Uneven pressure in the corner?
1 Like
Maybe; my bracket needs straightened and this cooler has a rounded cold plate.
I tried making it better; I failed to make it better. I need sleep.
The length of the exposed thread seems to differ a little bit. Or maybe it’s the perspective? Anyhow, the results speak for themselves. Maybe try a torque adjustable (dynamometric) screwdriver?
Also the length of that shim overhanging the plywood is different on either side, which means that the the applied torque is different(as in force applied by the screw on the lever, which is the overhang), thus strain and displacement differs as well. Put it simply - the shim might have bent a bit.
If you look closely you can even see that the plywood is more compressed on the right side.
Bent bracket, uneven threads, they kind of even out. I straightened it out anyway; then I got tired of lapping the heatsink and hoped it would be flat enough… apparently not.
Lapping the heatsink is only one part of the equation. Everything has to be even, otherwise you have a cascading effect and the mistakes add up. This is the kind of stuff that professional CNC machine shops get wrong. I’m not saying that improvised and ad-hoc solutions don’t work, but in my experience you have to do a lot of them to get things right or simulate the shit out of it before you get into pre-production prototypes(and even then it’ll need adjustments 90% of the time). I know this isn’t ever going to be production quality, but god damn it I want you to make it work 
Maybe try using standoffs(or a stack of glass fiber washers) between the pcb/metal plate and the noctua cooler? That that might even things a bit, but you have to make sure that the standoffs are of equal(and exactly right) height.
1 Like
A sheet of sandpaper on a hard flat surface. This is the 4th GPU I’ve done this to by the way.
Lapping it is then… I never said you lacked enthusiasm or elbow grease. I did check your yt channel before opening my mouth here.
1 Like
Tried again. Coverage looks to good for it to be that simple, unfortunately.
Fuck the Apple engineering philosophy.
Going to try 4 of my 60mm fans and hope I can turn them down quite a bit; otherwise looks like I’d need to throw a small wad of cash at a beefy cooler.
Well, better. Who ever said it’s just not enough cooling was right.
Right after I made this post, it black screened and rebooted.
not particularly but I’m running it as a frontier to get custom voltage and clocks
I do have a problem with this gt 740, I have no clock control and the driver is zombified but working
I’ll swap another 740 from the office
I am driving a 4k60 on HDMI 2.0 which its not supposed to have but does anyway
Well, I got the cooling better so I didn’t need to mess with voltage; besides, for what I’m doing, I can’t have display out not working. 85c isn’t great, but it wasn’t hitting 105c anymore (HMB was going up to just 51c). Still got occasional artifacting and black screen reboots. So sick of tech not working like it should.
Might make offers on WX4100s and check that EDID emulation stuff.
If you can somehow get the voltage lower somehow, that will give you higher amp ceiling while retaining the same watt ceiling, higher clocks means higher amps
When the ceiling is static undervolting less about temps and more about clocks
I have no idea if analog ways of reducing vcore will factor into it’s calculations, if it does, higher clocks, if it doesn’t then that means lower temps in this circumstance
Might be able drop the clocks to the floor to get it to be more useful than RGB case lights, would still require frontier which means no display, that combined with clocks dropped to the floor… doesn’t even seem worth the hassle of setting up. Just got my 2667 v2, and when I test it, it’ll be with the R5 240.
Trying to be diligent and see if I can cancel my return. Did a fresh install with new 2667 v2. Added an extra fan for base plate. Now to just run it until there’re problems, or until I decide I made an error in judgement.
Idk why the hotspot is so different, mines at most 10c hotter than core
At “stock” voltage?