Investigating Intel's CPU Socket Problems | Thermal Grizzly Contact Frame Benchmark
Вставка
- Опубліковано 5 чер 2024
- Sponsor: Linode Hosting - get $100 credit with GN (geni.us/UaKrUmF)
Intel's CPU socket for the Alder Lake (12th Gen) CPUs applies higher pressure on one side, causing reduced pressure between the cooler coldplate and CPU. We're replacing the stock Intel ILM (on the socket) with a Thermal Grizzly Contact Frame. Testing includes analysis of pressure on the 12900KS, 12900K, and other CPUs, core-to-core deltas, thermals, and more.
The best way to support our work is through our store: store.gamersnexus.net/
-- we currently have toolkits, modmats, mouse mats, and shirts in stock! The PC building modmats are VERY LOW stock, so if you want one, buy it soon! store.gamersnexus.net/product...
We're pushing a new Patreon bonus behind-the-scenes video tomorrow! Support us and also get access here: / gamersnexus
RELATED PRODUCTS [Affiliate Links]
Thermal Grizzly Contact Frame on Amazon: geni.us/YTTkS
Competing (Alternative) 12th Gen Contact Frame on Amazon: geni.us/hqOs5BX (we have not tested this one)
Arctic MX-4 Thermal Paste on Amazon: geni.us/k3kT6
RELATED VIDEOS
You'd also like our investigation into GPU power consumption issues: • The Brewing Problem wi...
And our CPU cooler reviews playlist: • Liquid Cooling vs. Air...
TIMESTAMPS
00:00 - Intel's Socket Problem
02:09 - How the Bending Happens
03:44 - Lots of Play in the Socket
05:28 - Bad Thermals Core-to-Core & Methods
06:40 - Pressure Scanning the IHS-to-Coldplate
07:17 - Extraordinary Results on Thermal Grizzly Contact Frame
07:42 - 12900KS Contact Frame Pressure
08:10 - The Best Results Yet
08:47 - Thermal Benchmarks of Intel Contact Frame vs. ILM
10:21 - Loose Socket Tests
11:35 - Alternatives to Intel's Socket
14:35 - Why Intel Doesn't Do This
15:20 - Damage Potential & Installation
19:50 - Conclusion: Fit & Finish Analysis
21:09 - Additional Concluding Thoughts
** Please like, comment, and subscribe for more! **
Links to Amazon and Newegg are typically monetized on our channel (affiliate links) and may return a commission of sales to us from the retailer. This is unrelated to the product manufacturer. Any advertisements or sponsorships are disclosed within the video ("this video is brought to you by") and above the fold in the description. We do not ever produce paid content or "sponsored content" (meaning that the content is our idea and is not funded externally aside from whatever ad placement is in the beginning) and we do not ever charge manufacturers for coverage.
Follow us in these locations for more gaming and hardware updates:
t: / gamersnexus
f: / gamersnexus
w: www.gamersnexus.net/
Test Engineering, Writing, Editing: Steve Burke
Test Technician, Research: Mike Gaglione
Video, Editing, Animations: Andrew Coleman - Ігри
Wow! Lots of interest from comments to see more of these. We're working on getting the cheaper Thermalright contact frame and some others from Amazon. Exciting to have a new review category! We'll work on a round-up. Tell us about other frames you've found!
You'd also like our investigation into GPU power consumption issues: ua-cam.com/video/wnRyyCsuHFQ/v-deo.html
The best way to support our work is through our store: store.gamersnexus.net/
-- we currently have toolkits, modmats, mouse mats, and shirts in stock! The PC building modmats are VERY LOW stock, so if you want one, buy it soon! store.gamersnexus.net/products/modmat-volt-large
And our CPU cooler reviews playlist: ua-cam.com/video/7VzXHUTqE7E/v-deo.html
It'd be cool do to a retrospective of all the socket mounting types from socket 3/7 to now :D
I'd love to see a comparison on a couple of things:
1. As a worst case scenario for manufacturing tolerances and material, I'd love to see the results with a 3d printed bracket like Luumi has tested.
2. I'd like to see how this would compare to the "washer mod" and the general advice you'd give to people who may be pursuing it
can u make a video about whats going on about z690 socket usb problems, on reddit have tread about this and ppl dont know what happening. i almost buy z690 but found that tread and dont know what to do, mb better buy 12700f and b660?
mdonaberger is such a funny username
Intel : Ty for tips for free ... it was worth a millions
So nice of Thermal Grizzly to do all the engineering Intel forgot to do.
ooooooooooof
...all this talk about brackets, clamps and paste as equalizer - are we sure it's not someone's dentures? 😆
@@EpicGamingEct Thermal Grizzly at least made sure their product worked *Before they sold it
@@EpicGamingEct ? What Intel does not send free stuft ever.. Which mail? Providence a link !!!
True
This is probably one of the biggest reasons why i keep coming back to gn. At this point, im a little burned out on the flashier parts of pc's like cpu's and gpu's. But the gn team shining light on some of the less flashy and more in-depth/niche parts like case fans, cpu sockets, cases, psu's and the like never fails to fascinate me. Keep up the good work!
Thanks so much! It keeps it really interesting for us too, and it's a great reminder of how cool PC DIY is when we get away from the usual path!
Totally agree. I've been burned out by the regular stuff but this was so interesting!
For people with just one computer they're relying on, or mission critical systems, this kind of product analysis is essential and I am personally grateful GN are covering it.
I agree with the general point, but I'm not sure case fans and cases fall into less flashy these days. Especially the cases. Often they're pretty hyped up too, and the especially flashy cases may even cost almost as much as the CPU you're popping into the build.
