edgetrading
Bronze Level Poster
Are there any undervolting tutorials for the ASUS B550 motherboard?
Undervolting tutorials?
- Thread starter edgetrading
- Start date
SpyderTracks
We love you Ukraine
It's the same process on any motherboard, just google, undervolting, it's a method more than something that can be shown.
I managed to find your build from one of your previous threads. I can't stress how important it is to seek advice from the forum before going ahead with such a purchase.
You have combined an absolute monster of a CPU with a very low end board (by comparison), a CPU fan that as much as is advertised to do a fine job will be at its absolute limit all inside a case apparently designed to cook bread in with a passive GPU to simply add to the temps. There is no surprise from me that you are having temperature/fan issues.
Each of these components in solitude will fit together absolutely fine but they have completely differing purposes. The case is fine for normal office/desktop use, the CPU cooler is excellent for mid-level chips (as is the motherboard).
Built together is very ill advised and as much as undervolting may help, I won't personally be getting involved in offering any advice on such a build as I wouldn't want to be linked to any failures or future issues.
You have combined an absolute monster of a CPU with a very low end board (by comparison), a CPU fan that as much as is advertised to do a fine job will be at its absolute limit all inside a case apparently designed to cook bread in with a passive GPU to simply add to the temps. There is no surprise from me that you are having temperature/fan issues.
Each of these components in solitude will fit together absolutely fine but they have completely differing purposes. The case is fine for normal office/desktop use, the CPU cooler is excellent for mid-level chips (as is the motherboard).
Built together is very ill advised and as much as undervolting may help, I won't personally be getting involved in offering any advice on such a build as I wouldn't want to be linked to any failures or future issues.
I take it back on the case, it was another spec that was being commented on that had the 175. The meshify is fine for use.
The rest stands though and it's no surprise. I don't even remember commenting to be honest, I have an awful memory, but it just goes to show how we don't suggest such combinations without a deeper understanding than may be apparent on the face of it.
The rest stands though and it's no surprise. I don't even remember commenting to be honest, I have an awful memory, but it just goes to show how we don't suggest such combinations without a deeper understanding than may be apparent on the face of it.
Is the 3600x ok for B550 using the Noctua? My started building this morning so I can't correct anything now, but I downgraded it late last night as I couldn't make my mind up and it had been said numerous times that my cpu (originally 3700x) was more powerful than it should be for my uses.
Thats a relief then
It's a bit of a strange combo. With the 3600X and the B550 you have no need for the Noctua. It's a bit of a waste of money as it won't add anything to the build.
I would have opted for the 3600XT though, with the stock fan (comes with the highly sought after RGB fan).
I got lost in your build thread though, was like Rachel making Trifle in Friends
Will the noctua be enough to keep it cool? I was going to ask this morning but they started building it before I had chance, unexpected as I got an email saying Monday was the build date. i only changed it as I'd read that my 3700x was too powerful for the 1660 super. so I thought downgrading the cpu and upgrading the gpu made it more even
Yes, it'll be more than fine. My point is more that it's overkill and un-necessary (It's a random choice to land on given your scrutiny in other areas and wanting to cut corners and save money).
Changing based on a comment from someone who, reading between the lines, is making stuff up as they go along isn't best advised. A CPU is not linked to a GPU in regards to their output or performance, so one cannot be too powerful for the other. What you can have is a situation where certain games are GPU or CPU bound, this is not a reason to lower performance of a system.... especially when it comes to frequency (which the 3700X is higher with). On the flip side though, if you decide to upgrade that GPU in a couple of years for something a bit more beefy..... you're 3600X is likely going to end up being the limiting factor..... where the 3700X would have given more life (or even the 3800XT which I think is about as good as you can get looking to the future right now).
This all seems moot now though, as it did at the time of discussion to be honest. We tried to go around all this in your thread with you but there wasn't a lot of coherent thought in the thread from yourself.... it was very erratic, randomly darting off in different directions depending which Reddit post was favoured that particular day..... at least that's how it came across.
