• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

5800x (and other Zen 3 chips) PBO settings/Temperature fix

Mussels

Freshwater Moderator
Staff member
Joined
Oct 6, 2004
Messages
56,281 (8.41/day)
Location
Oystralia
System Name Rainbow Sparkles (Power efficient, <350W gaming load)
Processor Ryzen R7 5800x3D (Undervolted, 4.45GHz all core)
Motherboard Asus x570 Gaming-F
Cooling Alphacool Apex UV - Alphacool Eisblock XPX Aurora + EK Quantum ARGB 3090 w/ active backplate
Memory 2x32GB DDR4 3600 Corsair Vengeance RGB @3866 C18-22-22-22-42 TRFC704 (1.4V Hynix MJR - SoC 1.175V)
Video Card(s) Galax RTX 3090 SG 24GB: Underclocked to 1700Mhz 0.737v (375W down to 250W))
Storage 2TB WD SN850 NVME + 1TB Sasmsung 970 Pro NVME + 1TB Intel 6000P NVME USB 3.2
Display(s) Kogan 32" 4K 70Hz + Gigabyte G32QC (1440p 165Hz) + Phillips 328m6fjrmb (1440p 144Hz)
Case Fractal Design R6
Audio Device(s) Logitech G560 | Corsair Void pro RGB |Blue Yeti mic
Power Supply Fractal Ion+ 2 860W (Platinum) (This thing is God-tier. Silent and TINY)
Mouse Logitech G Pro wireless + Steelseries Prisma XL
Keyboard Razer Huntsman TE (custom white and steel keycaps)
VR HMD Oculus Rift S + Quest 2
Software Windows 11 pro x64 (Yes, it's genuinely a good OS) OpenRGB - ditch the branded bloatware!
negative offsets are unstable at idle and low load



It is 0% likely that your system "doesnt work right" without it, and 100% likely something else is wrong - VRM's overheating, faulty hardware, incorrect settings - we cant know what until you figure it out either but avoid assumptions or you'll never get there.
 
Joined
May 29, 2014
Messages
11 (0.00/day)
You guys aren't answering my question and talking as if there aren't different levels of acceptable use case stability.
I mean, given the "tone" and ignorance, it's almost like people just want others to waste 48+ hours of time and electricity as a joke.
 
Joined
Dec 25, 2020
Messages
1,341 (1.74/day)
Location
São Paulo, Brazil
System Name Project Kairi v0.5 (temporary downgrade)
Processor Intel Xeon E5-4669 v3
Motherboard Gigabyte X99-Ultra Gaming rev 1.0
Cooling id-Cooling Frostflow X 360 + Gelid GC-Extreme
Memory 64 GB (4x 16) Corsair Dominator Platinum DDR4-2133 @ 9-11-11-21 1.35v
Video Card(s) ASUS TUF Gaming OC GeForce RTX 3090 24 GB GDDR6X
Storage WD Green SN350 480 GB
Display(s) Samsung The Frame 2022 32-inch (1080p60)
Case Cooler Master MasterFrame 700
Audio Device(s) EVGA Nu Audio (classic)
Power Supply EVGA 1300 G2 1.3kW 80+ Gold
Mouse Logitech G305 Lightspeed K/DA + Logitech G840 XL K/DA
Keyboard Logitech G Pro K/DA with GX Brown switches
Software Windows 11 Pro for Workstations 22H2
Benchmark Scores Older build pic (with R9 5950X installed): https://i.imgur.com/yxc0HrZ.jpg
You guys aren't answering my question and talking as if there aren't different levels of acceptable use case stability.
I mean, given the "tone" and ignorance, it's almost like people just want others to waste 48+ hours of time and electricity as a joke.

But there aren't. Your stuff is either stable or it has hidden flaws which will emerge due to how clock ramping, v/f curve etc. works.

Giving you any advice on how to achieve an unstable overclock is just that: giving bad advice that people on the forum might also follow.
 
Joined
Apr 30, 2020
Messages
581 (0.58/day)
System Name New One
Processor Ryzen 5 5600X
Motherboard MSI MGP Gaming WIFI
Cooling Stock
Memory Corsair Vengeance pro RGB 3200mhz 16Gbs
Video Card(s) EVGA 2080 Ti FTW 3
Storage Western digital Sata SDD 500gb
Display(s) HP X24i
Case Corsair 4000D Airflow
Power Supply Corsair 650M
Mouse Lifeworks gaming mouse lol
Keyboard lifeworks monster rgb light keyboard
You guys aren't answering my question and talking as if there aren't different levels of acceptable use case stability.
I mean, given the "tone" and ignorance, it's almost like people just want others to waste 48+ hours of time and electricity as a joke.
I use Prime95 to test stablity, but I also don't bother running high Infinity fabric I'm not anywere near 3800mhz In fact I'm below 3200mhz & I have Never had W.H.E.A errors, because of it. I also have never had a idle crash either.
My chip is horribly hot running prime95 small FFT's it hits like 82C, but I've just concluded I've got a reather hot sample as no one else gets the temperatures I get. Large FFT's 72C-74C, Heck Blend I only hit about 59-60C in prime.
I'm running LLC right now with pbo it did low temps bylike 2C, but like I said I have hot a chip it likes to use a max of 1.362 for some all cores loads most 5600x don't like anything above 1.288 or less.
 
