Did I Degrade or Destroy my 5950x? Now with a response from Asus on page 2!

kamikazi

kamikazi

[H]ard|Gawd
Joined
Jan 19, 2006
Messages
1,636
I was following along in the PBO Curve Optimizer thread with the new version of Ryzen Master PBO Curve Optimizer Thread. I updated my BIOS to 4201 (which is beta) and my BIOS settings were bone stock except for RAM timings. I hadn't messed with any PBO settings, voltages, boost, or anything. Then, I got the latest AMD chipset drivers installed and I ran the new Ryzen Master per core curve optimizer tool. I let the computer reboot, then ran some benches. Cinebench R20 scored close to 12000. Also, Because of the way my loop is set up to ramp up the pumps and fans based upon the air-water delta, cpu temps were hit 90C while running Cinebench because the coolant is not warmed up in such a short time (this will be important later. I also got my highest score so far in 3DMark. Again, I made no other changes. GPU is stock as well.

On Tuesday, there was a brief power blip at home. My computer was turned off when I got home. I booted it back up and looked in BIOS and the curve optimizer settings from Ryzen Master were still in there. I decided to retest everything and now Cinebench R20 scores in the 10000s. I run again and it gets lower, down in the 9000s. I watched in HWMonitor and the cores are all running around 3850 MHz. Now, the cpu temp is not getting out of the 60s when running cinebench. The fans and the pumps stay low because the coolant isn't getting heated up. It's like the chip just doesn't want to boost, but is using more power while somehow staying cool. This doesn't make any sense to me. Since then, I've cleared CMOS, reflashed the BIOS, flashed the BIOS down to version 4006, the last non-beta, and I've rolled Windows back to a restore point two weeks ago prior to me starting to mess with all this. BIOS is stock with no curve settings, just XMP on the RAM. I ran 3DMark again and my CPU score dropped by about 1000. Cinebench still running in the 9000s. Nothing has fixed it. It still will not boost. I re-installed Ryzen Master and watched everything while running Cinebench. It appears that EDC is now the limiting factor. It's hitting 100 percent (140A) while the PPT is hitting around 132 watts and TDC is not maxing out. Anyone have any idea about what is going on and how to fix it? Or is it just toast?
 
kamikazi said:
I was following along in the PBO Curve Optimizer thread with the new version of Ryzen Master PBO Curve Optimizer Thread. I updated my BIOS to 4201 (which is beta) and my BIOS settings were bone stock except for RAM timings. I hadn't messed with any PBO settings, voltages, boost, or anything. Then, I got the latest AMD chipset drivers installed and I ran the new Ryzen Master per core curve optimizer tool. I let the computer reboot, then ran some benches. Cinebench R20 scored close to 12000. Also, Because of the way my loop is set up to ramp up the pumps and fans based upon the air-water delta, cpu temps were hit 90C while running Cinebench because the coolant is not warmed up in such a short time (this will be important later. I also got my highest score so far in 3DMark. Again, I made no other changes. GPU is stock as well.

On Tuesday, there was a brief power blip at home. My computer was turned off when I got home. I booted it back up and looked in BIOS and the curve optimizer settings from Ryzen Master were still in there. I decided to retest everything and now Cinebench R20 scores in the 10000s. I run again and it gets lower, down in the 9000s. I watched in HWMonitor and the cores are all running around 3850 MHz. Now, the cpu temp is not getting out of the 60s when running cinebench. The fans and the pumps stay low because the coolant isn't getting heated up. It's like the chip just doesn't want to boost, but is using more power while somehow staying cool. This doesn't make any sense to me. Since then, I've cleared CMOS, reflashed the BIOS, flashed the BIOS down to version 4006, the last non-beta, and I've rolled Windows back to a restore point two weeks ago prior to me starting to mess with all this. BIOS is stock with no curve settings, just XMP on the RAM. I ran 3DMark again and my CPU score dropped by about 1000. Cinebench still running in the 9000s. Nothing has fixed it. It still will not boost. I re-installed Ryzen Master and watched everything while running Cinebench. It appears that EDC is now the limiting factor. It's hitting 100 percent (140A) while the PPT is hitting around 132 watts and TDC is not maxing out. Anyone have any idea about what is going on and how to fix it? Or is it just toast?
Click to expand...
Did you microwave it?


