![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
Such a shame... I mean they are expert in this regards, why they would let users do better? Because it's a risky operation and if something goes bad it's not warranty and they sell another CPU? Maybe next time they will solder like it was done before preventing people to play with it. :-/
Not at all. It's about cost and nothing more. It's a few pennies to us and potentially huge for them over the total production lifetime of a given CPU family. Not only that, but the TIM Intel uses was spec'ed to do a job at the lowest cost possible. It meets and even exceeds the thermal needs of the CPU. Overclocking doesn't factor in at all and we are the only ones who would benefit from the better TIM.
It's still a -K so overclocking should factor in... but you say TIM exceeds thermal needs and maybe they are happy with it as it is now.
It definitely exceeds the thermal specifications of the processor. If it didn't you would be closer to the TJunction max and see thermal throttling at close to stock speeds. You have to understand that unlocked processors are for the enthusiast community, but Intel doesn't care if you can get an extra 200MHz out of a better TIM. They do not want to reduce their profit margins on those CPUs.
Asus did a podcast with the guys from Intel and the issue of TIM was discussed, in some detail. As was the use of liquid metal, alternative TIMs, tolerances, all that stuff. They have to consider thermal pumping from thousands of thermal cycles (the die expands/contracts and pumps the wrong TIM out, leaving an air gap), having the stuff run out of the gap due to mounting/movement, lots of stuff end users never have to consider. The extreme overclockers are such a small segment of the market, they just aren't on the radar- except the X series processors that still get soldered heat spreaders.
Interesting! Do you have the link?
I use adaptive and the Asus guide is really helpful. I'd have never found the IA AC/DC Load Line settings without it, and without it your voltages will overshoot your turbo voltage by .04v or more.
My workload was:
- Set multipler to 46, sync all cores.
- Set voltage to adaptive, turboV to ~1.2v
- Set IA AC/DC LL as mentioned above.
- Set RAM to XMP voltage, freq and timings manually (leave sub timings on auto)
- Set LLC to 2
- Run the stress test that causes the most vdroop (ROG RealBench for me). See if it crashes, and at what avg voltage.
- If stable, increase multipler and try again.
- If almost stable or vdroop too severe, increase LLC.
- If unstable and minimal vdroop, increase turboV.
Repeat until thermal limit is reached (or if you have a beefier cooler than me, the voltage limit). Then run a mixed set of stress tests and see if it's still stable under varied workloads.
After tweaking on and off for a week, I hit 4.7GHz at 1.232v and LLC5 with Max temp of ~80c. They key is that vCore never drops below 1.216 under Max load.
Under manual, I'm stable at 1.216v and LLC3 but the temps are 1-2c higher.
I could probably push to 4.8, but I'm waiting until it gets hotter here to see how much headroom I have.
I don't use prime95 or synthetic stress tests that use an unrealistic level of AVX2 instructions (and consequently an unrealistic level of heat). If you do, and temps are too high under those tests, I'd use the AVX offset to handle that. If you need a setting higher than 2 or 3, maybe back off your OC.
Last edited:
nvgrim
[H]ard|Gawd
- Joined
- May 1, 2013
- Messages
- 1,747
I have a Z170 Impact thats been giving me issues since day one, i wonder if i should get this one. Im starting to hate ASUS components nowadays though for their hit or miss reliability.
I've reviewed a huge number of ASUS motherboards over an 11-year period. I can count on one hand the number of failures I've experienced. Out of all the mini-ITX motherboards I've looked at, ASUS' have been the best by far. The only one that's come close is the ASRock X99 mITX board.
nvgrim
[H]ard|Gawd
- Joined
- May 1, 2013
- Messages
- 1,747
The only issues ive had are weird ones with the BIOS. My current pc will boot into the BIOS on a cold boot no matter what you do, i have to ctrl+alt+del to restart the pc to boot into windows. Done everything i can and no fix. Also making F8 the boot load menu button is the most frustrating thing ever when trying to boot into safe mode.
Did you check the setting for the ASUS DirectKey feature? There is a button on the PCB for this and a setting in the BIOS related to this. I wonder if you have something laying against the board on that key, or if the DirectKey setting is set to boot direct to the BIOS. I believe you can do that. It's been awhile since I paid any attention to that specific setting.
Okay, turns out 0704 did *not* fix the cold boot bug. In fact, on occasion the thing will fail to post even when powered on. Upping voltage to the VCCIO and System Agent helps. Mine are both set to to achieve an actual voltage of 1.2 (settings are a bit less). It will fail cold boot 50% of the time, but when powered on works 100% again. At this point it's "good enough" for now. I could probably up the voltage a bit more and see if it fixes for good or makes it worse because too much V will mess with the signal, but meh.
Are you using XMP or have your RAM manually set to a similar speed? If so, try the following:
1) Power off computer completely. Switch off power supply or unplug power cord.
2) Wait 1 minute or more. (Problem can manifest without wait time, but the longer power down periods increase odds).
3) Plug back in/switch on PSU. Orange "OK" LED should light up on top of board, RGB LEDs on bottom should turn on if enabled. Press start/boot button on chassis immediately.
Repeat process 3 times. Per my observations, it should fail at least 2 of those times. If it powers on all 3 times, please share your build and BIOS settings you've changed.
Rig is down right now, undergoing conversion to water, but manual RAM settings, no XMP. I've powered down numerous times to "dead board" status (all power off, no LEDs lit) to clear cmos and never failed to boot back up. As I hardly ever totally power down (unplug PSU from the wall) during normal use, I'm not sure the bug you mention would ever manifest itself. I'm sure I'll have more chances to test it though once I commit to actual overclocking the rig.
Oh man is this making me want to make the terrible journey back to custom water: EK Monoblock for Strix Z270i, cools the CPU and MOSFETs. Has internal RGB LEDs that sync with AURA.
I haven't tried a fan on the pump header, but I will note that the pump header doesn't get the same graphic interface in bios, just high/med/low set points. I bet a fan will work on it since there are only 3 headers and not everybody watercools, but I can't back that up. The other 2 fan headers do accept pwm fans.
The issue is part MB, part weak memory controller on your CPU, which is why it doesn't happen to everyone. Set memory frequency and timings manually instead of using XMP auto setting. Increase CPU SA (system agent) voltage to ~1.2. or whatever it will let you bump it to before the number turns yellow on the screen.
Should fix that most of the time. If it still happens, just turn the PSU off and then tap your power button 3 times in a row. Then turn power back on and start normally. 3 buttons with power off will cause MB to boot into safe mode, which uses default settings (in particular: RAM). All you gotta do is hit F1 to enter BIOS then F10 to exit and it will boot normally with your OC settings.
Also, do not use the SSD heatsink on top of the board. If your SSD is cooking, just slap on some self-adhesive heatsinks from Amazon or eBay.
I used these and have temps 10c lower than when I used the built-in heatsink. Also, remove the sticker from the SSD unless the OEM says it will void the warranty. Or do it anyway and live without the warranty. Who am I to tell you how to live? I have a Crucial SSD, and they say jack about the sticker, so off it went.
https://www.amazon.com/gp/product/B00P6Z2BQU
As an Amazon Associate, HardForum may earn from qualifying purchases.
It doesn't matter. SSD is more or less the same temp with or without it. Get some stick-on heatsinks if 48-50c is too hot for your comfort.