|
Could use some help here guys, not sure if this is normal or not. I've overclocked my 4790K to 4.7GHz, I raised manual voltage slowly (up to 1.28V) until it would pass several hours of x264 stress test. In x264, temps max out at 83C. Idle temps at 28-30c. My cooler is a Noctua U12S. Is 83C OK for x264 testing, or is my cooling not sufficient? However, doing Prime95 small FFT, my temps shoot up to 100C IMMEDIATELY. I know P95 isn't good for testing haswell but that can't be normal is it?
|
#
?
Sep 21, 2014 17:05
|
|
|
|
| # ? Apr 19, 2024 09:31 |
|
|
Brain Issues posted:Could use some help here guys, not sure if this is normal or not. 1.28 volts is excellent for 4.7ghz. However small fft hitting throttling immediately is indicative of cooling issues to me at that voltage. I'd expect mid 80's. That being said smallfft and linpack (IBT) will produce the highest temperatures the fastest. In general I've been happy with mid 80's for these tests though. Despite that the important thing is your real loads never exceed a temperature you decide (say 75*), but its a little alarming that your temps jump so far with small fft imo.
|
|
#
?
Sep 21, 2014 17:12
|
|
|
Probably worth reseating the cooler with new TIM since mid-high 80's might be acceptable but not 100C.
|
|
#
?
Sep 21, 2014 17:20
|
|
|
cisco privilege posted:Probably worth reseating the cooler with new TIM since mid-high 80's might be acceptable but not 100C. 1gnoirents posted:1.28 volts is excellent for 4.7ghz. However small fft hitting throttling immediately is indicative of cooling issues to me at that voltage. I'd expect mid 80's. Thanks guys I'll try to reseat my cooler, I've got another tube of NT-H1. I was worried that I might have applied too much TIM when I put on my cooler, maybe I did.
|
|
#
?
Sep 21, 2014 17:22
|
|
|
Ok, so I re-applied NT-H1, using a small spot in the middle of the CPU and relying on the cooler to spread it around as I tighten it down. The first time I used more and spread it around the lid evenly. My temperatures are the same as before,, maxing out at 83C in x264 and spiking to 100C right away in P95 small FFT. Is a Noctua U12S just not enough cooler to run this chip at 4.7GHz? Should I just leave it alone? Considering it's stable enough using x264 for hours and hours, and I'm unlikely to ever load it up in normal usage a like P95 Small FFT load Brain Issues fucked around with this message at 18:20 on Sep 21, 2014 |
|
#
?
Sep 21, 2014 18:17
|
|
|
Brain Issues posted:Ok, so I re-applied NT-H1, using a small spot in the middle of the CPU and relying on the cooler to spread it around as I tighten it down. I've yet to overclock my 4790k, but P95 gives me 85c very quickly with a big rear end Thermaltake Frio bolted on top. Anything else I've run hasn't even hit 60c. I'm inclined to say P95 just isn't realistic. Or if you really want to be sure, there's always delidding.
|
|
#
?
Sep 21, 2014 18:58
|
|
|
I was gonna say, might wanna delid. My 4770k at 4.5ghz couldn't break 70c on air after I did that. Since then I switched to a 140mm CLC and I don't break mid 50s. I'm using NT-H1 on it as well.
|
|
#
?
Sep 21, 2014 19:37
|
|
|
Zero VGS posted:I was gonna say, might wanna delid. My 4770k at 4.5ghz couldn't break 70c on air after I did that. Since then I switched to a 140mm CLC and I don't break mid 50s. I'm using NT-H1 on it as well. Did you delid and replace the TIM with coolaboratory stuff, put the IHS back on, or did you mount direct to die? Also, how do you put the IHS back on, do you reapply adhesive or just set it back on top?
|
|
#
?
Sep 21, 2014 19:40
|
|
|
Brain Issues posted:
Set cpu in socket without IHS on. line up IHS ontop of cpu and close the retention bracket. You kind of have to fiddle a bit to keep the retention bracket from moving it too much. There is no need to reaaply adhesive.
|
|
#
?
Sep 21, 2014 19:57
|
|
|
Brain Issues posted:Did you delid and replace the TIM with coolaboratory stuff, put the IHS back on, or did you mount direct to die? I cleaned the IHS and chip completely of TIM and that silicone caulk. Then I put a rice grain of NT-H1 on the die, clamped it in, put another grain on the IHS outside, and clamped the fan onto that.
|
|
#
?
