First AMD build in 20 years!

III_Slyflyer_III

[H]ard|Gawd
Joined
Sep 17, 2019
Messages
1,247
Well, Microcenter had some 5950x in stock Saturday, so i decided it was time to upgrade my trusty 5960x haswell-e! First time on team red since my Athlon XP 2800+. Paired it with an MSI Meg ACE board (best MC had in stock). Board seems pretty good so far. Have a 3 RAD H150i XT cooling it and does damn good so far.

Any advise on how to tune these chips or if it is even worth the time and effort? I'm used to OCing where you could gain 20% or more, but times seemed to have changed quite a bit.

I dunno if I even want to OC past the RAM. Seems like this thing does better on its own. Hits 4.6~4.7 multi core gaming with spikes at 4.9~5.0 single core. Under 100% all core load (realbench burn test) its at like 4.1~4.2 all cores solid, temps under 60C.

When I tried a quick dirty PBO overclock (not knowing alot with AMD yet), my multithreaded score went up 10%, but single core suffered slightly (maybe ~2%) and temps were much warmer (80C+).

Everything i read says the performance gains on AMD Zen 3 are in flck/memory and thats about it. Guess I can still raise the minimum all core, but that keeps temps higher and robs from the single core boost when I need it in games from my understanding.

Its fun to watch coming from intel... my all core 100% load across all 32 threads at 4.2 is only 1.05V! When a single core boosts up to 4.8~5.0, Vcore will shoot up to like 1.48V for a split second of time.

It really does know what its doing i guess.

So far got my ram OCed up to 3400Mhz CAS14 with 1700Mhz FLCK. Gonna try for 3600Mhz+ later this week.

So far, pretty happy with this build moving from Intel. :)
 
Well, Microcenter had some 5950x in stock Saturday, so i decided it was time to upgrade my trusty 5960x haswell-e! First time on team red since my Athlon XP 2800+. Paired it with an MSI Meg ACE board (best MC had in stock). Board seems pretty good so far. Have a 3 RAD H150i XT cooling it and does damn good so far.

Any advise on how to tune these chips or if it is even worth the time and effort? I'm used to OCing where you could gain 20% or more, but times seemed to have changed quite a bit.

I dunno if I even want to OC past the RAM. Seems like this thing does better on its own. Hits 4.6~4.7 multi core gaming with spikes at 4.9~5.0 single core. Under 100% all core load (realbench burn test) its at like 4.1~4.2 all cores solid, temps under 60C.

When I tried a quick dirty PBO overclock (not knowing alot with AMD yet), my multithreaded score went up 10%, but single core suffered slightly (maybe ~2%) and temps were much warmer (80C+).

Everything i read says the performance gains on AMD Zen 3 are in flck/memory and thats about it. Guess I can still raise the minimum all core, but that keeps temps higher and robs from the single core boost when I need it in games from my understanding.

Its fun to watch coming from intel... my all core 100% load across all 32 threads at 4.2 is only 1.05V! When a single core boosts up to 4.8~5.0, Vcore will shoot up to like 1.48V for a split second of time.

It really does know what its doing i guess.

So far got my ram OCed up to 3400Mhz CAS14 with 1700Mhz FLCK. Gonna try for 3600Mhz+ later this week.

So far, pretty happy with this build moving from Intel. :)
The biggest disappointment for me with Ryzen is how PBO just sucks all the fun out of tuning the system. Unless you get crazy with water and mad voltages its just single digit percents available. Not really a bad thing, but overclocking just isn't what it used to be. I wish i kept my 300A, it was fun (gave it away, along with my voodoo 3 3500TV)

Congratz on getting a hold of that chip, hope you got to keep all your limbs.
 
Use PBO advanced settings and set up curve optimizer. Easiest way to maximize total performance across all cores. I started at -20 for all cores and was happy with that but you can tune each core individually if you want to spend the time to verify stability with every change. That gets very time consuming so I settled for -20 all cores.
 
