|
After another week I got another overclock related BSOD during gaming and I decided to take another route rather than add any more vcore. I typically don't mess with BCLK, at least not since earlier AMD days, but BCLK strap seems pretty interesting. However it's pretty hard to find a definition of it. From more or less vague references to it I understand that it's a way of changing your BCLK without messing with the PCI and DMI frequency, which is good, but I don't understand how it will directly effect the ram or how I should compensate. Sadly ram overclocking is one subject I haven't really dove into yet because probably because I have a likely-wrong "overclocking ram is pointless" mindset from 5 years ago. But I will. All I want to see is if I can get the same final numbers stable or with some wishful thinking, lower vcores. Say if I change it to 125 blck strap and the multiplier to 37, that's still about 4.6 ghz, but are there any other considerations for me here?
|
#
?
Apr 2, 2014 15:07
|
|
|
|
| # ? Jun 2, 2024 05:01 |
|
|
Go to 4.5.
|
|
#
?
Apr 2, 2014 15:31
|
|
|
ShaneB posted:Go to 4.5.  I am... for nowI think YOU should pop up to 1.399v and get 4.7
|
|
#
?
Apr 2, 2014 16:10
|
|
|
Ignoarints posted:
Or buy another Haswell and hope for a better lottery ticket payoff. 
|
|
#
?
Apr 2, 2014 16:11
|
|
|
ShaneB posted:Or buy another Haswell and hope for a better lottery ticket payoff. The moment I set money aside for that I will for sure be maxing it out lol. For the first time I set IBT to "maximum" and I almost touched 90 degrees. Made my computer run like complete rear end too after soaking up something like 15 gb of ram to start the test. I'm not sure what good that test does though over standard Edit: I had a long lunch so I tried out the blck strap thing at 1.25x, and while it worked (4620mhz), it doesn't seem to require any less vcore than before. It forced my ram clock to 1666mhz though. 1.66x wouldn't boot no matter how low I set all the frequencies. Oh well Ignoarints fucked around with this message at 19:55 on Apr 2, 2014 |
|
#
?
Apr 2, 2014 16:27
|
|
|
Ignoarints posted:
I've never had 4.6 stable even in the low 1.3's and I don't really want to go higher even though my temps are pretty chill... I could try it for the academic exploration, I guess.
|
|
#
?
Apr 2, 2014 19:56
|
|
|
Ignoarints posted:After another week I got another overclock related BSOD during gaming and I decided to take another route rather than add any more vcore. I typically don't mess with BCLK, at least not since earlier AMD days, but BCLK strap seems pretty interesting. However it's pretty hard to find a definition of it. From more or less vague references to it I understand that it's a way of changing your BCLK without messing with the PCI and DMI frequency, which is good, but I don't understand how it will directly effect the ram or how I should compensate. Sadly ram overclocking is one subject I haven't really dove into yet because probably because I have a likely-wrong "overclocking ram is pointless" mindset from 5 years ago. But I will. You have a good idea of how the BCLK strap works, and your instinct is pretty much on target not to mess with it. It doesn't have much use with Haswell. The main practical use of the BCLK strap is to get higher frequencies when your limiting factor is the clock multiplier PLL. In Sandy Bridge, this limit was usually around 4.9-5 GHz. Haswell really can't clock that high, so the PLL doesn't really need to be worked around. As for RAM overclocking, it has its own separate multiplier. Again, the BCLK strap can be used to help with ultra-extreme multiplier, but we're talking around DDR3-2933 here. The fastest Haswell will ever be is DDR3-2400 CL9 speed/timings, and that's only a percentage point or so above DDR3-1866 CL9. There was an edge case with the i7-3820, a non-unlocked Sandy Bridge-E chip. Using the BCLK strap could push that chip to 4.4 GHz despite it being "locked," so that was cool. But not required for an unlocked chip. So: BCLK strap is for setting extreme records. It's not for practical performance, and it's very likely not relevant to you.
|
|
#
?
Apr 2, 2014 20:02
|
|
|
Factory Factory posted:You have a good idea of how the BCLK strap works, and your instinct is pretty much on target not to mess with it. It doesn't have much use with Haswell. Yeah I actually edited a post above, I tried it (quickly) on lunch. It worked, but from my brief test of it there was no benefit, it seemed to require the same vcore and readily crashed in IBT with even 0.01v less. One weird thing it did was make my integrated graphics voltage 25% higher though. ShaneB posted:I've never had 4.6 stable even in the low 1.3's and I don't really want to go higher even though my temps are pretty chill... I could try it for the academic exploration, I guess. You're at 4.5 at less than 1.3 volts right? Your chip is very lucky that I do not own it. I'd probably being running 4.8 at 1.47 vcore or something dumb Ignoarints fucked around with this message at 20:36 on Apr 2, 2014 |
|
#
?
