3900x is a toaster oven (mine) SOLVED!

tangoseal

[H]F Junkie
Joined
Dec 18, 2010
Messages
8,622
Is anyone of you out there experience very high temps on these chips - 3900x?

Mine at wide open is hitting 90c in some benches even with an XSPC waterblock and custom loop. My old 2950x hit 75c peak at full load and it has more cores at double the process size.

I am curious if my processor might be bad.

The Wraith fan it came with also let it hit 90c

No way a waterblock with a dedicated loop is allowing the same thing? Really?

In power save profile the chip idles around 40 with my room being about 29c right now.

Wide open it avgs 77c but can peak in prime 95 to 90c. This is under water sigh

My Cinebench scores etc... are also lower than the reviews. I have my Gskill RipJaws Hynix CJR @ 3600 c16 2x8GB sticks in there right now.

When I disable half the cores to reduce the thermal load I can get over 201 points single thread in r15.

With all the cores available and higher heat I can only reach 196 peak

Yet reviewers out there are hitting 200+ / 3000+ with all cores on with a simple AIO.

I am really confused with this processor.
 

mnewxcv

Supreme [H]ardness
Joined
Mar 4, 2007
Messages
6,823
Try reseating the block. Are you running stock voltage? Try setting it manually.
 

Nobu

Supreme [H]ardness
Joined
Jun 7, 2007
Messages
4,326
Can you rotate the block 90°? might help a bit.
 

somebrains

[H]ard|Gawd
Joined
Nov 10, 2013
Messages
1,077
I would closely follow what Der8auer is doing right now.

Maybe shoot a quickie worklog and link him.
 

Gideon

2[H]4U
Joined
Apr 13, 2006
Messages
2,688
Is anyone of you out there experience very high temps on these chips - 3900x?

Mine at wide open is hitting 90c in some benches even with an XSPC waterblock and custom loop. My old 2950x hit 75c peak at full load and it has more cores at double the process size.

I am curious if my processor might be bad.

The Wraith fan it came with also let it hit 90c

No way a waterblock with a dedicated loop is allowing the same thing? Really?

In power save profile the chip idles around 40 with my room being about 29c right now.

Wide open it avgs 77c but can peak in prime 95 to 90c. This is under water sigh

My Cinebench scores etc... are also lower than the reviews. I have my Gskill RipJaws Hynix CJR @ 3600 c16 2x8GB sticks in there right now.

When I disable half the cores to reduce the thermal load I can get over 201 points single thread in r15.

With all the cores available and higher heat I can only reach 196 peak

Yet reviewers out there are hitting 200+ / 3000+ with all cores on with a simple AIO.

I am really confused with this processor.
Might want to check the voltage with Ryzen Master and make sure your motherboard is not over volting it.
 

ebduncan

[H]ard|Gawd
Joined
Feb 1, 2008
Messages
2,007
I don't have the 3900x but I just got my 3800x today. Dropped in my crosshair Hero VI. I'm chugging along at 4.3ghz @ 1.35volts. Temps are only hitting 80c under load. I also have a full custom water loop. bios version 7106

much higher than my old r7-1700. Only had a few hours of testing with it so far.
 

dragonstongue

2[H]4U
Joined
Nov 18, 2008
Messages
3,162
wonder if AMD used some "new type" of the Solder method they were using, not sure if exists or anything, but it could be possible they used something that might take a touch of time (use) for it to "set" properly.

I only state such as a curiosity that if they want the silicon to be able to handle a wide array of sub or high temp use stability, a standard solder has limits to avoid cracking though something that might take a bit more time to "mod" itself (with use) might be able to "age gracefully"

maybe take that wee bit extra time..also, correct me if I am wrong, they do not have dedicated Ryzen 3xxx liquid coolers, compatible, sure it is AM4, but, the chiplet and core being split with some "dead space" between says a "regular" liquid likely is the culprit, as 80 and 90s c temps for liquid sounds awful f hot as likely the bundled coolers are giving the same anyways (seems pointless for the expense of liquid if the blocks will not/cannot cool all that well.

granted Intel etc has had numerous issues in this regard and that is a small ass chip with a bunch of volts etc flying through it, so likely will not push out the same level of heat as say a 5-9xxx core i7 type deal )smaller, more condensed heat vs larger area)

^.^

beyond temps hope you all are truly enjoying them
 

tangoseal

[H]F Junkie
Joined
Dec 18, 2010
Messages
8,622
I figured. I meant, to use instead of the hard tube, for testing. That way you can just put it back without wasting tube on a hypothesis.
Yes I have some soft tubing but 1st I'm gonna hook a multi meter up to the test pads and try and see with the voltages are 1st and maybe a simple under volt will fix my issue
 
  • Like
Reactions: Nobu
like this

Dan_D

Extremely [H]
Joined
Feb 9, 2002
Messages
56,426
Using a 4.3GHz all core overclock mine never broke 78c. Using boost clocks it rarely went above 65c. Then again, it never would boost all the way to 4.6GHz on any core.
 

tangoseal

[H]F Junkie
Joined
Dec 18, 2010
Messages
8,622
Using a 4.3GHz all core overclock mine never broke 78c. Using boost clocks it rarely went above 65c. Then again, it never would boost all the way to 4.6GHz on any core.
What's your voltage under full load
 

TheHig

Gawd
Joined
Apr 9, 2016
Messages
635
What does the CPUz validation snapshot show? I have seen that show much more than even the stress test /HW info.

