Ryzen 7000 Offset Mount - Will this improve the high Thermals? (+B650E Launch)
Вставка
- Опубліковано 15 лип 2024
- Support me on Patreon:
/ der8auer
---------------------------------------------------------
Save 10% on your iFixit purchase: DER8AUER10
eustore.ifixit.com/der8auer
Pro Tech Toolkit: bit.ly/2JOFD8f
---------------------------------------------------------
Find my products at Caseking:
Delid Die Mate 2: bit.ly/2Rhv4y7
Delid Die Mate X: bit.ly/2EYLwwG
Skylake-X Direct Die Frame: bit.ly/2GW6yyC
9th Gen OC Frame: bit.ly/2UVSubi
Debug-LED: bit.ly/2QVUEt0
Pretested CPUs: bit.ly/2Aqhf6y
---------------------------------------------------------
My Equipment:
USB- Microscope*: amzn.to/2Vi4dky
My Camera*: amzn.to/2BN4h2O
(*Affiliate Links: If you buy something, I will get a part of the profit)
---------------------------------------------------------
Music / Credits:
Outro:
Dylan Sitts feat. HDBeenDope - For The Record (Dylan Sitts Remix)
---------------------------------------------------------
Paid content in this video:
- /
Samples used in this video:
- B650E Aorus Master, Ryzen 7950X, Corsair 6000C30 EXPO RAM
---------------------------------------------------------
Timestamps:
0:00 Intro
0:50 B650E Chipset & Offset Mounting
2:06 Gigabyte B650E Aorus Master
2:45 CPU Guard & Offset Bracket
3:54 BIOS-Update, EXPO & Cinebench Test Run
5:30 Good VRMs on B650E
6:08 Offset Mounting Bracket
7:45 Conclusion
10:06 Outro - Наука та технологія
Roman, unless I’m missing something it looked like the score actually went up by 9% from the first run to the second. If that’s the case, then I think it’s fair to conclude that the offset cooling worked as intended, but instead of reducing temperature it prevented the CPU from thermally throttling. Is that right?
thats what i see too. 1300 points makes perfect sense. unlike direct die it doesnt drop it far enough to get away from the throttling
Saw the same comment under my German video. Honestly I'm not quite sure how I missed it. I will look into this again :) Thanks a lot for paying so much attention to detail!
@@der8auer-en And yet again we see why he is the best.
Thanking people for finding possible errors in his videos.👍
I was going to comment this but it seems others saw it as well.
@@der8auer-en Akuter Koffeinmangel, ganz klar.
I clearly see the effect, 8% score improvement (14521 vs 15843). AMD has already said that it will boost as high and as long as cooler can handle temperatures, so if we think we will understand that offset WILL NOT improve temperatures but will allow to boost higher and we clearly see it in cinebench score. I'm surprised that Roman didn't notice this, power draw was also hidden from us while editing/shooting but it might change during offset mount test run.
maybe he doesnt want to go through with rnd of those little offset things again (^:
I did a manual OC on the CPU so I did/do not expect the CPU to act different with improved thermals. I honestly missed the improvement but it's awesome that you guys noticed it so I can just go back and repeat the testing! Maybe it will help in the end :)
@@der8auer-en This is why I love the OC community, I can't wait to see what you find out!
@@der8auer-en Oh! I didn't notice OC, my apologies if this info was in the video and I missed it. I assumed that offset was the only variable in this test, which is expected.
@@der8auer-en you should pin the comment to the top or make a comment of your own and pin it, so people don't miss it.
Excllent presentation as always. We love your nice cats!!!
Looks like it worked to me, that score improvement while maintaining very similar temps is a win in my books. That being said, I'm still spoiled by your delidding results.
Lovely cat, keep up the brilliant investigations you do on the hardware.
nice very helpful upon my recent purchase I wondered about this kind of thing given my current water block was designed around the last CCD layout. It cools exceptionally but of course any gains to be had I would want lol
being subbed to this guy gives me so much info man
this is exactly what I was looking for and der8auer read my mind
I quite like the main heatsink for the SSD, it actually has something resembling fins for a change
I'm subscribing to see more of that beautiful beasts (the cats!). Don't disappoint me.
