mix up the RAM?”, swapped the pair for the other Corsair, same issue. Went back to the GSkill, worked.

Ordered a new set of RAM, 64gb (2x32gb) Corsair 3600mhz C18, came yesterday… same issue. GSkill still works. Even tried running each of the Corsair sets at default (non-XMP/DOCP) settings as well as matching the (slower) GSkill ram, and going as far to attempt to run the Corsair as low as 1833mhz: Windows won’t load regardless, and same as before, trying Windows installer from a bootable USB gives me the same problem (even with the m.2 with Windows on it disconnected).

 

I’m at a complete loss on what the issue could possibly be.

 

Ones that can cause booting issues:

 

tRFC: this is pretty picky. Some sticks can do crazy low timings, others have to be 3x or 4x the timing. This also changes with frequency. At 3600 Mhz a "safe" value should for most sticks is 630 or more. At 2400 Mhz this can be >420, at 3200 MHz it can be >560.

tCL: A main timing, 16 or 18 should work across the board. Must be an even value when Gear Down Mode is enabled.

tRCDRD: Another main timing, some memory sticks can't go low. 20 or more is super safe, although really good mem can do 14.

tRC: Can vary wildly between different types of memory chips and frequency. Some can do 40, others need 60+. 80 is probably safe across the board.

 

tRDRDSD: This should only affect Dual-Rank sticks. Really good mem can do 1 or 2, but a super safe number is like 8.

tWRWRSD: This should only affect Dual-Rank sticks. Really good mem can do 1 or 2, but a super safe value is 8.

 

tRDRDDD: This should only matter when you have 4 sticks in the slots.

tWRWRDD: This should only matter when you have 4 sticks in the slots. 

 

tRDRDSCL: >5 is super safe, don't set this too low

tWRWRSCL: >5 is super safe, don't set this too low.

 

tRDWR

tWRRD

These two kind of work together. You usually can't set both of them to lower values. One needs to be kind of low, and the other higher. 

[Spork3245]

Out at the moment and can’t try the suggestions (likely not until tomorrow), but one question:

If I purchase a new AM4 CPU on Amazon to try a different CPU and see if that’s the issue, can I return it after use or is that a really bad idea?

[cusideabelincoln]

Not sure about return policy. Sounds risky. Good chance of getting charged a restocking fee at best.

[Spork3245]

1 hour ago, cusideabelincoln said:

Not sure about return policy. Sounds risky. Good chance of getting charged a restocking fee at best.

Well I’m almost certainly sending this 3600mhz RAM back since it doesn’t work either lol

[Spork3245]

Tried another 5900x. Everything works with this other CPU installed. Looks like it's the CPU's IMC going bad or something. Now I get to call AMD on Monday to see how their CS is. Joy!

 

Something interesting: I ran this 5900x with the new 3600mhz C18 RAM, rebooted multiple times, no issue, ran Cyberpunk (4k, max settings w/path tracing, DLSS at quality and DLSS at balanced two runs each), then installed my old 3000mhz C15 RAM, did the same including Cyberpunk... and Cyberpunk's bench results were consistently 2-3fps faster on avg with the 3000mhz RAM. I guess the 5900x prefers latency over speed. For awhile I've been wondering if I should "upgrade" to faster memory, and I guess that's a big N.O.  

[cusideabelincoln]

3 hours ago, Spork3245 said:

Tried another 5900x. Everything works with this other CPU installed. Looks like it's the CPU's IMC going bad or something. Now I get to call AMD on Monday to see how their CS is. Joy!

 

Something interesting: I ran this 5900x with the new 3600mhz C18 RAM, rebooted multiple times, no issue, ran Cyberpunk (4k, max settings w/path tracing, DLSS at quality and DLSS at balanced two runs each), then installed my old 3000mhz C15 RAM, did the same including Cyberpunk... and Cyberpunk's bench results were consistently 2-3fps faster on avg with the 3000mhz RAM. I guess the 5900x prefers latency over speed. For awhile I've been wondering if I should "upgrade" to faster memory, and I guess that's a big N.O.  

 

Interesting, I really hope my Ryzen doesn't do this as I've heard it can happen with any of the 5000 chips.

 

Overall Ryzen does prefer latency, since the bandwidth cap is around DDR3800 due to the Infinity fabric limitations. But 3000c15 should theoretically be the same latency as 3600c18 (10ns), assuming the other timings are proportionally similar. If you're running four dual rank sticks, that means you're in a quad rank configuration which effectively increases your bandwidth.

[Spork3245]

1 hour ago, cusideabelincoln said:

 

Interesting, I really hope my Ryzen doesn't do this as I've heard it can happen with any of the 5000 chips.

 

Overall Ryzen does prefer latency, since the bandwidth cap is around DDR3800 due to the Infinity fabric limitations. But 3000c15 should theoretically be the same latency as 3600c18 (10ns), assuming the other timings are proportionally similar. If you're running four dual rank sticks, that means you're in a quad rank configuration which effectively increases your bandwidth.

