|
Whoops, didn't realize that's what it was - my mistake. Yeah, 2x8GB DDR3-1600 is not a bad place to be even if you leave it at stock - a lot of systems from that generation would use 1333. You may realize, but to be clear the 0.5V is a maximum so you definitely don't need to keep it that far away - e.g. if your Vtt is up to 1.4V, please don't set the RAM from the stock 1.5 up to 1.9. This warning is mostly for older DDR3 which used 1.65 - you shouldn't use this stuff with the newer generations of DDR3 processors because if your Vtt is way down around 1.0 the large voltage gap can damage the memory controller. With boards I have worked with, RAM frequency is a product of BCLK and a separate, shown multiplier which you can also set - usually it starts at 6x or 8x and goes up in 2x (because it's DDR) steps. I don't think there's a way to set the RAM frequency in a more granular way due to the need for synchronization with BCLK. Your 1600MHz sticks would be 133x12 at stock, so if you set them down to 8x then you can go all the way up to a BCLK of 200 before you have to overclock the RAM just to keep going. Keep in mind that you may have to change latency too, depending on what your motherboard set as defaults once you set RAM timing to manual mode. It's been easiest for me to just take a picture with my phone of the SPD values in hwinfo or CPU-Z or whatever when the system is booted, and then I have a handy reference for what values are in spec while I'm looking at BIOS.
|
#
?
Jun 23, 2017 16:50
|
|
|
|
| # ? Apr 24, 2024 21:13 |
|
|
Eletriarnation posted:Whoops, didn't realize that's what it was - my mistake. Yeah, 2x8GB DDR3-1600 is not a bad place to be even if you leave it at stock - a lot of systems from that generation would use 1333. My DRAM frequency is changed directly in BIOS (same with QPI) AFAICT:  I don't know how exactly these figures are produced, and I wonder if they change when I mess with BCLK. But looking at that, it doesn't include 1600, so I'm actually not sure what is implied with "Auto". Looks like it's specifically for downclocking the RAM, but no idea if "Auto" will stick with the same frequency or the original multiplier and mess things up when I increase BCLK. e: Looks like another DRAM setting may open up once you enable overclock mode: https://www.youtube.com/watch?v=QFw_BPvaa5o. e: But the point being that the DRAM and QPI settings have an Auto option and some other frequencies that seem to be lower than the default clock (ie no 1600 option for DRAM), so I wonder what choosing "Auto" after I'm done overclocking would "revert" them back to. At least this looks like the only thing I have to figure out before proceeding. e: Looks like I can use the advanced settings to at least double-check on my RAM speed when I have to reset things by the end. I assume the AI Tweaker is just a simplified interface for this stuff. Now I just have to make sure I'll do right by QPI. ufarn fucked around with this message at 18:35 on Jun 24, 2017 |
|
#
?
Jun 23, 2017 19:23
|
|
|
My bet is that Auto for the RAM frequency setting is picking the highest SPD bin the motherboard supports, which I assume for your RAM is 1600MHz. Apparently they didn't feel like they needed to include a way to hard-set a bin higher than 1333MHz, at least in that UI mode. Often they'll hide more detailed settings like the multiplier unless you change some stuff from 'Auto' to 'Manual' like maybe CPU Level Up or Ai Overclock Tuner in what I see there - whatever gives you access to BCLK. I would bet that with that change, the RAM speed setting will also change to a multiplier instead of a frequency. If the interface is poorly designed it might continue to say '1333MHz' etc. but really mean "...at base BCLK", so x10. If there is still an "Auto" option at that point it will probably just scale up that highest SPD bin's clocks until it doesn't work anymore, if I had to guess. Hopefully it'll give you a multiplier since that will be simplest. I don't think that it's possible for the motherboard to hold the RAM at a given frequency while the BCLK moves around in any case, RAM and CPU frequency both have to be an integer multiple of that speed. Eletriarnation fucked around with this message at 21:00 on Jun 24, 2017 |
|
#
?
