Just wanted to test this 3900X with Stock Cooling on an Asus B550-F. I did a manual undervolt of 0.1250 as it was hovering around 1.424-1.440 stock. 1st set of pics show the start of blend testing 24 sets of Prime95. 2nd pic shows Prime95 moving to a more heating test. That is some difference in Vcore and VID. Test still running no errors. I haven't really messed with Overclocking lately since I got my used AMD 1600 back early this year. Still learning the ropes on the 3000 and 5000 Ryzen series. I read around that 1.325v was the sweat spot for this chip and that's where I started with the voltage offset.
-
Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!
Some fun with Ryzen 3900X and Stock Cooler.
- Thread starter dja2k
- Start date
- Joined
- May 17, 2005
- Location
- Hartwell, Georgia
I can't see your attachments. It's like a thumbnail was attached without a link to the full size images.
- Thread Starter
- #3
Sorry about that! Best I can do with the limitations of this forum's attachment process. I wanna see if I can set at fixed multiplier of 40 @ 4.0ghz and a fixed voltage also. This 3900X built is not mine, but a friend being short on cash for now to add a better Heatsink. I get my 5600X some time next week, that would should run cooler with my Fuma 2 Heatsink.
- Thread Starter
- #4
So I am gonna list my results cause I don't know how to add pics with links instead of embedded. I put everything back to defaults in the bios. Did a whole bunch of testing and reached 95 C max temps with stock Wrath Prism. This is on a Asus Strixx B550 with AMD 3900X (not mine which is listed in my signature).
CINEBENCH
Idle Frequency 3774MHz , VID 1.4812v (CPU TEMP), Core Voltage 1.68v (CPU-Z)
Testing Frequency 3998MHz , VID 1.2312v (CPU TEMP), Core Voltage 1.264v (CPU-Z)
My CPU Rank Score 18097 NO ERRORS, Cinebench Score from others +/- 18682
OCCT CPU
Idle Frequency 3774MHz , VID 1.4688v (CPU TEMP), Core Voltage 1.152v (CPU-Z)
Testing Frequency 4098MHz , VID 1.3062v (CPU TEMP), Core Voltage 1.328v (CPU-Z)
Results: NO ERRORS 1 HR
BLENDER
Idle Frequency 3774MHz , VID 1.4688v (CPU TEMP), Core Voltage 1.52v (CPU-Z)
Testing Frequency 3998MHz , VID 1.2312v (CPU TEMP), Core Voltage 1.280v (CPU-Z)
Results: bmw27 2m5s, classroom 5m33s, fishy_cat 2m45s,
pavilion_barcelona 5m45s, victor 10m11s
REALBENCH
Idle Frequency 3774MHz , VID 1.4562v (CPU TEMP), Core Voltage 1.68v (CPU-Z)
Testing Frequency 3998MHz , VID 1.2687v (CPU TEMP), Core Voltage 1.248v (CPU-Z)
Results: Results Hash Match! on all Tests.
PRIME95 BLEND
Idle Frequency 3774MHz , VID 1.4562v (CPU TEMP), Core Voltage 1.184v (CPU-Z)
Testing Frequency 3748MHz , VID 1.2687v (CPU TEMP), Core Voltage 1.104v (CPU-Z)
Results: NO ERRORS 1 HR
I see that different tests use different voltages to increase\decrease CPU frequencies. I did try Offset -0.1250 like I stated but didn't run all these tests. Can some one chime in and help if I should try a higher Offset or use a fixed Frequency. Any help is really appreciated and thanks.
CINEBENCH
Idle Frequency 3774MHz , VID 1.4812v (CPU TEMP), Core Voltage 1.68v (CPU-Z)
Testing Frequency 3998MHz , VID 1.2312v (CPU TEMP), Core Voltage 1.264v (CPU-Z)
My CPU Rank Score 18097 NO ERRORS, Cinebench Score from others +/- 18682
OCCT CPU
Idle Frequency 3774MHz , VID 1.4688v (CPU TEMP), Core Voltage 1.152v (CPU-Z)
Testing Frequency 4098MHz , VID 1.3062v (CPU TEMP), Core Voltage 1.328v (CPU-Z)
Results: NO ERRORS 1 HR
BLENDER
Idle Frequency 3774MHz , VID 1.4688v (CPU TEMP), Core Voltage 1.52v (CPU-Z)
Testing Frequency 3998MHz , VID 1.2312v (CPU TEMP), Core Voltage 1.280v (CPU-Z)
Results: bmw27 2m5s, classroom 5m33s, fishy_cat 2m45s,
pavilion_barcelona 5m45s, victor 10m11s
REALBENCH
Idle Frequency 3774MHz , VID 1.4562v (CPU TEMP), Core Voltage 1.68v (CPU-Z)
Testing Frequency 3998MHz , VID 1.2687v (CPU TEMP), Core Voltage 1.248v (CPU-Z)
Results: Results Hash Match! on all Tests.
PRIME95 BLEND
Idle Frequency 3774MHz , VID 1.4562v (CPU TEMP), Core Voltage 1.184v (CPU-Z)
Testing Frequency 3748MHz , VID 1.2687v (CPU TEMP), Core Voltage 1.104v (CPU-Z)
Results: NO ERRORS 1 HR
I see that different tests use different voltages to increase\decrease CPU frequencies. I did try Offset -0.1250 like I stated but didn't run all these tests. Can some one chime in and help if I should try a higher Offset or use a fixed Frequency. Any help is really appreciated and thanks.
- Joined
- Dec 15, 2008
- Location
- Your head, rent free!
The forum should auto resize, bud. Attach them to the forum... we'd love to see them.
EDIT: Worth noting, most links to image hosting sites go dead after some time... the forum wont.
