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

Help OC Ryzen 5 3600

Joined
Sep 3, 2019
Messages
1,669 (3.65/day)
Location
Thessaloniki, Greece
System Name PC alive since Aug 18th 2019
Processor Ryzen 5 3600 (PBO Enabled)
Motherboard Gigabyte X570 Aorus Pro, BIOS F22
Cooling Corsair H110i 280mm (Liquid metal for TIM)
Memory CorsairVengeanceLPX DDR4 2x8GB 3466MHz CL16-18-18-36 1T, B-die A0 PCB, @3800 CL16-18-18-36 1T(1.46V)
Video Card(s) MSI RX 5700XT Gaming X
Storage Samsung NVMe: 970Pro 512GB (2019) / SATA-III: 850Pro 1TB SSD (2017), 860Evo 1TB SSD (2020)
Display(s) 24" EIZO FlexScan S2411W 1920x1200, 16:10 60Hz samsung S-PVA 14-bit (16.7M/1.06B colors), 6ms G2G
Case None
Audio Device(s) Sound Blaster Z <--optical link--> Logitech Z5500 5.1 500W
Power Supply Corsair HX750i
Mouse Logitech MX Master
Keyboard Logitech G15 v2
Software Windows 10 Home 64bit (v2004)
Ok, the one you found is on:
Settings > Advanced > AMD Overclocking > AMD Overclocking > Precision Boost Overdrive

Try to look at:
OC > Advanced CPU Configuration > Precision Boost Overdrive
 
Joined
Apr 26, 2019
Messages
179 (0.30/day)
Location
Italy
System Name The Worker
Processor Ryzen 5 3600
Motherboard MSI b450 Tomahawk Max
Cooling Arctic Freezer 33 eSport One
Memory 16Gb DDR4 3200 Mhz Corsair vengeance LPX
Video Card(s) Gigabyte gtx 1660 Ti OC @2030Mhz core and 7400Mhz memory
Storage 2x 250Gb Samsung 850 Evo, 3TB HDD
Display(s) Asus mx259q
Case Corsair Spec Alpha
Audio Device(s) Onboard
Power Supply CoolerMaster G750M
Mouse Razer Deathadder Chroma
Keyboard Gamdias Hermes e2
Software W10 64bit, Some Linux distros
ok found and set
MSI_SnapShot_02.jpg
 
Joined
Mar 23, 2016
Messages
4,297 (2.50/day)
Processor Ryzen 7 3800X
Motherboard MSI B450 Tomahawk ATX
Cooling Cooler Master Hyper 212 Black Edition
Memory PNY Anarchy-X XLR8 Red DDR4-3200 16GB kit & PNY Anarchy-X XLR8 Red DDR4-2666 16GB kit
Video Card(s) MSI GeForce RTX 2060 GAMING Z 6G
Storage Samsung 970 EVO NVMe M.2 500 GB, SanDisk Ultra II 480 GB
Display(s) Samsung SyncMaster C27H711 OC refresh rate 110Hz
Case Phantek Eclipse P400S (PH-EC416PS)
Audio Device(s) EVGA NU Audio
Power Supply EVGA 850 BQ
Mouse SteelSeries Rival 310
Keyboard Logitech G G413 Silver
Software Windows 10 Professional 64-bit v2004
Core Performance Boost is the XFR option. Once you disable that option the CPU only runs at base clock speed (3.6 GHz for your 3600.)
 
Joined
Sep 3, 2019
Messages
1,669 (3.65/day)
Location
Thessaloniki, Greece
System Name PC alive since Aug 18th 2019
Processor Ryzen 5 3600 (PBO Enabled)
Motherboard Gigabyte X570 Aorus Pro, BIOS F22
Cooling Corsair H110i 280mm (Liquid metal for TIM)
Memory CorsairVengeanceLPX DDR4 2x8GB 3466MHz CL16-18-18-36 1T, B-die A0 PCB, @3800 CL16-18-18-36 1T(1.46V)
Video Card(s) MSI RX 5700XT Gaming X
Storage Samsung NVMe: 970Pro 512GB (2019) / SATA-III: 850Pro 1TB SSD (2017), 860Evo 1TB SSD (2020)
Display(s) 24" EIZO FlexScan S2411W 1920x1200, 16:10 60Hz samsung S-PVA 14-bit (16.7M/1.06B colors), 6ms G2G
Case None
Audio Device(s) Sound Blaster Z <--optical link--> Logitech Z5500 5.1 500W
Power Supply Corsair HX750i
Mouse Logitech MX Master
Keyboard Logitech G15 v2
Software Windows 10 Home 64bit (v2004)
Core Performance Boost is the XFR option. Once you disable that option the CPU only runs at base clock speed (3.6 GHz for your 3600.)
Isnt XFR the former name of PBO?
XFR (eXtended Frequency Range) bacame PBO (Precision Boost Overdrive)

