3800X ocing,coolnquiet? am I dreaming?!

learners permit

[H]ard|Gawd
Joined
Jun 15, 2005
Messages
1,766
Although I've only had a couple days to acquaint myself with a shiny new 3800X ,I feel compelled to share my current observations regarding the hardware features aforementioned in the thread title. Historically cool and quiet/ocing were typically not synonymous with AMD cpus. Apparently that is subject to change with the currently available product line. My personal goals with regards to these features have always been to maximize longevity and minimize latent heat while maximizing capabilities of whatever hardware flavor was best at the time. With this in mind here are my short term observations of my new x570 meg and 3800x.
Please keep in mind this system was was built primarily as a gaming rig and not for income related functions ,although I believe it could easily fulfill both obligations with an extra bios profile and another partition on the current 2 TB gen4 nvme raid 0 array.
First and foremost a big thank you to AMD for providing such a wonderfully insightful hardware monitoring utility called Ryzen master and of course the hardware to try it on. Bravo! These settings literally took about 5 minutes to implement after some time to get acquainted with the bios layout of this new MSI main board. Here are some screenshots of ryzen master as a picture is worth quite a few of my long winded mutterings. Forgot to mention this method doesn't require additional cooling as the stock cooler easily manages thermals.
 

Attachments

  • AMD screen save.png
    AMD screen save.png
    1.3 MB · Views: 0
  • AMD screensave2.png
    AMD screensave2.png
    1.3 MB · Views: 0
Last edited:
I gotta ask,did you actually do any testing on games and do you want to share them. Say 8 threads vs 16 threads on any CPU clocks.
Tip also for you,if you want better gaming performance ,fine tune your Ram ,at your present settings I would say your losing up to 20% FPS just in some games from the Ram alone the way you have it set.

Now it is the internet,did you ever use Prime95 before.You have core temp open showing CPU speed.Your Ryzen master shots are LOL.

Why not do something like this and post screens with some real information,such as BIOS settings,gaming with settings running,Prime95 with Ryzen open or HWinfo64 open.If you want to help people,then give them something to go on.

I am not picking on you.

Full Cinebench20 Single/Multi thred run with HWinfo64 open with all setting.


AMD 3800X All Core Overclock 4650Mhz Pc Gaming
 
You need to run a real benchmark with a score between stock and this undervolt you got going on. Cinebench 20 or 15. You are not actually running 4.6 at that low voltage. It's called clock stretching.
 
@ gerardfraser please go troll some other thread. @ thumper thanks for the tip but let me assure this is not the case unless the cpu is at 100% all cores and nowhere in my post did I mention undervolting my cpu. Actual voltage for the cpu is 1.5 volts. I'd be happy to share how this was achieved as long as you promise not to share this info with the obviously superior intellectual troll who enjoys feebly attempting to prove himself as such by berating those who fall short of his stringent standards.
 
Last edited:
What kind of memory are you running at 2667Mhz and Cas 20? I would think that would tank performance (relatively speaking).
 
Memory is here. The timings i let the board set cuz I figure bios engineers know a bit more than I do about proper memory timings. Regarding performance I do not notice any difference other than complete stability at these settings.
 
Last edited:
Memory is here. The timings i let the board set cuz I figure bios engineers know a bit more than I do about proper memory timings. Regarding performance I do notice any difference other than complete stability at these settings.
delete
 
Last edited:
Single core score is nice. Multicore score is very low. Like not close to a normal 3800x... I'll post mine later.
 
My multi. No idea what my single is. I'm too lazy to wait for it.
 

Attachments

  • cb1.jpg
    cb1.jpg
    259.2 KB · Views: 0
Something doesn't look right with your scores. If you're running it and the voltage is 1.1V due to a bios setting or whatever, then you most certainly aren't hitting a 4.6Ghz all core, and your CB score reflects that. THUMPer's posted video shows when Gamers Nexus was doing testing with low voltages like yours and the results were about a 20-30% drop in scores/benches.
 
Can you turn smt on for me? Pretty please? Curious of your score with it on. Bring out the big guns! ;)
 
You know ht loads cores harder and it will not work at this voltage but that was never claimed so after the discussion has begun you want me to try something I never stated was possible so you can what?
 
Just wanted to mention you leave a LOT of gaming performance on the table by running RAM at 2666 and CAS20 at that.
 