Sep 21, 2014 20:42
|
|
|
Zero VGS posted:I cleaned the IHS and chip completely of TIM and that silicone caulk. Then I put a rice grain of NT-H1 on the die, clamped it in, put another grain on the IHS outside, and clamped the fan onto that. Sounds pretty easy, I think I'll delid tomorrow at work and post here with my results.
|
|
#
?
Sep 21, 2014 20:55
|
|
|
FireSlash posted:So, a few weeks ago I was tinkering with overclock settings but quickly found that my cooling setup wasn't adequate. I have a thing for really quiet PCs, and that's generally not the domain of the overclocker. That is really cool and way beyond the scope of basically anything else anyone has tried here, even Veedubfreak.
|
|
#
?
Sep 21, 2014 23:33
|
|
|
Yeah I wasn't sure what to say, that is indeed cool.
|
|
#
?
Sep 22, 2014 19:40
|
|
|
FireSlash posted:
NEEEEEEEEEEEEEEEEEERD! but thats a really nice elegant way of doing stuff, kudos I did not go for brand name hose, just a company that sells sillicone that conforms to a bunch of standards, although I think liquid cooling with petrol would be counterproductive. I thought about a cooling chamber before c/gpu, brass, filled with ln, then remembered that its the rapid evaporation that cools with lm, not its a magical substance that is very cool, getting drunk or stoned gives me ideas, and the next day most of them are scrapped because science. I'd like a home made solution thats not as crazy bollocks as phase change but still gets below ambient, i should really stop trying to change the laws of physics from my armchair and build my loving water setup. I am just delaying because I want a bigger rad, and "moar fans", obviously its mainly going to be external. I have not even made anything with the stuff i have already because  guh.
|
|
#
?
Sep 24, 2014 00:48
|
|
|
You don't want to get below ambient because then you need to deal with condensation. You'd need to fill the socket with grease and paint the motherboard with nail polish and it'd still inevitably rust and short because of a tiny area somewhere you didn't notice. Phase change is also the cheapest/most sane way to get below ambient, it's why nearly all refrigerators and air conditioners rely on it. The only time they go out of that realm is for very specialized applications.
|
|
#
?
Sep 24, 2014 02:44
|
|
|
Ok, awhile back I asked about overclocking and I just got my new mobo and G3258. Now I'm having some trouble working out how to OC. The motherboard is an Asrock z97-Anniversary and it comes with a program called A-Tuning that does the OC, but I'm having trouble with how newbie I am at OCing. The advice I got was to start with 4.4Ghz and 1.299volts. Getting to 1.299 volts was no issue at all, I played with Vcore additional offset and put it to 1.299. CPU-Z said it was 1.299, though it kept going up and down. (I assumed that was alright). I then started moving up the CPU Ratio and the CPU Cache Ratio up by one slowly. From 32 to 40 there was no issue, but after that my computer restarted suddenly. What was I doing wrong? I didn't touch anything else and I thought I was following the instructions on the OP. EDIT: Googling has said A-Tunning isn't permanent, I have to use the BIOS. But I am still a little unsure of the Volts/VCore on the Bios screen. I did a tiny OC to 3300Mhz. Lord Windy fucked around with this message at 13:47 on Sep 25, 2014 |
|
#
?
Sep 25, 2014 13:31
|
|
|
Lord Windy posted:Ok, awhile back I asked about overclocking and I just got my new mobo and G3258. Now I'm having some trouble working out how to OC. Yes, but first update your bios. I dont know if they ship with the latest revision now but update it via MSI's utility in Windows. It's actually pretty straightforward and nice (sometimes it isnt). Also don't use the CD at all. Just download everything you need from MSI's page for the motherboard. Like chipset and everything. Then go into bios and set 1.299vcore then start at somewhere like 40x multiplier. Test and repeat from there. I wouldn't use software overclocking at all, just uninstall it.
|
|
#
?