I went from a 7700K to a 5800X. Last AMD rig I had was during the 939 days with an FX55. I overclocked the old fashioned way using the multi/voltage. Tried using the PBO/curve method and for the life of me can't figure the dam thing out.
 
Find the most negative value for all cores that runs stable and use that as a baseline. Then you can tune each core seperately if you aren't happy with the result. Expect to spend a lot of time for stability testing if tuning individually.
 
Is more to be gained tuning PBO (gonna have to read up on modifying these curves), or is more to be gained trying to get fclk/memory 1:1 as high as possible with lowest timings (probably 1800Mhz on my kit)?
 
Is more to be gained tuning PBO (gonna have to read up on modifying these curves), or is more to be gained trying to get fclk/memory 1:1 as high as possible with lowest timings (probably 1800Mhz on my kit)?
It's not a one or the other thing, you can do both. I have RAM at 3800mhz with 1900mhz IF and currently a -15 offset for all cores. The RAM and IF overclock will help as will the higher all core clockspeeds I get from the Curve Optimizer offset.

Keep in mind that the CO offset can lower the highest single core clockspeeds because the best cores will usually need a bit more voltage to hit the higher clockspeeds. This is one of the reasons people will do different CO offsets for different cores. Larger CO offsets for cores which don't boost high to begin with and smaller offsets for the "golden" cores which are likely to boost the highest in single thread or low thread loads.

I haven't bothered with setting individual offsets for different cores because I don't want to bother with the time it takes to test it and I almost always have high thread workloads going anyway.
 
So I decided to mess with the PBO curve... did a quick and dirty all core -20 voltage offset; then lowered PBO limits to: PPT = 200, TDC = 200, EDC = 150 to keep things cooler to allow for better boosting.

Currently have all core stable at 4.5Ghz @ 1.176V and it boosts single core at 5.0Ghz+ @ 1.448V. Temps stay under 65C in Realbench and Cinebench R20. Single Core test does not even raise temps much past idle... lol.

This a decent quick and dirty OC for these CPUs?

5950x OC.png
5950x OC All Core.png
 
I also picked up a 5950x and an Asus Crosshair Dark Hero board at Microcenter on Saturday and plan on building it Friday. I’m following your thread with great interest because while I have lots of OC experience with Intel, I have little with AMD and hope to learn from the guys in this thread.
 
Yea not bad at all but don't trust cpu-z accuracy because of the dynamic nature of PBO. Use HWinfo it better represents your actual cpu core speeds and voltages.
 
I also picked up a 5950x and an Asus Crosshair Dark Hero board at Microcenter on Saturday and plan on building it Friday. I’m following your thread with great interest because while I have lots of OC experience with Intel, I have little with AMD and hope to learn from the guys in this thread.
Wonder if it was the same MC as mine! lol. Yeah, OCing on this is way different than what I'm used to! :)

Yea not bad at all but don't trust cpu-z accuracy because of the dynamic nature of PBO. Use HWinfo it better represents your actual cpu core speeds and voltages.
Yeah, HWinfo matched for single and multi... cpuz was just an easy way to show it. I had both running.
 
Wonder if it was the same MC as mine! lol. Yeah, OCing on this is way different than what I'm used to! :)


Yeah, HWinfo matched for single and multi... cpuz was just an easy way to show it. I had both running.

I see you’re in Ohio, so definitely possible. :). I live outside of Indianapolis and drove over to the Sharonville store Saturday morning and was there around noon. They have now sold out of their large 5950x shipment - I probably should’ve bought a second to try to trade for a GPU.
 
I see you’re in Ohio, so definitely possible. :). I live outside of Indianapolis and drove over to the Sharonville store Saturday morning and was there around noon. They have now sold out of their large 5950x shipment - I probably should’ve bought a second to try to trade for a GPU.
Just go north on I71... ;)

Also, at least with my Microcenter, you could only buy 1 CPU.
 
