3700x big issues

[opethdisciple]

OK, so some of you know I've been complaining since Jan 2021 that my ram overclock of 3600MHz is now unstable.

As a bit of background for some of you reading this for the first time I ran my 8 pack 3200MHz ram at 3600MHz 14,15,15,30 @ 1.47v with no issues for about a year. I installed my RTX 3080 in Jan and all went pear shaped since then.

The conclusion eventually was heat. The heat from the RTX 3080 was pushing the ram over the edge of stability.

Fast forward to now and for the last couple of months I have struggled to run my ram at at any timings at 3600MHz. My ram is apparently officially rated to run at 3600MHz 16,16,16,36 @ 1.35v according to the product page. Even that proves unstable. I tried adding SOC and VDDG voltage but it doesn't make a difference.

Also the last two weeks something else has started to happen. My BIOS is really laggy.

Keyboard presses take a while to register there are pauses and it will also just continue to scroll as if I am holding down an arrow key.

Resetting the BIOS solves it for a while.

I had a theory today that perhaps the SOC or FCLK on the CPU is faulty. I set the FCLK to 1800 and left everything else on auto and I had some weirdness in windows with the UI. Hard to explain.

I know the FCLK and memory are meant to be able to run uncoupled but I thought maybe it might help if I made the ram run 1:1 with the FCLK so I applied the 3600MHz at 1.35v with loose timings.

Reboot and bios is laggy again, but Windows seems fine. Of course it's only temporary stability as I know I can no longer run at 3600MHz what ever I try.

On the subject of laggy BIOS I thought it was a BIOS issue so I was on the last stable bios form Jan for my board out of desperation to fix the lag I upgraded to the last beta. This hasn't solved my issue.

The saving grace here is that things seem OK, if I apply the DOCP profile of 3200MHz. I gamed earlier for about 2.5hrs with no issues.

Cant remember if BIOS is laggy though. I think it still is.

Yesterday I retested my ram at the CL14 3600MHz dram cal fast timings and it passes 400%. But I cannot game on it for long until it fails.

I am thinking now to just move the 3700x and ram to my workstation and get a 5600x and new ram.

I've never had a CPU go on me before. I assume I still have warranty as well on the CPU. But is this even covered by warranty? Because an FCLK of 1800 is not even a guarantee.

 

[darket]

OK, so some of you know I've been complaining since Jan 2021 that my ram overclock of 3600MHz is now unstable.

As a bit of background for some of you reading this for the first time I ran my 8 pack 3200MHz ram at 3600MHz 14,15,15,30 @ 1.47v with no issues for about a year. I installed my RTX 3080 in Jan and all went pear shaped since then.

Here's my very unprofessional take on this.

You Upgraded the Gpu, Your system is now being pushed harder and probably showing the cracks in an already weak IMC, as in anything over the official supported memory speed of 3200 and things start getting flaky.

Now I would do two things, one is to get my hands on some different 3600mhz memory to test, passing stability tests means nothing if the system is still unstable, the other is to make sure everything is stock and try again from there, if its fine at 3200mhz its either the memory or a weak imc on the chip.

As for the laggy bios EVERY gigabyte board I have used seems to have this issue, Asus is a little better but it seems adding a nice shiny gui created all sorts of resource issues in modern bios's , I'd be just as happy with the old txt only bios's as they didn't lag.

 

[DoneADougalOnSofa]

My 2700X suffered a similar fate. It was average at best when new in RAM terms, 3200c14 was doable, but not via DOCP, had to be manually tuned.
That started crashing, so I wound it down to 2933, same timings. All good for a couple of months... Then that started crashing too. As it stands, it is now unstable at anything other than stock 2400.

At some point following that, I thought 'sod it', and bought a 5800X.

 

[tommoT.Striker]

When I had a 1700x and 2700x they could only do 2933 and 3200 respectively. Never via DOCP all had to manually tuned. Only the 3900X i had for a while went up to 3600 stable as a table.

 

[opethdisciple]

Here's my very unprofessional take on this.

You Upgraded the Gpu, Your system is now being pushed harder and probably showing the cracks in an already weak IMC, as in anything over the official supported memory speed of 3200 and things start getting flaky.

Now I would do two things, one is to get my hands on some different 3600mhz memory to test, passing stability tests means nothing if the system is still unstable, the other is to make sure everything is stock and try again from there, if its fine at 3200mhz its either the memory or a weak imc on the chip.

As for the laggy bios EVERY gigabyte board I have used seems to have this issue, Asus is a little better but it seems adding a nice shiny gui created all sorts of resource issues in modern bios's , I'd be just as happy with the old txt only bios's as they didn't lag.