I LOVE your channel
When I installed my Heatkiller IV block I made sure the intake jet was over the dies as it can be mounted "goofy" like in skating. Now I get to see if it makes sense again.
I was paranoid about my 7950x install because of all these temp videos, but I just installed mine. Streamed and gamed at 55-60 degrees with an AIO. Running expo as well.
You're videos always seem to be right on point regarding our questions. I am very much interested in the B series Mobos as a choice for my new system. With the Exception of the ASROCK B650 Taichi, which has a VRM config of 24+2+1 most all the rest of the B series from the major Manu's are in the Mid to Low teens with VRMs although some of them have a very high AMP capacity. According to the VRM overview given by Hardcore Overclocking a few days ago his results after testing reveals the Capacity of the B series VRMs to be more than enough for most any Overclocking scenario.
Would it be feasible for you to engineer a replacement IHS for the 7000 series? You might be surprised at the interest generated if the results are a substantial improvement. I mean this wouldnt be for the Meta Budget crowd, but you might find a substantial interest. Who knows maybe AMD could arrange something for you like Lian Li did. Well, just a ramble. Thank You for your On Point Content. I knew I was going to like this video just by the smile on your face at the beginning of the video. 😁
I'd love to see if the other SKUs will also get that ~9% uplift in performance. Nearly 1300pts is a pretty big jump and I'd love to see this, plus lapping and see what a 7700x would put out.
You can mount a offset cooler but the IHS still only can spread the heat x mm in a certain direction. I think for 8000 they should try and swap IO DIE and chiplet location.
Would there be any benefit in sanding down the IHS by say 0.5 - 1.0mm? Would it still leave enough contact for the cooler without interfering with the CPU Bracket?
We made a lapping tool for this purpose which I'm going to send to Jay this week :)
@@der8auer-en Perfect! Thanks for the reply, that will be a very interesting video to look forward too.
@@der8auer-en I was wondering this as well, I am very curious what the impact will be on the temperatures
AMD still claims the IHS is not very thick, we will see soon how wrong they are on the IHS engineering design.
@@der8auer-en Awesome! I suggested this on the original delidding video. I think it should be close to half of the delid option, but much safer for the CPU.
[06:14] Interesting use of the Makita screw holder. 😸
the arctic liquid freeze II also has an offset mount for the hotspot idk how it would compare to yours tho.. looks super fancy
I'm early! Looking forward to the B650E series, everything is so expensive to the point of imbalance. I really want a board that does 16 lanes pcie5 AND 8 lanes (in x16 slot) pcie 4 (or 5, but that's pretty unlikely given that there are only 24 usable pcie5 lanes, and most would assign at least 4 to m.2)
What do you need PCI-E 5 x16 for? GPUs won't need that for >2% improvement in the next decade. You need it in an M2 slot more likely, since drives are already in production that use the bandwidth.
I would want a board that can take 8 lanes away from the GPU slot and give it to M2. PCI-E 4 x16 speeds for the GPU are plenty for the next few GPU generations. it will be similar to what happened with PCI-E 3 and the initial GPUs that supported it. They worked fine with 8 lanes for a long time. DirectStorage speed might suffer a bit perhaps, but it will still be fast.
The ideal would be 12 lanes for the GPU and 4 to another PCI-E 5 M2. But that is not possible.
@@marsovac In correct - You forgot the fact that DirectStorage API is now here on the PC platform.
That will give the GPU direct PCIe 5 M.2 speed access, as PCIe 5 M.2 SDDs are coming with 15GB/s read/writes.
For instance this will allow UE5 to stream in super large textures with no hitching!
And yes games that are 800GB to 1TB are around the corner, people just under estimate where PC gaming is going.
@@marsovac RTX 2080 TI saw up 15% loss in performance when using PCIE 2.0. The RTX 4090 is about 150% faster so there is a good chance that PCIE 4.0 is getting close to its limit.