So I decided to mess with the PBO curve... did a quick and dirty all core -20 voltage offset; then lowered PBO limits to: PPT = 200, TDC = 200, EDC = 150 to keep things cooler to allow for better boosting.

Currently have all core stable at 4.5Ghz @ 1.176V and it boosts single core at 5.0Ghz+ @ 1.448V. Temps stay under 65C in Realbench and Cinebench R20. Single Core test does not even raise temps much past idle... lol.

This a decent quick and dirty OC for these CPUs?

View attachment 347328View attachment 347329
Looks like you got a really nice sample. That voltage for all core is impressive.
 
I see you’re in Ohio, so definitely possible. :). I live outside of Indianapolis and drove over to the Sharonville store Saturday morning and was there around noon. They have now sold out of their large 5950x shipment - I probably should’ve bought a second to try to trade for a GPU.
I live 2 hours from the nearest MC in Columbus. :(
 
Just go north on I71... ;)

Also, at least with my Microcenter, you could only buy 1 CPU.

Yeah, they might have limited me on the CPU - I didn't really ask. They did enter my information into their system and included a printed form on checkout which had the video card information, as they told me I could only buy 1 card every 30 days.

I'm anxious to get my 5950X built on Friday. I ordered some more RAM and a new NVMe drive for the OS. Once I get everything loaded up, I'll try to do some OCs. I'm using my Noctua NH-D15S with dual fans with the Noctua thermal paste - is it worth going with a better paste?
 
I live 2 hours from the nearest MC in Columbus. :(

I'm about 2 hours from the Sharonville store. They had the 5950X in stock a few weeks ago and I passed but wasn't going to pass this time. They also had the Dark Hero board in stock for MSRP and with the CPU, I got the discount too. Hoping I can get it built Friday and then do some reading this weekend and pick the collective brains of the internet on overclocking. I'm using my NH-D15S as the cooler but if I don't get good results, definitely not opposed to buying a 360 mm or 420 mm AIO. I'm just a little gun shy with AIOs because my previous Arctic Freezer's pump died like a month after the warranty ended and when it happened, my system temp skyrocketed and it didn't seem like thermal shutdown happened fast enough and I've had occasional odd issues ever since. I'm hoping to compare the OP's H150 performance to my Noctua performance to see if I should upgrade.
 
III_Slyflyer_III, tons of overclockers on this forum but with the latest 5000 series AMD cpus, your chance of a substantial OC is almost nill.

I think AMD pushed these cpus hard at the factory even at stock so I doubt there is much OCing headroom.

I keep my custom watercooled 5900x and RX 6800 stock and enjoy the performance. You might get a bit more performance by installing DDR4-3600 ram but that's about it.

You already have a MONSTER rig that will serve you well for a number of years.
 
I'm about 2 hours from the Sharonville store. They had the 5950X in stock a few weeks ago and I passed but wasn't going to pass this time. They also had the Dark Hero board in stock for MSRP and with the CPU, I got the discount too. Hoping I can get it built Friday and then do some reading this weekend and pick the collective brains of the internet on overclocking. I'm using my NH-D15S as the cooler but if I don't get good results, definitely not opposed to buying a 360 mm or 420 mm AIO. I'm just a little gun shy with AIOs because my previous Arctic Freezer's pump died like a month after the warranty ended and when it happened, my system temp skyrocketed and it didn't seem like thermal shutdown happened fast enough and I've had occasional odd issues ever since. I'm hoping to compare the OP's H150 performance to my Noctua performance to see if I should upgrade.
The D15s is a beast. It's currently on my 3900x. It's large, however, and blocks my gpu retention clip. I have to remove the fans and use a plastic stick to disengage the clip. Soooo, I'm switching to an Arctic Liquid Freezer 420 when I replace the 3900x with a 5800x. It WILL NOT block the clip and should be more than enough for the eventual 5900x I purchase.
 
