Adding additional Memory; Bad idea to Mix same exact kit types?

[bigblueshock]

I know memory manufacturers don't advise to do this, but what's the general take on adding an additional memory kit of exact same kind to a system? I have two 8 gig sticks (16GB worth of RAM) but want to add another two 8 Gig sticks.

 

My thought process would be to buy the memory, take the original memory out, put new memory in, and test it for errors and stability. Then I would add the other two (all 4 dimms in at this point) and re-adjust clocks, timings, and voltages.

 

Is this a bad idea?

 

I'd just hate to eat the cost of the two 8 gig sticks I currently have (good rated b-die memory) and go either go two 16GB kit or four 8GB kit.

[speed_demon]

Sounds like a good way to do things to me. I run a somewhat similar Ryzen 3900X system and am planning to add another 2x8gb memory sticks later this week myself.

[axipher]

Generally adding another kit that is the same model number is fine as long as they are the same version and same timings.

 

For example, the same Corsair Vengeance Kit could be B-Die or something else based on it's version number.

 

Sometimes this is more obvious with different timings listed on the RAM stickers.

 

You can use Taiphoon Burner to check.

 

I would definitely make sure to run MemTest+ on the old kit on it's own, then the new kit on it's own, then all the sticks installed before you start using it full time. Ubuntu Desktop Images include MemTest+ on them as long as you boot from the USB stick NOT in UEFI more.

[Supercrumpet]

One or two PC building friends and I have done that with our rigs before and none of us have had any related issues, so that I'll throw in another vote to say that seems like a reasonable plan.

[ENTERPRISE]

If they are identical kits, but just two separate kits then they will be fine to use together. I have certainly never had an issue, nor would I expect to.

[486]

I do it all the time, adding a second kit later when I need more memory.

 

Its generally not suggested because the memory manufacturer, afaik, historically wouldn't guarantee all 4 sticks would function at rated specs and they likely aren't tested that way in QVL for your motherboard.

 

​​​​​​The only potential issue is the memory controller, there have been times where either more voltage or looser timings are required to run all 4 sticks at rated speed, however I haven't had that happen since DDR2 / intel core.. so I'd say go for it

[InverseTundra]

I havent had any issues on newer hardware no matter if its different speeds or timings personally. Though i would definitely recommend getting as similar a kit as possible for long term use and testing that kit solo then mixing is an even better way to verify long term reliability. Overall you shouldnt really have any issues.

[ENTERPRISE]

I tend to get kits just so it is a plug in and never think about it again scenario....literally the only reason.

[bigblueshock]

Awesome! I agree with you guy. I just wanted a second opinion

 

When I get my 2nd kit, I will take my first kit out and Membench it to make sure it's stable at the clocks I want, then I'll throw all four sticks in and tweak timings a bit.

 

Thank you ?

[NavJack27]

I've written a thing or two about memory.

https://thechipcollective.com/posts/navjack27/3950x-review-in-depth-memory-testing/

Things get confusing if you have Ryzen for sure.

Channels, ranks, interleaving... Also if you buy the same kit later there are no guarantees that the 2nd batch will use the same ICs. I've had that happen with my gskill kit. I could have more to say but I just woke up and wanted to put some sort of opinion in here if it helps.

[bonami2]

My 2600k run 1 stick of 4gb 1333 cl9 one of 8gb cl9 1600 and 2 8gb from another brand rated 1600 cl10.

 

Got to increase the imc controller voltage on the i7 my i5 did not need it. And it ran fine on auto at 1333 cl9 1.5v

 

 

Ran OCCT over 16 hour Prime 95 over 12 hour. And Asus burn Test.

 

Ran 24/7 for week long running ark server and folding

[ENTERPRISE]

My 2600k run 1 stick of 4gb 1333 cl9 one of 8gb cl9 1600 and 2 8gb from another brand rated 1600 cl10.

 

Got to increase the imc controller voltage on the i7 my i5 did not need it. And it ran fine on auto at 1333 cl9 1.5v

 

 

Ran OCCT over 16 hour Prime 95 over 12 hour. And Asus burn Test.