You're right but as I was in pre production I am relying on this info to try and make a decision and it was like I was on a timer before it was too late and now it is I couldn't justify the cost of a 3700x and 2060 super, the only way it became reasonable was by reducing the processor.. maybe I will regret it but I did take everyones thoughts on board though so appreciate it.
I couldn't justify the cost of a 3700x and 2060 super, the only way it became reasonable was by reducing the processor.. maybe I will regret it but I did take everyones thoughts on board though so appreciate it.
There's nothing at all wrong with the 3600, it's a highly recommended chip. My point is the logic you gave for choosing it is flawed. The 3700X is a better chip with a better future no matter which way you slice it.
The recommendation was based on your budget and uses, the same as the user you quoted was.
The recommendation was based on your budget and uses, the same as the user you quoted was.
edgetrading
Bronze Level Poster
Thanks for your responses. I did some reading on undervolting, and given the risk/complexity, think I'll just avoid it. But I still think the build is valid.
When under 100% load I'm getting CPU temps of 70C (using either my actual work code, Prime95 or Folding@Home) when I set fans to 50% (which keeps it quiet). Or 65C if I set them to 100%.
My CPU temperature is therefore in the same ballpark as the H150i in this review(which lists the CPU temp at 75C when under load), if you look at the "Stock CPU Temperature Under Load" chart, and that test is done with an expensive £549.99 Asus ROG Crosshair VIII Formula motherboard.
And from numerous reviews, such as this one, B550 motherboards tend to have capable VRMs. To be honest, I can't verify the temps of my VRM as I don't know how to monitor it.
The case is supposed to be a good airflow case. No one advised me not to get it, and I like its compact size.
The fan issues that I had were just a configuration issue and have been resolved.
My wildly fluctuating CPU temperatures turns out to be a feature of AMD chips and was a non-issue.
edgetrading
Bronze Level Poster
My boosts are fine. It gets 4.4-4.5GHz on single threads, and 3.7-3.8GHz across all cores.
Does anyone have comparable numbers for systems with better motherboards and coolers? It would be very useful for potential customers to know how different components can affect performance.
Single Core Speeds with Prime95:
All core speeds with Prime95:
Does anyone have comparable numbers for systems with better motherboards and coolers? It would be very useful for potential customers to know how different components can affect performance.
Single Core Speeds with Prime95:
All core speeds with Prime95:
SpyderTracks
We love you Ukraine
The motherboard doesn’t affect performance, so long as it’s the right chipset for the processor. Again with the cooler, so long as you’ve got the right cooler for the processor then boosts will reach what they’re supposed to.
If the cooler is inadequate, there’s no “expected temps” or comparable benchmarks, it all depends on the temps of the cpu within your system which will vary on everyone’s system.
What @Scott was referencing is that the temps you quoted aren’t at all possible on 100% load on air with fans at 50%, it’s just not possible. Unless your in a sub zero room or something that is.
I'm afraid that's entirely inaccurate from those Prime95 outputs.
You can see from your single core Prime95 test that the boost is severely hampered almost immediately, the boost will only apply if it has thermal headroom which is not possible on your setup.
Your multi core is also severely hampered.
Really you need something like hwmonitor to monitor your temps during a test.
But basically, it looks like your CPU is thermal throttling as it’s not reaching anywhere near stated speeds even on all cores (which will be far less than single core boost). This is completely expected, the platform just isn’t at all adequate for the CPU plus an extra negative that it’s on air cooling, it’s just not going to reach anywhere near what it’s supposed to.
The varying temperatures is down to the fact the B550 boards VRM's aren't able to handle the massive voltage regulation for that CPU, there's just absolutely no way. You HAVE to have X570 on that CPU to get even close to normal clocks. Plus you'll need an absolute minimum of the H115i Platinum.
I'm afraid that build just isn't going to work well whilst it lasts, and will very likely die quite quickly if the shortcomings aren't addressed. But it will cost about £400 to get it road worthy.
Last edited:
As above...... that's not running anywhere near at optimum.
IIRC the 3950X is 4.7Ghz single core and 4Ghz across all 16cores 32 threads.
If it's doing what you want OK though then there's no major issue. The advice is always put forward to get the most out of the system and have it running optimum in all situations.