Core Performance Boost is another name for PB (Precision Boost)

PB and PBO are 2 different functions
 
Joined
Mar 23, 2016
Messages
4,297 (2.50/day)
Processor Ryzen 7 3800X
Motherboard MSI B450 Tomahawk ATX
Cooling Cooler Master Hyper 212 Black Edition
Memory PNY Anarchy-X XLR8 Red DDR4-3200 16GB kit & PNY Anarchy-X XLR8 Red DDR4-2666 16GB kit
Video Card(s) MSI GeForce RTX 2060 GAMING Z 6G
Storage Samsung 970 EVO NVMe M.2 500 GB, SanDisk Ultra II 480 GB
Display(s) Samsung SyncMaster C27H711 OC refresh rate 110Hz
Case Phantek Eclipse P400S (PH-EC416PS)
Audio Device(s) EVGA NU Audio
Power Supply EVGA 850 BQ
Mouse SteelSeries Rival 310
Keyboard Logitech G G413 Silver
Software Windows 10 Professional 64-bit v2004
Isnt XFR the former name of PBO?
XFR (eXtended Frequency Range) bacame PBO (Precision Boost Overdrive)

Core Performance Boost is another name for PB (Precision Boost)

PB and PBO are 2 different functions
No, not on a MSI motherboard, CPB is XFR renamed. You can disable PBO, and leave CPB on for the advertised XFR boost (4.2 GHz.)

MSI_SnapShot_02.jpg
 

specopsFI

New Member
Joined
Sep 10, 2019
Messages
12 (0.03/day)
Just my two cents in this. Overclocking is and always has been about going over the specs and limits that the manufacturer has set. Zen2 is no different, except maybe for the headroom being very, very slim. I wouldn't worry one bit about going over the stock PPT, TDC and EDC limits. I am willing to bet (though can not prove, just like anyone outside AMD can't) that those are not set primarily for the silicon longevity, but for product segmenting. If 3600 was somehow substantially more fragile than for example 3600X, that would make Zen2 a very special case in the history of silicon chips. More likely is that the limits for the 3600 are set lower primarily because of the smaller stock cooler.

FIT is another thing, and that I would keep an eye on. As the Stilt has stated, the best way to get to know the guideline for the limits of your own chip is by removing PPT, TDC and EDC limits, keeping all other things at stock and putting the processor under load. And that load should be the heaviest load that you are going to be putting on the CPU when actually using it. For my own 3600, that is 1.313V (the lowest I've seen a core dip to under gaming load). And since my CPU/MB combo doesn't want to boost too well and I can get 4.2GHz stable at 1.297V, my choice is all core. If I did actual heavy stuff, I'd have to go for ~1.25V to feel comfortable.

Even for FIT, there is bound to be a certain buffer that AMD has seen fit (pun intented) to leave between FIT voltage and quick degradation. How big of an gap it is, we don't know. Most likely very small compared to pretty much any previous CPU we've seen. But still: going over PPT, TDC and EDC limits is nothing to worry about. FIT is relevant and a good ballpark to what you should set for voltage. And overall, gains by all core OC are very small but not completely irrelevant. Most gains are from IF and memory tweaking.
 
Joined
Apr 26, 2020
Messages
1 (0.00/day)
System Name Red-Queen
Processor AMD Ryzen 5 3600
Motherboard MSI B450 Tomahawk Max
Cooling ID-Cooling Auraflow X 240
Memory G.Skill Trident Z Neo 3600 MHz
Video Card(s) Sapphire Nitro+ RX580 8GB
Storage Crucial MX500 1TB
Display(s) Acer KG241Q
Case Phanteks Eclipse P350x
Audio Device(s) Logitech Pro Headset
Power Supply Silverstone ET750-G 750W 80 Plus Gold
Mouse Steelseries Rival 110
Keyboard Logitech G213
Ok... I expected a little longer than 9sec run... (at least a minute) but this will do as long as future screenshots will be taken the same way.
This will be your starting point.