One annoying problem I have been trying to figure out over the last couple of days is why under gaming load or heavy F@H load, my sound card will flip out and mess all the channels up. I have a 5.1 Surround system hooked up via optical to my SBXz through Tosslink, so it's digital in nature.

I updated all of my Chipset Drivers; Sound Card Drivers; Windows updates... etc. I just tried updating my BIOS to a Beta BIOS that has the AGESA ComboAM4v2PI 1.2.0.2 firmware. Hopefully that solves it, but I will keep everyone posted. It only happens under load too; which is odd.

As of right now; I have my Video Card on PCIe 4.0 x16 (CPU Lanes), 2 nvme drives (one is PCIe 4.0 and the other is 3.0); and i have my Sound card on the PCH PCIe 4.0 1x slot.

Some places suggested going to PCIe 3.0, but I'm trying the new BIOS first. Looks like I can set the PCH lanes to 3.0 and still keep the CPU to my video card at 4.0, which would be fine with me.

One important note: make sure to update your BIOS as AGESA ComboAM4v2PI 1.2.0.2 fixes some major L3 Cache performance issues I found out! Went from L3 numbers in the 350 GB/S to 950 GB/S+ now!
 
UPDATE: New Bios + Drivers fixed the issue (so far)! :) Also found some instability with my memory overclock; so I just defaulted to my XMP of 3200Mhz CAS 14 for now to rule that out as causing any issues. Was able to keep my custom PBO Curve and Boosting though!

Not sure if buying a 3600Mhz+ Kit is worth it though over my 3200Mhz CAS 14 kit (which I pulled from my X99 system)...
 
My first AMD build in over 40 years! :D (<-- Need a grinning old man version of this!)

The sound issue you were experiencing may have been related to the USB issues/problems that are now fixed as of AGESA 1.2.0.1 beta and higher. I use an external DAC/AMP and was previously also experiencing sound issues (mostly intermittent pops/crackles) that now seem to be fixed as of few BIOS updates ago.

Besides the previous tools mentioned, I've found that OCCT (Ver 8.1.1) is also a very useful for determining/discovering potential CPU errors that may result due to O/C'ing.

Very happy with the performance of my rig. Not doing anything crazy as to O/C, just have a good custom water loop on CPU/MB/GPU, with PBO turned on and lots of speedy peripherals in the form of PCIe 4.0 capable GPU and SSD.
 
Last edited:
Out of curiosity, RAM Speed on Intel was not too big of a factor. On AMD 5000 Series; am I better suited running at 3200Mhz CAS 13 (or lower with voltage); or am I better suited trying to get my RAM above 3500Mhz with looser timings (CAS 15 or 16)?

I have a Samsung B-Die Kit; but im having trouble getting anything above 3400Mhz stable with reasonable CAS and at 1.4V or less. However; with 1.4V on the DDR4, I can run 3200Mhz with CAS 13, and testing CAS 12 right now.

Does AMD prefer RAM Speed or Tight Timings for performance? Would it even be worth buying new sticks at 3600Mhz or is the performance gain negligible for gaming?
 
Out of curiosity, RAM Speed on Intel was not too big of a factor. On AMD 5000 Series; am I better suited running at 3200Mhz CAS 13 (or lower with voltage); or am I better suited trying to get my RAM above 3500Mhz with looser timings (CAS 15 or 16)?

I have a Samsung B-Die Kit; but im having trouble getting anything above 3400Mhz stable with reasonable CAS and at 1.4V or less. However; with 1.4V on the DDR4, I can run 3200Mhz with CAS 13, and testing CAS 12 right now.

Does AMD prefer RAM Speed or Tight Timings for performance? Would it even be worth buying new sticks at 3600Mhz or is the performance gain negligible for gaming?
speed is important because the infinity fabric speeds are tied to the ram bus. Timings are important because the ram throughput increases with tighter timings. For gaming you want both. But the increases aren't worth lots of money.

