Been reading a bit, it seems that even 64gb is troublesome\unstable on this platform (also goes for AMD?)
I’d like to go higher than that in my next 3D\creator oriented build, 128 or even higher if possible (been seeing some BIOS updates that allegedly allow up to 196 or even 256gb on some boards?) so I can avoid threadripper build cost.
Is it a real issue and how would you go about building a max RAM-loaded workstation in 2023\2024?
Currently rocking ancient i7-5820k with 64gb ram and it feels a bit tight on the ram side when doing Houdini simulations and even some renders.
I have 192GB 4 sticks 6400mt/s, running @5200mt/s with z790 and 13600k, so 14900k should run with good memory with 192GB with 5200mt/s.
awesome, thanks 4 info!
What is your use case? Do you need the single core that the 13900k has or will OG threadripper handle you just fine? Do you need PCIe lanes? I’ve gotten stable results on 4x 48 gb sticks on AM5 but it’s been at 5200 speeds with the higher end ASRock boards.
Are you on current BIOS versions? I am on ASUS but have not seen many good reports on 192GB speeds recently. I just got 2x48GB sticks and have them running well at 6200cl32 but might consider getting 2 more if 5200 is a decent possibility.
I need both single-threaded performance but also as many cores as I can get. So 14900 is my best choice I think without breaking the bank with threadripper platform.
Are you me? I currently have am running a 5820k with 4 sticks as well and upgrading to a 14900k. I am using it primarily for music production/work and gaming, and am weighing the benefits of going with a larger amount of ram and lower ram speeds, or higher ram speeds vs smaller amount.
I’m trying to find real world data on how ram speeds actually affect performance in the programs I use. It honestly seems like a large sacrifice being bound to 2x48 sticks just to be able to OC well. I would like 128gb or more ideally.
I have found some conversations (albeit anecdotal) that high ram speeds can allow you to be able to run a lower hardware buffer rate in a recording DAW. Trying to find more evidence towards this, but that could be a compelling argument towards ram speed for people in the audio world.
I was in a similar predicament, I went with a 13900KF and a MSI Z790 Edge WiFi DDR4 board (moved from AM4) kept my 4 RAM sticks and just went with the lower (CL16 3600) speed with XMP enabled. It’s like within 1% or less than a comparable rig running DDR5 at 4800-5000 speeds. I’m not to beat up about it. Stability is greater than the speed. Plus Intel isn’t as fussy about the fast RAM as AMD was. That (RAM stability, or lack there of) was what inadvertently got me to the new platform earlier than expected anyways.
It depends on the speed? If I recall. Unless you tweak but it’s still a problem
i’m using 192gb across 4 sticks of ram on a z790 formula and the 14900k is not stable at all. Only reducing clock speed to 5.5ghz works .
well that sucks… have you tried lowering RAM speed?
yeah but it still crashes even with the lower ram speeds. The only thing that seems to work is reducing cpu clocks. Do you think this 14900k is faulty?
no idea honestly but other guys in this thread seem to be able to run it with 4 sticks at lower speeds
I tried everything to get 4 sticks to run at 4800 on my 12900k. Switched to 2 6400 and things a good.
Theres some people running Micron B rev configs in either 2x32 or 4x16 which can still do 4300-4400 G1. 4xDR compatibility is also listed at 4400 on current DDR4 mobos.
The problem is that ram went EOL like 3 years ago and you can only get it second hand now. I totally screwed up by not getting 4x16 when it went on EOL sale.
This is actually a lot better for 128 Gb than 4800 DDR5 G2.
You can use as many sticks/gb you want however it’s the frequency of those sticks is the only issue. If you get 5600 speed ram it should work no issues. Anything above that is going to depend on cpu lottery/motherboard
Aren’t know… i have a 13900k and 4x32GB of ddr5@5600 won’t post
I’m a 13900k owner running 4 sticks of ddr5. Only 64gb, so take it with a grain of salt for 128+.
My experience is that 4 sticks and high capacities are absolutely stable… until you decide you want your RAM faster than JEDEC.
I finally got my 4-sticks of DDR5 running stable at their rated 6000MT/s, but it was… a painful process. I learned more about RAM overclocking than I ever wanted to, and it involved manually adjusting voltage settings and multiple overnight MemTest passes.
Jeez, I would love 4 sticks, for the look, but everything here is convincing me to go with 2 sticks when I upgrade to ddr5.
I just setup a 192GB 14900K workstation and it was pretty straightforward to get going compared to AMD by just installing the RAM and turning on XMP and that’s it.
5400MT/s seems stable from 80 hours of SAT but from my point of view, even 5200MT/s is a huge improvement over the 3600MT/s Intel and AMD spec for max capacity configurations which is what I would get if I settled for a Dell Precision 3660.
48GB sticks are less hard on the memory controller than 32GB sticks.
With dual channel you can get up to a pretty high speed, I am going to be getting 2x 6400mhz 32gb sticks, but I’m pretty sure that anything quad channel and above starts to be limited to 5400-5600 or so, or even lower. But if you’re doing creative stuff, that might not be what you’re worried about, and you may just want a bunch of 5200mhz sticks.
Exactly that. It’s not a problem if my simulations\renders are 4% slower or whatever, but if they can’t fit in the RAM, I have a major problem.