Thanks I'm aware of this but the added heat and instability are not worth the effort of working out tighter timings and fighting the instability. The gains to me are negligible. My wife makes me keep my office door shut when gaming and the vega 64 already makes it way too hot in here.
 
Thanks I'm aware of this but the added heat and instability are not worth the effort of working out tighter timings and fighting the instability. The gains to me are negligible. My wife makes me keep my office door shut when gaming and the vega 64 already makes it way too hot in here.

Wrong. Actually all-core overclocking Ryzen 2 CPU is not worth it, they are pretty much maxed out out of the box with possibly a little bit of headroom left for PBO, you just need to enable it and you are done.

Most of the performance gains is to be had from tweaking memory speed and timings because it has a direct effect on how fast the Infinity Fabric operates. It reduces the lag when cores "talk" to each other. With Intel memory did not matter very much, but with AMD Ryzen it is the very most important thing to tweak if you want the max performance the CPU can give.

And the heat output of memory chips is minimal.
 
With those memory timings you will be so memory limiting that the core would be waiting, less load, for memory writes and reads in multiple core applications. It would also significantly degrade gaming benchmarks but it would be up to you if your gaming experience is affected or not. Looking at your memory usage in Prime95, you are using the blend test, memory intensive which due to memory speeds and timings is basically throttling the CPU meaning not as high of voltage would be needed.

This is more like a magicians trick, yes 4.6ghz but really not doing 4.6ghz worth of work. The single thread test is mostly CPU cache workloads on CB and that does reflect 4.6ghz core speed, anything going beyond the cache of the CPU with your memory settings would tank.

If you are not going to use the cores potentials etc. because of heat, maybe a 65w 3700x vice the 105w 3800x would work better for you. Interesting behavior in the end and thanks for sharing.
 
Yea I tried tighter timing and higher speed on the memory and it reduced fps for gaming and caused hitching so I left it there.I finally settled on the 46.5 multiplier and have had 0 CTD or failures to post. Configured this way gaming performance is much smoother than my old 6950x. I'm very happy with this system and it never exceeds 73c with the stock cooler.
 
Memory is here. The timings i let the board set cuz I figure bios engineers know a bit more than I do about proper memory timings. Regarding performance I do not notice any difference other than complete stability at these settings.

Why did you buy 3600 memory to run it at 2667? Run the memory at 3600 so your infinity fabric runs where it is supposed to, 1800Mhz.
 
Because it was cheap enough to fit the budget I allotted for the build and has headroom if needed later. Why do I feel like I'm being interrogated/berated for a crime in this forum for posting my configuration choices???
 
Because it was cheap enough to fit the budget I allotted for the build and has headroom if needed later. Why do I feel like I'm being interrogated/berated for a crime in this forum for posting my configuration choices???

I asked because it would limit performance of the processor itself running the IF at a lower speed and the memory at a lower speed. It is your money though so if you want to spend it and not get the most out of what you spent then ok. I was just asking. This is an enthusiast board so sometimes I just assume people are trying to get the most out of their systems they built. Carry on.
 
And I do not believe for one second you would get worae performance by tinkering with lower memory timings unless you do now know what you are doing and have unstable timings. But if you trust the manufacturer more than yourself, then for the love of god just enable the XMP Profile (or DOCP some mobo's call it) that allows the memory run at the factory settings the memory was designed for, instead of slow failsafe mode that runs by default.
 
After some time to get better acquainted with the new rig I decided to try to get stable with smt on. Couldn't do 4.6 all core with smt on because its air cooled and thermals went ape shit but 4.5 was passable even though the cpu peaked @103C during the all core run. R20 smt on.png
 
You know ht loads cores harder and it will not work at this voltage but that was never claimed so after the discussion has begun you want me to try something I never stated was possible so you can what?

It's not Hyper-Threading. It's called SMT. Look it up.
 
Hehe cpu power peaked at nearly 170 watts but don't worry I installed a CO2 fire extinguisher in there. :) Anyway I'll keep smt off and go back to 4.65 for the all core and still remain at 107 watts or so for heat dissipation. That was just a run to prove the nay sayers wrong about being unable to run to run at said [email protected] volts without the clock stretching horseshyt they were spewing out.
 
Last edited:
Back
Top