/s
 
You could set edc to motherboard limit in bios since that's what's holding it back it seems.
 
mnewxcv said:
You could set edc to motherboard limit in bios since that's what's holding it back it seems.
Click to expand...
I thought of trying that. I'm trying to figure out the underlying cause. The numbers just don't jive, using more power to get lower performance without generating heat. It's as though something is being misreported internally.
 
Randall Stephens said:
Do you have another cpu to rule out the board?
Click to expand...
Technically yes, I have a 3900X in my server. However, I don't think I could swap out the CPU without draining my loop. That would be a full day minimum and I'll be dragged kicking a screaming as I finally have everything the way I want it. :D Prior to my most recent loop change, it was possible to get just the cpu block off and swap a cpu, but I don't think I can do it the way it is now, I'll have to look at that more closely.
 
kamikazi said:
I thought of trying that. I'm trying to figure out the underlying cause. The numbers just don't jive, using more power to get lower performance without generating heat. It's as though something is being misreported internally.
Click to expand...
how do you figure it is using more power than before? I never really liked ryzen master for this sort of thing. The 5000 series CPUs specifically are quite well optimized out of the box.
 
mnewxcv said:
how do you figure it is using more power than before? I never really liked ryzen master for this sort of thing. The 5000 series CPUs specifically are quite well optimized out of the box.
Click to expand...
Good question. I forgot to mention that I was watching it in HW Info at the same time to see the wattage on the CPU. It was showing close to 140 watts. So, I guess it's not using more power than before when it was boosting properly, it's using the same amount of power without boosting.
 
Sounds like a degraded CPU from overclocking too far. Probably time for a new CPU.
 
ZodaEX said:
Sounds like a degraded CPU from overclocking too far. Probably time for a new CPU.
Click to expand...
I'm beginning to think this is the case. How it got degraded is the question. I undervolted it via curve optimizer. I didn't activate PBO, all of those settings were stock. It worked great in very limited testing, then immediately crapped out after a reboot caused by a power outage. It actually got a cinebench score in the 6000s once I think. So strange. I did wreck a graphics card once by undervolting. I had an MSI Lightning Radeon 6970 that I slightly undervolted. It crashed under load and from that point on, it crashed under any 3d load. Perhaps not raising board limits with PBO while using curve optimizer caused the mobo to overcompensate with higher voltage?
 
Here's a screenshot taken while running CB20 Multi-core:

crappy chip.jpg


That nets a multi-core score of 9700 or so. Single Core will score 631. These are both with monitoring running, so I'm guessing they would be just a bit higher when running with nothing else in the background. Single-core is not far off, but multi-core is terrible.
 
did i miss where you reset everything to stock, ie a good bios clearing?
 
You could always pull it and see if you have some burned pins if you think the board over compensated.
 
pendragon1 said:
did i miss where you reset everything to stock, ie a good bios clearing?
Click to expand...
"Since then, I've cleared CMOS, reflashed the BIOS, flashed the BIOS down to version 4006, the last non-beta, and I've rolled Windows back to a restore point two weeks ago prior to me starting to mess with all this. BIOS is stock with no curve settings, just XMP on the RAM."
 
mnewxcv said:
"Since then, I've cleared CMOS, reflashed the BIOS, flashed the BIOS down to version 4006, the last non-beta, and I've rolled Windows back to a restore point two weeks ago prior to me starting to mess with all this. BIOS is stock with no curve settings, just XMP on the RAM."
Click to expand...
Did you pull the battery?
 
mnewxcv said:
"Since then, I've cleared CMOS, reflashed the BIOS, flashed the BIOS down to version 4006, the last non-beta, and I've rolled Windows back to a restore point two weeks ago prior to me starting to mess with all this. BIOS is stock with no curve settings, just XMP on the RAM."
Click to expand...
guess i did miss it.. but a do agree with the battery yank if you havent.
 
