OK to mix and match RAM?

beany_bot · 7 Dec 2015 at 00:00

HI, I currently have a MSI Z97 PC Mate with 4690K

I currently have 2x TeamGroup Xtreem LV 8GB (2x4GB) DDR3 PC3-21300C11 2666MHz Dual Channel Kit in slots 2 and 4.

Will I be OK putting 2x HyperX Savage Red 8GB (2x4GB) DDR3 PC3-19200C11 2400MHz Dual Channel Kit (HX324C11SRK2/8) in slots 1 and 3.

I realise I will slow all my RAM to 2400mhz but I am fine with that. Just as long as everything works as it should.

Many thanks as always.

EGuitarStar · 7 Dec 2015 at 00:08

You should be fine as long as you ain't overclocking the RAM.

beany_bot · 7 Dec 2015 at 00:09

EGuitarStar said:
You should be fine as long as you ain't overclocking the RAM.

no nothing like that. Thanks, good to know.

Limpnoodle · 7 Dec 2015 at 00:22

Yeah motherboard bios will set the setting's for you

joeyjojo · 7 Dec 2015 at 07:48

AFAIK you can only use one set of timings. The 2400 probably have slightly tighter timings than the 2666, so you need to experiment which work (best). Worst performance but best compatibility would probably come from using the slower speed (2400) and the looser timings (whichever they are, probably 2666). Then I'd try the slower speed and tighter timings, then overclocking the 2400 to 2666 (best performance). Memtest would be useful.

beany_bot · 7 Dec 2015 at 08:47

joeyjojo said:
AFAIK you can only use one set of timings. The 2400 probably have slightly tighter timings than the 2666, so you need to experiment which work (best). Worst performance but best compatibility would probably come from using the slower speed (2400) and the looser timings (whichever they are, probably 2666). Then I'd try the slower speed and tighter timings, then overclocking the 2400 to 2666 (best performance). Memtest would be useful.

I have literally no idea what any of that means I am afraid. I thought RAM was "plug n play"? I've just bought it anyway. I'll look into this memtest and try and swat up. Thanks.

Devrij · 7 Dec 2015 at 09:01

Basically the ram runs at a clockspeed of either 2400mhz or 2666 if the slower kit will overclock to match the faster one, but the timings of the ram (refresh etc) are also relevant. That's the C11 bit in the ram you mentioned. There are quite a few different timings, but the lower the number, the fewer clock cycles it takes to perform that action. So you can increase performance by either speeding up the number of cycles per second (clockspeed) or reducing the number of cycles for each action (timings) or both. Typically faster clockspeed ram runs at looser timings to maintain stability so he was saying to see whether you can get away with the tighter timings of the slower ram, or overclock the slower ram to 2666, or even both. With them both being c11 they might even be running the same timings.

beany_bot · 7 Dec 2015 at 09:19

Devrij said:
Basically the ram runs at a clockspeed of either 2400mhz or 2666 if the slower kit will overclock to match the faster one, but the timings of the ram (refresh etc) are also relevant. That's the C11 bit in the ram you mentioned. There are quite a few different timings, but the lower the number, the fewer clock cycles it takes to perform that action. So you can increase performance by either speeding up the number of cycles per second (clockspeed) or reducing the number of cycles for each action (timings) or both. Typically faster clockspeed ram runs at looser timings to maintain stability so he was saying to see whether you can get away with the tighter timings of the slower ram, or overclock the slower ram to 2666, or even both. With them both being c11 they might even be running the same timings.

OK I think I understand. In a nutshell though if I am happy running both RAMS at the speed of the slower pair then I can just Plug n Play yeah?

Devrij · 7 Dec 2015 at 21:58

beany_bot said:
OK I think I understand. In a nutshell though if I am happy running both RAMS at the speed of the slower pair then I can just Plug n Play yeah?

Yeah absolutely. The faster ram should be fine at the lower speed even if the timings are slightly tighter.

joeyjojo · 8 Dec 2015 at 09:53

Basically no, they're not plug n play. There's no way of knowing which timings the motherboard will pick when presented with different sticks. If you're lucky all the sticks will work fine at whatever settings it ends up at. If you aren't you'll get memory errors. This is why you should always match identical sticks.

