Wow! Is ddr5 very overclockable or did I just get lucky?

[ir_cow]

tRAS alone isn't going to make a big difference. In fact besides submitting HWBot scores, I rarely mess with it. Lots of other things are adding to the latency. tRAS is just one part of many.

"tRAS (row active time) is a term that is used in relation to dynamic random-access memory (DRAM). It refers to the minimum amount of time that a row of memory cells must remain activated before it can be deactivated and a new row can be activated."

 

[Dr. Dro]

Interesting info on tRAS, I'll try to set it to the sum of those three values and see how my system behaves. I was told that low means better on that one.

tREFI is the biggest single gain you can have but it makes the memory extra prone to corruption if it is too hot.

1.5v has been OK with my Trident Z5s, even without thermal pad on the pmic and without a fan all stress tests at my settings completed with a peak reported SPD hub temperature of 63C after several hours of constantly hammering the RAM

 

[ir_cow]

1.5v has been OK with my Trident Z5s, even without thermal pad on the pmic and without a fan all stress tests at my settings completed with a peak reported SPD hub temperature of 63C after several hours of constantly hammering the RAM

I'm more worried about part longevity. I haven't found much info on the PMIC when looking for the datasheets so far.

During my search I did find a unknown brand that makes PMIC for (someone?) and found that specific part had a maximum operational temperature of 90 for 120 minutes. One thing to note was the entire datasheet was in Chinese which was google translated.

Still haven't found the Richtek one yet. That one is the most common in DDR5 I've seen. We may eventually find out it's like 110c for 5000 hrs. That would translate to thermalpad is pointless. Lot of these buck converters found motherboards are rated to a maximum of 90 up to 110c. Also efficiency goes down the closer you get to the maximum temperature and can turn into like a runaway effect. At least with a thermal pad I know
generally speaking neither the PMIC or buck will never going to go above 60c. Thus extending the parts lifespan and avoiding any possible chance of thermal runaway scenario.

In my reviews I call out manufacturers for not putting a pad on there because the cost is probably just a few pennies. But at the same time most of them are offering lifetime warranties so really if the part fails you get a replacement either way. I think for DRAM as their own pad is always preferred. At least until we get the data sheets and maybe Igor does more testing to determine how impactful a themralpad really is.

 

[Dr. Dro]

Yeah man that makes sense, but so far I don't think I've had any cause for alarm. Seems solid, but as with any brand new tech, being one of the earlier adopters has this kind of setback.

It's fine, anyway, I got this 32 GB kit with room to experiment and have some fun with. Eventually, I may end up getting a slower/larger kit, even if that costs me a few fps, I enjoy a stable, predictable system

 

[ir_cow]

Just wanted to point out that I'm talking about two different things the PMIC / buck converter and the DRAM IC have different maximum operating temperatures. The DRAM spec sheets are easily available and they will list the maximum operating temperature of 85 c 4 1.1 volt at 4800.

The higher going frequency the higher you go in temperature the colder you need to be to maintain stability for the DRAM ICs. There is definitely some marketing shenanigans when it comes to DRM heat sinks and keeping the memory cold when reality most users are not overclocking their memory or going into the 1.4 + range.

 

[wolar]

Hynix M can reach higher, it just depends on CPU/motherboard for that part.
Mine are sitting easily at 7200-34-44-44-28 with 1.45v and secondary/tertiary timing as tight as they can get.
Saw others push above 7600 with same setup, just luckier on the ram/cpu.
Tbh above a point I wouldn't bother with it, you reach diminishing returns and potentially lower lifespan, pushing something that has 0 real world impact.

 

[WhenMusicAttacks]

Kingston seem to use only Hynix and lately also just A Die. I have been getting only A in 5600cl36 kits in the past 2 months. On the other hand, if you buy 5200 and get micron shit, you can be lucky to get 5400 working. It's not "ddr5" but "hynix ddr5" that is amazing.