PrimeGrid PPS LLR (CPU temp) challenge (for Ryzen 3000-series owners)! (Testing heat density of TSMC 7nm)

[Space Tyrant]

Ok, I've been tearing my hair out (well, thankfully not quite literally), trying to keep my R5 3600 CPU cool, while crunching 12 threads of PrimeGrid PPS LLR.

At this point, I would like to see other people replicate my setup, and see what their temps are.

I've used both a Gigabyte B450 AORUS PRO WIFI ATX, with both AGESA 1.0.0.2 (BIOS F40) and 1.0.0.3AB (BIOS F41b), and an Asus B450-F ROG STRIX Gaming ATX, with AGESA 1.0.0.2 (BIOS 2501).

I've tried both stock settings, as well as a mild 4.000Ghz manual OC with 1.365V, using the "AMD Overclocking" menu, "manual OC".

I've used both 4x8GB DDR4-3000 @ XMP, as well as 2x8GB GSkill RGB DDR4-3600 (with FCLK 1800 as well), in current setup (Asus B450-F).

I've used four different air coolers, as well as finally, on current rig, 240mm AIO WC (CoolerMaster). Still 97.5C running 12 PrimeGrid tasks. On water. Before the water even got a chance to "heat up", after being idle all night, 44C idle temp.

So if you would like to replicate, if you have a 3600 (or higher) 3rd-Gen Ryzen CPU, head on over to http://www.primegrid.com/ , sign up for an account, and download BOINC, from boinc.berkeley.edu , newest version for Windows, and then "Add Project", put in your PrimeGrid credentials. Note that you should also go to your PrimeGrid account page, on their site, and select "PPS LLR (CPU)" as your project type, and uncheck EVERYTHING ELSE.

Then in BOINC, under "Compute Preferences", make sure that "CPU count" and "CPU time" are both "100".

Edit: Oh yes, you should be running Win10 1903 64-bit, and download and installed the newest (7/7/2019 release or newer) AMD Chipset Drivers, and also, newest AMD Ryzen Master. Measure temps with Ryzen Master.

Edit: Btw, does anyone know, does the Ryzen R5 3600 use TIM/paste, rather than solder, on their lowest-end model? I suspect that they might.

You're running a 1.365v all core OC, running 12 (presumably) heavy compute threads and hitting 97C (in a cubby?) on your 3600?

That doesn't sound too shocking to me. You've OC'd the 3600 well beyond my 3600X base clock (3.8GHz, btw) on high voltage and your cooling is likely somewhat compromised from the cubby air pocket.

I have my 3600X in my desk cubby too and am running on peasant air with a lowly Wraith Spire 2 -- but it's augmented with a 120mm fan, FTW! I also have SMT disabled so I should be generating substantially less heat with my stock 3600x than your fire-breathing 3600.

I also have the door removed from the cubby; I'm guessing yours is also? My case is an open air Lian Li T60 so the cubby can stand in as a passively air-cooled outer case. Other than the cpu fan and the barely spinning PSU fan, there are no other fans in my cubby.

Anyway, I just did some temp testing. Starting from idle temp of 38C, I ran 5 consecutive ycrunchers (1B digits of Pi) and hit 86C in about 7 minutes. When I was running the cooler with the original 92mm fan, it hit 93C in the same test.

I don't think there is anything wrong with your CPU. If there were, my money would be on a substandard solder layer.

 

[Abwx]

I did post a screen shot. :|

Wasnt about you, you should have guessed...

Edit : It s quite surprising that there s no offset with AVX2 loads that get close to full throughput, even at base clocks.

 

[VirtualLarry]

Here's a pic (2, actually), of my B450-F ROG STRIX ATX board, with the 240mm CM AIO WC attached to the CPU.

Note that (before I took the RM pic), it was showing 103C. Which then I realized, wasn't too good, for water-cooling (Steam, pressure, etc.). But then I felt the water lines on the cooler, one felt warmer than the other, but neither one felt even remotely close to 100C.