Either way it definitely sounds like some boards pushing voltage or maybe cooling 12 tiny cores is a bitch when you lose surface area?
 

kilroy67

[H]ard|Gawd
Joined
Oct 16, 2006
Messages
1,039
I'm not sure what monitoring software people are using, but some like Core Temp or Real Temp did not work correctly in getting the temperature when the 2700X came out. I always found Hwinfo64 to work the best.
 

tangoseal

[H]F Junkie
Joined
Dec 18, 2010
Messages
8,622
Hooking my Multimetwr up to the pads Inwas getting 1.35 to 1.38 full load using CPU Z torture test. Temps were staying under 75.

I have no idea what was causing my astronomical temps earlier but I am going to lower voltage anyways

Ill.report back
 

tangoseal

[H]F Junkie
Joined
Dec 18, 2010
Messages
8,622
Ok just used a -0.025v offset and still using PBO for now.

Voktage under full load settles at 1.32 avg. Runs about 7 to 10c cooler. I bet I can get this chip under 1.3v stable hrmm that would be very nice.

Without starting a whole new thread can someone explain to me how to properly uncouple the ram from the fabric. I mean I k ow where the setting is but not sure how to do it and what target is going to let the cores communicate faster.
 

primetime

Supreme [H]ardness
Joined
Aug 17, 2005
Messages
6,605
just early adopter issues....i dont see the new cpu as being faulty but more of a bios/software glitch with your motherboard possibly
 

tangoseal

[H]F Junkie
Joined
Dec 18, 2010
Messages
8,622
I was able to negative offset it quite a bit. Rock solid @ 1.28v full load using PBO. Read via SVI2 TFN in hwinfo64.

I am pleased with that. About a 20c drop in temps across the board under full load.

I am running 3 simultaneous h265 recodes right now while im going to sleep. Peak temp is 71c and not a drop more.

Wow these boards are going to fry these processors in short order the way they are shipping. I heard Steve at Hardware Unboxed toasted his 3900x because of voltages and the chip is dead.
 

Dan_D

Extremely [H]
Joined
Feb 9, 2002
Messages
56,426
I'm not sure what monitoring software people are using, but some like Core Temp or Real Temp did not work correctly in getting the temperature when the 2700X came out. I always found Hwinfo64 to work the best.
I just use Ryzen Master on the AMD systems.
 

aduljr

n00b
Joined
Jan 25, 2017
Messages
29
I am wondering if we are going to need new water blocks because where the microfins are located and where the cores and IO die are located do not match up anymore. Has anyone seen any new water block designs?
 

mnewxcv

Supreme [H]ardness
Joined
Mar 4, 2007
Messages
6,823
I am wondering if we are going to need new water blocks because where the microfins are located and where the cores and IO die are located do not match up anymore. Has anyone seen any new water block designs?
I would think as long as you have a full coverage cold plate you should be fine. Even the ihs acts as a cold plate to a degree.
 

tangoseal

[H]F Junkie
Joined
Dec 18, 2010
Messages
8,622
I am wondering if we are going to need new water blocks because where the microfins are located and where the cores and IO die are located do not match up anymore. Has anyone seen any new water block designs?
I dont think so. Turns out that my motherboard and probably all x570 are just pumping excess power into these when left at auto.

I offset undervolted my 3900x like-0.03 volts and now full stable load is around 1.29 to 1.31v and I get higher boost clocks and far less heat.

My xspc raystorm is rapidly wicking away the heat now.
 
  • Like
Reactions: Boil
like this

RamonGTP

Supreme [H]ardness
Joined
Nov 9, 2005
Messages
7,993
I am wondering if we are going to need new water blocks because where the microfins are located and where the cores and IO die are located do not match up anymore. Has anyone seen any new water block designs?
That might add a marginal amount of efficiency but it won’t be a 10+C difference.
 
Joined
Mar 1, 2017
Messages
18
All stock here and 32 idle, 59 on the worst prime 95 torture. My boost is locked at 4.2ghz at the moment though. Running liquid metal on air (noctua). Something may be amiss
 

somebrains

[H]ard|Gawd
Joined
Nov 10, 2013
Messages
1,077
I was just about to link ^^^ but I was dumbfounded by the stock voltage settings and ended up watching the clip 4x to let it all sink in.

I’m curious what Der8auer is going to find with further testing.

Be interesting if BZ probed up a a CH8 and ran test loads to see what power behavior was happening over workload as well.
 

extide

2[H]4U
Joined
Dec 19, 2008
Messages
3,494
Without starting a whole new thread can someone explain to me how to properly uncouple the ram from the fabric. I mean I k ow where the setting is but not sure how to do it and what target is going to let the cores communicate faster.
Check this out: https://www.overclock.net/forum/13-amd-general/1728878-ryzen-3000-memory-fabric-x370-x470-x570.html

You probably want to keep the IF : RAM at 1:1 and get the RAM speed up even if you need to relax the timings a bit.
 
Top