PC Processors

[Lumberjack88]

I did a single core Prime95 test (no hypterthreading and the small FFT maximum mode was selected) and these are the results:

Here are my system specifications:

OS: Windows 11 Pro Latest Update 22H2

Case: NZXT H510i Elite (It's a little oven, hot as hell)

Motherboard: Gigabyte B550 VISION D (Latest BIOS version F15d)

CPU: AMD Ryzen 9 5950x

CPU Cooler: NZXT X63

RAM: 64 GB (4x16GB) @ 3600 MHz DDR4 G.Skill Trident Z Neo

GPU: Gigabyte 3080 Vision OC

Boot M.2 SSD: ADATA XPG Gammix S50 Lite 2TB (69% full)

PSU: Seasonic Prime TX-1000

Are 86° C normal for this case and CPU cooler combination while Prime95 is running?

Why isn't my CPU hitting anything close to 5.0 GHz during this Prime 95 single core test?

Do I have to enable Auto OC in Ryzen Master to get these speeds or should the default mode suffice?

In Ryzen Master's default mode, it still tells me that I should be able to reach 5000 MHz but HWMonitor is telling a different story with its max CPU core speed readings.

EDIT: I enabled Auto OC inside Ryzen Master, rebooted the PC and these are the new Prime95 single core stress test results:

The temperatures are a bit lower now, but I can see that the stress test is progressing a bit slower than under the default mode.

Is Auto OC even worse than default mode? The maximum speed has now been raised to 5.15 GHz, yet I still only get 0.1GHz more on top of the 4.5GHz. Still nowhere near 5GHz...

[misterj]

Lumberjack88, please post SSs of Ryzen Master in Advanced mode running Cinebench R23. I can't answer your questions about your case, you could try running with the sides removed. Auto OC enables PBO and probably tells you it will void your warranty. Thanks and enjoy, John.

[Lumberjack88]

I ran Cinebench in Default and Auto OC mode. I forgot to add HWMonitor to the Auto OC screenshots, but I hope it's still good enough to pinpoint the problem. These screenshots were made about halfway into the test, so that you can see what the values are during the test at a single point in time. I then wrote the final scores onto the screenshots in retrospect.

Auto OC has yielded almost 30 points more in the single core test and about 1000 points in the multi core test of cinebench 23.

I have still not seen any frequencies above 4.7GHz in the single core test and the task manager usually tells me that the frequency is around 4.5GHz on average and only a tiny bit higher in auto OC mode.

I'll first post the default mode results and then the auto OC results:

Here's an imgur album in case anyone prefers imgur picture view over the forum view:

https://imgur.com/a/4sbRGFm

So, what could be the problem? Is the PSU not delivering enough juice to propel the CPU into 5GHz territory or are some settings messed up?

[misterj]

Thanks, Lumberjack88. I am looking. Enjoy, John.

EDIT: I need to see SS for the Basic mode and do not need CB or HWMonitor SSs.

[Lumberjack88]

Just to make sure that I don't post anything irrelevant again:

Which basic mode do you mean?  Basic Mode inside Cinebench23?

I know there's a basic mode inside Ryzen Master, but you explicitly requested the advanced view inside Ryzen Master, so I guess that's not it, either.

And when I make a screenshot, is there a time that I need to wait before I should a screenshot or should the begining of basic mode already impact the Ryzen Master Advanced view values?

[misterj]

Thanks, Lumberjack88. I was talking about RM, but now remember you posted that at the top. My Bad - sorry. If you make a change in Basic view and apply it then I assume those changes would be reflected in Advanced view. I will look more seriously now and get back. Please post another SS of both RM Views running Cinebench, posting ONLY the RM SSs. Let CB get going a minute and then take the SS. You are throttling because of the red meters at the top of RM. You can increase the limits of these parameters, but may get voiding the warranty message. Thanks and enjoy, John.

[Lumberjack88]

I redid the cinebench test by first running the single core mode for a few minutes, then taking a screenshot of the basic view of RM and shortly afterwards, while still running the same test, taking a screenshot of the advanced view of RM. The same procedure was used for the multi core mode.

I did not use Auto OC mode at all, the whole test was done using the default mode. If you need the Auto OC mode as a comparison, I'll gladly provide that too.

Here's the single core test in basic and then advanced view:

Here's the multi core test first in basic and then advanced view:

Here's the imgur album if you prefer that to the uploaded photos:

https://imgur.com/a/8mpM8XO

[misterj]

Thanks, Lumberjack88. These are great, just what I asked for! The problem is what I said before. Here is a SS of my 3970X RM setting for upping the limits.

I would suggest you do it in a profile. In my case they are at their limit when I set PBO and I can only reduce them. I cannot find the SS for when I ran this some time ago, but am sure I reduced the limits a little. This substantially increased my CB score. Let me hear what you decide. Thanks and enjoy, John

[Lumberjack88]

Thanks for your recommendation, however, I'm not able to replicate these settings for my 5950x. The highest I can go in precision boost overdrive mode is this:

 https://i.imgur.com/1f4YgUJ.png 

I can only choose 720 as the PPT upper limit, TDC at 420 and EDC at 480.

The Peak Core voltage is set to 1.39375V and I'm not able to edit/change.

I also tried another PSU (beQuiet Dark Power Pro 11 1200W Platinum Rating) in place of the current one (Seasonic Prime TX 1000W Platinum Rating) that I'm using now and it made no change to the performance of the CPU, it's still not reaching 4.9GHz.

I also tried to set the Fans to PWM instead of Auto mode and this didn't help either.

