• 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

freeagent

Moderator
Staff member
Joined
Sep 16, 2018
Messages
7,546 (3.68/day)
Location
Winnipeg, Canada
Processor AMD R9 5900X
Motherboard Asus Crosshair VIII Dark Hero
Cooling Thermalright Aqua Elite 360 V3 1x TL-B12, 2x TL-C12 Pro, 2x TL K12
Memory 2x8 G.Skill Trident Z Royal 3200C14, 2x8GB G.Skill Trident Z Black and White 3200 C14
Video Card(s) Zotac 4070 Ti Trinity OC
Storage WD SN850 1TB, SN850X 2TB, Asus Hyper M.2, 2x SN770 1TB
Display(s) LG 50UP7100
Case Fractal Torrent Compact RGB
Audio Device(s) JBL 2.1 Deep Bass
Power Supply EVGA SuperNova 750w G+, Monster HDP1800
Mouse Logitech G502 Hero
Keyboard Logitech G213
VR HMD Oculus 3
Software Yes
Benchmark Scores Yes
215 EDC is a lot, more than I use on my 5900X. For 5950 I could see it..

I don't understand why my 5600X3D chip runs so cold. I thought for sure with x3d cache it would run hotter, but I never see it break 59 celsius in games, even very demanding x3d cache hungry games.
Yup, nice and cool :)

occt.PNG
 
Joined
Jul 30, 2019
Messages
2,379 (1.37/day)
System Name Not a thread ripper but pretty good.
Processor Ryzen 9 5950x
Motherboard ASRock X570 Taichi (revision 1.06, BIOS/UEFI version P5.50)
Cooling EK-Quantum Velocity, EK-Quantum Reflection PC-O11, EK-CoolStream PE 360, XSPC TX360
Memory Micron DDR4-3200 ECC Unbuffered Memory (4 sticks, 128GB, 18ASF4G72AZ-3G2F1)
Video Card(s) XFX Radeon RX 5700 & EK-Quantum Vector Radeon RX 5700 +XT & Backplate
Storage Samsung 2TB 980 PRO 2TB Gen4x4 NVMe, 2 x Samsung 2TB 970 EVO Plus Gen3x4 NVMe
Display(s) 2 x 4K LG 27UL600-W (and HUANUO Dual Monitor Mount)
Case Lian Li PC-O11 Dynamic Black (original model)
Power Supply Corsair RM750x
Mouse Logitech M575
Keyboard Corsair Strafe RGB MK.2
Software Windows 10 Professional (64bit)
Benchmark Scores Typical for non-overclocked CPU.
That is the big problem. Been discussing this with @ir_cow and motherboard reviewers on how they test VRMs and the answer came back that it's bascially a nightmare

Some VRM's can safely run at 120C, others thermal throttle at 80C.
Some can handle 50A at any temp, others are 80A at 25C and 20A at 80C.

On many boards, the hotspots are where you expect - others it must seep away through the traces in the PCB and heat up seemingly unrelated areas, especially in a blank spot between those heat sources where that heat may simply accumulate due to lack of airflow


@ir_cow made the silly mistake of getting me thinking on how to judge motherboards completely and fairly, so here's an in-depth teardown on that board above and why so many AM4 users had issues they couldn't resolve without a board swap.
This is that board with it's high VRM temps and shoddy marketing, but tended to be actually liked by it's owners. How can a high VRM temp board with weird claims not be trash?
I almost got that board.

I mean honestly, look at the marketing for this - DDR4 4666+ on AM4?
10 phase power design (later called 8+2) that's downright false
View attachment 308797
The dishonesty there sucks. I've never trusted any vendors ram OC claims although it's easy enough to look at the or RAM QVL and see what they are willing to admit they have tested.

We've got a dedicated thread to trying to get people to 4000 stable, let alone that huge number.
Quad crossfire?? I'm sensing a basic math problem with that
View attachment 308798

VRM's look beefy, but its a basic 4+1 setup (general rule is inductors divided by 2)
They say 8+2, but this is when people started to get annoyed if they failed to mention doublers in marketing
View attachment 308800
Next slide is about "premium" 60A chokes
View attachment 308801
Cheating a little here since someone else identified the parts for me but throwing in my interpretations on top.

A quick google on this board shows it has a PWM controller that can only do 7 phases

That article covers that it's been used in a 4+2 (doubled) setup
The x570 Taichi that people loved had the same VRM's but used 10 of them (and four weaker ones for the SoC)in a 6+2 setup covered here on TPU.
I have the x570, Taichi should I be worried?

Problem is that most VRM's are a rated a little dishonestly as Asrock says they're 60A, but the spec sheet says 50A - with 55A burst capabilities for only 10 milliseconds while the spec sheets show a realistic limit of 45A before they start pumping out heat

These are at least safe to high temperatures, showing they can handle upto 110C problem free - as long as you're under 50A. Full power, always? Nice.
View attachment 308804

The efficiency chart shows why that limit exists - because passing 20A, the sweet spot is gone - and past 30A you're at 15% of your power thrown to heat.
View attachment 308805

AM4's default PBO *on* limits are TDC: 95A and EDC: 142A with PPT: 142W
(At some point an "EDC" bug was introduced in AGESA that stopped boards passing these amperage limits, my guess as a failed safety limit - that's been corrected)

8 of those VRM's at AM4's stock 142A with a 5800x would be 17.75A each, quite happily in that sweet spot... before board makers bad decisions get involved with BIOS settings.

5800x uses the same power limit as 5950x with double the cores - so in theory if nothing like thermals or BIOS settings limited you a 5950x could use double the values of a 5800x.
(Same voltage, double amperage and wattage to twice as many cores)
But thermal density is an issue so doubling the limits isn't practical and won't work anyway under "normal" cooling.