Exactly! GN does everything others do, BUT they go deep into things that others don't ever mention; sometimes to foreign language depths!😂😂😂
When I don't know the words or acronyms, they're going deep.😂
I really appreciated Mike's installation walk-through, thank you guys. I work in China and the Thermal Grizzly contact frame isn't available here yet, so I got the Thermalright one for the equivalent of $6. Comes with a torque wrench as well, bonus. The quality is very good but it could use those witness marks the TG has. Installing it on my Aorus Z690 Master reminded me of the original Athlons. Back then they didn't have an IHS so users would use a contact frame to keep from chipping the actual chip. After installing the Thermalright my idle temps were only reduced by a degree or so, but my load temps dropped 7 degrees during Prime95 using a Corsair H150i Elite LCD. Definitely worth the small cost in time and money.
should we really be trusting a 6$ bundle torque wrench? would love if steve answered that for us.
@@Layarion It's not a torque wrench. It's an allen key. No idea what he's on about.
You don't need the witness marks if you are using a torque wrench. The witness marks are for those using a standard driver or Allen wrench.
Shout out to whoever is responsible for the new animations! Great work! Yall have gone a long way and absolutely love the content this channel always provides. So proud of yall! p.s. love the screwdriver set and coasters! I use them all the time now :)
The deep colors are really noticeable.
Literally the only channel I’ve watched where all the “Blender-Machines” are actually used. Others have graphic artists that’ll make intros and other still images for videos, or even use blender to make merch, and I always wondered why computer/tech/gaming channels dont use blender/etc to animate things. The more involved tech channels rely on existing animations pulled from published works or some internal developer videos.
Then I wondered why Steve put so much emphasis on the GN logo they made. They had A FEW animations but the main draw was animated bits like the 2020 end of year clusterfk, etc. I was so confused why Steve would bring up their 3d work so often when in their most opportune videos they weren’t used, like the AIO placement video, or factory tours.
Now I see. Took em a bit to get here, but yeah, please keep making these animations, and tackle complex topics with great 3d presentations. The best textbooks and smartest people cant explain simple topics without using 3d media/real life models.
Guys the content you've been releasing after your break is the best I've seen from any UA-cam tech channel recently. Keep up the great work!
Thank you! You're starting to see the stuff we were trying to get around to doing!
I've spent 9 days in intensive care, and watching Gamers Nexus has kept me largely sane :)
I love how you took the time to cover the strengths and weaknesses of the installation method for the thermal grizzly frame, showing alternative mass produced ILMs, letting us know that alternatives to the thermal grizzly holder exist and how much they cost.
Gamer's nexus and LTT fighting it out to see who can report the most accurate information about computers by buffing their own testing systems, down to the micron. This kind of accurate reporting, which removes as much bias as possible, is the kind of respectable reporting that i love to see.
unfortunately it's also harder to get because as you said, money for said equipment.
@@Layarion harder to get what?
@@MikoOhneHose equipment that lets you give accurate information.
I'm sooooo fkn here for it
@@Layarion with the views and shops that the two companies are running, it means they have the money for it.
1 year later - I've installed several contact frames now, and following these installation instructions I've had zero problems with posting, memory, or temperatures. Great job guys.
Have you seen cooler temps?
@@JB-ue6lf i did mine yesterday, dropped 20c
Am I correct here in following mikes advice on installation:
Reverse the screw until you feel the screw fall into correct place.
Once you feel the screw fall into place, rotate screw 90 degrees.
After rotating the screws diagonally 90 degrees, continue to add another 45 degree of rotation to all 4 screws (at this point you should feel small resistance).
After this point rotate a further 90 degrees diagonally to all 4 screws and that’s it??
@@James-fv1cm yes the first reverse to get the thread straight and not cross-threaded is good, but on my recent one i just screwed them in until near snug then lightly tightened them in x pattern
@@KentRoads thank you. So for you personally did you not worry about the exact measurements of rotation??
Looks like it needs some of the 2011 style ILMs with the double retention arms!
I have a pair of 2011 v0 systems and that retention system is really nice... altho somewhat disconcerting when using it for the first time sine it takes SO much force. I run Xeon E5 2697s and they stay amazingly cold for 150 watt CPUs
My thought exactly.
@KR4FT W3RK I am still rocking a 2690 V2 @ 4.0GHz at stock voltage. Temps never get over 70 with a NH-U12A. What a beast.
Yup, I have the dual socket version of the HP Z640 (2016 workstation) that uses this dual lever type retension on both CPUs (mobo & add-on riser). Recently upgraded old E5-2620v3 2.4GHz 80W CPUs to E5-2637v4 3.5GHz 135W and temps dropped (Noctua NT-H2 paste used) to average 39-50 degrees. My bottleneck now is my GTX 960. Been waiting so long, what's a few more months LOL. Best thing is this processor series is 1 gen prior to Win11 compatibility (even though I could update TPM) so no auto upgrade LOL.
Exactly what I thought. the 2011/2066 ILMs are a proven design as well.
Mike's installation segment was great. Such easy to follow instructions - true fitment for dummies stuff. Nice concise and well explained job!
Yeah, Mike is doing an Awesome job in front of the camera with his explanations
+1 on the Mike props. He's another great team member. Steve is building a heck of a team!
I got a Core i7 for my girlfriend's computer that ran hot. When I pulled the heatsink off again, I could see visible spots where the heatsink goop had never touched the die cover. I wet-sanded the die cover (glass tables are great for this, because they're almost perfectly flat) until I got through the nickel plating down to the bare copper, polished it mirror-smooth, and reinstalled it. The CPU ran 15°C cooler.
(on a side note, the code for the Degree symbol is ALT+0176.)
Why do huge corporations come up with these foolish mistakes?
@@cemsengul16 Money.
heatsink goop lmao.
So pleased you mentioned backing the screws out. This is something I aways do. It's very good practice to do this with plastic and wood so you don't cut new threads everytime you screw it together.
