• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.
  • The forums have been upgraded with support for dark mode. By default it will follow the setting on your system/browser. You may override it by scrolling to the end of the page and clicking the gears icon.

Adding 2 new Ram Sticks, stability

Exactly.
I love Gskill they just have the best RAM I have ever had even the RipJaws DDR3 is AMAZING!
Hi,
They have crappy memory too so don't think they don't lol
 
Hi,
They have crappy memory too so don't think they don't lol
YEAH? That is news to me I still have yet to find a PAIR that is as bad as the Corsair! LOL.
4 Systems and 3 with Gskill you do the math!
 
YEAH? That is news to me I still have yet to find a PAIR that is as bad as the Corsair! LOL.
4 Systems and 3 with Gskill you do the math!
Hi,
Hynix is hynix no matter the brand and it's hit and miss oc wise rated speed well I don't count that
Problem with corsair they mix up hynix and micron this is why some are okay and some aren't.

I guess I have to give corsair a little credit they do have ver.# to at least give you a clue
If the ver.# match you got a good chance of working if mixing kits.

I've had my share of good and terrible from corsair.
 
I like G.Skill too. Been using them since the DDR2 days, though not entirely exclusively. I do have to say I am really impressed with my Royal 3200C14 sticks. A little annoyed they wont run 3600 12-12-12- stable with 1.6v but cant win em all I reckon. I've had good ones and average one from them, you usually get what you pay for..
 
Would it matter ? For this problem?
Possibly. It would be worth trying 1.4V - all DDR4 should be able to tolerate it without degradation or damage, and extra voltage often helps with memory stability. It doesn't always help (my memory won't go above 3000MH even with 1.5V) but there's no real risk to trying it unless you go above 1.5V (depending on memory die type). Some hardcore overclockers go all the way up to 2V with Samsung b-die, which tolerates voltage very well, as long as you can cool it.

It may also be worth considering increasing your SoC voltage slightly. (though be careful, as the SoC won't tolerate voltages anywhere near what the RAM will) Having 4 memory sticks at 3600MHz is close to the limit of what Zen 2 memory controllers can handle at stock settings, and may be contributing to the stability problems.
 
Hi,
Corsair only way to bin those is the ver.# on each stick
Their terrible about using the cheapest parts at the time of manufacturing for products never know what you get until you look at the sticks.
Yes but i can't open the package before buying, unfortunatly!

I will try to order another one and match the 4 of the 6 sticks i have and hope for 4*8GB 3600Mhz CL18 would be ok...

Possibly. It would be worth trying 1.4V - all DDR4 should be able to tolerate it without degradation or damage, and extra voltage often helps with memory stability. It doesn't always help (my memory won't go above 3000MH even with 1.5V) but there's no real risk to trying it unless you go above 1.5V (depending on memory die type). Some hardcore overclockers go all the way up to 2V with Samsung b-die, which tolerates voltage very well, as long as you can cool it.

It may also be worth considering increasing your SoC voltage slightly. (though be careful, as the SoC won't tolerate voltages anywhere near what the RAM will) Having 4 memory sticks at 3600MHz is close to the limit of what Zen 2 memory controllers can handle at stock settings, and may be contributing to the stability problems.
my memory wont go pass 38/39 degrees celcius at full load is it alot for ram ?
 
Yes but i can't open the package before buying, unfortunatly!

I will try to order another one and match the 4 of the 6 sticks i have and hope for 4*8GB 3600Mhz CL18 would be ok...


my memory wont go pass 38/39 degrees celcius at full load is it alot for ram ?
Hi,
I'd hit the used market here/ ebay/....
 
Hi,
Corsair only way to bin those is the ver.# on each stick
Their terrible about using the cheapest parts at the time of manufacturing for products never know what you get until you look at the sticks.
Man, when will it ever stop....!? Or when ppl will realize finally Corsair is not a premium name when it comes to ram, like Adata total grab bag.
 
Man, when will it ever stop....!? Or when ppl will realize finally Corsair is not a premium name when it comes to ram, like Adata total grab bag.
Hi,
Well when ddr4 first came out corsair was pretty good very expensive but so was everyone else
Sadly they went for cheapest parts and now is mostly crap.
 
my memory wont go pass 38/39 degrees celcius at full load is it alot for ram ?
No, 38 degrees is quite low for RAM temperature - ambient temperature within a PC case is usually around that. I'm actually surprised it's so low, and (unless your room is quite cool and your case has a lot of airflow) I would suspect that the temperature isn't being detected properly, or that your "full load" isn't stressing the memory. If that's a real value, you should be pleased that your PC is well cooled.

