[SOLVED] Did someone else try to play with the 3700x core voltage?

[JayGau]

I received my Ryzen 3700x last week and I am pretty impressed. I used CPUZ to compare with my 2700x and the 3700x gets substantially higher scores in both multi-thread and single-thread. But that comes with a price: the temperatures at default settings are comparable to the 2700x with both PBO enable and performance enhancer at level 2 (and at default settings the 2700x has much lower temps). It was surprising for me since the new CPU has a much lower power consumption (but the maximum boost frequency is also higher so maybe it's why it heats more).

So I tried to lower the core voltage to see if I could get similar temps than the stock 2700x. I first put a -0.1 V offset. The max temp under full load went down (around 68-69C instead of 74-75C) but so the performance did especially for the multi-thread: with auto voltage I get above 5500 score on CPUZ and "perform way above expectation" at 96% on UserBenchmark but with the -0.1 V offset, I barely stay above 5000 and I get a "perform below expectation" on UserBenchmark. So the loss in performance doesn't worth the gain in temperature.

I also tried a fixed 1.35 V and I got similar results than the -0.1 offset. Moreover I tried to enable the performance boost option in the BIOS, hoping to force the clock frequency to go higher despite the lower voltage but I saw no effect at all.

And just for the fun of it (I didn't expect any good from that but I wanted to try) I set a -0.2 V offset and OMG... the temp went down by a lot, not even reaching 60 at full load but the score was terrible, going down to 4300 on CPUZ (my 2700x was scoring above 4800 at default settings), far from the original 5500.

With the 2700x, lowering the voltage was actually improving the performance (with a manually set 1.35 V I got slightly better results than auto voltage), most likely because the PBO was more willing to boost the frequency, but it doesn't seem to work like that with the 3700x and lowering the voltage decreases the performance.

I still want to do more tests with fixed voltages (I want to try higher values like 1.36 and 1.37) but did anyone else did similar tests and what did you get?

My specs:
Thermaltake Versa N21 Onyx Edition case
ASUS ROG Crosshair VII Hero
AMD Ryzen 3700x
NZXT Kraken X52 cooler
2x8 Gb Corsair Vengeance LPX DDR4 3000 C16
Samsung 970 EVO Plus - 1TB PCIe NVMe - M.2 Internal SSD (system drive)
Seagate Barracuda 2TB ST2000DM008 (backup drive)
Gigabyte GTX 1080
Corsair CX650 PSU

 

[InvalidError]

I'd say this falls within expectation: 3rd-gen PBO is more aggressive than 2nd-gen and needs sufficient voltage headroom to accommodate that aggressiveness. By adding a negative offset, PBO isn't getting the voltage headroom it is requesting from the VRM and lowers boosts to maintain stability.

 

[JayGau]

Ok it makes plenty of sense. I was suspecting something like that.

 

[CosmicDance]

I also upgraded to a 3700x from a 2700x a couple of days ago.

My 3700x temperatures are lower than my 2700x and I didn't use a negative voltage offset either - it also has PBO enabled for full boost.
It is about 58c in GTA V and 65c in Battlefield V which is just fine.
Also the all core boost is around 42.3 GHZ during gaming.

With the 2700x and PBO enabled my temperatures used to go to about 80c until I set a - .100 voltage offset to bring it down to around 70c during gaming.

They both use the identical cooler which is the bundled Wraith Prism.

I did use some decent quality thermal paste too and not the pre-applied stuff it came with.

Andy

 

[JayGau]

Your numbers are quite surprising Andy. Your temps for the 3700x and the stock cooler are lower than mine with a Kraken X52 (I am using Grizzly Hydraunot thermal paste).

When you write those numbers are they the average temp or the max peaks? Because me too my temp stays at 65C most of the time when I am gaming but I get some short bursts time to time (yesterday I got one at 78C while playing Battlefront 2). If you say you never see any peak higher than 65C while playing battlefield V I will have a very hard time to believe you.