Should I just use these max values in my current Ryzen Master profile 1 or would you recommend something else? Will I fry my CPU with those settings? I haven't clicked apply&test yet...

[misterj]

Thanks, Lumberjack88. We have quite different processors and boards. The limits are set by the MB vendor to reflect what they believe to be limits for their board. I assume these are the limits of the VRM and other components. I have a 3970X (Zen 2) on a Gigabyte MB. I think these limits are too high and they expect temperature to trigger throttling. Recently a user posted a SS with 4096 for one of them. This is rediculas and represents "no limit". The limits for yours seem more than large enough and I suggest you lower them and work up. Don't know why you changed power supply. Usually this is done if it is too small and voltages sag and the system crashes. I would not expect improvement in performance. Since you have plenty of headroom in temperature, I would not expect help from cooling unless the VRM is overheating, then I expect crashes. I do not think increasing the limits will damage your processor but remember I do not work for AMD. I have done it on my 3970X to see a nice increase in CB (never ran PBO), but I run stock setting except for XMP and NUMA mode. Please post SSs if you do try. Thanks and enjoy, John.

[Lumberjack88]

I followed this tutorial to enable PBO with some reasonable settings for the 5950x as recommended in this video:

https://www.youtube.com/watch?v=1TOoYc8fJIo

Here are my results:

Single Core Test with PBO enabled:

https://i.imgur.com/YBH0BgC.jpg 

Multi Core Test:

https://i.imgur.com/Q63GVDa.jpg 

The guy in the video is getting around 5GHz single core values just with these adjustments, and multi core frequencies of around 4.45GhZ. My single core results are very erratic, and HWMonitor shows that they've at least temporarily crossed the 5GHz mark, but on average, the task manager is still showing me frequencies of about 4.56GHz or so.

Can it really just be a thermal throttling issue, even in the case of the single core test, where the temperature was about 76°? I'm not really sure what else to try. I'd probably need to bump up the vcore and set the frequency multiplier manually, but seeing how the guy in the video reached 5GHz with ease, I'm not sure what's wrong with my system that I'd need to go the extra OC mile just to get some standard as-advertised speeds.

[warikooamit]

There are a few potential reasons why your AMD Ryzen 5950X processor might not be reaching its maximum clock speed of 4.5 GHz during a single core test with Prime95:

Thermal throttling: If the processor is overheating, it may be limiting its clock speed to prevent further overheating and potential damage. Check the temperature of the processor while it is running the Prime95 test, and make sure it is not exceeding its maximum temperature rating. If it is, you may need to improve your CPU cooling solution or ensure that your case has proper airflow to dissipate heat.

Power limits: Your processor may be limited by power delivery or power consumption. Check the power delivery to the processor, including the quality of the power supply unit (PSU) and the stability of the voltage. You may also need to adjust the power limits in the BIOS or UEFI settings to allow the processor to draw more power and reach higher clock speeds.

BIOS or UEFI settings: Make sure that the BIOS or UEFI settings are configured correctly to allow the processor to reach its maximum clock speed. This may involve enabling XMP (Extreme Memory Profile) for the RAM, setting the processor to "unlock" or "overclock" mode, and adjusting the voltage and frequency settings.

Hardware or software conflicts: There may be conflicts with other hardware or software on the system that are preventing the processor from reaching its maximum clock speed. Check for updates to the BIOS or UEFI, as well as any other drivers or software that may be causing conflicts.

Processor damage: In rare cases, the processor itself may be damaged or defective, which could prevent it from reaching its maximum clock speed. If you have ruled out all other potential causes and the processor is still not reaching its maximum clock speed, it may be worth considering the possibility of a damaged or defective processor.

[Lumberjack88]

If I want to check the PSU power delivery to the CPU, can I do it like this:

Use a different PSU, keep it outside the case and plug the motherboard, CPU and GPU cables and test run cinebench once again, all the while keeping the original PSU inside the case with the removed CPU, GPU and Motherboard cables.

Or would I need to remove all the cables from the current PSU, even the SATA power cables, and then mount the new PSU into the housing (I know that the grounding of the case is reliant upon the screws that we use to mount PSUs nside the case) and only then run the test again?

BTW: If you look at the first imgur album that I posted of the cinebench results, you can see HWMonitor registering max wattage values for the package as 134.5W and 101.89W for the Cores:

Are those values representative of a healthy CPU power delivery or would you say that these values are only fleeting in nature and can still allow for a faulty CPU power delivery?

https://i.imgur.com/tM5N64c.jpeg 

[Brian-Emmons]

Had a similar problem with my 5950X, MSI MPG X570 Gaming Plus (latest Bios), Corsair Vengeance 3000Mhz Dram 32GB (4x8GB), MSI MAG CORELIQUID P240 AIO 240mm CPU Cooler, Corsair PWM fans in front and rear of case, Corsair MP510 960GB Gen3 M2 primary drive, ASUS Strix RX480 and Corsair RM1000e power pack.

Found until I had set the fan profiles in Bios to PWM mode that the chip would not boost past 4.5Ghz and was running extremely cool on Hardware monitoring.

Since setting the cooling to the correct profiles now boosts to 4.9Ghz with temp between 50 to 64 Celsius depending on what I am doing at the time.

Using a fairly old full size tower with glass and solid side panel's with twin 140 front fans and single 120 rear fan, and twin 120 roof fans on cooling radiator. All fans PWM.

[Migz-DH]

Did you ever try pulling a panel off the case and blowing an external fan into the case to see if your performance improved with more airflow? If not, give it a try. It's a super easy and inexpensive test.