Ryzen

[Stereodude]

I'll be curious to get your feedback on the Scythe Fuma 2 when you're done your build and test it for a while. This CPU has a decent amount of heat to get ride of and I wanted to keep it reasonably quiet without dealing with AIO water coolers.

I think you pretty much already have the accepted king of the quiet air cooling hill. I was looking for something cheaper and I saw people saying good things about the Fuma 2. It seems to be as good thermally as the D15 at an equivalent noise level for almost half the price.

I went back and forth between the Fuma 2 and the Scythe Ninja 5 since they're about the same price (the Ninja 5 is 5% less due to a coupon at Amazon). The Ninja 5 is larger, looks to cool a bit better for a given noise level, but can't cool quite as well due to a lower max RPM for the fans. Frankly I'm a little confused by the TweakTown review. They have lots of good data, but they make a distinction between PWM control for cooling levels and noise vs. straight 12V. AFAIK from playing around with fans 100% PWM is the same as straight 12V so I'm confused by the distinction. Otherwise it seems they've simply introduced an extra variable of how the fans respond to the PWM duty cycle vs. CPU temp curve of their test setup by using an arbitrary PWM vs. temp curve that presumably doesn't get to 100% duty that then favors some cooler's fans vs. others. I read their article on their setup, but it doesn't really explain it (that I saw). It seems like their PWM only gets to say 80% (for example). So a cooler where the fan hits max RPM at 80% duty PWM looks to do better than one where the fan doesn't hit max RPM until 100% PWM duty.

If I understand the data correctly, what you might want is the Scythe Ninja 5. It can cool almost as well as the D15 with each running their respective fans at max RPM, but it's 5dB quieter. I decided I wanted the smaller setup of the Fuma 2 and went that route.

What utility/settings do you use when you encode? I see newer Handbrake versions now offers x265/x265 10-bit/x265 12-bit for options. I've still been encoding everything in x264 for most of my stuff.

I've been encoding from the command line using the x265 executable with the video stream coming from AVIsynth+. I haven't done too much HEVC encoding because it's so slow (mainly because I'm really picky about the quality). I'm generally not just recompressing the footage as is. I usually do additional processing, like motion compensated temperal denoising the footage first. It's all slow and pretty manual. I mostly encode things that need fixing. Like recovering the original 1080p25 video from something that was turned into 1080i60 (a fair number of concerts are like this). Or fixing something where the video levels are all bungled (glowing blacks) or improperly graded footage. Or if it's something where I want to knock down grain / noise and I like it enough to put in the effort.

I am not someone who is recompressing all my blu-rays to HEVC or transcoding everything to make them smaller or anything like that. My workflow is much to manual/tedious & slow to do that.

I've used Handbrake to turn stuff into smaller files to put on a tablet for watching on travel from time to time. I've used H.264 compression for that.

 

[Handruin]

I decided to upgrade/phase out my 500GB Samsung 850 Pro and ordered two Crucial MX500 2TB 3D. They're reasonably low price at $199 and I can move my games off the NVMe for other tasks.

 

[Stereodude]

I decided to upgrade/phase out my 500GB Samsung 850 Pro and ordered two Crucial MX500 2TB 3D. They're reasonably low price at $199 and I can move my games off the NVMe for other tasks.

What were you using that for? I only just finally got around to putting my 500GB Samsung 850 Pro into my HP Elitebook to replace the 180GB Intel that came with it last week.

I've thought about initially running this new Ryzen 9 build off the 256GB NVMe that I pulled out of my Dell XPS 13 (from when I swapped in the HP EX920) and waiting to see how the new batch of PCIe 4.0 NVMe drives that were announced/shown at CES perform and getting one of them. I may still do that. I need to go see if I can find where I put it before heading off to Microcenter.

 

[Stereodude]

I found the 256GB Samsung NVMe SSD. I'm going to use that for now. Microcenter didn't have the Seasonic Focus Plus Gold. They were selling the Seasonic Focus GX in it's place with the same SKU that they have for the Seasonic Focus Plus Gold on the website. Apparently it's a slightly tweaked successor model.