Also, the metal plates that attach the cooler/heatplate/pump to the CPU, the side edges, are exposed, and I can hold my finger to them, and it doesn't seem too hot to touch.

So, either this particular CPU, isn't transferring heat through the heatspreader (which would explain, why temps could jump from 44C to 95C in an instant, even under water), OR
My AIO WC already sprung a leak, and there's nothing circulating in it, OR
My temp sensor on my CPU is borked pretty hardcore.

Edit: Installed Asus "Ai Suite 3", installed "Dual Intellegent Processor 5", the CPU temps reported in that program, are 10C LOWER than Ryzen Master and HWMonitor. And given that I can touch the metal-attached heatsink mounting lugs, without having to pull away, indicates that they aren't getting THAT hot.

Also, I was finally able to check the pump RPMs using DIP5, it was showing 2280 RPM. So that's healthy.

Edit: Hmm, was on the cell, someone called, when they hung up, PC monitor went to blue-screen (lost signal), USB keyboard + mouse lights went out, front case LED intake fans still lit up, no response pressing a keyboard key or moving mouse (presumably since they were off, for some reason). Almost like a sleep mode, but I had disabled PC sleep, but left monitor sleep.

A little disturbing, but then again, I am pushing things a little bit. (3600 RAM, 1800FCLK, 4000Mhz fixed OC, 1.365V in BIOS, which CPU-Z and Asus DIP5 report as up to 1.417V.)

 

[GhostPirateLeChuck]

I wouldn't necessarily trust Ai Suite 3. On the desktop I just replaced, it would routinely tell me my case fans weren't active (despite being able to hear them run), and it would also alert me that the vcore had dropped to 0.00 while I was working.

Any chance something is gunking up the microchannels in the cooling plate, like those Enermax TR4 CLCs?

 

[Zucker2k]

I saw a similar post on reddit where the vrm temp and current spike were causing the cpu to also raise voltage and spiking temps. He was also on a B450 board. The culprit sensor was TMPIN5. Run your workload and observe your sensors.

 

[Space Tyrant]

Here's a pic (2, actually), of my B450-F ROG STRIX ATX board, with the 240mm CM AIO WC attached to the CPU.

Note that (before I took the RM pic), it was showing 103C. Which then I realized, wasn't too good, for water-cooling (Steam, pressure, etc.). But then I felt the water lines on the cooler, one felt warmer than the other, but neither one felt even remotely close to 100C.

Also, the metal plates that attach the cooler/heatplate/pump to the CPU, the side edges, are exposed, and I can hold my finger to them, and it doesn't seem too hot to touch.

So, either this particular CPU, isn't transferring heat through the heatspreader (which would explain, why temps could jump from 44C to 95C in an instant, even under water), OR
My AIO WC already sprung a leak, and there's nothing circulating in it, OR
My temp sensor on my CPU is borked pretty hardcore.

Edit: Installed Asus "Ai Suite 3", installed "Dual Intellegent Processor 5", the CPU temps reported in that program, are 10C LOWER than Ryzen Master and HWMonitor. And given that I can touch the metal-attached heatsink mounting lugs, without having to pull away, indicates that they aren't getting THAT hot.

Also, I was finally able to check the pump RPMs using DIP5, it was showing 2280 RPM. So that's healthy.

Edit: Hmm, was on the cell, someone called, when they hung up, PC monitor went to blue-screen (lost signal), USB keyboard + mouse lights went out, front case LED intake fans still lit up, no response pressing a keyboard key or moving mouse (presumably since they were off, for some reason). Almost like a sleep mode, but I had disabled PC sleep, but left monitor sleep.

A little disturbing, but then again, I am pushing things a little bit. (3600 RAM, 1800FCLK, 4000Mhz fixed OC, 1.365V in BIOS, which CPU-Z and Asus DIP5 report as up to 1.417V.)

I'm curious; how did you arrive at 1.365V for 4GHz? I was kinda hoping I could get by with much lower vcore than that. I've been putting off OC experiments until I get a better bios in one of my MB's, but now I'm starting to get impatient...