Joined
May 29, 2014
Messages
11 (0.00/day)
But there aren't. Your stuff is either stable or it has hidden flaws which will emerge due to how clock ramping, v/f curve etc. works.

Giving you any advice on how to achieve an unstable overclock is just that: giving bad advice that people on the forum might also follow.
After reading from Y-Cruncher's very own website I have concluded that it and P95 Blend are essentially equal in capability, with p95 simply needing more time for an equivalent result.

From My Personal Experience: 8 hours of p95 with the highest AVX mode your CPU supports is adequate as a stability stressor.
An 8 core cpu at 4.5Ghz is looking for 1 fault in 1e^15 cycles in this 8 hours. Moving up to 48 hours is not statistically significant enough to make a difference, and does not rule out thermal deviation induced instability or the dreaded "cosmic ray" bit flip.

So again, I don't understand the hardon for y-cruncher as a better proof over prime95.
 
Joined
Dec 25, 2020
Messages
1,341 (1.74/day)
Location
São Paulo, Brazil
System Name Project Kairi v0.5 (temporary downgrade)
Processor Intel Xeon E5-4669 v3
Motherboard Gigabyte X99-Ultra Gaming rev 1.0
Cooling id-Cooling Frostflow X 360 + Gelid GC-Extreme
Memory 64 GB (4x 16) Corsair Dominator Platinum DDR4-2133 @ 9-11-11-21 1.35v
Video Card(s) ASUS TUF Gaming OC GeForce RTX 3090 24 GB GDDR6X
Storage WD Green SN350 480 GB
Display(s) Samsung The Frame 2022 32-inch (1080p60)
Case Cooler Master MasterFrame 700
Audio Device(s) EVGA Nu Audio (classic)
Power Supply EVGA 1300 G2 1.3kW 80+ Gold
Mouse Logitech G305 Lightspeed K/DA + Logitech G840 XL K/DA
Keyboard Logitech G Pro K/DA with GX Brown switches
Software Windows 11 Pro for Workstations 22H2
Benchmark Scores Older build pic (with R9 5950X installed): https://i.imgur.com/yxc0HrZ.jpg
After reading from Y-Cruncher's very own website I have concluded that it and P95 Blend are essentially equal in capability, with p95 simply needing more time for an equivalent result.

From My Personal Experience: 8 hours of p95 with the highest AVX mode your CPU supports is adequate as a stability stressor.
An 8 core cpu at 4.5Ghz is looking for 1 fault in 1e^15 cycles in this 8 hours. Moving up to 48 hours is not statistically significant enough to make a difference, and does not rule out thermal deviation induced instability or the dreaded "cosmic ray" bit flip.

So again, I don't understand the hardon for y-cruncher as a better proof over prime95.

Generally, however that will also leave transient loads on the table. In addition to Y-cruncher or Prime, I also run memtest86 bootable and Realbench for some time to validate my configuration.
 
Joined
Sep 11, 2019
Messages
33 (0.03/day)
or the dreaded "cosmic ray" bit flip.
I have had people look at me cross eyed when I explain that to them. They say things like "Sir, this is a Wendy's"

When I was younger I would spend the time fine tuning, tweaking, and testing. Now? With the 5800X I'm typing from, I went in the UEFI, set it to ECO with +200MHz boost, done. I don't do anything harder than gaming though; no extended all core workloads. With a 240mm AIO in a full tower Phantek all fans on custom curve, in a large room at 23c, it might get in the high 60s during marathon gaming sessions while staying quiet. Like many others, the OOB experience was not so great.
 
Joined
May 29, 2014
Messages
11 (0.00/day)
I have had people look at me cross eyed when I explain that to them. They say things like "Sir, this is a Wendy's"

When I was younger I would spend the time fine tuning, tweaking, and testing. Now? With the 5800X I'm typing from, I went in the UEFI, set it to ECO with +200MHz boost, done. I don't do anything harder than gaming though; no extended all core workloads. With a 240mm AIO in a full tower Phantek all fans on custom curve, in a large room at 23c, it might get in the high 60s during marathon gaming sessions while staying quiet. Like many others, the OOB experience was not so great.
Yeh, tuning a chip has become it's own field of science, practically.
I dabble in some personal projects and the occasional use of video/image editing, but mostly use the PC as my boobtube.
As such, it will sit in the 60-70C range in the heaviest of games at my settings, and does most full-core work below 75C with the exception of truly horrifying in-cache AVX2 acceleration.