 

[snowhiker]

I see you are both using the Fractal Design Define R6 case. Looks even nicer than the R4 case I used for my now ancient 2013 3770k build:

 

[Stereodude]

I see you are both using the Fractal Design Define R6 case. Looks even nicer than the R4 case I used for my now ancient 2013 3770k build:

My E5-2687W v2 is in a Fractal Design Define R5. I wanted a case for this Ryzen R9 build that could hold at least four 3.5" HDDs and one 5.25 optical drive which takes a lot of current cases out of contention.

 

[Stereodude]

I got the system (mobo, PS, CPU, RAM, NVMe SSD, and an old fanless GeForce 8600GT) running open air LTT style on the motherboard box. I updated the BIOS to the latest version, and loaded the XMP profile for the memory. Right now it's running memtest86 from a USB stick.

I need to look into this a bit more. Does the Infinity Fabric run faster by default as the memory clocks go up or do you manually have to speed it up? From a quick bit of looking, it appears that it will run 1:1 up to 3733MHz automatically.

 

[Handruin]

What were you using that for? I only just finally got around to putting my 500GB Samsung 850 Pro into my HP Elitebook to replace the 180GB Intel that came with it last week.

Mostly for game installs and video recordings from the games. I wanted something faster than my 7200RPM HDDs but also didn't want to fill up my 1TB NVMe boot device. Even with a partial amount of games install, I have 380GB remaining out of 930GB so it's chewing through space quickly.

 

[Handruin]

I see you are both using the Fractal Design Define R6 case. Looks even nicer than the R4 case I used for my now ancient 2013 3770k build:

I'm a bit of a Fractal Design fanboy. Every computer I've built in the past 6-7 years has been put in a Fractal case. I have 2 R4s, 2 R5s, and now an R6.

 

[Stereodude]

I can't find any information anywhere if the AMD RaidXpert2 RAID mode support 4kN drives. I can't really find if it supports 4kN in normal AHCI mode either.

 

[Chewy509]

I would be very surprised if the disk controller didn't support 4kn drives in AHCI mode...

As for RAID mode, I don't even think the Intel RAID drivers support 4kn drives in RAID mode. (At least all drives would have to be the same as a minimum, eg no mixing 512 / 512e / 4kn drives, for it to work).

The fact you can't find any mention of it, suggests no support. (companies only like to define what they support, not what they don't support).

 

[Stereodude]

I would be very surprised if the disk controller didn't support 4kn drives in AHCI mode...

As for RAID mode, I don't even think the Intel RAID drivers support 4kn drives in RAID mode. (At least all drives would have to be the same as a minimum, eg no mixing 512 / 512e / 4kn drives, for it to work).

The fact you can't find any mention of it, suggests no support. (companies only like to define what they support, not what they don't support).

Why wouldn't they support it? Aren't most SSD's and NVMe drives 4kN now? If you can RAID them (you can), I don't why they wouldn't support RAID on a 4kN HDD. I didn't find anyone complaining about it. I found virtually no chatter about it online.

The LSI/Broadcom RAID cards I have in my servers claims to support 4KN drives and they're 5+ years old...

 

[Chewy509]

If the controller supports NVMe, then it should support 4Kn.... I did find Intel noting support for 4Kn on their RAID controllers on the latest chipset and firmware, but...

However, I'm not aware of any SATA based SSDs that are 4Kn (all the ones I have in the office are 512e).

 

[Stereodude]

If the controller supports NVMe, then it should support 4Kn.... I did find Intel noting support for 4Kn on their RAID controllers on the latest chipset and firmware, but...

However, I'm not aware of any SATA based SSDs that are 4Kn (all the ones I have in the office are 512e).

It's probably best to just buy the 512e versions of the drives since they're the same price as the 4kN for this PC build. There's no guarantee that the RAID controller uses the same paths for for NVMe that is uses for SATA drives.

