AMD R5 3600 dying again?

UkayJ · 10 Sep 2021 at 18:55

Hello, I just got a new cpu on warranty and after 4 days it's done the same thing that finished off the other one : computer freezes during gaming and both of my screens turn off so I had to hard reset it.

After a quick search : could it be the PSU? It's an XPG CORE Reactor 750W gold rated. It's only half a year old. I had no problems with the Corsair 600w gold rated that was in it for probably 3 years at least (but I put that in a second pc I was building for us).

Motherboard is an MSI B450 Tomahawk (the old one that I had to update the BIOS to accept the 3600 cpu).

pc is watercooled so cpu doesn't go over 65*C

I had it slightly undervolted via offset on the old broke one.

It's now fixed under 1.3volts (still going down 1 step at a time to see how low it will go) @ 4.1GHz.

Joxeon · 10 Sep 2021 at 19:07

Could always be the ram or GPU?

DoneADougalOnSofa · 10 Sep 2021 at 19:15

Why the undervolt? If it's crashing under load, my first guess is not enough power. I'd whack the CPU voltage back to Auto, and probably dial in some load-line calibration too.
What GPU are you using, and if applicable, are you using two separate cables from the PSU, as opposed to one with a splitter?

Cyber-Mav · 10 Sep 2021 at 19:15

Insted of messing with voltages run it at stock and see if its stable. Ryzen does have high failure rate so you may need to rma again if its not the ram causing the issue.

UkayJ · 10 Sep 2021 at 22:18

It must be the cpu and not the ram or gpu as I put my old cpu in second pc to test it and it wouldn't boot, also AMD confirmed it to be dead.

I'm undervolting to reduce heat and I wondered if the motherboard was feeding it too many volts and that's why the old one died. It reaches over 1.4v and I'm sure I read the safe limit is 1.35ish.

gpu is amd rx480 with a full cover waterblock on, its got one cable only.

I don't know what LLC is, I thought it was Intel only and never having owned an Intel cpu I never learnt more about it.

If it freezes again I'll change the undervolt back to auto and see how it gets on. The old negative offset undervolt worked for a long time though (since before the new psu).

DoneADougalOnSofa · 10 Sep 2021 at 22:28

LLC is on AMD too, it'll be somewhere in amongst the power settings in the BIOS. It gives a little extra voltage under heavy loads, to combat 'vdroop', which I reckon is what you've got going on...
There's often a sub-menu with stuff like power phases, it's usually in there.
I run a medium setting most of the time, my 2700X used to crash out without it, and my 5800X has done it once when I first got it. Worth a shot, maybe...

Joxeon · 10 Sep 2021 at 23:31

Should just set vcore and multi to auto as these CPUs are not like Intel and are designed to run up to 1.5v for light loads.

UkayJ · 11 Sep 2021 at 15:08

DoneADougalOnSofa said:
LLC is on AMD too, it'll be somewhere in amongst the power settings in the BIOS. It gives a little extra voltage under heavy loads, to combat 'vdroop', which I reckon is what you've got going on...
There's often a sub-menu with stuff like power phases, it's usually in there.
I run a medium setting most of the time, my 2700X used to crash out without it, and my 5800X has done it once when I first got it. Worth a shot, maybe...

Great explanation thanks, I'll have a look at that

Joxeon said:
Should just set vcore and multi to auto as these CPUs are not like Intel and are designed to run up to 1.5v for light loads.

As I said, I had read that 1.35v was the safe limit which is why I wanted to undervolt in the first place
but I'm going to try this LLC magic.

UkayJ · 11 Sep 2021 at 15:16

For LLC I have "Auto" and Modes 1-8, I presume 8 would give the greatest increase in voltage under load so I've set it to mode 2 for now.

DoneADougalOnSofa · 11 Sep 2021 at 15:34

2 or 3 should be about right. Personally, I tend to set SoC LLC one notch below the CPU, so for example level 3 on CPU, level 2 on SoC.
That's just me though, not sure it's really necessary!