My 5800x on an Aourus B550-i ITX ran at 130A in R20 and R23, while kitgurus review of a 5950x with unlimited PBO went upto 215A - close to that theory
View attachment 308808
You know my story. I will likely never trust a boards ability to do unlimited PBO ever again or Ryzen Master's stupid defaults. :banghead: That darn x470 Master SLI/ac was such a good stable board too it's a shame I made such a stupid error and destroyed it.

From an engineering perspective, that board gets a pass - that'd be very hard to overheat to 120C, and the VRM's should in theory handle even a power unlocked 5950x.
215A / 8 = 26.875W, within the VRM's happy place let alone their 50A maximum.

But you see 90C and you think oh god it's bad, because OTHER vrm's overheat at 80C or less
I'll likely never have the technical know how to understand what a VRM's limits are so I just have to take previous advice I have read that says keep your VRM's under 80c and you will probably be fine.

Ironically this chart shows an x570 steel legend with another 32c left on it's VRMs going by it's sensor readings while some of these others may be toeing the line to overheating or already throttling the CPU
View attachment 308809

As another metric that's ignored, the inefficiency of designs can lead to extra power wasted as heat - a 30W difference is huge for a small CPU like a Ryzen 2600
View attachment 308810


I get the feeling the checklist to actually test boards is a relatively large one with everything needing to be tested simultaneously to be useful

1. VRM specs, especially the temperature they start to lose efficiency
Not sure how or if this could happen but it would be nice if CPU's could coordinate better with VRM's to hit the best efficiency the board can provide.

2. All testing done with the highest power consumption CPU you can get, to find a boards actual limits
3. Also testing within recommended limits - AMD's stock PBO and Intels PL1/PL2 targets.

R23 test runs would need all the relevant data slapped together in one graph somehow:
  1. CPU's effective sustained clock speed (indicator of throttling)
    R23 score (performance loss vs competitors)
  2. Maximum amps drawn by CPU (HWinfo)
  3. Wattage used by system (Wall reading)
  4. VRM temps (and their rated maximums)


I also just ran into an issue on my ITX system where "PBO disabled" boosts harder than PBO with maxed out settings because ?!?!?
Crap like that is why people have issues with 90c at stock
I just started playing with PBO and CO again last Friday. (and re-read all 37 pages of this thread to prep for it) I was comparing my two systems after the chip swap and noticed my Taichi x570 boosts a bit harder on defaults than my Asrock B550 PG-ITX. (not sure if the VRM's on that board are crap or not but looked better then their ITX offerings for b450, x470, and cheaper B550) Of course I also found out my Power Reporting Deviation was reporting a bit over 120% on full load on my ITX. It didn't always do that so not sure when it started happening, perhaps a UEFI/BIOS update issue. Had to go and manually set PPT, TDC, and EDC to defaults to get it to behave, then I was like ok I might as well try PBO again (and CO now) and see what I can do. I managed to get my CPU temps down a bit (my ITX is on air) and peaked a CB R20 score around 10,000 but the clocks reporting didn't seem to get on par with the Taichi and something happened where I started to get inconsistent results so I have to start all over again from scratch and be a bit more careful in the setup and observations. Also I reckon my ITX being on NH-C14S might be thermally limiting me on that system vs. my water cooled x570 system so the idea I'm going to get some equality between the two systems when both are eventually finely tuned is a bit unrealistic but it was nice to see what appeared to be CO bringing my ITX cpu temps down on my 5950x with a -20 CO on all cores except the first which was -30. (this was on my previously well toasted chip)

1691982796466.png
 
Last edited:

Mussels

Freshwater Moderator
Staff member
Joined
Oct 6, 2004
Messages
58,413 (8.18/day)
Location
Oystralia
System Name Rainbow Sparkles (Power efficient, <350W gaming load)
Processor Ryzen R7 5800x3D (Undervolted, 4.45GHz all core)
Motherboard Asus x570-F (BIOS Modded)
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 1.15V)
Video Card(s) Galax RTX 3090 SG 24GB: Underclocked to 1700Mhz 0.750v (375W down to 250W))
Storage 2TB WD SN850 NVME + 1TB Sasmsung 970 Pro NVME + 1TB Intel 6000P NVME USB 3.2
Display(s) Phillips 32 32M1N5800A (4k144), LG 32" (4K60) | Gigabyte G32QC (2k165) | Phillips 328m6fjrmb (2K144)
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 ( Sexy white 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!
Benchmark Scores Nyooom.
215 EDC is a lot, more than I use on my 5900X. For 5950 I could see it..
Sure is, i just wanted to see the maximum - the worst case any board needs to be able to handle.

My ITX board seems to have broken BIOS settings and PBO disabled just disables all the safeties instead and flatlines to 90c, so i can absolutely see that happening to some poor bastard with a 5950x and something going pop

Power Reporting Deviation was reporting a bit over 120% on full load on my ITX
curve undervolt causes this

I have the x570, Taichi should I be worried?
No, i covered why it's weaker VRM friend is more than enough - it's overkill for AM4.

