AMD Ryzen 5000 Builders Thread

[B-Riz]

	Cores Threads	Base Freq	Turbo Freq	L3 Cache	TDP	MSRP
Ryzen 9 5950X	16c/32t	3400	4900	64 MB	105 W	$799
Ryzen 9 5900X	12c/24t	3700	4800	64 MB	105 W	$549
Ryzen 7 5800X	8c/16t	3800	4700	32 MB	105 W	$449
Ryzen 5 5600X	6c/12t	3700	4600	32 MB	65 W	$299

5600X comes with cooler, the rest do not.

above from https://www.anandtech.com/show/1614...en-3-on-nov-5th-19-ipc-claims-best-gaming-cpu

 

[DrMrLordX]

@Timur Born

I would guess that the scheduler wants to use lower-clockspeed cores for background threads wherever possible.

 

[Timur Born]

Indeed, albeit I saw a single thread of Prime 95 AVX load being schedules to the lowest core for quite some time in one test.

After lots of fiddling, testing, analyzing, measuring and wrapping my head around the whole thing I think that I now understand the mechanics and interoperability of PBO, Curve Optimizer, Vcore (offset), (maximum) frequency limit vs. single-core + multi-core performance.

The only thing that still confuse me are:

- The EDC (215 A) limit permanently hitting 100% during sustained load that is not even close to the TDC limit (210 A) or measured current (130 A).

- Some power read-outs of HWinfo that seem like either HWinfo errors of sensors reporting wrong values, specifically "Package Power" vs. "Core+SoC Power" (20 watts lower than Package Power) vs. PIn/POut (lower than Package Power).

Overall I am not convinced that messing with all these settings is worth over just enabling PBO in BIOS with either the AMD or mainboard preset. Even with CO offsets of -30 and a slight voltage bump (AMD overclock preset) the CB20 sustained multi-core increases only by 3.5 - 5% (average 4.65 - 4.7 Ghz vs. 4.5 Gz). And then you have to do stability tests.

Sustained single-core improvements (of maybe 2-3%) are even harder to achieve, because you have to use negative CO on the "best" cores, which are likely already running closer to the edge to begin with. This is where silicon lottery comes in again, which is what overclocking mostly is about anyway.

 

[moinmoin]

The EDC (215 A) limit permanently hitting 100% during sustained load that is not even close to the TDC limit (210 A) or measured current (130 A).

EDC is Electrical Design Current which is a value the motherboard tells the CPU and should represent the peak current its VRMs can handle short term.

Combined with sensors reporting wrong values either you got a not-ideal motherboard or a not-ideal BIOS version. Do you know what's the temp of your VRMs?

 

[Timur Born]

VR MOS temps peak at around 80°C for 24 threads of Prime 95 AVX load with PBO "Enabled" (preset). The VRM section on this MSI X570 Creation gets hotter than the one on my Gigabyte Z390 Aorus Master even with better airflow.

I am puzzled why EDC is measured as constantly hitting the limit without any dips when sustained load is applied. Measured sustained current is far away from 200 - 210 A, really not even close. So I obviously don't understand the EDC measurement yet.

 

[Det0x]

The deed is done, have finally gotten this fully stable
4 memory sticks + flat CL14 + T1 with GDM disabled is very rare with ryzen.. Could only run this after i had binned all my ram sticks for the different memory channels on the motherboard

• BIOS 3501 with AGESA V2 PI 1.2.0.2
• dual CCD 5950x
• 4x8GB gskill 3600 CL16
• 1900:3800 @ flat CL14 + T1 GDM-OFF

Screenshot of TM 1umus 25 cycle + Memtest 20000% stable (do notice this is my old bloaty windows install with lots of stuff running in background)

Newest OCCT 8.1.3 1 hour large dataset extreme + 4 iteration in y-cruncher with all tests (same boot as above)

Some performance number:
The SiSoftSandra v2021.31.12 (from Mar 5th, 2021). Not sure if this is a good match for CTR (?)
Intel latency checker

Next we have dram calc easy + normal bench together with cinebench r23

And lastly we have SotTR @ 1080p lowest as a gamebench running on my new 24/7 settings = 288 CPU average fps

Very happy with these results and my new 24/7 settings
Maybe i will try to push for higher fclk now 😎

Feel free to leave any comments or questions

 

[aigomorla]

I can confirm X570 Taichi + Sammy 980 PRO + 6900XT meaning i am on 100% Gen4 pci-e.... have not had usb issues after latest agesa updates.

