Yesterday, I finished my new build featuring the Ryzen 7 3700x, it contains:
Ryzen 7 3700X
MSI X470 Gaming Pro Carbon
Be Quiet! Dark Rock 4
16 GB DDR4 Tridenz Z RGB 3200 Mhz CL16
Fresh Copy of Windows 10
Fractal Design Meshify C with 2 intakes 1 exhaust
Now, after some testing around and monitoring the statistics I found the following:
When enabling PBO in the Bios, the voltage likes to spike up to 1.5V, so I set an offset of -.1v, now it does not exceed 1.4v.
Furthermore, I have enabled auto-oc, but I have not adjusted any settings, just set everything to "enabled".
When doing a stresstest, Aida64 for example, all cores go up to around 4.1, 4.2 Ghz with just under 1.4 Volts.
My main problem now are the temperatures.
When stresstesting, it goes well above 70 C, sometimes it spikes even in the 80s.
From all the reviews that I have read, noone had such high temps.
I have not adjusted any settings in Ryzen Master because I have not had the time to learn the software.
If there are any tweaks that I could do to improve things, I will do my best to try them!
Reseating the cooler will be the first thing that I will do, but I doubt that this will change anything at all...
Thank you for your help!
I can't really said where are the options on your MSI mobo, in my Gigabyte, I found one under AMD CBS menu, and the other two PBO options were found under AMD Overclocking menu.
Both AMD options were under the PERIPHERALS upper menu. So yeah they are not soo easy to find. (Maybe someone in MSI forum can give you a hint, or someone who owns the same motherboard you do since all BIOS are diferent from one mobo to another).
You are absolutely right, mate. I have miss my old quite system also, i5-3570k with RX480 was perfect as performance/silence. But for now, welcome to amd with performance and loudness...
Actually, with 3700X,
Idle and stresstest under Cinebench, with an old Thermalright true spirit 120M (200 watts), and thermal paste kryonaut.
Disabled PBO : idle 37/40 and stress 62/63.
Enabled PBO : idle 50 and singlecore stress 80.
Really day and night, under pbo there was always one core at 4.2ghz for no reason even on idle ...
Cinebench R20, in disabled PBO :
All CPU: 4595
Single CPU: 492
I disabled PBO but its the same. In idle aida shows my cpu is on 4275.0 MHz. I disabled PBO, and download new chipset drivers with ryzen ballanced plan but nothing work. My iddle tems are still from 40-50C. My motherboard is MSI B450 Tomahawk.
Have you tried setting the Windows Power Saving plan, if that doesn't lower your clock frecuency, then maybe, just maybe your motherboard is applying some kind of pre-factory overlock (most motherboards offer this pre-overclocking settings, I remember my old asus have it, and my new Gigabyte has it too, and this has nothing to do with PB, PBO or XFR, is a feature that motherboards offers). Best way to check this is loooking the normal view in BIOS (the non-avanced one). There should be a performance chart with some options, normal or standard is what it should be selected.
Just check that all the options are set like the pic: https://images.anandtech.com/doci/13371/MSI%20B450%20Tomahawk%20BIOS%20%2834%29.jpg (specially the game boost in the left top corner).
Core Performance is AMD PB which is the normal speed boost for AMD CPU, that shouldnt give you many problems with temps and voltages. If you disable it, then you will see that your Ryzen only reach 3600MHz as top speed.
You can enable AMD COOL N QUIET on your BIOs.... Do not set it AUTO, set it to ENABLED...With this feature (only this, even with PBO ON), im getting .80 idle voltages, and 40c temp on idle...
Its wallpaper engine that's preventing your CPU voltage from lowering at idle, I know this cause I also use wallpaper engine and noticed my 3900x would not drop voltages if its running in the background, even if you enable Cool'n'Quiet results would be the same (also be aware that even though Cool'n'Quiet will drop your idle voltage it will for some reason increase your voltages under load).