 

[Stereodude]

One thing that's slightly dissapointing is I can hear the fan on the motherboard more than I thought I'd be able to. I know some reviews said it was loud but I didn't think it would have the audible drone signature it has. I may look into ways to mod it with a bigger quieter fan down the road.

How loud is your motherboard aside from the fan? The Gigabyte Aorus Pro Wifi board is quite loud (at least outside of Windows) running open air. It's not the chipset fan or the the fans on the CPU HSF. The mobo stops the chipset fan most of the time. I guess it's some sort of "coil whine" except that it sounds like more white noise or a "shhhh" sound that you can hear pulsing and growing in volume and stopping as the CPU load changes. I've never had any piece of PC gear with this much electronic noise. I had a graphics card before that did something similar, but not nearly as loud. I hope the case blocks it. Otherwise I am going to be recording it with my Rode NT1 + Scarlett and be complaining to Gigabyte.

 

[Handruin]

How loud is your motherboard aside from the fan? The Gigabyte Aorus Pro Wifi board is quite loud (at least outside of Windows) running open air. It's not the chipset fan or the the fans on the CPU HSF. The mobo stops the chipset fan most of the time. I guess it's some sort of "coil whine" except that it sounds like more white noise or a "shhhh" sound that you can hear pulsing and growing in volume and stopping as the CPU load changes. I've never had any piece of PC gear with this much electronic noise. I had a graphics card before that did something similar, but not nearly as loud. I hope the case blocks it. Otherwise I am going to be recording it with my Rode NT1 + Scarlett and be complaining to Gigabyte.

I know the sound you're talking about where it's not quite a coil wine but more like the "shhhh" or a sizzle-like sound during certain PC operations. My board had a tiny amount of that noise that I noticed when I was working on it with the case open but I definitely cannot hear any of that type of noise when my R6 case is closed.

I'll be swapping out the FD R6 fans with Noctua NF-A14 PWN soon to give me a little more individual control on each fan and reduce the noise a bit more. I'll see if I can get a recording of the MB fan at different RPM levels. The software has a test to cycle through the RPMs which can illustrate the different noises and how bad it can get when the SB fan gets into the 3000+ RPM range. I've been tweaking the ASRock fan utility and I've gotten the noise to be a bit better but it's not 100% perfect.

 

[Handruin]

Somewhat related...if you look in your Windows Device Manager and see a few "Other devices" (see example image below) that have the yellow triangle warnings, you will want to the AMD drivers support page and pick your CPU. Their website does not allow me to direct-link the download which is stupid so you have to click through the menu in the second screen capture example I added. This should solve that issue in windows.

Also, if you're playing with the CPU timings and infinity fabric, take a look at the AMD Ryzen Master Utility.

Edit: also if your RAM has LEDs on it, here is the download link for the G.Skill RGB utility if you want to play with the colors, etc.

 

[Stereodude]

I know the sound you're talking about where it's not quite a coil wine but more like the "shhhh" or a sizzle-like sound during certain PC operations. My board had a tiny amount of that noise that I noticed when I was working on it with the case open but I definitely cannot hear any of that type of noise when my R6 case is closed.

The "shh"/sizzle sound is audible over the CPU HSF (when it's spun up under load) with one of the Prime95 test patterns (others test patterns weren't so bad). I'm skeptical that the case is going to mask it.

Somewhat related...if you look in your Windows Device Manager and see a few "Other devices" (see example image below) that have the yellow triangle warnings, you will want to the AMD drivers support page and pick your CPU. Their website does not allow me to direct-link the download which is stupid so you have to click through the menu in the second screen capture example I added. This should solve that issue in windows.

Also, if you're playing with the CPU timings and infinity fabric, take a look at the AMD Ryzen Master Utility.

Edit: also if your RAM has LEDs on it, here is the download link for the G.Skill RGB utility if you want to play with the colors, etc.

