|
The stock cooler is perfect for stock speeds, but you need to spend 30 dollars (or more if you want something quieter) on a 212 if you want to overclock. The stock cooler gets overwhelmed at even +100MHz, it's kind of neat how perfectly they made it.
|
#
¿
Oct 3, 2018 18:00
|
|
|
|
| # ¿ May 9, 2024 07:06 |
|
|
From other people the performance is similar but the noise is louder.
|
|
#
¿
Oct 6, 2018 14:18
|
|
|
Yeah, the 212 is notoriously the most annoying heatsink to install, but you only need to do it once.
|
|
#
¿
Oct 8, 2018 21:51
|
|
|
What's the popular option now? I thought the 212 was the cheapest good enough cooler.
|
|
#
¿
Oct 9, 2018 22:12
|
|
|
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
|
|
|
Increasing the VRM frequency on my P67 was the magic that got me from 4.2GHz to 4.6GHz. I think I only needed to raise it to 350, any higher didn't help.
|
|
#
¿
Oct 24, 2018 16:31
|
|
|
I lowered my Sandybridge PLL voltage to 1.6 or so, it lowered temperatures, maybe. Someone said it's also the voltage the temperature measurement uses for a reference, so there's less power going in, but it might make the temperature readings inaccurate.
|
|
#
¿
Oct 25, 2018 22:19
|
|
|
AIOs will be louder but fit into smaller spaces. I went with a D15 over a H110 for noise reasons.
|
|
#
¿
Nov 6, 2018 20:05
|
|
|
That does seem really high. I don't remember you mentioning it before, but the pump isn't on a speed controlled header, is it? The pump should run at 100% all the time. Maybe it's a faulty H115 that didn't get filled all the way? I think you would be able to hear air bubbles if it wasn't filled all the way though.
|
|
#
¿
Jan 6, 2019 14:56
|
|
|
Space and effort reasons. When I did watercooling I used a car heatercore outside of the case, but that involved lots of extra work that a lot of people don't want to put in. I also had a 12 inch or so 115V fan blowing though it which was very quiet and didn't put a load on the power supply, but again, lots of work to get it relayed up to run with the PC. AIOs are made to fit in weird cases with minimal effort, I used like 4 foot tubes with my loop, for people that don't want to keep the radiator outside of the case 4 feet of tubes would just be a mess.
|
|
#
¿
Jan 6, 2019 18:11
|
|
|
If you want it to last "forever" I'd keep it there, if 3 years or less is fine go to 1.5 V if it benefits. That's not to say it will die in 3 years at 1.5 V, but longevity will be affected and for some people a 200GE doesn't matter or some might want to keep it.
|
|
#
¿
Jan 9, 2019 16:27
|
|
|
Make sure your low power states are enabled, guides recommend disabling them for stability testing, but you should never need to keep them disabled. That should get you back your low power idle speeds. The memory is just a reporting thing, most programs report DDR at half speed to report the cycle speed, DDR means it functions on the rising and falling edges of the cycle so for marketing the cycle speed is doubled. There's also up to 5x per cycle transports, I think DDR 3 and 4 actually are 4X, but I don't know.
|
|
#
¿
Feb 6, 2019 00:16
|
|
|
XMP generally is tested. When I got my memory I ran it at the default 2400-15 for about a year, but I read somewhere XMP is tested and not a guess and have been running it at 2666-14 for over 2 years now. I'm sure some cheap brands might not test it, but good brands should be validating their XMP settings.
|
|
#
¿
Feb 20, 2019 11:46
|
|
|
Tools usually give you a good starting point and real world testing is still needed to dial in exactly what you need. There's only so many situations a tool can test, and you may have found a case it can't test for.
|
|
#
¿
Mar 11, 2019 18:58
|
|
|
If you want to hear the fans you can do what I did. I have a D15 but I replaced the stock fans with their 3000 RPM ones. If the temperature breaks 75C the fans go from silent to full speed acting as an alarm for me. You can make a more detailed fan curve with them making a little bit of noise at 60C or whatever.
|
|
#
¿
Mar 19, 2019 17:40
|
|
|
I don't like that because it's not cheap, it doesn't prove anything, and dirt will be a problem, not to mention corrosion.
|
|
#
¿
Mar 20, 2019 15:45
|
|
|
That's fine for a 9600K, but the effect on performance is a lot lower. Figure out the max the cores can do, then test for uncore overclocking stability. (I still have my Haswell at the default 3GHz because the change in performance was so small and not worth it to me to stability test it)
|
|
#
¿
Mar 20, 2019 18:59
|
|
|
Crotch Fruit posted:Good thoughts about watercooling It doesn't prove anything because we already knew water has a higher thermal capacity and conductivity than air. I think in a few months the fins will all be clogged up and watercoolers work in part because the don't have dissimilar metals, that's where most dirt and corrosion comes from. Water coolers are designed to keep forcing water though the entire block assembly so there's no dead zone that can get clogged up. There's also better ways to flow the water, but that's a more minor thing. An interesting video would be if corrosion preventing additives can work under such extreme conditions after a few months. Does painting the heatsink work and not severely effect the performance? It's like 2 decades ago now since I was in school and I haven't really kept up with new technologies. I went to college for thermodynamics so I might be too harsh on homebrew cooling projects. Don't get me wrong, I like when people think about cooling, it was a huge part of my life for a few years.