I always get a chill up my spine when per-core power wants to exceed ~13.5 watts. At less than 8mm^2 per core (cache slice included) that extrapolates out to a power density of 2 megawatt per meter square.
 
Joined
Oct 16, 2018
Messages
815 (0.52/day)
Location
Uttar Pradesh, India
Processor AMD R7 1700X @ 4100Mhz
Motherboard MSI B450M MORTAR MAX (MS-7B89)
Cooling Phanteks PH-TC14PE
Memory Crucial Technology 16GB DR (DDR4-3600) - C9BLM:045M:E BL16G36C16U4W.M16FE1 X2 @ CL14
Video Card(s) XFX RX480 GTR 8GB @ 1408Mhz (AMD Auto OC)
Storage Samsung SSD 850 EVO 250GB
Display(s) Acer KG271 1080p @ 81Hz
Power Supply SuperFlower Leadex II 750W 80+ Gold
Keyboard Redragon Devarajas RGB
Software Microsoft Windows 10 (10.0) Professional 64-bit
Benchmark Scores https://valid.x86.fr/mvvj3a
After reading from Y-Cruncher's very own website I have concluded that it and P95 Blend are essentially equal in capability, with p95 simply needing more time for an equivalent result.

From My Personal Experience: 8 hours of p95 with the highest AVX mode your CPU supports is adequate as a stability stressor.
An 8 core cpu at 4.5Ghz is looking for 1 fault in 1e^15 cycles in this 8 hours. Moving up to 48 hours is not statistically significant enough to make a difference, and does not rule out thermal deviation induced instability or the dreaded "cosmic ray" bit flip.

So again, I don't understand the hardon for y-cruncher as a better proof over prime95.
Y-cruncher Is better because it will find an Unstable CO in seconds or minutes compared to p95 that can take hours.
Example R23 almost -30 AC no issues. Ignore the score I was trying out different PPT/EDC/TDC
R23 5800X 15740.jpg


Same CO not even 1 minute into Y-cruncher - unstable. Running 8 hours off P95 is a waste of electricity IMHO when y-cuncher can do it in seconds. :p
YC error.jpg
 
Joined
May 29, 2014
Messages
11 (0.00/day)
Y-cruncher Is better because it will find an Unstable CO in seconds or minutes compared to p95 that can take hours.

Same CO not even 1 minute into Y-cruncher - unstable. Running 8 hours off P95 is a waste of electricity IMHO when y-cuncher can do it in seconds. :p

That's fine, so earlier today I ran Y-cruncher 5 billion a few times (got a 226.5 second run), then let it run the full test suite while I was doing my quarterly sunday house cleaning for about 3 hours.
No crashes, no reboots, no halts, no test failures.

If you look back, read all of my speculation, and take my settings into account it may very well be that my Mobo is mucking the power settings up, my use of LLC levels is allowing a slight overvoltage (or maybe the correct voltage), and my PPT/TDC/EDC setup in conjunction with all of this is allowing a -30 CO to simply run the chip without problems.

But I have to say most of the "knowledge" regarding Zen3's behavior seems to be a mix of "It works this way for me" and "This voodoo Saturn magic from the 11th parallel alter-universe is a fact of law"
A true answer to all of this speculation would be comparing fully manual volt and clock settings, as it removes ALL variables and allows the tester to force unstable voltage levels while forcing a locked clockrate.
Do you really KNOW your chip is stable at x clock and y volt?
I know my chip just cannot run R23 at 4.95Ghz, the voltage required slams into the 90C thermal limit, and running at 88C hotspot only gets me ~7 seconds before a reboot.
I know this because I test locked, manual settings to validate the automatic behavior I set up.
I set my PBO2 controls specifically because I want a 70-80C target, and from analyzing MANUAL locked settings -first-, I know my chip will do p95 (and now ycruncher) at 4.45Ghz at 1.135v at 75C.
From all of my testing I KNOW my particular 5800X is extremely efficient in the ~4.4-4.6Ghz range (heaviest AVX2 work to lighter AVX work), and every 50Mhz pushed requires climbing a voltage cliff that becomes dangerous after ~4600-4800Mhz.
From my hands-on manual testing I know what volt and what clock is unstable at which workload.
The "P##" word is apparently taboo and anathema to proof so I feel discouraged to talk any further about what I KNOW MY CHIP DOES.

I have not simply jacked random numbers into the settings because it gave me some R23 score I can be proud of. I set my 135/100/125 limits because it fits my cooling system and my desired temperature limits. I thus also set an all-core negative CO because after testing everything from a custom per-core based on the chip's FIT reporting, to everything from -15 to -30, the all-core -30 setting makes the voltage fall in line with the stable, known manual clockrates at such heavy benchmarking stress.

So what do you really know about how your chip functions?
 
