|
I wouldn't worry as much about VID. Difference between VID and Vcore is a given part of the design - that's Vdroop. VID represents the highest possible transient spike, not the steady-state voltage. While such spikes do stink, you're far off from Intel's steady-state recommendations (which were made taking VID and Vdroop into account). That recommendation is 1.38V Vcore. If you use Load-Line Calibration, you will get Vcore closer to VID at the cost of VID no longer representing the true maximum voltage. You should start hitting strongly diminishing returns in another .02V or so, at around 4.4-4.6 GHz - that'd be a good and safe spot to stop pushing.
|
#
¿
Nov 3, 2014 15:26
|
|
|
|
| # ¿ May 10, 2024 06:33 |
|
|
Twerk from Home posted:Thanks for all this. My temps overclocked on the Hyper 212 EVO are better than at stock clocks on the stock cooler. The stock cooler was letting it spend time at 75C, apparently? Shouldn't it be throttling to keep itself under 72? It had probably already stopped Turbo Boost at that point, yeah. Throttling below the base clock multiplier takes higher temps.
|
|
#
¿
Nov 3, 2014 18:00
|
|
|
Well... Yes and no. Yes because yeah, you would get a tiny clock bump with sustained turbo - like, 100 or 200 MHz - over the base frequency. No because 1) on a desktop chip, the difference between base frequency and max all-cores turbo is tiny (mobile chips it's much higher), 2) the chip may still throttle based on TDP even if the temperature is under control (but likely won't unless the integrated graphics are also in use), and 3) most overclocking-capable motherboards these days will affirmatively gently caress with the default Turbo behavior and force the chip to always run at the maximum turbo state under load, come hell or high water.
|
|
#
¿
Nov 4, 2014 03:24
|
|
|
Jawtramp posted:Crossposting from gpu thead, should have posted here first: The card does not run at its top clock speed all the time. It self-throttles and self-overclocks based on the load that's running. You will only see the effects of the voltage and clock rate sliders once you start an intensive program, e.g. Heaven or Furmark. Those crashing means your overclock isn't stable. You'll have to back it off until they (and any other 3D programs you run) don't crash. The card does not have "extreme power draw." 750W of power is plenty for your setup. Overclocking the CPU can harm stability in general, but if the CPU were the part causing problems you'd get Bluescreen crashes, not graphics problems. It might also be helpful to check the OP for a description of turbo boost behavior so you can understand what the card is doing already. tl;dr Your graphics overclock is too aggressive, back off.
|
|
#
¿
Nov 12, 2014 22:31
|
|
|
Is that Vcore (i.e. what CPU-Z reports) or just what's set in the BIOS? It's Vcore you want to keep at ~1.3V or lower.
|
|
#
¿
Nov 20, 2014 21:19
|
|
|
Heat is a HUGE problem there. 100 C is absolutely not acceptable. 80 C is, IMO, just on the irresponsible side of hot. Are you sure the cooler is mounted correctly and all the pushpins are pushed and locked? Even the stock cooler should be doing significantly better than that, given that it's a 50-ish watt chip stock.
|
|
#
¿
Nov 30, 2014 21:16
|
|
|
Apparently Corsair measured the clearance a bit wrong, because if you Google the combination of a Spec 03 with a Hyper 212, you get people saying "Yeah, it said it wouldn't fit, but it totally fits." It might be a Corsair push towards their in-house liquid coolers. If you can tolerate a rebate and more noise than a 212 EVO, Cooler Master's Seidon 120V closed-loop liquid cooler is on Newegg for $50 - $20 rebate = $30. Or, keep an eye out for the Phanteks PH-TC12LS that was announced last week - at $40, it'll probably be the new price/performance king for low-profile coolers.
|
|
#
¿
Dec 1, 2014 11:17
|
|
|
Eh, it's a 2011 processor first overclocked in 2014. By the time it dies from 1.4V instead of 1.38V, we'll be on whatever comes after Cannonlake. Does the processor properly idle with SpeedStep when you used fixed voltage? Like, it should drop to 1-1.1V and 1600 MHz when it's not doing much of anything. If it doesn't, consider redoing the overclock but with an Offset voltage. Start at +0.0 at stock clocks, see what the Vcore is when burning, and then add Offset to get it back to whatever 1.40 manual gives you. Nemesis Of Moles posted:Is the 760GTX worth overclocking? It runs pretty much everything I need it to but I figured if it's as easy as the CPU I might as well squeeze some more juice out of it. All GPUs are worth overclocking, and they're even easier than CPUs. First, max out the power target. Then set the temperature target to something you can tolerate - 80 C is good, or if you are confident in your cooling, 90 C (on the assumption it will never get there). Then set the boost logic to be temperature-dependent. That alone should provide a good boost. But if it's not enough for you, you can then adjust the base frequency upwards. Stress test with Unigine Heaven. If any games you play crash (check at least one of each for DirectX 9, 10, and 11), then dial back the clock offset in 13 MHz increments. If you feel salty you can overvolt it. I've never felt particularly salty - occasionally the gains are good (like when I put my Radeon 6850 to the 6870's stock voltage), but more often modern cards exhaust most of their voltage scaling just with the boost logic. Factory Factory fucked around with this message at 08:16 on Dec 5, 2014 |
|
#
¿
Dec 5, 2014 08:09
|
|
|
Both of those RAM overclocks seem pretty high. Not all R9-280s have 6 GHz RAM chips, and does the system memory really pass at least two passes of memtest86+ like that?
