|
I've got my vccio bumped up a bit with 3600 Ram on a skylake. I'll try it again at stock. (But the user agent has to be bumped, or it crashes out).
|
#
?
Apr 28, 2017 22:57
|
|
|
|
| # ? May 16, 2024 15:54 |
|
|
craig588 posted:IBT is not all that stressful or hot on modern processors. If you're hitting 100C there you really need to back off on some settings. What hits high 70s in prime95 for my Haswell only does in the high 60s for IBT. From what I've read several places you need to bump up VCCIO as well as VCCSA for memory stability and you generally want to have them within 10mV of each other. Having proper VCCIO and VCCSA settings can also supposedly help decrease the latency of RAM slightly, beyond just helping with stability so it's a good idea to do that even if it's already stable. Here is a handy chart for Skylake overclocking with VCCIO and VCCSA settings for different memory speeds:  It's about 2/3 of the way down on the right side. The chart is from this TweakTown Skylake overclocking guide and I did some testing with Aida64 and Geekbench 4 that showed slightly increased performance and decreased latency by using the settings they recommended. e: I decided to run the Aida64 and Geekbench 4 tests again to see what the differences in memory speed are between stock VCCIO and 1.2V at 3333MHz. I tested them each 5 times and took the middle result from each. Geekbench RAM speeds and latency were all within 2% of each other whether VCCIO was stock or 1.2V. Aida64 had the same results for RAM, but there was a 22% increase to L2 Cache read speed, an 8% increase to L2 write speed, and a 7% increase to L2 copy speed with 1.2V. There was also a 7% decrease in L3 Cache read speeds, while write and copy speeds remained within 2-3% with 1.2V. Stock VCCIO:  Single thread  Multithread  1.2V:  Single thread  Multithread 
Regrettable fucked around with this message at 03:07 on Apr 29, 2017 |
|
#
?
Apr 28, 2017 23:39
|
|
|
craig588 posted:IBT is not all that stressful or hot on modern processors. If you're hitting 100C there you really need to back off on some settings. What hits high 70s in prime95 for my Haswell only does in the high 60s for IBT. So with 16GB of DDR4 @ 3000 MHz I'd likely get away with leaving system agent voltage on Auto? I'm seeing about 100 C with about 6% throttling when I run IBT. Any other time, I haven't seen anything much above 80 C at 100% load. I'm worried I'd have to back off about 500 MHz to get reasonable temps with IBT when the system seems otherwise pretty drat close to stable for my purposes. I could turn things down a bit if 80 degrees is too hot. I'm still wondering about the importance of memory testing when using XMP and otherwise not touching any memory related settings.
|
|
#
?
Apr 29, 2017 01:47
|
|
|
zergstain posted:So with 16GB of DDR4 @ 3000 MHz I'd likely get away with leaving system agent voltage on Auto? As was previously stated, 100C is way too high for IBT and you'd run into serious issues when trying to test for real stability in prime95 as it will push it about 10C or more higher. There was someone who posted here just a few days ago saying they passed IBT with flying colors but prime95 would instantly shut down a core because of instability. That is not good and IBT is mostly used to see if you can continue with your testing to find an actually stable overclock. Also, based on the chart I put in the post right above the one I'm responding to you should put VCCSA at 1.15V for 3000MHz DDR4. I'm running at 3333MHZ and I used that chart to determine my VCCSA and it worked perfectly. Don't rely on auto voltage because a lot of motherboard manufacturers will have it use whatever voltage it thinks it needs to remain stable and can drastically overshoot what is actually necessary. e: If you decide not to take the advice that has been given here, I'm concerned that you are going to seriously damage your CPU on top of dealing with constant crashes. Regrettable fucked around with this message at 03:16 on Apr 29, 2017 |
|
#
?
Apr 29, 2017 01:56
|
|
|
I mentioned before that I've been using the x264 test to stress test. Specifically, the one from this thread, as it seems to be the recommended stress test.quote:x264 is the recommended and the default go-to stress test for this thread. If you feel the need to use a hotter test that is your right, but know that your overclock may be hampered by that choice. You could forego delidding in many cases simply by switching to x264. The downside to this method is that the overclocking process will take longer because we are replacing a very stressful program and a short test duration with a less stressful one and a longer duration. quote:My temperatures are through the roof! They also mention SA voltage and VCCIO, but only to the extent of try adding a little bit more if XMP doesn't work. I'll try boosting it to 1.15 V in the BIOS as recommended. I don't know how familiar people here are with that test, but last time I ran it temperatures hovered around 82 C and peaked at 84. I was able to run it for a good eight hours without crashing. I'm willing to accept that I might be asking too much of my Hyper 212 and back off a little bit. But I'd far rather base it on the x264 test than IBT. Or for another more real world test, after loading GTA 5 and driving around a bit, the CPU temp seems to be around 75 C, according to Afterburner.
|
|
#
?