@@marsovac I don't need a new computer in the first place - wanting and needing don't have to be the same thing. 😅 I agree that not many people will *need* pcie5 x16.
edit: although the mention of direct storage in the replies above may be a good point? 🤔
@@N0N0111 You misunderstand how DirectStorage works. It doesn't actually implement NTFS, so the data still has to pass through the processor before the GPU can use it, all it does is allow the CPU to store the data directly into the GPU instead of reading it into RAM first. No M.2 will be able to run at PCIe5 speeds for longer than a few seconds anyway, because the heat output would quickly become unmanageable. You'd have to fundamentally alter how NVME SSDs are manufactured, because thermal pads and PLCCs just aren't conductive enough to allow the kind of cooling you'd need. Instead of simple blocks of aluminium thermal glued to the chip you'll need fins, fans, springs, and thermal paste, and the controllers will need either an aluminium IHS or will become bare dies. It's not going to be a quick change. Even if a PCIe5 SSD could run at 100% speed, PCIe5x4 is the same bandwidth as PCIe4x8, so a PCIe5 SSD at full speed would still only be just enough data to saturate half the PCIe4 GPU's bandwidth.
Unless you're striping your storage between two or more PCIe5 SSDs we're a long way from needing more than PCIe4x8 on GPUs.
Thanks for the video, interesting to see the results of further testing.
Also, this B650E is showing as pre-order £440 in a few places in UK e-tailers, so I don't think the value proposition is going to get better anytime soon. Anything other than sole productivity needs, 5800x3d on B550, or perhaps team blue 12/13 gen with alder lake platform for better value?
ouch. Did not expect that :( That is just way too much
It's not called rip-off Britain for nothing
In the Netherlands at MeGekko you can get the ASUS Prime X670-P for 309 euro's, excl. shipping.
well aorus master is a top of the line model,they still put a lot of extra features even with the "lower" chipset.
back on z390 aorus pro was a lot more affordable.
It's almost like AMD dont want to sell very much of this gen at all. 440 quid for a B650 board? I'll be polite. The second word is off.
Love the cat grabbing at the screw, tech and cats, what's not to like 🐈🖥
Boy Cat,,, Left Pawed. 😄
I still want to get one of these just to lap the ihs and take alot of material off to see if it improves temps.
Can't wait for raptor lake! That's what I'm waiting for
Also enjoy seeing the Aorus products at work.
Love your videos, research, and products Roman. Only using Thermal Grizzly paste right now, but definitely a brand I look for and trust.
Thanks for your support!
@@der8auer-en I would love for your take on high end graphite thermal pads for use on the 7000 series. I have the Thermal Grizzy Kryonaut extreme paste but I'm thinking a very high end pad would be better especially with the 420 Artic Freezer II.
If you use the offset mount and don't file the AIO mount to allow the stock M.2 heatsink, you should use double sided thermal tape and a 20mm x 20mm heatsink for the PCIe 4.0 SSD controller.
would you try to see what lapping several millimetres off the IHS does?
go for it man, you have my permission.
Here is a thought:
Given the Zen4's thick IHS, can it be replaced with a vapor chamber and if so, would it improve cooling vs a direct-die cooling solution?
Obviously it would depend on type of cooling solution (ie air or water) but given that with an air cooler only 1 or two pipes actually touch the CCD, it might improve.
Could you try how the Arctic Liquid Freezer II 420 (specifically the 420) will do with the new 7000-series? Since this cooler also has a possibility for making it an offset mount.
that or the IceGiant cooler he has, wonder if that can be configured for AM5
is the offset in the same direction?
@@aphysically6071 I believe it is, I have it configured as such for my 4750G and I did the same for my old 5900x
I've done the test with my Arctic Liquid Freezer II 420 as you request here. Using the offset mount does improve temperature and/or boost blocks. Should be used when cooler provide this feature! PS: Results posted on Reddit
@@Luca-pb8vt Where? 😅
Please test the thickness reduction of the ihs by means of aggressive lapping to see what the improvements are
These cpus are meant to push to max temp. If cooler is better, cpu will just push more watts, which means higher clock. (according to Gamersnexus at least) So just by looking at the watts and clock speed, should be able to see if cooler is better. Temps gonna be max no matter what.
6:12 "yes, it smells apprrroved"
Roman JazeTwoCents asked a really good question regarding your last video with direct if you have a piece of metal between the die and the cooler that have the same thiknes as a normal ihs will it still go so hige in the heat is The Heat in the new Ryzen's only so high because of the new thik ihs shild ?
Der Bauer,, would it be possible for you to config your CNC to shave off 50% of the IHS table to achieve better thermal conductivity? Ok, maybe the masses with AM4 cooling couldnt just slap their solutions on but Im certain there could be an easy solution for that as well.