Apr 2, 2014 20:13
|
|
|
So I bought a QNIX and overclocked it to 96hz, which is all well and good. But now even while just at the desktop, my graphics cards don't idle. If I set the monitor back to 60hz they go into their idle state. The increased noise of the cards at desktop is really quite a big deal (I have two 570s). If I force an idle PState with nvidiainspector it works fine, but I'd like it to work automatically. Anybody got any ideas?
|
|
#
?
Apr 3, 2014 12:17
|
|
|
Well for kicks I tried to hit 4.7 on my i5. Even at 1.36vcore and 2.06 in it eventually blue screened. I tried the same settings for 4.6 and no dice. Uncore is at 38. Since I can run 4.5 at like 1.26 and 1.95 or something like that I'm just going back to that... Failure!
|
|
#
?
Apr 3, 2014 14:42
|
|
|
ShaneB posted:Well for kicks I tried to hit 4.7 on my i5. Even at 1.36vcore and 2.06 in it eventually blue screened. I tried the same settings for 4.6 and no dice. Uncore is at 38. Since I can run 4.5 at like 1.26 and 1.95 or something like that I'm just going back to that... Failure! I don't know if this is happening to you too, but my uncore would always go to 4.0 ghz if I set it to 34 multiplier with turbo boost off. It was really confusing until I found that and it was always a source of instability. For whatever reason I have to set it to 35. One "benefit" of blck strap I noticed was the uncore seemed to be exactly what I wanted it to be at. It was especially weird that it was 4.0ghz since that number doesn't really apply to anything, its higher than turbo boost. I know you're probably not super interested in getting over 4.5ghz but once I set my ring bus to 1.20v, uncore to 35 multiplier, vrin to ~2.10 the sky was the limit on clock speed. Well, sky being vcore. Before that I had quite a bit of trouble at any settings above 4.4. That vrin is high, but that was for 4.7, I typically run 1.950 for 4.6 and 4.5ghz so far. It can probably be lower on 4.5 but I haven't worked on lowering that one more yet.
|
|
#
?
Apr 3, 2014 14:50
|
|
|
Ignoarints posted:I don't know if this is happening to you too, but my uncore would always go to 4.0 ghz if I set it to 34 multiplier with turbo boost off. It was really confusing until I found that and it was always a source of instability. For whatever reason I have to set it to 35. One "benefit" of blck strap I noticed was the uncore seemed to be exactly what I wanted it to be at. It was especially weird that it was 4.0ghz since that number doesn't really apply to anything, its higher than turbo boost. HWiNFO is reporting my uncore at 3800MHz. I could try to drop that I suppose. Edit: man that guy just ain't going faster than 4.5. I jacked the vcore to 1.35, dropped uncore to 35, jacked vin to 2.1, and still wouldn't even hold 4.6. I'd rather just run a way lower voltage and be stable and sub-75C even with synthetics at 4.5. ShaneB fucked around with this message at 17:15 on Apr 3, 2014 |
|
#
?
Apr 3, 2014 16:55
|
|
|
ShaneB posted:HWiNFO is reporting my uncore at 3800MHz. I could try to drop that I suppose. high five lottery winners  who am I kidding, this is still incredibly fast
|
|
#
?
Apr 3, 2014 17:19
|
|
|
BurritoJustice posted:So I bought a QNIX and overclocked it to 96hz, which is all well and good. But now even while just at the desktop, my graphics cards don't idle. If I set the monitor back to 60hz they go into their idle state. The increased noise of the cards at desktop is really quite a big deal (I have two 570s). If I force an idle PState with nvidiainspector it works fine, but I'd like it to work automatically. Anybody got any ideas? You're overclocking a monitor? What the gently caress?
|
|
#
?
Apr 4, 2014 19:56
|
|
|
atomicthumbs posted:You're overclocking a monitor? What the gently caress? My reaction when I found this out as well. I'm glad its possible though, 120hz monitors are expensive as poo poo
|
|
#
?
Apr 4, 2014 20:18
|
|
|
Has anyone just filled the entire void under the IHS with TIM before?
|
|
#
?
Apr 11, 2014 22:39
|
|
|
Wait, it's not normally filled? I thought the problem came from too much TIM resulting in poor conductivity, not that there wasn't enough to fill the space.
|
|
#
?
Apr 11, 2014 22:44
|
|
|
craig588 posted:Wait, it's not normally filled? I thought the problem came from too much TIM resulting in poor conductivity, not that there wasn't enough to fill the space. No it's not even close, the die itself is super small. It's been proven more or less (by one guy, but very thoroughly) that the glue is too thick and it just keeps the IHS too high off the cpu. Either way, when you replace the TIM you just cover the cpu itself. I was just thinking the other day what if you just filled it completely with non conductive stuff. Most likely nothing, but who knows. I'd have done it right then and there but I don't really want to replace my CLU (and I can't fill it with that)
|
|
#
?
