Overclock.net banner
1 - 20 of 57 Posts

·
Registered
Joined
·
2,963 Posts
Discussion Starter · #1 · (Edited)
These brackets, if all made about the same, really do work, check out how even the pressure is, and improved thermals.

I installed the Thermalright LGA1700-BCF bracket, and also went from cheap ebay TIM, to some waxy TIM, TF-7 , and my temps did improve by 5-10C. Way more than lapping did. I was thinking it was mainly the TIM, I'll compare the TIM's on my old CPU someday.
Rectangle Font Fashion accessory Gadget Metal


I could probably tighten the bracket down more evenly, and a little less, but I tried to get it even.

Before that I also had lapped my CPU, but not the convex cooler. In the thermal wax left on my CPU, I could see the concave pattern (reverse) . So I think it made pretty good contact. I could lap the air-cooler, and my old CPU too.

Now I have a ALF2 420mm on there, and it doesn't tighten down anywhere near like my air cooler did. I never checked how flat it was either, but I'll remove it soon to see what the TIM spread is like. I'd lap that too, but it's so new, I'd like to keep it's warranty for awhile.

The TF-7 was really hard to spread, I got MX-5 iirc with the ALF2, so I should try that too.
 

·
Registered
Joined
·
719 Posts
I think you need to have a problem, before it can fix anything. Not all boards have this problem.
 

·
Registered
Joined
·
3,698 Posts
I'm scared of AMD's AM5 as it's going to be LGA. I've never in my life installed any of those.
G34 was a good AMD socket, and it was LGA. The retention mechanism worked and I never had any issues with good thermal transfer. Same with LGA 3647 for Intel (although that doesn't actually have a frame to clip down)... The Threadripper/Epyc sockets are decent too - although if you were rushing or not paying attention the thin plastic frame around the CPU could miss the guide sleds and if you tightened it all down you were in for a world of pain.

I don't know whether Intel are just cheaping out on their consumer sockets, or whether the board manufacturers/socket ODMs are, or what.

I wouldn't worry about AM5 until we see what it's like.
 

·
Madonion.com
Joined
·
339 Posts
You've probably botched lapping the first time like most people do. In order to do it right, you need to cut out the socket from a Z690 board, secure your chip in it and then lap. If you do it outside, it will bend again as soon as you put it back in the socket.
Now that you have a custom frame that puts (hopefully) an even pressure on your chip, I'd advise you to lap it again.
 

·
Registered
Joined
·
916 Posts
I ended up with about an overall 7-8C temperature decrease with the frame that gave me some modest additional overclocking headroom, but still have about the same core to core deltas (odd numbered cores seem to run hotter than even).
 

·
Registered
Joined
·
719 Posts
There are to many variables:
Lapping or no lapping IHS
Lapping or no lapping cooler
in or outside frame
IHS condition
clamp pressure, for example my Z690 Pro was a LOT stiffer than my Unfiy-X
Bracket positioning
Air, AIO or custom?

Not sure if one can make general statements
 

·
Overclocking Enthusiast
Joined
·
6,887 Posts
I'm scared of AMD's AM5 as it's going to be LGA. I've never in my life installed any of those.
AMD has been using LGA almost as long as Intel. Intel's 775 was 2004. And AMD's 1207 was 2006. However AMD uses LGA mostly for server sockets. So it's AM5 socket should be built well.
 

·
Registered
Joined
·
719 Posts

·
Registered
Joined
·
1,899 Posts
It's made me actually prefer having the pins on the chip lol

Sure, if you are careless and break a pin on a CPU it's more costly than on a mobo, but the price of mobos has been ever increasing.
 

·
Registered
Joined
·
17 Posts
It's made me actually prefer having the pins on the chip lol

Sure, if you are careless and break a pin on a CPU it's more costly than on a mobo, but the price of mobos has been ever increasing.
There's nothing inherited wrong with PGA design, it's shown to be well capable and reliable over decades. People have re-bent or even re-put pins back in place and CPUs worked. More so, the pins are nowhere as easy to be bent as those tiny pins on LGA sockets. A LGA socket like 2066 is simply unrecoverable once you mess it up. I did do mess my X299 and now it barely boots / 3 RAM slots work. On the other hand, I've luckily didn't screw up my 3900X... I once pulled it straight out of the socket due to using too glue'ish TIM. The pins are pretty tightly put down there, I'd say.
 

·
Premium Member
Joined
·
6,774 Posts

Now I have a ALF2 420mm on there, and it doesn't tighten down anywhere near like my air cooler did. I never checked how flat it was either, but I'll remove it soon to see what the TIM spread is like. I'd lap that too, but it's so new, I'd like to keep it's warranty for awhile.

The TX-7 was really hard to spread, I got MX-5 iirc with the ALF2, so I should try that too.
Hello there. Could you post some HWINFO CPU FREQ/TEMP data while/after running a Cinebench R23 loop? I am curious what your particular rig is showing for min/max/avg with the 420mm.

Thanks!
 