 

Ran 24/7 for week long running ark server and folding

 

Would be interesting to see a memory bench of the mixed modules running vs a matched kit at the same speed and timings to see what kind of impact you are looking at.

[xlen]

I know memory manufacturers don't advise to do this, but what's the general take on adding an additional memory kit of exact same kind to a system? I have two 8 gig sticks (16GB worth of RAM) but want to add another two 8 Gig sticks.

 

My thought process would be to buy the memory, take the original memory out, put new memory in, and test it for errors and stability. Then I would add the other two (all 4 dimms in at this point) and re-adjust clocks, timings, and voltages.

 

Is this a bad idea?

 

I'd just hate to eat the cost of the two 8 gig sticks I currently have (good rated b-die memory) and go either go two 16GB kit or four 8GB kit.

 

I bought an ADATA XPG Z1 16GBx1 and a while later bought another one of those with the same spec, what I ended up is with MFR and B die in my PC that doesn't clock past 2933CL14 with my first-gen Ryzen 1600.

If you're willing to take the gamble you should have no issues with it, otherwise, trade-in your current RAM and buy a new kit.

[bonami2]

 

Would be interesting to see a memory bench of the mixed modules running vs a matched kit at the same speed and timings to see what kind of impact you are looking at.

 

Am in to try it but with what program haha

[bonami2]

 

I bought an ADATA XPG Z1 16GBx1 and a while later bought another one of those with the same spec, what I ended up is with MFR and B die in my PC that doesn't clock past 2933CL14 with my first-gen Ryzen 1600.

If you're willing to take the gamble you should have no issues with it, otherwise, trade-in your current RAM and buy a new kit.

 

 

 

Yea they do that with ssd too. Change controller configuration and keep the same model number.. Kinda a scam.

[ENTERPRISE]

 

 

 

Yea they do that with ssd too. Change controller configuration and keep the same model number.. Kinda a scam.

 

It comes down to supply channels as to what components they can still get but I agree that ideally they should note the revision with a new model number.

[xlen]

 

 

 

Yea they do that with ssd too. Change controller configuration and keep the same model number.. Kinda a scam.

 

They don't do it with SSDs, and chips on RAM are determined by what's available for the most reasonable price, so you can have any chips on any series unless manufacturer specifies what they use...

[bonami2]

 

They don't do it with SSDs, and chips on RAM are determined by what's available for the most reasonable price, so you can have any chips on any series unless manufacturer specifies what they use...

 

Yes they do Kingston got hated for that for putting a slower controller on a ssd. So the review where all faster than the updated product.

 

Yea it like buying a car expecting a 2.0l and getting a 1.2l for the same price. I think it should be illegal

 

 

https://www.extremetech.com/extreme/...r-good-reviews

Edited April 27, 2020 by bonami2

[xlen]

 

Yes they do Kingston got hated for that for putting a slower controller on a ssd. So the review where all faster than the updated product.

 

Yea it like buying a car expecting a 2.0l and getting a 1.2l for the same price. I think it should be illegal

 

 

https://www.extremetech.com/extreme/...r-good-reviews

 

that's just kingston pulling a kingston...

[bigblueshock]

 

Would be interesting to see a memory bench of the mixed modules running vs a matched kit at the same speed and timings to see what kind of impact you are looking at.

 

Just wanted to follow up with the thread...

 

I got my 2nd kit of F4-3600C15D-16GTZ

 

I'd like to share what I did for Testing Purposes:

 

 

1) I pulled my original set of memory and replaced it with the new set. Same timings - 3733 Cas 14. Passed 60+ minutes with MemBench no issue.

 

2) I threw all 4 sticks in there and re-loaded Memory Defaults (2133, no idea what CAS or Voltage, but fully stock) Passed 60+ minutes with no issues.

 

Here's the fun part, I started tweaking now...

 

3) I ran a DRAM Calc for 4 sticks at 3600. Plugged everything into the BIOS. Computer booted, but got errors in less than 10 seconds in Membench.

