Overclock.net - An Overclocking Community

Overclock.net - An Overclocking Community (https://www.overclock.net/forum/)
-   AMD CPUs (https://www.overclock.net/forum/10-amd-cpus/)
-   -   CPU related issue or GPU. (https://www.overclock.net/forum/10-amd-cpus/1734972-cpu-related-issue-gpu.html)

Alastair 10-21-2019 12:26 AM

CPU related issue or GPU.
 
So guys I need some help trouble shooting a problem I am experiencing. I recently delidded my FX-8370. Yes I delidded a soldered cpu. I did it as I wanted to learn how to do it and I could learn on a CPU that because of its age isn't worth very much any more. Long story short. Everything seemed to go smoothly. I delidded the cpu without damaging any of the little caps and resistors that surround the die. I applied Liquid metal and mounted my Raystorm block directly to the die. System booted first time. And passed stress tests at stock clocks with flying colours. CPU package temps reaching 2C above ambient. All ram shows up and is in dual channel. So all looks good.

So awesome. Time to get the overclock dialled in. I crank the cpu to 5050MHz (266x19) @ 1.5375 with CPU/NB at 2660MHz 1.3875V and HTT at 2660MHz with my memory at 2127MHz 9-10-9-27 1T. Passes my quick dirty X10 very high IBT run. And passes longer runs of realbench overnight for about 14hours.

Satisfied that this should be good enough to game and bench on I decided to load up Timespy to see if I could break 4K physics with my cpu at this clockspeed. I have the advanced Ed so I can skip the demo. Go straight into GT1 and it finishes loading aaaaaannnnnnnndddd black screen. Complete hardlock with a Blackscreen. So I think to myself ok. That's odd. We will try again. As I power down and reset the pc. As it logs back into Windows I get a display driver stopped responding error. But whatever I try again and same result. Long story short this black screen of death is plagued me for about a week and I have been tearing my hear out trying to solve it. Here are the steps I have taken thus far.

DDU'ed and reinstalled drivers.
Rolled back to older drivers. Tried 19.10.1, 19.9.2 and 19.5.2
Lowered cpu overclock to a very safe overclock of 4.5Ghz
Reinstalled windows.
Run optimised defaults in bios (passes)
Run an insanely low voltage for my OC (passed once)
Run 2x memory sticks. Swapped them around reseated.
Swapped my Sata cables.
Run stock speeds at elevated Vcore (crashes)

So interestingly enough I found a pattern emerge. Cpu overclocks were crashing the system. Even though I am pretty sure we are stable. So cpu that's fine. It passes stock clocks which is good to know as well. But the interesting thing i have found is this.

Vcore seems to crash the system.

At stock clocks for everything, except setting the Vcore a bit higher at 1.35V. And the system Blackscreens in a 3d load and I have to reset the machine. So it's Vcore related? 4GHz stock clock at slightly higher voltage and the system falls on its face.

Previously before the delid I ran 5GHz on my cpu at 1.5Vcore.

I don't understand maybe I damaged something in the delid process. But visually everything looks good. Temps are good. For the most part.

Under Avx IBT
CPU at 5050MHz sits in the mid 60s on the package.
Motherboard cpu socket gets hot (mid 80s) but no throttling. I have upgraded the VRM cooling on my motherboard so it throttles later when it comes to socket Temps.

Stock Temps are 2c above ambient for core temps.
So yeah. So rig specs are in sig.. But some things have changed.
So

FX 8370 at 5050MHz 1. 537v / 2660MHz [email protected]/ 2660 HTT. (safe clocks 4.5GHz 1.35V) (elevated Vcore stock clocks 4Ghz @1.35V )
Asus M5A99FX Pro r2.0 with VRM fan. Revised VRM heatsink mounting mechanism, back side socket fan and heatsinks on the driver ICs on the back of the board.
Corsair Dominator plat 32GB 4x8GB 2666MHz @ 2127MHz 9-10-9-27
AMD Vega 64 @ stock.
Antec HCP-1300
Full custom loop with total of 640x45mm rad space. Cpu direct die with liquid metal to raystorm cpu block. Vega 64 with full cover Barrow waterblock.

At this I can only imagine I screwed something up on my cpu in the delid process. But I don't know what it could be as there were no damaged caps or resistors. The die looked pretty good with some minor surface scratches after removing the remaining solder.

BroadPwns 10-21-2019 02:55 AM

I've never OCd Bulldozer but check if your PCI frequency doesn't rise up.

Alastair 10-21-2019 03:38 AM

pcie is not coupled to Baseclock.