I'll likely never have the technical know how to understand what a VRM's limits are so I just have to take previous advice I have read that says keep your VRM's under 80c and you will probably be fine.
One user i'm helping in PM's, his VRM's are worthless at 80c

His board has a 3+1 setup of these

The main takeaway is that these are only capable of ~11A each at 80C or 15A at 25C (which they'll never be at)
I need to learn to interpret some of this, but i highlighted the ones i know are important to why the asus B450M series was considered trash tier

At 80C (which they reached fairly easily) the phases could only output around 11.2A each or 16.8A for under 10 seconds - compare that to the 50A each at 120C on the asrock boards of the same era.
1691995969754.png


This is most likely a 4+2 setup with doublers (8+4)
1691996140591.png


So at 80C for 10 seconds you can get 8x16.8A for 134A total - notice how this could choke even on a 5800x (Mine runs 140A at stock), let alone the 210A figure that 5950x threw out?
For greater than 10 seconds you're capped to 8x11.2 at 89A - which can JUST run a stock 65W chip since 90A is their default EDC limit.

Thing is these could easily get hotter than 80c since they're passively cooled and not really capable of handling anything beyond a 65W CPU, and they'd only get worse with an AIO reducing airflow or GPU heat added on.
Heck even high ambients like an Aussie or texas summer would ruin this boards day.



As long as the VRM's are documented, we can find out some basics (like the max temperatures) and at least figure out what CPU's they're rated for.
I'm definitely seeing the worst boards were designed without the actual CPU's so things like aiming for 90A - and then marketing wanting to claim it supports a 5950x.

The BIOS support for that CPU Says it's good to go and a mere 105W CPU - not 142W.
How hard would it have been to actually put a 105W default PBO limit in the BIOS for these and stop all the problems?
1691998429169.png
 
Last edited:

tabascosauz

Moderator
Supporter
Staff member
Joined
Jun 24, 2015
Messages
7,593 (2.35/day)
Location
Western Canada
System Name ab┃ob
Processor 7800X3D┃5800X3D
Motherboard B650E PG-ITX┃X570 Impact
Cooling NH-U12A + T30┃AXP120-x67
Memory 64GB 6000CL30┃32GB 3600CL14
Video Card(s) RTX 4070 Ti Eagle┃RTX A2000
Storage 8TB of SSDs┃1TB SN550
Display(s) 43" QN90B / 32" M32Q / 27" S2721DGF
Case Caselabs S3┃Lazer3D HT5
The BIOS support for that CPU Says it's good to go and a mere 105W CPU - not 142W.
How hard would it have been to actually put a 105W default PBO limit in the BIOS for these and stop all the problems?
View attachment 308934

The CPU list is just listing TDP. Of course it's 105W=142W.

Unheatsinked 4-phase (doubled lo-side or not) is pretty unconfident even with 65W (88W) CPUs if ambient temp is higher and bad airflow, so to actually be safe probably it'd have to go down to 65W (76W). RDSon are not identical between Sinopower SM4336/4337, SiRA12/14, 4C06/4C10 but in unheatsinked 4-phase they are all equally trash.

I used to say that 4-phase doubled lo-side discretes with a good heatsink comprising lower RDSon discretes would be good, but bad implementations of good discretes like the Carbon put an end to that.

I think at this point we can just conclude that the only boards that will 100% confidently handle 142W are 6+ phase DrMOS boards, no exceptions. Which in X470's case just leaves 9 non-ITX boards (counting the Taichis which use high phase count dual N-fets), and in B450's case leaves a whopping total of 1 non-ITX board (itself nothing more than a PCH-swapped B550M TUF so not even a contemporary).

I'll likely never have the technical know how to understand what a VRM's limits are so I just have to take previous advice I have read that says keep your VRM's under 80c and you will probably be fine.

Assuming they were gracious enough to give you a software sensor, this doesn't solve a fire hazard VRM because:
  • How do you know it's an accurate sensor?
  • Is it even placed in the Vcore power plane? Or is it situated somewhere useless like in the middle of the SOC plane?
  • Where amongst the Vcore phases is it placed? Location matters.
  • Bad AM4 boards are always 4-layer boards, and lower layer count magnifies PCB temps and hotspots.
I just started playing with PBO and CO again last Friday. (and re-read all 37 pages of this thread to prep for it) I was comparing my two systems after the chip swap and noticed my Taichi x570 boosts a bit harder on defaults than my Asrock B550 PG-ITX. (not sure if the VRM's on that board are crap or not but looked better then their ITX offerings for b450, x470, and cheaper B550) Of course I also found out my Power Reporting Deviation was reporting a bit over 120% on full load on my ITX. It didn't always do that so not sure when it started happening, perhaps a UEFI/BIOS update issue. Had to go and manually set PPT, TDC, and EDC to defaults to get it to behave, then I was like ok I might as well try PBO again (and CO now) and see what I can do. I managed to get my CPU temps down a bit (my ITX is on air) and peaked a CB R20 score around 10,000 but the clocks reporting didn't seem to get on par with the Taichi and something happened where I started to get inconsistent results so I have to start all over again from scratch and be a bit more careful in the setup and observations. Also I reckon my ITX being on NH-C14S might be thermally limiting me on that system vs. my water cooled x570 system so the idea I'm going to get some equality between the two systems when both are eventually finely tuned is a bit unrealistic but it was nice to see what appeared to be CO bringing my ITX cpu temps down on my 5950x with a -20 CO on all cores except the first which was -30. (this was on my previously well toasted chip)

C14S is not really capable of sustaining Ryzens whether 1CCD or 2CCD past about 160W for 12-core, maybe closer to 180W for 16-core. Even with good airflow, on an open bench, let alone in a NR200. You can run your 5950X but stock is about it. It simply does not have the chops, you'll need water or a dual tower for playing with higher PBO limits.