Oh wow! I know you mentioned only microns but putting that squaring block on it really visually shows the gap!😮It's huge!😮😮
7 degrees is not a laughing matter, especially if you go head to head with a competitor. Saving what? $5? Should not be worth it.
Right?! 7 degrees is pretty crazy good for a simple swap.
@@GamersNexus How'd you reckon combining this with the copper IHS might fair? Even better, or the same?
@@mitlanderson hope he answers lol
@@SouthBayLA1310 why? He will say they haven't tested it....obviously they would have showed and takes about it if they had
@@mitlanderson The ihs is soldered on these so risk of destroying the processor is probably not worth it for most.
My experience with the Thermal Grizzly Contact Frame... I was an early adaptor of the 12900K and have tried 240, 360, and multiple air cooling solutions to only ever come up short. Immediately upon launching Cinebench R23, the CPU package temp would spike to 100c and thermal throttle for the length of the test. The only solution was to turn down the max core clocks and limit the power draw. As a final attempt to tame the beast, I ordered the IceGiant cooler. In my communications with them, they said they had a new mount for the 1700 socket, so I patiently waited. It arrived in about a week for no charge. I started an email thread with their techs about my situation; they were EXTREMELY helpful. I'm not new to PCs, having built my first PC about 30 years ago, but I've never been an overclocker or someone that goes in and mucks with the BIOS. I expect things to just work. (as a note, I also have an AMD 5950x system that has always "just worked." Intel should take notes..) Anyway, the IceGiant folks were super helpful and recommended the TG contact frame, which they were gracious to sell and ship to me for less than I could get on eBay. I installed it last night (super easy, by the way.. if you are apprehensive about the installation, it's easy.. take a breath, take your time, and you'll be set). The beast is now tamed. I can run cinebench r23 with the power limits removed, and the CPU temps start at about 80c and by the end of the test are in the mid-90s, with maybe one or two instances of momentary thermal throttling. I'm running a cheap 360 aio as I wanted to see how the contact frame improved my current setup. I'll be installing the IceGiant cooler next. Power draw during the test is hitting about 241 sustained. Not only can I run the test, but I'm able to run all core at 5.1 and maintain good temps. My motherboard is the gigabyte aorus pro z690. The techs at IceGiant hit the nail on the head with their assessment that the stock ILM was causing my thermal throttling issues. I'm super happy now that my 12900K can run like it was meant to. So thank you to TG for creating a high-quality fix and thank you to IceGiant for all their help. And, Thank You Steve for putting in the time to show us the TG Frame and the generic option as well. If you have a 12th gen, do yourself a favor and just get a contact frame. Your CPU will thank you.
Cheers
Rick
Did they fix it with the 700 series motherboard? They seem pretty quick in discontinuing the older 600 series board and the ILM was pretty strong (way stronger than LGA1151). I have a B760 + 13700K with Thermalright PA120SE and in cinebench R23, I max out around 5.2 GHz with temps of around 85°C.
@@pixels_per_inch Are your temps without the contact frame in this video ? cheers
You certainly failed in summaries and essays...
Amazing coverage, illustrations, animations, and quality. Props to the entire GN Crew 👏
I'm really impressed by the quality of the last few videos, GN has taken a big leap forward in animations and content. Not that is was bad before, it's just much better now.
21:20 If you're talking about Thermalright, I've never seen any of their products have less than excellent build quality. They were really popular in 2001-2003 and have been under the radar lately, but they represent the best value in air cooling, particularly for SFF builds.
I still have thermalright true copper on my phenom 955, and axp-100 c65 full copper on my 5800x for my sffpc.
No, was talking about some one-off on Amazon.
SFF is exactly why I bought this actually. There's only so much you can do with a low profile cooler.
@@GamersNexus Thermalright one was 10$ on aliexpress with free Thermalright paste [so seller made money, he paid a cut to aliexpress and Thermalright made money, i can only imagine that the real cost of this plate is UNDER 1USD], its obvious that derbauer loves 2000% profit, asking 35EURO is a theft
I can confirm that my silver arrow SBE “extreme” is an absolute champ of a cooler.
I need to upgrade my cpu finally, but it’s kept my i5-6600k at 40-45degrees on a 4.3ghz OC for 6 years now.
I’m either gonna try for 4.5-5ghz or just sell it and not pop my vrm’s 😂
4:26 love the animation, the attention to detail of this channel is second to none
Amazing work once again
This is the type of content that I crave that only GN seems to be providing. I think you're leading the way for other mainstream PC media outlets to explore these smaller niche topics. Please keep it up, you guys are awesome.
Great product! Proper CPU pressure is very important, and even if the mechanism looks simple, A LOT of effort goes into making such a product. Anyone who has used early interposed BGA->LGA 1151 CPUs, should be able to tell how much of a headache it was to get proper pressure on some-- I'm not even talking thermals, but memory channels or even booting everything up. Many relied on the Intel cooler to do the job properly, but obviously that's not what was done here, which makes it even cooler. Nice job guys!
The cooler itself also told the tale of how well it was attached after it was too late as well (as in, you removed it and saw how well the paste spread out). Sometimes it would be perfect, sometimes it would be all on one corner and you can kinda guess that thermals were trash.
caramba,olha quem esta aqui
Thanks Steve
Mike did a really good job with that walk through. His attention to detail and intelligent approach is fantastic. Thanks guys!
Videos like this are why I love this channel. Such awesome info and always so thorough. Keep doing what you guys are doing we’re all here for it!
I was an early adapter of the LGA775 platform that I believe introduced this mechanism back in 2005. It got known pretty early that the spring load was causing that bend and I opted to remove it and only hold the CPU down with the waterblock. Worked great and helped a bunch on my temperatures. I honestly thought they solved this issue years ago.
I think they did with the 1100/1150 era ones, they just reintroduced it after swapping to a rectangle cpu instead of square.