My brother is loving his system.
He says its wicked fast, and he is trying to tax his GPU on games he plays which is sadly @ 1440p.
I told him he needs a 4k, to tax that beast.

Make sure you guys update that AGESA.

 

[Timur Born]

I will soon start extensive USB and PCIe audio tests on both old and new AGESA, including looking for workarounds for those not getting the new version for their mainboard.

 

[aigomorla]

I will soon start extensive USB and PCIe audio tests on both old and new AGESA, including looking for workarounds for those not getting the new version for their mainboard.

They said most of the issues were on the PCI-E 4.0
Not many people were running all gen 4, because video cards are next to impossible to get, and people who did get them did not run them on a Gen 4. as they were used to MINE crypto's.

 

[Timur Born]

USB on AMD is Asmedia based, who have a mixed history of USB 3 (audio) compatibility. USB 3.1 chipsets/hubs usually worked, USB 3.0 ones were unusable for professional audio.

If I can reproduce problems then I will also test my Texas Instruments based USB 3.0 hub in between as workaround, because that helped in the past. These are hard to come by, though, mine is part of a Dell U3014 display.

 

[B-Riz]

I can confirm X570 Taichi + Sammy 980 PRO + 6900XT meaning i am on 100% Gen4 pci-e.... have not had usb issues after latest agesa updates.

My brother is loving his system.
He says its wicked fast, and he is trying to tax his GPU on games he plays which is sadly @ 1440p.
I told him he needs a 4k, to tax that beast.

Make sure you guys update that AGESA.

What setup does your brother have?

My 1080 fps went 200 consistent now with Zen3 and the 240Hz monitor, about to jump to 2560x1440 today if the monitor is good, no dead pixels or anything.

I have the 5900X on an X570 Master with what should be the latest BIOS, an SN850 2TB and a 5700XT on PICE4, no USB issues I have found, but, I don't have the bus loaded up, just keyboard, mouse and might put a Creative Labs X-Fi USB on it.

 

[Dave3000]

My Asus B550-F Strix (Wi-Fi) motherboard's BIOS sets the TRdwr for Channel A to 8 and Channel B to 7 if using the DOCP profile. I'm using 32GB Ballistix Gaming DDR4-3200 memory. Should both memory channels be the same timing for the TRdwr? I set it manually to 8 so both memory channels are 8 but I'm not sure if both channels should be set to 7 or 8. I'm using the 1804 BIOS. I sometimes have been getting crashing to the desktop when playing Sunset Overdrive and the audio would start chopping up just before crashing to the desktop and I'm wondering if it's because of the different timing for both memory channels that the auto setting did for the TRdwr.

 

[Shmee]

@Det0x your memory setup is very similar to mine, though your latency is much tighter. Also you are putting a lot more voltage across that memory

I am running 2 different 3200 Gskill Bdie kits, one Trident Z the other Flare X. Mine run at CL16, but I only put a little over 1.41V across them. My motherboard tends to put more memory voltage than what is actually set.

 

[Det0x]

@Det0x your memory setup is very similar to mine, though your latency is much tighter. Also you are putting a lot more voltage across that memory

I am running 2 different 3200 Gskill Bdie kits, one Trident Z the other Flare X. Mine run at CL16, but I only put a little over 1.41V across them. My motherboard tends to put more memory voltage than what is actually set.

Yes above ~1.5volts on samsung b-die you need to start considering to put a fan on them to keep them from overheating.. (should try to keep temp below 40-45 degrees)
I can maybe lower the voltage a little bit to around ~1.55 while retaining stability, but these tight timings require alot of power..
But i dont think i'm in the dangerzone yet, when i keep the temperature and the others settings in check, and there are many b-die sets out there which require 1.5volts at stock settings.