If your still seeing constant CPU usage on your processor in ryzen master at idle (also encountered something similar), I would suggest doing a windows refresh using windows 10s 'Reset this PC' feature and opt to not backup existing files (only way to make sure it's a fresh install) and only install chipset driver and drivers you need, as I've also had problems recently with dragon center & StoreMI also preventing CPU voltages from dropping at idle.
Would also advice you first try to reset to defaults your bios and only enable xmp for your ram, also use AHCI and disable CSM and test if you see your idle voltages and temps go down at idle in windows with Ryzen Balanced power plan before you try enabling other stuff in the bios or adding offset voltages. And do note that it normally takes about a minute at first boot into windows before you start seeing voltages dropping at idle with nothing running, you would usually only see your voltages start rising when you have anything open in the foreground, watching videos in chrome in particular does that.
As for the temps, that looks about normal seeing that even with my 240 AIO with Push/Pull fan setup I'm still reaching late 70s under load (room temp of 29c) with the radiator mounted as intake to best cool it with Idle temps going 31-35c+ with voltages as low as 1.06v then jumps to 40-50c+ with voltages increasing upto 1.46v whenever i open any app.
You are actually right with WE, when closing it it goes down to just under 1v.
Thank you for sharing your thoughts, I won't do a fresh install because it is a brand new system and without wallpaper engine it seems like it goes down just fine.
that still does not explain my boost issues, but I guess that is up to AMD / MSI to fix.
Yea depending on what wallpaper you are using it could be either using more CPU or GPU resources. Pausing tends to negate all that though. Its good to set application rules to have it pause or stop for certain applications.
Its been a few weeks now since I instaled AGESA 1003ABB + new AMD chipset drivers on my Ryzen 5 3600, as long as I have the PBO options OFF in BIOS everything is quite good, yeah stock fan gets louder but temp and vcore are way better.
A week ago I went and got my self a Deep Cool Gammaxx GTE air cooler (don't care about LED, I just wanted a screwed mounting system). Temps went down and noise too, and overall my CPU now boost higher (with only PB enable) probably cause of the better temps headroom.
On the other hand boosting higher means temps, on some scenarios, didn't go as low as I wish. For example on Blender complete test (took about hour and 13 mins to finish, no crushes), temps reading on hwinfo were 76.3° C MAX (a lot less than on stock cooler anyways, was about 86°C I believe for the short test).CPU now boost all cores to 4050 MHz most of the time while rendering, thats about 100~125MHz higher than with stock cooler.
In any case with this new cooler I never seen temps over those 76.3°C which is great. With stock one temps were always above 86°C (Cinebench R20) and sometimes above 93°C (prime95).
On cinebench with scythe mugen 5 rev.b, I have 80 degrees C, but boost for test is 4250 MHz.
Hi, I don't have CR20 numbers here at work, but I did a consecutive 30 mins test (you can set this on preferences), but I do remember it was about 73°C boosting about 4080Mhz all core.
OK, I got back home, so with the new cooler:
- CR20, 30 mins of consecutive runs, MAX reading temps I got was: 72.8°C
- Prime 95 small packages, 37 mins (then I got bored) 79.8°C, used to be 93,3°C after 10 mins running with stock cooler.
- Shadow of The Tomb Raider (got an RTX 2070, 1920x1080 and graphic details are on highest), 70°C after 18 mins (I believe the poor result here is due to the GPU dropping a lot of heat inside the case). In any case this is better than 78.4° C with stock cooler.
I'm using Thermalright Macho Rev B cooler, max temperature in aida64 stress test is 85 C, average is 75 (ambient is 28 C), CPU power is around 90 W, voltage up to 1.38 V, frequency >=4.1 GHz (gigabyte aorus elite, bios F5b).
After a month with 3700X I am pretty sure that something is wrong with the 3000 series.
OK, frequencies are not as advertised, it barely reaches 3.4 GHz only when it is cold and under 60°C, but how can you keep this thing under 60°?
With PBO Enabled and everything on auto, these are my results.