For testing memory temperatures, it's sometimes worth running a GPU benchmark at the same time - because GPUs put out a lot more more heat than RAM, they can significantly raise the temperature of the air in your case, and indirectly increase the memory temperature. This is likely to be a more realistic "worst case scenario" that's more representative of situations like gaming, rather than only loading the CPU.

Adding voltage will result in the memory putting out more heat, but 39 degrees is well within the range you should expect for good stability and reliability. There is a decent amount of variation between memory die types, but you shouldn't see any major drop-off in stability from temperature unless your RAM is consistently around 45C or more. Lower temperatures are always better, and there is a trade-off between directly improving stability and indirectly harming it, and the ideal point of that trade-off is different for different memory modules, depending on the dies and heatsink.

Because of this, it's often difficult to predict how a given memory module will react to increasing or decreasing voltage. You can try either and see if it provides any useful benefit. Memory OC is a lot of trial and error.
 
just because your buying gskill doesnt mean anything.
2 kits wont boot, while 3 corsair kits did (all had micron dies), is enough proof for me that i wont believe ppls ram recommendation except for specific kits (sticker)..
and 2 of 5 corsair kits i ordered (to replace the gskill) had samsung b-die, which is known to be the crappy quality ram corsair makes, lol
 
just because your buying gskill doesnt mean anything.
Well.. it does mean one thing..

You didn't buy Corsair :D

I had a good set of Dominators once.

I still have them, pretty sure they don't work though.
 
just because your buying gskill doesnt mean anything.
2 kits wont boot, while 3 corsair kits did (all had micron dies), is enough proof for me that i wont believe ppls ram recommendation except for specific kits (sticker)..
and 2 of 5 corsair kits i ordered (to replace the gskill) had samsung b-die, which is known to be the crappy quality ram corsair makes, lol
I have a Corsair kit of CMW32GX4M2C3200C16 (ver 4.32, dual rank, 2 sticks, 32GB) internally it says b-die (which seemed unlikely) but found some posts that suggested it was actually c-die and with that die lowering voltage actually allowed it to OC better. After struggling to get it past DDR4-3600 I tested the theory and was able to OC that kit to DDR4-3800 on 1.33v and FCLK 1900 (keeping 1:1 ratio) on my 3800x. After some tweaking the timings were not earth shattering but for the most part followed the kits XMP rated timings in nanoseconds at the higher frequency.
 
just because your buying gskill doesnt mean anything.
2 kits wont boot, while 3 corsair kits did (all had micron dies), is enough proof for me that i wont believe ppls ram recommendation except for specific kits (sticker)..
and 2 of 5 corsair kits i ordered (to replace the gskill) had samsung b-die, which is known to be the crappy quality ram corsair makes, lol

Except for the fact that Corsair revision numbers don't even indicate what you get, they lump multiple ICs under the same revision codes now so you literally don't know unless you peel off the heatspreaders (and even then it depends on if Corsair erased the markings, so more like unless you mess with VDIMM and tRCD/tRP/tRFC you just won't know).

G.Skill has a 042 code for a reason on their stickers - unlike Corsair it actually clearly tells you what you have. IC density, vendor, vendor revision. That's all the info you need to know what to do with VDIMM and timings behaviour.
 
Anyone recommend looking at SoC voltage yet? Often, especially with four sticks, the memory controller needs a small bump in voltage.
 
Except for the fact that Corsair revision numbers don't even indicate what you get, they lump multiple ICs under the same revision codes now so you literally don't know unless you peel off the heatspreaders.

G.Skill has a 042 code for a reason on their stickers - unlike Corsair it actually clearly tells you what you have. IC density, vendor, vendor revision.
It was bad enough resellers don't even list the revision number so now the revision number is useless to identify the ICs. Ouch. That hurts.

Anyone recommend looking at SoC voltage yet? Often, especially with four sticks, the memory controller needs a small bump in voltage.
There was a documented sweetspot in the Ryzen DRAM overclocking guide. I don't recall what that was off hand but when exploring I tend to set it to 1.1v and work my way down.
 
I didn't built a computer for 10 year, just lame the state where ram is... also rgb... what maddness is this every new brand doing something does it different..
 
Pretty much all manufacturers do this. Most only guarantee stock speeds and timings for a reason. They do it to save money. Corsair bins their ics very tightly to maximize the amount of segments they can use their ics in. This obviously allows them to stretch profits. It also means most of their lower tiered ics dont clock well (b-die included). This makes them seem greedy when compared to a company like Gskill who generally leave more meat on the bone for ocing no matter the tier.
Also keep in mind not all b-die oc well. You have to consider where they come from. So even if you happen to find a set of your original b-die... Well, youd better do some testing of the set you have before you bother with that.
If your not familiar with memory ocing or dont plan on learning how to with what you have. Your far better off buying a 4x8gb kit of the specific speed you want and calling it a day.
 
