|
Fhqwhgads posted:Is the 212 still the gold standard air cooler for non-extreme overclocking?
|
#
?
Oct 9, 2018 21:53
|
|
|
|
| # ? Apr 24, 2024 15:17 |
|
|
What's the popular option now? I thought the 212 was the cheapest good enough cooler.
|
|
#
?
Oct 9, 2018 22:12
|
|
|
Fhqwhgads posted:Is the 212 still the gold standard air cooler for non-extreme overclocking? I think the noctua dh15 is the current hotness
|
|
#
?
Oct 9, 2018 22:56
|
|
|
Cryorig H7 gets mentioned a lot in the PC parts thread, it's basically just a slightly fancier 212. There are also some competing ultra-budget 120mm tower coolers like the Deepcool Gammaxx 400 (so budget that installs with stock cooler style push pins, but should still perform about as well). Also with current CPUs now having 50-100% more cores what was "good enough" before might be cutting it too close now, so it could make sense to move up one or two tiers.
|
|
#
?
Oct 9, 2018 22:57
|
|
|
The DH15 is a price no problem cooler. It's very good (I own one) but it's not a cheap non extreme option. The Cryorig and Deepcool look like they perform a little better than the 212 and cost a little more, depending on sales that week I'd go with the cheapest one, I was worried the 212 was hopelessly outclassed and I was giving people wrong advice. If they're easy to install the extra 5 dollars might be worth it over the 212. craig588 fucked around with this message at 23:16 on Oct 9, 2018 |
|
#
?
Oct 9, 2018 23:01
|
|
|
NH-D15 (Quiet/Silent version) remains the gold standard, but the current lack of non-fugly colours lead some to buy a Chromax instead. Noctua wil be releasing a black version of the D15 some time next year when they figure out how non-brown colours work. If you're on Ryzen, the stock Wraith cooler is pretty awesome apparently. ufarn fucked around with this message at 23:54 on Oct 9, 2018 |
|
#
?
Oct 9, 2018 23:52
|
|
|
craig588 posted:I was worried the 212 was hopelessly outclassed and I was giving people wrong advice.
|
|
#
?
Oct 10, 2018 00:24
|
|
|
AIO are so affordable now I don't see myself ever going back to an air cooler.
|
|
#
?
Oct 10, 2018 00:36
|
|
|
mewse posted:I think the noctua dh15 is the current hotness Noctua is over 2x as expensive. The 212 does *great* for non-extreme overclocking.
|
|
#
?
Oct 13, 2018 19:54
|
|
|
I've had a hyper 212 evo for years now, I slapped a second fan on the heatsink and I get about 80c when all 8 threads of my i7 3770k are at 100%, but I think that's mostly a limitation of the crappy TIM at this point, delidded temps would probably be 20-25c lower. It's telling that it'll jump up 20c within a second and drop back down 20c a second later, like the heat from the die just isn't being transferred enough to dissipate it over time. That's at 4.5ghz at 1.3v. Though I think these days cryorig seems to have the better coolers around the 212's price bracket. Like the H7 or something.
|
|
#
?
Oct 15, 2018 10:15
|
|
|
Only reason to buy a 212 is if you need 20xx support or the cryorig h7 is out of stock.
|
|
#
?
Oct 15, 2018 15:55
|
|
|
Only 120mm fans? Ew
|
|
#
?
Oct 15, 2018 16:34
|
|
|
cat doter posted:I've had a hyper 212 evo for years now, I slapped a second fan on the heatsink and I get about 80c when all 8 threads of my i7 3770k are at 100%, but I think that's mostly a limitation of the crappy TIM at this point, delidded temps would probably be 20-25c lower. It's telling that it'll jump up 20c within a second and drop back down 20c a second later, like the heat from the die just isn't being transferred enough to dissipate it over time. That's at 4.5ghz at 1.3v. Yeah, that sounds like partly a TIM issue. My 2500K at 4.4GHz/1.39V never went over 70C with a 212+ using only the stock fan. Don Lapre posted:Only reason to buy a 212 is if you need 20xx support or the cryorig h7 is out of stock. Well, they go on sale for $20 a lot and it's not actually that hard to put on if you have the mobo flat and have done it before; I've bought like six of them with uniformly great results. The system I'm typing this on right now has dual X5670s with 212s and even with Prime95 going on all cores I'm looking at 50-55C, so it's not bad unless you start trying to exceed around the 120-130W it was designed for. It isn't actually good at this point either though, so I'd only buy another if it was on sale and I had a similar <=100W locked clock situation. Might as well shoot a bit higher if performance could actually be affected.
|
|
#
?