@@WayStedYou that makes sense. I don't think my current 2011-3 socket has the problem either, so it's quite surprising by Intel to repeat an almost 20 year old design flaw.
@@LenniZ1337 That is what happens when bean counters get involved
As a Engineer I can say that one of the causes of the difference in pressure between the CPU and the socket is caused by the U point of the lever where it makes contact with the lip of flap that pushes the CPU down.
By using the lever as a spring and only having that spring on one side, as soon as that U shape of the lever spring makes contact with the lip of the CPU flap all the torque will mostly be applied on the side of the lip closest to the lever spring. In that case the right side as you said here 4:27. Due to its shape and attachment points as the spring elastic deforms it could alto try to rotate on itself and that will alleviate the force / lift the side of the U that is opposing to the lever spring.
A solution to that problem is tightening the that shape of that U making it less horizontally wide (orientation reference 5:29) or turning it into a I where it have only a single contact point on the center of the lip of the CPU flap. You could make a V cut on the lip of the CPU flap to help align the new I shape on the lever to the lip of the CPU flap. Changing it into a I could make it harder to manufacture.
Another solution would be to have a lever being used as a spring on both sides. In short 2 levers, one on each side of the socket. This can be accomplished very easily and cheaply by extending the original lever to the other side. In short you literally mirror the right side of the socket (orientation reference 5:29). They would add a ~10cm of spring wire as overall material for manufacturing the socket. And since there is 2 springs they have to reduce the spring tension or there will be too much torque on the socket.
Exactly what we were pointing out! Glad it made sense because it was somewhat difficult to verbalize and adequately show what we identified since everything is so small! Although we were thinking a wider contact would be better as it would distribute the load farther and reduce pinpointing the highest pressure to one spot, but interesting thoughts and can see how you idea would work also!
That was veru fun to read,and i can only immagine intel is not interested in this because potential cost in production,and they dont care for the end user.
I don't think the lever mechanism is the main problem here. The lacking cooler mounting pressure was always on the RAM side, which suggests the CPU IHS is bending down towards the RAM and VRM. That can only happen if pressure is applied to the IHS from two sides: the RAM side and the VRM side. That's exactly the case with the Intel ILM. It has two narrow tabs which are the only parts making contact with the IHS pushing it down, and those tabs are on the RAM and VRM sides of the CPU. They press down on the IHS while the CPU is being pushed up by the 1700 pins of the socket. See 2:40 and 4:21 in the video. Nothing else makes contact with the CPU. That's what bends the CPU and IHS, causing uneven mounting pressure for CPU coolers.
The lever mechanism is probably the reason why we see lacking mounting pressure always on the RAM side, but it's not the root cause of this issue. The main issue is that the IHS bends because it is being pushed down from two points on the opposite sides of the IHS instead of being pushed down evenly around the entire IHS or at least from all four sides. It basically comes down to cost as said on the video. It's cheap and easy to make a mechanism which only applies pressure to the IHS from two small areas which are opposite of each other. It would cost a lot more to make a mechanism which applies even pressure to all sides of the IHS. That's what the Thermal Grizzly Contact Frame does. It's shaped so that it pushes down on the entire lip of the IHS all around it.
TL;DR: The flap pushing down on the IHS makes contact with the IHS from only two points allowing the IHS to bend, the lever mechanism causes the IHS to be pressed down harder on the RAM side of the socket.
Great testing and I appreciate that you did extra testing with Intel ILM around 11:00!
So cool! Very thorough explanation and general information on this.
I would've never thought about how uneven pressure can cause thermal distribution issues. Thank you for putting this video together. Huge value here!
It's so interesting to see how much influence this ILM has even though the cooler already puts so much pressure on the CPU.
When derBauer was showing this a couple of months ago on his English channel he used an Allen/Hex wrench to tighten it down. The design of the Allen wrench with its 90-degree handle also gave a good visual indication of how much movement you are applying on the final tightening of the screw i.e. quarter of a turn.
Really loved how specific and easy to understand that demonstration of the process for the installation was. I'm eventually getting a completely new system and I'm planning for it to last for years, in spite of making it do some heavy lifting, so something like this to help prolong the life of the CPU as much as possible is definitely useful info.
The demonstratives were really well made. Great job!
I'm curious how AMD will approach this problem for their next gen CPUs.
Same! Can't wait to test AM5.
They have extra holes and it will be much better, it is already a known fact
Well considering AM5 is the same size as AM4, probably the same way they do now.
@@WayStedYou but am5 is lga so they’ll probably have to rework in the end
@@WayStedYou nope, it is a completely new system
As an engineer, I simply like to know that what I'm building is tight and as intended. So even if it isn't for extreme OC, I can see the value in this thing for people like me. I want something to do its job, if the Intel pressure thing basically fails there, I want something that doesn't. 35 Dollars is then a bit expensive for sure though.
We definitely need a lot more data on failures before we can really conclude anything other than a temp benefit. We can certainly speculate about lower failure rate, and there is definitely a place for the TG ILM in the market, but I wouldn't expect much change from intel without someone doing large scale testing and finding provable failure rate increase with the stock ILM vs the TG ILM while also testing the competing(AMD) ILM's for failure rate.
The 7c is certainly nice, but is this video enough to make intel engineers go "Shit better improve it because someone showed a 7c benefit with a $35 part". I'm not completely convinced. Bring up this conversation at an investors meeting and unless the new system provides that same 7c at say less than a 30% cost increase(per ILM), i doubt they will go for it, especially with an impending global recession.
@Brantyn a double lever ilm like Intel has used for a decade with their bigger sockets (like 2011, 2011-3, 2066) should be implemented, especially on high end mobos. I think it would make a compelling feature. Also making the socket more compact would help. Meteor lake (14th gen) will come with over 2000 pins so seeing what Intel uses will be interesting.
sorry to inform you will have to buy a new IHS and bracket or lap your existing IHS. The IHS products are atrocious. Every CPU I had made terrible contact. IF you want really fancy you will mirror finnish your cooler and IHS. wont need thermal paste.
