Question: is delidding and lapping my cpu going to give me some nice gains or is it going to be a lot of work for nothing? Overclocked @1.16v I can get to 4ghz and my temps with prime95 are 65c. @1.25v I can get to 4.4ghz, but then the temps shoot up to 80c, which is a tad bit too high for me, nevermind going to 4.6ghz at even higher voltage. So what gains could delidding and/or lapping give? Or what has it done for you? These are the stats: What I have to work with (CPU cooler=CNPS10x with Venturi HF-14 fan) (and yeah, cablemanagement is impossible in this case :bang: )
lapping both cooler and ihs will result in 5-6c usually, just cpu maybe 2-3c. Delid + liquid ultra will drop temps by 20-30c. I got about 28c on my 3770k
I'll tell you about my de-lidding experience and try to help you come to a conclusion. My 3770K is a bad clocker with "high" current leakage and consequently high temps when overclocked (my cooler is the Corsair H70). I found that higher RAM frequency and DRAM voltage further limited my max CPU overlock, with the extra heat it produced causing the CPU to exceed the safe temperature limit (Tjmax of 105°C) when under load. Before de-lidding I ran Linx at stock clocks with default Turbo settings, stock voltage (VID 1.160 V under load with no LLC) and RAM at 2400 MHz (11-13-13-35 2T) as you can see below; Here is a Linx run with all cores set to Turbo to 4.2GHz with stock voltage and RAM at 2400 MHz (11-13-13-35 2T), before de-lidding; As you can see, the hottest core was already hitting 84°C at only 4.2 GHz. After reading about the pros and cons of de-lidding I went ahead and de-lidded my CPU nine days later using the razor blade method. I applied Coolaboratory Liquid Ultra to the CPU core and "lapped" the nickel off the underside of the Heatspreader to reveal the copper surface, but left the top of the Heatspreader alone. I re-mounted my Corsair H70 and re-ran the Linx test at 4.2 GHz with stock voltage and RAM at 1333 MHz (9-9-9-24 1T) using SPD; As you can see, the hottest core only got to 70°C, that's 14°C lower than before I de-lidded! I then enabled XMP on my G.SKill DDR3-2400 and reran Linx; I gained a couple of Gigaflops and shaved 4 minutes off the test run. Unfortunately I failed take a screenie of the temps but as I re-call the core temps were higher by about 3 to 4°C. I run my 24/7 overclock at 4.5GHz @ 1.36 V with RAM at 2666 MHz (11-13-13-35 2T @ 1.65 V), although it throttles when I run Linx (or do long video encodes with Avisynth) I have found it to be stable when gaming. To get 4.6 GHz stable I have to lower my RAM to 1600 MHz and increase my Vcore to 1.40 V, which is too much voltage to run 24/7 IMO. If you have a chip that is throttling or hitting really high temps then the chances that de-lidding will help are a lot greater. If you have a good clocker that stays cool then I would say de-lidding is pointless.
When I delid my 3570K I dropped temps about 23C! Didn't do much for OC headroom, but its nice to see such low temps compared to before.
Those are some nice gains! Is Coolaboratory Liquid Ultra to be recommended over for example Gelid Solutions GC-Extreme? I currently use that. Dropped 5C on my CPU and GPU versus my older solution (can't remember what that was) @Chinobino: you lapped the inside of the heatspreader? I've seen plenty of heatspreaders lapped on the outside, but not the inside. What did you use?
question is, is it worth the risk, ive seen people kill 3770k's by delidding, and i really wouldnt worry about 80c, it probably isnt going to do anything to harm the chip. youd be at more risk delidding in my opinion...
IMO it probably doesn't make that much difference. Gelid Solutions GC-Extreme is non-toxic so it's probably better to deal with. Well technically I only lapped the raised inside edges and then hand sanded the concave middle section with some coarse sandpaper (200 grit) and then finished with some wet & dry (4000 grit) IIRC. I put the IHS in a vice and used my thumb to sand the nickel off, I didn't want to make it too concave or uneven so I just removed the nickel to expose the copper. The IHS had so much space when I put the Coolaboratory Liquid Ultra in that it didn't make contact with the IHS, hence lapping the edges to ensure the CPU core made good contact.
I did delid my 3770k. Running it without headspreader with watercooling now. It runs at 4.7ghz and doesn'tget hotter than 75 degrees C anymore. It went into the 90's before deliding.