If you want to mix sticks, at the very least make sure the timings match the weaker sticks (the Kingston I think, although I can't find the exact TeamGroup specs) and do several hours of memtest. As it says below start with 11-13-14 (CL-tRCD-tRP). tRAS isn't listed but I imagine 35-38 would be OK. Then I'd recommend burning a copy of UBCD to CD (or memory stick) and running memtest for several passes to test them.

http://www.kingston.com/dataSheets/HX324C11SRK2_8.pdf

If you aren't interested in learning about memory clocking then I suggest you return one set of memory and buy a matching set for the other.

beany_bot · 8 Dec 2015 at 09:57

joeyjojo said:
Basically no, they're not plug n play. There's no way of knowing which timings the motherboard will pick when presented with different sticks. If you're lucky all the sticks will work fine at whatever settings it ends up at. If you aren't you'll get memory errors. This is why you should always match identical sticks.

If you want to mix sticks, at the very least make sure the timings match the weaker sticks (the Kingston I think, although I can't find the exact TeamGroup specs) and do several hours of memtest. As it says below start with 11-13-14 (CL-tRCD-tRP). tRAS isn't listed but I imagine 35-38 would be OK. Then I'd recommend burning a copy of UBCD to CD (or memory stick) and running memtest for several passes to test them.

http://www.kingston.com/dataSheets/HX324C11SRK2_8.pdf

If you aren't interested in learning about memory clocking then I suggest you return one set of memory and buy a matching set for the other.

I can't buy a matching set for the other because my teamgroup ram is high profile and I need low profile for slots 1 and 3 because of my CPU cooler. So that is out of the question.

P.S. I looked into this memtest but it seems to be 32bit only? unless I am missing something? Couldn't find a 64bit version

joeyjojo · 8 Dec 2015 at 11:17

beany_bot said:
P.S. I looked into this memtest but it seems to be 32bit only? unless I am missing something? Couldn't find a 64bit version

Not sure which version you're thinking of but the versions of MemTest86 and MemTest86+ on the UBCD are both 64 bit.

Note they have slightly different features: the former is a newer version (May 2014 vs Sep 2013) but the latter has some missing features e.g. XMP support.

I've changed my mind actually and would suggest just getting the latest free MemTest86, it's more modern.

Devrij · 8 Dec 2015 at 11:20

You can always check the timings for the two sets of RAM and set them manually in the bios. While joey is correct, it's unlikely the faster ram has tighter timings than the slower stuff you have, so I'd be surprised if it isn't stable after just dropping the clocks down to match the slower pair. It is worth running memtest once you are all set up just to make sure, as unstable RAM can really cause a lot of grief.

Memtest runs in DOS so don't worry about that

beany_bot · 8 Dec 2015 at 11:40

Thank you both. Downloading MemTest now. Will test it when I get home and install new RAM. Good to be sure as you say.

joeyjojo · 8 Dec 2015 at 12:32

Devrij said:
While joey is correct, it's unlikely the faster ram has tighter timings than the slower stuff you have, so I'd be surprised if it isn't stable after just dropping the clocks down to match the slower pair.

Surprisingly the TeamGroup stuff probably does have tighter timings even though it's faster. The only product OcUK have that's similar is 11-13-13-35 @ 2666, vs the Kingston 11-13-14-? @ 2400.

But as you say that's actually good news for beany_bot because it means the TeamGroup stuff will be slacker and slower if it's made to match the Kingston stuff.

Let us know how it goes beany_bot. :cool:

beany_bot · 8 Dec 2015 at 15:38

joeyjojo said:
Surprisingly the TeamGroup stuff probably does have tighter timings even though it's faster. The only product OcUK have that's similar is 11-13-13-35 @ 2666, vs the Kingston 11-13-14-? @ 2400.

But as you say that's actually good news for beany_bot because it means the TeamGroup stuff will be slacker and slower if it's made to match the Kingston stuff.

Let us know how it goes beany_bot.

OK will do. Sadly the RAM didn't arrive today. I ordered it yesterday morning first class, I thought it would be here today but it isn't.

I'm not in work till Monday either so won't get it till then. Ahh well. I have memtest ready to go. anyway.

arknor · 8 Dec 2015 at 16:02

Basically no, they're not plug n play. There's no way of knowing which timings the motherboard will pick when presented with different sticks

er they always fall back to the lowest speed ram settings?
I've never had a problem with miss matched ram in dual channel

beany_bot · 9 Dec 2015 at 21:09

joeyjojo said:
Surprisingly the TeamGroup stuff probably does have tighter timings even though it's faster. The only product OcUK have that's similar is 11-13-13-35 @ 2666, vs the Kingston 11-13-14-? @ 2400.

But as you say that's actually good news for beany_bot because it means the TeamGroup stuff will be slacker and slower if it's made to match the Kingston stuff.

Let us know how it goes beany_bot.

All is good!, plugged it in and it went to windows fine, restarted and ran memtest and it was fine, no errors on 7 passes. Chrome with multitabs opening much faster when I am in game too. Very happy. Thank you all for advice. CPU cooler going on tomorrow.