@@prydzen You're high on cannabis my friend. So long as you believe there's any circumstance where you don't need thermal paste...
Anything that can increase OC potential, even for a basic level overclocker like me, is great news! 35 bucks and probably another 100mhz on my cpu with lower temps is a no brainer.
Cool you guys taught me something new. I’m not a stats person and didn’t know you can average averages in certain cases. You guys are incredible, introducing honest and meaningful data into reviews in a professional way. Excellent content as usual
I know its not the purpose of the video, but This video helped me resolve a memory issue I was having where the machine would not post with any speeds higher than 3000mhz. The play in the standard mounting hold down bracket was causing an issue where I had poor contact and the system wouldn't post. Thanks for the documentation!!!
Many users in Japan have also reported the 12 Gen Intel CPU's bending in the middle because of the 2 contacts in the middle.
Many Japanese Tech UA-camrs have been reporting this.
You can search UA-cam using the phrase "インテル12世代cpu 曲がる"
the middle bending issue was aready known and expected
thats crazy thank for the tip!
The thumbnails of some videos are already memes, omg.
Fantastic video guys. The quality of your content just keeps getting better and better! Thank you for not shying away from the more technical stuff!
Great work as always, but I really like the new animations. They are really well done and really help convey what you guys are presenting.
It's so awesome to see the new studio bringing dividends. This was a really high quality video guys- and on a super interesting product no less!
CPU contact is important, especially in LGA. Surprisenly they can sometimes still boot without some of the pins but you lose functionality like some of the RAM slots and this can lead people to believe it's a motherboard issue and they end up replacing it for no reason.
I had exactly this problem with a previous Intel-based computer. The memory was behaving very erratically, and it turned out to be an over-tightened cooler. It would boot and function normally for a while, but eventually there would be a random blue screen. It was maddening to troubleshoot.
Any evidence for booting without pins?
@@dfgdfg_ If you're lucky the missing pins can be either not in use or ground pins. Apart from that you can still have partial fucntionality like missing memory lanes like mentioned in the video.
I have my first two RAM slots nonfunctioning with an i5-12600k, asus-proart b660 mobo. With your comment in mind, if nothing changes after retightening or checking for pins, would it be better to send just the CPU back?
@@caison8482 It would be best if you could test with another CPU to make sure it's the issue but I guess people don't usually have extra CPUs just lying around, this was one of the perks in my previous job. Of course testing the CPU on another motherboard would also work but isn't any more likely to available.
CPUs are rarely defective, altho that can happen too of course. I have a CPU that won't let me use anything faster than NVidia GT430 (maybe issue with some PCIe lanes). Fortunately I don't actually need a GPU on that machine.
This is really cool and deeply informative information, I admit I've never really thought about these sorts of things and it was really shocking to see the level of difference between the Thermal Grizzly frame and Intel's. Your content is really unique and I learn a lot from it, so thank you!
i've always felt like the ILM mechanism was putting vastly uneven pressure on the IHS. There is no doubt that Gamers Nexus keeps all hardware manufacturers up at night, due to your lengthy and in-depth analyses of literally everything inside a computer. I think your conclusions and opinions are the buoys by which both manufacturers and consumers guide their ships. 🖤
The quality of this video and all your testing is incredible!
The quality of the journalism of the last couple of videos and this one is just absolutely outstanding. Loved the 3d animations, there isn't a better way to demonstrate how things work than a good 3d model. Keep up the good work!
Thermal Grizzly: “fine, I’ll do it myself” lol
YES! I'm glad someone finally addressed this, this is how my last cpu socket broke. I was cleaning out my computer and ended up having to get a new mobo+cpu
I had an old Z68 board go bad after 6 months because of something like this. Everything worked fine and then I packed up the computer in it's original packaging to move. After I had time to set it up again, I lost a RAM channel. ASrock refused to help me...
I am impressed how you keep improving production quality. Mike is also a great addition to the team. Overall great video and the product itself is worth a look to improve temps.
This video made me aware of a category of PC parts that I was never really thinking of, but now am incredibly intrigued in. I probably won't get the Contact Frame, at least not for now, but the ILM is now a part I'd like to consider moving forward with future builds. If this type of coverage continues, at least as long as it is necessary and when needed, then I'd be totally on board for it.
The testing you guys do is amazing. Instead of just giving your opinion you give actual data that none of us could ever get at home.
Eagerly awaiting your other brands bend correcting frame test video!
Steve, thank you and the rest of the team for doing what you do. If it weren't for you guys digging down and scientifically testing this stuff we the consumers wouldn't have any means to be aware of issues like this.
I am fascinated how companies that are specialized in manufacturing very specific narrow range of products for long periods of time and many iterations and revisions of the same thing still manage to introduce new problems into their products.
gotta say this is one of your highest quality videos. and shout out to mike for what I think is his on camera debut in a video like this. excellent work all around.
After trying to research this issue for the last few weeks, this is exactly the video I needed! Thanks GN!
That thing kinda reminds me of the old Athlon XP DIE protector shims from the Socket A days. For those that don't know, back then AMD Cpus didn't have a heatspreader, so the CPU DIE Silicon was directly exposed and could easily be damaged or destroyed if you weren't careful when mounting your cooler. So there was a quite substantial addon market that sold stuff like Copper DIE protectors/spacers that helped with this issue.
Wasn't this normal decades ago? On both sides?
@@slimal1 There were a few Intel CPUs that had exposed dies, like some Socket 370 chips, but they pretty quickly moved to using an IHS. AMD kept using exposed dies for much longer.
Years before that, before they moved to "flip chip" packaging, the actual CPU die faced down towards the socket, not up towards the cooler.
