Originally Posted by eyecrave
Actually im using 2 x 16gb sticks specifically these ones https://www.gskill.com/product/165/1...5V32GB-(2x16GB
) as they recently went on sale at newegg. I was actually really debating getting the trident z neo 3600 c16 64gb kit but that would cost me almost as much as my cpu mobo so decided against it. Like i said before i'm really new to overclocking ram. I will try some timings with 3600 and 3733 to see where i could get them but i'm pretty happy where i am now. I honestly was going to stick with 3733 cl16 but knowing i could push the IFclk to 1900 mhz i wanted to stick to 3800 and tune them as much as i can and i posted the results. My temps on ram can get pretty hot though specially when testing i got up to 48c but no errors so i was thinking of maybe adding another fan to blow over the ram. How are your temps with the corsair aio? I have a h1151 pro still in shrink wrap from an rma but decided to install the nhd15s instead thinking that i will get better airflow on my motherboard. Your rig looks pretty good and would be a bad look if you went blue for an amd rig.
I just wanted to point out to you first, that in the last post you made showing your AIDA64 memory bandwidth along with HWINFO64, when you were at 3800MHz CAS16 - https://www.overclock.net/forum/28119718-post791.html
I missed this last night as I was posting on my phone, but, you may want to check on that if you haven't already and run MemTest or something. Your memory write is horrible and close to half what it should be (says 30GB/sec). It should be much closer to 55GB/sec. If you guys didn't know, memory write is actually more important than any RAM statistic on that test aside from latency- the CPU will read a lot of stuff from RAM once it's been written in, and it will also copy it around, and these are almost always faster. However, writing to memory usually takes more time and slows the whole process down. It sounds like you said you backed off since (to 3733 CAS16), but I'd urge you to take a look at your memory write speeds, because even if the rest of your RAM bandwidth is high or higher, and the latency is lower, the system will be slow as a dog and possibly corrupt data if there is something wrong with your memory write speeds or it's throwing errors.
The Trident Z Neo isn't anything special aside from the RGB LEDs as far as I can tell, so it's good you didn't waste your money on it. Those stats (3600 C16) aren't that great, even for 64GB (why do you need this much memory?) This is the kit I got
, color or RGB doesn't matter to me, speed and timings do. And they do 3600 CAS14 no sweat with some tweaking and voltage. I also only paid $119 for them I believe. Tbh you'd be best off buying a different kit that is single rank and 2x8GB (like 20nm Samsung B-die or 18nm Hynix) and really researching it first at some of the links I posted, if you want to get stable at very high RAM and IF clocks. As for me, I did more messing around with the system last night and my 3900X absolutely refuses to POST at anything over Fclk 1800MHz. Even 1833 is a no go and I've tried both Safe and Fast presets from Ryzen DRAM calc, I've tried various combinations of SoC voltage (not exceeding 1.125v), VDDP, VPP, PLL, VTT, CLDO VDDP, CLDO VDDG and at a bunch of higher frequencies and timings.
That seems to be the best score I've pulled so far but given that it seems competitive or better than many other screenshots I've seen of AIDA64 (with Memclk above 3666 and Fclk above 1833) I'll take it. It might be because I have the R9 3900X, or it might be something I'm missing but I can't get past 1800MHz Fclk...
Anyway, as far as the H115i goes, Corsair/Asetek makes SO many different coolers that it's sort of ludicrious to even try and keep up. I was actually going to purchase a new one (the H100i Pro...) until I read and saw that my older H100i V2 is pretty much one of the best they've ever made and has a really high powered (but failure prone) pump. Well mine has been running for years on end and is fine and apparently cools better than the more recent ones. Though, I have some pretty unheard-of (and cheap!) fans on it, the Arctic Bionix F120 that go to 2000rpm and push above 90CFM. I also have similar 140mm fans in the front of my case, and the rear. I would definitely suggest not using the stock fans if you decide to use the H115. Invest in some quality, high CFM/static pressure fans and use a fairly aggressive fan profile on your board (I would set it manually in the bios, and connect the headers directly to your motherboard, and run them at 40% minimum always). Also I'm not sure what thermal paste you use but my recommendation would be Prolimatech PK-3 Nano Aluminum, a thermal paste no one's ever heard of, but is very similar performing to Arctic MX-4, Gelid GC Extreme, IC-7 Diamond or Thermal Grizzly Kryonaut for a lot less money. See: https://www.tomshardware.com/reviews...on,5108-8.html
(PK-3 is up there with the best and actually quite a bit better than Arctic MX-4. It's also $30 for a giant 30 gram tube
whereas a single 1 gram tube of Kryonaut is $10....) I just bought a fresh tube of PK-3 as the previous one I had lasted like 6 years and did 100+ applications on various things and still hadn't run out, but it had dried out a bit. I got the 30 gram tube from Amazon...
The reason I'm talking about fans and TIM is just that yeah, I had heat issues causing thermal shutoff originally when I put my 3900X + board + RAM in and started overclocking it. I was trying for a manual 4450/4425/4275/4250 overclock with 1.375v though using Shamino's beta bios that allows for per-CCX overclocking through the BIOS. Doing it this way, the cores are basically locked at that specific speed and you get vastly improved performance compared to letting the chip boost and jump all over the place on it's own. Depending on ambient temps, my rig was shutting off under these tests but they are fairly extreme. I settled on a stable 4400/4400/4200/4200 OC w/ 1.36v and it wasn't overheating and shutting off and completed Cinebench at only slightly less of a score, and CPU-Z bench showed my single core speed was 545 instead of 555... fine. Anyway, it was still running around 77-80C under load on the cores (as per the temp readout on the 7-segment POST code display), and I'm guessing the cores were around 90C. No good. Well, when I mounted the block I had used the old (6 or 7 year old, partially dried out) tube of PK-3 Nano TIM I've had forever. I remounted the block and realized I had used too much- so I used less, and used it from a brand new tube. With all the settings exactly the same, my temps were much better- the delta between my core temps and socket temp was more like 5C difference instead of 15C, and it ran cooler overall. So now, with 72F ambient (22.2C) temperatures I'm seeing temperatures around 73C core and 82C socket with the same workload (Cinebench R20) at the same OC on the CPU and RAM. This is also after turning the machine on and letting it idle in Windows for around 20 minutes to warm up, allow the coolant in the CLC to heat up, and let the idle temperatures stabilize.
I would say that unless your temperatures are causing issues like mine (e.g. thermal shutdown/exceeding 95C) that it probably isn't worth switching your current cooler for it- unless you already have some pretty great fans laying around and a solid TIM (Arctic MX-4 or any of the others I listed, see the Tom's Hardware link). You might only really gain 3C better temps or something and it also depends on what chip you have.
I use an AIO/CLC because I have two 1080tis that can pull up to 350W each overclocked (depending on workload- tested with a Kill-a-Watt meter), get above 70C on 100% fan, and exhaust all of that hot air up onto the CPU socket area. Heat rises. With the rad in the top and the fans on it pulling hot air up and out of the case, while also cooling the CPU, it's a much better setup than having a giant hunk of metal (that will absorb that 700W of heat from the 1080ti's) in the way. Also, I set up my front and rear fans to form basically a wind tunnel going over the socket and the top card- the front fan goes to 3k RPM and the rear goes to 2k- it works well for me. (I am yet to test something like Fire Strike Ultra combined test looped with this CPU overclocked and my 1080tis overclocked and see if it's adequate, but realistically, no game I play is going to hammer my CPU or my GPUs like that. I'd honestly be surprised if my cooling is good enough to handle that specific test without the CPU overheating, unless I take my side panel and top and front fan filters off.)