See that PPT/TDC/EDC values? (raw value and %)
The default CPU PBO limits are

-----default-----(yours)
PPT: 88W--------(83W)
TDC: 60A---------(46A)
EDC: 90A---------(77A)

PPT: Power Package Tracking (total CPU socket power draw)
TDC: Thermal Design Current (max Current draw when throttling >95C)
EDC: Electrical Design Current (max Current draw when normal)

Under the same workload mine was:

-----default-----(mine)
PPT: 88W--------(86W) +2
TDC: 60A---------(49A) +3
EDC: 90A---------(78A) +1

The difference between them is due to temp.
Yours 66~67C
Mine 62~63C

And all these are monitored and regulated by silicon manager. The lower the temp the more watt and ampere the CPU is allowed to draw.
But your temps are more than acceptable.
What you can do now and may try to improve things, is to cap EDC.
If you bring that 77A EDC down the silicon manager will see potential less silicon stress, and may try to increase power draw with more clock and voltage. Some how it tries to take the stress headroom created by EDC reduction.

So, go to BIOS under XFR Enhancement and set PBO to manual like this

PPT: 90
TDC: 0 (default)
EDC: 70
PBO Scalar: Auto

...and run the R20 again.
Hi Zach,

below are my TDC EDC and PPT readings. with PBO enabled. what manual settings would you recommend to squeeze alil more juice out from the processor via PBO. im running a Ryzen 5 3600 and MSI B450 Tomahawk Max motherboard. Thanks!

1588181837455.png
 
Joined
Sep 7, 2020
Messages
1 (0.01/day)
System Name Custom
Processor AMD Ryzen 3600 @ 4,4 GHz all core @ 1,24 V
Motherboard MSI MPG X570 Gaming Plus
Cooling Scythe Mugen 5 ARGB Plus
Memory Corsair Vengeance LPX schwarz DIMM Kit 16GB, DDR4-3200, CL16-18-18-36 (CMK16GX4M2B3200C16)
Video Card(s) GeForce® GTX 1080 GameRock Premium Edition
Storage Samsung SSD 860 EVO 500GB
Display(s) Sony KD 55XD8577
Case Corsair SPEC-02
Audio Device(s) Realtek® ALC1220 Codec
Power Supply Be Quiet Pure Power 600 watt
Mouse Logitech G500
Keyboard Logitech Classic Keyboard 200
Software Windows 10
Benchmark Scores Cinebench R20 : 4003 Cinebench R15 : 1748
Hi Guys,

My Ryzen 3600 runs now at 4400 MHz at 1,24 Volts with air cooling (Scythe Mugen 5), it needs only 15 watt in ideal. It scores 4000 Points in Cinebench R20. It stays cooler than with the standard settings and needs of course less Watts under load. I not tuned my Ram timings yet, I only activate the XMP profile (3200 MHz cl16). I tested it with Prime 95 1 hour. (max temp 75 °C, v drop 1,244 - 1,231, max 91 Watt SMU)

I'm not sure if my CPU EDC is holding me back, it maxed out at 90 A under heavy load, cause of the low CPU Voltage.

I'm pretty sure that this OC is better than run standard settings.

I tested 4,5 GHz at 1,3 Volts, it was stable in all benchmark test, (Cinebench R20, SiSoftSandra, UserBenchMark, 3DMark TimeSpy, FireSrike) but I don't tested it with Prime, it was getting a bit hot without an AIO (85 °C till Cinebench), and I don't want to toast it directly, so I would recommend using an AOI at those Voltages.

I tested 4,5 GHz at 1,3 Volts, it was stable in all benchmark test, (Cinebench R20, SiSoftSandra, UserBenchMark, 3DMark TimeSpy, FireSrike) but I don't tested it with Prime, it was getting a bit hot without an AIO (85 C° till Cinebench), and I don't want to toast it directly, so I would recommend using an AOI at those Voltages.

At my 4400 MHz OC, I started at 1,2 Volts, but it was completely Prime stable at 1,24 Volts.

I tested out the max CPU voltages at standart settings under Full load (Prime 95), with a x570 chipset and bios from August 2020, my Voltages jumpes around at 1,26 - 1,32 Volts, but most of the time it's near to the 1,26 Volts at 3990 MHz (max CPU Temp 80 °C).

So I think 1,26 Volts is the minimum all core Voltage under maximum CPUload from AMD, so completely Save.

1,32 Volts was the maximum Voltage under maximum CPU load, but it was not often there and only short.

So I think 1,26 Volts Is total save all core voltage, I would not go over 1,3 Volts, if you want to be sure that your CPU don't degenerates fast and keep your Temps under 80 °C.

Nobody knows exactly how fast these chips degenerate, but every CPU degenerates over time of using, also with stock voltages and settings, heat and voltage makes it happen faster and nobody knows with how much years we are ending up, at what heat and voltage, without bumping it a bit up to be stable again, because if u set the perfect voltage where your CPU is stable, you have no headroom for degeneration, like with stock voltages.
 

