|
willroc7 posted:By watch for do you mean use the AVX offest? Thanks for the reply. Right now I'm using a Hyper T4 with some cooler master clotted cum thermal paste, but I've got a Noctua D15 and some conductonaut on the way FYI, I found the Noctua thermal paste to actually be really good. I am running the D15 and it is wonderful. So basically, factor in the AVX. So if you are at 4.6GHZ and your AVX offset is set to 4, and you do a benchmark uses AVX you are now benchmarking 4.2. Similarly, if you are not using AVX offset and your test is using that instruction set you will be generating more heat than any "typical" workload. You mentioned you were running 70 degrees under load, and I was wondering load that was. Overall, that is well within the thermal specs, so I'd say you have plenty of headroom especially if you are putting on a D15.
|
#
?
Nov 17, 2017 20:12
|
|
.
|
|
| # ? Apr 26, 2024 01:39 |
|
|
NT-H1 is pretty good stuff, Gelid GC Extreme is a little better but twice as much per tube. These days pretty much everything beats the old Arctic Silver 5 and such, non-conductive pastes have come a really long way. Liquid metal spanks everything else, of course, but it's also a royal pain to deal with.
|
|
#
?
Nov 17, 2017 20:39
|
|
|
I got the conductonaut for delidding. Should I use that for the noctua as well? And the load I was talking about was Prime95 blend.
|
|
#
?
Nov 17, 2017 20:41
|
|
|
willroc7 posted:I got the conductonaut for delidding. Should I use that for the noctua as well? And the load I was talking about was Prime95 blend. Latest Prime95? Try small FFT. If your in the 70s with no AVX offset you are in really good shape. I originally wanted to delid, but I am running pretty well at up to 5GHZ so there wasn't much point other than academic curiosity, really.
|
|
#
?
Nov 17, 2017 20:54
|
|
|
Lockback posted:Latest Prime95? Try small FFT. If your in the 70s with no AVX offset you are in really good shape. Hm, I'll have to check what version.
|
|
#
?
Nov 17, 2017 21:09
|
|
|
Lockback posted:Latest Prime95? Try small FFT. If your in the 70s with no AVX offset you are in really good shape. I checked your posts to the thread and see you've got an 8700k and asrock board. Could you tell me what you've got your clock and vcore set to, and what is the setting called for vcore in the asrock bios? I did a "ez overclock" thing in the bios for 4.5 but it thermal throttles using old non-avx prime95, I think because the mobo auto sets the voltage too high. The default avx offset on my pro4 seems to be 300mhz which is relatively modest compared to the ~1ghz you were seeing on your board.
|
|
#
?
Nov 17, 2017 21:41
|
|
|
mewse posted:I checked your posts to the thread and see you've got an 8700k and asrock board. Could you tell me what you've got your clock and vcore set to, and what is the setting called for vcore in the asrock bios? The EZ overclock really over-cranks the voltage, I would set it manually. I think they rely on a very high voltage and a high vdroop. I was seeing some really odd behavior with the AVX offset in the original BIOS, and the EZ overclock also sets it (I think one of the times I set it to 0 it showed default but was using the last value from the EZ Overclock). ASRock actually sent me a custom 1.2 bios that allows you to set default to 0 if you want, which was pretty cool. As for selection, it's under the Voltage Configuration, CPU Core Voltage. I just set an absolute value, I don't use the offset or Auto. I run 4.7 at 1.28v or 5GHZ at 1.375 (which is probably higher than I need, but the temps were pretty reasonable). I got it stable at 5.1 @ 1.39 but I couldn't get 5.2 stable (though, I haven't gone any deeper trying to adjust crap. I suspect I might be able to if I spend the time). I can get some exact menu picks when I get home later, but don't use the EZ Overclock.
|
|
#
?
Nov 17, 2017 23:11
|
|
|
Lockback posted:The EZ overclock really over-cranks the voltage, I would set it manually. I think they rely on a very high voltage and a high vdroop. I was seeing some really odd behavior with the AVX offset in the original BIOS, and the EZ overclock also sets it (I think one of the times I set it to 0 it showed default but was using the last value from the EZ Overclock). ASRock actually sent me a custom 1.2 bios that allows you to set default to 0 if you want, which was pretty cool. If you don't use vcore offset, does it keep the voltage high even when idle? I might have read about that causing huge power consumption
|
|
#
?
Nov 17, 2017 23:19
|
|
|
Edit: Nevermind, read that wrong.
|
|
#
?