As you are finding, everything you have selected will be compatible with each other but the chip is definitely hampered with the surrounding choices.
Not sure if it makes any odds but the link you posted previously shows the ROG Strix B550 which, I would hazard a guess, has better VRM capabilities than the bog standard B550 board. Not the best £40 saving I can think of to be honest.
IIRC the 3950X is 4.7Ghz single core and 4Ghz across all 16cores 32 threads.
If it's doing what you want OK though then there's no major issue. The advice is always put forward to get the most out of the system and have it running optimum in all situations.
As you are finding, everything you have selected will be compatible with each other but the chip is definitely hampered with the surrounding choices.
Not sure if it makes any odds but the link you posted previously shows the ROG Strix B550 which, I would hazard a guess, has better VRM capabilities than the bog standard B550 board. Not the best £40 saving I can think of to be honest.
edgetrading
Bronze Level Poster
What sort of performance gain could I expect if I replaced the motherboard with an X570 and got an AIO? The budget isn't really an issue as it's a work machine. But I am really struggling to find any genuine evidence that an X570 and AIO will improve things.
As an example Kitguru benchmarked the 3950x chip to take 123 seconds to run the BMW blender test using a Gigabyte X570 Auros Master and Corsair H100X AIO here, using Blender 2.79b. And reported a 79C temperature for Blender Classroom here.
Anadtech reports a time of 128 seconds using an X570 Taichi and Kraken X62 AIO here.
They don't specify what case they use. Perhaps they use open benches without fans/cases.
I downloaded the software, and it takes 109 seconds on my machine, with CPU temps up to 61C. Ambient temperature in my home is around 25C. I'm not sure why it's a little faster than the published figures, but I checked that I'm using the correct version of Blender. I'm using Linux, so may that causes a small difference.
I also ran the Classroom test, and my max temperature was 62C.
Anyway, given the above, my system seems to be performing in line with other systems.
Do you have any evidence that systems similar to mine fail or are thermal throttled? Have you seen similar systems failing? Is there a chance that you could have made a mistake?
As an example Kitguru benchmarked the 3950x chip to take 123 seconds to run the BMW blender test using a Gigabyte X570 Auros Master and Corsair H100X AIO here, using Blender 2.79b. And reported a 79C temperature for Blender Classroom here.
Anadtech reports a time of 128 seconds using an X570 Taichi and Kraken X62 AIO here.
They don't specify what case they use. Perhaps they use open benches without fans/cases.
I downloaded the software, and it takes 109 seconds on my machine, with CPU temps up to 61C. Ambient temperature in my home is around 25C. I'm not sure why it's a little faster than the published figures, but I checked that I'm using the correct version of Blender. I'm using Linux, so may that causes a small difference.
I also ran the Classroom test, and my max temperature was 62C.
Anyway, given the above, my system seems to be performing in line with other systems.
Do you have any evidence that systems similar to mine fail or are thermal throttled? Have you seen similar systems failing? Is there a chance that you could have made a mistake?
Sounds like your system is running perfectly. I wouldn't touch it.
You were the one that started the thread worried about thermals. If the thermals have now sorted themselves (and 62C would definitely suggest you have no problems) then I don't see a reason for changing anything.
When things are working OK just leave them as they are IMO. No point in fixing something until you break it.
JohnWalton981
Silver Level Poster
I have a 3900XT, so a step down from you. My other components are; an X570 and H115i AIO, with a 2080S, in a Corsair 680x case.
I've just tested on Prime95 and my max temp was 78°C, with an average of 76°C and an average clock across all cores of 4215Mhz, and showed 87% average usage.
I also ran the BMW and Classroom render on Blender, 56 and 163 seconds respectively - max temp 67°C.
I don't see any issues with mine and I'm not sure how or why I'd want my temps and/or performance to be any different.
I've just tested on Prime95 and my max temp was 78°C, with an average of 76°C and an average clock across all cores of 4215Mhz, and showed 87% average usage.
I also ran the BMW and Classroom render on Blender, 56 and 163 seconds respectively - max temp 67°C.
I don't see any issues with mine and I'm not sure how or why I'd want my temps and/or performance to be any different.