Apr 29, 2017 16:54
|
|
|
I'm not sure I trust that guy. quote:if I made a program that crashes you at stock clocks, you would feel compelled to underclock your CPU, even if that application in no way represents real-world usage? You use synthetics to test what might be a worst case scenario because you have no idea how programs will change in the future. I'm downloading that 264 test to give some reference for how much it heats up and stresses my CPU. If it's really less than IBT it's really worthless because a lot of people already consider IBT worthless. (I still like it like how that chart shows, run it for like 20 minutes for a quick sanity check to see if it's worth continuing) Most people go the other way and would rather base stability on P95 than IBT, not going to progressively less and less stressful tests. Results edit: Bad news about that 264 test. After a loop it maxed out hitting 61C. That's not the full story either because my fans are locked to 20% speed until 65C, so IBT in the high 60s is with the fans sped up a bit and they hit 100% by 80C so prime95 hitting high 70s is with the fans nearly maxed out. It's practically just a step above idle. After a few more loops I'm sure it'd get a couple degrees hotter, but after a single P95 or IBT loop it's already a lot higher and spinning up my fans. craig588 fucked around with this message at 18:39 on Apr 29, 2017 |
|
#
?
Apr 29, 2017 18:32
|
|
|
craig588 posted:I'm not sure I trust that guy. You mean the CPU? If it were possible to write a program that would crash at stock clocks but run fine if you underclocked, that would be a hardware flaw, wouldn't it? I still have no intention of lowering my multiplier to the point where IBT isn't burning up, since for my purposes, this is mostly stable as it is. I did mention the temperatures I observed after playing GTA V for a bit in my last post, I suppose that's probably a bit too high, though as I understand things, and maybe if my CPU was running like that all day it would die an early death. Luckily I'm not using the computer like that. I tried prime95 with small FFTs just now. It gets up to about the same temperature as IBT, but I'm noticing far less throttling. Though maybe that's because one thread instantly fails. At this point, I'll probably just lower the multiplier down a notch or 2, and go back to playing my games. zergstain fucked around with this message at 01:59 on Apr 30, 2017 |
|
#
?
Apr 30, 2017 01:57
|
|
|
zergstain posted:for my purposes, this is mostly stable as it is. You're really getting into edge cases of luck when it comes to stability now. quote:You use synthetics to test what might be a worst case scenario because you have no idea how programs will change in the future. Playing a different game might cause it to crash instantly. You might end up triggering steam anti cheat for irregularities. You might end up with all sorts of weirdness from calculation errors that don't result in a crash but actually let apps continue running in an unexpected way. Crashing is actually preferred because it stops everything rather than risking unexpected writes doing something completely out of control. Usually people that have a "stable enough for me" idea is when it's crashing after 8 hours of prime but eh, that's starting to get into real edge case territory for their use. Crashing instantly with "hot" code is real bad because the second something hard happens you might crash.
|
|
#
?
Apr 30, 2017 02:37
|
|
|
craig588 posted:
How hot a test gets a processor is not a definitive indicator of whether that test is a valid stability test or not. Prime95 is using AVX instructions to do a few specific tasks that will get a processor about as hot as it's ever going to get, but it may not touch some possibly flakey transistors. I've had a couple systems that were "prime stable" but would crash playing games. On one of them, x264 would crash after a few loops despite being 20° cooler than Prime95. That doesn't mean you should use x264 to the exclusion of Prime95 or IBT, but it isn't worthless just because it runs relatively cool. Also 100C under IBT means your voltage is way too high or your cooler is insufficient or improperly installed. "One thread instantly fails" means your voltage is too low or overclock too high, regardless of whether you think it's "stable enough". Theris fucked around with this message at 03:22 on Apr 30, 2017 |
|
#
?
