Issues overclocking TR 2990WX using MSI x399 MEG Creation MB

F

fibonacc

n00b
Joined
Sep 12, 2018
Messages
4
Hello, I'm a new member, although I've been visiting this place quite often.

I have issues overclocking either manually or with Precision Boost override. Here are some of the specs before going into details: TR 2990WX, MSI x399 MEG Creation, Seasonic 1000W Gold PSU, Noctua TR4 cooler, 64 GB RAM at 2933 Mhz. OS is Windows 10 for Workstations (basically Pro).

The problem is that if I set PBO at anything above 300W (TDC/EDG at 350), the system freezes after a few iterations of IntelBurnTest High. It also freezes after 3-4 minutes in Prime95 small FFT. Cinebench, WPrime, Aida64 seem fine. Temperature is not an issue, the CPU and VRM barely gets to 64C before it freezes. I expected to be able to do at least 400W with PBO using this motherboard. I also tried manual overclocking from BIOS, setting CPU at 3.6 Ghz and voltage at 1.35 and same issue. The memory is not the problem, tested at 2400 Mhz and the freezing persists, and I had 4 sticks (32 GB) in a Ryzen 2700x system and it was stable.

Could there be an issue with the CPU? I find that quite unlikely, it's supposed to be top binned. I had a MSI X399 Carbon AC MB before, but it had multiple other issues besides not being able to clock high and sent it back to Newegg.

Thanks!
 
I used the same motherboard and CPU on an aging Thermaltake 1200 watt PSU. It couldn't handle the strain and the system would shut down almost immediately during testing. I switched to a newer Corsair 1000watt unit which handled the strain fine. All that said, freezing isn't usually a sign of a power supply problem in my experience. When i had similar problems it was related to memory settings. I also needed to increase the SOC voltage to 1.1v to push that CPU. Load-line calibration needed to be set for that CPU where as I didn't on the 1920X and 2950X CPUs.
 
Regardless of temps, unless you're chilling air or something that Noctua is nice but not enough for an OC 2990WX.

Also what ram exactly? If its not the good stuff, well something something pay for.

Also the temp monitor stuff is still behind the curve, had to update the few that read the sensor right and the one that works well (hwinfo64) will reboot the system if I leave the monitor tool running more than a few hours. Might be triggering a watchdog overflow or something, because I can kill it but leave everything loaded and it made the last week just fine.
 
Aluminum said:
Regardless of temps, unless you're chilling air or something that Noctua is nice but not enough for an OC 2990WX.

Also what ram exactly? If its not the good stuff, well something something pay for.

Also the temp monitor stuff is still behind the curve, had to update the few that read the sensor right and the one that works well (hwinfo64) will reboot the system if I leave the monitor tool running more than a few hours. Might be triggering a watchdog overflow or something, because I can kill it but leave everything loaded and it made the last week just fine.
Click to expand...

That's a good point. Overclocking the 2990WX required a triple radiator and a massive water block to pull off. Even then, I couldn't get all the cores above about 4.0GHz.
 
Dan_D said:
That's a good point. Overclocking the 2990WX required a triple radiator and a massive water block to pull off. Even then, I couldn't get all the cores above about 4.0GHz.
Click to expand...
I'm not trying to overclock it to anywhere near 4Ghz, I want to keep it around 3.6 Ghz, where it's using around 350W when compiling C++. This fails even when the temps are not getting close to the limit of 95 C. For example when doing manual overclocking, setting all cores at 3.6 Ghz with voltage at 1.35, the machine freezes when it's only at around 75 C. I would expect to see at least the temp. raising until the system shuts down. I'm going to try raising the SOC voltage too, maybe it helps. The RAM is some G-Skill CL 15, not the greatest, but works fine at 2933 and changing it to 2400 doesn't make the freezing disappear.
 
fibonacc said:
I'm not trying to overclock it to anywhere near 4Ghz, I want to keep it around 3.6 Ghz, where it's using around 350W when compiling C++. This fails even when the temps are not getting close to the limit of 95 C. For example when doing manual overclocking, setting all cores at 3.6 Ghz with voltage at 1.35, the machine freezes when it's only at around 75 C. I would expect to see at least the temp. raising until the system shuts down. I'm going to try raising the SOC voltage too, maybe it helps. The RAM is some G-Skill CL 15, not the greatest, but works fine at 2933 and changing it to 2400 doesn't make the freezing disappear.
Click to expand...