Apr 11, 2014 23:16
|
|
|
Ignoarints posted:No it's not even close, the die itself is super small. It's been proven more or less (by one guy, but very thoroughly) that the glue is too thick and it just keeps the IHS too high off the cpu. Either way, when you replace the TIM you just cover the cpu itself. I was just thinking the other day what if you just filled it completely with non conductive stuff. Most likely nothing, but who knows. I'd have done it right then and there but I don't really want to replace my CLU (and I can't fill it with that) It wouldn't act as a larger heatsink, if that's what you are thinking. It's meant to act as a super thin layer between the die and the heatspreader/heatsink to fill the very minute air gaps, because air is an excellent insulator. It's not going to like... spread the heat from the die across a larger area into the heatspreader.
|
|
#
?
Apr 11, 2014 23:20
|
|
|
ShaneB posted:It wouldn't act as a larger heatsink, if that's what you are thinking. It's meant to act as a super thin layer between the die and the heatspreader/heatsink to fill the very minute air gaps, because air is an excellent insulator. It's not going to like... spread the heat from the die across a larger area into the heatspreader. My "reasoning" was that relative to the surface area of the top, the sides of the die amount to a good amount of area. I'm quite sure some of the TIM squeezes over the edge, but it seems like it has no real efficient way of moving that heat anywhere. This actually started off in my head wondering about how if the space was under pressure heat would conduct better. Whether or not that would matter (as in if the capacity to apply heat to the IHS was already greater than the ability to remove it, which would be logical) is another issue, but it got me thinking that perhaps if the void was just filled with TIM would it allow the heat to spread slightly more efficiency to the IHS, mostly due to the sides of the die. Would a cooler work less well if the heat was actually more concentrated in the center of the IHS due to the cpu's location under it? Like a hot spot. If so, would spreading it out (even if a little) help at all? I have no idea, well I pretty much doubt it would help even if it were true because of the cooler, but I really don't know.
|
|
#
?
Apr 11, 2014 23:34
|
|
|
Ignoarints posted:My "reasoning" was that relative to the surface area of the top, the sides of the die amount to a good amount of area. I'm quite sure some of the TIM squeezes over the edge, but it seems like it has no real efficient way of moving that heat anywhere. Just make a non-conductive metal shim that goes around the die perfectly to increase the surface area.
|
|
#
?
Apr 11, 2014 23:41
|
|
|
ShaneB posted:Just make a non-conductive metal shim that goes around the die perfectly to increase the surface area. Haha about all I could pull off is taking a dump on my cpu die with TIM. gently caress it I might try it later when I get my monitor
|
|
#
?
Apr 11, 2014 23:51
|
|
|
Thermal paste is practically an insulator compared to copper. I don't remember the exact figures, but copper has a thermal conductivity somewhere around 300 while even the best pastes are somewhere around 10. ShaneB is right. Compared to air gaps it's hundreds (Probably thousands) of times better, but heat will spread through the IHS so much better than it'd pass through a mound of paste that all you'd be doing is making a mess. Edit: It's why you always see custom machined aluminum blocks serving as heatsinks for weird shapes in prototypes, never custom formed thermal epoxy. craig588 fucked around with this message at 23:59 on Apr 11, 2014 |
|
#
?
Apr 11, 2014 23:53
|
|
|
craig588 posted:Thermal paste is practically an insulator compared to copper. I don't remember the exact figures, but copper has a thermal conductivity somewhere around 300 while even the best pastes are somewhere around 10. ShaneB is right. Compared to air gaps it's hundreds (Probably thousands) of times better, but heat will spread through the IHS so much better than it'd pass through a mound of paste that all you'd be doing is making a mess. Hmmm. Well I understand it is no way ideal, but it is very simple and easy to do. I wonder if it will be better. At the moment a (small) portion of the cpu itself is just exposed to air. It might be one of those technically better but practically immeasurable things. A mess is guaranteed. I wouldn't mind that, I don't want to actually gently caress it up in some other way though. Can't imagine how though, unless the TIM is electrically conductive even if it says it isn't.
|
|
#
?
Apr 12, 2014 03:38
|
|
|
Kinda on subject, MSI announced that their new Z97 XPower AC motherboard will come with a shim for directly mounting a heatsink to a CPU die. This has been done before with the EK Precisemount Naked Ivy before, but this new system allows the socket bracket to remain on.
|
|
#
?
