I just succesfully baked DDR4

[Shrek]

I have a small heatgun for reflow so I can use flux under the chips.

 

[A&P211]

Thats great and all, but I've personally been kind of hesitant to ever try the oven trick just since you know... I use the oven to cook food.

OP uses the oven to cook RAM. Everyone life choices.
You cook food, OP cooks ram and I dont do anything with my oven.

 

[Jism]

I have a subscription to Glovo and Uber. What do i need an oven for? Oh wait, Bake DDR4!

16 hours later, memtest passed.

 

[Steevo]

Awesome.

 

[Jism]

Can anyone explain what is going on tho?

I used the sticks ; one of them gets faulty at a random moment. I bake it; everything seems fine and runs for the next 3 years (lol). same happening again, i bake it and it runs fine again.

Is it the heat cycles?

 

[Shrek]

Yes, thermal fatigue of the solder; if you used flux it would last a lot longer, but avoid an oven used for food.

 

[Jism]

Interesting.

 

[iameatingjam]

OP uses the oven to cook RAM. Everyone life choices.
You cook food, OP cooks ram and I dont do anything with my oven.

Never too late to try something new. Bake one of those defective intel cpus and tell us what happens.

jk

 

[Steevo]

Can anyone explain what is going on tho?

I used the sticks ; one of them gets faulty at a random moment. I bake it; everything seems fine and runs for the next 3 years (lol). same happening again, i bake it and it runs fine again.

Is it the heat cycles?

The solder.has too much Tin, and tin forms crystals in the solder matrix since its a anamorphic blend, the " cure is to replace the solder or bake every few years.


It's was probably the beginning or end of a batch where the balls were out of spec or just a few were.

 

[Jism]

Yep; might be a true reason on why it happens. Not that i care; i'm wanting to replace the kit with a QVL specified one that at least operates at the intended speed(s).

I also noted yesterday that setting 1.35V in BIOS yields 1.39V on RAM. I know some DDR4 can really do bad once with too high voltages but 0.4v should be far within spec right knowing DDR4 is up to 1.45V ~ 1.55V or so.

 

[tabascosauz]

Can anyone explain what is going on tho?

I used the sticks ; one of them gets faulty at a random moment. I bake it; everything seems fine and runs for the next 3 years (lol). same happening again, i bake it and it runs fine again.

Is it the heat cycles?

I'd test with something worthwhile that isn't memtest86, before coming to the conclusion that things are stable. HCI memtest, TM5, Karhu

Assuming from your running 3200CL14 that it's B-die (technically CJR can also do it but loose 14-17-17). B-die is temp sensitive, so anything that doesn't stress it enough to bring the temps up doesn't deserve to be called a memtest, because you don't get a complete picture as to load stability.

DRAM overvolt is common, but different on every board. Some overvolt as much as 0.06V+.

DDR4 is most certainly not good for 1.55V, or even beyond 1.45V. That's only certain ICs (B-die, Rev.E/B, DJR) and still it's a tossup as to whether it will cause you problems down the line, if you have a bad bin B-die for example you should not be so daring with volts.

 

[Jism]

It is a 3200Mhz CL14/16/16/5 something kit; its capable of doing well above 3600Mhz at a tad loser timings. There's OC potential in that because i had that kit with a Asus X470 motherboard and the same CPU. It would just operate fine at 3466Mhz with CL14 and such. The bad thing is that the Intel NIC (onboard) was faulty and suddenly stopped working, and approx 2 weeks later took the whole board with it. As a replacement i've bought a X570 Aurus Elite of Gigabyte - it does the trick but it refuses to work above anything 3133Mhz. I can reach to 3166Mhz using Bclk but anything higher is non-post.

It's a memory QVL issue - no matter the timings i set (manually, automaticly or put timings down) it just refuses to boot beyond a certain speed. When i replace that kit with a valid one i should be able to run as intended. 3000Mhz vs 3200Mhz is'nt the world.

 

[tabascosauz]

It is a 3200Mhz CL14/16/16/5 something kit; its capable of doing well above 3600Mhz at a tad loser timings. There's OC potential in that because i had that kit with a Asus X470 motherboard and the same CPU. It would just operate fine at 3466Mhz with CL14 and such. The bad thing is that the Intel NIC (onboard) was faulty and suddenly stopped working, and approx 2 weeks later took the whole board with it. As a replacement i've bought a X570 Aurus Elite of Gigabyte - it does the trick but it refuses to work above anything 3133Mhz. I can reach to 3166Mhz using Bclk but anything higher is non-post.

It's a memory QVL issue - no matter the timings i set (manually, automaticly or put timings down) it just refuses to boot beyond a certain speed. When i replace that kit with a valid one i should be able to run as intended. 3000Mhz vs 3200Mhz is'nt the world.

May want to be careful about everything else that's been in that board. When one of my Gigabyte boards died it was the only thing that completely gave up the ghost but both CPU (5700G) and B-die kit were noticeably damaged and had to be RMA'd later as well. RAM was quite obvious, and the CPU took a huge hit to achievable fabric and memory speeds

Older B-die can be A0 like the 4000CL19 kit I killed from voltage, but although your max speed will be limited just above 4000, it should still be able to do some respectable B-die things as to timings.

Anyway, just make sure you memtest properly (certainly not memtest86) before entrusting stability to it

 

[Jism]

I had it run for at least 16 hours yesterday - zero issues. Since i had the system on for weeks without any reboot and no notable chrashes or issues, i can kind of assume it's stable. Memtest is quite good at finding issues or errors - no need to hammer on those sticks. This is not ram that needs to be in a server or something.

 

[The King]

Can anyone explain what is going on tho?

I used the sticks ; one of them gets faulty at a random moment. I bake it; everything seems fine and runs for the next 3 years (lol). same happening again, i bake it and it runs fine again.

Is it the heat cycles?

The oven bake trick is actually only a temporary solution. The Micro fractures in the solder will eventually return after sometime after several heat cycles.

If the RAM is 3200 CL14-16-16 @ 1.35V it is possibly older Samsung B-die.
Most of these RAM kits have life time warranty if you find the local distro for your RAM they may replace them for you for free.

I would not tell them about baking it though!

 

[Jism]

Nah, no need to. It's hynix btw. And it's just not operating at intended speeds as it should (explained above).

I could not care less; a new kit of up to 64GB is only around 200 to 250 euro. That's nothing for the amount of hours i use my PC for.

 

[Shrek]

The oven bake trick is actually only a temporary solution. The Micro fractures in the solder will eventually return after sometime after several heat cycles.

Unless ... one uses flux.

 

[Jism]

Kind of impossible to flux things up while doing it in the oven, right?

 

[Shrek]

I use liquid flux (Kester 186) and make sure it wicks below the chip (before heating); but the fumes will mess up an oven.

 

[Jism]

Yeah that's not the intention, to make the oven useless at that point.