Joined
Oct 16, 2018
Messages
815 (0.52/day)
Location
Uttar Pradesh, India
Processor AMD R7 1700X @ 4100Mhz
Motherboard MSI B450M MORTAR MAX (MS-7B89)
Cooling Phanteks PH-TC14PE
Memory Crucial Technology 16GB DR (DDR4-3600) - C9BLM:045M:E BL16G36C16U4W.M16FE1 X2 @ CL14
Video Card(s) XFX RX480 GTR 8GB @ 1408Mhz (AMD Auto OC)
Storage Samsung SSD 850 EVO 250GB
Display(s) Acer KG271 1080p @ 81Hz
Power Supply SuperFlower Leadex II 750W 80+ Gold
Keyboard Redragon Devarajas RGB
Software Microsoft Windows 10 (10.0) Professional 64-bit
Benchmark Scores https://valid.x86.fr/mvvj3a
I have been saying some of the same things if you look at my earlier post. I did say I had to use a combination of PBO and LLC to get the maximum out of my chip as well.

In my case I am limited with my Dual tower Air cooler so have not pushed my CPU beyond 4650 all core was able to break 16K in R23 but still hit over 90 degrees so was considering getting an AIO.
I found that my chip does not drop clocks at 90C but only when it goes closer to 95C. I have not adjusted my temp limits in PBO. Still best to keep it under 90 at all times IMHO.

If I use one of the ECO mode option in the BIOS 45W/65W/95W and -CO at 30 that is much more stable than running it at MAX PBO so that is a factor to consider.
You could probably get -30 on all cores to work by adjusting PBO and other voltages


Another factor is the motherboard/BIOS/PSU power delivery etc. Some are better than others, some can deliver clean power low ripple many other factors can affect overclocking.
 
Last edited:

tabascosauz

Moderator
Supporter
Staff member
Joined
Jun 24, 2015
Messages
5,475 (1.97/day)
Location
Western Canada
Processor 5800X3D | 5700G
Motherboard ROG Impact | B550I Aorus AX
Cooling TR FC140 | Scythe BS3
Memory 32GB 3733CL14 | 16GB 4333CL16
Video Card(s) RTX 4070 Ti Eagle | Vega 8
Case Caselabs S3 | Lone L5
Power Supply Corsair HX1000 | HDPlex 200W
A true answer to all of this speculation would be comparing fully manual volt and clock settings, as it removes ALL variables and allows the tester to force unstable voltage levels while forcing a locked clockrate.
Do you really KNOW your chip is stable at x clock and y volt?

If this whole rant is just to disparage Precision Boost 2 in favour of fixed all-core OC, just say so lol. This is not a new topic since 2019 lol. All-core OC is neither dead nor useless, it's just unnecessarily restrictive for clocks, and not how the chips are designed. You have yourself proven that PB2 gives you lightly threaded boost that can't be reached with manual OC. But if you're happy with a fixed 4.5GHz, no one's stopping you.

How do we know it's not stable? Because the stress tests crash and error out when it's not, lol. Boost algo means there's always some variability in clocks, but it's not as if one iteration runs 4650MHz and the next one runs 4900MHz. It just needs a bit more time in testing due to that variability.

No one said you have to spend 1000 hours running corecycler with P95 or ycruncher, but there's hundreds of pages of threads of other people who put in the time to figure it out. That's not "this voodoo Saturn magic from the 11th parallel alter-universe is a fact of law" just because you don't like it :laugh:

CoreCycler - tool for testing Curve Optimizer settings | Page 53 | Overclock.net

The "P##" word is apparently taboo and anathema to proof so I feel discouraged to talk any further about what I KNOW MY CHIP DOES.

You do know that corecycler runs Prime95 by default, right........? You know, the same default config that basically ensures "good enough" stability and only takes a couple hours at most........? lol

All everyone is saying is that CO negative offset undervolt generally is a question of single-thread stability.
To test that, you run a single threaded test that maximizes boost clock.
Running vanilla P95 Blend is not a single-threaded test. That's why we use a script that runs P95 in a useful manner.
 
Joined
Aug 15, 2016
Messages
378 (0.16/day)
Processor Intel i7 4770k @ 4.5 GHz 1.294V
Motherboard ASUS Sabertooth Z87
Cooling Corsair H100i
Memory Patriot Viper 3 RedD 16 GB @ 1866 MHz
Video Card(s) XFX RX 480 GTR 8GB
Storage 1x SSD Samsung EVO 250 GB 1x HDD Seagate Barracuda 3 TB 1x HDD Seagate Barracuda 4 TB
Display(s) AOC Q27G2U QHD, Dell S2415H FHD
Case Cooler Master HAF XM
Audio Device(s) Magnat LZR 980, Razer BlackShark V2, Altec Lansing 251
Power Supply Corsair AX860
Mouse Razer DeathAdder V2
Keyboard Razer Huntsman Tournament Edition
Software Windows 10 Pro x64
My chip is horribly hot running prime95 small FFT's it hits like 82C, but I've just concluded I've got a reather hot sample as no one else gets the temperatures I get. Large FFT's 72C-74C, Heck Blend I only hit about 59-60C in prime.
I believe that's related to the stock cooler you have.
 