215 EDC is a lot, more than I use on my 5900X. For 5950 I could see it..

Yup, nice and cool :)

4 watts of SOC???? What the hell??
 

freeagent

Moderator
Staff member
Joined
Sep 16, 2018
Messages
7,546 (3.68/day)
Location
Winnipeg, Canada
Processor AMD R9 5900X
Motherboard Asus Crosshair VIII Dark Hero
Cooling Thermalright Aqua Elite 360 V3 1x TL-B12, 2x TL-C12 Pro, 2x TL K12
Memory 2x8 G.Skill Trident Z Royal 3200C14, 2x8GB G.Skill Trident Z Black and White 3200 C14
Video Card(s) Zotac 4070 Ti Trinity OC
Storage WD SN850 1TB, SN850X 2TB, Asus Hyper M.2, 2x SN770 1TB
Display(s) LG 50UP7100
Case Fractal Torrent Compact RGB
Audio Device(s) JBL 2.1 Deep Bass
Power Supply EVGA SuperNova 750w G+, Monster HDP1800
Mouse Logitech G502 Hero
Keyboard Logitech G213
VR HMD Oculus 3
Software Yes
Benchmark Scores Yes

tabascosauz

Moderator
Supporter
Staff member
Joined
Jun 24, 2015
Messages
7,593 (2.35/day)
Location
Western Canada
System Name ab┃ob
Processor 7800X3D┃5800X3D
Motherboard B650E PG-ITX┃X570 Impact
Cooling NH-U12A + T30┃AXP120-x67
Memory 64GB 6000CL30┃32GB 3600CL14
Video Card(s) RTX 4070 Ti Eagle┃RTX A2000
Storage 8TB of SSDs┃1TB SN550
Display(s) 43" QN90B / 32" M32Q / 27" S2721DGF
Case Caselabs S3┃Lazer3D HT5

freeagent

Moderator
Staff member
Joined
Sep 16, 2018
Messages
7,546 (3.68/day)
Location
Winnipeg, Canada
Processor AMD R9 5900X
Motherboard Asus Crosshair VIII Dark Hero
Cooling Thermalright Aqua Elite 360 V3 1x TL-B12, 2x TL-C12 Pro, 2x TL K12
Memory 2x8 G.Skill Trident Z Royal 3200C14, 2x8GB G.Skill Trident Z Black and White 3200 C14
Video Card(s) Zotac 4070 Ti Trinity OC
Storage WD SN850 1TB, SN850X 2TB, Asus Hyper M.2, 2x SN770 1TB
Display(s) LG 50UP7100
Case Fractal Torrent Compact RGB
Audio Device(s) JBL 2.1 Deep Bass
Power Supply EVGA SuperNova 750w G+, Monster HDP1800
Mouse Logitech G502 Hero
Keyboard Logitech G213
VR HMD Oculus 3
Software Yes
Benchmark Scores Yes
Joined
Jan 18, 2012
Messages
237 (0.05/day)
Location
Somewhere in the middle of my universe
System Name The Dark side of the room
Processor AMD Ryzen 9 5900X
Motherboard MSI MEG X570 Unify
Cooling Custom loop watercooling (Bykski CPU-XPR-POM-M-V2, Alphacool Eisblock GPX, Freezemod PU-PWM5B18W)
Memory GSkill Ripjaws V DDR4 3600 CL16 (4 x 16GB)
Video Card(s) XFX Speedster QICK 319 Radeon RX 6700 XT
Storage 1 x Kingston KC3000 1024GB + 2 x Kingston NV2 2TB
Display(s) LG 34WP65C Ultrawide 3440x1440 @ 160Hz freesync premium
Case Thermaltake Core P90 TG
Audio Device(s) onboard Realtek® ALC1220 with Logitech Z906
Power Supply MSI MAG A850GF 80 Plus Gold
Mouse Generic
Keyboard Trust Asta GTX 865
Software Windows 11 pro
Benchmark Scores It's a form of exhibitionism...;-), but fun Software programs just have to run nicely....
Eventhough I (yet) do not fully comprehend the interaction between PPT, TDC, EDC, the AMD PBO & CO-algorithem, it is only logical that for playing with these values you need a solid and stable base.
I always say: you can build a house but first you need a proper foundation. If not, in time things will sag, bend and/or crack. Same goes for a PC system.

That is why at the time I purchased my MB, I specifically looked at reviews which went deeper than only connectivity, extra's, looks and rainbow vomit. I liked the MSI MEG X570 Unify very much for it's features and looks, but was hesitant due to the fact that the first (affordable ;-) ) MSI X570 boards where horrible on the VRM's regarding to the reviews.

But then I found this review which took away my worries. They actually have quite a lot in depht video's which might be useful.
Seemed that MSI learned from their faillure at that time.

For me personally up to this day a decision I do not regret, a great mb with everything I need without having to buy a really high end one. When friends ask me for building advise, I always tell them not to cheep out on the MB and do some research first, especially when the're on a budget and want to have headroom for future upgrades.

My settings atm are:
  • PPT 210
  • TDC 140
  • EDC 165
  • CO -30 all core
  • PBO + 100 MHz boost overdrive
everything stable, low temps and decent bench scores on CB R23 (22.670 MT & 1657 ST) for a R9 5900X.

