Question Ryzen 7 3700x Underperforming

[hypernovae555]

I've recently built a gaming pc (for the first time) for a friend, and I'm running several benchmarks. My R7 3700X is doing poorly, but not so poorly as to be unusable, and I want to figure out if it's just poor silicon or something else. In Cinebench R20, it's about ~25 points 1T and ~250 points 16T short of what cgdirector reported a normal R7 3700x bench to be. Userbenchmark puts it at 12th percentile of all 3700x's.

During these tests it peaks at about 85C (set limit to 90), with an all-core boost of 4.05 GHz. Base clock is still 3.6.
I've enabled PBO, I used Ryzen Master to wipe all settings MSI Dragon Center or anything else might have set (then deleted it) and changed all settings either in Master or in the BIOS. I set core voltage to 1.35V--which it holds, and I'm using the Ryzen Balanced power plan.

Specs here:
CoolerMaster MasterAir MA410M
MSI MPG X570 Gaming Pro Carbon Wifi
G.Skill Trident Z Neo 3600MHz 16-19-19-39
Seagate 3TB HDD
Crucial 1TB M.2 drive (OS)
MSI RTX 2070 Super Gaming X (which gets great thermals, no more than 70C)
Win10 1909 (I assume, I installed it two weeks ago)

If it's thermals: the cooler is seated right, and the thermal paste is only about a year old, properly sealed. I have the fan curve in the BIOS set to 100% at ~72C.
Any other info I can provide may take some time, as I don't have immediate access to the computer itself. Thank you.

 

[bmockeg]

Sounds pretty decent for those temps. Want better performance, then you need a good aftermarket CPU cooler.

 

[Phaaze88]

Sounds pretty decent for those temps. Want better performance, then you need a good aftermarket CPU cooler.

OP has a decent cooler...

MSI MPG X570 Gaming Pro Carbon Wifi

That might have something to do with it. Msi's lower end X570s don't have a very good reputation right now...
That, and perhaps tweaking the memory timings a little with Ryzen Dram Calc.

 

[bmockeg]

OP has a decent cooler...


That might have something to do with it. Msi's lower end X570s don't have a very good reputation right now...
That, and perhaps tweaking the memory timings a little with Ryzen Dram Calc.

I see that now. Temps still seem a bit high, probably due to the voltage.

 

[dorsai]

85c is too hot for a 65w cpu...it needs better cooling, the memory timings are not good, and the PC should be on the "high performance" power plan.

Additionally you could set a core voltage offset in the bios that may help with temps and performance...generally -0.75 to -0.100 will work but you'll have to set it and run a stress test to be certain. The MSI MPG X570 Gaming Pro Carbon Wifi is a solid board and more than capable of handling a 3700x.

Also it's recommended to run R20 three times and average the score as the 1st one can be lower than expected.

 

[drea.drechsler]

...
I set core voltage to 1.35V--which it holds
...

If you did that by setting a VCore over-ride to 1.35V that's probably why right there.

If you're going to use an over-ride voltage, especially one that low, try also setting a manual all-core 'overclock' about 4.15-4.2Ghz. Assuming it holds stable, you might find CB 20 MT scores improve. Maybe up in ~4950-5050 range, but ST scores will probably hold similar to just slightly higher.

Otherwise put voltage back to AUTO and let the processor manage itself.

 

[hypernovae555]

Additionally you could set a core voltage offset in the bios that may help with temps and performance...generally -0.75 to -0.100

So a core voltage offset is a separate setting than manual core voltage?

 

[hypernovae555]

If you're going to use an over-ride voltage, especially one that low, try also setting a manual all-core 'overclock' about 4.15-4.2Ghz.

Do you mean by upping the base clock, like with Ryzen Master?

Otherwise put voltage back to AUTO and let the processor manage itself.

The auto core voltage was 1.45V, which made it idle at ~60C and throttle easily. My research found this voltage was unnecessarily high for PBO. Is that wrong? Should I try reseating the cooler with different thermal paste?

 

[drea.drechsler]

Do you mean by upping the base clock, like with Ryzen Master?


The auto core voltage was 1.45V, which made it idle at ~60C and throttle easily. My research found this voltage was unnecessarily high for PBO. Is that wrong? Should I try reseating the cooler with different thermal paste?

Yes...except I'd set a manual CPU clock in BIOS. But you could in RM too, but see below first.

Core voltage should pulse even at idle and it's perfectly normal to see peeks up to 1.45, even to near 1.5 at times. AMD has said it's OK, several times, it's perfectly normal.

As far as temperature... it might be OK to have short peeks at 60 as a bursty load is being processed but it should lower right away and only occasionally longer if a bursty load sustains a little bit longer. If it's holding steady 60 C at idle, your cooling isn't working very well. Also if it is I wouldn't explore manual overclocking unless you're very careful. You've got to fix cooling first.