I just delidded my i5 and ended up with a broken CPU (vice method). http://forums.guru3d.com/showthread.php?t=401378 So I'd say not worth the risks, outcome is somewhat a matter of luck I guess.
Aw man, what a shame. IF I was going to do the delid I was going to use the vice method, as that seemed the least risky of the three I've seen (sharp object/razorblade, vice and hammer or only a vice) I'm really doubting if I'm going to do it. @4ghz my temps are fine, but it's not much of an overclock. For decent temps 4.2ghz is the limit for me. I think I'm going to postpone the delidding until the performance of my CPU is becoming a bottleneck in the games I want to play. Thank you all for answering my question! If and when I'm going to delid, I'll be sure to update this topic with some pictures
Based on my experience, if you want to delid, the razor method may be better after all. You needn't "twist" or hammer the CPU that way and there aren't any ICs or similar near the glue that might be damaged if you do it carefully. The only thing "protruding" from the surface is the core in the middle of the chip. Check some delidding videos on Youtube to see for yourself. But, again, risky and certainly not worth destroying your CPU, as I've recently learned. I even got a deal on my current CPU, cost "only" 190 EUR. The replacement (4690K) set me back 250 EUR. So that's a total cost of 440 EUR for this little operation. Thankfully I have some savings to spend on this.
I did the Delid hammer method with my 4670k and with the IHS removed, cleaned off the glue seal, lapped the underside of the IHS so it sits lower and actually contacted the CPU core, then lapped the top surface. Dropped temps over 30'c. At 4.2ghz I was hitting almost 90'c and the max I hit now is around 60'c. Idle is low 30's.
gains like that, so tempting, id do it if i had the guts lol, the older haswells will gain more from delidding because of that crappy TIM, and you say 90c, what voltage and what cooler?
Well, almost 2 years later and I did it. In all the stress I forgot to take pictures I used the Vice only method, which was very quick and easy. The hardest part was taking the glue off. For now I used some GC extreme and that already gave me 10'c lower temps. For the overclock it doesn't help much, but the temperature is much more stable now. It would spike like crazy, especially now that games are actually using close to 100% of the CPU. I would even get some crashes because of sudden extreme spikes. Maybe it's because my CPU is getting it's 5th birthday pretty soon, but the intel tim was VERY rubbery at this point. The CPU die was almost clean and the tim stuck to the heatspreader when I took it off. I could peel it off in a single move.....
Excellent well done. Recently did my 4770k and shaved ~20'c on load temps. I bought a cheap kit off eBay to do it - 2 pieces of plastic 3D printed, pop in CPU ( squeeze in vice, tap with hammer etc etc) chip is delidded nice and safely. I cleaned of all old Tim and glue and used some nail polish lacquer to cover up the small row of caps next to die on this chip, then applied Liquid Ultra to CPU die. Popped back in socket, the heat spreader is held in place by mobo retaining strap. Didn't use glue as wanted best contact between CPU and heatspreader. Gelid GC between Heatspreader and heat sink. Load temps before were 85'c, now they are 65'c - and this is max load stress testing Prime CPU at 4.7ghz One of the best mods I've ever done, really happy
Use a liquid metal TIM. Best atm is Thermal grizzly Conductonaut. You can expect to drop another 5-15c off again. I had a 3770k on water, would do 4.8GHz relatively hot under load in p95 at 1.32v. I was able to to 5ghz at same voltage with a 28c reduction in temps.
Bad photo Must say, there isn't thát much improvement, but that could be because ambient temps are about 10°C higher than during the other test.
Glad you did it in the end, wouldn't bother re sealing ihs back to pcb, half the problem is the distance between die and ihs. When I tear down my system again I'm going to redo my TIM on my 4770K die with conductonaut liquid metal then put kryonaut on the IHS.
I know this may be beyond the reach of most people, including myself now since I no longer work in a toolroom, but the way to do this is to use a surface grinder. Standard surface grinding wheels (20mm) will make a perfect line down the middle of the underside of the IHS to accommodate the die. If you know of a machining shop then I'm sure they will do it for very little money as it's such a quick and easy job. Of course, while you are at it, you should measure the clearance from board to die top and get the IHS ground with a couple of thousandths clearance. You know, all the time I worked there I never made anything for myself, now I don't work there and I have numerous projects that need machining Such as my custom DNA200 ecig mods, bits for a friends trike he is building, mods to my pc, ecig tanks, the list goes on and on.