Ryzen Owners Zen Garden

[Arrakis9]

well that really sucks, so what your telling me is i should have gotten the X model

 

[biffzinker]

Doesn't matter which Ryzen model in the product stack. They all shut off CnQ if you overclock.

Abnormal behaviour that CnQ was turned on while overclocked.

 

[Arrakis9]

so its relatively safe running these chips at a set voltage? the chip does not run hot at all just a little bit worrying that the voltage wont come down with the clocks.

 

[Mussels]

i think thats a board limitation, as i have the same thing

you need to use P-state OCing with a different voltage per state to get it working the way you seek

 

[eidairaman1]

Inb4trolls

 

[Arrakis9]

i think thats a board limitation, as i have the same thing

you need to use P-state OCing with a different voltage per state to get it working the way you seek

there was something in the bios under ryzen tweaks that sounded like that. i'll have a look and see what it does, thanks.

 

[msroadkill612]

i think thats a board limitation, as i have the same thing

you need to use P-state OCing with a different voltage per state to get it working the way you seek

Well done Mussels. u give good advice.

 

[TheGuruStud]

Anyone here have Strix X370-F mobo? I'm really tired of Asus's garbage, but I didn't buy it. LLC absolutely does not work at all. Auto, level 1 or level 5 makes no difference. Vdroop from 1.393 goes all the way down to 1.33 or less (but I assume it's too low realistically) and locks the machine up. Clearly, the VDD result is a bogus as it shows 1.38x. All of the sub VIDs show the crazy drop. Unless it's dropping below 1.38 and not showing, b/c the V bounces all over.

 

[TheGuruStud]

BTW, I found a bug. Hwinfo cannot be ran during stress testing with this setup. It will randomly reboot even at stock speeds and V doesn't matter. I blame Asus, it's junk. RAM CL setting doesn't work on odd numbers, too.

Had to crank V with LLC set to 5 to pass OCCT, b/c of the droop.

 

[Johan45]

Make sure you're using the latest BIOS and version of HWInfo. ASUS has made some changes to their sensors, not sure it applies to your board but most likely

 

[YautjaLord]

@eidairaman1: So cores, much Ripper. Will still cost me 3 to 4 monthly salaries if not more. Will watch it more thoroughly regardless.

'ssup all. Got me F22 BIOS/EFI update for my mobo, so far so good, 3.9GHz/1.4v, 1711cb in R15, 12,000+ Kb/s WinRAR, 4000+ multi/450 single-core in CPU-Z bench, haven't tested Prime95 for 2h, but i guesstimate it'll run in 75-80 degrees C range. Now my question is: has anyone tested 1700X on same mobo as me with the same F22 applied to it @ 4.0GHz?

BTW: 35 degrees C @ this 3.9GHz OC @ idle. My room is below 25 C when i turn on ACU, 29/30 C when off.

*UPDATE*

Prime95:

One hour for now, F22 BIOS/EFI update applied, 1h 7min but who gives a f*ck if it's not 3 or 4, let alone 24h run. Cheers. Will probably do 3 or 4h run on this frequency next week, i got me EKWB's Thermal Grizzly Hydronaut TIM, will see how it compares to what i currently use (Arctic MX-4).

P.S. It's 1.4v in BIOS, CPU-Z applies some weird voltages to what is set in BIOS, 39.00 multi/1.40000v vCore.

P.P.S. New options added with this EFI update, like P-State & individual voltage tweaking sections to existing, will post a screenie here, have any one of you tweaked these? Pic:

It's like being 1st grader in OC'ing yet again, the 1st one at the top (highlighted) is where i found all the P-State related stuff, it's f*ckin' bananas! The fact it's also in the Peripherals makes it even more nuts than it already is.

 

[Mussels]

Finally got some real RAM instead of second hand generic stuff
lowered my CPU clocks a little for drastically lowered voltage (and waaaaaay colder temps), and satisfied my OCD by having matching clocks on CPU and ram (3600)

Wiring has been tidied since these were taken, ram was too tall for top mounted rad so i had to re-organise everything to fit it in the front intake instead

 

[TheGuruStud]

@eidairaman1: So cores, much Ripper. Will still cost me 3 to 4 monthly salaries if not more. Will watch it more thoroughly regardless.

'ssup all. Got me F22 BIOS/EFI update for my mobo, so far so good, 3.9GHz/1.4v, 1711cb in R15, 12,000+ Kb/s WinRAR, 4000+ multi/450 single-core in CPU-Z bench, haven't tested Prime95 for 2h, but i guesstimate it'll run in 75-80 degrees C range. Now my question is: has anyone tested 1700X on same mobo as me with the same F22 applied to it @ 4.0GHz?

