cancel
Showing results for 
Search instead for 
Did you mean: 

Processors

tarvaln
Adept I
Adept I

3900x Max Boost Consolidated Testing/Results

*AMD has said that they are releasing a new BIOS to address the issues here on Sept 10th. Now we are waiting "2-3 weeks" for MB vendors to release new BIOS for their boards. You are welcome to post your results from your testing as we wait. 

This Question/Topic is to consolidate Max Boost Testing/Results regarding the 3900x only. Please do not post if you cannot test a 3900x. If you have any advice on settings please post a pic of your test/results following the guidelines below.

Please only post results. This topic is for proof of results, not speculation.

Please post a picture of your test results if you can or cannot get your 3900x to boost to 4.6 on any core.

Please us Ryzen Master to view your cores. Select "Show Histogram" under Settings.

Be sure to click "Peak" in the "Histogram" section.

Please use either CPU-Z, CinebenchR20, or Geekbench for testing. Remember to run single core test.

Please list your Motherboard, BIOS version, and RAM speed.

If adjusted, please tell us if you used Ryzen Master, Bios setting, or another tweaking tool.

Please list your Highest Single Core MAX Boost

Your pic needs to include Ryzen Master and what test you are running.

If you can get a MAX Boost of 4.6 or higher then please post the information above and what settings you changed to get there.

Questions/Statements:

Why this forum?

There are many people who like many forums who have many different MB's. Everyone involved has an AMD CPU so this is the most common forums between users.

Well, if you (did you try to) adjust (overclock) X, Y, Z then it should work. (?)

Post a pic of your 3900x CPU max boost at 4.6 with the adjustment you advise.

Why these test?

There are common test that should test your MAX Boost. If you have another that will, then post it with a pic of your results. We want to duplicate your results.

Why Ryzen Master and not HWiNFO?

There seems to be a difference in reporting between Ryzen Master and HWiNFO. Ryzen Master is a first party software where as HWiNFO is a third party. I don't know which is accurately reporting so it is safer to go with first party until it is shown which is incorrect.

The images are small and hard to read.

Open the images in a "new tab" and you can view them at full size. If it's still too small, use "save image as" and view them in your picture viewer. If you just click on them they will pop up in a small window.

Example:

Cannot get any core to Boost to 4.6 on any setting. ALL set with Ryzen Master. MB: Asus Prime x570 Pro BIOS: ver 1005, RAM: G.SKill 3200c14. Highest MAX Boost: 4.350

cinebenchr20 default 09022019.png 

Lastly, thank you to all who post results.

54 Replies
shinkojiro
Miniboss
Miniboss

Re: 3900x Max Boost Consolidated Testing/Results

2019.08.25-23.20.png

The 532 you see here is my manual 4.6 in CCD0. PBO wont allow the processor to achieve this. Temperature, which is the 'CPU package' temperature listed in Ryzen Master, averaged around 46c with all PBO limits effectively removed. "CPU" Temperature is lower around 35/36c.

2019.08.25-23.21.png

The high score here is also my manual cept with all cores at 4.3ghz with 1.35v. EDC gets maxed at 200A so i doubt it can get higher than that. (I can set the EDC higher, but non-AVX loads don't hit 200A so Im not concerned about it).

I put HWiNFO along with it to show some differences. According to ryzen master, the max boost is the 4.4ghz that those single core scores represent, while HWiNFO is showing 4550mhz. I was pretty sure it was never maintaining something like that for more than a split second because any single-core score i get is in line with manually clocking all cores to 4.4ghz.

Now, I will switch back to my manual setup and show a single-core run with the histogram open in it.