Joxeon · 11 Sep 2021 at 18:33

UkayJ said:
Great explanation thanks, I'll have a look at that

As I said, I had read that 1.35v was the safe limit which is why I wanted to undervolt in the first place
but I'm going to try this LLC magic.

The lower limits only apply for manual vcore, auto setting vcore is fine and the limit for that is 1.5v

jigger · 12 Sep 2021 at 14:44

Over voltage and over amperage can kill electronics. Over amperage is by far the most likely killer as you have no control over a CPU’s current draw and it has a nasty relationship with heat.

Not plain and simple at all. Just more Cyber-Mav misinformation.

UkayJ · 12 Sep 2021 at 22:26

So I set the CPU back to Auto, it's crashed twice I think already (screens freeze then go off so I have to hard reset).

The only other thing I've done in bios is set RAM to 3600MHz (what it's rated for) and Infinity Fabric clock? to 1800MHz.

It can't be the CPU OC/undervolt or anything else that's causing this?! XPG Core Reactor PSU?

Joxeon · 12 Sep 2021 at 22:34

UkayJ said:
So I set the CPU back to Auto, it's crashed twice I think already (screens freeze then go off so I have to hard reset).

The only other thing I've done in bios is set RAM to 3600MHz (what it's rated for) and Infinity Fabric clock? to 1800MHz.

It can't be the CPU OC/undervolt or anything else that's causing this?! XPG Core Reactor PSU?

Did you set the vcore back to auto also?

Cyber-Mav · 12 Sep 2021 at 22:35

UkayJ said:
So I set the CPU back to Auto, it's crashed twice I think already (screens freeze then go off so I have to hard reset).

The only other thing I've done in bios is set RAM to 3600MHz (what it's rated for) and Infinity Fabric clock? to 1800MHz.

It can't be the CPU OC/undervolt or anything else that's causing this?! XPG Core Reactor PSU?

possible imc on cpu not able to handle the ram speed. set to lower speed like 2666 and test form there. will help to isolate where the issue lies.

heatonpkmassive · 12 Sep 2021 at 22:46

Underclocking your RAM to a lower speed to test is a good shout, as is checking what actual voltage your memory is running at. One or two boards sometimes need a little extra juice there.

Have you looked into software logs? You may well have done but I see no mention of it here.

UkayJ · 13 Sep 2021 at 07:12

Yes vcore back to auto.

I'll lower the ram speed, should I keep the Infinity Fabric(?) Clock at half the ram speed as well?

Software logs-wise I think I tried looking but wouldn't know what to look for and because it's just freezing unexpectedly I can't have a quick look at Hardware Info for example.

Hardware Info otherwise seems OK temp wise etc.

Cyber-Mav · 13 Sep 2021 at 08:17

lower ram speed and keep fabric clock 1:1 for now and see how things are.

Trig · 13 Sep 2021 at 08:57

UkayJ said:
So I set the CPU back to Auto, it's crashed twice I think already (screens freeze then go off so I have to hard reset).

The only other thing I've done in bios is set RAM to 3600MHz (what it's rated for) and Infinity Fabric clock? to 1800MHz.

It can't be the CPU OC/undervolt or anything else that's causing this?! XPG Core Reactor PSU?

I had some RAM in an AM4/B450M board with a 1600AF, RAM was rated but anything other than AUTO wouldn't let me do much before it crapped out.
I suggest defaulting the BIOS, set the boot order and leave everything else.

Joxeon · 13 Sep 2021 at 11:15

Which bios version are you running currently?

UkayJ · 13 Sep 2021 at 20:02

Trig said:
I had some RAM in an AM4/B450M board with a 1600AF, RAM was rated but anything other than AUTO wouldn't let me do much before it crapped out.
I suggest defaulting the BIOS, set the boot order and leave everything else.

I'll try that now thank you.

I just had fabric clock (FCLK) on Auto and RAM @ 3200MHz, it seemed lag free but suddenly froze so I set it to 2666MHz which was quite laggy, only had one game though on Apex Legends.

Bios version : 7C02v1D - this was the one to use to get the 3xxx series working on older boards so I haven't updated BIOS again, should I?