Will be playing again with the values, but still have a lot of reading to do regarding that. Guess it's true when the say: 'you're never to old to learn'
 
Joined
Oct 15, 2011
Messages
1,982 (0.43/day)
Location
Springfield, Vermont
System Name KHR-1
Processor Ryzen 9 5900X
Motherboard ASRock B550 PG Velocita (UEFI-BIOS P3.40)
Memory 32 GB G.Skill RipJawsV F4-3200C16D-32GVR
Video Card(s) Sapphire Nitro+ Radeon RX 6750 XT
Storage Western Digital Black SN850 1 TB NVMe SSD
Display(s) Alienware AW3423DWF OLED-ASRock PG27Q15R2A (backup)
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 23H2

freeagent

Moderator
Staff member
Joined
Sep 16, 2018
Messages
7,546 (3.68/day)
Location
Winnipeg, Canada
Processor AMD R9 5900X
Motherboard Asus Crosshair VIII Dark Hero
Cooling Thermalright Aqua Elite 360 V3 1x TL-B12, 2x TL-C12 Pro, 2x TL K12
Memory 2x8 G.Skill Trident Z Royal 3200C14, 2x8GB G.Skill Trident Z Black and White 3200 C14
Video Card(s) Zotac 4070 Ti Trinity OC
Storage WD SN850 1TB, SN850X 2TB, Asus Hyper M.2, 2x SN770 1TB
Display(s) LG 50UP7100
Case Fractal Torrent Compact RGB
Audio Device(s) JBL 2.1 Deep Bass
Power Supply EVGA SuperNova 750w G+, Monster HDP1800
Mouse Logitech G502 Hero
Keyboard Logitech G213
VR HMD Oculus 3
Software Yes
Benchmark Scores Yes

tabascosauz

Moderator
Supporter
Staff member
Joined
Jun 24, 2015
Messages
7,593 (2.35/day)
Location
Western Canada
System Name ab┃ob
Processor 7800X3D┃5800X3D
Motherboard B650E PG-ITX┃X570 Impact
Cooling NH-U12A + T30┃AXP120-x67
Memory 64GB 6000CL30┃32GB 3600CL14
Video Card(s) RTX 4070 Ti Eagle┃RTX A2000
Storage 8TB of SSDs┃1TB SN550
Display(s) 43" QN90B / 32" M32Q / 27" S2721DGF
Case Caselabs S3┃Lazer3D HT5

freeagent

Moderator
Staff member
Joined
Sep 16, 2018
Messages
7,546 (3.68/day)
Location
Winnipeg, Canada
Processor AMD R9 5900X
Motherboard Asus Crosshair VIII Dark Hero
Cooling Thermalright Aqua Elite 360 V3 1x TL-B12, 2x TL-C12 Pro, 2x TL K12
Memory 2x8 G.Skill Trident Z Royal 3200C14, 2x8GB G.Skill Trident Z Black and White 3200 C14
Video Card(s) Zotac 4070 Ti Trinity OC
Storage WD SN850 1TB, SN850X 2TB, Asus Hyper M.2, 2x SN770 1TB
Display(s) LG 50UP7100
Case Fractal Torrent Compact RGB
Audio Device(s) JBL 2.1 Deep Bass
Power Supply EVGA SuperNova 750w G+, Monster HDP1800
Mouse Logitech G502 Hero
Keyboard Logitech G213
VR HMD Oculus 3
Software Yes
Benchmark Scores Yes
I'm at like 7W at 1.03V, and that's a good result because on my Unify-X was like 10W
Well to be fair fclk and mems are only at 1600, so it’s not really doing much :confused:
 

tabascosauz

Moderator
Supporter
Staff member
Joined
Jun 24, 2015
Messages
7,593 (2.35/day)
Location
Western Canada
System Name ab┃ob
Processor 7800X3D┃5800X3D
Motherboard B650E PG-ITX┃X570 Impact
Cooling NH-U12A + T30┃AXP120-x67
Memory 64GB 6000CL30┃32GB 3600CL14
Video Card(s) RTX 4070 Ti Eagle┃RTX A2000
Storage 8TB of SSDs┃1TB SN550
Display(s) 43" QN90B / 32" M32Q / 27" S2721DGF
Case Caselabs S3┃Lazer3D HT5
Well to be fair fclk and mems are only at 1600, so it’s not really doing much :confused:

dang I need jedec to get to 3W

you are a wizard after all :)
 

Mussels

Freshwater Moderator
Staff member
Joined
Oct 6, 2004
Messages
58,413 (8.18/day)
Location
Oystralia
System Name Rainbow Sparkles (Power efficient, <350W gaming load)
Processor Ryzen R7 5800x3D (Undervolted, 4.45GHz all core)
Motherboard Asus x570-F (BIOS Modded)
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 1.15V)
Video Card(s) Galax RTX 3090 SG 24GB: Underclocked to 1700Mhz 0.750v (375W down to 250W))
Storage 2TB WD SN850 NVME + 1TB Sasmsung 970 Pro NVME + 1TB Intel 6000P NVME USB 3.2
Display(s) Phillips 32 32M1N5800A (4k144), LG 32" (4K60) | Gigabyte G32QC (2k165) | Phillips 328m6fjrmb (2K144)
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 ( Sexy white 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!
Benchmark Scores Nyooom.
The CPU list is just listing TDP. Of course it's 105W=142W.
But that's what the motherboards tend to assume they're actually running - instead of it being a time limited boost, they just run that boost at all times by default.

Uh oh! I wouldn't trust that power configuration on anything more than a Ryzen 5!
Doublers aren't neccesarily bad, it's really about the VRM's themselves.

Some can provide their full power under 25c and less than half at 80c, while some provide their full power until 120c
I'd trust those ones that work at 120c over any other, even if i did my best to keep them a lot colder than that.

I wanted to know how you managed to get down to 4 watts of SOC power at 1.0V lol
The RAM itself matters - more ranks/sticks uses more power. Figured that out when my SoC wattage dropped when i removed my dummy LED sticks.
VENGEANCE RGB PRO Light Enhancement Kit — Black

But then I found this review which took away my worries. They actually have quite a lot in depht video's which might be useful.
They did indeed learn and fixed this board, then went back to their budget crap shenanigans right after with some bad B550 and intel boards.
Throwing my interpretations on top and pasting his details here for those just reading.