Jun 24, 2017 20:57
|
|
|
Deuce posted:I upgraded from the 6600k to the 7700k, and decided to sacrifice the 6600k and practice delidding for the very first time. I own both a 6600k and a 7700k and I think I'm gonna bite the bullet and delid, thanks for the write-up.
|
|
#
?
Jun 24, 2017 22:33
|
|
|
MaxxBot posted:I own both a 6600k and a 7700k and I think I'm gonna bite the bullet and delid, thanks for the write-up. I'd suggest a delid kit and nonconductive paste. Lots of people use liquid metal for a couple extra degrees but that seems like a terrible idea.
|
|
#
?
Jun 25, 2017 20:56
|
|
|
Any reason why I shouldn't go with RealTemp instead of Core Temp? Core Temp has a bad habit of freezing temporary during stress tests. Is there also a good way to gauge whether you did a decent job of applying the paste? I'm getting the 38--41C with fans on lowest speed, and I think I'm getting the same idle temps as pre-re-application, for better or worse. ufarn fucked around with this message at 10:52 on Jun 29, 2017 |
|
#
?
Jun 28, 2017 18:08
|
|
|
This is probably the best thread for this... I found a small youtube channel that tested a whole lot of fans and made audio calibrated videos of them. Very useful to compare how CPU/case fans sound at various voltages and throughput levels. It certainly saved me from buying a bunch of Noctua 3000RPM industrial fans.  https://www.youtube.com/user/CoolingTechnique/videos there's also this calibration video, you're supposed to set your volume so that you can barely hear the background noise. https://www.youtube.com/watch?v=iL5AlyDOEds
|
|
#
?
Jul 2, 2017 12:36
|
|
|
I use those 3000 RPM Noctua fans pretty much exclusively as long as the budget permits. They can PWM down to 600RPM and I feel like using a high speed fan slowly is better for longevity, but beyond that they serve as an alarm, I have them set to ramp up to full speed by 75C so I know if I've gotten a bunch of dust clogging my heatsink or something weird happened with the contact to the CPU. The normal fans are fine to run at 100% speed all the time and you can barely hear them, but it's nice to have the headroom to ramp up 3000 RPM fans for testing instead of having fans already maxed out. I really, really like them, the only problem is their cost, you can get 90% of the way there in terms of noise to air moved with fans that cost half as much so they're hard to recommend for people building something cheap.
|
|
#
?
Jul 2, 2017 15:22
|
|
|
I just replaced my old 2500k with a 2700k which is just a binned 2600k to my understanding. I was able to get it up to 5.048 GHz Vcore 1.352 Are 2700k binned that well or is this a golden chip?
|
|
#
?
Jul 8, 2017 03:11
|
|
|
It's pretty golden, I was only able to get 4.7 on a 2700K at 1.32v, which was about par for the course at the time.
|
|
#
?
Jul 8, 2017 03:25
|
|
|
Definitely above average but Sandy Bridge overclocked well, 5GHz wasn't super uncommon. Mine could hit 5 but it needed a lot of voltage and my cooler wasn't up to the task.
|
|
#
?
Jul 8, 2017 03:31
|
|
|
Fauxtool posted:I just replaced my old 2500k with a 2700k which is just a binned 2600k to my understanding. That's good. My 2500K did 4.9 at 1.35V. Shame my cooler wasn't up for it.
|
|
#
?
Jul 8, 2017 03:59
|
|
|
The way I did it was pretty janky. I did the bios easy OC which gives you 4.3 then with all other settings the same I put LLC at medium and set multiplier at 49. The stats I pulled are from when im stress testing. The vcore isnt spiking any higher than 1.352 so i feel safe I cant figure out what settings were used because when I OC the "right" way with 24/7 turbo and a locked voltage I cant get it anywhere near stable
|
|
#
?