Interesting data
Seems this might offer some benefit? I'm excited to hear more!
Something else I'm curious about: How much of Ryzen's thermals are due to the increased IHS thickness? Direct die showed great improvement. I wonder if a thinner IHS would offer some benefit. Either a custom tooled replacement, or shaving a couple mm off with a cnc to test/evaluate
Ryzen 7000 targes a 95C thermal limit before a power limit, so you cannot lower temps unless you dissipate more than the raw power it is consuming off the wall.
the offset might be a good idea for a different CPU or direct to die, this new one have less contact surface
Offset mount helped. Look at the cinabench score difference, like 1k difference. It boosted the cpu more as it had more temperature head room. How did you miss it?
I guess he didn't use his notepad and completely missed it haha.
I also want to ask the video editor how did she miss it lol.
So about the IHS thickness, is it possible to lap it down half way, and adjust the mounting mechanism to allow a closer mount? I don't mind that kinda change as I don't risk breaking the die. I would be fine doing direct die as I used to forever ago before IHS but CPUs cost a lot more and might be more fragile, idk.
@der8auer when are you releasing your ZEN 4 delidding tool and where can i buy it?
I'm just wondering ... since the IHS is so thick that it causes problems can you just sand 1mm of the material off to improve heat transfer?
I wonder how much better the thermals would be if the dies were offset so that the i/o die was up against the edge of the IHS and the CCDs were closer to the middle. Currently it seems to be the other way round.
Non-pro tip - you can ctrl + left-click to highlight multiple items in HWiNFO (such the two die temps), same as with a text document (only in the same column though).
Hi der8auer!
Is there any ETA on when the cpu guard for zen 4 going to be available in the US?
Thank You!
6:13 your white kitty is just purrfect
bro where chalk up u been, it is so cool
Are you going to do a year long stress test for the new Ryzen chips? just to see if the 95C operating temp is going to somehow degrade the silicon or not.
I think someone should try that, doing Cinebench loop for a whole year while the CPU is pegged at 100% load.
Have you tried shaving 2mm off of that extra THICC ihs (a way safer option compared to removing it altogether)? If loading mechanism becomes an obstacle - swap it with holding bracket like you did with de-lidding...
If there is a case of using back-plated cooling solutions, would there be a chance of cutting the AMD backplate yo have the "AM4" holes available for mounting?
I've looked at cad models of am5 with cpu installed and it seems like ihs sticks out a bit, would grinding it down a bit be a less risky alternative to direct die? or would that be negligent?
How are the PCIe lanes on this motherboard? I got a B550 for my 5900x and now I am a little regretting it because I have an additional GPU for encoding and the other 16x slot is actually a PCIe 3 x4
Get x670e for that
Is it possible to use a combination of ECO Mode and Undervolting in order to build a small form factor 7600X build?
If the chiplets are like this aren't the heat pipes of the cooler in horizontal position less effective than in vertical?
What about aggressive lapping to thin out the lid? How to get coolers to mount properly if we pull a mm out of the lid? No pins on the cpu so lapping is less risky.
You could probably modify the mounting hardware of the cooler to make up for that 1mm of difference. Some shorter pins or clip arms should do the trick.
TBH, i would buy something like this offset bracket for 'server' applications if you can reduce the CPU temperatures by just a few degrees by getting better heat transfer, the fans can run slower, meaning less wear on the bearings.
On the subject of boards having everything you could want
I'm trying to build a server right now with a 7950x, but i find the X670 boards lacking for home-lab use.
I cant wait for a board with
IPMI(i've got a few X470 boards with IPMI, and dedicated management LAN, its awesome)
integrated 10G+
3-4 PCIE 1x slots (I'll be using these for tuners)
2PCIe 16X slots one in 8x or at least 4x (I'll be using this for a SAS HBA and
ECC uDIMM support(hopefully 256GB once 64GB DDR5 u?DIMMs becomes available)
6 M.2 slots, with 4 of them being 2x (2 for boot, and 4*2x for optane or any drive you dont need high bandwidth)
8 SATA ports so i can have 2 drive controllers.
@SharQ Short of integrated video Ryzen 3000/5000 was perfect, the down side was that to get integrated video, it dropped a significant number of cores and PCIe lanes(as well as cut the cache)
Meaning that if i wanted an all-in one server, i either got 16 cores, but needed to drop in a GPU(taking up valuable PCIe lanes and a valuable slot) or be limited to 8 cores, drop valuable PCIe lanes from the CPU, but retaining the slot that would have been used by an add in GPU
But Zen4 has more bandwidth, and integrated graphics, without dropping PCIe lanes.