I cant find the reference, but there is a sweet spot for speed, where going higher brings in a new clock divider and lower infinity fabric speeds, I think it was around 3600, but I dont recall.
 
I recently also went back to AMD for the first time since the Athlon days. I went with a 5800X at Microcenter. So far, so good...although frankly I don't notice a ton of speed difference vs. my previously OC'd 8700K. My wife was in need of something new, so the upgrade made sense since she gets the hand-me-downs. The joys of most of my work (and 4K gaming) being almost entirely GPU-focused. Still, it is interesting going back to team red. Seems a lot of folks are. I'll have to tinker with the OC'ing options. I got pretty used to how easy it was on Intel and will have to figure out how the curve stuff works.
 
Out of curiosity, RAM Speed on Intel was not too big of a factor. On AMD 5000 Series; am I better suited running at 3200Mhz CAS 13 (or lower with voltage); or am I better suited trying to get my RAM above 3500Mhz with looser timings (CAS 15 or 16)?

I have a Samsung B-Die Kit; but im having trouble getting anything above 3400Mhz stable with reasonable CAS and at 1.4V or less. However; with 1.4V on the DDR4, I can run 3200Mhz with CAS 13, and testing CAS 12 right now.

Does AMD prefer RAM Speed or Tight Timings for performance? Would it even be worth buying new sticks at 3600Mhz or is the performance gain negligible for gaming?
Read this: https://hardwarecanucks.com/memory/choosing-the-best-amd-ryzen-5000-memory-a-beginners-guide/

Bottom line, having really tight timing seems a bit less important than overall RAM speed when it comes to performance on the AMD 5000 series. Granted you'll eek out a bit more having both speed and tight timings as travm mentions above, but for the gains vs price paid, it doesn't seem like it is worth chasing. The price/performance sweet spot really seems to be with 3600mhz RAM for the 5000 series.

I have 3600Mhz/CAS 14 RAM.

Comparison from the article above for the Doom Eternal benches, avg frame rates are:

3600Mhz/CAS 14 = 400.4 FPS
3600Mhz/CAS 16 = 397.2 FPS
3600Mhz/CAS 18 = 395.2 FPS
3600Mhz/CAS 20 = 393.3 FPS

Compared with RAM Speed, when they are all CAS 18:
4000Mhz/CAS 18 = 405.8 FPS
3800Mhz/CAS 18 = 402.8 FPS
3600Mhz/CAS 18 = 395.2 FPS
3200Mhz/CAS 18 = 381.7 FPS

So jumping from 18 CAS down to 14 CAS with 3600MHZ RAM garners roughly a 5 FPS increase.
But jumping from 3600/18 RAM up to 4000/18 RAM brings with it a roughly 10 FPS increase.

I could switch to a 4000Mhz/CAS 18 RAM kit which would outperform my 3600Mhz/CAS 14 kit by 5 FPS, or 1.3%. Doesn't seem worth the cost/effort though, especially given that hitting Infinity Fabric 1:1 isn't always guaranteed on a 4000Mhz kit... To me at least, 3600Mhz with tight timings seemed like a good/safe choice.
 
What happens if you change the speed and hold latency constant? Keeping CAS 18 means longer latency at the slower speeds. Are you able to run 3200 CAS 16, 3600 CAS 18, 3800 CAS 19, 4000 CAS 20 holding the latency at 10 ns?
 
I’ve got 64 GB of Gskill Trident b-die (3200, CAS14) in 4 DIMMS which I’m going to use with my 5950x and Asus Crosshair Dark Hero board. What do you guys think is a reasonable overclock? I don’t want to get too crazy and want to be stable, but I’d like to bump it up if possible.
 