Jul 8, 2017 04:04
|
|
|
I’m currently overclocking, and everything’s been going smoothly - a little too smoothly. My OC to 200 BCLK finally failed, and I went into BIOS to up the voltage. I go into my BIOS, and it’s reset my CPU voltage setting to Auto, and the current CPU voltage is currently at 1.36V, which is above my max, but not life-threatening. What’s weirder, though, is that even as I set it to Offset and +0, it changes the offset voltage to 0.00625 (basically 0), but shows the current CPU voltage as 1.36V still. Has my BIOS become self-aware, and what do I do about it? I was able to hit a max BCLK of 220@1.225V, and my CPU temps are currently 68C with all fan speeds set to low-ish, so it’s not like there’s not more headroom, so what’s going on with the CPU voltage? I assume the Current Voltage figure updates dynamically, so why won’t it go down? It doesn’t change as I change the voltage offset, though. e: Tried setting it to a manual offset and lowered it to 190 BCLK, and now it’s stuck at loading “ASUS EXPRESS GATE”; so guess I got to reset the CMOS. e: What I did beforehand: Lower DRAM and QPI. Find max BCLK with CPU multiplier at 12. Raise CPU multiplier to Auto, set it to Offset +0. Benchmark. At the point of reaching 200, I hadn’t had to raise my voltage. e: Things went kind of south, but nothing a CMOS reset and System Restore couldn’t fix. Still have no idea what the hell went wrong. Guess I can try again, but seems like something for vcore automatically reset or configured itself in BIOS. e: Got everything back to normal, and BIOS definitely resets my Offset settings. It doesn’t reset as soon as I restart, so it probably reacts to an increased load from my stress tests in Windows 10. e: I’ve more or less got it under control, but something’s just wonky about the interface. Think I’ve ended up with 180 BCLK vs 133. Assuming the offset, +0.075V, is added to the “Current CPU voltage”, I should be right below 1.35V, although I’m well below the potential maximum of 220 BCLK I managed. Only hit 62C with minimal cooling. e: It's now stable at 170, but I'm not sure how to turn back up DRAM and QPI. Here are the current BIOS settings and their original frequency pre-OC: https://imgur.com/a/C2AsS     I still don't QPI, but either way, it seems like I'd be better off finding out what it's supposed to be and if there's somewhere I can tweak a multiplier to bring it and DRAM closer to whatever the QPI was before. Pre-OC, the DRAM options were Auto, 800, 1066, 1333, and QPI Auto, 4270, 4800. ufarn fucked around with this message at 12:55 on Jul 15, 2017 |
|
#
?
Jul 14, 2017 14:05
|
|
|
Kazinsal posted:I believe VID is the stock/default voltage and Vcore is what you've set yourself. VID is a preprogrammed table of voltage/clock frequency pairings. You can't change it. There is a second VID table for voltage/turbo clock frequency pairings. You can think of VID as the voltage requested by the cpu for a given clock frequency and Vcore as the voltage delivered. Mobo auto voltage is sort of based around VID with an optional manual offset. I'm pretty sure adaptive Vcore applies to turbo clocks only.
|
|
#
?
Jul 15, 2017 20:54
|
|
|
ufarn posted:e: It's now stable at 170, but I'm not sure how to turn back up DRAM and QPI. Here are the current BIOS settings and their original frequency pre-OC: Your RAM was running at 1333 before - it says 661 in the CPU-Z screenshot, but that's the clock rate so for DDR the data rate would be 1322. It's not 1333 exactly because your BCLK dipped from 133MHz to ~132 while you were looking at it, which is fine. Your QPI is going to be very overclocked no matter what you do because you don't get enough multiplier options to go down much, but it should be fine to just stick with the lower setting from what I know. I've never heard anyone say that overclocking the QPI has a noticeable effect on system performance, but I don't know that I've ever seen much other discussion on it either. If you were able to get higher BCLK levels stable by lowering your CPU multiplier, then I think QPI must have also been higher which suggests that you still have some headroom where you are at 170. Eletriarnation fucked around with this message at 05:32 on Jul 16, 2017 |
|
#
?