Mussels

Freshwater Moderator
Staff member
Joined
Oct 6, 2004
Messages
56,281 (8.41/day)
Location
Oystralia
System Name Rainbow Sparkles (Power efficient, <350W gaming load)
Processor Ryzen R7 5800x3D (Undervolted, 4.45GHz all core)
Motherboard Asus x570 Gaming-F
Cooling Alphacool Apex UV - Alphacool Eisblock XPX Aurora + EK Quantum ARGB 3090 w/ active backplate
Memory 2x32GB DDR4 3600 Corsair Vengeance RGB @3866 C18-22-22-22-42 TRFC704 (1.4V Hynix MJR - SoC 1.175V)
Video Card(s) Galax RTX 3090 SG 24GB: Underclocked to 1700Mhz 0.737v (375W down to 250W))
Storage 2TB WD SN850 NVME + 1TB Sasmsung 970 Pro NVME + 1TB Intel 6000P NVME USB 3.2
Display(s) Kogan 32" 4K 70Hz + Gigabyte G32QC (1440p 165Hz) + Phillips 328m6fjrmb (1440p 144Hz)
Case Fractal Design R6
Audio Device(s) Logitech G560 | Corsair Void pro RGB |Blue Yeti mic
Power Supply Fractal Ion+ 2 860W (Platinum) (This thing is God-tier. Silent and TINY)
Mouse Logitech G Pro wireless + Steelseries Prisma XL
Keyboard Razer Huntsman TE (custom white and steel keycaps)
VR HMD Oculus Rift S + Quest 2
Software Windows 11 pro x64 (Yes, it's genuinely a good OS) OpenRGB - ditch the branded bloatware!
But there aren't. Your stuff is either stable or it has hidden flaws which will emerge due to how clock ramping, v/f curve etc. works.

Giving you any advice on how to achieve an unstable overclock is just that: giving bad advice that people on the forum might also follow.
^ This

I changed my ram OC and had roughly one game crash a week. It passed any testing i threw at it, but something about the combination of loads from 3D gaming could trigger it over time that synthetic testing couldnt.

Just because I wasn't aware of it, didn't mean the instability wasn't there all along.
 
Joined
May 29, 2014
Messages
11 (0.00/day)
:snipped:
You missed the entire context of my post.

Start with manual settings to learn where your chip's de facto hard limits are (i.e. "what does it take to run 11 watts per core, or what voltage rund AVX2 at 4.2Ghz without fail), and if it's more leaky or more "tight". Use this to know where your chip starts falling off the cliff, how it behaves under various loads.
From there, tune the PBO limits for a desired temperature min-max range, and then fine tune your CO to match the known stable voltages at the known stable clocks within those limits.

Since I know my chip doesn't work at 4.95Ghz and I've recorded the single core boost behavior at that and higher clocks, and since regardless of what AMD says is safe I want my chip to never exceed 1.35v per any core at any clock, I turn the boost down so the PBO controls match my criteria.
It's as simple as that. Do some heavy manual testing to -know- how the chip functions, and then set your PBO stuff to match what you've recorded as functional within your desired power/voltage/temp limits. From there do your final stability checks.

That, at least, is the method I used and it seems to have worked quite well, irrespective of whatever is common knowledge of what is proper for Zen 3.
 
Joined
Feb 3, 2019
Messages
1,861 (1.27/day)
Location
Chicago Land
Processor 13600KF
Motherboard Asus B660-G (BCLK enabled!!)
Cooling Stock Air
Memory Patriot C36 series 5200mhz
Video Card(s) Gigabyte RTX 2060
Storage Samsung 980 pro m2
Display(s) 21" - 55"
Case None
Power Supply Antec CP series 850w
Mouse Razar Mamba Tournament Edition
Keyboard Logitech G910
If Arctic could make an AS5 update I'd buy it in a heartbeat. I'll gladly trade a couple C for a "apply it once, forget it for 8 years" reliability. Thermal Grizzly makes fantastic OC enthusiast products, but those products just aren't as dependable as a "lesser performing" TIM.


I still think motherboards are a huge culprit in behavior anomalies. While I understand a Ryzen 5000 should -not- work stable at -30 CO all core, my setup doesn't work -the way it's supposed to- unless I have it set to -30 CO.
I've been having a hard time looking for absolute top-end benchmark run results, as such I have zero real statistics to compare the quality of my 5800X with.
With that said, my chip just cannot do R23 at 4.95Ghz. I attempted a 14C ambient run at (fully reset/manual control) 1.365v (SVI2 input) and it ran for a whole 7 seconds before a full system crash. WHEA suggests a cache-level fault (i.e. too low voltage for the entire CCD), but I'm suspicious of temp-limit shutdowns due to the op-temp of 88C, who can even know what the real hotspot spikes are.