@@RobertHancock1 interesting. Thanks for sharing that
Not only that, but the stock cooler had that leaf sprung mechanism that you levered over the edge with a flat bladed screwdriver. Very disconcerting.
@@ic_trab I remember that! I think mine was a Coppermine Celeron, Socket 370.
Overclocked to 850MHz
LOVING this type of content. This and the last GPU power video really hit it out of the park for me. This is the sort of thing that gets forgotten in the mundane building of computers, but it's the kind of details that we need to be paying attention to and that are actually important that could make or break a build. They become even more important if we are building machines for others, so that we can reduce the friends and family tech support tax LOL.
Outstanding presentation....well done all around!
I just did the mod with my D15 and the difference is huge! The variance is similar to yours, 5-7C cooler on the exact same settings and test run (simple 3xCinebench run with ambient at 25C).
The best computer hardware channel on UA-cam. Methodology in testing is sooo important and not to knock down other channels but you guys understand that the most. And not only that, but you guys go a step further and test stuff that noone else even realizes is a thing. Keep leading.
Based on the threadripper mount, I doubt AM5 will have issues (read: "AMD is aware of the possibility and will take steps to avoid it"). But it could, so I'll be interested to see how that pans out.
If they use the TR casset system it's probably worth that small premium just so everything works right.
Sure hitting those cheap boards is nice but if you're building a system on a shoe string budget, maybe it's best to wait so you can get a better board.
LGA 1718 (Socket AM5) mechanism looks exactly same as LGA 1700 (Socket V). It's mainstream, so cannot be expensive.
Check out the Thermalright AM5 secure frame!
This is a new level of empirical content from you folks. This is amazing!
Love this kind of content. This installation method was very easy to understand and mine installed easily first time.
Two rando thoughts: 1. Intel isn't just being cheap. (I mean they are, but...) Intel is also doing their ILM quick and dirty because of system builders. The faster and less crisis-prone a system can be built, the more money Dell and such save building systems. Even at the cost of crap thermal performance. (I mean, we've all seen Dell builds, right?) Fiddling with degrees turned or torque wrenches, or even two pull handles instead of one, all costs them money. And let's face it, Intel sells A LOT more CPUs to Dell than it does to you or me. 2. If I personally were building a new Intel PC with their current ILM, I would DEFINITELY be adding a Thermal Grizzly CPU Contact Frame to my build parts order. No question. No doubt.
Man, the complete lack of any torque tools in most factories does terrify me
I think i ordered close to $135k in air/pneum/hand torque screwdrivers once. They initially only gave me 15k for it but made it back in maybe 3 weeks from the decrease in bad parts and increased throughput.
@@angrydragonslayer thermal grizzly has a video that shows the perfect way to tighten the screws by hand
Hey Steve I'd love to see you go back in time and do these pressure tests on the very very old lga and pga sockets showing how things have changed over time
this video just demonstrated that its gotten worse over time
@@scalz420 we can’t say that without knowing how things were before
Loved the portion with Mike too in addition to the new part category and analysis
Awesome video. Thank you for your in depth coverage once again!
a return to the lga 2011 mounting mechanism would be cool
The fact you now have to buy a 3rd party socket to fix thermal issues that clearly Intel most likely knew about but couldn't be bothered to fix before release.
Let's see if AMD don't have anything like this with AM5 as I assume it would use a simular mechanism. Great video GN really good investigation
I might be wrong, but given that AM5 may be a smaller socket in the first place and those weird IHS edges, we saw on press releases, they may have less issues with uneven pressure to begin with
I wonder if this is intel settings themselves up for a 4770 > 4790 situation, where they make something intensionally mediocre so they can show a massive improvement next generation (much like apple with the last few intel macs)
According to IgorsLab who made a video (in German) about the AM5 socket, it's going to have a better solution that shouldn't mirror any of these concerns.
If you run into thermal issues , there's 5 other factors that are causing the problems before the socket. Sure , i'd like to see a better socket too , but the average user who is buying 12600/12700K's and runs in to thermal issues - should take another look at their cooler or PC case long before they need to worry about the socket.
It's annoying for people like me who are into min-maxing temps though , this video made me consider spending the $35 on the thermalgrizzly socket while my temps are completely fine.
And if i was really in to overclocking it'd be an even bigger annoyance.
Bought a 12600k at launch. Just had a look yesterday and fitted a Thermalright “fix” out of curiosity. Couple of things worth mentioning. Firstly the cpu itself shows zero signs of been bent. Checked on flat surface and against a level. Secondly, the before and after temperatures are identical. No difference what-so-ever.
This is not a universal problem. If you have the 12th gen system and it works fine, you do NOT need to buy third party mounts.
This is why I buy GN's merch. Actually interesting testing, not just benchmarks. Awesome video!
It will be very interesting to see if these same sorts of issues happen with AM5 when that gets released to the public. Keep up the great testing, amazing information!
Intel must be loving this free research and development/product testing. Great work on this one, pretty neat that each cpu has its own finger print so to speak.
Intel did all the research, knew all this and determined that the customer dis-satisfaction rate would not be high enough to warrant a better performing, more costly system.
I think the dual lever design of the X99 was very good, I was surprised not to see it on the 1700 socket, considering it's size.
Especially given the fact that their pin count is not that far off. Intel could've just made it a similar socket and it would've been a much better design. I wonder how CPUs or motherboards will look like after like 6 years or so with a cooler not originally intended for 1700 (smaller contact area), but adapted to it. Those pins develop quite a good amount of force if you think about it tho..
installing a contact frame now for a 13th gen cpu, I appreciate the installation instructions here - since I don't have a precision torque screwdriver laying around
Did it work for the 13th gen too?
OMG, I'd ordered everything for my first Intel build when I saw an ad for this over on Hardware Unboxing. I immediately came here. Thank FSM, I found out about this before I started putting it together. Have one of these on the way now.