Nov 17, 2017 23:29
|
|
|
mewse posted:If you don't use vcore offset, does it keep the voltage high even when idle? I might have read about that causing huge power consumption Right, but I usually shut my PC off when I'm not actively using it so offset is basically identical in my case. If you keep yours running offset is worth considering, though I am not sure if the idle voltage delta will really be that much difference in kilowatt hours. How your case handles heat would also be a worry. I have literally never seen a CPU that had it's lifespan cut down to below 6 years due to overclocking, so I'm usually not super worried about that.
|
|
#
?
Nov 18, 2017 01:40
|
|
|
Conductonaut between ihs and cooler is riskier and doesn't have a big benefit anymore, like a few degeees compared to the 20C you can get with just delid. Also someone said it melts the text off your ihs, and your cooler depending on what it is made of. Also bigger risk for it to get to wrong places.
|
|
#
?
Nov 18, 2017 04:54
|
|
|
I'm not sure Conductonaut is worth the hassle even for under the IHS - like I think you're supposed to redo it yearly. Most of the gains come from shaving away a bit of the adhesive - Intel has a Z-height problem, not really a paste problem. Even a regular thermal paste is going to be a lot better after delidding. Hot take: delid and Gelid GC Extreme.
|
|
#
?
Nov 18, 2017 05:54
|
|
|
Ihmemies posted:Conductonaut between ihs and cooler is riskier and doesn't have a big benefit anymore, like a few degeees compared to the 20C you can get with just delid. Also someone said it melts the text off your ihs, and your cooler depending on what it is made of. Also bigger risk for it to get to wrong places. In my experience, there really isn't much reason to worry about liquid metal moving around as long as you're not using altogether too much of it. It does have some effect on the condition of the IHS though. When I changed out coolers on my 7700k that had liquid metal between the IHS and the copper plate, I noticed a sort of cloudy grey quality to the surface that I couldn't remove completely with isopropyl alcohol or light scrubbing.
|
|
#
?
Nov 18, 2017 06:01
|
|
|
Paul MaudDib posted:I'm not sure Conductonaut is worth the hassle even for under the IHS - like I think you're supposed to redo it yearly. Most of the gains come from shaving away a bit of the adhesive - Intel has a Z-height problem, not really a paste problem. Even a regular thermal paste is going to be a lot better after delidding. From what I’ve seen liquid metal still nets you an additional 5-8C vs regular paste between the die and the IHS. You also don’t need to redo the LM yearly, silicon lottery stated they had come CPUs that were a few years old that were still maintaining the same temps are a few years use. I’d agree that removing the adhesive and using better paste will still be better than stock but LM still gives better temps over all under the IHS. I agree that LM isn’t worth it between the IHS and the CPU cooler.
|
|
#
?
Nov 18, 2017 16:55
|
|
|
Liquid metal is hugely impactful underneath the heatspreader. Conductonaut is more themally conductive than solder, you can get improved temps on soldered CPUs just by heating off the solder and replacing it with LM. De8auer has videos on the subject. I'll always recommend LM when delidding.
|
|
#
?
Nov 18, 2017 17:04
|
|
|
I am getting the i5 8600k and I don’t really plan to overclock it - I don’t need to take the CPU apart and replace that thermal material, do I?
|
|
#
?
Nov 18, 2017 17:22
|
|
|
Three-Phase posted:I am getting the i5 8600k and I don’t really plan to overclock it - I don’t need to take the CPU apart and replace that thermal material, do I? Why get a K series and not overclock it? If you didn't overclock, there would be no real reason to delid, I don't think.
|
|
#
?
Nov 18, 2017 17:24
|
|
|
No
|
|
#
?
Nov 18, 2017 17:24
|
|
|
Yeah use lm and seal with silicon and never touch it again. Leave a small gap though in the silicon or just do the four corners.
|
|
#
?
Nov 18, 2017 17:37
|
|
|
Three-Phase posted:I am getting the i5 8600k and I don’t really plan to overclock it - I don’t need to take the CPU apart and replace that thermal material, do I? Even if you do overclock (and you should, even a small oc), delidding isn't something you need to do. Intel left some thermal transfer on the table with how they connected the ihs, but it's all still well, well within spec and the CPU is more than happy stock. But seriously, clock it to 45x on all cores with default voltage. Zero risk, barely more heat and you should be stable unless your case is a sauna.
|
|
#
?
Nov 18, 2017 18:11
|
|
|
You could leave it stock and delid it so your fans or coolers run less and are quieter.
|
|
#
?