And to reiterate some previous posts - I still haven't had any idle/low load/light MT issues with my tested setting of 135PPT/100TDC/125EDC at -30CO. I'd expect a real world scenario of multiple RAM heavy browser + network multitasking + virtualization usage + mixed-load gaming to create SoC related errors, as I've seen in literally every PC pushed "too hard" with concurrent low/light MT RAM heavy loads in the past 15 years of my experience. These issues have not happened (yet).
Motherboard... Agesa used and also the memory kits. Not neccesarily 2nd'ary timings but thirds also. Primaries, obviously a pretty narrow selection on all kits such as 16-18-18 ddr4 as an example.

The Whea errors are going to be produced from two things.
Heat = cpu/chipset & memory.

So not going to focus strictly on a cpu temp, the SOC might be the culprit with high temps and whea errors (stability in general).

Just as a note, I try to remind people 70c.

Yes 70c. That's your cpu high temp alert. Designed to prevent a heat soak and throttle on default systems. This temp I recommened for many to focus on while under load.

But that temp is near impossible to obtain without chilling or super low ambient temps.

Transistors.
It's a balancing act.

How low can you adjust vcore before too many transistors stop working and create instability and lower cpu scores?

Why bother.... the vcore is not the main issue. On all 4 generations, self awareness of its running temperature reflects its boost and vcore use.

In theory, you'd use more vcore or soc or vdimm to obtain stability for any given overclock. Be it forced through pbo or not.

The draw back is cooling always.

Why cruncher?? It uses a bit of cache resources more so than CB23 for example. And most cache (memory also) doesn't particularly like overclocking and heat at the same time.

I've seen some posts of very decent overclocking. All they had in common is simply outstanding cooling solutions.
 
Joined
Oct 15, 2011
Messages
1,357 (0.33/day)
Location
Springfield, Vermont
System Name KHR-1
Processor Ryzen 7 5800X
Motherboard ASRock B550 PG Velocita
Memory 32 GB G.Skill Trident Z Neo F4-3600C18D-32GTZN
Video Card(s) Sapphire Nitro+ Radeon RX 6750 XT
Storage Western Digital Black SN850 1 TB NVMe SSD
Display(s) Samsung Odyssey G40B 27-inch
Case Corsair 275R
Audio Device(s) Technics SA-EX140 receiver with Polk VT60 speakers
Power Supply eVGA Supernova G3 750W
Mouse Logitech G Pro (Hero)
Software Windows 11 Pro x64 22H2
I just got my Ryzen 7 5800X today and I saw these values in Ryzen Master when running Cinebench R23 with all cores loaded:

PPT=142W
TDC=95A
EDC=140A

Is this because of an issue with the ASRock B550 PG Velocita? Looks like PBO was set or sort of was set.
 
Joined
Sep 23, 2022
Messages
286 (2.15/day)
System Name Windows | Linux
Processor 7700X | 5800X3D
Motherboard Pro X670-P | B550 Unify
Cooling NH-D15 | NH-D15s
Memory 32GB 6000 CL30 | 32GB 3200 CL14
Video Card(s) PNY RTX 4080 | Red Devil 6900 XT Ult.
Power Supply Prime TX-850 | Prime GX-1000
Software Win 11 Pro 22H2 | Debian Bookworm
I just got my Ryzen 7 5800X today and I saw these values in Ryzen Master when running Cinebench R23 with all cores loaded:

PPT=142W
TDC=95A
EDC=140A

Is this because of an issue with the ASRock B550 PG Velocita? Looks like PBO was set or sort of was set.

Those are the factory defaults for the 5800x (and 5900x/5950x) without PBO. It was pretty over-cooked out of the box. Nothing wrong with your board.

Check out the first post in this thread for some alternate settings to cool it down.
 

freeagent

Moderator
Staff member
Joined
Sep 16, 2018
Messages
5,167 (3.23/day)
Location
Winnipeg, Canada
System Name Street Meat
Processor AMD R7 5900X
Motherboard Asus Strix B550-XE
Cooling Thermalright Frost Commander 140 White
Memory 2x8 G.Skill Trident Z Royal 3200C14, 2x8GB G.Skill Trident Z Black and White 3200 C14
Video Card(s) EVGA RTX 3070 Ti FTW3 Ultra
Storage WD SN850 1TB, SN750 1TB, SN750 500GB, Asus Hyper M.2, 2x Intel 545S 256GB, 1TB Toshiba spinner
Display(s) LG 50UP7100
Case Fractal Torrent Compact RGB
Audio Device(s) JBL 2.1 Deep Bass
Power Supply EVGA G+ 750, Monster HDP1800
Mouse Zowie EC2 Evo
Keyboard Logitech G213
Software Yes
Benchmark Scores Yes
I use PBO differently than everyone else it seems. I find the absolute maximum power that my CPU can draw, Then I change my power limits and test again, and I continue until I cannot get the CPU to draw more power. 235 PPT, 160 TDC, 190 EDC in the case of my 5900X using 1203. And with those numbers I apply my curve, whatever that may be... And I get the full boost, 5150MHz on multiple cores, and on something like Linpack it will run at 4500MHz. For WCG it will run at 4600-4650, same with F@H. No crashes, no errors, nothing. It will run Pi32M on all cores right to their limit, but always fails stock core cycler in the same spot... core 4. I think lol. But I did the same thing with my 5600X and it will pass core cycler at 4800MHz. Fuck that program lol :)
 