 

Spent a couple of hours trying to adjust Timings, Voltages, you name it. Nothing worked. I then did some Googling and apparently there's NO way on my motherboard you can run four sticks at 3600+ with Gear Down mode disabled (daisiy chain topology). Unfortunately the calculator had it Disabled. I enabled it and it was good. Not satisfied at this point due to knowing there's a bit more potential.

 

4) I switched the DRAM Calc to 4 sticks at 3733. It now has Gear Down mode enabled with nice tight timings. Plugged everything in, and went to work.

 

In the end, Passed for 240 minutes with 4 sticks at 3733 Cas 14 @ 1.48v. Heat isn't an issue luckily because I have a nice G.skill RAM cooler. Even after the 4 hours, Max memory temp is 35c. I probably don't need the cooler (Would have probably been around 45c) but it doesn't hurt to have!

 

 

[xlen]

 

Just wanted to follow up with the thread...

 

I got my 2nd kit of F4-3600C15D-16GTZ

 

I'd like to share what I did for Testing Purposes:

 

 

1) I pulled my original set of memory and replaced it with the new set. Same timings - 3733 Cas 14. Passed 60+ minutes with MemBench no issue.

 

2) I threw all 4 sticks in there and re-loaded Memory Defaults (2133, no idea what CAS or Voltage, but fully stock) Passed 60+ minutes with no issues.

 

Here's the fun part, I started tweaking now...

 

3) I ran a DRAM Calc for 4 sticks at 3600. Plugged everything into the BIOS. Computer booted, but got errors in less than 10 seconds in Membench.

 

Spent a couple of hours trying to adjust Timings, Voltages, you name it. Nothing worked. I then did some Googling and apparently there's NO way on my motherboard you can run four sticks at 3600+ with Gear Down mode disabled (daisiy chain topology). Unfortunately the calculator had it Disabled. I enabled it and it was good. Not satisfied at this point due to knowing there's a bit more potential.

 

4) I switched the DRAM Calc to 4 sticks at 3733. It now has Gear Down mode enabled with nice tight timings. Plugged everything in, and went to work.

 

In the end, Passed for 240 minutes with 4 sticks at 3733 Cas 14 @ 1.48v. Heat isn't an issue luckily because I have a nice G.skill RAM cooler. Even after the 4 hours, Max memory temp is 35c. I probably don't need the cooler (Would have probably been around 45c) but it doesn't hurt to have!

 

 

God no, don't use the calculator, it's bad. The calculator doesn't calculate anything it's simply a bunch of stolen profiles...

I recommend spending a day or two on tweaking, 1)put timings ridiculously high like 20-20-20-20-42-64 and find the highest frequency that works, with this setup I'd probably stop at 3800 and then tweak the timings and test stability, change 1-2 timings between runs that way you will know what works and what doesn't, I should add a reminder that there are formulas for many timings and you should visit DDR4 OCing guide from integral, I find that it's the most complete and reliable guide and comes from somebody who actually knows what's what: https://github.com/integralfx/MemTestHelper/blob/master/DDR4%20OC%20Guide.md

[bigblueshock]

 

God no, don't use the calculator, it's bad. The calculator doesn't calculate anything it's simply a bunch of stolen profiles...

I recommend spending a day or two on tweaking, 1)put timings ridiculously high like 20-20-20-20-42-64 and find the highest frequency that works, with this setup I'd probably stop at 3800 and then tweak the timings and test stability, change 1-2 timings between runs that way you will know what works and what doesn't, I should add a reminder that there are formulas for many timings and you should visit DDR4 OCing guide from integral, I find that it's the most complete and reliable guide and comes from somebody who actually knows what's what: https://github.com/integralfx/MemTestHelper/blob/master/DDR4%20OC%20Guide.md

 

Wow great reference. Thanks! Taking a look now.

 

I don't know if my chip can hit 1900 FCLK 24/7 stable, I'd like to keep it at a 1:1 ratio. I may try 1900 FCLK and 3800 and start at high cas like you said.