Oct 15, 2018 16:49
|
|
|
cat doter posted:I've had a hyper 212 evo for years now, I slapped a second fan on the heatsink and I get about 80c when all 8 threads of my i7 3770k are at 100%, but I think that's mostly a limitation of the crappy TIM at this point, delidded temps would probably be 20-25c lower. It's telling that it'll jump up 20c within a second and drop back down 20c a second later, like the heat from the die just isn't being transferred enough to dissipate it over time. That's at 4.5ghz at 1.3v. I had the exact same setup but with a 4790k, FYI when I swapped to a corsair 280mm AIO it dropped me only about 10C across the cores. It's definitely a limitation of the TIM. 80C is fine and sounds optimal with the hyper 212 EVO.
|
|
#
?
Oct 15, 2018 17:30
|
|
|
Can someone recommend a good Micro ATX (or ITX) z370 board that's good for overclocking? Is there any reason to go z390 other than 9th gen compatibility? Last night I was playing with my system just getting it tuned, was happily humming along at 5ghz (on my 8700k w/ AIO cooler) doing stress tests with OCCT and POV-Ray. I was using ASrock's "F-Stream" utility to adjust vcore from within windows, trying to find a good balance between stability/temps. I made what I consider to be a rather innocuous change (vcore from 1.355 to 1.325), and the exact moment I clicked on "apply" my machine went dark. It powers on now, but nothing will come up on the screen. No POST beeps, either. Nothing. I've cleared the CMOS and removed all peripherals, nothing seems to matter. I'm hoping my CPU is ok because that really shouldn't have hurt it, but then again I don't see why it would have killed my mobo either. It doesn't make any sense. It's an ASRock z370 gaming-ITX. The only bios changes I've ever made on it are clock ratio (never beyond 50 / 5ghz), LLC (changed from level 5 to level 1), and vcore which has always been somewhere between 1.25 - 1.4. Anyways back to my original question. Is this board worth a poo poo? Looking for something I can get quickly.
|
|
#
?
Oct 16, 2018 19:24
|
|
|
https://www.youtube.com/watch?v=wpSMW27gzzo GN live (not anymore) 9900k OC stream. VelociBacon fucked around with this message at 03:58 on Oct 20, 2018 |
|
#
?
Oct 20, 2018 03:55
|
|
|
Question about LLC (load line calibration). I have an Asrock z370 Extreme 4 and have been experimenting with LLC settings. Normal setting is 3 (medium) and at 100mv offset I get right under 100c throttle temps at avx loads. I tried LLC 1 and temps skyrocketed yet the voltage went down to around 1.25v. Should I try say LLC 4 to drop temps even more? Should I be concerned with VID being higher than LLC 3 max at 1.44ish I figured LLC 1 would just stop the vdroop but it seems to crank temps to insane levels. I don't get it.
|
|
#
?
Oct 20, 2018 05:35
|
|
|
redeyes posted:Question about LLC (load line calibration). I have an Asrock z370 Extreme 4 and have been experimenting with LLC settings. Normal setting is 3 (medium) and at 100mv offset I get right under 100c throttle temps at avx loads. I tried LLC 1 and temps skyrocketed yet the voltage went down to around 1.25v. I have a rudimentary understanding of LLC so please take this advice with a grain of salt and maybe look around more online or wait for someone else on here who knows better to confirm but: The sharp end of LLC settings will actually create a slope that is positive, so your Vcore ends up higher than your VID. Are you monitoring your VID and Vcore? Adjust your LLC until they are close to the same. I would err on the LLC setting that gives you the closest Vcore just under your VID.
|
|
#
?