Are motherboard manufacturers required to use Intel's design entirely? If not, it would be interesting to see something like this implemented on the more enthusiast type boards. Or better yet a proper latching mechanism that existed on other platforms for improving the contact while maintaining ease of installation.
I would think including that on the overclocking enthusiast targeted boards by Asus/MSI/Gigabyte would be a good idea/selling point. Just not sure if Intel would allow them to do something like that, especially since they have to take into account voiding warranties and such.
They'd be _allowed_ to do whatever they want, but this would most likely prevent them from being able to put the Intel label on their packaging or claim to officially support Intel processors because it isn't tested and approved by Intel's engineers. So Intel isn't forcing motherboard manufacturers to do anything, but the market would do that because not "officially" supporting Intel chips on a motherboard designed for Intel chips is going to scare consumers away from buying the product since there would be no guarantee everything would work as expected.
The motherboard manufacturers would have to compensate for that by doing their own certification and staking their own brand reputation on the reliability and footing the bill for any warranty-related issues. That's because Intel's warranty would be void by not going with what they've certified as compatible with their product, so either consumers would have to settle for risking burning up a $600+ CPU for a couple degrees cooler operation, or would have to pay the motherboard manufacturer for footing the bill to backstop the CPU warranty that Intel would no longer provide on their own dime.
These added expenses will keep motherboard manufacturers from doing this, and even if they did, would keep a good percentage of consumers from ever buying those more expensive products in the first place. And Intel won't do it because the demand for that additional manufacturing cost just isn't there. The only place they're going to spend that kind of money making this sort of improvement is on enterprise-grade hardware where the price point is substantially higher, and that's why they already do it in that market segment.
The best solution for both Intel and motherboard manufacturers, as well as for consumers, is to do it just like this. Anything else would necessarily cost more both to the companies doing it, as well as the consumers using it. The lower cost of manufacturing it at scale would be more than offset by the additional cost to the motherboard manufacturer and the end consumer footing the bill for filling in those missing warranties.
@@keithd.2722 That's why they would just send it in the box with the motherboard for the consumer to install once they receive their board. Easy work around
Nice explanation Mike, also appreciate the angelic lighting top down - hope to see this as a mainstay for all future appearances.
This plus your new GPU power draw video earned my sub. This channel is way more professional than even my university undergrad lab experience.
Great analysis, nice work!
I’m currently running this with my 12900K. With an undervolt applied to stock frequencies (zero clock/performance degradation), I got my all core stress test thermals to hit steady state at around 80C average with spikes to 84C fairly consistently when rocking the stock Intel ILM on a 240mm AIO vs. 97C at bone stock. With the contact frame installed with the same settings and test conditions, the highest recorded temperature/average was 77C for the duration of the test. I’m happy with this result for the money spent. It’s just too bad Intel’s engineers can’t be bothered to design their ILM better, and their voltage profile to not be completely balls to the wall.
How much of an under volt did you apply?
If you're in US... Where the heck did you buy the frame from? I can't find it anywhere?!
Holy balls. Love the data and new testing tools, though!
The Stream Professor!
If I were a design engineer for these, I would make a socket where the mounting is pretty much like the contact frame, but instead of screws, the four holes would be replaced by a kind of fixed metal nut of sorts, one end closed, and the surface open. The inside of the nut would not be threaded, instead smooth, but with a pair of L-shaped canals carved inside on opposing surfaces (like one on the inside left and one on the inside right). The contact frame would then be fixed onto the nut by non-threaded bolts that have a special "tooth" protruding on both sides made to pass into the L-shaped canals.
The mounting process is simple:
1. Insert the CPU onto the socket.
2. Place the contact frame over the CPU.
3. Insert the non-threaded bolts into the holes and into the respective nut. They will only go in while the "tooth" on both sides match the canals so there's no making a mistake there.
4. While keeping the contact frame in place with one hand, push in each non-threaded bolts and turn them until the "tooth" on both sides of the bolt locks into place of the L-shaped canals, in an alternate manner similar to when mounting a CPU cooler.
Since the bolts can only reach the bottom of the L-shaped canals via the special "tooth" being in the right position, there is no danger of too much torque or too little torque. As long as the board manufacturers build the sockets properly, the nut should keep everything even, and all in the same equal or leveled position.
I'm glad you aren't the engineer of these
Good luck machining bolts with tight enough tolerances that even with a single canal/thread, every piece is 100% identical and provides consistent mounting pressure.
I'm really glad for the detail you all went into for this video which is extremely important and helpful with something like this. I am curious if the cooler you used was new out of the box or one that had been used for a while as I know there are some odd concave/convex variance with cold plates and this can change over time/platform.
I did read that the socket 2011 ILM costs something like 3-5x more than 115x so I understand across the board why they did it, but it sure seems like they could do a higher quality one as an option. Perhaps it's too big of a hassle for motherboard manufacturers assembly lines, but it just doesn't seem like it would be given the broad range of boards already available.
I wonder how the addition of a cooler on top of the original socket affects contact... presumably if overtightened on the ram side it may compensate somewhat?
The how-to segment was extremely well done. The explanations for "why" each step actually made the execution clearer, a very rare occurrence. Too many youtubers go off on a tangent halfway through instructions. That was well laid out, thought out, and presented. If you screw it up after watching this video your shoes probably have Velcro instead of laces, through necessity. 5 of 5, highly recommended.
Thank you guys for this epic in depth content! I was very interested in this when Derbauer announced it. Without a doubt, this willl be a day one buy for me, and I will employ the technique that Mike used for installation.
I paid $45 for this contact frame last weekend. The installation was easy and everything appears to be working rock solid. My 12900K appears to be working flawlessly.
Always disliked those spring-loaded mechanisms and wondered why we don't secure the CPU in the same way as its cooler.