Yes i know that motherboards that support PCIE bifurcation are expensive, i paid ~$500 for an Asrock Rack X470 board, the thing had integrated graphics, but it didnt support any sort of video encoders/decoders, dont get me wrong, i liked being able to have my file server run most things i needed without needing to drop in a GPU, but i needed a second machine for hardware accelerated video.
Now with Zen4, assuming they make boards like they did for Zen+, Zen2, and Zen3, i can have an all-in-one server, which means i only need 3 total machines instead of 6+, my main, my redundant, and my off site.
Right now for under $150, there is a B550 AM4 motherboard with 5 16X slots(yes i realize only the top slot is 16x, but i really only need 1*8x and 3*1x) It supports ECC uDIMM(limited to 128GB because of DDR4) but only has 8 PCIe lanes for M.2 slots (instead of the 12 i want), 6 SATA ports, and 2.5G LAN. All of the desired changes are possible on AM5, as i have a similar featured AM5 board, with 4 full speed M.2 slots(i'd like 2 full speed and 4 1/2 speed)
My point is that all the way back with Ryzen 2000 and 3000, this was possible, it was just that no one offered a motherboard with these features. You either got a gaming motherboard with a 16X slot and a few 1x slots, or you got a mining motherboard with a bunch of physical 1x slots leaving you with nowhere to plug in a SAS controller or high speed networking.
Thankfully companies like Asrock Rack had things close enough that often had 2 16X slots that were electrically 8x, and then 1-2 other slots that were usually physically 1x. The downside being that you had to choose between 16 cores, or 8 cores with an iGPU for hardware transcoding
With Zen4 and X670 now having extra PCIe lanes, and extra IO for fast on board networking, a motherboard meant for a a 7950x, can now have most anything a SMB can want other than ECC RDIMM or DCPMM, there are probably issues with IOMMU but that could in theory be changed with firmware.
With the right PLX/redrivers the 16 lanes of PCIE 5.0 could be converted to 32 lanes of 4.0 for 2 16x slots, wired at 16x speeds, allowing for both high speed storage, and 100Gbps networking
Since you'll revisit to see the scores can you do it on a 7600x as well and see the perf improvmnt too
Will you make a lapping block for the 5800x3d? I'm curious since they are a little warm. Ditto for if you'll ever put out a relocation bracket, say for the aio coolers?
I tested 5800x3d with just a 280mm arctic aio with the offset mount, and the temps are nowhere near warm. Even with hard AVX stressing...
@@royalgm would you be willing to say which offset mount it would be? There's two listed and I'm on a b550f mobo
Would it be feasible to make a block that fits the heat spreader?
Then not only the top but also part of the side would be cooled.
And I would insulate the capacitors with high temperature silicone.
Well, my English isn't so good, hope you guys understand.
Wonder if the offset + direct die makes a difference?
that's on the to do list for the next weeks :)
Could you try Peltier cooling
Offtopic question, but linked to the AM5: is it possible to use either Thermal Grizzly AM5 backplates (M4 or short) to simply replace the standard AM5 backplate, which flexes visibly? Are their AM5 SAM threads the same as the standard AM5 backplate's? Thank you!
Why the m2 slots are covered with that metallic cover?
Can this offset work well for 5000 series CPUs?
EK waterblocks take quite a long time to be shipped this time..I wonder if it's related to all those mounting problems
Is there any info on the release date of the 650 boards?
Do you have a link for purchase of the thermal grizzly cpu guard please
Don't you think, the IHS on AM5 processors are intentionally made thicker ? Because next time when they release the chips with 3D-v chache they won't have to do anything about height, they will just shave off few millimetres of copper from inside of IHS. Plus, next time they can brag about 10-12% more cooling, while all they do is what you've done already by just lapping the IHS.