EDIT: If you're using RyzenMaster to monitor temp and voltage, just know it's not good as it does some sort of averaging that nobody has completely figured out. Instead go get HWInfo64 and look at the SVI2/TFN sensor for core voltage and the Tdie/Tctl sensor for CPU core temperature.

 

[jon96789]

The MSi X570 MPG Gaming Pro Carbon does suck... I had one and it ran extremely hot, the VRMs would hit 90-95 degrees and the CPU would throttle down. I ended up buying an ASUS X570 ROG Crosshair VIII Hero and the board runs ~50 degrees cooler.

But I know that an air-cooled AMD CPU will have a lower CPU speed than a water-cooled one. But the differences in speed is minimal, about 50-100 Hz.

 

[Karadjgne]

For some odd reason, the midrange x470 msi boards did extremely well when it came to the dual phase setup it uses in the VRM's. Then came the x570's and msi is in trouble. The VRM's run way hot, considerably higher than Asrock, which is somewhat higher than asus or gigabyte.

View: https://youtu.be/OMRUhtMs9Ok

View: https://youtu.be/zuyuS04lD4o


If you are having issues, I'd say it points towards a combination of things, cooling, bios settings, motherboard, even PBO settings might be a little low and setting power limits.

 

[dorsai]

The MSi X570 MPG Gaming Pro Carbon does suck... I had one and it ran extremely hot, the VRMs would hit 90-95 degrees and the CPU would throttle down. I ended up buying an ASUS X570 ROG Crosshair VIII Hero and the board runs ~50 degrees cooler.

Buildzoid has a video on youtube discussing the x570 Gaming Plus, which I own, and it has zero problem running a power hungry 2700x above 4ghz on a 24/7 all core workload...with the vrm's staying around a reasonable 70c with good airflow. People tend to forget watercooling removes the fan that would normally sit over the cpu and exhaust air over the vrm...a simple case fan blowing on the vrm remedies the situation.

 

[dorsai]

So a core voltage offset is a separate setting than manual core voltage?

Yes it will limit the voltage based on the offset you select...for example if you're seeing 1.45v and set a -50mv offset your voltage will drop to 1.40v. I recommend running a fast benchmark like R15 while using HWinfo64 before and after the adjustment to monitor the cpu voltage to verify the change.

The option is available in the MSi bios OC tab under Advanced CPU config.

 

[dorsai]

For some odd reason, the midrange x470 msi boards did extremely well when it came to the dual phase setup it uses in the VRM's. Then came the x570's and msi is in trouble. The VRM's run way hot, considerably higher than Asrock, which is somewhat higher than asus or gigabyte.

View: https://youtu.be/OMRUhtMs9Ok

View: https://youtu.be/zuyuS04lD4o


If you are having issues, I'd say it points towards a combination of things, cooling, bios settings, motherboard, even PBO settings might be a little low and setting power limits.

I have no problem with Hardware Unboxed...I watch their content all the time...but the testing method they use is on an open bench with no airflow on the vrm. When installed in a case with proper airflow the boards handle the 8 core chips fine...I know because I own the Gaming Plus and ran a 2700x on it for months without any throttling issues or temp concerns.

 

[drea.drechsler]

Buildzoid has a video on youtube discussing the x570 Gaming Plus, which I own, and it has zero problem running a power hungry 2700x above 4ghz on a 24/7 all core workload...with the vrm's staying around a reasonable 70c with good airflow. People tend to forget watercooling removes the fan that would normally sit over the cpu and exhaust air over the vrm...a simple case fan blowing on the vrm remedies the situation.

My beef with MSI's low end X570 boards is that they are still asking a premium price (compared to previous gen X470) for them. So is the competition, of course, since much of the premium comes from the cost of the chipset plus the high board construction spec level AMD demands.

But the competition also includes near-premium quality VRM's even in their low-end X570 models. It's genuinely annoying MSI didn't... especially since they really nailed it in B450 and X470 and were decent enough even in B350/X370 (considering). The reaction is inevitable, I think, and mostly deserved. It's not just annoying you pay this price and have to put a fan on the VRM if you use an AIO. I don't have to for my B450M Mortar, fercrisake, with a 3700X sitting under a 240mm AIO!

But still, come on. Put it in perspective: they are still perfectly functional, just as you note. So there's no need to put the fear of the electron Gods in someone just cause they have one.

 

[Karadjgne]

@dorsai

So let's see. Open test bench, so thermals are already @ 7°C cooler, since case temps even with good airflow in a 22-23°C ambient still average around 30°C or more.

Open test bench, unrestricted airflow from the cpu cooler, which happens to blow directly across the VRM's without the detriment of gpu exhaust.

Honestly, that's a best case scenario for testing VRM's and the MSI boards are failing. Asus and Gigabyte in exactly the same scenario, with exactly the same cpu and setup are 20°C cooler or better? Please, reality check. VRM cooling beyond passive cooling should be a luxury, not a necessity, shouldn't have to need direct airflow at all, as proven by Asus and Gigabyte and even ASRock.