Ok. Thanks for letting me know that the laggy bios isnt symptomatic of a component failure.

Still very odd though.

As for putting everything to stock. Last night I gamed at 3200MHz on the ram using docp for about 2.5hrs and it was fine.

That's when I decided to see if it was an fclk issue and upped the fclk to 1800 both leaving the docp 3200mhz in place and with out and it was not having any of it at all. It booted fine but major graphics corruption in Windows and freezing.

I've tried applying volts to soc and vddg bit that doesn't help.

 

[opethdisciple]

Ok I been thinking on this.

If my ram passed 500% HCI and I ran it for a year at CL14 3600Mhz the fact it became unstable once I installed the RTX 3080 is it safe for me to deduce it's the IMC on the CPU that's the issue?

I'm thinking of picking up a 5600x but keeping my ram which I know is a good clocker.

 

[Troezar]

Ok I been thinking on this.

If my ram passed 500% HCI and I ran it for a year at CL14 3600Mhz the fact it became unstable once I installed the RTX 3080 is it safe for me to deduce it's the IMC on the CPU that's the issue?

I'm thinking of picking up a 5600x but keeping my ram which I know is a good clocker.

The only real way to tell is to swap out the CPU and see if you still get the issue. Not a great option given the cost, but if it is ok you can still keep the 5600X and sell the 3700X to offset the cost.

 

[Destination]

Try reseating everything especially the cpu, in case a poor contact is overheating the imc.

Any chance you shorted the memory when placing the gpu?

 

[opethdisciple]

I suppose the next question is to ask is running my ram at 3200MHz losing me much performance with my RTX 3080?

I have it in my head the minimum acceptable ram speed is 3600MHz on Ryzen 3000 and up.

Upgrading to a 5600x would probably only cost me £100 after I sell the 3700x and hopefully (if the fclk gods smile on me) I might even be able to run my ram as high as 3800MHz.

 

[nlel1975]

I suppose the next question is to ask is running my ram at 3200MHz losing me much performance with my RTX 3080?

Not much at all, at 1080p more than 4K but there’s not a lot of difference.

 

[CAT-THE-FIFTH]

If installing an RTX3080 is causing issues its unlikely it's the CPU. Those memory tests are going to stress your CPU IMC far more than a random game IMHO.

It's more likely it's an issue with heat or a PSU problem. Remember,a RAM test is unlikely to cause as much heat production,or spike power consumption as gaming using the RTX3080.

Remember,also RAM and IMCs can degrade if you are not careful with voltage - so realistically you need to use minimal voltage increases on both,and again its not to say a Zen3 CPU won't have the same issues longtime also. Remember,they have been out for less than a year.

Both Zen2 and Zen3 use the same memory controller which is the same 12NM//14NM I/O chip,and you could hit the same issue down the line.

Once more games start using RT,and move over to the new generation of consoles as a primary development platform,the Ryzen 5 CPUs are probably going to start increasingly falling behind the Ryzen 7 CPUs. Then you might ditch the Ryzen 5 5600X,and then spend more on a new CPU. All for what?? Around £100 more for a Ryzen 7 5800X?? So I wouldn't go from a Ryzen 7 3700X to a Ryzen 5 5600X.

Moreover,the current Zen3 CPUs are being replaced by the B2 stepping ones too,and as the current ones are B0,its telling me the current CPUs might have some hidden problems,for AMD to literally jump two steppings. Usually its a one stepping jump - the B1 stepping was canned,for AMD to never release it. That is the thing though - as much as there are people with no Zen3 issues,you do also hear people have their share of problems too.

 

[Game]

motherboard issue over cpu ? Going by my own problems in the past its been the culprit more often

 

[opethdisciple]

If installing an RTX3080 is causing issues its unlikely it's the CPU. Those memory tests are going to stress your CPU IMC far more than a random game IMHO.

It's more likely it's an issue with heat or a PSU problem. Remember,a RAM test is unlikely to cause as much heat production,or spike power consumption as gaming using the RTX3080.

Remember,also RAM and IMCs can degrade if you are not careful with voltage - so realistically you need to use minimal voltage increases on both,and again its not to say a Zen3 CPU won't have the same issues longtime also. Remember,they have been out for less than a year.

Both Zen2 and Zen3 use the same memory controller which is the same 12NM//14NM I/O chip,and you could hit the same issue down the line.

Once more games start using RT,and move over to the new generation of consoles as a primary development platform,the Ryzen 5 CPUs are probably going to start increasingly falling behind the Ryzen 7 CPUs. Then you might ditch the Ryzen 5 5600X,and then spend more on a new CPU. All for what?? Around £100 more for a Ryzen 7 5800X?? So I wouldn't go from a Ryzen 7 3700X to a Ryzen 5 5600X.