I seem to recall reading somewhere that it was done to keep the height consistent between AM4 and AM5 and allow cooler compatibility between the two, but your theory is certainly plausible.
@@xHighPotencyx If the had kept AM4 & AM5 cross-compatibility in consideration, it might have been in development since the design & modeling phase, where the socket have new shape & mechanism according to function, but the dimensions could coherent. I think that was very finely detailed technical nonsense just to cover-up before anybody points it out.
Do you think by lapping a lot of the Ryzen 7000 IHS thickness will have any benefit in terms of cooling and thus in boosting? Lets say removing 1mm of the IHS? thanks.
In a previous video Der Bauer already lapped one. 2+celsius improvement.
@@halrichard1969 I know he dellided one, but I can't find a ZEN 4 lapping clip
Can you make a video on shaving down the ihs and see if there is any temp difference?
Wish to see the AM5 Delidding and off set mouthing temperature result
he already did and reduced temps by 20c.
Can u try lapping surface of proc? Since it thicker than 5000 series
man do you think there will be a consumer platform using liquid metal on direct die? It seems feasible with a sponge style socket like the PS5
Would Lapping the IHS help improve temps?
Don't know if it would be safe, but if one would take thermal camera shot of cpu without any cooler, is the die pattern specially visible? Like is one side far hotter.
This looks like an ideal board for me as it has both gen 5 for gpu and m.2 as standard.. I'm building a simple emulation machine which I want everything lol upto ps4 and if I can ps5 and Xbox series X (yeah I know not asking for much are we.) What is the best GPU for emulation or is it a per system basis of NVIDIA will be better for PS games and ATI/AMD for Sega etc? I pretty much want an easy plug and play setup with around 32GB ram, 7600 CPU to start and leaning towards NVIDIA 3060 unless someone has some feedback which could sway my thinking (not to say I'm unhappy, just not completely certain as of yet.)
Thanks and if you leave feedback, give as detailed an answer as you like. If I have follow-up questions I'll gladly ask.
Thankyou
Technically the science is heat travels from high temperature to low temperature. Whether the heat sink is left or right doesn't affect the flow, but here I think contact has a negative impact on heat flow. Processors are reaching high wattages that our cooling methods are needing to be premium quality, i.e fans, lapping, and 100% mounting contact. The alternative is to buy it and use it at 90% efficiency.
In France the Gigabyte B650 Aorus Elite AX listed at 360€, its exact X670 counterpart is at 410€ while the Ryzen 7600X is listed at 430€ all in the same webshop.
Quite the confusing prices tbh... looks like the B650 Aorus Master is gonna be 550€ at this rate.
To improve thermals we need to use PBO2 and set a thermal limit.
I use a different method: Instead of moving the cooler south, I use a bigger cooler. Went with the Alphacool 420 Pro - which is designed for much bigger CPUs.
Pls have sanding results.
Der 8auer when you previously de-lidded the CPU, did it offer you any insight that offsetting the cooler on this architecture of CPU would benefit its cooling at all?
The cores in ryzen are off center since I think Zen 2, and most coolers are actually domed cold plates, meaning they apply the most mounting pressure in the center. if you offset the dome so the cold plate is closest to the IHS over where the cores are, you can gain thermal performance as the cores can have heat more efficiently removed.
With the crazy electricity prices, atm, i'm more interested in downclocking videos :D Knowing how much you can downclock a cpu/gpu and still get great performance. Like the best performance per watt, or smth like that would be a great video.
The easiest way is ECO mode. No doubt you can learn from here how to tune it better than default. If it can be done.
check actually hardcore overclocking video on the 7950x...he does 29'000 cinebench r23 for 65watts O_O when my 5950x used like 170+ for the same score, if you limit the watts you lose 25% perf but spare 75-80% of the power it's bonkers
Where can you buy the cpu guard from?
Beautiful cat
Can I say that i still don't get the new fusion hub architecture? So now we have 3?
I have the Gigabyte x670e and 7950x offset cooling with the provided offset bracket for the Artic 420 AIO.
The included instructions (online) indicated that Artic has seen real world gains in lab with the 7000 series and I have mounted as per their instructions.
Stock, no overclock, I am posting a 37,600 to 37,800 score on R23 which is only slightly worse then a custom cooling setup that can post a 38k score.