Apr 30, 2017 03:16
|
|
|
zergstain posted:You mean the CPU? If it were possible to write a program that would crash at stock clocks but run fine if you underclocked, that would be a hardware flaw, wouldn't it? I warned you about this exact scenario in my last response to you. That is still a very unstable overclock. Enjoy your constant headaches since you're just going to do whatever because you obviously know better than anyone else here. It makes me wonder why you were even asking for advice. Oh, that's right, because you were constantly crashing because your overclock was hilariously unstable, lol. Theris posted:How hot a test gets a processor is not a definitive indicator of whether that test is a valid stability test or not. Prime95 is using AVX instructions to do a few specific tasks that will get a processor about as hot as it's ever going to get, but it may not touch some possibly flakey transistors. I've had a couple systems that were "prime stable" but would crash playing games. On one of them, x264 would crash after a few loops despite being 20° cooler than Prime95. This is why you're supposed to add 10-15mV to your Vcore after passing prime95.
|
|
#
?
Apr 30, 2017 04:55
|
|
|
quote:Prime95 stopped and says there's an error. Okay, so I saw the rounding errors, then saw that, and figured rounding errors might not matter to me. craig588 posted:Playing a different game might cause it to crash instantly. You might end up triggering steam anti cheat for irregularities. You might end up with all sorts of weirdness from calculation errors that don't result in a crash but actually let apps continue running in an unexpected way. Crashing is actually preferred because it stops everything rather than risking unexpected writes doing something completely out of control. I was under the impression that VAC worked via fingerprinting. I'd be a little more concerned if I knew of any cases where someone was falsely banned due to system instability. Regrettable posted:I warned you about this exact scenario in my last response to you. That is still a very unstable overclock. Enjoy your constant headaches since you're just going to do whatever because you obviously know better than anyone else here. It makes me wonder why you were even asking for advice. Oh, that's right, because you were constantly crashing because your overclock was hilariously unstable, lol. Mostly I was trying to ask about the "real world" temperatures I was observing. I didn't entirely reject all advice that was given here either. The only thing I'm really refusing to do is drastic changes when to me it seems mostly useable. I'll admit that before I tried a different game, I hadn't considered these crashes were due to my OC config. They might be less frequent than you think though. At any rate, I have lowered the core multiplier and adjusted Vcore a bit. I'm using adaptive voltage. Hopefully that alleviates your concerns about me killing my CPU at least somewhat. I got prime95 to run without an instant rounding error now. And IBT is definitely hotter for me, by about 5 degrees.
|
|
#
?
Apr 30, 2017 15:47
|
|
|
I don't think anyone was worried about the CPU dying, it was more about having a miserable computer experience from random unexplained crashes.
|
|
#
?
Apr 30, 2017 16:56
|
|
|
Regrettable posted:e: If you decide not to take the advice that has been given here, I'm concerned that you are going to seriously damage your CPU on top of dealing with constant crashes.
|
|
#
?
Apr 30, 2017 17:34
|
|
|
That was assuming you were somehow stable enough to run a long prime95 stress test, since you were already hitting 100C in IBT and some versions of prime95 can push temps 10C+ beyond that. Skylake shuts itself down at 110C to try to prevent damage so you could have been getting very close to that point. Also, damaging a CPU is not the same as killing it outright. I'd be more worried about voltage doing that than temperatures. More likely it would start having stability issues and degraded performance. Sorry, I know my last post was pretty snarky, but I was starting to get frustrated with some of the responses to what I considered good advice in this thread.
|
|
#
?
Apr 30, 2017 21:33
|
|
|
Regrettable posted:That was assuming you were somehow stable enough to run a long prime95 stress test, since you were already hitting 100C in IBT and some versions of prime95 can push temps 10C+ beyond that. Skylake shuts itself down at 110C to try to prevent damage so you could have been getting very close to that point. Also, damaging a CPU is not the same as killing it outright. I'd be more worried about voltage doing that than temperatures. More likely it would start having stability issues and degraded performance. Sorry, I know my last post was pretty snarky, but I was starting to get frustrated with some of the responses to what I considered good advice in this thread. Isn't temperature how voltage does damage though?
|
|
#
?
Apr 30, 2017 23:36
|
|
|
GutBomb posted:Isn't temperature how voltage does damage though? Sometimes, but it also increases electromigration which causes resistance to build up so it requires more voltage to maintain the same clocks and can eventually render hardware completely unable to function because of broken transistor interconnects. A combination of high heat and high voltage increases electromigration even more. Regrettable fucked around with this message at 00:22 on May 1, 2017 |
|
#
?