It also depend on RTT resistance how much amps the memory get.. Before i was running 6-3-3 RTT timings but with the latest agesa they seemingly doubled the resistance, so now iam running 7-3-1 instead. (tCKE also matter)

 

[aigomorla]

What setup does your brother have?

CPU:: 5800X
Ram: Crucial Ballastix (32GB x 2) DDR4 3600mhz - 64GB Total
Motherboard: Asrock X570 TaiChi
GPU: Sapphire Nitro+ 6900XT
PSU: EVGA Supernova Gold 850W
Storage Primary: 1TB Samsung 980 Pro. - nVME
Storage Games: 4 x 4TB Samsung 860 QVO In Raid - 0 - SATA

Yeah, its a beast...

Creative Labs X-Fi USB

if im going DAC id get a real DAC.

 

[Hitman928]

Got my 5900x installed. Looks like 2 cores will boost up to 4.9 GHz at stock, so 100 MHz over rated boost. Everything at stock I get 20813 multi and 1574 single in CBr23 so seems to be right where it should be. This is with just changing the CPU (2700 previous) and not reinstalling Windows or anything. I have some additional personal testing I can go through over the weekend when I have more time. At least at stock, the 5900x uses less power, runs cooler, and obviously performs much better than my overclocked 2700 (4 GHz). Happy so far.

 

[Kenmitch]

Got my 5900x installed. Looks like 2 cores will boost up to 4.9 GHz at stock, so 100 MHz over rated boost. Everything at stock I get 20813 multi and 1574 single in CBr23 so seems to be right where it should be. This is with just changing the CPU (2700 previous) and not reinstalling Windows or anything. I have some additional personal testing I can go through over the weekend when I have more time. At least at stock, the 5900x uses less power, runs cooler, and obviously performs much better than my overclocked 2700 (4 GHz). Happy so far.

Nice!

You should play around with Ryzen Clock Tuner and compare results. The latest offering looks to work pretty good with my 5900x.

ClockTuner v2.1 for Ryzen (CTR) Guide - Introduction (guru3d.com)

I'm still playing around with it, but it shaved around 19-20c off my temps using a suggested profile. 4475/4400 @1.175v's got me 22.649 in CBr23 for multi-score with app running and surfing the web. I haven't tried the single core with hybrid-OC yet.

 

[B-Riz]

CPU:: 5800X
Ram: Crucial Ballastix (32GB x 2) DDR4 3600mhz - 64GB Total
Motherboard: Asrock X570 TaiChi
GPU: Sapphire Nitro+ 6900XT
PSU: EVGA Supernova Gold 850W
Storage Primary: 1TB Samsung 980 Pro. - nVME
Storage Games: 4 x 4TB Samsung 860 QVO In Raid - 0 - SATA

Yeah, its a beast...



if im going DAC id get a real DAC.

LOL, nice, I will get a 6800 XT someday, for MSRP, someday, lol.

Did he have any issues with the 980 Pro?

https://www.reddit.com/r/buildapc/comments/n8868i

That is why I went with an SN850.

It is actually the Sound BlasterX G5, I have been eyeballing some DAC's on Drop though.

Sound BlasterX G5

The pocket-sized Sound BlasterX G5 delivers massively superior gaming audio performance in a compellingly tiny form-factor. This virtual 7.1 24-bit/192kHz, 120dB USB Digital Audio Converter (DAC) and headphone amplifier delivers for the ultimate headphone gaming audio experience - whether it...

us.creative.com

 

[Hitman928]

Nice!

You should play around with Ryzen Clock Tuner and compare results. The latest offering looks to work pretty good with my 5900x.

ClockTuner v2.1 for Ryzen (CTR) Guide - Introduction (guru3d.com)

I'm still playing around with it, but it shaved around 19-20c off my temps using a suggested profile. 4475/4400 @1.175v's got me 22.649 in CBr23 for multi-score with app running and surfing the web. I haven't tried the single core with hybrid-OC yet.

