Apparent memory issues... but only while compiling cpp code?!?

[BoyC]

Hi.

I built a new pc earlier this year, I made it top of the line so it can last me a decade if I'm lucky.

Motherboard: ASUS PRIME Z790-P

CPU: INTEL Core i9-13900KF

GPU: Gigabyte 4090

Memory: G.SKILL 64GB Ripjaws S5 DDR5 6000MHz CL30 KIT F5-6000J3040G32GX2-RS5K
PSU: SEASONIC Prime PX 1600W
Main storage: GIGABYTE 2TB AORUS NVMe GEN4 SSD M.2 PCIe M.2 2280
And an arctic cooling AIO for good measure

 

This is my personal PC and I go through phases of gaming a lot and developing a lot on it (sort of a relax phase and creative phase). I started developing heavily this august for a deadline and I noticed an odd behavior. Randomly Visual Studio, especially when doing a rebuild all of my project, will say that it's out of heap memory, or throw up issues that make absolutely no sense, like complaining about some random character in a header file that belongs to the windows sdk, stuff like this. Compiling multiple times kinda can get around the issue, so I limped along to my deadline, but after that I immediately started investigating the problem.

The same projects that are not building on this pc build without any issues everywhere else on the same software stack, including an intel nuc, my laptop, my work pc, and various other pcs from the last 13 years of development that I used. A coworker had similar issues with his new work pc, everything was fine until a project build, for him changing the memory timings a bit helped.

 

Interesting facts about the issue:

* Setting the compiler to be 64 bit helps somewhat, but the issues only get more rare, and a full rebuild usually brings them out.

* There are three issues the compiler spouts randomly: out of heap error, internal compiler error or random failures complaining about code that should be fine, as if there was some sort of memory corruption

 

Here's what I tried.

1. fresh windows install

2. bios update

3. did a windows memory test

4. turned off xmp

5. turned off the efficiency cores

6. tried shuffling the memory sticks around

7. tried various memory frequency and timing settings, but i'm not really versed in this so i definitely didn't do an exhaustive test

8. tried the sticks one at a time

9. saw that my exact memory kit isn't showing up on the motherboard's supported list anymore, so i actually got me a new kit just today: KINGSTON FURY 64GB Beast DDR5 5600MHz CL40 KIT KF556C40BBK2-64

Same issue, all the time.

At this point I'm thinking the memory may actually be fine and the culprit is either the cpu or the motherboard.

 

Odd thing is that for gaming this machine has been absolutely perfect and obviously a beast. It's just when I try to use it as a build server when it craps the bed.

Could you guys give me some tips on how I could test the system more thoroughly in order to locate the problem?

Thanks.

[Zomeguy]

Could try software like memtest86 or Aida64 to test the system and see if you get some more details on what's going on.

 

The fact that you switched memory and it's still happening would rule out the memory unless you're really unlucky with memory but lean toward the CPU or motherboard as the issue.

 

[Tetras]

14 minutes ago, BoyC said:

4. turned off xmp

You could try turning off XMP (I'd still keep the DRAM's voltage) and the CPU's turbo modes, so that if there's some kind of instability issue at high clocks it should help to rule it out.

 

Have you tested the SSD at all?

 

When you get the compiler errors, is there anything in the event viewer?

 

17 minutes ago, BoyC said:

A coworker had similar issues with his new work pc, everything was fine until a project build, for him changing the memory timings a bit helped.

Are there any common features between the two builds?

 

Do you have Windows security features enabled (virtualisation/HVCI) that maybe Visual Studio is not fully compatible with?

[BoyC]

1 hour ago, Tetras said:

You could try turning off XMP (I'd still keep the DRAM's voltage) and the CPU's turbo modes, so that if there's some kind of instability issue at high clocks it should help to rule it out.

VERY good call, disabling turbo boost actually makes the problem completely go away.

Is this an RMA kinda situation then?

[Tetras]

1 hour ago, BoyC said:

VERY good call, disabling turbo boost actually makes the problem completely go away.

Is this an RMA kinda situation then?

You could try increasing the stock voltages a bit with an offset and see if that helps, with these CPUs I'm not completely convinced how much they're tested to ensure 100% stability with the stock profiles and presumably there's still some variance to account for.

 

If it passes Intel's diagnostics (here), then I'd assume it would not be considered broken, unless it reproducibly fails something like Prime 95.

[BoyC]

2 hours ago, Tetras said:

You could try increasing the stock voltages a bit with an offset and see if that helps, with these CPUs I'm not completely convinced how much they're tested to ensure 100% stability with the stock profiles and presumably there's still some variance to account for.

 

If it passes Intel's diagnostics (here), then I'd assume it would not be considered broken, unless it reproducibly fails something like Prime 95.

It did pass the Intel diagnostics and just about any diagnostics I threw at it. I guess compiling code is just that much more of an all-around test. With the turbo boost revelation I could find a bunch of people complaining of similar issues with the exact cpu-motherboard combination, so back to the shop it goes for a round of RMA.

Thanks for pointing me in the direction of the next step, I was stuck on this for weeks now.

[BoyC]

Ok I did some more tests and as soon as I start a build the CPU temp spikes instantly, not sure if it's the AIO not being able to keep up or a CPU issue, but man.

Also, doing a bunch of rebuilds for about 2 minutes resulted in the first BSOD i've seen, but with numbers like that 113 that's no wonder.

[Tetras]

2 hours ago, BoyC said:

Ok I did some more tests and as soon as I start a build the CPU temp spikes instantly, not sure if it's the AIO not being able to keep up or a CPU issue, but man.

Also, doing a bunch of rebuilds for about 2 minutes resulted in the first BSOD i've seen, but with numbers like that 113 that's no wonder.

Did you re-enable the turbo modes for this testing?

 

I suspect the reason why it happens is because of the voltage increase, a 13900KF has two additional boost technologies: thermal velocity boost and adaptive boost, alongside turbo boost max 3.0.

 

You could try disabling these, or a combination of them, to see if it curbs excessive boost behaviour.

 

I'm not sure how an AIO would respond to the temperature of only one core increasing, you may be able to change what it responds to (and how fast it responds) in the BIOS.

[BoyC]

Update:

It's been 4 weeks since the PC was sent back and today the supplier finally acknowledged the issue and I'm getting a full refund for the CPU. Which is cool since in the meantime 14th gen came out and since the original cost covers the price of the 14900kf the question of "you really shouldn't be upgrading from 13th gen" became a moot one. Thanks for the help everyone, hopefully in a week's time I'll have my pc back.