I'd try 1.1v SOC. That worked for me. A system hard locking doesn't have to be due to temperature. My test system was locking up before it got hot enough to be a problem. The issue was relating to memory settings. There are many possible settings and speed isn't the only one. The DRAM clock may not be your problem. With the G.Skill FlareX RAM, I could run tighter timings than I can with the Corsair stuff I've got. The Corsair RAM has to have the voltages set slightly higher than 1.35v to work right. I usually need 1.36v or higher on Corsair RAM. You can also try loosening the timings or checking to make sure they are detecting and being set properly.
 
Dan_D said:
I'd try 1.1v SOC. That worked for me. A system hard locking doesn't have to be due to temperature. My test system was locking up before it got hot enough to be a problem. The issue was relating to memory settings. There are many possible settings and speed isn't the only one. The DRAM clock may not be your problem. With the G.Skill FlareX RAM, I could run tighter timings than I can with the Corsair stuff I've got. The Corsair RAM has to have the voltages set slightly higher than 1.35v to work right. I usually need 1.36v or higher on Corsair RAM. You can also try loosening the timings or checking to make sure they are detecting and being set properly.
Click to expand...

Tried 1.1 V for SOC and relaxed RAM timings, doesn't help. I probably give up, it would be only a few percent faster anyway and the TR from next years should be quite a big improvement. Still, it's quite disappointing not to be able to overclock a bit more having this motherboard. Thanks for the help!
 
Last edited:
Have you tried compiling without the overclock and letting pb2/pbo do their thing?

It has been shown that sometimes a locked over clock doesn’t really net any gains.
 
  • Like
Reactions: Dan_D
like this
lightsout said:
That's insanity I had no idea a cpu could draw that type of power.
Click to expand...

Yes 500 to 750 watts over clocked even slightly. And the other 200+ is ram. Storage. Gpu and other subsystems.

A power supply that is rated for 1kw is really nominal at less than that.
 
Hakaba said:
Have you tried compiling without the overclock and letting pb2/pbo do their thing?

It has been shown that sometimes a locked over clock doesn’t really net any gains.
Click to expand...

I'd let PBO take care of it. It won't necessarily clock as high but it might run better as some cores will run slightly faster. I got 4.25GHz on some cores using a 2950X. This was actually faster in some things than running all 16 cores at 4.2GHz.
 
Dan_D said:
My 1,000 watt Corsair unit did fine with it.
Click to expand...

Yeah but you need overhead.

Its safer to recommend a 1kw plus size because if the CPU alone is using 750 watts overclocked, toss a dozen hard drives, a few solid states, maybe a good bit of cooling fans and a GPU or 2 and you can rapidly surpass a kw.

I used to run 4 hd6970s on my x79 board with a 6 core 3930k. It would pop 15 amp breakers in my house. I was using a 1200 watt psu.
 
Seems to be from the memory after all, with the clocks lowered it's fine at 350w PBO (didn't try more yet). The PSU should be more than enough, max. 400w for CPU, the rest is below another 400. At this point it's unlikely I will get some faster RAM, would need to sell these 64 GB on Ebay. I'll keep the system like this until next year when I will get a Zen2 TR.
 
Last edited:
b-dies are not the be-all/end-all for everyone AMD.

I'm running b-die. dual rank 16GB/dimm x 4 dimms (there ain't no single-rank 16GB sticks that I know of). Can't run at stock XMP specs. Tried 3 different bios (newest one was really unstable). Tried upping the ddr voltage as high as 1.38v. Tried upping the vddr. Every time I would get random bsod crashes. At the end of the day I had to drop the speed down a notch and leave xmp voltages and other stiff as defined by the spd xmp settings. After that not a single app crash or bsod.

And yes, GSkill trident z 3200 14,14,14,14, 34,48,1T timings and I verified the b-dies with taiphoon. Perhaps it is a mobo issue (trying to force 2T command rate is apparently ignored) or simply a TR 1950x issue. At the end of the day running my memory at 3166 does not cause any subjective grief. It is just that after paying almost $1200 CA for the memory it is pretty frustrating not to get their entire goodness - forget overclocking them (though suppose I could if I slack the timings to 18 or so, but c'mon!).

--- Edit - Power, dropped on me before I finished... --

I know with AMD that XMP specs are more of a guideline than engraved in stone. I just expect the expect GSkill's memory to be the go-to memory for all things AMD.
 
Last edited:
Dan_D said:
I'd let PBO take care of it. It won't necessarily clock as high but it might run better as some cores will run slightly faster. I got 4.25GHz on some cores using a 2950X. This was actually faster in some things than running all 16 cores at 4.2GHz.
Click to expand...


PBO on a 2950x should keep the full 4.4ghz across 4 cores, at least it does on mine.
 
xmp setting are for intel platforms and dont always work on intel. manually set all the timings and voltage. also that ram can take 1.4v, all ram can take +10% no prob.
 
You must log in or register to reply here.
Back
Top