View attachment 1412

View attachment 1413

No RGB LEDs on the RAM. I was a little surprised to find there are LEDs on the back side of the motherboard to light up some narrow translucent strip on the motherboard so it has a little glow from the front. Thanks for the tip on the chipset drivers. I didn't load any drivers or do anything more than a 1909 Windows install so far and a short Prime95 session.

 

[Stereodude]

I didn't have any yellow triangle warnings, but I installed the AMD chipset drivers anyhow.

The Gigabyte Smart Fan section in the SIV application is pretty slick. You can define a custom PWM vs. temp. curve for each fan header as well as defining what thermal sensor it uses to control the PWM for that fan header.





All the standard profiles seem to have the fans at full speed by 65C.

 

[Handruin]

I got the system (mobo, PS, CPU, RAM, NVMe SSD, and an old fanless GeForce 8600GT) running open air LTT style on the motherboard box. I updated the BIOS to the latest version, and loaded the XMP profile for the memory. Right now it's running memtest86 from a USB stick.

I need to look into this a bit more. Does the Infinity Fabric run faster by default as the memory clocks go up or do you manually have to speed it up? From a quick bit of looking, it appears that it will run 1:1 up to 3733MHz automatically.

I don't know if this gets you closer to an answer but in the AMD Ryzen Master utility you can adjust some of what you are asking. I need to run some more tests with the Coupled Mode On vs Off to see if there is a notable difference with the added latency when it's uncoupled (per the description in the utility).

I'm also playing around with the PBO to see how well that works out.

 

[Handruin]

This was about the highest I got tonight in cinebench R20 playing with the Ryzen Master and auto-overclock.

 

[Stereodude]

This was about the highest I got tonight in cinebench R20 playing with the Ryzen Master and auto-overclock.

Is that a ASRock BIOS feature or something in the Ryzen Master software?

I only tried the two different power profiles in Windows so far. The Ryzen balanced and the Ryzen performance. There was a difference of a few percent between them in the x265 benchmark I ran.

I will have to install Ryzen Master and see how things are set.

 

[Stereodude]

So on mine the Infinity Fabric and the Memory Clock were automatically locked together:



My quick tests showed that PBO + Ryzen Performance profile in Windows 10 gave the best performance in the x265 FHD benchmark. Auto OC mode gave worse performance than Default or PBO. PBO is a few percent faster than default and the Ryzen Performance profile in Windows is a few percent faster than the Ryzen Balanced profile in Windows. The two are additive.

 

[Newtun]

So according to the manual

The AMD Ryzen Master application is supported on the following operating systems:

• Microsoft Windows 10 64-bit version

Yes, that's the complete list. Wonder if it would work under Wine, for *X enthusiasts?

 

[Stereodude]

So according to the manualYes, that's the complete list. Wonder if it would work under Wine, for *X enthusiasts?

I think that might be a stretch. You can probably make all the same tweaks in the BIOS though.

 

[Stereodude]

I'm skeptical that the case is going to mask it.

I don't really think the case masks it quite enough. I made a recording with my Rode NT1.



Here's the m4a of it. I did nothing to the recording other than compress it to AAC and trim a little off the front and end of the recording. No EQ, no compression, no level adjustments, etc.

Here's the relevant time stamps

0:01 turn on PC
0:02-0:07 chipset fan spins up and down
up to ~0:32 PC booting
0:32-0:37 enter password & log in
1:00 start Prime95 stress test #2
1:36 stop Prime95 stress test #2
1:49 start x265 FHD Benchmark
2:56 start x265 FHD Benchmark with PBO turned on
3:55 start Prime95 stress test #1? with PBO turned on
4:35 start Prime95 stress test #2? with PBO turned on
5:06-5:52 scrolling up and down in AMD Ryzen Master application with mouse scroll wheel (no programs are running)
5:54 Press PC power button to turn it off
5:59 PC switches off

I'm nearly certain I'm going to try to exchange the motherboard for another of the same model at Microcenter. I figure they can't all be this bad given that this is the first piece of computer gear I've bought with anything approaching this level of audible electronic switching noise.

