***The Official 5900X \5950X owners thread***

[Audioboxer]

My PBO was on auto, I think that is just within standard parameters, I set mine to motherboard which upped the watts and A for the various sections and all of a sudden my CPU suck down some proper juice.

CB23 went from 1616/25300 to 1680/30000

and now under sustained stressed load it holds ~4.6 all core and hits 90C at times with my weedy W/C setup but that's fine it'd rarely see that load

Can't wait to get this in my proper system and see what it'll actually do with good RAM and cooling

Yeah that's what I did, except instead of using motherboard for PBO I did 280/225/225. Read that's a good starting point to let the chip rip. Though for optimisation you really need to be finding your best settings as the video above shows to reign in the temperature.

All core when running an AIDA64 stress test for me is around 4.55~4.6 at 80 degrees. I'm now testing a -20 curve offset. Being a lazy overclocker and brute forcing the curve till I find my first instability

 

[sandys]

Yeah that's what I did, except instead of using motherboard for PBO I did 280/225/225. Read that's a good starting point to let the chip rip. Though for optimisation you really need to be finding your best settings as the video above shows to reign in the temperature.

All core when running an AIDA64 stress test for me is around 4.55~4.6 at 80 degrees. I'm now testing a -20 curve offset.

Not running a curve offset yet but it's taking quite a bit more power with PBO limits tweaked

My system idles at approx 80w, with PBO on Auto Aida stress load was about 220w, Watt numbers from wall, not software chip numbers, so complete system with PSU efficiency etc.

With the mobo PBO settings Aida stress load is 300w so quite a bump, so as it heats up the clocks do drop but it seems to manage that nicely with a gradual clock drop, it doesn't suddenly chop off 2Ghz like some systems can do.

 

[Audioboxer]

What do you guys prefer to use for testing curve optimiser core stability? I've set up my WHEA error only log custom view and been using AIDA64 as historically its been decent for me catching instability. TestMem5 good for memory, but my memory is 100% stable anyway. Its the 5950x I'm working on.

For example I slapped a -25 all core curve on for a laugh expecting some crashing but an hour of AIDA64 has so far not dumped its own instability reading or any error logging in windows event log.

I've got a feeling this might mean AIDA64 is not going to be hard enough on the CPU to weed out problems. Back to Prime95 or is OCCT good? I've never really used it.

Even with a -25 curve this chip is still sustaining 82-84 degrees CPU Package temp in AIDA lol. Toasty. This is with 270/150/190 on power settings.

 

[Grim5]

What do you guys prefer to use for testing curve optimiser core stability? I've set up my WHEA error only log custom view and been using AIDA64 as historically its been decent for me catching instability. TestMem5 good for memory, but my memory is 100% stable anyway. Its the 5950x I'm working on.

For example I slapped a -25 all core curve on for a laugh expecting some crashing but an hour of AIDA64 has so far not dumped its own instability reading or any error logging in windows event log.

I've got a feeling this might mean AIDA64 is not going to be hard enough on the CPU to weed out problems. Back to Prime95 or is OCCT good? I've never really used it.

Even with a -25 curve this chip is still sustaining 82-84 degrees CPU Package temp in AIDA lol. Toasty. This is with 270/150/190 on power settings.

Let it Idle on desktop for 8 hours, for Ryzen 5000 desktop idling is a better stress test than any benchmark

 

[sandys]

Let it Idle on desktop for 8 hours, for Ryzen 5000 desktop idling is a better stress test than any benchmark

Funny you should say that I set a curve, ran some heavy load tests etc all seems fine, machine is never powered down as it is a NAS and basically idling most of the time, machine rebooted in the night last night, so need to find a way to discover that earlier, will probably put it stock and confirm it is stable, I don't actually need all the power, the chip it replaced has had 40 odd days on solid, only reboots due to me, I've just been having fun playing with the 5950 in the machine as its my only am4 platform, I'll wait 'til I pop it in my main machine before exploring it further.

 

[Audioboxer]

Let it Idle on desktop for 8 hours, for Ryzen 5000 desktop idling is a better stress test than any benchmark

Yeah, good bit of advice lol. When I was using FMax Enhancer with my 3900XT the most frequent instability I would get, or hardest to diagnose, would be reboots when idling.

Presumably due to the short single core burst spikes of these processors being incredibly sensitive to instability. Whereas heavy stability tests tend to check multicore stability.

 

[Audioboxer]

I decided just to see what would happen if I ran this powerplan with the 5950x https://www.computerbase.de/forum/threads/energiesparplan-zen2-ryzen-3000.1934824/

"Fixes" my idle temps to 36~38 degrees lol. This is what I used on my 3900xt. So it just seems that the Windows balanced power plan AMD wants people to use with the 5xxx processors is likely what keeps temps on idle a bit higher. Probably still pumping in more voltage/drawing more power than it needs to.

The issue using these tweaked powerplans whilst gaming or doing any heavy tasks is they probably hinder performance on these new processors. I'll give them a go anyway. There is something satisfying about not seeing your processor spike to 50 degrees simply because... you're using a web browser lol.