Alastair 10-21-2019 06:14 AM

I dunno. Vcore seems to trigger it. Even at stock clocks raising vcore above 1.35v triggers it. I'm going to switch from my benching OS on HDD to my gaming OS on my SSD and see if the issue persists there.

Snaky115 10-21-2019 06:50 AM

Comedic but true story from a friend with reference Vega64 - His V64, at stock clocks, needs to run -10% power limit in wattman to not crash in 3D loads. (I still think it is his Gigabyte gold 750w PSU but who knows, it works)
Just some almost related food for thought since you have a Vega. Maybe throw in a different videocard?

But this is pretty bizarre, that voltage alone can cause crashes. I can assume you had it pretty well overclocked even before delid? **ponders with his coffee mug**

Alastair 10-21-2019 08:20 AM

Quote:

Originally Posted by Snaky115 (Post 28166240)
Comedic but true story from a friend with reference Vega64 - His V64, at stock clocks, needs to run -10% power limit in wattman to not crash in 3D loads. (I still think it is his Gigabyte gold 750w PSU but who knows, it works)
Just some almost related food for thought since you have a Vega. Maybe throw in a different videocard?

But this is pretty bizarre, that voltage alone can cause crashes. I can assume you had it pretty well overclocked even before delid? **ponders with his coffee mug**

It ran pretty well before delid. 4.95GHz in summer months and 5GHz in winter months. Both around about 1.5ish vcore.

I decided to delid to 1. Learn how to do it firstly. Might as well learn with something that isn't expensive and new for fear of breaking it. Also might as well lean how to do it on a hard cpu that's soldered because then other cpus are probably then a bit easier.

And 2 just to eek a little more out of the FX so I can have some benching fun.

But yeah. Vcore seems to trigger it. Even at stock clocks. Psu rails at least according to software appear to be at the right voltages. It's a Delta made Antec HCP1300 so I can't see it being the issue. It's not very old either. I'll try a bit more power on the V64 but I mean it was doing fine before the delid.

cssorkinman 10-21-2019 09:47 AM

Takes a real enthusiast to delid an FX . :)

Is there a setting for NB PCI-E voltage on your motherboard bios ( or overclocking utility you might use)?

Seems like a power issue but the fact that it seems to involve both the cpu and gpu is what led me to that particular setting. I have to bump that voltage a bit on my 990 GD 80 in order to run certain gpu setups properly. I think I had to fiddle with it on my FX 790 GD 70 as well when I had quadfire 4870's -geez that was a lifetime ago....lol.

Alastair 10-21-2019 12:37 PM

1 Attachment(s)
Quote:

Originally Posted by cssorkinman (Post 28166378)
Takes a real enthusiast to delid an FX . :)

Is there a setting for NB PCI-E voltage on your motherboard bios ( or overclocking utility you might use)?

Seems like a power issue but the fact that it seems to involve both the cpu and gpu is what led me to that particular setting. I have to bump that voltage a bit on my 990 GD 80 in order to run certain gpu setups properly. I think I had to fiddle with it on my FX 790 GD 70 as well when I had quadfire 4870's -geez that was a lifetime ago....lol.

Yes and as you can see from the picture it went according to plan. I cound 4 naked pads for caps or transistors but from other pictures I can find online those are unpopulated and I saw no loose pieces after the lid came off.



I am using an asus board so I dont know. I have:
NB voltage: so I am pretty sure thats voltage direct to the 990FX chipset.
VDDA: which if I remember is an auxiliary voltage to the CPU to support high base clocks.

NB HT: which from what I know is HyperTransport voltage.

NB 1.8V voltage: No idea what this is.
and SB voltage which is for the SB chipset. SB750?

Alastair 10-21-2019 12:40 PM

I have my Vega 64 on a vertical GPU mount with a Riser when the Furys used to be direct in the board. think I need more PCIE volts cause of this?

cssorkinman 10-21-2019 01:01 PM

Quote:

Originally Posted by Alastair (Post 28166562)
I have my Vega 64 on a vertical GPU mount with a Riser when the Furys used to be direct in the board. think I need more PCIE volts cause of this?


That's what I keep coming back to in my mind.

Edit: Beautiful job with that de-lid man!


All times are GMT -7. The time now is 01:17 PM.

Powered by vBulletin® Copyright ©2000 - 2019, Jelsoft Enterprises Ltd.

User Alert System provided by Advanced User Tagging (Pro) - vBulletin Mods & Addons Copyright © 2019 DragonByte Technologies Ltd.
vBulletin Security provided by vBSecurity (Pro) - vBulletin Mods & Addons Copyright © 2019 DragonByte Technologies Ltd.

vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2019 DragonByte Technologies Ltd.