 

[Stereodude]

Oh, I also tried another power supply earlier and got the same results. It's definitely the motherboard that's to blame.

I posted in the Aorus X570 thread at overclock.net asking for feedback if this is typical or not for the board.

 

[Stereodude]

They're not proving to be very helpful...

Edit: They're even dumber than I first thought.

 

[Handruin]

I'm with you on that one. You clearly explained the troubleshooting steps and they didn't bother to read that "Focusrite Scarlett Solo (2nd Gen) into a laptop running on battery power".

Sucks but I guess maybe returning/swapping it out for a new one and seeing if the problem persists. It's a bit of a pain, but if the noise really is that bad then I agree. I'll see if I can try the same with mine and see if it has some kind of VRM sizzle noise under load.

 

[Stereodude]

Sucks but I guess maybe returning/swapping it out for a new one and seeing if the problem persists. It's a bit of a pain, but if the noise really is that bad then I agree.

I've got it all disassembled, packaged up, and ready to return. It wasn't too bad since I'm careful when unboxing, saving all the packaging, and what not. The stupid NVMe heatsink doesn't even make contact with the NVMe drive. There's some sort of partially tacky thermal pad on the heatsink that was covered with a blue release liner (which I had and put back on to return it), but there were no signs on the thermal pad that it had even been touching the SSD.

I was impressed by how thin the thermal compound between the HSF base and the lid of the Ryzen 9 was. I was a little worried that I didn't apply enough since I was a bit stingy when applying it, but plenty squeezed out on all sides. I will be at least similarly stingy with the thermal compound when I put it back together.

I'll see if I can try the same with mine and see if it has some kind of VRM sizzle noise under load.

If yours sizzled like this one did you couldn't have missed it so far unless the room it's in is always pretty noisy. I pretty much noticed it from the second I first powered up the board. I had hoped the case would contain it, but I figured that wasn't really likely.


What sort of temps are you seeing under load and at what ambient room temp? With PBO off I saw mid to upper 60's (C) with Prime95 in a ~17C room. With PBO on, the first Prime95 stress tests would increase the temps ~10C higher as the CPU would run a bit faster during it (the other P95 stress tests ran at the same speed and didn't pull extra power). It looks like PBO lets the CPU run up to ~150W whereas with it off only ~100W (as shown in Ryzen Master). The CPU fans are running PWM 100% by 65C.

I wish the fan speed controller had more hysteresis and/or was averaging temp readings over time. The temperature can jump up and down by significant amounts quite rapidly and the PWM tracks it too closely IMHO. Like something in the background will spike the CPU usage for a second or two on a single core, the reported CPU temp will jump by 20C (or more) for those few seconds, and the fans ramp up to chase it and then settle back down after the momentary load disappears. The reported temps on the Intel systems I've had don't jump around by the same magnitude with transient loads. As far as I'm concerned they could run something like a 10 second average of the CPU temp and adjust the fan speeds off that and it would work much better.

 

[Handruin]

My MB has a similar piece of tacky thermal tape that should be making contact with my NVMe when I put the motherboard cover plate back on. I haven't confirmed that mine makes contact but I was assuming it did while reattaching the plate because it felt like there was a tiny amount of resistance when pushing the late down. My NVMe temps are around 44C for my 970 EVO.

My office is pretty quiet most of the time except when the heat or A/C is running. I don't notice the sizzle or any kind of coil whine from anything within the case. I can hear the HDDs when they spin up from being idle or the south bridge fan when it used to ramp up. I've since change the fan profile to keep the RPM lower.

I've seen some wildly-jumpy temps with Prime95 running. My room temp is around 18C right now and with PBO on I get temp fluctuations throughout the stress test. The cycle I've observed is something like 60C for 5-7 minutes then 85C for 5-7 minutes, then back to 60C.

The CPU power usage does jump up to 173W according to Ryzen Master; I don't know how accurate that is. My CPU fans are running around 30-40% as configured in my fan software when the temps are in the 60C range and around 75% fan speed in the 85C range. It's possible if I turned up the case fans that more fresh air might reduce some of the temps. I still need to mess with the profiles some more. I realize it's not an easy comparison to your setup given I prioritize keeping the fans slower for reduced noise with a byproduct higher temps.