·
WaterCooler
Joined
·
5,088 Posts
There's nothing inherited wrong with PGA design, it's shown to be well capable and reliable over decades. People have re-bent or even re-put pins back in place and CPUs worked. More so, the pins are nowhere as easy to be bent as those tiny pins on LGA sockets. A LGA socket like 2066 is simply unrecoverable once you mess it up. I did do mess my X299 and now it barely boots / 3 RAM slots work. On the other hand, I've luckily didn't screw up my 3900X... I once pulled it straight out of the socket due to using too glue'ish TIM. The pins are pretty tightly put down there, I'd say.
Oh I ripped my 5900X out of the socket when removing a cooler once and it bent several pins. I took my time and bent them back. Was very nerve racking. First time I ever dealt with a PGA style board/CPU.
 

·
Registered
Joined
·
17 Posts
LGA is worse tbh, unless there are very little amount of pins laid down on a wide open area, which is like never the case. With PGA I'd recommend not using TIM that literally glues your IHS to the heatsink... some cheaper TIM do this. Even if you did it, try stuff like heating that spot with a hair drier or something before going around and doing a strongman pull on the heatsink... lol.
 

·
WaterCooler
Joined
·
5,088 Posts
LGA is worse tbh, unless there are very little amount of pins laid down on a wide open area, which is like never the case. With PGA I'd recommend not using TIM that literally glues your IHS to the heatsink... some cheaper TIM do this. Even if you did it, try stuff like heating that spot with a hair drier or something before going around and doing a strongman pull on the heatsink... lol.
I just learned how to take the cooler off without ripping it up. Now I am using MX-5 and doesn't seem to glue it.
 

·
Premium Member
Joined
·
259 Posts
Oh I ripped my 5900X out of the socket when removing a cooler once and it bent several pins. I took my time and bent them back. Was very nerve racking. First time I ever dealt with a PGA style board/CPU.
Did this too with my first round of PGA since... probably the Athlon64 X2?

Problem in this case was an ITX board with heatsinks on three sides of the socket and a heatsink that just fit. Rotating was not an option, and the CPU came out glued to the bottom of the HSF. One row of pins along an edge bent, which when bent back resulted in a still-functioning CPU.

On the other hand, I've bent LGA pins, and there's no coming back from that. Whatever damage done is likely to be permanent.

but the price of mobos has been ever increasing.
So, I wanted a Z690 board with... well, Thunderbolt 4 built in.

Ruling out Gigabyte's and ASUS' creation-oriented boards, with the Gigabyte board breaking three kits of DDR5 (written about elsewhere) and ASUS having wonky PCIe configurations, I went with the next cheapest MSI board - the MEG Z690 ACE. Cost more than a 12900K, and I'm using a 12700K in it.
 

·
Registered
Joined
·
719 Posts
I ordered 3 thermalright frames from Ali, let’s see if it does anything.
 

·
Registered
Joined
·
183 Posts
Achieved average of 7C, up to 10C improvement with the Thermalright frame. With the socket removed I was able to properly tighten my EK block backplate. The block backplate kept the ILM backplate in place while I replaced it. Installation was simple and while I ensured I tightened all screws evenly in an X pattern, I didn't go crazy over it. Had no issues with dual-channel or posting after installation which can be caused by uneven contact with LGA pins. I decided to use MX-4 instead of GELID. Always used it before, but when I removed my i7-6700k while upgrading to Alder Lake, the paste had basically become cement. Had to use much more force than I would have liked to in order to remove waterblock. MX-4 seems like a good choice, not really concerned with the 1-2C I could get with best thermal Grizzly non-LM TIM. MX-4 should be able to be removed without as much force of Gelid which becomes a very sticky putty over time.

One important thing to remember is to press CPU down when removing the bracket. A few users described the CPU coming out of socket with the frame. It creates the potential to have the CPU fall onto the pins. I believe there is a report of one user destroying the board by this occurring. Overall I am very pleased with the contact frame as it was $6 to have it shipped. Saving 7C and preventing socket warping in the future makes it well worth it IMO.

I'd be interested if GN looked at TG vs Thermalright bracket or at least examined the difference in dimensions and manufacturing tolerances. I expect that TG would be more consistent, but costs approximately 5X+ what the TA bracket costs.
 

·
Registered
Joined
·
2,963 Posts
Discussion Starter · #20 · (Edited)
Hello there. Could you post some HWINFO CPU FREQ/TEMP data while/after running a Cinebench R23 loop? I am curious what your particular rig is showing for min/max/avg with the 420mm.

Thanks!
Ok but I haven't tuned up the voltage at all, I just went back to adaptive.

The main difference tho, is how much higher I can set the power limit. Before, ~170W was sure to be hitting 95-100C, and now it must be up around 260W, that some cores will be hitting 100C. It's summer here, and my room gets fairly hot, like 25-26C, before benchmarking that is.

Right now CB23 should run at 50x and 1.36V, back in a minute.

Here's 10min run
So that's with turbo offset ratio OC mode, 8P at 50x, Ecores at 39x, I should put them up again too.

The CPU down clocks a lot, during the start of each new run, not for throttling tho, I think it's just the TRO mode, power doesn't ramp until a moment in either. Otherwise it's at 50x in the test.
Property Building Rectangle Chair Window

Font Parallel Terrestrial plant Rectangle Screenshot


Maybe thats not readable. Also where's the pause button on HWinfo, the average power was ~253W
Font Screenshot Electronic device Multimedia Software

Font Parallel Pattern Screenshot Number

Font Parallel Number Pattern Document


What it calls the pump there, is just 2 other fans. My ALF2 doesn't have the separate controller. Or any RGB, but thats good, no extra heat or balance issues.
 
1 - 20 of 57 Posts
Top