May 1, 2017 00:20
|
|
|
GutBomb posted:Isn't temperature how voltage does damage though? Higher temperatures increases the amount of electromigration. But there's always some temperature there, even at ambient with voltage it'll occur.
|
|
#
?
May 1, 2017 03:24
|
|
|
I've been messing around with my 5820k, when I built this pc I just set it to 4.1 at 1.050v which was nice and stable after testing. I bumped it up to 4.4 at 1.250v which worked but it got very hot, so I backed down the voltage while testing until it became unstable which lead me to 4.4 @ 1.170v. Temps in realbench and IBT are around 80c with ibt being a little bit more, this is in a fairly cramped case using an NH-U14S, i'm just wondering if it's worth upgrading the cooling as my case can just about squeeze an aio with a 240mm rad in at the top.
|
|
#
?
May 1, 2017 18:05
|
|
|
More cooling will always help with getting faster and staying cooler, but have you considered staying at just 4.3GHz? You could probably do 4.3 with your existing cooling and keep temperatures cool. 4.4 would probably be possible with good temps with better cooling, but that's probably going to be at least a 100 dollar upgrade for maybe a few percent performance increase.
|
|
#
?
May 1, 2017 18:46
|
|
|
Yeah i'll give that a go, I can probably bring the voltage down a bit more at 4.3 too.
|
|
#
?
May 1, 2017 19:06
|
|
|
So, I put back the underclock on my 7700k (1.08v @ 4200mhz), saved changes and rebooted my ASRock Z270M-ITX, and now I can't post. I've tried every input on the monitor, every input on the GPU, removing the GPU and using motherboard HDMI, I absolutely can't get an image of anything, BIOS or otherwise. When I turn on the PC the fans spin and USB stuff lights up but that's it. I held down the CMOS clear jumper with the power unplugged and plugged in too, that doesn't help either. Pretty stumped here.
|
|
#
?
Jun 16, 2017 05:02
|
|
|
I've had to actually take the CMOS battery out, unplug the power cord and turn the PC on, then wait 10 minutes before the drat cmos would actually clear.
|
|
#
?
Jun 16, 2017 06:53
|
|
|
JnnyThndrs posted:I've had to actually take the CMOS battery out, unplug the power cord and turn the PC on, then wait 10 minutes before the drat cmos would actually clear. Yeah I'm officially pissed, I've left the CMOS battery out all day, used the reset jumper, changed power supplies, reseated the RAM and all PSU cables on the PSU and motherboard side, tried with and without the GPU completely removed, tried all GPU HDMIs and DisplayPort, tried both motherboard HDMI, and tried all the inputs on my monitor. When I power on, the fans for CPU and GPU spin up and stay on and all USB stays lit, but absolutely nothing will get me a picture.
|
|
#
?
Jun 17, 2017 03:46
|
|
|
I have an i5-760//Asus - P7P55D-E Deluxe, and because I don't want to splurge on an Intel chipset and socket that may not be futureproof, I want to see what I can do with overclocking. Another alternative is getting a Xeon CPU, but I haven't seen benchmarks that makes it stand too much apart from my current one. And if I have to go through the same overclocking hoops, I might as well give it a shot with this one. Not that I'd mind paying $79 to boost performance significantly if that's an option. I've read these two guides as well as the OP so far: * http://www.masterslair.com/vcore-vtt-dram-pll-pch-voltages-explained-core-i7-i5-i3/#cpu-phase-locked-loop-pll * http://www.techreaction.net/2010/09/07/3-step-overclocking-guide-bloomfield-and-gulftown/ One of the things that trip me up is seeing the OP saying to not, for the love of god, tamper with PLL, while the first link above encourages me to use it - albeit to just get some slight extra mileage out of the CPU. Another is that Vcore and VTT seem to be conflated. From what I understand, with my Lynnfield, BCLK*Multiplier = CPU speed, and I can only change the BCLK in this case. When an OC doesn't work, I should step up the Vcore, but do I increase VTT by the same amount? On top of that, what really complicates the picture for me is that I have to tamper with stuff like PLL Turbo Boost, CPU ratio, memory speed, QPI frequency, CPU Spread Spectrum, and set VTT to a default value. On top of that there's managing memory when I step up the BCLK. I'm not really interested in improving the memory, just my ancient CPU. If it were just a matter of increasing a multiplier and boosting Vcore with 0.025V every time I ran into an issue up to 1.35V, I wouldn't have any issue, but the discrepencies and conflated language makes it all a little confusing to say the least. Basically, the steps seem to boil down to: 1. Disable or lower a bunch of BIOS settings 2. Set some defaults for now 3. Step and increase voltage until I reach the goal 4. Re-enable or raise some of the settings from #1 The second link is not hard on the theory; it's more of a cookbook I can just follow depending on how aggressive I'm feeling, but I'm basically just mindlessly following orders with that. I don't necessarily mind that, as long as there's no risk attached. ufarn fucked around with this message at 18:00 on Jun 17, 2017 |
|
#
?