Joined
Oct 15, 2011
Messages
1,357 (0.33/day)
Location
Springfield, Vermont
System Name KHR-1
Processor Ryzen 7 5800X
Motherboard ASRock B550 PG Velocita
Memory 32 GB G.Skill Trident Z Neo F4-3600C18D-32GTZN
Video Card(s) Sapphire Nitro+ Radeon RX 6750 XT
Storage Western Digital Black SN850 1 TB NVMe SSD
Display(s) Samsung Odyssey G40B 27-inch
Case Corsair 275R
Audio Device(s) Technics SA-EX140 receiver with Polk VT60 speakers
Power Supply eVGA Supernova G3 750W
Mouse Logitech G Pro (Hero)
Software Windows 11 Pro x64 22H2
I use PBO differently than everyone else it seems. I find the absolute maximum power that my CPU can draw, Then I change my power limits and test again, and I continue until I cannot get the CPU to draw more power. 235 PPT, 160 TDC, 190 EDC in the case of my 5900X using 1203. And with those numbers I apply my curve, whatever that may be... And I get the full boost, 5150MHz on multiple cores, and on something like Linpack it will run at 4500MHz. For WCG it will run at 4600-4650, same with F@H. No crashes, no errors, nothing. It will run Pi32M on all cores right to their limit, but always fails stock core cycler in the same spot... core 4. I think lol. But I did the same thing with my 5600X and it will pass core cycler at 4800MHz. Fuck that program lol :)
My 5600X seems to be a dud, as Chromium was randomly giving me access violation error reports, including when just scrolling down a web page, IIRC. I was randomly getting crashed tabs, when higher than 4.7 for single-core boosting. :(

Your 5600X seems to be a golden sample, compared to mine.
 

freeagent

Moderator
Staff member
Joined
Sep 16, 2018
Messages
5,167 (3.23/day)
Location
Winnipeg, Canada
System Name Street Meat
Processor AMD R7 5900X
Motherboard Asus Strix B550-XE
Cooling Thermalright Frost Commander 140 White
Memory 2x8 G.Skill Trident Z Royal 3200C14, 2x8GB G.Skill Trident Z Black and White 3200 C14
Video Card(s) EVGA RTX 3070 Ti FTW3 Ultra
Storage WD SN850 1TB, SN750 1TB, SN750 500GB, Asus Hyper M.2, 2x Intel 545S 256GB, 1TB Toshiba spinner
Display(s) LG 50UP7100
Case Fractal Torrent Compact RGB
Audio Device(s) JBL 2.1 Deep Bass
Power Supply EVGA G+ 750, Monster HDP1800
Mouse Zowie EC2 Evo
Keyboard Logitech G213
Software Yes
Benchmark Scores Yes
when higher than 4.7 for single-core boosting.
You probably just needed to raise your power limits, not mobo but your own in the advanced settings.


Edit:

I have some crap Adatas in there right now, but I still have it set to boost at 4850 even though it failed core cycler lol. It passed at 4800. I will run it again just to make sure i am not full of you know what :)

Edit again...

No problems either and it's been a couple of years now
 
Joined
Oct 15, 2011
Messages
1,357 (0.33/day)
Location
Springfield, Vermont
System Name KHR-1
Processor Ryzen 7 5800X
Motherboard ASRock B550 PG Velocita
Memory 32 GB G.Skill Trident Z Neo F4-3600C18D-32GTZN
Video Card(s) Sapphire Nitro+ Radeon RX 6750 XT
Storage Western Digital Black SN850 1 TB NVMe SSD
Display(s) Samsung Odyssey G40B 27-inch
Case Corsair 275R
Audio Device(s) Technics SA-EX140 receiver with Polk VT60 speakers
Power Supply eVGA Supernova G3 750W
Mouse Logitech G Pro (Hero)
Software Windows 11 Pro x64 22H2
You probably just needed to raise your power limits, not mobo but your own in the advanced settings.
That was with CTR under Windows 10. I was at a loss at what to change. I loved how fast it seemed, then boom, a sudden error report!
 