That took a little surgery, but it's out. I'm gonna wait an hour or two to put it back, does that sound right?
 
learners permit said:
Memory VDDIO and VTT read 0V. Your chipset driver and/or Ryzen master are borked. I'd remove and reinstall both and check RM again for proper representation of the registers.
Click to expand...
I believe those have read 0v for as long as I've used ryzen master with this board. But, that will be my next step, remove the chipset drivers and clean out ryzen master.
 
No dice. Still the same. I did run that 1smus clock tuner tool. Said my cpu is a silver sample. Hah.

I may tru upping the tdc, ppt, and edc and then try to add boost to get it back to stock performance I guess.
 
chameleoneel said:
I would suspect that your board got hurt from the power outage, before I would suspect the CPU.
Click to expand...
That makes sense. Someone else in the thread mentioned I should try a different cpu to see if I get the same results on the motherboard. The computer is connected to a power strip with surge suppressor and it wasn't tripped. It used to be on a nice Cyberpower UPS, but at full load, it was tripping the alarm on the unit as it just pulls to much power. I did get a GFCI outlet in the basement that tripped. Nothing else in the house did though. No breakers were flipped. It was odd.
 
kamikazi said:
Here's a screenshot taken while running CB20 Multi-core:

View attachment 470841

That nets a multi-core score of 9700 or so. Single Core will score 631. These are both with monitoring running, so I'm guessing they would be just a bit higher when running with nothing else in the background. Single-core is not far off, but multi-core is terrible.
Click to expand...
Your power settings for a 5950X on PBO are REALLY LOW. For comparison, I have posted mine below; I also manually set my temperature limit to 85C (vs. 90C). I sit at about 80C solid all the time when running simulations in my VMs, been doing it for close to a year with no issues or degradation.

EDC is the limiting current when under full load; 140A is the sweet spot for boosting balance, but 160A and higher has been fun in benches. Your TDC is extremely low at 95W, that's the limit on current determined by the thermals of your chip; I run at 200A perfectly fine and it looks like you are maxing that out at 95A. The highest you can go is between your motherboard and CPU, so up that if you can for sure.

PPT is your total package power under load allowed; you can safety raise this to your motherboards limits. I keep mine at 250W and never come close to hitting it... I'm under heavy load right now with this snapshot, and you can see I'm still not close.

1651860694846.png
 
III_Slyflyer_III said:
Your power settings for a 5950X on PBO are REALLY LOW. For comparison, I have posted mine below; I also manually set my temperature limit to 85C (vs. 90C). I sit at about 80C solid all the time when running simulations in my VMs, been doing it for close to a year with no issues or degradation.

EDC is the limiting current when under full load; 140A is the sweet spot for boosting balance, but 160A and higher has been fun in benches. Your TDC is extremely low at 95W, that's the limit on current determined by the thermals of your chip; I run at 200A perfectly fine and it looks like you are maxing that out at 95A. The highest you can go is between your motherboard and CPU, so up that if you can for sure.

PPT is your total package power under load allowed; you can safety raise this to your motherboards limits. I keep mine at 250W and never come close to hitting it... I'm under heavy load right now with this snapshot, and you can see I'm still not close.