Oct 20, 2018 05:41
|
|
|
VelociBacon posted:I have a rudimentary understanding of LLC so please take this advice with a grain of salt and maybe look around more online or wait for someone else on here who knows better to confirm but: I was measuring VID. Which at LLC level 4 is now up to 1.48v?! at lidle, 1.44 at load. I am not even sure how to figure out what vcore is set to since I have the BIOS set to offset voltage.
|
|
#
?
Oct 20, 2018 05:49
|
|
|
redeyes posted:I was measuring VID. Which at LLC level 4 is now up to 1.48v?! at lidle, 1.44 at load. I am not even sure how to figure out what vcore is set to since I have the BIOS set to offset voltage. What software are you using? Try HWinfo64.
|
|
#
?
Oct 20, 2018 05:53
|
|
|
VelociBacon posted:What software are you using? Try HWinfo64. Was using CoreTemp but I have HWinfo64. I can't seem to find vcore in here.. [edit] Ah found it under motherboard sensors. Idle is around 1.0v to 1.28. AVX load is around 1.36v vcore. VID reports like 1.44v Sound safe? redeyes fucked around with this message at 06:01 on Oct 20, 2018 |
|
#
?
Oct 20, 2018 05:58
|
|
|
redeyes posted:Was using CoreTemp but I have HWinfo64. I can't seem to find vcore in here.. Yeah sounds like you're drooping from 1.44v to 1.36v. You could try one setting more LLC if you want your Vcore to be closer to your VID I think but I haven't checked if 1.44 is the safe voltage for your chip or anything. e: If you look at your vcore in HWinfo64 when it's idling and then see it drop when you start a stresstest that'll also show whether you're drooping. This won't work if you have c-states enabled I think so you'll have to disable them first. VelociBacon fucked around with this message at 11:01 on Oct 20, 2018 |
|
#
?
Oct 20, 2018 06:43
|
|
|
Don't ever disable C-states. All that does is make the CPU use more power even when it's not actually doing anything. If you want to check how big the droop is at max boost, the easiest way to do that is probably to set the Windows power management to maximum performance. That should make it idle at the max boost frequency and never clock down. However, there is no real reason for doing that. Knowing exactly how big the droop is at idle isn't very interesting. To review some things you probably already know: at a hypothetical zero load (no current going through the CPU), VID is equal to Vcore. With low or no LLC, the more current you use, the bigger the difference between VID and Vcore gets (Vcore drops while VID stays the same). The higher the frequency, the higher the VID. Using AVX will also increase VID. If you get the CPU to sit at its max frequency, the VID will be quite high - how high depends on what you've done in the BIOS, but in this case 1.44v. Now, with C-states enabled, even though the frequency is high, if there's no load on the CPU, the current will be low and the actual Vcore will be quite close to VID. If you put load on the CPU, VID will stay the same but the current will be higher and Vcore will drop, in this case down to 1.36v. For a daily overclock on Coffee Lake, keeping Vcore below 1.4v under full load is generally accepted as safe. At low loads though you can allow Vcore to be slightly higher - it's the combination of high current and high voltage that's really dangerous. The reason to use LLC is for scenarios where you've pushed the idle Vcore as far as you dare (to, say, 1.45v) but you're not stable under load because it drops down too far. Then you use LLC to bring the Vcore under load closer to idle Vcore. You basically never want the highest LLC settings that invert the droop so you get a Vcore higher than Vcore under load. Anyway, the point here is that as long as you stay away from the highest LLC settings, all you really need to care about is Vcore under load. As long as you keep that below 1.4v you're good. VID isn't a real voltage so that being high doesn't matter, and idle Vcore isn't that important as long as it's not super high. TheFluff fucked around with this message at 14:39 on Oct 20, 2018 |
|
#
?