0 Kudos
shinkojiro
Miniboss
Miniboss

Re: 3900x Max Boost Consolidated Testing/Results

2019.08.25-23.50.png

And this is my manual. Notice it is less voltage than what AMD feels needs to be applied for single core boosting to the advertised speed. 1.4625 is nothing. That solid red line shows the core holding the peak of 4.6ghz throughout the test, (Save for the crazy activity shown in the next picture) and temperature was average of 42c the whole time, which is lower than the 46c average for only 4.4ghz. The process also focused on what is listed as the fastest core of CCD0 the whole time as well instead of bouncing around between core 1 and core 3, but i'm sure any of the cores in CCD0 for me can do 4.6ghz. I recon the same would be for CCD1, which I believe I can test by denying Cenebench the ability to use certain cores so it will migrate usage to other CCD's for single core tests.

2019.08.25-23.47.png

That crazy heap of activity during the test is from leaving background programs open, but notice above I actually got a better single core score than the last time I ran it.

The strange difference in our 2 Ryzen 3900x systems is my cenebench runs always use CCX0 cores for single core scores whether stock or manual, even tho my 'fastest core of CCD0' is in CCX1, while yours uses CCX1 which contains the 'fastest core in CCD0'. What is your motherboard and AGESA version? Mine is Crosshair VIII Formula AGESA 1.0.0.3ABB.

0 Kudos
tarvaln
Adept I
Adept I

Re: 3900x Max Boost Consolidated Testing/Results

Cannot reach 4.6Mhz on Max Boost. Test with Auto Over Clock set with Ryzen Master. 

MB: Asus Prime x570 Pro BIOS: ver 1005 (AGESA 1.0.0.3 ABB), RAM: G.SKill 3200c14. Highest MAX Boost: 4.355

cinebenchr20 AOC 08252019.png

Cinebench and CPU-Z usually use CCD0,CCX1, C05 and C06 on my PC. Sometimes CCD0, CCX1, C04 at around 4.2 Mhz I think one time, during a test a few days ago, it used CCD0, CCX0, C01 and hit 4.3Mhz. That was the only time I ever saw it use a different Core group.

0 Kudos
tarvaln
Adept I
Adept I

Re: 3900x Max Boost Consolidated Testing/Results

Cannot reach 4.6Mhz on Max Boost. Test with Auto Over Clock set with Ryzen Master. 

MB: Asus Prime x570 Pro BIOS: ver 1005 (AGESA 1.0.0.3 ABB), RAM: G.SKill 3200c14. Highest MAX Boost: 4.324

SuperPi32M08262019.png

Trying some tests I saw in this review AMD Ryzen 9 3900X Review | TechPowerUp. My Boost Mhz seem to be the same.

Please google the review. I keep having issues with external links on this forum. My post stay in "Currently being moderated" even when I get an email saying it's approved.

0 Kudos
shinkojiro
Miniboss
Miniboss

Re: 3900x Max Boost Consolidated Testing/Results

If you as me, all the Nay-Sayers are running away from this thread. They aren't testers, they are ex Intel use to 'it just working' because it was the same architecture for about a decade. Zen was new, and had its new-arch woes just like any will. Zen 2 is new yet again and it should be expected.

0 Kudos
ajlueke
Exemplar
Exemplar

Re: 3900x Max Boost Consolidated Testing/Results

Cannot reach 4.6Mhz on Max Boost. Tested with PBO enabled through UEFI. 

MB: Asus Crosshair VII x470, UEFI: ver 2703 (AGESA 1.0.0.3 ABB), RAM: Corsair Dominator @ 3600c16. Highest MAX Boost: 4.5

Histogram.JPG

I do hold 4.5 GHz steady, but I am periodically switching between core 2 and core 3.

High boost.JPG

0 Kudos
ajlueke
Exemplar
Exemplar

Re: 3900x Max Boost Consolidated Testing/Results

Just some of my observations from testing.

1.  PBO appears to do nothing.  If I enable it in any fashion there is no tangible effect on core performance.  All it seems to do is raise voltage, but not apply that voltage to doing any additional work. 

2.  The precision boost algorithm doesn't take into account differing CCD quality.  The power budget is wasted by boosting less efficient cores.

Figure 1:  CCX Overclock using Ryzen Master.