I had the same annoyance as you with regards to the HSF temp readings and how far and fast the fans change RPM over a short period of time. I find the fluctuationg noise distracting and I wish that the change in RPMs was more of a weighted move (or averaged like you suggested) over a longer period of time so that they were less jumpy and rather more consistent. I'd rather them sit at a slightly higher speed and not fluctuate than to jump all over. I 100% agree that the PWM tracks/polls too closely and the RPM change is annoying. My last Intel system with an Asus MB did not jump around nearly at all like this system does.

I'm going to look around for other fan software management and if I can't find anything that does what we both want, I'm considering trying to write my own that will allow a more customizable fan speed averaging.

 

[Stereodude]

The new motherboard after the exchange and MC seems about like the last one in terms of noise. Maybe it's a tiny bit quieter, but not much.

I haven't run P95 for real long periods of time so I can't definitively say if it alternates between two speeds or not. So far, it seems to hold the high speed indefinitely (15+ minutes). It looks like I'm getting higher clocks than you at a lower power.



It looks to have settled around 78C, 4025MHz, and 140W with the side of the case open.

 

[Handruin]

Wow, interesting. I did a test and set all my fans to 100% fixed speed (case fans also) to see if I'm just not configuring the fan profile to provide enough cooling and the temps won't drop below 83C-85C after running P95 for 10+ min. I'm still having lower clocks than you and much higher wattage.

I may need to remount my HSF and see if I did a bad job with my thermal paste. The Noctua NH-D15 is supposed to be top end and comparable to AIO in some cases. I also have 3 x 140mm Noctua for the case and the motherboard temp never moved from 38C.

I also don't know what happened to my RAM timings...I wonder if something got messed up in my BIOS after I played with the auto overclocking and now my timings are stupid.

Here is P95 in the first 5 few minutes.



Here is 10+ min in with 100% on all fans including case fans:

 

[Stereodude]

You're running the first P95 test (small FFTs)?



Ignore the # of threads I took the screenshot on my E5-2687W v2.

It seems like you're running a mixed test or something. The In-place large FFT test runs for me like your first screenshot with higher clocks and less power draw even with PBO on. The first one has higher power draw with the lower clocks (at least with PBO enabled). With PBO off the power draw of the two is similar (clocks are still lower on the first test vs. the second though).

 

[Handruin]

Yeah that explains the large temp changes; I was running the mixed/blended test to also test RAM during the stress. When I run the Small FFTs the temps go right up to 85C like in my 10+ minute test and the frequency drops back down to 3,954MHz. I'll have to try resetting the HSF and thermal paste.

Something did mess with my RAM timings and the XMP profile was deselected in the BIOS. Now they're back to

What thermal past did you use in your build? I am using an older tube of Arctic Silver 5 and now as I look at some reviews, I may be able to shed about 5C off my temps if I get some Thermal Grizzly Kryonaut.

 

[Stereodude]

Yeah that explains the large temp changes; I was running the mixed/blended test to also test RAM during the stress. When I run the Small FFTs the temps go right up to 85C like in my 10+ minute test and the frequency drops back down to 3,954MHz. I'll have to try resetting the HSF and thermal paste.

Something did mess with my RAM timings and the XMP profile was deselected in the BIOS. Now they're back to

What thermal past did you use in your build? I am using an older tube of Arctic Silver 5 and now as I look at some reviews, I may be able to shed about 5C off my temps if I get some Thermal Grizzly Kryonaut.

I just used what came with the Scythe Fuma 2 HSF. It looks to be a silver based compound.

I'm also currently using the fans that came with the Fractal Design case. They don't seem to be very high RPM (under 1000RPM at full speed). Despite only being 3 wire fans the Aorus motherboard is able to control their RPMs. I may swap the fans with the Noctua fans I put in my Fractal Design R5 that my E5-2687W v2 is in.

