And throw in stock CPU coolers or any CPU air coolers not in push-pull horizontal direction, which, all what they do is pumping hot air down around CPU cooler.
@@sidewinder86ify Not enough. I know, my Asus x570-I Strix is almost being cooked by stock cooler on 3900x paired with 250Watt GPU (in my case 1080ti) under load. Under Linux wifi adapter starts to even throttle when it reaches 80, 85 degrees due to amassed heat around. There must be more than 70 degrees inside the case. I was planning bigger twin 14cm push-pull horizontal cooler anyway, just wanted to see how stock cooler fares, and perhaps it would better with different motherboard, different case (Fractal Nano S) and different GPU. And not under load... Stock cooler pushes air down the motherboard, before it gets out, it hits everything around: nvme, memory, chipset on mini-itx.
My favourite story about this is people on reddit buying a Gigabyte ITX board where they complained that their nvme drive was horribly slow. Well, the nmve drive goes over the chipset. And the video card sits right on top of that. It's a damn furnace. And some people were putting 3900x's in these things where the vrm's had very weak cooling. I dunno. I just laughed.
@@ActuallyHardcoreOverclocking That too. Although I think the same could be said about horizontal mobo mount cases, if heat convection is enough to drive the heat away from the memory modules.
Your channel has been above and beyond other resources online. You've gotten yourself several likes and a sub from me. Thank you very much for putting the effort for putting detailed and helpful information all in one place.
Buildzoid be like "I pass memtest but then I start Blender render and crash" Meanwhile in "me" camp - "I pass memtest and bluescreen the moment Windows boots".
Thank you so much for discussing this! For over a year I could not figure out why my system would crash while gaming but be stable any other time. After watching this video, I started collecting sensor data from all components, including RAM, looked for ways to reproduce the crashing, and looked for correlations between time of crash and sensor data. From your video I focused my attention on RAM temps. Turns out that once my RAM hits 60C, it'll BSOD or go black screen. Knowing this, I could reproduce the crashes and found that simply placing a fan in front of ny RAM stopped the crashes and be stable up to 55C. I always thought RAM could work up to 85C based on things I read on forums. I think there is a knowledge gap out among most PC builders where most of them only focus only on GPU or CPU temperatures and troubleshoot by RMAing those, their PSU, or Motherboards. Or, they will RMA their RAM only to get the same crashes again, back to square one and think there is a weird compatibility issue with their components when the truth is the only problem they have is a mismatch between what is acceptable in reality and what they think is acceptable based.. I think there needs to be more awareness around RAM stability at different temperatures, that a GPU can be fine at 80-100C, but RAM is happiest up to 45C or so. Hope to see more videos about this in the future!
Wow, I didn't know system RAM was so sensitive to temperatures. VRAM on Gpu's can easily get twice as hot and still work fine. I remember Kingston used to have water cooled DDR3 RAM called Hyper X h2o. But now days manufacturers seem to be more interested in making flashy RGB RAM than focusing on cooling.
"But now days manufacturers seem to be more interested in making flashy RGB RAM than focusing on cooling" From what I have observed, Flashy RGB ram tends to have bigger and thicker heatsinks.
One more thing you might want to add which might be limited to HEDT systems only is the heat output from the VRM's. On my X299 system, I found that the small fans on my VRM would push the heat into the surrounding area causing the RAM temperature to rise.
Lol obviously it wasn’t. He didn’t even comment on my comment yesterday. I was being facetious because of the similarities. But I will accept the burn.. great content man. I have been learning a lot about memory oc lately.
Quite funny that this video comes up right now, after I bought those 2x8 B-Die dimm for fun testing and noticed I needed to put lower clocks when temps > 50 C. Hynix CJR doesn't seem to mind as much so I'm considering removing the B-Dies since single-rank , even b-die, doesn't cut it to dual-rank CJR.
thank you buildzoid, i have no idea where corsair sourced the cheap DDR4 in one of my off builds but stability became trash once the gpu and cpu came into play until i placed a small fan on the ram.... despite as you said; memtest dumping it with many instances. Gonna blame the 3080ti for that. A 4 year belated thanks for saving me from self-committing to the looney bin.
Hey Buildzoid! As usual thanks a lot for the video. I am still using x99 and enjoying it a lot with a 5960x. What ram is the end game for this platform? I have a Gigabyte X99 Ultra Gaming. Since I wont be upgrading my x99 probably until ddr5 (and not even right away due to new ddr usualy being crap for at least a few months) i want to squeeze all I can from this mobo. Thus why I ask the one and only king to let me know what in your opinion is the best Ram for x99? Big thanks in advance.
if you have B-die, just point a fan straight at it. No ifs, no buts, just strap an 80mm or so fan straight on top of the sticks so that they have direct airflow. There, problem solved.