 

[Kenmitch]

I'm curious; how did you arrive at 1.365V for 4GHz? I was kinda hoping I could get by with much lower vcore than that. I've been putting off OC experiments until I get a better bios in one of my MB's, but now I'm starting to get impatient...

Running stock settings, pbo off, voltages on auto, b=die @3600 CL14 , x570 MB....My 3600 boosts to 4000-4025MHz @ 1.365-1.380v's under all core load. I tend to believe it's hard coded into the power tables.

My MSI MB's uefi seams to be pretty decent so far. The PBO stuff looks to be more or less non functional as far as I can tell. Like you I'm waiting for a more mature uefi to dabble around some more.

 

[maddogmcgee]

I'm curious; how did you arrive at 1.365V for 4GHz? I was kinda hoping I could get by with much lower vcore than that. I've been putting off OC experiments until I get a better bios in one of my MB's, but now I'm starting to get impatient...

A quick stability check had my 3700x passing 4200 all core at 3.125 (Ryzen Master)......did not have time to check if it was really stable and voltages were accurate but those volts do seem very high/dangerous for an all core.

 

[IEC]

A quick stability check had my 3700x passing 4200 all core at 3.125 (Ryzen Master)......did not have time to check if it was really stable and voltages were accurate but those volts do seem very high/dangerous for an all core.

Considering I can do 4.2GHz AVX2 stable on my worst chiplet and 4.35GHz AVX2 stable on my better chiplet at 1.25V those voltages are absurd for 4GHz.

But hey, don't take my word for it, Silicon Lottery's absolute worst binned 3900X is guaranteed stable under their conservative binning at 4GHz all-core at 1.20V.
https://siliconlottery.com/collections/matisse/products/3900x40g?variant=29472043499606

1.365V (and it looks like 1.40V is being applied from Larry's screenshots) is absurdly high for 4GHz manual OC on Ryzen 3000.

 

[Kenmitch]

A quick stability check had my 3700x passing 4200 all core at 3.125 (Ryzen Master)......did not have time to check if it was really stable and voltages were accurate but those volts do seem very high/dangerous for an all core.

That's crazy high voltage....Even for a typo!

 

[VirtualLarry]

1.365V (and it looks like 1.40V is being applied from Larry's screenshots) is absurdly high for 4GHz manual OC on Ryzen 3000.

Well, that's basically what it was getting, stock, more or less, so I don't see why that's so absurd for an all-core manual OC either.

And I tried 1.25V, it booted Windows, but was giving "Calculation Error"s on PrimeGrid tasks.

I just remember trying 1.350V on my B450 AORUS PRO WIFI board, and it being unstable, and having to bump it up a notch.

Also, all of you that are trying to compare a bottom-binned 3600 CPU, with a top-binned 3900X CPU, as far as voltages and clocks, need to understand that the voltages that those CPUs need, are not the same.

Edit: Tried 1.30V, "Computation Error". Not sure why you guys don't believe me on the voltage.

I had the same issue with my G3258 CPUs. People on here, were like, 1.2V for 4.0Ghz, I was like, no way dude. I needed 1.280V-1.300V for 4.0Ghz, and higher for 4.2Ghz (which was never fully stable for me).

Of course, other peeps were just gaming, I was running DC, which demands a whole 'nother level of stability.

 

[Space Tyrant]

Considering I can do 4.2GHz AVX2 stable on my worst chiplet and 4.35GHz AVX2 stable on my better chiplet at 1.25V those voltages are absurd for 4GHz.

But hey, don't take my word for it, Silicon Lottery's absolute worst binned 3900X is guaranteed stable under their conservative binning at 4GHz all-core at 1.20V.
https://siliconlottery.com/collections/matisse/products/3900x40g?variant=29472043499606

1.365V (and it looks like 1.40V is being applied from Larry's screenshots) is absurdly high for 4GHz manual OC on Ryzen 3000.