So glad someone with enough power in the industry stepped forward with an obviously more practical solution.
Here's hoping it becomes the standard.
Well, GN already elaborated it. The installation method is so damn finicky you probably will send everyone to "just get a pre-built". I mean, "screw it tight, but not too tight, and slowly but not too slow...." If you screw Noctua's too tight, its ok, but screwing CPU too tight can cut your memory by half or worse. I built my own PC, but I will not dare to do it if I'm told to do the ThermalGrizzly way
As pointed out, on the older HEDT, workstation and server platforms with the double levers, it is not an issue, neither is it on smaller CPUs with the one lever. Intel have just hit that area where the CPUs in the desktop platform are almost as large as the HEDT platform at that is why it becomes an issue. Having it automatically apply the right tension for the pins underneath is real important for LGA sockets and doing that without a precision torque driver is hard.
@@ArchusKanzaki It's a risk anyway - either the risk of overtightening or the risk of the chip degrading unevenly. I choose to rely on myself, whether it be to position the cooler nigh perfectly or to screw things up ;)
Intel's solution is very good from the end user's perspective, because it leaves little room for screw ups. There's no risk of the mechanism falling on the pins and there is no way to get the torque wrong. They turned a task that could require a lot of attention and a specialized tool into a trivial and straightforward step.
When you overtighten a heatsink, it's only going against its own bracket. You're not adding pressure on the socket because the height is set by the bracket. Imagine if you had to manually set the correct height on a heatsink. That's kinda what this is.
The method for tightening showed in this video is very finicky and when done by millions of people, there would be a lot of those who get it wrong. Even if you provide the correct torque value, most people who own a torque wrench or screwdriver never actually had it calibrated and while it can be good enough for bicycle parts, you need to be very precise with a CPU.
Incorrect pressure can lead to all kinds of problems, some of which actually aren't immediately obvious. Imagine if part of troubleshooting every obscure problem was removing the heatsink and checking the torque of your CPU socket. They could provide a purpose built torque bit holder, but those are relatively expensive and would substantially increase the cost of the product.
@@CanIHasThisName How many torque drivers can be even set that low? 0.3-0.6NM requires precision ones at best I would think.
I appreciate you guys doing all of this for the consumers. On the end of intel they need to get this right, because on the other hand consumers shouldn't have to do this.
I went into this video without a huge amount of interest in the topic. How much of a difference could it actually make? I've got to say, I'm only 10 minutes in but blown away by the results. Excellent work by Der8auer and by the GN team. These kinds of investigations push the tech environment to ever greater heights
Thanks Steve. long time fan from England, Britain!
@8:55 Just some feedback, it would be helpful to use 'freeze top row' in excel to freeze the column headings. Very helpful info, Intel need to seriously re-work the ILM for 13th Gen and in regards Thermal Grizzly's contact frame 8 to 10 degrees is quite abit especially for warmer climates and since the Intel's ILM is distorting the shape of CPUs over time. The contact frame is well worth the purchase as it's much cheaper than buying another CPU imo.
Thanks to Mike for the little tutorial on mounting it. Alot of good points!
It would be interesting to see a hybrid between this smaller socket and the double-lever mounting mechanism found on their old X99 and X299 chips. Obviously when you're manufacturing millions of these, pennies of added cost matters, but it'd be a nice feature on the higher-end boards where overclocking is expected.
I've never considered myself an enthusiast, I never try to see how far I can push any part of my pc. I take a lot of time testing, daily driving, and monitoring before I start to make adjustments. I'll take a 5% performance boost that's completely stable over a 25% boost thats 85% stable every time, I also would never allow anyone else to build or repairing my pc. Your thorough and in depth coverage is extremely helpful to me and also entertaining. I truly want to thank you and your team for the amazing content!
Not sure why anyone has not mentioned the flex on the back of the CPU socket. The brackets supplied on LGA1700 boards are inadequate for the pressure applied.
Worked well for a square socket, but Intel have used the same design for a rectangle socket now and the leverage is too much for that thickness of metal.
With a thicker bracket that has no flex you can tighten these frames down hard/level with the motherboard and still have all memory channels.
Way better temps too.
The thread method does not apply equal torque either, just equal thread distance, if one screw is a touch longer than the other your frame will be unlevel.
Also the Thermalrite brackets wobble on a glass surface slightly, not the best aluminum used id say.
so is the thermal grizly contact frame good or not?
what do you want to say with your comment regarding to the contact frame presented in this video?
@@johnmachter40 I went and purchased the 7mm heavy duty watercool bracket and used it with the thermalrite bracket.
The rear one is so strong that the CPU now is convex! And so is the thermalrite bracket.
@@syntaxzedex4316 you mean watercool backplate? huh, i just wanted to order one.
so you would suggest to get only the thermalright contact frame without the watercool backplate ?
@@johnmachter40 I have not bench tested that system yet waiting on a psu.
But I do like the rigidity of the heavy duty black 7mm bracket. It's the only one that ties both cpu and cooler together.
@@syntaxzedex4316 whats the name of the watercool 7mm bracket?
I'm surprised you wouldn't be interested in doing a comparison of the more budget friendly and actually available frames. I know 2 dozen people that use the Thermalright frame, still haven't seen anyone have the thermal griz yet.
I'm guessing because thermal grizzly has the international patent for this retention mechanism, and the thermalright clones are unlicensed clones
We never said we "wouldn't be interested" in doing that. We just haven't done it yet. That doesn't mean we aren't interested.
Correlation doesn't equal causation
@@insu_na Are you sure they even have a patent? Nothing about this strikes me as non-obvious. At any rate, Thermalright was first to market. There are allegations that they misappropriated a Chinese designer's work, but nothing about TG.
@@concinnus Roman at least claimed that they had a patent. Didn't actually look for a filing.
Very interesting, great video as always guys :)
Great video guys! Very interesting product.