Jul 16, 2017 05:28
|
|
|
Alright, cheers. I've stress tested the crap out of the setup, and there's just a very steep increase in required voltage at this point. I had a bunch of times where I thought I was good at 180/190, but on the fourth or eighth hour of Prime95, it finally gave out.
|
|
#
?
Jul 16, 2017 15:10
|
|
|
can anyone recommend a guide for overclocking a 7700k? I just built a new PC and I have a 240mm CLC on the way. I want to read up before I install and overclock
|
|
#
?
Jul 20, 2017 23:26
|
|
|
Fauxtool posted:can anyone recommend a guide for overclocking a 7700k? I just built a new PC and I have a 240mm CLC on the way. I want to read up before I install and overclock http://www.tweaktown.com/guides/7481/tweaktowns-ultimate-intel-skylake-overclocking-guide/index3.html It's for skylake but everything there applies.
|
|
#
?
Jul 21, 2017 03:50
|
|
|
thats a pretty excellent guide. My ram is rated at 3000 but my mobo says it can go up to 3866 on the box. Does that mean I can potentially overclock that ram up to 3866 or do I have to buy ram rated at 3866?
|
|
#
?
Jul 21, 2017 07:44
|
|
|
In theory, your RAM can absolutely go up to 3866. The 3000 rating on it just means it's been successfully tested to run at that speed by the manufacturer. As such, pushing it up further than that is really not recommended unless you've got a very good idea of what you're doing.
|
|
#
?
Jul 21, 2017 08:22
|
|
|
the benefits are almost none going from 3000-3866 so i wasnt even gonna try
|
|
#
?
Jul 21, 2017 10:43
|
|
|
Honestly it's something I'd be tempted to try for shits and giggles, but it would also require cranking the gently caress out of my voltage.
|
|
#
?
Jul 21, 2017 11:13
|
|
|
Apologies if this isn't the right thread to ask, but what's a good app to monitor CPU/GPU temps 24/7 that might also give me a warning if my temps go above some safety limit? Thanks!
|
|
#
?
Jul 21, 2017 16:37
|
|
|
I like using built in firmware functions for it so I have my fan speeds set to ramp up at 75C and stay at 600RPM before then. The fans ramping up to 2,000 RPM acts as the alarm for me and I don't need to run any extra software. If you have a Pascal you can't do it for the GPU because of their locked bioses though.
|
|
#
?
Jul 21, 2017 17:20
|
|
|
Fauxtool posted:thats a pretty excellent guide. If your RAM ran with any stability at a higher speed, they'd have sold it at a higher speed. They may well retard it by one tick from stable (e.g. tests at 3200 maximum, sell at 3000) but there won't be any significant gains anywhere.
|
|
#
?
Jul 21, 2017 21:51
|
|
|
Mister Fister posted:Apologies if this isn't the right thread to ask, but what's a good app to monitor CPU/GPU temps 24/7 that might also give me a warning if my temps go above some safety limit? Thanks! HWInfo64 is pretty good. I set it to start with windows and open sensors and minimize everything, but I have the sensor readouts for GPU temp, CPU voltage and Core Max in the system tray (in the big sensors window you can right click and add to tray). I don't think it has a feature for warning you about overheating, but your CPU will throttle at 99C and I'd hope the GPU has some hard limit like that, too.
|
|
#
?
Jul 22, 2017 00:25
|
|
|
Is there any info out there yet on safe voltages for the X299 series chips? I see that Silicon Lottery is binning chips in the 1.25-1.275 v range even for their highest clocks, which is a fair bit less than the ~1.4v you see on some of the Kaby Lake chips.
|
|
#
?
Aug 3, 2017 03:33
|
|
|
Tom's Hardware published a big new thermal paste review http://www.tomshardware.com/reviews/thermal-paste-comparison,5108.html
|
|
#
?