Oct 20, 2018 14:36
|
|
|
VelociBacon posted:Yeah sounds like you're drooping from 1.44v to 1.36v. You could try one setting more LLC if you want your Vcore to be closer to your VID I think but I haven't checked if 1.44 is the safe voltage for your chip or anything. On my board LLC 4 is less than LLC 3. LLC 1 is the least Vdroop. LLC 2 and 1 are WAY hot on my CPU, it throttles instantly at AVX load. quote:As long as you keep that below 1.4v you're good. VID isn't a real voltage so that being high doesn't matter, and idle Vcore isn't that important as long as it's not super high. For reference, this is an 8700K with no fixed voltage set, just an offset of 150mv @5Ghz AVX also at 5Ghz, Cache clock at 4.7Ghz. I am also running Windows 10 pro for Workstations and no joke, running benchmarks with Ultimate Performance mode nets say %1-%5 performance increase vs Always On/Performance power mode. I run aircooling which is a Cryorig H5 with a 140mm static pressure fan. I did get my Corsair DDR4 3200Mhz (el cheapo) memory to CAS 14 @ 3200 which made me a little happy. Now I am on to Memory Command Rate. 2T vs 1T. I am not sure if this is even worth messing with but I have tried a bunch of settings at 1T and bluescreen at login. I am not sure if say CAS 16 at 1T is better or worse than CAS 14 at 2T. Maybe a goon will know. redeyes fucked around with this message at 15:47 on Oct 20, 2018 |
|
#
?
Oct 20, 2018 15:22
|
|
|
1T command rate has always been a marginal latency improvement in exchange for a variable but decidedly not-worth-it hit to stability. Don't do it unless you're playing with benchmarks only.
|
|
#
?
Oct 21, 2018 14:42
|
|
|
Sidesaddle Cavalry posted:1T command rate has always been a marginal latency improvement in exchange for a variable but decidedly not-worth-it hit to stability. Don't do it unless you're playing with benchmarks only. Got it. I couldn't get it to work anyhow. I settled on CAS 14 @ 3200.
|
|
#
?
Oct 21, 2018 17:14
|
|
|
Is there a really good write-up or video anywhere about the advanced BIOS stuff in relation to overclocking? Stuff like cache ratios, SVID, that kind of thing that is common across boards.
|
|
#
?
Oct 22, 2018 02:11
|
|
|
So I finally got around to getting my 3770k delidded, I've been meaning to do it for a while. Operating temps are crazy good, it seems like I can throw as much voltage at this thing and temps remain quite manageable. I attempted some more overclocking since I had it at 4.5ghz, but I slammed into a wall hard at 4.8ghz, seems like without some real expert level overclocking skills I wouldn't be able to get that stable. I can get into windows fine but everything in windows sorta behaves funky and it eventually blue screens. I've got it at 4.7ghz now and I'm having some issues getting it 100% stable. I bumped load line calibration to level 2 and it's hitting around 1.4v at least according to CPU-Z. Anyone have any tips getting this rock solid?
|
|
#
?
Oct 23, 2018 18:18
|
|
|
cat doter posted:So I finally got around to getting my 3770k delidded, I've been meaning to do it for a while. Operating temps are crazy good, it seems like I can throw as much voltage at this thing and temps remain quite manageable. What is your Vcore before/during load? What temps are you at under load? Sorry it's not clear if that 1.4v is during load or not.
|
|
#
?
Oct 23, 2018 18:25
|
|
|
VelociBacon posted:What is your Vcore before/during load? What temps are you at under load? Sorry it's not clear if that 1.4v is during load or not. Load temps are about 68c, and vcore during load hits 1.392 but seems to stabilise at 1.384, that's using core offset. Idle vcore seems to be about 1.2v.
|
|
#
?
Oct 23, 2018 18:34
|
|
|
So I've been playing with my 9700k trying to see what I can get out of it, and I'm running into a weird issue that I'm hoping someone can help me understand. This is at 4.9ghz all cores: Scenario A: I set vcore to 1.32v and leave LLC on 'normal' which allows a lot of drop under load. When loaded, vcore drops to 1.24v and my temps are in the high 60s/low 70s. Scenario B: I set vcore to 1.24v and set LLC to 'turbo' which keeps vcore steady. When loaded (exact same load as before), vcore stays a constant 1.24v but the temps are now in the high 70s. What is the difference here? Why does it run so much hotter using LLC when the voltage under load is the same? In both scenarios I can't drop the voltage any further, it's unstable if I let it drop below 1.24.
|
|
#
?