I’ve got 64 GB of Gskill Trident b-die (3200, CAS14) in 4 DIMMS which I’m going to use with my 5950x and Asus Crosshair Dark Hero board. What do you guys think is a reasonable overclock? I don’t want to get too crazy and want to be stable, but I’d like to bump it up if possible.
I found this to be a helpful start... don't expect to get much more out of your chip though, but its still fun to get that extra 1~3%!!!! 😆

https://www.overclock.net/threads/r...izer-to-5-1-ghz-pbo-and-overclocking.1774434/

As for Ram, the more I read, the more I'm just keeping my 4 sticks of 32GB 3200Mhz and running them at 13-13-13-26 1T. I game at 4K and there is not even a statistical difference for me above 3200Mhz at that resolution. Your talking within the margin of error for a test run in a game... lol. Check this article....

https://www.techspot.com/article/2140-ryzen-5000-memory-performance/
 
I found this to be a helpful start... don't expect to get much more out of your chip though, but its still fun to get that extra 1~3%!!!! 😆

https://www.overclock.net/threads/r...izer-to-5-1-ghz-pbo-and-overclocking.1774434/

As for Ram, the more I read, the more I'm just keeping my 4 sticks of 32GB 3200Mhz and running them at 13-13-13-26 1T. I game at 4K and there is not even a statistical difference for me above 3200Mhz at that resolution. Your talking within the margin of error for a test run in a game... lol. Check this article....

https://www.techspot.com/article/2140-ryzen-5000-memory-performance/

Yes, the performance differences I was pointing out (very low single digit percentages) really only applies to 1080p gaming... And even then, you would be very hard pressed to notice the difference between say 380 fps vs 405 fps. :D
Once you shift up to 4K gaming, that gap also becomes much smaller.

Unless you are really into memory benching or running really big render jobs, the difference in performance between 3200Mhz RAM and 4000Mhz RAM for regular gaming/productivity on a 5000 series AMD CPU is going to be extremely negligible.
 
Thanks guys, that’s what I suspected. I probably won’t devote too much time to tweaking the RAM at this time.
 
Well, Microcenter had some 5950x in stock Saturday, so i decided it was time to upgrade my trusty 5960x haswell-e! First time on team red since my Athlon XP 2800+. Paired it with an MSI Meg ACE board (best MC had in stock). Board seems pretty good so far. Have a 3 RAD H150i XT cooling it and does damn good so far.

Any advise on how to tune these chips or if it is even worth the time and effort? I'm used to OCing where you could gain 20% or more, but times seemed to have changed quite a bit.

I dunno if I even want to OC past the RAM. Seems like this thing does better on its own. Hits 4.6~4.7 multi core gaming with spikes at 4.9~5.0 single core. Under 100% all core load (realbench burn test) its at like 4.1~4.2 all cores solid, temps under 60C.

When I tried a quick dirty PBO overclock (not knowing alot with AMD yet), my multithreaded score went up 10%, but single core suffered slightly (maybe ~2%) and temps were much warmer (80C+).

Everything i read says the performance gains on AMD Zen 3 are in flck/memory and thats about it. Guess I can still raise the minimum all core, but that keeps temps higher and robs from the single core boost when I need it in games from my understanding.

Its fun to watch coming from intel... my all core 100% load across all 32 threads at 4.2 is only 1.05V! When a single core boosts up to 4.8~5.0, Vcore will shoot up to like 1.48V for a split second of time.

It really does know what its doing i guess.

So far got my ram OCed up to 3400Mhz CAS14 with 1700Mhz FLCK. Gonna try for 3600Mhz+ later this week.

So far, pretty happy with this build moving from Intel. :)

Eh...

It used to be that overclocking got you some free performance. As others have pointed out, there's not much headroom on these chips. Since these chips are already wicked fast it's not much of an issue (at least for me).

I'm a gamer but it's way down the priority list compared to my professional activities. I've not "needed" to overclock since running a Phenom II x6.

These days, it seems to me, that the real advantages of the platform is stability, speed, and the option of ECC RAM in many configurations.

That would be my take on it... YMMV...
 
Just to chime in here, but you enabled smart access memory//resizable bar support, right? The RTX 3000 series just got it and you might get a 5-12% increase in FPS gaming. It looks like you got most of the other tweaks mentioned.
 