Sep 25, 2014 15:15
|
|
|
Lord Windy posted:Ok, awhile back I asked about overclocking and I just got my new mobo and G3258. Now I'm having some trouble working out how to OC. I set my voltage by using the Max Vcore option. This way it still goes up and down with load but doesn't exceed what I set as max. I set the offset voltage to the lowest possible, .001, so it doesn't go over max voltage when under extreme load. Once you have the CPU stable at the speed you want/can get, start lowering the max voltage by .005 and testing until you find the lowest stable voltage.
|
|
#
?
Sep 25, 2014 16:07
|
|
|
So when I run small FFT chunks on prime95 with a 4790k clocked to 4.5, this happens Spikes up to the 80s-90s for the duration of that test. Happens at 4.4GHz, too. Haven't checked at lower than that. Anyone know why this is happening/if it's normal? E: I are smart and can't read above me in the thread. I'll try reseating the cooler tomorrow when I can buy some iso Moatman fucked around with this message at 01:09 on Sep 26, 2014 |
|
#
?
Sep 26, 2014 01:05
|
|
|
Probably need to reseat, though what's your vcore set at?
|
|
#
?
Sep 26, 2014 01:17
|
|
|
cisco privilege posted:Probably need to reseat, though what's your vcore set at? 1.1v Tried fiddling with it but nothing helped e: gonna run x264 overnight, see if that does anything. I'll probably grab a can of air tomorrow, too. Dust cover was filthy. Moatman fucked around with this message at 01:54 on Sep 26, 2014 |
|
#
?
Sep 26, 2014 01:49
|
|
|
Moatman posted:1.1v That is way too hot for 1.1 volts regardless of mhz. I wouldn't run anything overnight you already know its too hot.
|
|
#
?
Sep 26, 2014 02:03
|
|
|
Just did an x264 pass at 4.4 and it was sitting at about 70. Definitely gonna need to reseat
|
|
#
?
Sep 26, 2014 03:04
|
|
|
Moatman posted:Just did an x264 pass at 4.4 and it was sitting at about 70. Definitely gonna need to reseat I reseated mine, its definitely installed correctly, and it still gets way too hot on Small FFT. I think this chip really just doesn't play nicely with Prime95. I've run x264 overnight (8 hours) on 16 threads and its fine and never goes above 83C, and that's at 1.28V @ 4.7ghz.
|
|
#
?
Sep 26, 2014 04:00
|
|
|
Brain Issues posted:I reseated mine, its definitely installed correctly, and it still gets way too hot on Small FFT. I think this chip really just doesn't play nicely with Prime95. Yeah that's fine, especially with the stock cooler. 1.1v and 99 degrees is not
|
|
#
?
Sep 26, 2014 05:01
|
|
|
Ok, after some testing 4.5Ghz at 1.299v seems to be very stable on the G3258. But 4.6 and everything is not. I'm really unused to OCing, should I be bumping up the voltage. Is there an area where it is no longer safe for long term use?
|
|
#
?
Sep 26, 2014 05:53
|
|
|
Lord Windy posted:Ok, after some testing 4.5Ghz at 1.299v seems to be very stable on the G3258. But 4.6 and everything is not. I'm really unused to OCing, should I be bumping up the voltage. Is there an area where it is no longer safe for long term use? If your temperature isn't too high you can pretty much play around until 1.350 before you need to start worrying about the CPU. However, I'm not 100% on that motherboard... but its probably fine. There isn't much choice  the deal is too good. If I recall the text turns RED and scary after but that has nothing to do with the capability of the CPU. Perhaps its an early warning for the vrms on the mobo itself? Perhaps the manual will tell you. I personally ran it at 1.375 but it was more or less for validation I believe at 4.8 ghz
|
|
#
?
Sep 26, 2014 06:17
|
|
|
I might go to 1.349 as my max volt, head for 4.8 and see what the temp is like. Go down for stability and temp as needed, and once I find a sweet spot just reduce the max volt until I hit some kind of sweet spot. EDIT: Impossible to get to 4.6 or greater no matter what I do. I am happy enough at 4.5 and 1.32v Lord Windy fucked around with this message at 09:20 on Sep 26, 2014 |
|
#
?
Sep 26, 2014 06:43
|
|
|
Lord Windy posted:I might go to 1.349 as my max volt, head for 4.8 and see what the temp is like. Go down for stability and temp as needed, and once I find a sweet spot just reduce the max volt until I hit some kind of sweet spot. Might be vrin, cache voltage. Or chip stability, or the mobo can't keep a steady voltage. But 4.5-4.6 seems to be the sweet spot anyways. The difference between stock and 4.5 is ridiculous for games for this chip, 4.5 - 4.8 not as much.
|
|
#
?