I’ve actually had issues with crashing due to this old G Skill memory cooler, I got it off of a triple channel Trident X kit for x58 I still use, the fans bearings are kinda in rough shape so the fans vibrate for a couple minutes until they warm up and it sometimes causes the system to crash. It took months before I made the connection. Random freezing on a cold startup that was completely unpredictable and unreplicateable. It always boots on the second attempt and stays stable lol
this is why i warm up the case with furmark and even run furmark at the same time as my memory test. but yes i came to this same conclusion a long time ago especially with my little mini 1070 ti which gets pretty hot. plus if the car gets hot a lot of that probably transfers over the slot to the board and then to ram as well. side panel fans help a lot. i need to test side panel fans exhaust vs intake. everybody says intake but i think either front fans as exhaust and side as intake or front as intake and side as exhaust makes more sense. why would having all that intake be better? has to mess with air flow right? anyway, had to replace the optical 5.25 bay with a hand cut plastic panel and cut it out for a fan because i have an old case design that i keep using. why spend money on cases right? so my only front fan was the bottom one. this has helped a lot. have the z390 aorus pro and it has ambient sensors. nothing goes above mid 40's c and theres a lot of sensors. but i have fans EVERYWHERE. my mini card gets so loud with its fan speeds that i just let the other fans ramp up to 1000rpm or so, why not? i wont hear it over the gpu anyway all this is why i just bought 3200 14-14-14 and ran it at 3600 16-16-16. all this finickyness (a word?) combined with the aorus pro's mediocre memory oc'ing means its not worth it for a 24\7 machine. if i really want a bit more ill just get another matching kit and go for dual rank. might have to bump voltage but not much. its not worth it for a normal machine. a bit more would be fine if the board supported it but wont make a big difference. has anybody tried front fans as exhaust and side panel as intake (also rear as intake too i guess?)? how did it work? edit: BZ mentions this in the vid. will draw hot air over the ram sticks. but i dont think that would really be a problem for me
I just OCed my mem sticks to 4133Mhz 17-17-17-37 (GSkill Trident Z) Rock stable on daily tasks and memtest, BSOD on gaming with SLI. You probably solved my problem. Better air flow near sticks or lower freq on them. Thanks :)
@@musicsouth2033 Those sticks are GSkill F4-4266C19-8GTZR. Their XMP profile is 4266 Mhz 19-19-19-39 on 1,4V but xmp profile isnt working on my z390 aorus master because these sticks arent on QVL list for this board. So i tried to manually oc these sticks. 4133 17-17-17-37 on 1,45V is my final stable result. I made two runs in memtest. Zero errors.
Also you might not have any overheating issues but if you have 4 sticks, PCIE 4 graphics card and an NVME you might also overwhelm the SOC if its starved for voltage. This can happen if you have a stable system and then introduce more ram sticks, NVME or a new graphics card to the mix.
4:15 sys_check_exception iirc, I was getting it a lot when I was being brain-dead and using 4 sticks on a daisy-chain board that weren't on the QVL. I switched to the 2x8GB Patriot Viper Steel 4400mhz cl19 kit you recommended and that issue has never occurred again.
My RAM temp can hit 80C during gaming. That is in a case with really good airflow, AIO on the CPU, and a 5700XT GPU. RAM is not OC, running at 3200. I even undervolted the RAM a bit (1.35V). GPU temp is well controlled at about 50C...maybe 60C Junction Temp.
BTW, would you recommend a RAM cooler of some sort when overclocking Bdie? I'm not so sure if it would help given the fact that the GPU will be throwing some of it hot air into it. thanks in advance.
That's why i want to mount my GPU horizontally with fans look down. I have got case, HAF XB Evo (from Cooler Master), but i can't find proper bracket, look's like it's not exist and i have to make it by myself. Yeah, i also need PCI Express riser, but it's smallest problem...