Nov 18, 2017 18:57
|
|
|
Lockback posted:Right, but I usually shut my PC off when I'm not actively using it so offset is basically identical in my case. If you keep yours running offset is worth considering, though I am not sure if the idle voltage delta will really be that much difference in kilowatt hours. How your case handles heat would also be a worry. I have literally never seen a CPU that had it's lifespan cut down to below 6 years due to overclocking, so I'm usually not super worried about that. Thanks for all your posts yesterday. I spent hours last night figuring out what was going on with my system. I built this machine with a bigass 280mm AiO radiator so I couldn't understand how you were getting 4.9/5.0 on air cooling. What I discovered is that the voltage regulator module (VRM) on this motherboard is loving terrible for overclocking. The Asrock pro4 doesn't just have a bad VRM for overclocking, the top side of chips aren't even heatsinked:  There's a thread on overclock.net where they rank the VRM implementations on the various z370 boards. The pro4 is literally on the bottom with the advice "not advisable":  At the end of last night I was just trying to get it stable at 4.7 but after like 10 minutes of non-AVX prime95 torture, it was throttling with "vrm temperature throttling" triggered. I think if I stick some small aluminum heatsinks to those chips I might be able to get away with it, but I'm basically fed up with the motherboard at this point. This is my first time overclocking and I feel like I got burned a bit - I guess the lesson is live and learn and don't buy the cheapest motherboard. (sorry this prob belongs in the overclocking thread but all the discussion was in this thread)
|
|
#
?
Nov 18, 2017 19:05
|
|
|
mewse posted:Thanks for all your posts yesterday. I spent hours last night figuring out what was going on with my system. I built this machine with a bigass 280mm AiO radiator so I couldn't understand how you were getting 4.9/5.0 on air cooling. I guess I don’t have much say after all as far as overclocking my 8600k goes. (Should I just return my shiny new ASRock z370m Pro4 motherboard that arrived this week?) Unless the Overclocking you’re doing is a bit more extreme... Whelp this build is sort of going form bad to worse.  EDIT: Can the VRM’s be delidded? Three-Phase fucked around with this message at 19:20 on Nov 18, 2017 |
|
#
?
Nov 18, 2017 19:11
|
|
|
Three-Phase posted:EDIT: Can the VRM’s be delidded? Lol, no. The only workaround I can see is sticking chinese aluminum heatsinks to them like I mentioned. I think these heatsinks could work. The problem is even if the VRM chips have heatsinks, it's a poor implemenetation and it needs to provide more juice to the cpu to keep things running, elevating temperatures, preventing higher overclocks. The higher end mobos say "digital VRM" and I don't know if that's a thing -- is the pro4's VRM analog?! I'm still new to this too. If you want to do a modest overclock like 4.6 it should work fine. If you want to push things into 4.9/5.0 like the coffee lake reviews all discussed, you need a better mobo.
|
|
#
?
Nov 18, 2017 19:27
|
|
|
Three-phase stop trolling these poor guys.
|
|
#
?
Nov 18, 2017 19:28
|
|
|
3.6GHz to 4.6GHz is a “modest overclock” nowadays?  VulgarandStupid posted:Three-phase stop trolling these poor guys. Dammit when I got my first 486/66 it had a heatsink and fan that must’ve been about the size of a half-dollar.
|
|
#
?
Nov 18, 2017 19:29
|
|
|
Three-Phase posted:3.6GHz to 4.6GHz is a “modest overclock” nowadays? Stock turbo on a 8700k is 4.7GHz. So the silicon is designed for at least that, power and temperature permitting.
|
|
#
?
Nov 18, 2017 19:46
|
|
|
mewse posted:Thanks for all your posts yesterday. I spent hours last night figuring out what was going on with my system. I built this machine with a bigass 280mm AiO radiator so I couldn't understand how you were getting 4.9/5.0 on air cooling. Wow, I did not realize the Extreme4 and the Pro4 were that different. Can you return the Pro4? The Extreme4 has been super easy and it really doesn't cost that much more.
|
|
#
?
Nov 18, 2017 21:03
|
|
|
There are holes for screws directly to the left and right of that bank of ICs. So if you measure things accurately it may be possible to fabricate a heatsink and mount it on them. But it would take some work and my concern is that you may not have even contact across all the chips. Have you considered immersing the whole system in oil or FlorInert pfleurocarbon (like they did with the old Crays)?
|
|
#
?
Nov 18, 2017 21:41
|
|
|
Three-Phase posted:Have you considered immersing the whole system in oil or FlorInert pfleurocarbon (like they did with the old Crays)? Not sure how practical it is but it sure looks cool. https://twitter.com/tomshardware/status/930830838681538560
|
|
#
?