He covers it at the 10 minute mark, in detail like always.

He seperates the VRMs into CPU and SoC and explains it's an 'old 12 phase design'
The power controller is an IR35201 an 8 phase or 6+2 controller according to it's datasheet "(8+0/7+1/ 6+2)"
The doublerd are 6 IR3599-3599ADS-1.8 "able to drive up to four independent power stages from one PWM input" on the back of the PCB

6 phases doubled to 12 on the CPU fits that description, with the remaining 2 for the SoC.
Because it's got an active doubler it's aware of what it's doing and avoids the problems that caused users to be come aware of doublers in the first place - where they were used as dumb devices and resulted in a lot of inefficient heat waste and voltage spikes/droops.

He calls it a 6+2 x2 - I like that description
"It's still a full 12 phases, just with a slower transient response time"
1692773469152.png


The doublers are on the back of the board, not the front.
They're basically used as timers to choose which phase the power gets sent to, fairly dumb devices overall.
This isn't bad like the designs that made users aware of VRM doublers because the PWM controller itself is *aware* it has doublers and can send a signal to control which of the phases the power goes to - it can send to group A or group B as it wishes.
1692777412593.png

The downside is a small delay in power being received, so LLC can be slower to catch up.
This is why boards send slightly more voltage than needed to a CPU by design, to cover up that delay safely - or just bigger capacitors.


Something new I just learned - I love him for this stuff - is that this socket is missing a bunch of capacitors on the back and he calls that out for being part of the reason the transient responses were slow on the board. Lower wattages/LLC levels simply don't need them all - but it's an interesting fact to look for on boards.
1692774077777.png


Apparently, Gigabyte fill them all out while MSI use around half of them - heres an Asus Tuf board (a budget one) for comparison
1692774335397.png

These can smooth out voltage overshoot spikes, but buildzoid does claim it may only be 5 millivolts or so and overall not that important. Just one of many small details that get overlooked.


skipping on, the power stages are the power stages, which are IR3555's with a 60A max rating
These are multiple components in one, which is why theres less parts here than some boards.
"IR3555 powIRstage (has high side MOSFET, lowside MOSFET and driver inside"
He mentions a datasheet that I can't find, which is unfortunate. He mentions the 3575 is the same thing with a metal tab integrated for better cooling and that they were the best of the best of their era.
Googling the 3575 takes you to a spec sheet and the name of that spec sheet is... the IR3550. Derp?
IR3550 (infineon.com)

Heres the power loss at various Amperage outputs so you can figure out the heat levels you're dealing with
1692779128250.png

Any given amperage can be at various CPU voltages, so the wattage value can vary here at the same amps!
Amperage tends to scale up with more cores.


Any value up there can be divided by 12 for the phases in use, so at 100A from the CPU divided by 12 you're at 8.33A per phase - under 1.5W per phase which is easy to passively cool without needing a heatsink.

Going to buildzoid who must have zoomed in further to give more accurate values:
At 100A, he mentions you're looking at just 11W of heat out of the VRMs - nothing.
at 150A, from something like a 3900x or 5800x and above - about 15W. Spread over all those FET's, you barely even need a heatsink.

The biggest danger would be if those FET's were running at 100C which is within their specs, nearby capacitors are the devices in danger from the heat
1692776067167.png


At 200A the board finally hits it's inefficiencies, peaking at 19W.
This has room for the 5950x overclocked that peaked at 215A last page - and you'd finally need a heatsink at ~20W output.
After this isn't worth mentioning because nothing on AM4 goes that high.



He really liked the polymer capacitors on the SoC VRMs and got uspet the board has no graphics outputs as that alone made him want to try APU RAM overclocking. (The three devices in the bottom right)
1692776452099.png


Summary: Oh no doublers! But actually, a board better than AM4 can even use with decent overhead on the CPU and overkill on the DRAM.


Why am i suddenly posting a lot of motherboard stuff? You'll figure it out soon.
 
Last edited:

Mussels

Freshwater Moderator
Staff member
Joined
Oct 6, 2004
Messages
58,413 (8.18/day)
Location
Oystralia
System Name Rainbow Sparkles (Power efficient, <350W gaming load)
Processor Ryzen R7 5800x3D (Undervolted, 4.45GHz all core)
Motherboard Asus x570-F (BIOS Modded)
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 1.15V)
Video Card(s) Galax RTX 3090 SG 24GB: Underclocked to 1700Mhz 0.750v (375W down to 250W))
Storage 2TB WD SN850 NVME + 1TB Sasmsung 970 Pro NVME + 1TB Intel 6000P NVME USB 3.2
Display(s) Phillips 32 32M1N5800A (4k144), LG 32" (4K60) | Gigabyte G32QC (2k165) | Phillips 328m6fjrmb (2K144)
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 ( Sexy white 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!
Benchmark Scores Nyooom.
Commenting on the post i made above, with more learning on my side.

Doublers are meant to be used in alternating phases - so while a board may only be four phases doubled to 8, they take turns and have a longer cool down time to boost their efficiency.
A PWM controller that's aware of this means that the original controller can send a signal to time when they do that shift for maximum efficiency, and those boards are like the asrock taichi where doublers are used effectively and work extremely well.

The really bad boards like the asrock x470 master/SLI did not use doublers, meaning the actual electrical component to alternate those phases - they slapped them in, in parallel making it almost entirely worthless. As people repeated that information around, it got misinterpreted as "this board has doublers used wrong" to "doublers bad"



The irony is that asrock made some shitty boards, then made great ones to earn back trust... then went right back to shitty ones. It's like they have separate development teams that don't share information.
 