Apr 12, 2014 04:28
|
|
|
That reminds me a lot of the copper shim I had on my 1.4ghz AMD thunderbird so as not to crack the core with a bigass heatsink (heat pipes were new and not really in the market yet)! It still works, but the 38x80mm fan on it sounds like a vacuum.
|
|
#
?
Apr 12, 2014 04:33
|
|
|
BurritoJustice posted:Kinda on subject, MSI announced that their new Z97 XPower AC motherboard will come with a shim for directly mounting a heatsink to a CPU die. This has been done before with the EK Precisemount Naked Ivy before, but this new system allows the socket bracket to remain on. Lol I love the pictures they added. It's like someone just needed something to do.
|
|
#
?
Apr 12, 2014 04:38
|
|
|
BurritoJustice posted:Kinda on subject, MSI announced that their new Z97 XPower AC motherboard will come with a shim for directly mounting a heatsink to a CPU die. This has been done before with the EK Precisemount Naked Ivy before, but this new system allows the socket bracket to remain on. This owns.
|
|
#
?
Apr 12, 2014 06:41
|
|
|
ShaneB posted:This owns. Ironic this is for the next chipset, which will probably greatly reduce the desire to delid. But I still would most likely since it would be the bare chip
|
|
#
?
Apr 12, 2014 06:51
|
|
|
The Haswell Refresh K-CPUs are supposed to be soldered on again, aren't they? Wouldn't delidding those shred the die?
|
|
#
?
Apr 12, 2014 08:01
|
|
|
I have an I5 2500 3.3 ghz. I'm just looking too up it slightly to try and get more fps in tf2 and arma2 (both very cpu dependant games). Could I just get away with using the software that came with my mobo for a slight overclock?
|
|
#
?
Apr 12, 2014 09:17
|
|
|
Sir Unimaginative posted:The Haswell Refresh K-CPUs are supposed to be soldered on again, aren't they? Wouldn't delidding those shred the die? That's just speculation, nobody knows.
|
|
#
?
Apr 12, 2014 16:24
|
|
|
All the tests I read running bare die vs. CLP between the die and heatspreader were basically within 1-2C or something like that, anyway. I just like the idea.
|
|
#
?
Apr 12, 2014 16:47
|
|
|
Jippa posted:I have an I5 2500 3.3 ghz. I'm just looking too up it slightly to try and get more fps in tf2 and arma2 (both very cpu dependant games). Yes for slight overclock sure. Keep in mind it might just go insanely overboard and then crash if you autotune. I imagine its okay if you simply manually set what you need. A long, long time ago I used software for overclocks. The bios seems annoying to learn, but in the end its actually easier. And better. If you have a non-k though you can only go so far and you can probably get away with software stuff, as long as you understand its not ideal. Frankly even then I'd just set the multiplier to 41 (max for non k for that if I understand) and let it turbo boost how it wants. Auto voltage would likely be fine in that instance, if a little wasteful.
|
|
#
?
Apr 12, 2014 23:34
|
|
|
I feel like I bump this thread a lot lol I want to try my retarded make-a-mess under my IHS, but does anyone know how to remove CLU? I have a thermal paste removal "set" which is just one thing that melts TIM then one bottle of what smells like rubbing alcohol. But I'm not sure that would work the same way, and I'd like to avoid hassle or delays. I saw a video from the CLU people but there was no sound to explain what they were using (looking like a rubbing alcohol pad?).
|
|
#
?
Apr 15, 2014 15:32
|
|
|
Ignoarints posted:I feel like I bump this thread a lot lol Yeah it's rubbing alcohol, the comments say so.
|
|
#
?
Apr 15, 2014 15:56
|
|
|
deimos posted:Yeah it's rubbing alcohol, the comments say so. Oh nice, thanks. Will report my results/disaster/ruined CPU (and/or 30* drop. Well, maybe, I'd probably sell Intel the "technology" for 10 million first) Ignoarints fucked around with this message at 16:01 on Apr 15, 2014 |
|
#
?
Apr 15, 2014 15:58
|
|
|
there is no way this will work how you think it will
|
|
#
?
Apr 15, 2014 16:14
|
|
|
ShaneB posted:there is no way this will work how you think it will If you think that's bad, look what I just made in paint at work because I'm bored  It's a sealed cooler that boils water (or something) in a vacuum (28 inches makes for 30* boiling point or so) that then rises to an air/air cooler and then returns back as water. edit: because as we know from 7th grade, water that is boiling doesnt rise above its boiling point (I know this is not true)
Ignoarints fucked around with this message at 16:30 on Apr 15, 2014 |
|
#
?
Apr 15, 2014 16:28
|
|
|
|
| # ? Jun 2, 2024 05:01 |
|
|
That's just a heat pipe without the capillary material so it'll only work in one orientation.
|
|
#
?
Apr 15, 2014 16:46
|
|