Just to chime in here, but you enabled smart access memory//resizable bar support, right? The RTX 3000 series just got it and you might get a 5-12% increase in FPS gaming. It looks like you got most of the other tweaks mentioned.
Yeah, actually flashing my 3090 tonight... wanted to do some tests first to see differences in the platform change without that being a factor.
 
Will I need to flash my Asus Tuf 3070 for it to utilize the resizable bar?
 
Will I need to flash my Asus Tuf 3070 for it to utilize the resizable bar?
Yes - all GPU's that can support it will need a BIOS re-flash/update before they actually work with it. (MB Bios update as well). Eventually, new GPU's will come with the necessary BIOS baked in out of the box.
 
It's not a one or the other thing, you can do both. I have RAM at 3800mhz with 1900mhz IF and currently a -15 offset for all cores. The RAM and IF overclock will help as will the higher all core clockspeeds I get from the Curve Optimizer offset.

Keep in mind that the CO offset can lower the highest single core clockspeeds because the best cores will usually need a bit more voltage to hit the higher clockspeeds. This is one of the reasons people will do different CO offsets for different cores. Larger CO offsets for cores which don't boost high to begin with and smaller offsets for the "golden" cores which are likely to boost the highest in single thread or low thread loads.

I haven't bothered with setting individual offsets for different cores because I don't want to bother with the time it takes to test it and I almost always have high thread workloads going anyway.

I'm currently using -15 on my two fastest cores and -5 on the rest. Based on your post you are recommending a lower negative offset for the best cores vs higher?
 
So I decided to mess with the PBO curve... did a quick and dirty all core -20 voltage offset; then lowered PBO limits to: PPT = 200, TDC = 200, EDC = 150 to keep things cooler to allow for better boosting.

Currently have all core stable at 4.5Ghz @ 1.176V and it boosts single core at 5.0Ghz+ @ 1.448V. Temps stay under 65C in Realbench and Cinebench R20. Single Core test does not even raise temps much past idle... lol.

This a decent quick and dirty OC for these CPUs?

View attachment 347328View attachment 347329

I'm struggling with temps now. All I've done so far on my Dark Hero is enable XMP and turn PBO on - I haven't done any other tuning. Idle temps are fine IMO (mid 30s to low 40s) but I hit 88 on Cinebench R20. I've got a few guides to read about which settings to tweak but I'll probably pull the Noctua NH-D15s off next weekend and apply some new thermal paste. I really would be surprised if that's the issue though.

Your temps look great IMO and I may borrow some of your settings. I've been doing a lot of reading and it seems Asus boards might be pushing too much voltage, but I need to dig into it more.
 
^^^ I’ve been on a custom water loop for 10+ years now... feels good to never have had to even consider using the phrase “I'm struggling with temps” :D
 
I'm struggling with temps now. All I've done so far on my Dark Hero is enable XMP and turn PBO on - I haven't done any other tuning. Idle temps are fine IMO (mid 30s to low 40s) but I hit 88 on Cinebench R20. I've got a few guides to read about which settings to tweak but I'll probably pull the Noctua NH-D15s off next weekend and apply some new thermal paste. I really would be surprised if that's the issue though.

Your temps look great IMO and I may borrow some of your settings. I've been doing a lot of reading and it seems Asus boards might be pushing too much voltage, but I need to dig into it more.
If your on a 5950x, your not cooling that with anything less than a H2O AIO.

I think all AMD boards over-volt from what I have been reading. I have manually set almost all of my voltages now. My mobo was putting 1.2V on the SOC (NB/Chipset)! Way above normal, so I set that to 1.05V along with a bunch of other voltages that were quite a bit higher than stock.

I recommend starting maybe with my settings (or close) as a quick OC. The lower current limits seem to have helped a lot with maintaining boosts and keep temperatures in check.

Pretty happy with my CPUz scores for now :)

Capture4.JPG
 
Back
Top