3600 18-22-22-42

vs

3000 15-17-17-35

 

It could be that the 4x 3000mhz are running in quad rank as you suggested, which can apparently be “hard on the IMC” 

[cusideabelincoln]

Increasing the rank, from single->dual->quad, is harder in the same way that increasing frequency is harder. If the memory sticks were to be of equal capabilities, it's easier for the IMC to run a higher frequency with a single rank set than it is with dual rank than it is for quad rank.

 

It's probably a combination, because those other timings are pretty bad too on the 3600 kit and maybe it has even worse other ones.

[Reputator]

IIRC an unstable Infinity Fabric clock can manifest as memory instability.

[Spork3245]

10 hours ago, Reputator said:

IIRC an unstable Infinity Fabric clock can manifest as memory instability.

The FCLK has been running at the default clock set by the XMP/DOCP profile for years afaik. It also wouldn’t work with it set to “default” settings instead of XMP/DOCP, and would only work with the GSkill RAM installed (which we’re assuming is single rank vs the dual rank Corsair). It could certainly be that the Infinity Fabric became unstable over the years as well, though!

[Spork3245]

Trying to fill out a warranty form and the part number for my 5900x isn't showing up. I googled and it looks like the PN on mine is for an OEM CPU... but, I bought it from an Authorized Retailer and it came in retail packaging (not OEM packaging).  
I emailed AMD's warranty support to see what the deal is.

[Reputator]

4 hours ago, Spork3245 said:

The FCLK has been running at the default clock set by the XMP/DOCP profile for years afaik. It also wouldn’t work with it set to “default” settings instead of XMP/DOCP, and would only work with the GSkill RAM installed (which we’re assuming is single rank vs the dual rank Corsair). It could certainly be that the Infinity Fabric became unstable over the years as well, though!

 

Setting XMP/DOCP shouldn't do anything to your Fabric clock.

 

But yeah it looks like you already arrived at the same conclusion I probably would have, which is an issue with the CPU.

[Spork3245]

On 12/3/2023 at 11:48 AM, Spork3245 said:

Trying to fill out a warranty form and the part number for my 5900x isn't showing up. I googled and it looks like the PN on mine is for an OEM CPU... but, I bought it from an Authorized Retailer and it came in retail packaging (not OEM packaging).  
I emailed AMD's warranty support to see what the deal is.

Me: *explains the issue noting every bit of troubleshooting attempted and explains how using a different 5900x fixed the issue so the most likely culprit is the CPU being faulty*, *also mentions that I’m unable to fill out for an RMA request because the part number on my CPU does not show up on their required selection to submit one, despite my CPU being bought by an authorized seller and coming in retail packaging and includes pictures of the CPU + receipt*
 

AMD Support: I understand you’re having trouble booting your PC. If you’re not getting a video signal please check your graphics card. If you’re still having issues, please refer to our troubleshooting guide.

 

Me: Thanks, I have already tried everything in there and then-some. Unfortunately, the issue persists unless I use a different CPU, which seems to indicate that the issue is with the CPU.  I’m concerned that the part number for my CPU seems to be for an OEM model even though it was purchased retail and came in retail packaging as AMD.com is not allowing me to fill out for a warranty claim with this part number.

 

AMD: Please try the CPU in a different system, if the issue persists fill out our warranty claim form here.

 

Me: *tries calling AMD’s tech support instead of emails*

 

AMD phone: (automated message when attempting to direct to CPU warranty claim) All warranty claims must be filled out online via our warranty claim form located at AMD.com/warranty

 

Me: 

[Spork3245]

On 12/5/2023 at 2:38 AM, Spork3245 said:

Me: *explains the issue noting every bit of troubleshooting attempted and explains how using a different 5900x fixed the issue so the most likely culprit is the CPU being faulty*, *also mentions that I’m unable to fill out for an RMA request because the part number on my CPU does not show up on their required selection to submit one, despite my CPU being bought by an authorized seller and coming in retail packaging and includes pictures of the CPU + receipt*
 

AMD Support: I understand you’re having trouble booting your PC. If you’re not getting a video signal please check your graphics card. If you’re still having issues, please refer to our troubleshooting guide.

 

Me: Thanks, I have already tried everything in there and then-some. Unfortunately, the issue persists unless I use a different CPU, which seems to indicate that the issue is with the CPU.  I’m concerned that the part number for my CPU seems to be for an OEM model even though it was purchased retail and came in retail packaging as AMD.com is not allowing me to fill out for a warranty claim with this part number.

 

AMD: Please try the CPU in a different system, if the issue persists fill out our warranty claim form here.

 

Me: *tries calling AMD’s tech support instead of emails*

 

AMD phone: (automated message when attempting to direct to CPU warranty claim) All warranty claims must be filled out online via our warranty claim form located at AMD.com/warranty

 

Me: 

Called tech support again and just kept going through prompts until I could speak with someone, the guy I spoke with gave me a part number to use. He said that the part number shouldn’t matter for an RMA and explained that sometimes CPUs pressed for OEM get moved to be sold at retail instead when there are shortages (like in 2021 when I bought this).