To me, it's looking pretty obvious as to the cause, Asus, Gigabyte and ASR are using the same mosfets as found on their top line boards, with the same power phase ability, and MSI mid-grade boards are not, they are running the same setup as their low end mobos instead, so any high demand cpu is gut-punching the VRM's, forcing them to work harder, therefore hotter.

You are running a 2700x, not a 3700x, there's a significant amount of differences in voltage regulation, current application and consequent VRM thermals with those 2 cpus. It'd be interesting to see just how close to instability and high 90's your VRM's really are running, consistently.

 

[jon96789]

But why limit yourself to a mediocre board in the beginning? If you are going to buy a new board, buy one that has good reviews and good VRMs. Granted a 2700x does not suck a lot of power, but what happens if you want to upgrade down the road and the board can't handle the heat? It's a waste of money...

 

[Karadjgne]

It's not uncommon. Ppl get used to brands and name recognition. It's very prevalent in psus especially. The midgrade msi B350, B450, X470's are really quite good, have a good variety of bells and whistles compared to other vendors and have had many decent professional and consumer reviews, so it's a natural assumption the X570's will be also. Unfortunately it's not the case this time.

 

[TJ Hooker]

Open test bench, unrestricted airflow from the cpu cooler, which happens to blow directly across the VRM's without the detriment of gpu exhaust.

I can't find proper test setup info for HW unboxed X570 thermal testing video (which seems annoyingly common for youtube reviews, one of the reasons I dislike them), do we know for sure whether they were using an air cooler or liquid cooler?

 

[dorsai]

@dorsai

You are running a 2700x, not a 3700x, there's a significant amount of differences in voltage regulation, current application and consequent VRM thermals with those 2 cpus. It'd be interesting to see just how close to instability and high 90's your VRM's really are running, consistently.

I don't want to hijack the OP's post any further over MSI low end x570 vrms...it would make a great discussion on another thread.

I will only say two things and leave it at that...the 2700x uses more power than the 3700x...and since I've had no issues with throttling on the Gaming Plus I don't believe that's the OP's issue. I certainly could be wrong, but based on his post I will stick with the 1st response I gave...cpu cooling, memory timings, and power plan settings all need to be looked at before moving to other issues that may or may not be present.

 

[Karadjgne]

Ryzens do not do well at all on high performance power plan, they need to be on Balanced or Ryzen Balanced.

16-19-19-39 isn't at all "bad timings" for 3600MHz ram.

The CoolerMaster MasterAir MA410M isn't bad at all, it's actually a decent cooler in either of its versions, equitable to a beQuiet Dark Rock 4 easily, if not better.



Which leaves PBO/bios settings as the most likely suspect, limiting voltages by themselves isn't all that good because when the cpu wants the power, it'll draw amperage instead (P=IV) upto the PBO current limit in order to get the wattage, which will drive the VRM's nuts.

What does the 3700x being a 65w TDP cpu have anything to do with temps? It'll easily top out at 88w-90w limits. The i9 9900k is a 95w TDP cpu that'll easily pull 250w and hit 100°C+ in seconds on a budget cooler. As is, the 3700x can easily hit 85°C gaming, but this can be fixed (somewhat) possibly by microcode updates with latest bios revisions.

PPT: 88w, TDC: 60A, EDC: 90A is what PBO should be maxed at.

 

[MasterMadBones]

AMD admits that PBO doesn't offer a lot of value for Ryzen 3000 because of how those CPUs are binned. Tom's confirms that the performance improvement is usually less than 2%, probably with optimal settings. Add in some sub-optimal settings, such as a lower Vcore limit, and you will lose performance.

Reset everything to default, disable PBO and run those benchmarks again to see if we fall within expected performance margins for a stock 3700X.

 

[hypernovae555]

Well, it took a bit of ram tightening with DRAM calc, a more aggressive fan curve, an auto core voltage, and a manual 4.2GHz base clock, but my benchmark scores are now truly average, with thermals no worse for the wear. I guess PBO just shouldn't be bothered with in my case.

 

[Karadjgne]

Many ppl have come to the same conclusion, that PBO may not be all its cracked up to be. It's a good idea, just needs some tweaks and some more user friendly adjustable settings, because it's curves can be quite aggressive, and while that's good for performance, it's not necessarily so good if just looking for more than base, but still decent.

 

[my_pc_build]

Buildzoid has a video on youtube discussing the x570 Gaming Plus, which I own, and it has zero problem running a power hungry 2700x above 4ghz on a 24/7 all core workload...with the vrm's staying around a reasonable 70c with good airflow. People tend to forget watercooling removes the fan that would normally sit over the cpu and exhaust air over the vrm...a simple case fan blowing on the vrm remedies the situation.

Hi dorsai - could you please elaborate 'a simple case fan blowing on the vrm remedies the situation': do you mean, such a fan sitting inside the case, blowing air directly onto the vrms? Or a case fan in a standard position (e.g. bottom of the case, intake directed towards the VRMs)? Sorry for jumping in on the OP.