It's nice to see a video touch on IMC temps and the impact on ram overclock. In a lot of the circles I'm in this is a seldom touched on subject. I used to also experience a similar issue quite frequently when overclocking back in the AM3 days (AMD Opteron 1389 and AMD Opteron 3380) while pushing the BClk and IMC frequency up. There would get to be a point that no amount of voltage bump would stabilize the IMC at a high enough BCLK (265~282 in my case) even though the ram was already clocked below stock ranges. Once stabilized it would often take a small amount of heat to destabilize it beyond normal outside a memtest even though the cpu was otherwise 'cool'. Artificially limiting cooling (removing the CPU fan or covering it with a rag) to force a certain temp range I found helped find a stable max temp range in worst-case scenarios and ensure stability on the long run. I also found that at such extreme edge cases loosening timings resulted in worse stability as well and would only stay stable within a very specific timing range. 🤷
Just tried this out on my own system, just to get somewhat of an idea how much power it's using when i am watching this video, turns out, ~150 Watt's total. CPU: AMD 3900X RAM: Corsair Dominator Platinum RGB @3600Mhz CL16 - 4x 8GB modules Mobo: Gigabyte X570 Aorus PRO (Non WiFi) GPU: An almost antiek EVGA GTX 760, LoL no joke (But then again, i'am not a gamer, and it is going to be replaced not to long from now with a much newer one, so for now i don't realy care)
finally someone talking about the elephant in the room, that damn gpu backplate cooking your components and making all top mounted cpu radiator stupid (I have one because my current case can only have it there no space at the front) I have corsair vengeance pro rgb that have temp sensors I honestly never saw the memory go above 50°C but I have good airflow, the ideal tough is to have either a front mounted radiator or even cool the cpu with the bottom radiator nothing should be above your gpu backplate except fans exhausting air
This is why I use ASUS Realbench as the final arbiter of stability - it hammers everything (CPU, RAM, and GPU). If had systems pass everything _but_ Realbench, including Memtest.
Reminds me of HBM2. My Vega64 is technically stable at 1150 MHz just fine, though when HBM2 hits over 72 degrees, i start to get artefacts, at 75 degrees, anything will just crash. Anything above the drivers will usually crash.
What when I have Short PCB card with long cooler that blows air through the heatsink completely, example RX Vega 56 Red Dragon, or Sapphire Pulse, they are both Nano PCB
While stress testing my ITX case is 61c ambient around the ram. RAM peaks at 66c. My AIO is side mounted as an intake blowing up to 220watts from the CPU at the motherbaord. LOL. I could swap it to exhaust but then it pulls heat from the 3090 to cool the CPU with. 3733CL14 is stable. Never been able to get over 1866fclk 100% stable, so I don't really need faster ram speed.
For that reason, as memtest stable memory can crash very fast in actual use, i decided its worthless to do long memory testing. I just do quick one and then go test different things. I feel like that for me its actually caused by lower temperature than in memtest, because my memory actually never gets as hot as in memtest in other uses. Also i have been able to get settings that give errors very early in memtest, but once memory heats up errors stop. I don't know for sure but this could be due to asus prime x370 pro being really crap in memory overclocking. I thought that maybe i should try memory watercooling as i feel like memory on my system is very sensitive to temperatures, but then i just got triple 40mm fans to cool memory, but im not even sure if that helped memory overclocking, at least it got cooler.
Hello Buildzoid , i ordered a Patriot Viper Steel 2x8 4000Mhz kit ( TeamGroup and Patriot need to start paying you royalties man hahaha ) and i plan to run in an ITX sandwich layout case ( Nouvolo Steck ) so GPU won't be that much of an issue but i will have to remove the heat-spreader in order to make them fit under my NH-L12S cooler . Do you recommend to slap some sort of lower profile heatsink on them or can i leave them naked ( they will have an 120x15mm fan just above them ) , knowing that i plan to obviously OC them ( i mean underclock but tighten the timings ) so i might go as high as 1.45V ? Best Regards and Happy New Year bro !
Alb 92 If you plan to go that high on voltages, you should have some kind of heat sink, especially because mITX cases aren’t known for their good airflow around the CPU, even with blow down coolers. I suggest that, if you do it the way you want to, and encounter problems after some time (meaning your OC works, but heat screws you), try to get better cooling on your memory.
@@nanolog522 Well for this kit 1.45V is xmp if im not mistaken , and on top of that according to Buildzoid DDR4 should be able to hadle 1.45V without a heatsink with some minimum airflow if i recall correctly , but yeah he also said B-Die is sensitive so i will see . For sure i will pay attention to stability and temps and see how it goes from there . If a heatsink is needed ive found the generic heatsink used on corsair vengeange lpx on eBay so i will pull the trigger if needed .
@@alb9229 In mITX cases it is really hard to get airflow over the ram, even using coolers like the black ridge or l9i/a, because often the air has no way to go from there, resulting in a buildup of hot air in that area. You are correct, that RAM should work with little airflow, but if you can archive that is unlikely. Due to the close proximity of the RAM to the CPU on ITX mainboards you might also get thermal bleed from the CPU.