Ryzen 3700X on an ASUS GAMING PRO, latest BIOS, all auto, PBO Enabled
32 GB of RAM@3200MHz
Cooler Dark Rock Pro 4
Stressing it with prime95
Under 55°C it sits at 4350MHz all cores and occasionally some cores jump to 4.4
@65°C it draws 80Watts and sits at 4150MHz all cores
@75°C it draws 90Watts and sits at 4100MHz all cores
@85°C it draws 100-110Watts and sits at 4050MHz all cores
@90°C it draws 150-160Watts and sits at 3980MHz all cores
@94°C+ almost 170watts, throttling rules and it sits at 3900MHz all cores
So, It reaches 4400MHz on some of the cores only when it is under 60°C, but with the slightest utilization temps jump to 60s, the question is, what can keep this thing cooler than 60-75°, to my belief nothing. So this CPU is almost totally depended on temperature and works @4100-4250GHz, much much lower than expectations!
Of course, you can always manually overclock it but to reach 4.4 and to be stable you need to set a high vcore, something that looks forbidden now due to very high temps produced.
After applying metal liquid between the cooler and IHS things improved by almost 20°C in the range of 65-80°, that was when the CPU draws from 80 to 120watts, (80° became 60 at 120watts!) but improved only by 3°C when the wattage jumps over 150watts, (90° instead of 93-94°), also frequencies improved accordingly.
With the stock cooler things and behavior are pretty much the same, so it is not a matter of cooling system.
There is a serious thermal problem with these processors, they cannot effectively spread out the produced heat to the cooler.
I've noticed that when the power is around 100watts and temps at 60-64°C(with liquid metal instead thermal paste, otherwise temps are much higher), the cooler measures 40°C
When the wattage is at 150watts and CPU temps rise even more at 89-90°C, the temperature in the cooler is not rising as well as it should! it starts to drop slowly to 37-35-32-31...what the f@ck is happening here? It is impossible to have lower temperatures in the cooler when in CPU temps get higher!
In my opinion, either there is a very poor heat dissipation or...every existed BIOS reads false high temperatures due to a CPU internal manufacturing defect! and therefore starts an unnecessary throttling!
Look around, the internet is beaming of people complaining about hot processors but cold coolers!
I don't want YET to dellid the CPU to check the thermal conductivity between the dies and IHS.
Surely there is a problem, maybe after some months, AMD will be forced to redraw all sold CPUs and replace them with new ones unless they find a way to fix the problem with a BIOS update.
Despite this, though benchmarks are really good and reveal a powerful processor, although not the processor we all expected.
How do you measure power consumption? Did you raise power limit? In my case (for now) I have < 70 C in prime95 (Macho Rev B + MX4, ambient temperature is 27 C), frequency is 4.15 GHz, power is around 90.2 W. Load is stable at all cores. AIDA64 stress test rises single core up to 4.4 GHz and temperature up to 73 C now (load unstable, temperature jumps from 65 to 73 C).
With HPUID HWMonitor
Of course, if you lower the voltage or manually overclock power consumption can be much lower. (bear in mind that I stress with prime95, this is absolutely different than AIDA stress)
Also, I can get stable manual overclocking @4.2 on all cores with 1.33v vcore, then temps stays low and power consumption stays under 90watts, but I want to use the PBO overclocking and leave the chip to do whatever is best for it, this way it can jump to 4.4 in some cores when needed, also I don't mind about high idle vcore, this is absolutely normal for this CPU and not a reason for any concern.
There are many things you can do to lower the temps and power consumption, but this, as I found out, is always against performance.
But you got me wrong I am afraid, I don't care about the high temps, if CPU works at 30s 60s or even 100s don't bother me at all, silicon is not carbon, 100 C for silicon is like spring for humans.
What is my concern is that heat doesn't spread out of the CPU to the cooler, every cooler on this CPU stays almost cold, this is the only problem, I want to feel the cooler warm.
Also, AIDA stress measures the CPU temperature and not the CPU diode temp which is at least 10°C higher!
I am talking always about the diode temp.
See the difference?
AFAIK this difference depends on motherboard/BIOS. I and my colleague have almost the same data about power consumption(Gigabyte X570 Aorus Elite F10a) - limit is 90 W according to different software. It is strange to have almost 170W without changing power limit. In my case AIDA heats CPU more than prime95 as I wrote before.