Attachments

CynicalCreator

New Member
Joined
Sep 12, 2020
Messages
1 (0.01/day)
I've been tweaking my 3600 over the last week or so too, i could run cinebench and games with it at 4.5/1.3v but it wasnt stable after the 1st hard cycle of prime95, not crashing, just workers stopping, and it was only scoring 4025 in cinebench while 4.4 was getting 4000.

Been trying to dial it in to full overnight prime stability. Thought i was good but had 1 stop an hour in at 4.325/1.3v, currently at 4.325/1.30625v a little farther. 4.3/1.3v made it through 2 hours of prime no stoppings, but it couldnt break 3900 in cinebench, close, but whole numbers though... I'm on the 4th hard cycle with no workers stopping now, hopefully this is the final tweaking.

Its definitely running games smoother less stutter than the normal boost that maxed at 4."2", and the actual use overall average power consumption looks the same if not better, only hitting the edc cap, 10-20 over ppt and tdp during prime hard cycles though, all within range with games, just a gaming rig so when its on its running full power anyway, not too worried about a little degradation.

But, something i noticed about the volts, like in the screenshot above, my unstable settings always had some cores dropping to 0, and everytime i had prime failures it was on the same cores, pretty sure thats not good

I had thought i was good at 4.4 without doing a real long full stress test, but then i had a game crash and started checking hwinfo meticulously, seen the 0 volts and went back to the drawing board, where i noticed the stopped worker/0 volt core correlation. I havent had any crashes since i got the settings to where the volts dont ever hit 0 though, lowest they droop is 1.244 now

...

Well, as i was writing, a worker failed, almost 2 hours into it, going to try 4.325/1.3125v, i need that totally useless 3900 cinebench score damn it! Does it really matter if only 1 worker is failing? I mean, aside from the stress tests its fine aint it?
 
Joined
Feb 20, 2019
Messages
1,752 (2.69/day)
System Name PowerEdge R730 DRS Cluster
Processor 4x Xeon E5-2698 v3
Cooling Many heckin screamy bois
Memory 480GB ECC DDR4-2133
Video Card(s) Matrox G200eR2
Storage SD Card. Yep, really no other local storage.
Display(s) It's probably a couple of boring Dell Ultrasharps and a sacrificial laptop.
Case 39U 6-rack server room with HEVC and 44KVA UPS
Mouse Maybe
Keyboard Yes!
Software ESXi 6.5 U3
Benchmark Scores I once clocked a Celeron-300A to 564MHz on an Abit BE6 and it scored over 9000.
As much as overclocking is fun, unless you are doing it solely for the joy of the overclocking experience I'd highly recommend just dialling it back plenty.

If you're almost stable at 4.5, then stick with 4.3 and call it a day. It's easy to waste so much time trying to dial in the perfect overclock but your chip will slowly degrade as it ages anyway, so trying to find the exact knife-edge of an OC's stability isn't a long-term plan anyway. If any application crashes you'll always have that nagging "I wonder if it's my overclock" at the back of your mind and you'll go back through another 20 minutes of testing, voltage tweaking, and then another overnight stress test to make sure.

My new 3600XT (don't worry, I didn't make such a foolish purchase with my own money) is actually 1h stable at 4.7@1.375V running an 1866 FCLK but does require more cooling noise than I'm prepared to tolerate and that voltage isn't a good idea long term, nor do I think that's stable enough for a daily driver.

Since this chip will probably live in this particular PC for a good couple of years at least, I've dialled it back to 4.4@1.275V and running a very modest 1733 FCLK with DDR4 tuned to CL14. I'm 100% confident the OC is stable now, and will be still be stable in 3 years from now, all whilst it's running extremely quietly compared to stock PBO.

Sure, my Cinebench R20 score is only 3980 now instead of 4170 but that's barely a 4% performance drop for peace of mind and peace and quiet. I'd likely need to average a few benchmarks to even show that difference over the run-to-run variance, and realistically I'm not going to be able to tell the difference between two experiences within 4% of each other.
 
Joined
Jul 8, 2019
Messages
115 (0.22/day)
My Ryzen 3600 runs now at 4400 MHz at 1,24 Volts with air cooling (Scythe Mugen 5)
How the hell. I have the same cooler, same cpu and my chip can't hit 4.2 stable with 1.35V during prime95 small FFT 128k size. 4.1 is fine. AVX prime95 load completely obliterates it and make an instant crash at 4.2. My CPU is early bin from late july 2019, maybe that's why its so crap?
 

lordcarloske

New Member
Joined
Sep 30, 2020
Messages
1 (0.02/day)
I've been pushing my 3600x over 1.3v since I got it, no degradation whatsoever (although I'm not benching it day in day out). Gaming gets it up over 60 degrees but not by much, benching with p95 pushes it up over 80 degrees but that's to be expected, it's a torture test.

