Ryzen Overclock thread. JUST OC results, and hardware/settings

[EXCellR8]

I got 3.9ghz on 1.35v with water... no problems so far. will try for 4.0 again once I flash a better BIOS. RAM still limited to 2667 for the time being

 

[Chiropteran]

Asrock x370 killer sli a/c, g.skill tridentz 3200mhz cl14 8GBX2, ryzen 1800x

Max stable overclock is 4050 on all cores. This is with 1.375vcore and LLC 1 (the highest setting). default soc voltage w/ LLC 1 as well, and XPM settings for RAM except clock reduced to 2933mhz. Temps top out at 93C (which is actually 73C given the 1800x 20 degree offset)

However, going down to 4025 I can set LLC down to 2 and temperature is a good 5C lower at full stress test, so this is my daily usage setting for now.

I have a feeling I could get higher with a better thermal solution, even at 4100 it boots and runs windows fine, only fails under the stress test after temperatures get above 90C. When my TAICHI arrives I'll be rebuilding with proper arctic silver and see if that does better than the default heat paste.

 

[Elfear]

Pushed it a little higher.

1700@3.8Ghz 1.25V in bios (1.19-1.21V under load according to CPU-Z and HWinfo)

Trident 3200CL14 @ 3200Mhz 14-14-14-34 1.35V in bios (1.373V according to HWinfo)

SOC Voltage = 1.1V

CPU Temps = 42C under load

The above settings passed 10hrs of Prime95 Small FFT.

So far my experience with Ryzen has been very positive. A lot of that is probably due to the fact I started tweaking weeks after the boards launched which gave manufacturers a chance to work out some of the bugs. I’ll keep pushing just to see what the chip will do but I’ll probably settle on 3.8Ghz for 24/7 use.

 

[cryosx]

changed out my cooler and can do 3.9 Ghz at 1.35 volts. RAM at 2933.

 

[Crono]

Ryzen 1600 @ ~3.9GHz (3892.11MHz), 1.392v, MSI X370 SLI Plus, with 16GB DDR4-3200 Corsair Vengeance LPX RAM currently running at 2933MHz (16-18-18-36).

 

[Markfw]

OK, I think this is my final settings, and a little disappointed.

1800x, 3.95 ghz, 1.4 vcore, 89.5c (69.5c real ?) on H100i Corsair AIO water (X370 Taichi mobo)
1700x 3.75 ghz 1.375 vcore 89c (69c real?) on Noctura DH-15 (AB350 Pro4 mobo)

BUT these are 24/7 100% loaded CPU's, and the 1700x also has a 980TI running full load.

Edit: one good thing. Both are running the Gskill TridentZ 3200 samsung b-die PC 3200 CAS14 at rated speed,

 

[Rifter]

OK, I think this is my final settings, and a little disappointed.

1800x, 3.95 ghz, 1.4 vcore, 89.5c (69.5c real ?) on H100i Corsair AIO water (X370 Taichi mobo)
1700x 3.75 ghz 1.375 vcore 89c (69c real?) on Noctura DH-15 (AB350 Pro4 mobo)

BUT these are 24/7 100% loaded CPU's, and the 1700x also has a 980TI running full load.

Those temps dont look horrible but the speeds are a little low, looks like you didnt win the silicon lottery this time around.

 

[Markfw]

Those temps dont look horrible but the speeds are a little low, looks like you didnt win the silicon lottery this time around.

I tried for 4 ghz on the 1800x, and I just kept upping the vcore, and it still was not stable, and the temps got really high, so I settled at 3950. The 1700x was sort of the same deal, they really want a LOT of vcore increase past a certain speed. I might even go down 50 hz, and get a lot of vcore decrease(and temps). For now, just letting them both run. 24 hours stable@100% load.

Its like the CPU hits a wall, and you need a lot of vcore to overcome it. If you stay a little under the "wall" speed, you can go really low on vcore, and still get some OC.

 

[richierich1212]

Well you are really putting those systems to the limit. Maybe the BIOS updates will help more with stability and you may soon be able to push a bit higher.

 

[Malogeek]

Got my AM4 kit today for my Noctua NH-U9S. My main focus was achieving lower temps at 3.8Ghz in Handbrake loads and I got about 8c lower. Due to the stupid BF1 audio bug with Ryzen software overclocking, I have to BIOS overclock which means I can't change profiles so I need a good middle ground for both gaming and full loads.

Ryzen 1700 3.8Ghz@1.325v 0.95v SOC

Handbrake caps out at 72c with the CPU fan at 85%, which is right where I want it.

 

[badcoder]

Pushed it a little higher.

1700@3.8Ghz 1.25V in bios (1.19-1.21V under load according to CPU-Z and HWinfo)

Trident 3200CL14 @ 3200Mhz 14-14-14-34 1.35V in bios (1.373V according to HWinfo)

SOC Voltage = 1.1V

CPU Temps = 42C under load

The above settings passed 10hrs of Prime95 Small FFT.

So far my experience with Ryzen has been very positive. A lot of that is probably due to the fact I started tweaking weeks after the boards launched which gave manufacturers a chance to work out some of the bugs. I’ll keep pushing just to see what the chip will do but I’ll probably settle on 3.8Ghz for 24/7 use.

This is awesome..... But what is your full specs...