I just got back from a pstate oc session. I wound up with 4.0GHz at 1.18~ vcore, tested stable with 5 consecutive runs of ycruncher 1B Pi test. CPU hit a peak of 72C (down 14C from the stock test). Both tests were done with SMT off so a bit more voltage (~1.2v?) would be required with SMT and a bit higher temps would result as well.

This is a 3600X with a Spire 2 cooler, btw, in an open case inside a desk cubby -- but the 92mm fan has been replaced with a 120mm.

Pretty consistent with your numbers, all in all.

 

[maddogmcgee]

That's crazy high voltage....Even for a typo!

Considering I can do 4.2GHz AVX2 stable on my worst chiplet and 4.35GHz AVX2 stable on my better chiplet at 1.25V those voltages are absurd for 4GHz.

But hey, don't take my word for it, Silicon Lottery's absolute worst binned 3900X is guaranteed stable under their conservative binning at 4GHz all-core at 1.20V.
https://siliconlottery.com/collections/matisse/products/3900x40g?variant=29472043499606

1.365V (and it looks like 1.40V is being applied from Larry's screenshots) is absurdly high for 4GHz manual OC on Ryzen 3000.

I am sure 3.125 is way too high. Unfortunately I did not have time to do anything other than dial in what I assume is a safe voltage, have a stab at a reasonable clock and stability test for a few minutes. Oh the joys of a newborn!

From what I have read anything above low 1.3's is probably dangerous for a sustained load and appears to be the rough limit for a stock CPU under load (mine hits 1.375 under less stressful stability tests). 1.25v does seem like a safer aim for permanent overclock though even if my CPU does not make it as high as yours.

If my CPU did need 1.4v for 4GHz I would be sticking with 3.9 or stock PB I think!

 

[coercitiv]

Also, all of you that are trying to compare a bottom-binned 3600 CPU, with a top-binned 3900X CPU, as far as voltages and clocks, need to understand that the voltages that those CPUs need, are not the same.

A difference from binning is to be expected, but not as high as 0.2V. More importantly, your system appears to be unstable even with this voltage. (that's assuming the problems come from Vcore and not memory OC / FCLK / SoC voltage)

Between your tests with different motherboards, unusually high idle and load CPU temps and suspiciously high (and likely unstable) 1.4V Vcore, I'd say there's something wrong with your CPU. If I were you I would take the necessary steps to replace it.

If you still want to hang on to this one, the next logical step would be to "collect" VID table information and compare that with what's needed on your side to keep the system stable without memory OC.

 

[GhostPirateLeChuck]

Whelp. It was fun while it lasted. Looks like the motherboard went tits up about an hour ago. Not even getting the little reset switch on the motherboard to light up. I pulled the PSU out of it, and connected it to the old desktop, and it powers up. I tried the PSU from that PC in the new one, and no dice there. I've been slowly disconnecting things from the motherboard, but the only things left at this point are the RAM, NH-D15, and the CPU.

It was going so well, too. I was getting some last minute software installed for work tomorrow before bed, enabled Hyper-V in Win10 and rebooted, then....nothing.

Looks like I get to yell at NewEgg and Gigabyte in a few hours. Limping along with this laptop sucks hard. I should've just gone with my gut, and grabbed an Asus X570 Strix-E, or something.

 

[VirtualLarry]

A difference from binning is to be expected, but not as high as 0.2V.

And you know the intricacies of TSMC's7nm process and AMD's binning... how? How are you so sure of that? Surely, you must have some sort of factual basis? (*)

Like I said, at stock, in both my B450 mobos, one Gigabyte, and with AGESA 1.0.0.2 and 1.0.0.3AB, as well as an Asus B450 with AGESA 1.0.0.2, at "stock" settings, the CPU boosts to nearly, but not quite, 4000Mhz, and it does so around 1.35V, give or take. So I don't think that I'm somehow magically "out of line" to set my stock vcore at that range, for a manual OC around that frequency.

(*) I see to recall Q6600 CPUs, with stock VIDs of ~1.3xV, through ~1.5xV. So a binning range like that isn't unexpected, on some processes. Granted, the vast majority of chips are withing a normal distribution, around a median (?), but Intel's ARK used to give voltage ranges, for their entire bin range, for some of their older CPUs. (Probably scrubbed from ARK by now.)