Joined
Oct 15, 2011
Messages
1,982 (0.43/day)
Location
Springfield, Vermont
System Name KHR-1
Processor Ryzen 9 5900X
Motherboard ASRock B550 PG Velocita (UEFI-BIOS P3.40)
Memory 32 GB G.Skill RipJawsV F4-3200C16D-32GVR
Video Card(s) Sapphire Nitro+ Radeon RX 6750 XT
Storage Western Digital Black SN850 1 TB NVMe SSD
Display(s) Alienware AW3423DWF OLED-ASRock PG27Q15R2A (backup)
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 23H2
Looks like I'm on to something with my 5800X:

Looks like my 8th core and the 2nd core are the best ones. (They are boosted the most consistently)
 

freeagent

Moderator
Staff member
Joined
Sep 16, 2018
Messages
7,546 (3.68/day)
Location
Winnipeg, Canada
Processor AMD R9 5900X
Motherboard Asus Crosshair VIII Dark Hero
Cooling Thermalright Aqua Elite 360 V3 1x TL-B12, 2x TL-C12 Pro, 2x TL K12
Memory 2x8 G.Skill Trident Z Royal 3200C14, 2x8GB G.Skill Trident Z Black and White 3200 C14
Video Card(s) Zotac 4070 Ti Trinity OC
Storage WD SN850 1TB, SN850X 2TB, Asus Hyper M.2, 2x SN770 1TB
Display(s) LG 50UP7100
Case Fractal Torrent Compact RGB
Audio Device(s) JBL 2.1 Deep Bass
Power Supply EVGA SuperNova 750w G+, Monster HDP1800
Mouse Logitech G502 Hero
Keyboard Logitech G213
VR HMD Oculus 3
Software Yes
Benchmark Scores Yes
Looks like I'm on to something with my 5800X:

Looks like my 8th core and the 2nd core are the best ones. (They are boosted the most consistently)
Give more power :)

Give more power :)
For me I find my older Zen 3 parts respond well to lots of power and curve.. less power and curve makes things unstable.
 
Joined
Oct 15, 2011
Messages
1,982 (0.43/day)
Location
Springfield, Vermont
System Name KHR-1
Processor Ryzen 9 5900X
Motherboard ASRock B550 PG Velocita (UEFI-BIOS P3.40)
Memory 32 GB G.Skill RipJawsV F4-3200C16D-32GVR
Video Card(s) Sapphire Nitro+ Radeon RX 6750 XT
Storage Western Digital Black SN850 1 TB NVMe SSD
Display(s) Alienware AW3423DWF OLED-ASRock PG27Q15R2A (backup)
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 23H2
Also, looks like ASRock B550 PG Velocita, does require an SoC voltage increase for >16 GB of RAM and/or higher Fclk!

Hardware monitors are showing 0.9V, usually no higher than 0.97V. I guess I now can start to manually bump up some voltages.
 

Mussels

Freshwater Moderator
Staff member
Joined
Oct 6, 2004
Messages
58,413 (8.18/day)
Location
Oystralia
System Name Rainbow Sparkles (Power efficient, <350W gaming load)
Processor Ryzen R7 5800x3D (Undervolted, 4.45GHz all core)
Motherboard Asus x570-F (BIOS Modded)
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 1.15V)
Video Card(s) Galax RTX 3090 SG 24GB: Underclocked to 1700Mhz 0.750v (375W down to 250W))
Storage 2TB WD SN850 NVME + 1TB Sasmsung 970 Pro NVME + 1TB Intel 6000P NVME USB 3.2
Display(s) Phillips 32 32M1N5800A (4k144), LG 32" (4K60) | Gigabyte G32QC (2k165) | Phillips 328m6fjrmb (2K144)
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 ( Sexy white 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!
Benchmark Scores Nyooom.
Give more power :)


For me I find my older Zen 3 parts respond well to lots of power and curve.. less power and curve makes things unstable.
Lower voltage makes things less stable, less power with PBO limits cant - unless you've got the curve voltage too low for an intermediate step
(Might be stable at all core 4.5GHz and single core 4.85GHz, but unstable with 3 cores at 4.7GHz, etc)

That's what usually trips people up as they can only really test full load on all cores or a single core, testing multiple cores at every possible load level is impossible and generally takes time to find out.
 

freeagent

Moderator
Staff member
Joined
Sep 16, 2018
Messages
7,546 (3.68/day)
Location
Winnipeg, Canada
Processor AMD R9 5900X
Motherboard Asus Crosshair VIII Dark Hero
Cooling Thermalright Aqua Elite 360 V3 1x TL-B12, 2x TL-C12 Pro, 2x TL K12
Memory 2x8 G.Skill Trident Z Royal 3200C14, 2x8GB G.Skill Trident Z Black and White 3200 C14
Video Card(s) Zotac 4070 Ti Trinity OC
Storage WD SN850 1TB, SN850X 2TB, Asus Hyper M.2, 2x SN770 1TB
Display(s) LG 50UP7100
Case Fractal Torrent Compact RGB
Audio Device(s) JBL 2.1 Deep Bass
Power Supply EVGA SuperNova 750w G+, Monster HDP1800
Mouse Logitech G502 Hero
Keyboard Logitech G213
VR HMD Oculus 3
Software Yes
Benchmark Scores Yes
SuperPi 32M, it’s a good quick way to test cores for max speed, just gotta do one at a time :)
 
Joined
Feb 3, 2019
Messages
4 (0.00/day)
Just sharing my PBO Limits+Curve Optimizer settings...