Joined
Jun 27, 2011
Messages
6,352 (1.50/day)
Processor Ryzen 5950x
Motherboard Asus Tuf Gaming x570-plus
Cooling Scythe Mugen 5 Rev B with Noctua push pull fans
Memory Gskill 4x16gb 3600mhz Cas 16
Video Card(s) MSI RX 6750 XT MECH 2X 12G OC
Storage 2x Adata SX8200 1tb
Display(s) Asus 144hz
Power Supply Cougar 850w
Software Windows 10 64 Bit
I use PBO differently than everyone else it seems. I find the absolute maximum power that my CPU can draw, Then I change my power limits and test again, and I continue until I cannot get the CPU to draw more power. 235 PPT, 160 TDC, 190 EDC in the case of my 5900X using 1203. And with those numbers I apply my curve, whatever that may be... And I get the full boost, 5150MHz on multiple cores, and on something like Linpack it will run at 4500MHz. For WCG it will run at 4600-4650, same with F@H. No crashes, no errors, nothing. It will run Pi32M on all cores right to their limit, but always fails stock core cycler in the same spot... core 4. I think lol. But I did the same thing with my 5600X and it will pass core cycler at 4800MHz. Fuck that program lol :)
I do something like that when I want maximum performance regardless of power consumption. I have tested at a variety of settings.
 
Joined
Oct 15, 2011
Messages
1,357 (0.33/day)
Location
Springfield, Vermont
System Name KHR-1
Processor Ryzen 7 5800X
Motherboard ASRock B550 PG Velocita
Memory 32 GB G.Skill Trident Z Neo F4-3600C18D-32GTZN
Video Card(s) Sapphire Nitro+ Radeon RX 6750 XT
Storage Western Digital Black SN850 1 TB NVMe SSD
Display(s) Samsung Odyssey G40B 27-inch
Case Corsair 275R
Audio Device(s) Technics SA-EX140 receiver with Polk VT60 speakers
Power Supply eVGA Supernova G3 750W
Mouse Logitech G Pro (Hero)
Software Windows 11 Pro x64 22H2
I've got a 5800x, played around a bit with it.
PPT 110W
TDC 80A
EDC 100A
Your ampere and watt settings made Cinebench R23 multicore score drop!
 

Mussels

Freshwater Moderator
Staff member
Joined
Oct 6, 2004
Messages
56,281 (8.41/day)
Location
Oystralia
System Name Rainbow Sparkles (Power efficient, <350W gaming load)
Processor Ryzen R7 5800x3D (Undervolted, 4.45GHz all core)
Motherboard Asus x570 Gaming-F
Cooling Alphacool Apex UV - Alphacool Eisblock XPX Aurora + EK Quantum ARGB 3090 w/ active backplate
Memory 2x32GB DDR4 3600 Corsair Vengeance RGB @3866 C18-22-22-22-42 TRFC704 (1.4V Hynix MJR - SoC 1.175V)
Video Card(s) Galax RTX 3090 SG 24GB: Underclocked to 1700Mhz 0.737v (375W down to 250W))
Storage 2TB WD SN850 NVME + 1TB Sasmsung 970 Pro NVME + 1TB Intel 6000P NVME USB 3.2
Display(s) Kogan 32" 4K 70Hz + Gigabyte G32QC (1440p 165Hz) + Phillips 328m6fjrmb (1440p 144Hz)
Case Fractal Design R6
Audio Device(s) Logitech G560 | Corsair Void pro RGB |Blue Yeti mic
Power Supply Fractal Ion+ 2 860W (Platinum) (This thing is God-tier. Silent and TINY)
Mouse Logitech G Pro wireless + Steelseries Prisma XL
Keyboard Razer Huntsman TE (custom white and steel keycaps)
VR HMD Oculus Rift S + Quest 2
Software Windows 11 pro x64 (Yes, it's genuinely a good OS) OpenRGB - ditch the branded bloatware!
I use PBO differently than everyone else it seems. I find the absolute maximum power that my CPU can draw, Then I change my power limits and test again, and I continue until I cannot get the CPU to draw more power. 235 PPT, 160 TDC, 190 EDC in the case of my 5900X using 1203. And with those numbers I apply my curve, whatever that may be... And I get the full boost, 5150MHz on multiple cores, and on something like Linpack it will run at 4500MHz. For WCG it will run at 4600-4650, same with F@H. No crashes, no errors, nothing. It will run Pi32M on all cores right to their limit, but always fails stock core cycler in the same spot... core 4. I think lol. But I did the same thing with my 5600X and it will pass core cycler at 4800MHz. Fuck that program lol :)
the 8 core chiplet CPU's dont work that way, the heat density is too much

the 5600x and 5900x are 6 cores each and far, FAR easier to cool

Your ampere and watt settings made Cinebench R23 multicore score drop!
That's kind of the point - you drop the heat and temps at the loss of some benchmark points that mean nothing in reality.
You might be faster in a 5 minute R23 run, but in real world use that lasts longer a heatsoaked cooler would result in a performance drop at some point in time

You also usually want to combine the curve undervolting with the limited PBO settings, you'll need to take time to tweak those values
 
Top