 

[loafbred]

I'm too lazy to load another benchmark, but here's Prime95 small FFT, after two or three minutes. 3600X, Gigabyte x470 Aorus Ultra Gaming, bios F40, and Noctua NH-D15 cooler. I used the 41b bios for a couple of days, but it runs ~100 MHz slower (with slightly lower vcore). All bios settings are default with the exception of RAM at 3600 MHz CL16. I tried several of the Windows power plans, and settled on the Windows "balanced". I would share the behavior of each, but I didn't keep records, and thinking about it makes my head hurt.

 

[coercitiv]

And you know the intricacies of TSMC's7nm process and AMD's binning... how? How are you so sure of that? Surely, you must have some sort of factual basis? (*)

Wrote by TheStilt, AFAIK reflects the info we got from AMD reps as well.

According to FIT, the safe voltage levels for the silicon are around 1.325V in high-current loads\ and up to 1.47V in low-current loads (i.e ST), depending on the silicon characteristics.

 

[IntelUser2000]

Edit: I noticed that you're using an X570 board, I wonder how much that factors in, as far as "mature" BIOS, versus all of the beta-esque BIOS releases that we've been getting for the older B450 and X370 boards.

The board could be the issue. Keeping compatibility between different architectures and power requirements are challenging.

B450 is a lower end so its understandable. I know the motherboards nowadays come with all sorts of buzzwords(especially with durability and stability)but really I don't think they are are good as they say they are.

Your hardware sensor seems wonky too. You make a good point about 103C being weirdly cool on the mounting bracket.

If you can't keep it stable with an overclock, don't. Test everything with stock first.

 

[VirtualLarry]

Your hardware sensor seems wonky too. You make a good point about 103C being weirdly cool on the mounting bracket.

If you can't keep it stable with an overclock, don't. Test everything with stock first.

Yeah, Asus AiSuite 3 / DIP5, shows temps of 89C at most thus far, not 103C, and it basically IS stable, @ 1.365V, just not at 1.25V, or 1.30V.

 

[VirtualLarry]

I mean, if my coolant temp was actually 100C or higher, wouldn't I hear bubbling/boiling, cavitation on the pump, etc.? Or at least, hear something out of the ordinary?

I think that my sensor might be borked too. Asus DIP5 / AiSuite3 says 75C, running 12 threads of WCG. Like 89C doing 12 threads of PrimeGrid PPS LLR (uses AVX2, very intensive tasks, probably as bad or worse than Prime95 does).

 

[Space Tyrant]

I mean, if my coolant temp was actually 100C or higher, wouldn't I hear bubbling/boiling, cavitation on the pump, etc.? Or at least, hear something out of the ordinary?

I think that my sensor might be borked too. Asus DIP5 / AiSuite3 says 75C, running 12 threads of WCG. Like 89C doing 12 threads of PrimeGrid PPS LLR (uses AVX2, very intensive tasks, probably as bad or worse than Prime95 does).

No way you're gonna boil water with your AIO without knowing. You wouldn't be able to touch the radiator without getting burned, for example, if the coolant is actually flowing. And, if your fans are actually blowing air through the radiator, the temperature delta between the coolant and the CPU (package, not cores) would be much larger than a few degrees.

The heat density of these things, even at stock, is crazy. I think they'll need to go to a full-width vapor-chamber-IHS before the next shrink -- if they can't get a major power consumption reduction in the same process.

 

[VirtualLarry]

The heat density of these things, even at stock, is crazy.

I fully agree!!!

 

[Space Tyrant]

I fully agree!!!

If you can see your CPUTIN temp, there should be a big delta between that and the actual core temps. That temp is closer to what your cold plate is encountering, so your coolant should be even cooler than that -- assuming both water and air are flowing, of course.

 

[Elfear]

3900X with PBO off. Ran Prime Grid for a few minutes using the settings you specified. That's on custom water.