So, you have CPU and CPU Diode temps almost identical, that's really strange...
In my case, the difference is almost 12 degrees, and AIDA stress is very soft compared to aggressive prime95.
In my case both of prime95 and aida bumps power limit (90W, on 3800X and above it is 140W), so total heat power is the same. As I see CPU temperature is max of die sensors temparature (like hotspot on GPU), so bad load balance causes voltage & frequency increasing on single core and rises local temperature. 3700X die is very small, so heat density is really big. In my case cooler is not hot, so the bottleneck is thermal conductivity of very small die. It is physics, nothing personal.
I got you now, you have disabled PBO and power consumption is on 90watts limit.
And was wondering about what was this 90watts limit you were talking about...
I did the same and results are identical to yours and temps even better, and indeed now AIDA stress CPU harder than prime95, but this is because with prime95 when PBO is disabled, frequencies stay low around 3600.
So, you are working your CPU in shallow waters, because like this performance is worse, an example: CINEBENCH20 gives me 5000 with PBO enabled and around 4650 with PBO disabled.
Your point about heat big density due to very small die makes sense, this explains why cooler remains cool all the time, of course, its physics but this isn't something good, in fact, is something very very bad.
Honestly, if it really reaches 90C, Where all this heat goes? it goes to cooler but very very slowly, it is a problem!
Tomorrow I will try to measure the temperature at the side of the IHS with a laser thermometer when the sensor shows 90, I bet it won't be more than 30-40
I can see underneath of cooler plate because my box's top is removable.
putting my finger down there I can feel no heat at all, it is like dead!
Still, bet that sensors give false readings.
No, PBO is enabled. Power limit for 65 W CPU is almost 90 W and 140 W for 105 W TDP CPU (Power Consumption & Overclocking - The AMD 3rd Gen Ryzen Deep Dive Review: 3700X and 3900X Raising T... for example), frequency up to 4.4 GHz as expected. Prime95 massively uses AVX, so frequencies is lower (AVX is power hungry), so frequency in my case around 3.85 GHz, in aida it is 4.1 GHz. Power in both cases around 90W, the only difference is power balance between cores. What motherboard are you using now?
Asus TUF Gaming X570-Plus
Yes, prime95 frequencies are the same as yours, with AIDA is about 4.1
There is a difference in power consumption where it seems that there is no limit on my motherboard, maybe it is a setting somewhere that I missed.
Are heatpipes of your cooler a little bit warm (not hot) when you stressing cpu? In my case thet are, but almost nothing else. As I understand die to small chip size there is a big termal gradient inside, so the cooler contact with relatevely cool surface. The second thing is to warm even my cooler the power should be around 200 W, but you cooler is even bigger - so may be it is normal thing. Just curious: what is you motherboard chipset temperature? Does chipset cooler stop on idle?
gastheo, the only truly valid readings come from Ryzen Master (RM). Please post a screenshot of RM. AIDA64 also gives correct CPU temperatures if you use tCtl. HWMonitor and HWinfo are among the worse, sending commands to the processor to effect the readings. For some unknown reason AMD moved Robert Halloc's reporting to Reddit. He is AMD's front for performance. He posted about HWinfo and the release of the AGESA update to correct the maximum boost concerns on Gen 3 Ryzens. Please go to Reddit and search for his explanations. AMD and MS have made changes to the W10 1903 Scheduler to help deal with these concerns. AMD released the AGESA to MB vendors, so bug your MB vendor to make sure you are running the fix and make sure you are running up to date 1903. Enjoy, John.
EDIT: Liquid Metal is the best TIM but is also likely to seriously corrode your processor. The Internet has pictures of badly corroded Ryzens. Tom's HW posted a extensive study of TIMs, from liquid metal to toothpaste. Outside the first and the last, there was little difference. If you really got a 20C improvement after changing to liquid metal, I suspect your previous TIM/Cooler were seriously misapplied.
No point for a screenshot, all software including Ryzen Master shows the exact same temps.