@freeagent
its more about the fact that "our" experience is statistically irrelevant, and not proof of anything.
even if someone works as a tech and installs 20 kits/week, thats just ~1000 kits/year, vs millions being sold planet wide (in total).
so telling ppl not to buy a certain brand because they worked with a handful of items is ...

@tabascosauz
true, if your able to look at them in a store. online, not so much difference as no one show you the labels..
 
@freeagent
its more about the fact that "our" experience is statistically irrelevant, and not proof of anything.
even if someone works as a tech and installs 20 kits/week, thats just ~1000 kits/year, vs millions being sold planet wide (in total).
so telling ppl not to buy a certain brand because they worked with a handful of items is ...

@tabascosauz
true, if your able to look at them in a store. online, not so much difference as no one show you the labels..
I don't have the numbers like that, I wasn't arguing at all., just being a bit of a smart ass I guess.. sorry! I would not count out your experience as irrelevant because I use everyone's feedback in my consideration for purchases. I don't know everything, that's why I come to you folks :)
 
@maxfly
at least for ryzen, your still better off with 2 sticks of SR (clock/compatibility wise)..

@freeagent
lol. sorry, didnt want to make it sound like you were, and i do like to be a smart ass and argue as well :D

just saying, as long as its not info that goes into the thousands (of kits used), its not representative for the whole brand.
i get it when ppl say "i dont recommend/get something different" etc, but saying a brand is crap just because not all their kits are tier one b-die, or arent the best ocers (when was that ever guaranteed?), or that it takes some research to find the proper sticks, all based on a "couple of kits" ppl used, is just ignorant.

example
fill a box with 100 bills (90x 10$ and 10x 100$) and you ask 2 ppl to pull 10 pieces (without them able to see whats in the box)
lets say by chance the first person grabs all the100s, the next one grabs all 10s.
if asked what's in the box, the first would say "its all hundreds", the 2nd would say "its all 10s", when in reality it was neither,
but their sample size is too low to make a proper assumption...
 
Last edited:
Ok guys so just got another 2 stick of 8GB they are samsung..

Tested them all ok

Then joined all 4 sticks i have Corsair 8GB*4 3600Mhz CL18 with XMP speed ok latency ok, windows memory test which always failed passed with 0 erros or pages missing.


What can I do to improve now ?
First step ? and how?

I already read pages and pages of this but is kinda theorical i need to know how and where i need to change the stuff to improve most?


Maybe try CL16 ? But how ?

Thanks for any kind soul!
 
Ok guys so just got another 2 stick of 8GB they are samsung..

Tested them all ok

Then joined all 4 sticks i have Corsair 8GB*4 3600Mhz CL18 with XMP speed ok latency ok, windows memory test which always failed passed with 0 erros or pages missing.


What can I do to improve now ?
First step ? and how?

I already read pages and pages of this but is kinda theorical i need to know how and where i need to change the stuff to improve most?


Maybe try CL16 ? But how ?

Thanks for any kind soul!

Here are two resources to start if your interested in overclocking your ram however you might consider if you want to spend hours and hours enjoying your new PC "as-is" or hours and hours trying to overclock it.

AMD Ryzen Memory Tweaking & Overclocking Guide

DRAM Calculator for Ryzen

I use https://www.memtest86.com/ to test but there are other programs out there to test as well.

Are all your sticks ver 4.32? (look at the sticker on the back of the ram next to the model number)
 
Last edited:
Here are two resources to start if your interested in overclocking your ram however you might consider if you want to spend hours and hours enjoying your new PC "as-is" or hours and hours trying to overclock it.

AMD Ryzen Memory Tweaking & Overclocking Guide

DRAM Calculator for Ryzen

I use https://www.memtest86.com/ to test but there are other programs out there to test as well.

Are all your sticks ver 4.32? (look at the sticker on the back of the ram next to the model number)
Thank you, i will study this.
Going from CL 18 to CL 16 is also overclocking?
I was reading and seems that i would benefit more from latency reduction that Mhz increase, cause of infinity fabric 3600mhz is the sweets spot?

About the calculator, does it work properly for ryzen 5000 ? I read other guides and they say only goes up to 3000 the "auto-calculators"

:D
 
Thank you, i will study this.
Going from CL 18 to CL 16 is also overclocking?
I was reading and seems that i would benefit more from latency reduction that Mhz increase, cause of infinity fabric 3600mhz is the sweets spot?

About the calculator, does it work properly for ryzen 5000 ? I read other guides and they say only goes up to 3000 the "auto-calculators"

:D
Since memory controller for Zen3 is the same or better than Zen2, the DRAM Zen calc will help you up to a point so why not to give a go?
 
Back
Top