|
|
#
¿
Dec 13, 2014 22:59
|
|
|
Temperature walls are more related to voltage than clock speeds (though clock speeds are also related to voltage). And voltage (well, Vcore) looks sufficiently low. I'm tempted to say that the cooler might not have been installed correctly - maybe you pre-spread the thermal paste and there are air bubbles trapped? Best way to improve temperatures would just be to delid the thing and clean up the IHS adhesive.
|
|
#
¿
Dec 15, 2014 06:05
|
|
|
1gnoirents posted:Which brings me to why I'm here. I've been stable for months and months at 4.5 ghz and 1.35 vcore (4670k). That's actually my lazy multiplier, I know its stable at 4.6 and ~1.40 or 1.385 (I know, not great). This is with IBT, smallFFT for many hours, and regular use for at least half a year. I never blue screen. Video encoding is practically as tough on a CPU as a stress test. Video encoding uses a bit of everything on a CPU, but is especially heavy on floating point, including AVX instructions. It's also very reliant on memory bandwidth and highly parallelizable, so as much of each core as possible is going to be lighting up from out-of-order execution running a whole bunch of stuff in parallel on each core, even without hyperthreading. Temperature-wise, it can match Prime95 or even IntelBurnTest. So if any common real-world workload would bring an overclock to its knees, it'd be video encoding. But that said, that's not the first time I've seen a pattern like that and I'm wondering if I shouldn't mention it in the OP. Agreed aggressively volted his CPU (1.38V), and now it can't hold its old clock. My own CPU, despite having an excellent voltage response (i5-2500K 4.6 GHz @ 1.28V), had a low upper limit, and now it's even lower - 4.4 GHz at the same voltage. Electromigration isn't purely a "suddenly it stops" phenomenon; physically, it's progressive degredation. High volts will knock poo poo around.  I don't know, but it seems plausible that this could be the physical erosion of linkages to the point where the chip can't handle such an aggressive switching speed cleanly. The bottleneck was already the weakest link, after all.
|
|
#
¿
Dec 15, 2014 18:55
|
|
|
AS5 is not meaningfully different than most standard thermal paste.  What's load temp like? Idle is much less important, but you could always try bumping the fan speed if it bothers you.
|
|
#
¿
Jan 7, 2015 22:24
|
|
|
1gnoirents posted:Paste at full load can be a matter of like 3 degrees unless there is something seriously wrong with the factory paste. I am in a firm minority where I think those 3 degrees can make a difference (sometimes it wont matter at all though) but thats only when youre pushing it to the max. That being said, I also think 45 degrees is a tad high for idle, but like factory said check your full load temps which is the important information. I don't disagree that paste can make a 3C difference. It's just that AS5 doesn't live up to its hype and is nothing special any more.
|
|
#
¿
Jan 8, 2015 05:19
|
|
|
Get a Cryorig H7 instead - easier to install, smaller, cools better, quieter, and only like $5 more.
|
|
#
¿
Jan 18, 2015 00:50
|
|
|
Even if a little extra voltage fixed the card, crashing at stock settings means the card is faulty. Try rolling back the drivers, and if it's still as misbehaved, then get an RMA. -- I used my overclocking knowledge for something strange today: I undervolted my Surface Pro. Took 50 mV off the top, resulting in a 10 C temperature drop under LINPACK and a small but noticeable improvement in battery life. Crazy. Supposedly the real payoff will be when I game on the thing, as the lower heat will allow higher turbo clocks.
|
|
#
¿
Jan 21, 2015 02:49
|
|
|
Oh dear God. Well, because it works, that was definitely a filtering cap. If your PSU and motherboard have great ripple control, the CPU will be fine. But ripple increases wear and tear, and your CPU now has less protection against it. How do you feel about getting a replacement cap and an SMD soldering iron? 
|
|
#
¿
Jan 27, 2015 14:32
|
|
|
SMD soldering is really hard, even with a lot of experience. Taking a hammer to the chip is practically normal and easy in comparison. 
|
|
#
¿
Jan 28, 2015 08:05
|
|
|
I don't think you need to delid a G3258. It's a Devil's Canyon part, after all. I, on the other hand... I just got a shiny new i7-3770K. I gave it an easy clock, 4.4 GHz at around 1.1V, but it's already hitting ~70 C. I could choose to be okay with that, I guess... 4.6 GHz will need more than 1.13V, which was the highest I could go while staying below 80 C. Right now, I'm letting it Prime95 overnight at the 4.4 GHz clock to make sure the chip works fine, before I go voiding the warranty.
|
|
#
¿
Jan 31, 2015 05:39
|
|
|
I am finding it super difficult to delid a 3770K without an actual vise to work with.
|
|
#
¿
Feb 1, 2015 19:58
|
|
|
|
| # ¿ May 10, 2024 06:33 |
|
|
This 3770K is remarkably resistant to delidding. I give up. 4.4 GHz at 1.1V it is.
|
|
#
¿
Feb 17, 2015 18:53
|
|