If the only way to run this chips cool is by limiting consumption at the expense of worse performance, then this zen2 is not meeting the expectations influenced by the exaggerated reviews before zen 2 platform release, remember they were talking about overclocks near 5GHz and other unreal things.
Liquid metal doesn't corrode any processor, after many years of use I never saw the slightest corrosion in copper or nickel, also it is very easy to remove it without leaving any trace on IHS, everything they say about removed letters and scratches is a myth driven by competition or scared novices who never attempted to give it a try, I never sanded the surfaces prior to applying because there is no need for it.
Also, I didn't have a 20C different, settings like boost and consumption were slightly different and I was exaggerated a bit as most people in forums do, my apologies for that, there was an obvious difference though!
I found a screenshot from 2 days ago when liquid metal was used, and I took another shot today using the paste provided by Dark Rock Pro 4 which I applied last night, see both and compare, I must admit that there is no difference between liquid metal and Dark Rock Pro 4 paste, which is proven to be much better than MX-4.
AIDA Stability test, with PBO Enabled and liquid metal. (Don't mint about the max, there were higher because I had stressed with prime95 before AIDA)
AIDA Stability test, with PBO Enabled and Dark Rock Pro 4 thermal paste.
Conclusions are easy, liquid metal doesn't worth the hassle, although I enjoyed the time spent.
Another thing is, this unnamed thermal paste provided by BeQuiet is really decent, I will ask them where can be found.
Today I tested Arctic MX-2, no difference at all, absolutely the same, finally it's not worth the time with these thermal pastes, there all about the same.
With MX-2 AIDA Stress(even better than LM!)
you´ve got to remember that a stress test is nothing like everyday use.If you´re not close to 90 c you´re golden. Everyday use doesn´t strain your cpu like a stress test,it´s kinda in the word....
There is a real problem with those AMD Ryzen 7 CPUs - my 3700x with a Wraith Prism stock cooler and all default BIOS options have 90+ degrees (AIDA64 stress test). My CPU crystals are really unstable at this temperature and while playing games I have freezes and reboots from time to time. I've tried to change my Windows power settings, set 99% maximum CPU, but no effect in my case.
So I've changed some BIOS settings to take control over this situation:
1. Disabled CPU Core Boosting
2. Set CPU Frequency manually to 4.0 GHz
3. Changed my CPU Core Voltage to 1.188V (I have strange power management on my MB and voltage is changing from time to time from 1.176V to 1.188V while CPU and GPU are 100% stressed with AIDA64).
4. Wraith Prism cooler has a little switch on it, so I turned it on to boost cooler power (now it sounds like an airplane taking off)
5. Set my Case coolers to react to the CPU temperature and to have their maximum after 60 degrees (3 x Corsair 120 mm coolers 1500 rpm). It additionally helps to move hot air from the CPU and the GPU (-4 degrees difference for both CPU and GPU).
As a result, I have a stable CPU with a maximum temperature of 75 degrees, but with some performance loss. I tested it with AIDA64 1 hour stress test and while playing Mechwarrior 5: Mercenaries with 4K maximum settings and have no visual differences in performance, so I settled on this result.
P.S. I think that frequency and core voltage can be slightly higher, and this CPU can be stable and not so hot on 4.2GHz. I will try this later.
I have tried to raise frequency and voltage and currently have a stable system with 4.3GHz and 1.296V. The temperature is 93 degrees maximum, 84 degrees average during 1-hour AIDA64 stress test.
I found this thread because my Ryzen 3700x was going above 90C in HWMonitor. Following web advice, I got Ryzen Master. Though not testing it on a game or doing a stress test, I was seeing 60C idle temperatures.
like roboticron, i found changing the min temp for full fan power helped. I did this for case fans as well as the CPU cooler. It alleviated the problem - 45C idle and never above 80C for longer than a quick peak. I will do some sustained gaming to see how it performed. I did not power power or clock speed. I just made the fans proactive (65 for high speed, 40 for mid, 30 for low, if i remember correctly). It was easy to do in my Asus BIOS; others may differ.