BTW: 35 degrees C @ this 3.9GHz OC @ idle. My room is below 25 C when i turn on ACU, 29/30 C when off.

*UPDATE*

Prime95:


One hour for now, F22 BIOS/EFI update applied, 1h 7min but who gives a f*ck if it's not 3 or 4, let alone 24h run. Cheers. Will probably do 3 or 4h run on this frequency next week, i got me EKWB's Thermal Grizzly Hydronaut TIM, will see how it compares to what i currently use (Arctic MX-4).

P.S. It's 1.4v in BIOS, CPU-Z applies some weird voltages to what is set in BIOS, 39.00 multi/1.40000v vCore.

P.P.S. New options added with this EFI update, like P-State & individual voltage tweaking sections to existing, will post a screenie here, have any one of you tweaked these? Pic:

It's like being 1st grader in OC'ing yet again, the 1st one at the top (highlighted) is where i found all the P-State related stuff, it's f*ckin' bananas! The fact it's also in the Peripherals makes it even more nuts than it already is.

FWIW, IBT and OCCT are way more stressful in my experience on ryzen if you're going for ultra stable. 4hr+ with OCCT or several runs of IBT with 10GB ram (to murder it) should be good to go.
It could chew through Prime, but would fail after 1 hr+ on the others. 1.4V should be more than enough with high LLC. 1.3875 should work for 3.9 (FAR easier than 4.0, they sure did bin em).

 

[YautjaLord]

FWIW, IBT and OCCT are way more stressful in my experience on ryzen if you're going for ultra stable. 4hr+ with OCCT or several runs of IBT with 10GB ram (to murder it) should be good to go.
It could chew through Prime, but would fail after 1 hr+ on the others. 1.4V should be more than enough with high LLC. 1.3875 should work for 3.9 (FAR easier than 4.0, they sure did bin em).

Prime95, OCCT & IBT are heaters, Aida64, ASUS's RealBench, etc... - realistic (if you can call it that) stress tests, rest is of no significance. DL'd OCCT, the real kicker is both Prime95 & OCCT (correct me if i'm wrong) use AVX & OCCT also use AVX 2 instruction sets, both rape CPU real good; Prime95's Small FFTs test is the one to test if the cooling is actually good, 3h run of it will see if the OC is stable. I'll give it (OCCT) a try tomorrow, bout a time to see if Thermal Grizzly TIM is better than my Arctic MX-4. Will give it 1h run for now.

 

[TheGuruStud]

Prime95, OCCT & IBT are heaters, Aida64, ASUS's RealBench, etc... - realistic (if you can call it that) stress tests, rest is of no significance. DL'd OCCT, the real kicker is both Prime95 & OCCT (correct me if i'm wrong) use AVX & OCCT also use AVX 2 instruction sets, both rape CPU real good; Prime95's Small FFTs test is the one to test if the cooling is actually good, 3h run of it will see if the OC is stable. I'll give it (OCCT) a try tomorrow, bout a time to see if Thermal Grizzly TIM is better than my Arctic MX-4. Will give it 1h run for now.

Indeed, they are heaters. IBT was hotter than even prime small FFT. But, I value absolute stability over power consumption (had to up V a tick for IBT/OCCT). Of course, I know it's overkill, b/c package power was reported at 180 watts at times under OCCT and IBT, which is cuckoo for cocoa puffs at 3.95 ghz lol

At least it never downclocked itself in fear of Chernobyl 2

 

[YautjaLord]

Medium Data Set or Large Data Set? Also CPU OCCT or CPU Linpack? Number of logical threads - 16. Also applied Thermal Grizzly Hydronaut few mins ago.

 

[TheGuruStud]

Medium Data Set or Large Data Set? Also CPU OCCT or CPU Linpack? Number of logical threads - 16. Also applied Thermal Grizzly Hydronaut few mins ago.

I always use large and run linpack afterwards, as it's even harder on the cores, I believe (IBT uses linpack). I run Aida for funsies, too, I'm not sure what it does. IMO, Large data sets should always be tested, b/c if you're going to find an instability, it's going to be in the mem controller/cache. Some people might find linpack easier to pass, but that's why I said I use 10GB+ in IBT, b/c it's quite the workout routine LOL.

If you can pass those for over 4 hrs, then I wager it's very stable. I've been using my personal machines for many years each with this testing methodology (and of course all the ones I build and OC for people) and never had a crash. There's been a few unrelated BSODs that were fixed with updates due to early windows versions being so crap.

Historically, I found P95 to be the most stressful, but Zen is a different ball game than the old architectures and it hasn't worked out that way for me.