View attachment 471065
Click to expand...
I agree, my settings are low. They are the defaults for the processor. I was getting good boosting with those settings for the whole time I've had this CPU up until a couple of days ago. It would boost and CPU temps would be pegged at 90 as my cooling system isn't kicked in until the coolant warms up. I was getting 11800 or so in CB20. I have played with manually setting them higher since and it did improve the score, but it still just doesn't boost the same. I think I set EDC to 160, TDC to 120, and PPT to 165. I got around 10500 in CB20 with that. I'm sure I can play with it and get some performance, but I can't understand how it worked so well at stock and then suddenly went to crap. It makes me leary of hammering with PBO. I'm going to really work on it this weekend and see if I can somehow move the tubing around enough to swap my 3900X in there without disconnecting everything. That will definitively answer the question about whether or not it's the CPU or the board. Right now, the 3900X is in an Asus TUF x570 board. I should be able to bench it there and verify what it will do and then run the same settings on this system. I dread having to take everything apart as I've spent months off and on getting this set up just the way I wanted.
 
For giggles, I messed with PBO a little. PPT at 165, TDC at 120, and EDC at 180 with Boost Clock Override at +200 with Curve Optimizer set to -5 for all cores, I get a CB20 score of 10612. The core speed is around 4000 for all cores I think. CPU temps stay below 60 with the cooling system idling.
All the same settings except setting PPT, TDC, and EDC to board limits nets a CB20 score of 11129 with all cores running around 4375. CPU temps stay in the low 60s with the cooling system idling.
 
kamikazi said:
For giggles, I messed with PBO a little. PPT at 165, TDC at 120, and EDC at 180 with Boost Clock Override at +200 with Curve Optimizer set to -5 for all cores, I get a CB20 score of 10612. The core speed is around 4000 for all cores I think. CPU temps stay below 60 with the cooling system idling.
All the same settings except setting PPT, TDC, and EDC to board limits nets a CB20 score of 11129 with all cores running around 4375. CPU temps stay in the low 60s with the cooling system idling.
Click to expand...
Seems low considering your temp of only 60C... CB20 nets me 11530 on my 5950x (or thereabouts roughly) but my CPU temp will easily hit 75C+ during the run. In CB23 I score around 30200.

Also, not familiar with the ASUS bios, but anything AGESA 1.2.0.5 or above has a PBO bug with EDC above 140 in the bios. Id recommend setting 140A in the bios for EDC and only up it when you need too in Ryzen Master for runs. The bug somehow locks the max voltage lower, which hurts PBO boost clocks horribly.

EDC of 140 my voltage will go up to 1.5V as it should for boosting. If I set anything higher in bios, its locked to like 1.43V or something silly and kills performance but temps will also be very low. I half wonder if your running into this.
 
III_Slyflyer_III said:
Seems low considering your temp of only 60C... CB20 nets me 11530 on my 5950x (or thereabouts roughly) but my CPU temp will easily hit 75C+ during the run. In CB23 I score around 30200.

Also, not familiar with the ASUS bios, but anything AGESA 1.2.0.5 or above has a PBO bug with EDC above 140 in the bios. Id recommend setting 140A in the bios for EDC and only up it when you need too in Ryzen Master for runs. The bug somehow locks the max voltage lower, which hurts PBO boost clocks horribly.

EDC of 140 my voltage will go up to 1.5V as it should for boosting. If I set anything higher in bios, its locked to like 1.43V or something silly and kills performance but temps will also be very low. I half wonder if your running into this.
Click to expand...
When I set BIOS to stock, I score 9600 to 9800 in CB20. Before the sudden change, it would score a little higher than yours when set to stock. I may roll back to BIOS 3601 as I had it there for the last year, just to be sure.
 
I haven't done anything more in regards to re-flashing BIOS or anything else yet. I haven't tried gaming yet, but I've been running CB20 off and on. Still scores around 10000. Here's a screenshot of HWInfo64. Looks like I'm getting great speeds on individual cores, four cores maxing at 5050, one at 5000, three at 4975, and only one below 4900. I know that really means nothing for performance, but I do find it interesting. Every core except core 0 is getting the full 1.5+ volts at some point over the last 5 days.

5950x problems.jpg
 
Last edited:
You also had erroneous registers in Ryzen master which leads me to belive there's more to it than software.
 
You must log in or register to reply here.
Back
Top