Sep 26, 2014 16:37
|
|
|
I stuck with a lower vcore and a "lower" overclock. I could only get 4.7ghz with 1.240v. No matter how much I raised the voltage or screwed with other settings it couldn't get 4.8. Lower voltage is better anyway, runs cooler and uses less electricity.
|
|
#
?
Sep 26, 2014 16:41
|
|
|
r0ck0 posted:I stuck with a lower vcore and a "lower" overclock. I could only get 4.7ghz with 1.240v. No matter how much I raised the voltage or screwed with other settings it couldn't get 4.8. Lower voltage is better anyway, runs cooler and uses less electricity. Well I wouldn't worry that is a fantastic speed for that voltage.
|
|
#
?
Sep 26, 2014 16:44
|
|
|
1gnoirents posted:Well I wouldn't worry that is a fantastic speed for that voltage. I wonder now if I should delid. The CPU hits high 90s almost immediately with prime 95. But its fine if I use the intel's official test tool. CPU runs in the low 70s and doesn't crash. Is it just prime95 crushes haswell or will delidding help?
|
|
#
?
Sep 26, 2014 16:46
|
|
|
r0ck0 posted:I wonder now if I should delid. The CPU hits high 90s almost immediately with prime 95. But its fine if I use the intel's official test tool. CPU runs in the low 70s and doesn't crash. Is it just prime95 crushes haswell or will delidding help? If you mean Intel XTU (?) that has always been pretty soft for me, but also very comparable with the first 15 min of P95 blend (or in place fft I believe). I'd delid personally just because but frankly its not for any practical reasons unless you're pushing for more speed. 90's is certainly a thermal limit I'm uncomfortable with even if its small fft, in fact its higher than average for the voltage. But heat can be taken care of, voltage per mhz like this is just luck, so this is a good chip imo. By uncomfortable I mean I wouldn't push any more voltage but it seems fine for everyday use. If I had that chip I'd be salivating to delid it just to see if I could hit 5.0 ghz, there's a better chance than most if you're at where you're at now. But again, not for any real reasons (although its not as if this chip doesn't produce results)
|
|
#
?
Sep 26, 2014 17:12
|
|
|
1gnoirents posted:Yeah that's fine, especially with the stock cooler. 1.1v and 99 degrees is not I'm using a Noctua U12s
|
|
#
?
Sep 26, 2014 17:28
|
|
|
r0ck0 posted:I wonder now if I should delid. The CPU hits high 90s almost immediately with prime 95. But its fine if I use the intel's official test tool. CPU runs in the low 70s and doesn't crash. Is it just prime95 crushes haswell or will delidding help? Both, I think. The Overclock Haswell thread thinks Prime95 is pretty rough on Haswell, but I dunno if that includes Refresh CPUs too. Brain Issues posted:I'm using a Noctua U12s I hope our heat problems aren't related to that. e: Reseat seems to have fixed it mostly. Still doesn't like prime95 small fft all that much (high 80s), but a couple other stress tests were in the 60s or lower, and OCCT small data sets broke 80 like once. That was at 4.6GHz & 1.28V Moatman fucked around with this message at 21:59 on Sep 26, 2014 |
|
#
?
Sep 26, 2014 17:34
|
|
|
Speaking of de-lidding, I thought I'd give that a try on my 3570k (I've decided to leave it at 4.4Ghz). Quick trip report; temp difference was -3°C at 4.4Ghz on the hottest core, around -5°C or a little over when I was pushing the voltage trying to get to 4.5Ghz. Difference between the hottest and coolest core was more stark too, before the de-lid, it was around 5°C, now it's closer to 10°C. Even though it didn't yield too much in the way of cooler temps, it was a fun little project, and it does perform better, though there were a few things which could have improved the results if I had avoided them. First, I think I clamped the heat-spreader too tightly, and the teeth of the vice gouged the sides of the heat-spreader, causing some of the metal to rise above the plane of the heat-spreaders surface. This was on the top and bottom, and I "fixed" this by sanding it down by hand. Secondly, there was a permanent stain on the underside of the heat-spreader where the factory TIM had been. Try as I might, I couldn't get it off, so I think this residue may be causing less efficient heat transfer. Lastly, though this is easy to fix (making the effort to remove the cooler from my TJ08-E is more effort than I can be bothered with more than once in a day), I think I put too much thermal paste between the CPU and heat-spreader, and I just used some Artic Cooling MX-2 I had lying around, perhaps a better paste, in conjunction with a less liberal application, might have helped too. tl;dr; just de-lid, it probably won't be worse.
|
|
#
?