 

[YautjaLord]

Failed. Cause? After fiddling with Command Prompt & rest - you won't believe it, HOLD ON TO YOUR SOCKS!!!!!! - gdrv.sys. What is it you ask? Gigabyte Tools. Dafuk is this shit? All kinds of BSODs & stuff, System Recovery fixed it, but now i'm stuck with no OCCT cause last System Restore checkpoint was @ 28.6.2018. Gigabyte Tools? AppCenter? Are you f*ckin' joking me? Even funnier, AppCenter said there's the new version of it, same that caused all the trouble. Poor ol' GB, they are just bunch of nutcakes aren't they? Anyway, i'll resort to Small FFTs (heat), In-place large FFTs (FPU & RAM) & Blend (a f*ckload of RAM) tests to check for stability, none of the apps i have are raping the CPU &/or GPU 100% like this heater is. Video editing in OBS/Blender? 100% CPU utilization? Prime95 rapes both & than some.

P.S. Chose Large Data Set & CPU: OCCT, not CPU: Linpack, gave me error 2+ mins into test & popped all bunch of error logs. Error turned out to be GB app, after i ran verifier.exe in Command Prompt. (verifier.exe /standard /all followed by reboot) Will use OCCT probably, but not now. I still want RGB Fusion. Cheers.

 

[TheGuruStud]

Failed. Cause? After fiddling with Command Prompt & rest - you won't believe it, HOLD ON TO YOUR SOCKS!!!!!! - gdrv.sys. What is it you ask? Gigabyte Tools. Dafuk is this shit? BSODs.

LOL. It took me a while to figure out that HWinfo was rebooting the PC when testing. Idk even know how that's possible. I would expect a freeze, but the freaking reboots had me worried it was hardware.

 

[YautjaLord]

No hardware related now, 33 degrees C idle. Will run Small FFTs tomorrow, for 2 or 3h. Will probably switch from waterblock to EK's Aorus X370 monoblock i purchased, for my current CPU (1700X) & for 2700X later this year, 2nd GTX 1070 G1 rev. 2.0 & LG's 23.8" 4k IPS monitor joining the fray.

*UPDATE*

Got rid off of BIOS Setup (GB's bloatware #1) & USB DAC UP 2 (even bigger GB's bloatware #2), see if it'll make my CPU, OC & Prime95's Small FFTs happier/snappier. Prime95 Small FFTs 2h run coming tomorrow. CPU tDie temp is 30.8 degrees C idle in HWInfo64.

P.S. Prior to getting rid off of these two joke "utilities" had a pop up message to update AppCenter every time i booted into OS, not anymore. Yay. jk RGB Fusion is an exception: quite neat to choose individual LED RGB regions of mobo & graphics card, but i digress.

 

[biffzinker]

Anyone attempt disabling cores for a higher core clock?

Disabling a pair of cores although only one becomes unstable at 4.2 GHz in a heavy FPU load. I've managed 4.250 GHz was working on stress testing earlier.

 

[Mussels]

Anyone attempt disabling cores for a higher core clock?

Disabling a pair of cores although only one becomes unstable at 4.2 GHz in a heavy FPU load. I've managed 4.250 GHz was working on stress testing earlier.

i did, but found my boards VRM's were more the limit than the CPU - was having serious Vdroop past 1.3v, and higher voltages made the VRM's run at absurd temps

 

[Apocalypsee]

My new rig. Notice the cooler? Yep that's the old trusty Arctic Freezer 64 Pro that used to cool Athlon X2 3800+ overclocked 3.2GHz. The clip fits perfectly to AM4 stock bracket and provide good clamping force. It can even handle the CPU overclocked 3.8GHz with only around 64C in heavy games like Witcher 3. Not bad for a decade old cooler that only meant to cool dual core CPU's.

The RAM is Crucial cheap DDR4 2400 RAM rated at CL17-17-17-39. Its a Micron H-die chip single rank. At first I have no hope of clocking these. Just upping the RAM speed in BIOS just increase the latency to over 20 at 3000MHz. After scouring around reddit I found a compatible timing and managed 3200MHz at not-bad 16-17-16-36 1T at 1.3V. I'm happy with it

I kinda regret buying mATX even though I wanted a smaller footprint case. It have 5 case fans to cool down the 250W 980Ti and 65W CPU. I undervolt both, the 980Ti runs now at 1152MHz 1.024V (by my calculation its TDP should be around 200W) while the CPU is undervolt by 0.1V at stock clock. Runs cool and stable.

All in all I'm happy with the purchase. Games that hogged on old 4-core i5 4670K runs great on this without any overclock.

 

[GoldenX]

Update my bios for the B350 board, and got PBO, couldn't be happier.

 

[biffzinker]

Update my bios for the B350 board, and got PBO, couldn't be happier.

I was just looking for an update from MSI a couple days ago. Thanks for the post, update pending.

Updated