As long as you're using good cooling you can really push these chips hard, max tdp is 95 degrees so as long as it's not shutting down during use you're fine.

The default voltage in bios is like 1.44v but that's using the fluctuating frequency not a set one.
i can doit easy 43.50 to 1.325 voltios
 

Attachments

craxton

New Member
Joined
Dec 30, 2019
Messages
4 (0.01/day)
Don't know where to post it so do forgive. My 3600xt is set to 4.5 manual oc @ 1.28volts. I have 3466 mhz ram, infinity fabric is set to 1733. In fsb ratio on Aida and cpuz both say it's like 3:58 or something like that not at home atm. Omw to work (7p.m./7a.m.) I'm wondering how am I to get a 1to1to1 ratio for these? Running an msi mpg x570 gaming wifi edge board. PBO is off, there's no cool and quite inside the bios on 3000 series chips as my 1600af did have this option. Which it clocked at 4.1ghz 1.27 volts...another day. My cin r20 score is 4099 app core, single is 527.. anyhow yea considering my base clock isn't 3600mhz and instead is 4500mhz that would mean I would need an infinity clock of 2250 to get a 1/1/1 ratio or? New new new to zen 2 just got my 3600XT yesterday for 220 bucks...
 
Joined
Sep 3, 2019
Messages
1,669 (3.65/day)
Location
Thessaloniki, Greece
System Name PC alive since Aug 18th 2019
Processor Ryzen 5 3600 (PBO Enabled)
Motherboard Gigabyte X570 Aorus Pro, BIOS F22
Cooling Corsair H110i 280mm (Liquid metal for TIM)
Memory CorsairVengeanceLPX DDR4 2x8GB 3466MHz CL16-18-18-36 1T, B-die A0 PCB, @3800 CL16-18-18-36 1T(1.46V)
Video Card(s) MSI RX 5700XT Gaming X
Storage Samsung NVMe: 970Pro 512GB (2019) / SATA-III: 850Pro 1TB SSD (2017), 860Evo 1TB SSD (2020)
Display(s) 24" EIZO FlexScan S2411W 1920x1200, 16:10 60Hz samsung S-PVA 14-bit (16.7M/1.06B colors), 6ms G2G
Case None
Audio Device(s) Sound Blaster Z <--optical link--> Logitech Z5500 5.1 500W
Power Supply Corsair HX750i
Mouse Logitech MX Master
Keyboard Logitech G15 v2
Software Windows 10 Home 64bit (v2004)
Infinity Fabric is unrelated to CPU speed. It’s related to DRAM speed. Since you run DRAM at 3466, IF should be 1733 along with memory controller, also 1733.
Check ZenTimings software to see all 3 speeds aligned at 1733MHz.
MCLK:FCLK:UCLK
 
Joined
Feb 20, 2019
Messages
1,752 (2.69/day)
System Name PowerEdge R730 DRS Cluster
Processor 4x Xeon E5-2698 v3
Cooling Many heckin screamy bois
Memory 480GB ECC DDR4-2133
Video Card(s) Matrox G200eR2
Storage SD Card. Yep, really no other local storage.
Display(s) It's probably a couple of boring Dell Ultrasharps and a sacrificial laptop.
Case 39U 6-rack server room with HEVC and 44KVA UPS
Mouse Maybe
Keyboard Yes!
Software ESXi 6.5 U3
Benchmark Scores I once clocked a Celeron-300A to 564MHz on an Abit BE6 and it scored over 9000.
3466 at 1:1:1 will show up as 3:52. I don't know why they use daft fractions, but they all do. AIDA, CPU-Z etc.

FSB runs at 100MHz, and your base clock of 4500MHz just means a 45x multiplier of the 100MHz FSB. Like @Zach_01 says, core clock is unrelated to the memory speed and infinity fabric, they are linked only by the common 100MHz FSB

So your CPU multiplier is 45x
Your fabric, mem controller, and RAM's muliplier is 52/3, and dividing by three gives you 17.33 - so when you multiply that by the 100MHz FSB you get your 1733MHz speed that matches the DDR4-3466 (which runs at 1733 and is only called 3466 because it transfers data twice per clock).

I'm not even sure CPU-Z shows FCLK though, unless it's the NB - but that could just be the memory controller clock as the Northbridge is effectively the IO die on the CPU these days that handles the memory controller duties among other things....

1605098600892.png
 
Last edited:
Top