My 2017 Rig:
Idle load
Ryzen7 1700
Wraith Spire (set upto 2250rpm - 89% - 71degreecelsius)
Msi B350 Tomahawk
Gskill TridentZ 3200mhz 16gb(8*2)
Msi RX 480 Armor 8GB OC
WD Black 1tb 7200rpm
Samsung SSD EVO 850 240gb
DEEPCOOL MID TOWER CABINET (ATX) - DUKASE V2
Case Fan - 1 (running at 100% = 1200rpm)
Antec f19 fan - 2 (each running at 100% = 1200 rpm)
Arctic MX-4 Thermal Compound
LG 22MP68 22" Full HD IPS SLIM LED MONITOR (1920x1080)

 

[Markfw]

BTW, I was playing around, and went back to stock to see the effect on speed and power. Both of the below are at 100% CPU usage doing BOINC

Overclocked=250 watts (full system, 5 fans and AIO pump and 550ti video card) 1.4 vcore LLC 3 3.950 ghz
Stock=183 watts 1.2 vcore. 3.7 ghz

So 67 extra watts to get just 250 extra hz, thats extreme !

 

[coercitiv]

So 67 extra watts to get just 250 extra hz, thats extreme !

The 16% voltage jump should have warned you about the outcome, that's a 35% increase in power just from higher vcore.

 

[IEC]

Adding 0.2V to vcore does that to you.

6.75% increase in draw due to frequency assuming linear scaling (more than this in reality)
36% increase due to voltage (1.4V/1.2V)^2 - this is an approximation as CPU is not a resistive load, but it's close enough

I'm quite happy with my 3.8GHz @ ~1.23v for 24/7 usage. I'll probably reattempt 4.1GHz for benches at some point, though.

 

[Malogeek]

1700@3.8Ghz 1.25V in bios (1.19-1.21V under load according to CPU-Z and HWinfo)

How did you get such a low voltage at 3.8? Is it just temp based? Or is your sample golden and mine is brick?

Anything below 1.325v at 3.8Ghz on my 1700 and in Handbrake it will crash and reboot.

Also under load CPU-Z and AI Suite are showing between 1.34 and 1.375v

 

[coercitiv]

How did you get such a low voltage at 3.8? Is it just temp based?

Remember when people where asking what the point is to buy a 1800X over 1700? There you go, on average of course.

 

[Malogeek]

Remember when people where asking what the point is to buy a 1800X over 1700? There you go, on average of course.

But we both have 1700's

 

[coercitiv]

But we both have 1700's

Yes, and you're disappointed by your overclock/voltage: that's exactly what 1800X was offering, a better chance for better voltage/frequency curve. (no guarantee though)

For what it's worth I think you made the correct choice though. Enjoy the beast!

 

[Malogeek]

Yes, and you're disappointed by your overclock/voltage: that's exactly what 1800X was offering, a better chance for better voltage/frequency curve. (no guarantee though)

What? This is nothing to do with his opinion of his overclock. I'm comparing required voltages to get the same frequency between the two systems, that's all. You make no sense.

Anyway I'll take it to the builders thread to stopping crapping all over this one.

 

[unseenmorbidity]

Remember when people where asking what the point is to buy a 1800X over 1700? There you go, on average of course.

I actually ran the numbers from Silicon Lottery's initial stats, and the difference in the avg clocks between the 1700 and 1700x was basically negligible. The 1800x had a delta of about 100 mhz on avg, but there were several caveats to it.

Supposedly, he had only tested 30 cpus from each bin, which isn't enough of a sample size, but it's the only data I know f that exist.


The 1700x was tested at lower voltages for whatever reason, and the 1800x was the only cpu with a 4.1ghz category. Perhaps none of his 1700x and 1700 cpus could hit 4.1?

The way they test is by using 3 voltages. A cpu might fail, but only need a slight bump in voltage to actually reach the given speed. This bias lowers the average for lower binned cpus, as Ryzen has a huge voltage wall around 3.9GHz.

 

[Faffo]

Hi, first post on this forum.

I'm currently running a 1700x @3.9Ghz, 1.35v LLC3 AIDA64 stable for over 30mins.
Ram is corsair vengeance lpx with white leds 3200MHz @2993 16-18-18-36
MoBo is an ASUS PRIME x370 pro.
Cooler: corsair h100i.
Temps reach aroud 67-68C (cpu reported) and 64-65 in AIDA in an h440 (veeery chocked case) and with around 24C ambient temp.

Only thing bugging me is the vcore reports.
In HWinfo64 the SVI2 TFN valure (it should be the one reported by the cpu) is consistent and sensible: 1.35 on idle and around 1.331 on load.
The Vcore from the mobo is instead all over the place. 1.35@idle and overing around 1.4v under load.
It makes no sense to me for the vcore to acutally increase so much under load insetad of showing vdroop (the LLC3 is halfway in the LLC settings)

 

[Soulkeeper]

My 1800x refuses to do 4GHz (turbo off) with 1.35v or lower
These things seem voltage hungry.
I'm still trying to dial in the lowest vcore required for 4GHz

 

[Markfw]

My 1800x refuses to do 4GHz (turbo off) with 1.35v or lower
These things seem voltage hungry.
I'm still trying to dial in the lowest vcore required for 4GHz

I went 1.4, and it was still only stable for a few hours, Enable LLC, it helps. It went to 1.42 a lot on the auto-upvolt. 3900 is pretty easy though.

 

[Soulkeeper]

It just seems odd that a chip with a 4GHz turbo can't hold 4GHz locked with good aftermarket cooling and a slight voltage bump.

 

[imported_jjj]

It just seems odd that a chip with a 4GHz turbo can't hold 4GHz locked with good aftermarket cooling and a slight voltage bump.

That's turbo with 2 cores only and the default V go quite high in such a scenario.

"On the high-end models the actual (effective) voltage for the base frequency (e.g. 3.6GHz on 1800X SKU) can be anything between 1.200 - 1.300V. Meanwhile the actual (effective) voltage for the highest single core boosted PState (XFR, e.g. 4.1GHz) can be as high as 1.47500V."
https://forums.anandtech.com/threads/ryzen-strictly-technical.2500572/