 

[sedrosken]

Well, damn. I can't say I saw this coming -- I caught a sale on some things, and now I'm the proud owner of a Ryzen 5 1600AF, ASRock B450M-Pro4, and 16GB of DDR4-3000 CL16 RAM. This is strictly a platform upgrade, nothing crazy otherwise, keeping everything else. Right as I started messing with stuff that's ever so slightly beyond my 3570K's power, I have the keys to the castle practically handed to me. Seeing how everything else is still plenty capable, even rather cushy (I particularly love my NH-D15S cooler, even though I barely push it at all) I see no point in pushing further than board/CPU/RAM.

For context, for those of you who haven't heard, the 1600AF is a weird refresh of the 1600 with Zen+ wafers. It's officially a 2600 but slightly down-binned, but unofficial testing shows it usually clocks about as well as a 2600 and for 85 bucks I really couldn't say no. As for the board, the selection process was simple -- I didn't want to spend much over 70ish for one, and the Pro4 had the best VRMs in that bracket, so in the cart it went. The RAM is a weird brand, OLOy, but a friend of mine who I trust over myself when it comes to newer parts (he's had his hands on more new stuff than me, hah) says he's used and trusts their stuff considering the price of course.

I wanted to go for a bit more but I had to stop myself before my spending went too far out of control -- particularly, I wanted a small NVME boot SSD and another 16GB of RAM -- I'm not spending money that isn't mine, but I did have it earmarked for other things (not bills, just other things I've been meaning to do), so I'm kind of a weird mix of ashamed, elated, and frustrated. Maybe I'll consider that around the time of my birthday. Then again, there are perks to being single after all.

At the time of writing I'm currently recompiling my ck-patched 5.5.2 kernel with graysky1's GCC optimizations for the Zen hardware as opposed to my current Ivy Bridge optimizations.

 

[Handruin]

There is some weird and wild stuff going on with PBO and the values you enter into the fields for PPT, TDC, and EDC under Ryzan Master (or through your BIOS). More specifically if you do NOT take the max values for all fields and go much, much lower, things get interesting.

In my case, I eventually got to a value of 233 for all three fields, and now...I'm thermal throttling the CPU. My chip was hitting 95C in Prime95 at some points but I didn't run it for long. I've backed the fields down to 210 for now until I can figure out why my temps are so high.

I saw this video after a little searching which shows similar behavior. You might want to see what happens on your motherboard.

I checked and my Taichi X570 has the AGESA Combo-AM4 1.0.0.4 Patch B installed via the latest BIOS.

 

[Stereodude]

There is some weird and wild stuff going on with PBO and the values you enter into the fields for PPT, TDC, and EDC under Ryzan Master (or through your BIOS). More specifically if you do NOT take the max values for all fields and go much, much lower, things get interesting.

In my case, I eventually got to a value of 233 for all three fields, and now...I'm thermal throttling the CPU. My chip was hitting 95C in Prime95 at some points but I didn't run it for long. I've backed the fields down to 210 for now until I can figure out why my temps are so high.

I saw this video after a little searching which shows similar behavior. You might want to see what happens on your motherboard.

I checked and my Taichi X570 has the AGESA Combo-AM4 1.0.0.4 Patch B installed via the latest BIOS.

View attachment 1427

Interesting. I'll have to play with that later.

Which P95 test numbers were you using for the screenshot above? The 128 custom FFT from the video or the small one?

 

[Stereodude]

Case closed up my temps settled at about 80.5C with PBO enabled running the small FFT P95 stress test after more than an hour. It was showing 4gHz in Ryzen Master and 135-140W. It felt like the room had warmed up some by that point too.

Frankly, the efficiency loss PBO and overclocking this thing incurs, I'm not sure it's worth it. A ~5-10% performance increase for 50% more power draw (or more) seems like a bad trade-off unless your time is very valuable/expensive or you need that little extra to do something real-time that you can't quite do without the extra boost.