Would you recommend watercooling bdie? I already got a custom loop and without a fan bc of the front radiator my ram hits 52c even with fan 45 with 1.37vdimm
I had this issue with my new PC running on 4400mhz, first I was blaming my GPU for crashes then I blamed PSU and the last thing was the ram... but I noticed that when I open the PC case then there is no crashes anymore... I was undervolting the GPU and got less crashes then i was wasting more and more time until i put the gpu to other PC and tested there 3 hours without crashes. I removed xmp and cpu oc on my main pc and finally got rid of crashes, then tried xmp 4000mhz and seems no crashes anymore. Damn rtx 3080 gives a lot of heat.
I suspect that heat might be my problem with RAM-overclocking, the system is stable up to 3666 MT/s when I run Memtest86 (2 times 4 passes with no error), and I can use my system for light tasks or only CPU-intensive tasks 24/7 without any crash, yet my system crashes with certain games within 15 minutes. I had to lower the RAM clock frequency to 3400/2 MHz to get it fully stable.I use a Fractal S with stock fans, a D15 for the CPU and RAM with Samsung B dies. I consider to put a second front-intake fan, one full on the graphics card and one full on the RAM.
@UCYm0trd8r08vgZCrBmdVRvg If I get no error in 2 times 4 passes of Memtest86 then the memory should be stable or at least suffiiciently stable to rarely ever get a crash in games due to memory instability as long as heat is not a problem, as far as I understand. I only use Memtest86 to test stability of my memory. I have two modules of 16 GB (so dual rank) of Corsair Vengeance LPX, rated at 3333 MT/s and CL=16..
Try using TestMem5 with anta777's config, it finds errors memtest86 misses very easily. If you can get errors with testmem5 in the 3666 config, you can tell it's unstable for sure.
@Nuno Gomes I have that second fan which comes with the D15 but at the moment I can't plug it in, I don't have a suitable splitter. My motherboard has 2 case-fan connectors and 2 PWM-connectors of which one is for watercooling (not sure what that means).
@@qpwoeiruty668 It is possible that Memtets86 misses some errors but would it cause frequent crashes with a game - Rainbow Six Siege - which is not that heavy on the memory?
@@peterjansen4826 Nuno is right....your motherboard manual should confirm if the PWM connector for water cooling is a fan connector for an AIO water cooler. It is fine to connect a second CPU fan to it. Alternately you could buy a 4 pin PWM fan splitter cable from Noctua if you want to run both CPU cooler fans from a single PWM controlled motherboard header.
I believe it's something that is happening inside my PC. I have ~150 W Polaris RX 570, R5 1600 4 GHz and 2x8 Ballistix AES. When I overclocked and tested them they did 3400 MHz CL14. However when gaming, they caused blue screens at 3400 CL14 and 3200 CL14. 3200 CL16 works fine, but let's be honest it kinda sucks. Now I'm testing 3400 MHz CL16. I can't go higher becouse of shitty IMC, but probably becouse of heat CL14 I also consider as unstable. Scythe Infinity sits on my CPU and it has to be mounted vertical, and my front air intake is also low, so everything figures. The only difference, I have one 14 cm low RPM fan on side panel and I was shocked how great difference it makes at first.
That's me yesterday. I had gotten some garbage bin aliexpress Samsung memory on Zen+ 500% memtest stable at 3400mhz 1.4v with a fan on the top helping the airflow around it, decided it was too noisy at idle and changed that fan's curve and bam, got several bluescreens while just watching youtube. Memory overclocking is hard.
What about being stable under hci memtest 1000%, gaming and all type of different loads etc... turn off the pc, and not being stable on a cold boot the next day?
I have Hynix memory. Corsair Lpx 3000 kit. System is 2600x on b450 mortar titanium. Memory fails on anything above 3066, is it my settings or memory/hardware limit?
Could watercooling ram be effective then? Years ago I heard many times that ram just doesn't produce enough heat to be worth it, but I guess that might had been just ddr3? Also would chilled watercooling have additional positive effect on the ram?
If my cpu 10900k cores are sub 70c and my ram sticks are being actively cooled and I'm still getting memory related errors, I'm assuming it's the IMC getting to hot, is there a way to reduce the temperature of that component?
Do these GPU heat based errors appear fast? I had my 4x8 GB RAM running fine with 3733 MHz in my Ryzen 3000 system for more than a week and suddendly got several crashes in a row. Changed to 3600 MHz and no problems anymore.