Moreover,the current Zen3 CPUs are being replaced by the B2 stepping ones too,and as the current ones are B0,its telling me the current CPUs might have some hidden problems,for AMD to literally jump two steppings. Usually its a one stepping jump - the B1 stepping was canned,for AMD to never release it. That is the thing though - as much as there are people with no Zen3 issues,you do also hear people have their share of problems too.

Good points. But what made me suspect my IMC is two things:

1. My ram is officially meant to come in 3 configurations. 3200MHz, 3600MHz and a 3800MHz. You can find all this on the website.

It should run at 3600MHz using the CL16 base timings at 1.35v with no issues.

At the moment it cant.

2. To test if it was FLCK issue I set the FCLK to 1800 and left the ram at 3200MHz docp. Having got in to windows I experienced major graphical corruption. Windows went brown!

Cntl alt delete to bring up task manager though looked normal but windows and the rest of the shell was brown!

So to me this rules out a PSU heat issue. This was with no gaming at all. Just on windows.

Not only was I able to run my ram at 3600MHz 14,15,15,30 prior to the 3080, for a while I actually ran it faster 3600MHz 14,15,14,28.

Using both of these configurations I was completely fine for at least one year.

 

[opethdisciple]

Perhaps I just need to push the SOC voltage?

The internet says up to 1.2v is fine. But I thought PCIe-4 devices don't play well at those voltages?

 

[Joxeon]

Since you game at 1440P I doubt it's worth the hassle of trying to run the ram any faster than the DOCP profile if you're having problems since the difference is only likely to be 1 or 2 fps and sacrificing stability for that isn't worth it.

 

[DoneADougalOnSofa]

Personally, I prefer to run tighter timings at 3200 than looser at 3600 with my 5800X.
Negligible difference in anything, and I prefer a little less strain on everything. The two settings are very close, performance-wise in my experience.

 

[opethdisciple]

Pushing the SOC voltage seems to have had a positive effect.

I only applied the primary timings 14,15,15,30 @ 3600MHz using 1.47v with an SOC of 1.15v and it allowed me not only to run the benchmark in Valhalla (before it was crashing) but I also gamed on it for about an hour and a half.

To early to tell of course if this is going to be stable long term.

But it's encouraging.

 

[CAT-THE-FIFTH]

Good points. But what made me suspect my IMC is two things:

1. My ram is officially meant to come in 3 configurations. 3200MHz, 3600MHz and a 3800MHz. You can find all this on the website.

It should run at 3600MHz using the CL16 base timings at 1.35v with no issues.

At the moment it cant.

2. To test if it was FLCK issue I set the FCLK to 1800 and left the ram at 3200MHz docp. Having got in to windows I experienced major graphical corruption. Windows went brown!

Cntl alt delete to bring up task manager though looked normal but windows and the rest of the shell was brown!

So to me this rules out a PSU heat issue. This was with no gaming at all. Just on windows.

Not only was I able to run my ram at 3600MHz 14,15,15,30 prior to the 3080, for a while I actually ran it faster 3600MHz 14,15,14,28.

Using both of these configurations I was completely fine for at least one year.

It seems really weird for the 12NM/14NM I/O die to suddenly have issues like this after a year. So I wonder if the motherboard is doing something funky with voltages to cause electromigration issues.

 

[opethdisciple]

It seems really weird for the 12NM/14NM I/O die to suddenly have issues like this after a year. So I wonder if the motherboard is doing something funky with voltages to cause electromigration issues.

My theory is that the GPU is putting more demands on the CPU including the memory controller.

Hence why I thought maybe I will try bumping the soc voltage.

In one game Valhalla it seemed to make a difference. But one game for one and a half hours isn't a great test.

The point is I know my ram is good for CL14 3600 which is why I got the idea that it was something else that was the problem.

Heat... Well it could be. But I am not convinced. Because all the reviewers said the 3000 FEs don't affect ram temps and I have a decent case for airflow.

 

[Journey]

Have you actually considered it could be the motherboard that is causing the issue, bad power stability from the CPU VRM (or memory VRM)? A laggy BIOS is odd behaviour, but certainly doesn't identify the CPU as the cluprit.

I'd RMA the CPU to AMD, they foot the bill for shipping and collect via DHL from your door, and the TTAT is about 1 week. Once you've got the new CPU installed that will rule out the CPU if you are still seeing the same faults, but I'd also prepare to be testing your motherboard, RAM and even PSU.