Aug 5, 2017 10:49
|
|
|
eames posted:Tom's Hardware published a big new thermal paste review Nice. Thermal Grizzly Kryonaut won out of all the non-electrically conductive pastes I saw in a previous line up too, it's what I'm using because of that. Comes with a really nice applicator, as well. HalloKitty fucked around with this message at 19:20 on Aug 5, 2017 |
|
#
?
Aug 5, 2017 18:29
|
|
|
Yeah, I use the same paste and can't complain. Liquid Metal doesn't seem to be worth the 2-3°C difference at all. My only concern that its long term rating is max. 80°C and I put it on my Macbooks which see 100°C CPU temp from time to time, because that's how they're designed. Granted that doesn't mean that the paste is also being heated to 100°C but I'll see how it holds up...
|
|
#
?
Aug 5, 2017 19:08
|
|
|
I think liquid metal works best with delidding. Saw a 20C drop in temps replacing it and using mx4 between the IHS and heat sink.
|
|
#
?
Aug 5, 2017 20:13
|
|
|
Now I'm wondering what would happen if you delid and use the Coollab thermal pad in between the die and IHS.
|
|
#
?
Aug 6, 2017 04:37
|
|
|
There's no real way to delid and use air, is there? Even with a shim.
|
|
#
?
Aug 6, 2017 10:40
|
|
|
you mean direct die cooling without a IHS? AFAIK it's not possible because the substrate is now so thin that it would crack under the load of a modern heatsink putting all of its force right in the (unsupported) center of the package. a thunderbird-era shim might work but then you'll still have to modify the heatsink baseplate so it clears the socket and overall mounting height might also cause issues. technically possible but not recommended.
|
|
#
?
Aug 6, 2017 10:45
|
|
|
eames posted:you mean direct die cooling without a IHS? That's it. The shim from aqua computers looks good, but my baseplate 
|
|
#
?
Aug 6, 2017 12:59
|
|
|
EK has a direct die kit for ivy and haswell https://www.ekwb.com/shop/ek-supremacy-precisemount-add-on-naked-ivy May or may not work with later cpus.
|
|
#
?
Aug 6, 2017 15:23
|
|
|
Don Lapre posted:EK has a direct die kit for ivy and haswell The later CPUs have a PCB that is like a millimeter thinner so I wouldn't think so. Edit: okay like .3mm thinner but that can still throw it off. By the way, this is an excellent article on why Intel doesn't solder their CPUs: http://overclocking.guide/the-truth-about-cpu-soldering/ Zero VGS fucked around with this message at 19:54 on Aug 9, 2017 |
|
#
?
Aug 6, 2017 16:10
|
|
|
I'm overclocking my laptop i7 4720HQ to play Grim Dawn, which only uses one core more smoothly. (Really seems like only desktops with 4+GHz CPU's can run it at solid 60 fps without turning lighting to lowest setting.) I've succeeded in making it run at solid 3.6 GHz but I really don't get what is really going on here:  (default settings on the left, maxed out multipliers on right) Isn't it supposed to turbo up to 3.6 GHz when only one core is active by default? It only went to 3.4 GHz before I fiddled with XTU. I'm not hitting any power, thermal or current limits so performance shouldn't be held back by anything. Played few days like that, but today I see this with same settings:  I don't get it, My CPU ran ~100MHz faster today. And how is active core count determined? If I check task manager I can always see some usage on all the cores so it can't be based on that. But even when playing this game my active core count bounces between 0-2. I know this CPU isn't meant to be overclocked so it's very limited, but I really want as much single core performance as possible for this game. Don't need to OC for any other game and I know that it's pointless to even attempt for ones that use more than one core due to current and power limits. Sininu fucked around with this message at 23:21 on Aug 12, 2017 |
|
#
?
Aug 12, 2017 23:13
|
|
|
|
| # ? Apr 24, 2024 21:13 |
|
|
How much power can a PCIe slot push by itself? I thought it was 75 watts, which would preclude the 1050ti from overclocking, but manufacturers make overclocked 1050tis. Also, how aggressive is MSI with their binning?
|
|
#
?
Aug 15, 2017 17:23
|
|