EDIT2: You can resize them in Paint...easy breezy.
EDIT: Worth noting, most links to image hosting sites go dead after some time... the forum wont.
EDIT2: You can resize them in Paint...easy breezy.
- Joined
- May 17, 2005
- Location
- Hartwell, Georgia
Provided they meet the max upload MB requirement, large images will auto-resize when embedded.
- Thread Starter
- #11
Update
Sorry my bad, I didn't have voltage at fixed 1.3500. Had with -offset 0.1000 and testing for Cinebench, Blender, and OCCT have dropped average 5'C temp. I did notice that CPU doesn't throttle down to 3773MHz and stays stable at 4048MHz and even goes higher to 4223MHz in testing. Prime95 at default voltage was the only test that would hit 95'C.
Setting fixed voltage to 1.3500 throttles down to 3200MHz running Cinebench. I stopped texting there. Seems the offset voltage was better, honestly I don't understand why.
Sorry my bad, I didn't have voltage at fixed 1.3500. Had with -offset 0.1000 and testing for Cinebench, Blender, and OCCT have dropped average 5'C temp. I did notice that CPU doesn't throttle down to 3773MHz and stays stable at 4048MHz and even goes higher to 4223MHz in testing. Prime95 at default voltage was the only test that would hit 95'C.
Setting fixed voltage to 1.3500 throttles down to 3200MHz running Cinebench. I stopped texting there. Seems the offset voltage was better, honestly I don't understand why.
- Thread Starter
- #12
dja2k said:Today 03:02 PM
Yeah fixed voltage 1.350 cripples the cpu and runs at 3200MHz. I played more with offset and have it -0.1062v , MB set that when I entered 0.1050v. Cinebench multi core runs at a steady 4000MHz +/- a few MHz, voltage on Ryzen Master shows 1.36v - 1.37v fluctuating, and temp steady at 81°C max. On single core test, it's all over the place, frequency fluctuating from 2.6GHz - 4.3GHz with high voltage at 1.449v, and temp 62°C. Power Plan Ryzen balance vs Ryzen high performance have no effect
Yeah fixed voltage 1.350 cripples the cpu and runs at 3200MHz. I played more with offset and have it -0.1062v , MB set that when I entered 0.1050v. Cinebench multi core runs at a steady 4000MHz +/- a few MHz, voltage on Ryzen Master shows 1.36v - 1.37v fluctuating, and temp steady at 81°C max. On single core test, it's all over the place, frequency fluctuating from 2.6GHz - 4.3GHz with high voltage at 1.449v, and temp 62°C. Power Plan Ryzen balance vs Ryzen high performance have no effect
- Joined
- Jun 21, 2002
Yeah I was going to say fixed 1.35v is way too much on air. These are a bit of a different beast, we're getting used to them, but the super high idle voltages, tbh I wouldn't think much of, might be spurious anyway. I had some good luck using an offset on the 2600x, I haven't had much luck using one on the 3800x. I'm pretty sure the newer CPUs "see through" it and pretty much pull as much voltage as they can. Of course if you can get things to run a little cooler then you get a bit more headroom.
Here's a decent article explaining how these things manage their frequency: https://www.gamersnexus.net/guides/3491-explaining-precision-boost-overdrive-benchmarks-auto-oc
It can be interesting to play with, but aside from sub-ambient setups, they are pretty much fastest on auto settings. Sometimes you can get 50-100mHz better all core but then you don't get the high single core clocks for lightly threaded workloads.
Here's a decent article explaining how these things manage their frequency: https://www.gamersnexus.net/guides/3491-explaining-precision-boost-overdrive-benchmarks-auto-oc
It can be interesting to play with, but aside from sub-ambient setups, they are pretty much fastest on auto settings. Sometimes you can get 50-100mHz better all core but then you don't get the high single core clocks for lightly threaded workloads.
- Thread Starter
- #14
This same friend has a similar setup, a 3700X on Asus B450-F and max temp were 65- 70°C with top 1.45v single core on stock Wrath Prism Cooler also. So with seeing that, I just left this 3900X run its single core up to 1.45v and mid 85°C temp which hit max 4300mhz or so. All this with an offset of 0.1062v in bios and didn't see 90°C+ like I did with Stock voltages.. Later I will probably upgrade both his systems to a better cooler and play more with fixed voltage and frequencies. Basically Cinebench scores where on par with default scores for that specific processor.
- Joined
- Jun 21, 2002
I stand corrected. I still wouldn't want to run Prime95 Small FFT with those volts, but looking back the OP was running blend, not small.
Prime is overrated anyway. I'd rather test in real world usage. The benchmark was an after thought.
On my 3600XT 1.2675 got me 4400MHz, and that was good for all loads, and by all loads I mean Linpack Xtreme.. 4500MHz came t 1.3375, and yeah I could run Linpack, temps were right at the limit, not something I would run daily. On my 5900X I am ok with 1,25v all loads for 4500MHz, 4600MHz comes at 1.35v, and like with my 3600XT, not ok for all loads.. but almost all loads It will run it, but the CPU is right at its limit. With your 3900X.. I dunno. I would have to play with it, but you should find your FIT voltage, and call that your all load voltage. I am using a top end heat sink with stronger fans than what it came with and a ****load of case flow vs your stock cooler.. I would probably just run it at stock
I dont do stock though.. I would have bought a Dell if I did..
It will run it, but the CPU is right at its limit. With your 3900X.. I dunno. I would have to play with it, but you should find your FIT voltage, and call that your all load voltage. I am using a top end heat sink with stronger fans than what it came with and a ****load of case flow vs your stock cooler.. I would probably just run it at stock I dont do stock though.. I would have bought a Dell if I did..
Similar threads
