Question Ryzen 7 1800x only reaches 69C under full load. Doesn't reach its full potential/performance

[SlimE_1]

My ryzen 7 1800x doesn't get its full potential under full load. Yesterday I have let my cpu render a picture with 1-16 threads and my professor, my colleagues and me saw that these results weren't the usual and that there must be something wrong with my cpu. A friend of mine took my exact code for the picture which we render in java and he got a different, expected result on his 1-12 threads (pictures in attached files).

My result: Y-Axis = the time it needed to render the picture in minutes, X-Axis = the number of threads used

https://cdn.discordapp.com/attachments/639096874040819722/656922118134693945/a09-results.png

My mate's result: Expected outcome, he has a ryzen 5 2600 with 6 cores and 12 threads, he used my exact same code and got the expected result

https://cdn.discordapp.com/attachments/639096874040819722/656597334263595041/a09-results-test.jpeg

As you can see, the picture renders a lot faster when you compare 1 thread and 2 threads but it gets slower when I use 4-5 threads which makes no sense. 16 Threads is almost the same outcome as 2 threads which makes absolutely no sense.

My mate however got the expected result, where the times are shorter, the more threads are being used.

I benchmarked my cpu in cinebench r20. There will be pictures in the attached files of me showing the core temperatures, clock speeds and the cinebench result. It only reaches 69C on every core on 100% load which shouldn't be the case. It should reach at least 85C (I assume) and there must be something wrong.

https://cdn.discordapp.com/attachments/639096874040819722/656921967429156899/cinebenchr20.png

https://cdn.discordapp.com/attachments/639096874040819722/656921998265417729/cinebenchr20-2.png

Thanks for every help I can get.

 

[TJ Hooker]

I changed all that stuff and I still get the same results. The picture renders slower when using 4-5 threads than 2 or 3 threads

Are you talking about cinebench here, or your Java render?

Aren't my voltages odd here? 1.2v and lower?

You want to look at Vcore, not VID. And use hwinfo or Ryzen master for monitoring, as mentioned previously.

 

[SlimE_1]

Look at Settings … System ... About and make sure your version of Windows 10 is updated to version 1903 OS Build 18362.535 to be sure the AMD Ryzen Balanced Power Plan and Windows Scheduler work together properly with your Ryzen CPU.

If it isn't, update Win 10 then run that test again.

I was also told that the cpu might be broken and not functioning properly :/

 

[SlimE_1]

Are you talking about cinebench here, or your Java render?


You want to look at Vcore, not VID. And use hwinfo or Ryzen master for monitoring, as mentioned previously.

I got the same results in the java rendering process

I used hwinfo, the pictures are from hwinfo.

I changed the power settings to amd. Are there more things I could do or is my cpu just broken

 

[DMAN999]

Check the CPU VCore Voltage (SVI2 TFN) reading in HWInfo while running CB20 multicore test:
https://www.hwinfo.com/files/hwi_621_4040.zip

If it doesn't get to around 1.3v to 1.325v with your CPU clocks up around 4 Ghz while the test is running, I'd check your BIOS settings and make sure that they are set properly.
You will need to Google for a guide to setting your MB BIOS up with a Ryzen 7 1800x.

 

[SlimE_1]

Check the CPU VCore Voltage (SVI2 TFN) reading in HWInfo while running CB20 multicore test:
https://www.hwinfo.com/files/hwi_621_4040.zip

If it doesn't get to around 1.3v to 1.325v with your CPU clocks up around 4 Ghz while the test is running, I'd check your BIOS settings and make sure that they are set properly.
You will need to Google for a guide to setting your MB BIOS up with a Ryzen 7 1800x.

https://imgur.com/a/41eWUeW

View: https://imgur.com/a/41eWUeW


it says that my memory clock is 1600 mhz even though I overclocked it to 3200?

 

[TJ Hooker]

I got the same results in the java rendering process

I used hwinfo, the pictures are from hwinfo.

I changed the power settings to amd. Are there more things I could do or is my cpu just broken

The pictures in your post where you've looking at voltage during cinebench are hwmonitor, not hwinfo.

Are you actually having any performance issues in Cinebench? Because your all core score is fine. I'm trying to understand why you're still focusing on CPU behaviour while running cinebench if everything is performing fine...

 

[TJ Hooker]

it says that my memory clock is 1600 mhz even though I overclocked it to 3200?

That's normal. DDR means double data rate. So a data rate of 3200 MHz corresponds to a memory clock of 1600 MHz.

So I noticed that in your java render time graphs, the 16 thread render time for the 1800X is longer than the single thread render time for the 2600. I have a very hard time believing that both CPUs were running the same workload.

Can you post screenshots of hwinfo with the clock speed and Vcore while running the Java render with 16 threads?