Jun 17, 2017 17:57
|
|
|
Zero VGS posted:Yeah I'm officially pissed, I've left the CMOS battery out all day, used the reset jumper, changed power supplies, reseated the RAM and all PSU cables on the PSU and motherboard side, tried with and without the GPU completely removed, tried all GPU HDMIs and DisplayPort, tried both motherboard HDMI, and tried all the inputs on my monitor. I just changed out my motherboard with an entirely different brand and I'm still getting the same thing, fans spin up but no video. I get video using the same HDMI cable and monitor if I plug in a different PC, so the only component left is the CPU. Is it possible to fry a 7700k with an undervolt+underclock? Edit: unlike my old motherboard this one has POST LEDs and they are indicating a CPU issue. So yes, I apparently undervolted my CPU to death, 1.08v was too low for it I guess... Zero VGS fucked around with this message at 21:19 on Jun 17, 2017 |
|
#
?
Jun 17, 2017 20:58
|
|
|
Got a new 7700k and we're back up and running. Now I'm curious how an undervolt could have fried a 7700k in the first place. The only thing I can think of is maybe if the voltage was too low, the capacitors on the CPU tried to compensate and self-destructed instead?
|
|
#
?
Jun 18, 2017 01:21
|
|
|
That sounds really weird and I've never heard of undervolting breaking a CPU (not booting or crashing though, sure). Maybe you just had a bad chip.
|
|
#
?
Jun 18, 2017 04:18
|
|
|
Less Fat Luke posted:That sounds really weird and I've never heard of undervolting breaking a CPU (not booting or crashing though, sure). Maybe you just had a bad chip. Yeah I mean I hit apply on 1.08v and bam couldn't even get the BIOS screen anymore. By the way, on the new motherboard and CPU I tried to apply 1.08v, I saved and booted to Windows, and get this:  CPUID says "VID" is hovering around 1.17v, but CPU-Z says Core Voltage is around the 1.08 I set. Are those two different values?
|
|
#
?
Jun 18, 2017 05:20
|
|
|
I believe VID is the stock/default voltage and Vcore is what you've set yourself.
|
|
#
?
Jun 18, 2017 06:01
|
|
|
Check under motherboard readings in HWMon. Other software will also report vcore under mobo as well
|
|
#
?
Jun 18, 2017 07:40
|
|
|
Intel was pretty cool about warranty over the webchat, I didn't keep the box with the serial, but there's a "FPO" number on the chip that they used to verify it. They said I can mail it at my expense, or pay $25 to have a working 7700k cross shipped to me, if anyone was wondering. They made me walk through a basic troubleshooting script with them, didn't ask me if I overclocked or not (which I technically didn't since I only underclocked).
|
|
#
?