Nov 18, 2017 21:52
|
|
|
mewse posted:At the end of last night I was just trying to get it stable at 4.7 but after like 10 minutes of non-AVX prime95 torture, it was throttling with "vrm temperature throttling" triggered. I think if I stick some small aluminum heatsinks to those chips I might be able to get away with it, but I'm basically fed up with the motherboard at this point. I'm running these in my Titan XP VRAM chips. They helped my overclocks when I went from the reference cooler to a hybrid solution. https://www.amazon.com/dp/B00P8BAZ9W/ they are 10mm square and 14mm tall. It's a pack of 8 and come with thermal sticky pads. Not entirely sure how big VRM chips are, but you may be able to stick 1 sink on each chip. They make some smaller ones that look like they may be 5mm square, but they require additional thermal adhesive or sticky pads purchase. SlayVus fucked around with this message at 21:56 on Nov 18, 2017 |
|
#
?
Nov 18, 2017 21:52
|
|
|
MaxxBot posted:Not sure how practical it is but it sure looks cool. I get that the flourinert is vaporizing inside the box to draw energy away from the components, but how does it cool down enough to go back to liquid? Does the gaseous flourinert get run through a heat exchanger with the atmosphere?
|
|
#
?
Nov 18, 2017 22:37
|
|
|
VostokProgram posted:I get that the flourinert is vaporizing inside the box to draw energy away from the components, but how does it cool down enough to go back to liquid? Does the gaseous flourinert get run through a heat exchanger with the atmosphere? You have a custom loop sitting just out of frame. You have a 120*240mm rad inside the box, and you pump water through it to two more 120*240 rads outside. The water ends up basically at 20-25C after the 2nd rad, goes in through the pressure bulkhead to the internal radiator and all the fins act as a shitload of condenser area for the gas, it works quite well, and even if your room is 35C because you're a buttcoin business magnate, it will still work quite well in any case that was actually designed for a 10 PSI overpressure. My 'this is stupid, awesome, but stupid PC build', will be a wall mount, like the ones you see at gaming case competitions, with a bigass plexiglass box in the center that holds the core system. The stuff is about $250/gallon if you buy 5 gallons, which is enough to do a full ATX build without too much issue. I'm gonna be a wiseass and take apart the power supply and bury it in blue aquarium gravel like it's some kind of pirate ship, possibly have a plastic fish floating in the tank. Methylethylaldehyde fucked around with this message at 22:58 on Nov 18, 2017 |
|
#
?
Nov 18, 2017 22:50
|
|
|
After some exhaustive resaearch I am going to grab a Corsair H60. It is in fact compatible with 1151, the sockets have a little slider you need to adjust for the four bolts. The big Cooler Master air coolers are too large for the Corsair Carbide 250 mATX case I am getting. Now just waiting in the Microcenter line... 😪
|
|
#
?
Nov 19, 2017 18:00
|
|
|
Lockback posted:Wow, I did not realize the Extreme4 and the Pro4 were that different. Can you return the Pro4? The Extreme4 has been super easy and it really doesn't cost that much more. Come on now. I am sure you can rig up a decent heatsink on those VRMs. Probably work 10x better than a stock piece of poo poo.
|
|
#
?
Nov 19, 2017 18:09
|
|
|
I got the ASRock Pro4 mATX which probably has the same meh vrm cooling, but it was so cheap (just over $100 after rebate) that I will just go to a new z390 mobo when I want to try some exxxtreme o/cs after delidding etc.
|
|
#
?
Nov 19, 2017 18:50
|
|
|
FWIW the latest Asus Z370 Maximus BIOS update (802) completely changed the LLC slopes to have more vdroop. They also adjusted various secondary voltages (Vccsa, Vccio, Standby) to more sane values. The update wipes all saved profiles and loading a saved profile from a previous version from a USB stick is not possible. Best you can do is save your current settings to a USB stick in a text file (ctrl-F2 in the bios backup tab) and manually reenter them after the update. Blergh.
|
|
#
?
Nov 19, 2017 19:06
|
|
|
eames posted:FWIW the latest Asus Z370 Maximus BIOS update (802) completely changed the LLC slopes to have more vdroop. They also adjusted various secondary voltages (Vccsa, Vccio, Standby) to more sane values. The newest BIOS for the Strix Z370-F and E boards also fixes a whole bunch of stuff, including the issue I was having with XMP or memory frequency changes overriding some voltage settings. Everything working as intended now.
|
|
#
?
Nov 19, 2017 19:25
|
|
|
So the MSI Z370M and the ASUS Z370-G are the better MATX boards?
|
|
#
?
Nov 19, 2017 19:48
|
|
|
|
| # ? Apr 26, 2024 01:39 |
|
|
Ok I think it works. Getting Ubuntu loaded for preliminary testing before committing to the rest of the hardware.
|
|
#
?
Nov 19, 2017 21:55
|
|