So, I filled out the warranty claim form, referenced the case number from the previous emails in my description along with my full specs and a description of the issue, as well as a picture of the receipt and pictures of the CPU which were required. I received an obviously copy and pasted email hours later asking for a picture of the invoice/receipt, picture of the CPU, my name/address/phone, a description of the issue, my PC’s specs, which product I have, and it’s serial number… literally all the information that their form requires from you to attempt warranty services. 
I replied with new pics of the CPU + invoice, all questions/details answered, etc - asked if there is a way to expedite this process as I use my PC for work, even if I have to have a hold put on a credit card that’s released once they receive the faulty processor and also offering to pay the difference for another CPU should they be out of my current. They replied at 2am…WITH THE SAME COPY & PASTED EMAIL, literally asking for the same information I just sent them yesterday, nor did they answer or even acknowledge anything I asked in my reply.

This is literally the worst CS and RMA experience I’ve ever had. Why in the fuck is this email only and why is their warranty CS obviously over seas just copy and pasting emails? ASUS’ RMA was near instant and a simple 10-15 minute phone call, EVGA was two-three emails over a day or two, Corsair for fan failures took two days of emails on both occasions, Cooler Master for a failed RGB controller in my case was two days on Reddit, Lianli was one message to CS who told me my RGB strimmer was out of warranty but are going to send me a new RGB part anyway, PNY was a 30-45 minute phone call as they had me trouble shoot for a bit but then set up an RMA. Like holy fuck: AMD is bigger than ANY of these fucking companies and have the most shit-tastic warranty department ever. It’s like they purposely have a bad warranty experience to try to get people to just give up and but new parts.

 

Seriously, I might never buy an AMD product ever again after this shit.

 

[Nokra]

4 hours ago, Spork3245 said:

Called tech support again and just kept going through prompts until I could speak with someone, the guy I spoke with gave me a part number to use. He said that the part number shouldn’t matter for an RMA and explained that sometimes CPUs pressed for OEM get moved to be sold at retail instead when there are shortages (like in 2021 when I bought this).

So, I filled out the warranty claim form, referenced the case number from the previous emails in my description along with my full specs and a description of the issue, as well as a picture of the receipt and pictures of the CPU which were required. I received an obviously copy and pasted email hours later asking for a picture of the invoice/receipt, picture of the CPU, my name/address/phone, a description of the issue, my PC’s specs, which product I have, and it’s serial number… literally all the information that their form requires from you to attempt warranty services. 
I replied with new pics of the CPU + invoice, all questions/details answered, etc - asked if there is a way to expedite this process as I use my PC for work, even if I have to have a hold put on a credit card that’s released once they receive the faulty processor and also offering to pay the difference for another CPU should they be out of my current. They replied at 2am…WITH THE SAME COPY & PASTED EMAIL, literally asking for the same information I just sent them yesterday, nor did they answer or even acknowledge anything I asked in my reply.

This is literally the worst CS and RMA experience I’ve ever had. Why in the fuck is this email only and why is their warranty CS obviously over seas just copy and pasting emails? ASUS’ RMA was near instant and a simple 10-15 minute phone call, EVGA was two-three emails over a day or two, Corsair for fan failures took two days of emails on both occasions, Cooler Master for a failed RGB controller in my case was two days on Reddit, Lianli was one message to CS who told me my RGB strimmer was out of warranty but are going to send me a new RGB part anyway, PNY was a 30-45 minute phone call as they had me trouble shoot for a bit but then set up an RMA. Like holy fuck: AMD is bigger than ANY of these fucking companies and have the most shit-tastic warranty department ever. It’s like they purposely have a bad warranty experience to try to get people to just give up and but new parts.

 

Seriously, I might never buy an AMD product ever again after this shit.

 

That's super annoying, sorry man. Not to go all Karen on them, but could you explicitly ask to speak to a supervisor? Sounds like they're just passing you around to lower-level people who don't know what they're doing.  

[Spork3245]

1 hour ago, Nokra said:

That's super annoying, sorry man. Not to go all Karen on them, but could you explicitly ask to speak to a supervisor? Sounds like they're just passing you around to lower-level people who don't know what they're doing.  

 

I'm not even getting passed around - they're just copy and pasting the same email (it's the same person's signature) and AMD doesn't do warranty/RMA over the phone so I can't even speak with anyone about it. When I tried asking a phone tech support guy, they said they don't even have access to the RMA system to check status or look up a service request number. 

Like how can AMD have the worst CS ever? It's incredulous to me! 

[SuperSpreader]

If you're getting the spinning windows dots it makes me think it's a Windows or hard drive issue. 

 

Also yeah AMD sucks.

[Spork3245]

1 minute ago, SuperSpreader said:

If you're getting the spinning windows dots it makes me think it's a Windows or hard drive issue. 

It’s not  

[SuperSpreader]

48 minutes ago, Spork3245 said:

It’s not  

 

[Spork3245]

1 hour ago, SuperSpreader said:

 

If I remove all current hdd/ssd/m.2 and try a new + blank drive Windows Installer does the same thing.

Different CPU stopped the issue.

[SuperSpreader]

16 minutes ago, Spork3245 said:

If I remove all current hdd/ssd/m.2 and try a new + blank drive Windows Installer does the same thing.

Different CPU stopped the issue.