Jun 19, 2017 22:01
|
|
|
ufarn posted:I have an i5-760//Asus - P7P55D-E Deluxe, and because I don't want to splurge on an Intel chipset and socket that may not be futureproof, I want to see what I can do with overclocking. Another alternative is getting a Xeon CPU, but I haven't seen benchmarks that makes it stand too much apart from my current one. And if I have to go through the same overclocking hoops, I might as well give it a shot with this one. Not that I'd mind paying $79 to boost performance significantly if that's an option. You are correct that BCLK * Multiplier = core frequency, and there's the additional wrinkle that BCLK* RAM multiplier = RAM frequency and BCLK * QPI multiplier = QPI frequency. I haven't worked with any ASUS P55 boards, just an Intel DP55WG and a Gigabyte X58 board which is a pretty similar platform to overclock, but it looks like your Vtt is what other systems might call the Uncore voltage. You need to increase this if the memory controller or QPI can't keep up, but the former shouldn't be an issue - your RAM's speed will limit you before the controller more than likely. However the QPI speed will increase in lockstep with the core unless you can lower its multiplier (I didn't have the option, except for "slow mode" which is ruinously bad) and you may need to bump Vtt to keep it stable especially as you pass 166+ BCLK and head towards 180-200. The second link you gave in particular is a pretty good guide and the voltages it gives are safe in my opinion - 1.4V for CPU/Vcore and Uncore/Vtt. Here is what I did: 1. Find roughly the fastest core speed stable at stock voltages. Adjust RAM multiplier to keep the frequency within spec and minimize instability; you can optimize the RAM later. Feel free to just set it all the way down so you don't have to think about it until you're done with the core, if you prefer. Do the same with QPI if you can - but not "slow mode". 2. Begin to step up gradually, maybe 5-10 to BCLK at a time, and if it's not stable first try boosting Vcore a bit and then Vtt separately - you'll get an idea which is your limiting factor as you go. While feeling out the curve like this I don't usually do intensive stability testing, just boot and run something for 5 minutes to make sure it doesn't bluescreen. Continue to keep RAM speed in check by lowering the multiplier as you go. 3. Once you get to whatever voltage you consider your limit, you'll have an idea what the curve looks like and what speed you want to try to keep - if you can get 3.8GHz at 1.25V and it takes 1.4V to get to 4.0, you may decide that you'd rather have the lower speed to keep it cool and quiet for example. 4. At this point you can figure out if your RAM is stable at a higher multiplier if you wish. I personally don't mess with DRAM voltage much - I'm sure a little bit is OK just like anything else, but getting faster RAM usually seems like a better option. 5. Make sure you do at least one lengthy stability test with whatever you end up on, of course. By doing this I was able to get an i3-550 (3.20/24x) up to around a BCLK of 180 if I remember correctly, so ~4.4GHz, without exceeding 1.4V. I then replaced the i3 with a Xeon X3440 (2.53GHz/19x with turbo to 21x/20x on one/all cores) and was able to get that to about 185/~3.8GHz + turbo, but even disabling turbo wouldn't allow me to push higher than 190 so that may have been the limit of the motherboard. It's probably quite possible to get your i5-760 to a similar range, anyway. e: It's also possible to just approach it in reverse - start at whatever Vcore/Vtt you think is the max you'd be comfortable with, figure out what CPU clock speed you can make stable at that while spiking RAM multiplier so that it's a nonfactor, and then tune the RAM and voltages further from there if you want. It might be a faster approach if you just want to get as fast as you safely can and don't care what the curve in between looks like. Eletriarnation fucked around with this message at 17:53 on Jun 20, 2017 |
|
#
?
Jun 20, 2017 15:50
|
|
|
It should be noted that Nehalem is pretty much the only Core platform it's safe to crank the BCLK like that on. Anything newer, you've maybe got two or three MHz you can squeeze out of the BCLK before the system becomes really unstable due to the BCLK also being used for the PCIe lanes. And on SB-E and IB-E, you have the magic that is BCLK Strap overclocking, which changes your CPU clock formula to be BCLK * BCLK Strap * Multiplier.
|
|
#
?
Jun 20, 2017 17:07
|
|
|
I upgraded from the 6600k to the 7700k, and decided to sacrifice the 6600k and practice delidding for the very first time. Put the 6600k in an old motherboard, let it run at stock settings. Both Intel Burn Test and Prime95 would almost immediately shove the poor bastard up to ~90C. (92 highest reported) This was using an old Hyper 212 Evo with two fans on it running at 100%, open-air test bench. I was going with the razor blade method, and this was pretty tough. Even added some blood to my sacrifice on the altar of science when I nicked myself with the razor. With all the grunting, scraping, blade slipping, and scratching sounds, I was already pretty much 100% certain the thing would be dead as disco when I pushed the start button. My concerns were more or less confirmed when the startup gave me the message "New CPU detected! Press F1 to enter bios" or whatever. It's the same CPU, computer! Must be screwed. Load the bios, everything at default, reset. Monitor does a weird flicker, like a boot loop. Yep. Killed it. Suddenly I'm looking at the windows login screen. Holy gently caress is this thing working!? 15 minutes into IBT and the highest temp reported across all cores is 66. Now, this is just at stock settings, but it is an IBT load that was ~25C higher before the delid. Jesus christ is this normal!? I can't wait to see what this thing will do on water.
|
|
#
?