I'm surprised that 50C is destructible for DDR4 memory. On my X99 system I have 8 x 16 GB Samsung M393A2G40DB1-CRC4Q ECC dual rank D-Die. Idle 35-47C and under load (prime) it can go up to 55-70C depending which stick. I have no issues now. I'm planning to play with it to get better timings. Do you think it's save ? I mean my case has average airflow all other components are cool enough. Is that normal for those sticks ? Is it save to start working on timings with those temps ?
I have trident z royals 1.55 volts xmp and they all get around 65 while gaming I spoke to a g skill rep through email he told me that’s absolutely normal. I have never gotten errors
Just a question not regarding the video. I bought new RAM (Patriot Viper 4 Blackout Series 16GB KIT DDR4 3600MHz CL17 2x8GB) which IS NOT supported officialy by my mobo (MSI B350 Gaming Plus + Ryzen 2600) and it wont post. From what I read the RAM usually works it just isnt stable at higher clocks etc. but should work. I tried both stick invidually in each slot and to no avail, while old RAM works just fine. My question is how likely are both sticks defective. Could clearing CMOS help?
The reason why the sticks won't work, isn't strictly because they haven't been approved by MSI, but its related. When you put a stick of Ram into your Motherboard, if it is on the QVL List, this means not only has the Memory Kit been tested, but the Motherboard Vendor has actually set Auto Timings for your Kit that will work with your kit. If it is not on the QVL, then the Auto Timings it is drawing on could be timings that just plain won't work for your specific kit. What you want to do to get the kit to work is go through and set every single Timing, Sub Timing, and Tertiary timing manually, as well as all Terminations and Voltages Manually, then you will boot just fine. For most people that is way too much work, in that case buying a Kit that is on the QVL is better suited for that type of user.
Yeah my 980 Ti with custom BIOS will do >500 watts on air and it was a bad bad time for my case temps. Had to pull off my sidepanel to stop the memory errors.
@@speedfastestsolo dude i was running an overclocked FX-8370 + OC'd 980ti (1440-7600) and i don't ever remember pulling anything close to 500w and i have a Corsair RM 750i which has gives you input/output wattage (total system)
Would you run Samsung B-die at something like 1.50 volts on the DRAM using a motherboard like the MSI MEG godlike Z390? of course it would be for daily usage.
@@m.streicher8286 Oh okay, that's good to know, so 1.50 volts on the DRAM should be fine? and your saying that SOC voltage would affect the IMC? that's system agent voltage right? what about VCCIO? Does that affect the IMC too?
Why mainboards aren't designed with the heat flow in mind, the hotest elements at the top, coolest at the bottom - e.g. pciex slots at the top of mb, then cpu and memory at the bottom.
Because they don't care if we fry our components because of bad design + bad airflow, they actually WANT us to upgrade our machines - yearly if possible :P
huh, i have a ryzen 5 3600x and during gaming or idle it just runs at around 50 to 55 with its stock cooler. in cinebench, where all cores are used, it runs at 80 degrees (and i bet, with a good cooler you could get it down to 75 or so). isn't 90 degrees a bit high for a cpu?
not even 30 minutes, but i have some constructive input, all consumer cpus/gpus use flip chip designs so generally closer to the pcb is hotter like you said, however i have a vertical mounted gpu and use a desk fan (big one) aimed at an open side panel (so i can see my rig) and ive had instability in games as opposed to in memtest. my belief is that the memory controller itself is surrounded by more heat while gaming hence throwing off stability. 10 bucks says this gets covered in the rest of the vid
so how do you stress the GPU and RAM? I'm having this same concern the GPU will cook my B-die RAM but i can't figure out a way of memtesting the RAM while simultaneously ramping up GPU power and heat. (basically the CPU is bottlenecking the GPU)
I'm having problems with this right now, with a 3960x at stock, one side of the memory sticks is about 10-15c hotter than the other, it will error out and crash apps when I'm running cpu rendering and gpu load gaming at the same time. Irql not less or equal 90% of the time. Put fans to push air onto the ram sticks but there is still a delta on one of the sticks.. Bitch to fix
Or you could have the problem I was having yesterday... The heat spreader falling off my Patriot Viper Steel because they never secured the thermal pad to the ICs properly. Edit: turns out that kit is the good one. I purchased a matching set and it was just a bad kit. Occasional errors even at 2133 default BIOS settings.
Yo Buildzoid! I watched your overclocking ram video on the master a few times and I managed to get 4133 17 17 17 34 450 trfc .... it seems to be working. I'm at 5giga 1.345 volts / high loadline, 1.44 ram voltage 1.2 vcc / vsa ............. Does this sound normal? Is it ok? Seems to be working.... I wish you were here to talk about it in person and you could teach me even more dude!