Ryzen 7 5800X​

PBO LimitsCurve Optimizer
PPT: 110 WCO: All Cores
TDC: 77 AAll CO Sign: Negative
EDC: 100 AAll CO Magnitude: 30

image.png
image2.png
 
Last edited:
Joined
Dec 13, 2022
Messages
19 (0.04/day)
System Name CYX PC
Processor Ryzen 7 5800X
Motherboard Asus Rog Strix B550 E-Gaming
Cooling Arctic Liquid Freezer II 360 - x11 total case fans - Arctic P12 120mm (x9) - Arctic F8 PWM 80mm (x2)
Memory GSkill TridentZ 3200mHz 16CL (32GB - 4x8GB)
Video Card(s) Gigabyte RTX 3070 Ti - Gaming OC (Rev 1.0)
Storage Samsung 970 Evo Plus M2 (500GB) - Samsung 970 Evo Plus M2 (2TB) - Toshiba P300 HDD (3TB)
Display(s) Acer Nitro XV342 CKP (IPS 144hz - 1ms - Flat Ultrawide - GSynch Compatibility)
Case LianLi Dynamic O11 Midi (Black)
Audio Device(s) Fenton Studio Speakers 120W/Yamaha Sub 80W - Logitech G Pro X Gaming - AudioTechnica ATH-M50x Studio
Power Supply Corsair RMx 650W (2018)
Mouse Razer Deathadder Elite
Keyboard Motospeed CK-104
Software Windows 11 Pro
5800Χ (PBO 2 UnderVolt PER CORE (not all cores)
After hundreds of tests I found the golden ratio in terms of maximum speed - stability and above all very good temperatures.
Definitely have more stability with UV Per Core vs All cores but you have to take your time -many test.
After a year and a half (1.52years) there were no problems even with Windows 10-11experimental ones and stable Windows.
The Results is better than Stock 5800X for thermal reasons the CPU boost smooth and faster.
Max Temp: 78C extreme bench 2hours and 68C in AAA.
Iddle: 32C

SO! My Settings
PPT: 120 TDC: 75 EDC: 130
Scalar: Auto (It's same with x1 but for better results use Auto)
F-Max Ench for Asus bios use: Auto (Auto = Disable by default)
C-State: Auto

@0: 28
@1: 28
@2: 28
@3: 28
@4: 28
@5: 5 (Problematic Core)
@6: 18 (Best Core)
@7: 18 (Second Best Core)

I use one of the B550 mobo with best VRM, PWM Controller, Mosfet etc..the ASUS ROG Strix B550 E-Gaming
AIO: Ofc ARCTIC Freezer II 360 with x6 120mm fans (push/pull direction) and for thermalpaste: The new MX6 (work with same best results with MX4 too).
The MX6 is better in heavy loading vs MX4 best iddle temps.

AIO Fans Curve: 20C - 40%
TSensor (Inside - Max 28C
 

Attachments

  • 20221221_045656.jpg
    20221221_045656.jpg
    470.9 KB · Views: 37
  • CB-R23 TEST 10.05.23.jpg
    CB-R23 TEST 10.05.23.jpg
    359.7 KB · Views: 51

izy

Joined
Jun 30, 2022
Messages
930 (1.39/day)
I always wondered why the best core needs more voltage, my CO (on 5700x, AutoOC +200) is -20 , -25, -25, -25, -25, -20, -25, 25 , the -25 were working fine at -27 too and i used to have the -20 cores to -22 stable in all stress tests but i had some random errors at some point (it could had been the RAM) so i tuned the CO a bit up to -20/-25 to be sure its stable, my best core is the one with -20 (first -20 is second best core) but i dont understand why the best core cant go as low as the other cores in CO because they all boost to the same frequency (they all doing even 4700Mhz all core in some stress tests and all can boost up to 4850Mhz), a simple logic tells me that the best core should be the worst core if it needs more voltage to run at the same speed as others, i guess im missing something here.

Edit: an explanation that i found on reddit , not sure if its right or wrong but makes sense:

"... The best cores have lower VIDs for each frequency step.
A good core might have, say, 1350mV for 4600MHz, while a bad core might have 1400mV. These are programmed in the chip at the factory, and these are the values that you're offsetting when setting a curve optimizer value."
 
Last edited:

Mussels

Freshwater Moderator
Staff member
Joined
Oct 6, 2004
Messages
58,413 (8.18/day)
Location
Oystralia
System Name Rainbow Sparkles (Power efficient, <350W gaming load)
Processor Ryzen R7 5800x3D (Undervolted, 4.45GHz all core)
Motherboard Asus x570-F (BIOS Modded)
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 1.15V)
Video Card(s) Galax RTX 3090 SG 24GB: Underclocked to 1700Mhz 0.750v (375W down to 250W))
Storage 2TB WD SN850 NVME + 1TB Sasmsung 970 Pro NVME + 1TB Intel 6000P NVME USB 3.2
Display(s) Phillips 32 32M1N5800A (4k144), LG 32" (4K60) | Gigabyte G32QC (2k165) | Phillips 328m6fjrmb (2K144)
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 ( Sexy white 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!
Benchmark Scores Nyooom.
I always wondered why the best core needs more voltage
Because higher clocks needs higher volts, it's as simple as that.

It could be as simple as the other cores leak too much heat, and that's the one that handles the higher voltage the best.
 

izy

Joined
Jun 30, 2022
Messages
930 (1.39/day)
Because higher clocks needs higher volts, it's as simple as that.

It could be as simple as the other cores leak too much heat, and that's the one that handles the higher voltage the best.
That much i could think of too , higher clocks = higher voltage but the best cores need more voltage for the same speed , i guess they are programmed with lower VIDs from the factory, i thought they all came with the same VID.
 
Top