 

[SlimE_1]

Check the CPU VCore Voltage (SVI2 TFN) reading in HWInfo while running CB20 multicore test:
https://www.hwinfo.com/files/hwi_621_4040.zip

If it doesn't get to around 1.3v to 1.325v with your CPU clocks up around 4 Ghz while the test is running, I'd check your BIOS settings and make sure that they are set properly.
You will need to Google for a guide to setting your MB BIOS up with a Ryzen 7 1800x.

https://imgur.com/a/VbBGdLL

View: https://imgur.com/a/VbBGdLL


cinebench with hwinfo

 

[DMAN999]

https://imgur.com/a/41eWUeW

View: https://imgur.com/a/41eWUeW


it says that my memory clock is 1600 mhz even though I overclocked it to 3200?

1600 x 2 = 3200, that is correct.
But When you start HWinfo choose Sensors only like this:

https://imgur.com/pt2RsUj

View: https://imgur.com/pt2RsUj


And look at this reading while CB20 is running:

https://imgur.com/lrWd8K2

View: https://imgur.com/lrWd8K2

 

[SlimE_1]

That's normal. DDR means double data rate. So a data rate of 3200 MHz corresponds to a memory clock of 1600 MHz.

So I noticed that in your java render time graphs, the 16 thread render time for the 1800X is longer than the single thread render time for the 2600. I have a very hard time believing that both CPUs were running the same workload.

Can you post screenshots of hwinfo with the clock speed and Vcore while running the Java render with 16 threads?

https://imgur.com/a/VbBGdLL

View: https://imgur.com/a/VbBGdLL

this is cinebench

 

[DMAN999]

1.2v is definitely on the low side.
I would look for a good guide for your BIOS settings for an 1800x and see what you have vs what a good guide says.

But HWinfo does show a few of your cores hitting 4 GHz so it does seem to be performing like it should for the most part.


Like TJ Hooker said, it could just be an issue with your Java Test setup.

 

[SlimE_1]

That's normal. DDR means double data rate. So a data rate of 3200 MHz corresponds to a memory clock of 1600 MHz.

So I noticed that in your java render time graphs, the 16 thread render time for the 1800X is longer than the single thread render time for the 2600. I have a very hard time believing that both CPUs were running the same workload.

Can you post screenshots of hwinfo with the clock speed and Vcore while running the Java render with 16 threads?

this is java

https://imgur.com/a/MCcSqV2

View: https://imgur.com/a/MCcSqV2

 

[DMAN999]

That shows you running 3.7 GHz on all cores which seems normal as far as I know.

The AMD Ryzen 7 1800X is the top of the line, 8 core, 16 thread CPU AMD released today, March 2nd, for $500.
The CPU has a base clock of 3.6 Ghz (although I typically see it sit right at 3.7 Ghz under load) and will boost all the way up to 4.1 Ghz with XFR provided there is ample temperature headroom

From here:

AMD's Ryzen 7 1800X: Performance testing and overclocking of the brand new CPU!

I have been kindly sent the Ryzen 7 1800X and the MSI X370 XPOWER Gaming to review from AMD... and the performance is something to be reckoned with. R7

cubedcorps.com

 

[TJ Hooker]

The CPU does seem to be working as it should.

Which version of Java do you have installed, and does your colleague have the same version? Same version of Windows?

 

[SlimE_1]

The CPU does seem to be working as it should.

Which version of Java do you have installed, and does your colleague have the same version? Same version of Windows?

my java version is 11.0.4, my friend, who tested it, has java 12

 

[SlimE_1]

The CPU does seem to be working as it should.

Which version of Java do you have installed, and does your colleague have the same version? Same version of Windows?

I tested it on java 13, still same output

 

[TJ Hooker]

So those are the jdk (java development kit) versions. What is the jre (java runtime environment) version? It should be of the form version 8 update 231, or build 1.8.0_231.

 

[SlimE_1]

So those are the jdk (java development kit) versions. What is the jre (java runtime environment) version? It should be of the form version 8 update 231, or build 1.8.0_231.

java.runtime.name=Java(TM) SE Runtime Environment
java.runtime.version=11.0.4+10-LTS

 

[DMAN999]

I would look into the differences between your 1000 series CPU and his 2000 series CPU when coding/rendering in Java.
It might just be a case of his 2600 being optimized to use multiple threads in a different manner than your 1800x is.

 

[TJ Hooker]

java.runtime.name=Java(TM) SE Runtime Environment
java.runtime.version=11.0.4+10-LTS

Ah, sorry, my mistake. I wasn't aware Java had been revamped for Java 11 such that the JRE was no longer a standalone thing. All I can say is that after trying to read up a bit on Java versioning it seems like an utter mess

Anyway, all I can really say is that this seems much more likely to be a software thing than a hardware thing. Unless you start to see the 1800X underperforming in other workloads.