CCX.png

CCX0 is at 4.5 GHz, CCX1 is at 4.4 GHz for CCD0. 

CCX0 and CCX1 are at 4.1 GHz for CCD1.  Voltage is at 1.3V.

The downside to a manual overclock is that you cannot apply additional voltage on a lightly threaded, low current workload.  Here are my Cinebench scores with this setting.

Figure 2:  Cinebench Result CCX Overclocking (1.3V)

CCX+over.png

Now, here are my numbers again if I use PB, XFR and PBO enabled.

Figure 3:  Ryzen Master, Percision Boost, Extended Frequency Range, Precision Boost Overdrive Enabled.

PBO.png

Figure 4:  Cinebench Results, PB, XFR and PBO Enabled.

PBO+cine.png

Discussion:  What is immediately striking to me about these results is the single thread score, not so much that I'm not reaching 4.6 GHz.  When PB and XFR are applied, the single core voltage is boosted to 1.47V and increase of 170mv over my manual setting.  The result performance however, is exactly the same.  That extra voltage doesn't appear to be used for anything.

Figure 5:  Comparison PB, XFR vs Ryzen Master CCX Overclock.

OverclockingCB20 (single)VoltageCB20 (multi)Voltage
PB, XFR5071.4873211.3
Manual (CCX)5091.375251.3
%Difference0.4%-12.2%2.8%0.0%

As we see in the table a 12% drop in voltage nets the same performance.  This leads me to conclude that PB is somehow broken on the 3000 series.  When the number of cores being used declines, PB should add additional voltage an boost the few cores up higher, which it does, but somehow that doesn't translate to increased performance. 

Now the issue with the multicore boost is a bit different.  When boosting a multicore load, PB and XFR boost all cores simultaneously, which was fine for the 2000 and 1000 series when all cores were relatively the same.  However, AMD has already revealed that the 3900X especially has one CCD of 6 cores that performs better than the other.  Trying to push the "bad" CCD to higher clocks takes more voltage than pushing the same clocks on the "good" CCD.  This causes quite a bit of the power budget to be wasted on the "Bad" CCD.  It would be a better use of the available power budget to only push all cores together to say 4.0 or 4.1 GHz, and then only push CCD0.  I was able to increase my Cinebench multicore score by 3% simply by doing exactly that. 

I would recommend right now disabling PBO as it appears to do nothing, and overclocking via CCX controls in Ryzen master.  Sure, you lose the high voltage boost on lightly threaded loads, but right now that extra voltage isn't doing anything as far as I can tell.

0 Kudos
sqwerty
Adept II
Adept II

Re: 3900x Max Boost Consolidated Testing/Results

OP you ask people to run single thread but your screenshot is showing a multi run.

Fwiw, I know this isn't a 3900X, 3700X did hit boost clock on 1.0.0.2 based bios but boost clock was rubbish imo. As others have stated it cannot sustain it and switches to another core, costing performance. 4.4GHz  "boost" was slower (reproducibly so) than 4.3GHz fixed.

On later bios it cannot hit boost and still cannot beat a lower clocked manual underclock/undervolt.

0 Kudos
shinkojiro
Miniboss
Miniboss

Re: 3900x Max Boost Consolidated Testing/Results

that looks like a single core run to me . I didnt stay on the initial 1.0.0.2 bios long enough to see how it performed. immediately flashed to 1.0.0.3AB.

But yea, Hwu did their own research into the boost clock deficit, and while that was all well and dandy im betting even the "best" test board of the bunch arent maintaining the boost clock long enough to give representative  performance. he didnt show any scores but im betting none of them scored as high as my manual.

Knowing what i know tho, not being scarred to test certain things, Being able to run a 4.6ghz at ~43c and constant  with a manual while AMD's whatever the hell they are doing in agesa runs 4.4 at ~46c, I know this will be fixed with a simple agesa update, just as soon as Amd gets rid of that possible hostile insider.. I just dont understand how it could go so wrong at launch otherwise.

0 Kudos