Jun 22, 2017 01:35
|
|
|
Deuce posted:I upgraded from the 6600k to the 7700k, and decided to sacrifice the 6600k and practice delidding for the very first time. How high can you get it on air now?
|
|
#
?
Jun 22, 2017 01:39
|
|
|
GRINDCORE MEGGIDO posted:How high can you get it on air now? Sadly, this chip's problems getting past 4.6ghz seem to persist. 4.6 will run at 1.35v. 4.7 is hard to stabilize, not stable at 1.385. (part of the reason I was willing to sacrifice it, apparently it's a fairly mediocre chip for skylake) But I'm not an expert chip tweaker. I just set core voltage manually and the multiplier and try a test edit: 1.4v failed as well. Deuce fucked around with this message at 02:49 on Jun 22, 2017 |
|
#
?
Jun 22, 2017 02:41
|
|
|
I think I might get down to overclocking somewhere in the weekend, because I think I've got it sorted out just about. First, I find my max BCLK by 1a. turning down frequencies of RAM and QPI (to keep BCLK from messing with things) 1b. lower my CPU multiplier 1c. start overclocking BCLK and adding voltage to the Vcore (to maximum of 1.35V) until I hit a wall Then I find my ideal BCLK by 2a. disabling Load Line Calibration 2b. resetting BCLK to the original default 133 (or 150). 2c. reset CPU multiplier to original pre-OC 2d. keep Vcore at the level from before 2e. start overclocking BCLK and adding voltage to Vcore/IMC (to maximum of 1.35V) until I hit a wall After finishing 3a. reset frequencies of RAM and QPI to original (as close as possible) Here's the one thing I'm not sure about. Some guides say I'll "just know" when my memory voltage becomes a problem, but I don't see where outside of keeping it within 0.5V of the VTT/IMC I figure this out, so how does the gradual fine-tuning of the memory voltage work, and when does it take place in this? From the get-go, I'll set all voltages except PLL and PCH to Manual, ie CPU, VTT/IMC, and DRAM. PLL and PCH will remain at Auto. If that's not possible, they'll all obviously just be set to Manual with no changes to PLL and PCH. ufarn fucked around with this message at 13:16 on Jun 22, 2017 |
|
#
?
Jun 22, 2017 13:12
|
|
|
So I'm pretty bored with my i5-4670k running at stock. I wanna pump it up to 4ghz, maybe higher but 4 sounds easy to achieve without turning my box into the sun. One concern is my ASRock Z87 extreme4 board has these default oc profiles that change a buncha settings in the menus. Beyond these funky new settings that I've not seen before, I'm wary of trusting the OC profiles. I've never seen these before! Preliminary googling has said that they're not to be trusted for stability and heat concerns. Weird they built em in then. One of these settings that is changed when I compare differences is the Cache multiplier. Some googling really didn't reveal many results on that. I've never heard of this one before, but it's a pretty central value in my board. From my research, changing the vcore to manual and the clock multiplier should be enough. I assume it's still recommended from ye olde generations to disable speedstep, turbo boost, and all that other stuff, but I've also read conflicting information that nah, it's fine. So which is it? I also heard Prime95 isn't the go-to for these types of chips. What is the preferred stresser? This is a gaming box, not a video conversion box, mind you. I don't think I need a super burner. Do I? The guide in the OP was pretty slick, as well as some 'quick' OC guides I've looked at about Haswell, but I'm curious about those specific concerns in case information I've read in guides that are mostly many years old is dated.
|
|
#
?
Jun 23, 2017 05:56
|
|
|
ufarn posted:I think I might get down to overclocking somewhere in the weekend, because I think I've got it sorted out just about. This all sounds good to me. I am not sure if you'll be able to get enough real benefit from overvolting your RAM to be worth the risk of going very high. How much RAM do you have, and of what speed?
|
|
#
?
Jun 23, 2017 06:05
|
|
|
|
| # ? May 16, 2024 15:54 |
|
|
Eletriarnation posted:This all sounds good to me. I am not sure if you'll be able to get enough real benefit from overvolting your RAM to be worth the risk of going very high. How much RAM do you have, and of what speed? I don't want to tamper with the RAM at all, just keeping it the req 0.5V from VTT. The CPU is by far the bottleneck. Assuming that's how it works; I assume picking the original frequency in BIOS will keep around factory settings, different BCLK aside. I assume the frequency setting in BIOS is a product of BCLK and a hidden multiplier.
|
|
#
?
Jun 23, 2017 14:16
|
|