is it a MYTH that mixing RAM hampers performance?
can i fill all my RAM slots with completely different RAM?
is it a MYTH that mixing RAM hampers performance?
can i fill all my RAM slots with completely different RAM?
No. It's not.
OP picture is much less impressive when you notice that all the modules on the picture are pre-1995, except for that one SDRAM module (1997).
depends on the mixing.
a given channel will operate at the slowest / lowest common denominator of all dimms.
You can mix brands though (though god only knows if XMP will work properly with mixed brands, I've never tried).
Thought I had 16gb (4gb x 4) of identical g.skill RAM in my media server, and it'd segfault/shit itself every so often and was driving me nuts. Turns out two 4gb modules had different timings and it worked--but would throw errors every so often. I removed one pair, and the server has run reliably ever since.
Most of your problems will come from stability and compatibility. If it works and stays working, you are past 99% of things that can be worried about.
only if you want to overcock it.
but ram can even run unlink now.
>avadirect.com
>2008
>almost 2008 + 10
ive read that nowadays problems are minimized
dont know if it is true
It depends. Most incompatibility comes from different latency that the CPU can't account for. Other than that, it's down to whether your motherboard supports having multiple kinds of RAM sticks, although the hardware interface should be the same.
don't do it
I bought a dual channel kit to add to my existing kit for an upgrade
They are the exact same models and they both use Samsung B-dies but bios won't even post once XMP is enabled
>can i fill all my RAM slots with completely different RAM?
In fact, you should do this. Diversity, you know.
Now this is autism
In theory, since most modern hardware has at least some NUMA aspects, where some memory is closer to some CPUs than others, it should be possible to mix fast and slow RAM having some nodes with fast, and some with slow; and let the OS figure out how to use the nodes with the slow ram as a cache for infrequently accessed data.
I have that much laying around too
I see at least two SODIMMs and a regular DIMM. Also how is a pile of two dozen RAM sticks "impressive"?
I don't have any reason to believe anything would be different nowadays. Modern mobos still default to the shittiest timings and frequency.
>is it a MYTH that mixing RAM hampers performance?
No. Mixing RAM can cause all sorts of problems from poorer performance to kernel panics.
At least when they are running in dual+ channel. If it's all single channel then there are usually no problems.
What the fuck. This is an unironic thread.
This is how LOW Sup Forums has gone.
This is even more important these days with quad channel memory.
Yes you can.
No it's not a myth.
All the sticks will run at the timings and frequency of the slowest stick you have there.
Also if the sizes are different, they will only run in single channel.
Not that autistic when you realize he's wrong though
>Also if the sizes are different, they will only run in single channel.
I think it'll go into flex mode and the lower sized RAM amount will be dual channel while the leftover amount will be single channel.
Also if the amount of RAM on each channel is the same you can still get dual channel so shit like 8 4 8 4 should work without any performance impact.
No, it isn't a myth. I used to run 2x2GB + 2x1GB DDR2, 2s timings were 5 4 4, 1s 5 5 5. I had to manually set them both to slower, and despite that I was getting micro stutters with audio on core 2 duo, and occasional bsods.
Is that dual socket phase change cooler on the pic?
>can i fill all my RAM slots with completely different RAM?
If it runs at same configuration (rated frequency and timings) it doesn't matter. But having same ram units allows you to overclock ram better because all units will have close limits.
How it's impressive at all?
By adding more dimms you multiplied the number of ranks for each channel by 2. Multiple ranks hinder overclock. That's all there is to it. There is a reason 4 stick kits have higher timings on their XMP profiles than 2 stick kits.
Yeah it is back in the day when southbridge and northbridge still exist, i got thrown a lot bsod back then.
Northbridge still exist. They are inside the CPU die now.
>By adding more dimms you multiplied the number of ranks for each channel by 2.
No shit? So is best performance achieved with 2 sticks of RAM then?
>impressive
You must be really easily impressed then.
I don't even consider my own box of memories remotely impressive.
I remember watching linus tech tips say something like not only does the ram have to be the same but you should purchase the sticks at the same time in the sense that differing mileage hampers compatibility. Is there anything to this?
Sure, if you're running a high end enterprise server.
Not really. Performance in different tasks is influenced by different factors. If lower timings give you higher score than yes, less ranks per channels allow more tight timings. Note that there are a lot of dual rank dimms and even some 4-rank dimms. There are no single rank 16GB dimms on the market.
Some reviewers indicate that some tasks benefit from multiple ranks, but afaik there is no theoretical ground there and I think it's bullshit.
ram can only perform as fast as the slowest ram, that's what they're getting at. check yoru ram speed and bus speed, just don't let it get below 55mph
Meme dimm manufacturers like g.skill/geil/corsair/team/whatever often sell dimms with different dies (can be even different manufacturers like samsung/hynix/micron/nanya/noname) under the same part number. When you buy a set at least it's guaranteed to work (but in fact not really, if you read reviews on newegg you will see that return rate for meme overclocked g.skill ram is somewhat like 5%).
> less impressive
> all the modules on the picture are pre-1995
That makes it more impressive because RAM was more expensive back then.
Old memory is pretty pricey these days too with all the retro movement.
That SODIMM is not from 1997 through or pre 1995.
Why are you so wrong?
It was mostly more expensive per byte, not per module (or per chip back when modules didn't exist). I remember sticking eight 30-pinn SIMMs into a 486 board and I was a poorfag back then.
Let's ask this then. Assume 8 available slots on the board (max is irrelevant)
>All single dimm sticks
>All same range of speeds
>All different manufacturers
Would this be worse or better than having 4 sets of 2-pair dimms
Would this be worse or better than 2 sets of 4-pair dimms (sets of 3 are retarded)
...
it's not autistic when it's wrong
the point of being autistic is to be factual
No BIOS cant run different ram at the same time, you need to put at least the same frequenxy and timings. I did the mistake of putting different timings and now i can only use my pc if BIOS skips memery modules and lets the CPU handle them so my custom OC to 3,6 GHz is gone. ;_;
define hampers
each ram module has a tiny ROM which then BIOS reads and tunes memory parameters during the memory initialization that they are compatible across all modules in your system. It's true that if one module can support faster rate and other can't, system will make the faster module to work at the same rate as the slowest one. Usually it's not that huge difference.
If you're a Linux, oh sorry, GNU/Linux fag that only rices his desktop, repairs X.org and shitposts then stick in what shitty memory you can find in the dumpster. It won't matter.
If you actually do productive tasks and enjoy interactive entertainment as computer video games, stick to one type of RAM.