Thanks, I'll probably be starting a fresh Windows install and tweaking memory for a while as I have time, but I'll look more into it after that.

From 1 of my personal tests, using the exact same memory and timings as the 2700 (for now), one of the CAD tools I use for work performs 2.72x faster on the stock 5900x versus the overclocked 2700. Even if you equalize for cores, that would be a 1.81x speedup though with fewer cores the Zen3 cores would be able to also run faster so that speedup would again increase.

 

[aigomorla]

Did he have any issues with the 980 Pro?

He has not reported anything to me in slowing down.
Its probably also due to the fact he has not hit 1/4th on that drive.
The 980 is pure windows and Apps drive, no games will be stored on it, unless it requires a nVME.

 

[Timur Born]

My curiosity is served for the time being and I decided that overclocking the 5900X is not worth the cost and stability testing efforts.

Compared to stock settings with PBO disabled (LLC mode 8):

- Less than 7% CB23 score improvement increases power consumption by over 37% (PBO Advanced + CO). We are talking less than 135 watts CPU Package power vs. over 185 watts.

- Less than 5% improvement increases power draw by 25% (negative Vcore offset -0.0625 V, which decreased performance compared to the above settings accordingly).

Question: Does anyone know why CPU Package Power (and PPT) is always measured 20 watts higher than CPU Core + SoC, both idle and at full load? What other parts of the CPU use 20 watts (other than heat)?

 

[B-Riz]

Question: Does anyone know why CPU Package Power (and PPT) is always measured 20 watts higher than CPU Core + SoC, both idle and at full load? What other parts of the CPU use 20 watts (other than heat)?

What software are you using to measure?

I set my PPT to 120, and Ryzen Master shows 120 W limit.

Oooo, I got 142 fps installing it, lol.





Prime95 Small FFT all core all threads shows using 120 W right now.

 

[Timur Born]

Yes, and both your PPT and "CPU Package Power" are 20 watts higher than your "CPU core + SoC Power". You will also notice that these extra 20 watts are always present, both at peak load and full idle load.

So there is no meter telling us where these 20 watts went, which would be nice to know more about.

 

[Timur Born]

HWiNFO's author Martin replied:

CPU Package Power and PPT account for additional rails that can't be measured, but are estimated.

I will discuss with him if said "estimation" may be too high when it regularly exceeds the VRM's own POUT/PIN measurement (which cannot fully be trusted neither).

 

[Kenmitch]

Yes, and both your PPT and "CPU Package Power" are 20 watts higher than your "CPU core + SoC Power". You will also notice that these extra 20 watts are always present, both at peak load and full idle load.

So there is no meter telling us where these 20 watts went, which would be nice to know more about.

HWiNFO's author Martin replied:

I will discuss with him if said "estimation" may be too high when it regularly exceeds the VRM's own POUT/PIN measurement (which cannot fully be trusted neither).

No idea what's included in the PPT in the end, but the readings aren't consistent as far as the difference goes. I'm leaning towards something else lops into the PPT's reading that isn't viewable?

Here's some snaps of my B550/5900x running in different clocks, ECO modes, etc. You can see the value isn't fixed at all.












All of the above readings were running CBr20 with the same memory clocks/timings/voltage.

Edit: Dropping my memory clocks down from 3600MHz to 3200Mhz with the same timings while dropping the memory voltage from 1.46v's to 1.36v's increases the gap for some reason? Use the image just above this one to compare the results.



Conclusion? Further testing required.

 

[Timur Born]

I looked further. Package/PPT delta decreases to about 14 watts when C-states (sleep) are active, aka idle time for cores. Under load I get 20 watts. According to HWiNFO's Martin "it's usually VDDIO_MEM, VDD18 and RoC. "

I asked about the memory controller being part of the cIOD, he answered "SoC power covers the VDDCR_SOC rail only, VDDIO_MEM is a separate rail. ROC = Rest-of-Chip - a pre-calculated sum of all minor rails. "