|
|
#
¿
Mar 21, 2019 00:29
|
|
|
I think you can raise the voltage a lot. 1.25 is like what people are giving Coffee Lake. I think 1.38 is the unofficial limit for Bloomfield. I have my Haswell at 1.29 right now. As long as your temperatures stay in check I wouldn't worry about raising the voltage at all. Crashing on startup is just classic stability problems, if it happened after a few hours of Prime it could be temperature, but when it's booting I'd go straight to voltage. Of course it could just be a weak chip and there's nothing you can do about that. It slows down at idle to save power. They've gotten really good at handling the speed changes and there should be virtually no performance or stability impact at normal levels. My install of Windows 10 reports my Haswell as running at both 3.3 GHz and 4.3 GHz depending on what I'm checking. I'm going to guess that Windows has started to read a string in the processor that says what the stock speed is and reports that to avoid reporting problems with people seeing like 800 MHz if they check the field while it's idle.
|
|
#
¿
Mar 25, 2019 15:11
|
|
|
Is the memory speed based off the QPI speed? Memory getting unstable used to be what caused weird things to happen, but it has been generally fixed for a while now, but Bloomfield was one of the last processors you could change the FSB on. You could try setting the memory divider one lower so the bios displays that the memory is running slower, but with the overclock it's the same speed. I think with X58, you can't change the multiplier but it's very capable of FSB changes. If I'm remembering wrong, yeah, don't touch the bus and push the multiplier. craig588 fucked around with this message at 23:57 on Mar 25, 2019 |
|
#
¿
Mar 25, 2019 23:54
|
|
|
Someone else'll have to help on specifically what to change, I've resigned to paying extra for Asus motherboards so I don't need to learn different settings and I can get used to everything being called and sorted generally similarly. You want to set the timings to 9-9-9, generally memory runs best at the timings it was specced for. The performance benefits from tighter timings are generally small and instability happens from small changes. Right now it's running at 1333, but with 7s that could be what causes the problems when the CPU is running fast enough to stress it some more. I think the memory speed is fixed, if it wasn't you'd see a speed something like 830 MHz with the 165x21 overclock.
|
|
#
¿
Mar 26, 2019 00:19
|
|
|
The 930 has a stock FSB of 133 MHz, setting it to 100 MHz will cause problems. I've always found CPU-Z to be extremely reliable for the limited set of things it reports. Clock speeds and timings, VID sort of works, but the CPU voltage can be adjusted by the motherboard so much it doesn't mean much. Normally it grays out the field if it doesn't trust the reading it's getting.
|
|
#
¿
Mar 26, 2019 01:09
|
|
|
This doesn't help anything, but I have all of the terms floating around so I might as well get them out. tCL is column address strobe delay, tRCD is row to column delay, tRP is row precharge, tRAS is row address strobe delay. None of the names matter to consumer use of memory, but I wanted to get them out since I knew them. I'd start with setting everything to 11 except the tRFC to like 30. Not a rule, but a good guideline is to start out setting the tRFC to about the first 3 values added together minus 1. 32 will probably work too but is probably higher than it needs to be. Timings matter very little for performance and they more need to match what the memory was designed for. The best timings versus the worst timings at the same clock speed will only be a few percent of performance difference. Command rate is more down to the memory controller, but with 4 sticks you might need 2 there for stability. Usually (But not always!) 1 works for 2 modules and 2 is for when you've fully populated the memory. If the memory has a SPD or XMP profile built in that's already those settings go with that. There are like 20 other timings that also don't really matter for performance that the SPD can probably have that will be better than the motherboard making its best guess at. Google the model number off one of the memory chips and see if anyone else has raised the voltage on them. I'd go to 1.65 volts with no worry, but if other people didn't see gains I wouldn't bother and keep them at 1.5 volts. The brand or rating doesn't matter in this case, whatever some chips were binned to if they respond well to voltage than everything from that line should.
|
|
#
¿
Mar 29, 2019 12:21
|
|
|
Memtest is reliable, memory is hard to test overclocked because there's no way to heat it up and test a variety of operations. For memory overclocking kind of the one real stability test is just using all of the apps you use.
|
|
#
¿
Apr 4, 2019 14:35
|
|
|
|
| # ¿ May 9, 2024 07:06 |
|
|
Up to 1.65 memory voltage is fine. (For DDR3, DDR4 is down to 1.35) The memory voltage only hits parts of the CPU capable of handling it.
|
|
#
¿
Apr 13, 2019 09:05
|
|