Oct 23, 2018 21:28
|
|
|
That sounds pretty weird. What are you using to read out vcore? What about reported power consumption, is that the same too? Also, dumb question, but are you using the same benchmark for testing? There's a lot of difference between AVX and non-AVX workloads, for example.
|
|
#
?
Oct 23, 2018 21:37
|
|
|
Yeah hwinfo64 had two vcores listed for my mobo and I kinda had to trial and error to figure out which was for the processor
|
|
#
?
Oct 23, 2018 21:38
|
|
|
hwinfo64 was giving me two readings for vcore as well and I wasn't sure which to go by, so I switched to SIV (system information viewer) and have been going by that. The load is from cinebench and pov-ray in both scenarios, they both seem to peak around the same temps for me.
|
|
#
?
Oct 23, 2018 21:46
|
|
|
n.. posted:So I've been playing with my 9700k trying to see what I can get out of it, and I'm running into a weird issue that I'm hoping someone can help me understand. I have to say I'd expect a "turbo" mode LLC to be a positive slope situation so I'd be careful and maybe use a couple different pieces of software to see what you're actually delivering.
|
|
#
?
Oct 23, 2018 21:50
|
|
|
VelociBacon posted:I have to say I'd expect a "turbo" mode LLC to be a positive slope situation so I'd be careful and maybe use a couple different pieces of software to see what you're actually delivering. Yeah, about that - So my bios shows a load/voltage line graph and highlights the line for whichever option you have chosen - interestingly none of them have a positive slope, the highest setting "UltraExtreme" shows as a steady line, but in reality I get more voltage under load using that setting. It was through trial and error that I landed on "turbo" (which is a slight downward slope a couple notches down on the graph) actually performs as though it's a straight line. I think they're just mislabeled in gigabyte's lovely bios. I don't think I checked vcore in cpu-z so I'll try that when I get home and see if the readings are any different.
|
|
#
?
Oct 23, 2018 21:58
|
|
|
Is that an MSI board?
|
|
#
?
Oct 23, 2018 22:00
|
|
|
n.. posted:Yeah, about that - I really don't think CPU-Z is reliable when it comes to vcore reporting. I don't know what would be though if HWinfo64 isn't being reliable. :/ You could try Intel XTU, maybe? Definitely check the package power reported by HWInfo though. e: also, vcore as reported via software monitoring is typically read out in 0.016v increments. On my board vcore keeps bouncing back and forth between 1.328 and 1.344v at full load, which probably means the actual vcore is somewhere between 1.33 and 1.34v. Then again it's probably not perfectly accurately measured in the first place. TheFluff fucked around with this message at 22:10 on Oct 23, 2018 |
|
#
?
Oct 23, 2018 22:02
|
|
|
VelociBacon posted:Is that an MSI board? Gigabyte z390 aorus pro (i know i know, gooncensus is that gigabyte is a garbage fire but if you watch buildzoid's video on the new z390 boards they're actually good, with the same VRM components as much higher end boards (actually better on the pro than the master)). Admittedly the bios is pretty bad, though. TheFluff posted:I really don't think CPU-Z is reliable when it comes to vcore reporting. I don't know what would be though if HWinfo64 isn't being reliable. :/ You could try Intel XTU, maybe? Definitely check the package power reported by HWInfo though. It's not that hwinfo64 is being unreliable, it's that there are two mb sections on the sensor screen with two different 'vcore' readings, and I have no idea which is the correct one for the CPU.
|
|
#
?
Oct 23, 2018 22:10
|
|
|
|
| # ? Apr 24, 2024 15:17 |
|
|
n.. posted:It's not that hwinfo64 is being unreliable, it's that there are two mb sections on the sensor screen with two different 'vcore' readings, and I have no idea which is the correct one for the CPU. Huh. Odd. Do they seem to follow each other but with an offset from each other, or is there something weirder going on?
|
|
#
?
Oct 23, 2018 22:12
|
|