About the 2700x, with PBO and performance enhancer enabled I got some high temps too with bursts at 80C but with everything at default I was barely reaching 70C, even for short bursts. It's why I said that my 3700x is running hotter when compared to the 2700x at default settings but similar when PBO was enable. Do you remember what was your temps with the 2700x and default settings?

 

[CosmicDance]

Yes I sometimes get peaks with the 3700x just as I did with the 2700x.
My figures are average temperatures or the temps that are most consistent if you like.

The 3700x has a much lower TDP than the 2700x so temperatures should be lower.
I was surprised by how much though as I reset my BIOS to defaults before installing it then expected to have to use a voltage offset to lower temps but this wasn't necessary.

There are so many posts on here and the AMD forums of the 3000 running with higher than normal temps.
Most of these posts are people benchmarking with Cinebench or Prime 95 though which absolutely hammer the CPU and raise temps to levels you never experience during gaming.

With my 2700x if I had default settings i.e. PBO on Auto and no voltage offset my temps were around 65c for gaming.
However, it would only boost to 3.9 GHZ so I set PBO to Enabled which boosted to 4.2 GHZ but temps went up to 75-80c.
After the minus voltage offset of - .100 they were around 68c but the usual spikes up to around 76c every so often.

I use a B450 series motherboard as well so, like everyone else using these with the Ryzen 3000 series, the BIOS are still effectively in Beta stages and temperatures vary a lot depending on the manufacturer.

 

[JayGau]

Ok so overall your temps are similar to mine in the same conditions. I still find it a little bit surprising since you use the stock cooler but on the other hand I have to say that my 3700x seems to perform more than the average. I got 97% percentile on UserBenchmark yesterday ("performs way above expectation") and everything is at default settings. So it makes sense that it runs a little bit hotter than the average too.

Oh and my room temperature is currently quite high too (around 27C) so at a normal 20-21C I would most likely stay below 70C most of the time (spikes included).

Also I am not sure the 3700x is supposed to run at lower temp than the 2700x despite the lower TDP for two reasons: first it has an additional 100 MHz max boost and as InvalidError wrote the PBO is more aggressive for the third generation chips so it boosts higher and more often than the 2700x. If you combine these you have better performance but also more heat despite the lower TDP.

 

[hftvhftv]

Ok so overall your temps are similar to mine in the same conditions. I still find it a little bit surprising since you use the stock cooler but on the other hand I have to say that my 3700x seems to perform more than the average. I got 97% percentile on UserBenchmark yesterday ("performs way above expectation") and everything is at default settings. So it makes sense that it runs a little bit hotter than the average too.

Oh and my room temperature is currently quite high too (around 27C) so at a normal 20-21C I would most likely stay below 70C most of the time (spikes included).

Also I am not sure the 3700x is supposed to run at lower temp than the 2700x despite the lower TDP for two reasons: first it has an additional 100 MHz max boost and as InvalidError wrote the PBO is more aggressive for the third generation chips so it boosts higher and more often than the 2700x. If you combine these you have better performance but also more heat despite the lower TDP.

The average is going to look real good if you start it when your package temp is only at 30C. I'd take them with a grain of salt. I would be more interested in the max temperatures.

 

[hftvhftv]

I saw very little difference in temperatures between auto-overclocking within Ryzen Master and manually setting voltage to 1.400V @ 4.2GHz with my R5 3600. But my voltages got quite a bit higher when using Ryzen Master. Haven't had any stability issues or BSOD even with Prime95 blend tests.

 

[TechyInAZ]

May I add a suggestion?? Everyone should list their memory speeds they're using.

When i was overclocking my Ryzen 7 1700X, I noticed a significant amount of temperature difference from different memory speeds, and especially different memory sub-timings. For example, using the Ryzen DRAM calculator, the difference in temperature between the "safe" and "fast" sub-timing configurations was as much as 5-7C during P95 testing.