Sep 27, 2014 08:00
|
|
|
1gnoirents posted:Yes, but first update your bios. I dont know if they ship with the latest revision now but update it via MSI's utility in Windows. It's actually pretty straightforward and nice (sometimes it isnt). Can you do BIOS overclocking and still get idle states? I started that route but running 4ghz constantly wasn't what I was after. Running MSI's Control Center gives me 1.6ghz idle and 4ghz active, with each core only on as needed. The only thing I bothered to do was select the ratio multiplier and set it to x40. It picked 1.27V vCore and actually sets it to 0.992 when all four cores are idle. This is my experience on a P67A board with a 2500k chip. I could probably push it a lot higher if I wanted but I've never felt CPU-bound. Edit: I'm only hitting 63c on prime95 small on air (212 EVO). Does it change settings after 15 minutes and get hotter or is that a different test that I'm thinking of. Either way, I can push it a lot harder than I have so far. Harik fucked around with this message at 03:10 on Sep 30, 2014 |
|
#
?
Sep 30, 2014 03:01
|
|
|
Harik posted:Can you do BIOS overclocking and still get idle states? I started that route but running 4ghz constantly wasn't what I was after. Running MSI's Control Center gives me 1.6ghz idle and 4ghz active, with each core only on as needed. The only thing I bothered to do was select the ratio multiplier and set it to x40. It picked 1.27V vCore and actually sets it to 0.992 when all four cores are idle. Small fft comes in at 15 min, you can just run that instead. Its called a ctate (or just c1, c3, c6 etc) usually for the variable frequency but I dont remmeber if it was on that board's bios. I have no idea if the clock speed is actually degrading at all though, I was only concerned about voltage. This chip benefits from overclocking in a non linear way for games so its worth it to go higher imo
|
|
#
?
Sep 30, 2014 05:46
|
|
|
Kintamarama posted:Thanks for the input, you guys are awesome. I thought I'd add that the system is rock solid at 4.4ghz with the +.02 offset, my max Vcore in AIDA64, and PS2/BF4 was 1.304, so no shortfall. I've noticed in AIDA64's stress test, you can tick a box to stress the GPU too, so maybe I should add a tick to that box from now on. I'll keep tinkering, but I think that my wall might be 4.5ghz, and I feel as though my low end motherboard might be playing a roll in that, but I can live with 4.4Ghz, I'm not missing out on much if I can't get that last 0.1Ghz. Bit late replying but it you're still having problems I have the same board and CPU and its backwards to what most other boards use for CPU llC, 0% is meant to give you the least compensation but on mine it gives the most. I have mine set with Intel Speedstep Technology = enabled Additional turbo voltage = + 0.031v Cpu offset mode = -0.005v CPU llC = 50% CPU pll = 1.709v Enhanced Halt State (C1E): Enabled CPU C3 State Support: Disabled CPU C6 State Support: Disabled Package C State Support: Disabled This gives me 0.84v VCore @ 1600 when idle and 1.232v at 4600, under load with prime 95 it drops to about 1.216v and bounces around slightly either side of that, with LLC set to 0% it goes way over 1.232v and at 100% it drops down to about 1.18v and is unstable. I increase the Additional turbo voltage to get up to 4.7 and 4.8 but decided the increase in voltage to keep it stable at those speeds over 4.6 is not worth it.
|
|
#
?
Sep 30, 2014 17:28
|
|
|
|
| # ? Apr 19, 2024 09:31 |
|
|
Woah, you've just blown my mind, I had never considered adjusting the turbo voltage instead of the offset. Is this how Asrock names their fixed voltage option? I feel like I've gotten a second wind, I'll definitely give it a shot on the weekend.
|
|
#
?
Oct 1, 2014 10:21
|
|