 

[Raiden85]

For me on the balanced or ultimate profile with a 5950x the idle temp on a Corsair H150i Pro XT (ML RGB fans) is 30c to 32c (21c room temp) even with plenty of crap running in the background, as I write this there is 273 processes running so 40c+ on balanced seems quite high. This on Windows 10 build 20H2 and the new 21H1. Still using ultimate profile as the CPU still idles nicely while the ultimate gives me better write benchmark performance by quite a bit consistently on a gen 4 SSD.

When I rebuilt this PC in Dec 2020 I'm still using the same build of Windows that all my Intel CPUs ran so it never had the Ryzen power plans installed which AMD scrapped for the 5000 series.

Ryzen like my 9900k did does like to spike when doing light tasks such as opening google in chrome when chrome is already open, easily spikes 10c+, even moving the mouse that's been idle for a few minutes does it as well, which my 9900k didn't do on that one.

 

[Grim5]

I'm just using the performance power plan

but yea regardless of your power plan Ryzen is very spikey, just open a web browser as you said and temps will jump to like 45c

 

[Audioboxer]

For me on the balanced or ultimate profile with a 5950x the idle temp on a Corsair H150i Pro XT (ML RGB fans) is 30c to 32c (21c room temp) even with plenty of crap running in the background, as I write this there is 273 processes running so 40c+ on balanced seems quite high. This on Windows 10 build 20H2 and the new 21H1. Still using ultimate profile as the CPU still idles nicely while the ultimate gives me better write benchmark performance by quite a bit consistently on a gen 4 SSD.

When I rebuilt this PC in Dec 2020 I'm still using the same build of Windows that all my Intel CPUs ran so it never had the Ryzen power plans installed which AMD scrapped for the 5000 series.

Ryzen like my 9900k did does like to spike when doing light tasks such as opening google in chrome when chrome is already open, easily spikes 10c+, even moving the mouse that's been idle for a few minutes does it as well, which my 9900k didn't do on that one.

Can't imagine my Windows balanced powerplan has been edited at any point, but who knows. I think when Windows 11 releases it's clean install time and we'll see if MS has made a good job of it.

Even today after all these years Windows 10 can still be quite janky when updates roll around.

 

[Audioboxer]

Found some more powerplans for people to use because the Windows ones are currently trash https://www.overclock.net/threads/r...for-windows-10-balanced-and-ultimate.1776353/

Hopefully Windows 11 isn't as buggy a mess as Windows 10 has been lately from performance, to audio issues to other bugs.

 

[tazztone]

I had a Deepcool Captain 240 that decided to empty itself a few years ago. It cost me £1000+ to replace the damaged components. Never again will I use an AIO

that cooler is advertised as having "Anti-leak tech" tho

now i am worried my liquid freezer II will leak too

 

[mickyflinn]

that cooler is advertised as having "Anti-leak tech" tho

now i am worried my liquid freezer will leak too

Anti leak tech thats hilarious.

You cant go wrong with top air cooler but i wouldnt worry the likely hoid of a leak is slim.

 

[tazztone]

Anti leak tech thats hilarious.

i know right? https://www.deepcool.com/product/LiquidCooler/Anti-leakPatentInside.shtml their logo is such a rip-off of the intel logo

 

[mickyflinn]

i know right? https://www.deepcool.com/product/LiquidCooler/Anti-leakPatentInside.shtml their logo is such a rip-off of the intel logo

All credit if there trying to minimise leaks but if ifileaks nearer the block your going to get water just maybe a little less.

 

[Audioboxer]

Probably posted a few times but just want to iterate this is a great script for testing individual cores when doing your curve https://github.com/sp00n/corecycler/releases

Currently have

Two best cores = -10
Next two best cores = -5 and -15
Rest = -25

Funnily enough it was one of my 2nd best cores that needed to be dropped down to -5 for stability. Might work at -6~-9 if I test.

My next question is on all core speeds when stability testing/benching. I seem to top out at 4.55ghz across all cores at once. I've seen a few people say they reach 4.6ghz. Wondering if it's my PBO settings or 4.55ghz is quite normal for all the cores to sit at in HWINFO during heavy loads?

 

[Kirkyl85]

Just got my system up and running. Man this is a big jump from my Q6600 ten years ago.

 

[Deaf Spacker]

I've given up trying to tame my 5950X, if it wants to hit 90 degrees then it can.

I tried PBO with a all core negative offset of 10, 200Mhz boost and I tried with scaler set to Auto and X3, it crashed whilst idling on desktop, changed the offset to 5, still crashed.

Then I tried Clock Tuner 2.1, the less said about that the better.

I haven't got the spare time needed to do PBO for each core.

 

[Mr Evil]

...I tried PBO with a all core negative offset of 10, 200Mhz boost...

Don't bother increasing the clock frequency. Cores will only reach the max speed for very short periods of time, if at all, so increasing it won't do anything except reduce stability.

Try just the all core negative offset of 10 without touching any other PBO settings. If you want to go further without spending too much time tweaking, then keep the two "best" cores at -10 and set the rest to -20.

 

[Deaf Spacker]

Don't bother increasing the clock frequency. Cores will only reach the max speed for very short periods of time, if at all, so increasing it won't do anything except reduce stability.

Try just the all core negative offset of 10 without touching any other PBO settings. If you want to go further without spending too much time tweaking, then keep the two "best" cores at -10 and set the rest to -20.

I'll give that go today thank you.