I can not verify, but I would say that offset may be beneficial.
Can you please test single core with this offset? Something weird is happening, I am currently posting a single core score in R23 of 2080-2090, which is slightly higher than reviewers scores of non offset mounting… the same reviewers who have a 38,2k score multicore, higher than mine by a measurable margin.
You should make AIO mounting kits for delidded Ryzen's. I am highly thinking about it but refrain because my EK elite AIO mounting posts use a rubbery spacer plus the thread distance is just enough for cpu's with heat spreaders. If I do it i'll need a tap and die for that size and to find how much of the rubber I will need to cut off. That's some cash right there for just myself, now if I ran my own shop, and offered it as a service, different story.
What if you chop down the top portion of the processor to make it finner, as thin as the 5000 series, then get a cooler with a different shape to be offset downward to compensate that gap?
I'm curious to see how the top layer actually make the temp worse, without having to run direct die cooling with a metal plate in between like some youtuber asked for in order to know what was lost here (but which wouldn't translate to the real heat transfert lost compared to chopping down the top of the cpu die cover).
You could lap the top of the IHS and remove about 1mm of material, maybe 1.1 or 1.2 if you want to get really in there. Your best bet from there would be to modify the mounting hardware to be the same distance shorter, so an existing cooler would fit.
I was waiting for this.. I guess that thick IHS doesn't make any difference, I wonder with a thinner IHS how it would behave :P
Does this work with an air cooler because I have a Dark Rock Pro 4
it's a shame you weren't provided with the pricing the B650's seem to be the more budget friendly options and it is more intriguing, if the pricing was right, I'd probably go for a Ryzen 7000 for my rebuild I am planning.
thanks for saying what I needed to hear - "B660E is enough"
What is safe voltage for 5nm will 1.26v all core cause degradation or not
When the delliding tool for am5 will be available? 7950 is too hot🥵
I expected a better result. Looking at a delidded CPU, the upper half of the IHS is exclusively the IOD while the lower half is both CCDs. Half the cooling capability of a standard water block then goes to the IOD, which barely needs cooling, and only half goes to the CCDs, who desperately need as much as possible. You're probably right about the thick IHS.
If I had the tools to machine copper (let alone to make microfins) I would try a custom block with fine fins only over the CCDs and a very narrow jetplate, and thicker fins over the IOD before the outlet, as a sort of zen4-specific direct-die block.
I do wonder if that's something EK, Alphacool, or Corsair are considering. It's essentially the same setup used in GPU blocks, so it's impossible they haven't thought of it before. You can't do it on Intel's monolithic dies and it's been pointless on earlier Ryzen because those just didn't produce enough heat to make it worth bothering with, but with the low efficiency of these new Ryzen chips and AMD's ever growing market share it begins to make more sense.
It was a better result. R23 score was higher 2nd run. Zen 4 will boost all the way to thermal limit always so this result makes sense.
@@shadow7037932 When I wrote my comment those hadn't been posted yet, I just read them. You're right, when you look at cinebench the result is more like what I had expected, but that knowledge only makes me more eager to try direct die cooling.
I'm tempted to try making a custom jetplate for one of my old blocks so all the cool water is directed over the fins that cover the CCDs, just to see if that has a similar improvement to Roman's offset mount. It would be a lot easier than offsetting the cooler mount enough to get the same effect, modern motherboards are very cramped in that area. I would probably have to use an Intel frame and make my own clamping mechanism to move the block far enough.
My motherboard has to arrive before I can start playing though. I have the CPU, but waiting for the board is killing me.
@@CycahhaCepreebha I remember IBM was playing around with direct die water jet cooling long ago. Would be a cool project if you've got the machining tools and skill to make a prototype.
@@shadow7037932 Unfortunately I lack both. I do have the CAD skills to hire Barrow to do a custom block for me, but for a hobby project I don't think it's worth the bother and I've no doubt they charge more than I'd be willing to pay. Feature creep would also come into it at that point, if I'm buying a custom block I'll also try to squeeze a D5 pump into it because that's another thing I've been wanting to do for ages, and then it will just never happen.
I do have the tools to make jet plates and to CNC acetal, so what I might do is make a custom top and jetplate for my old EK Velocity block. I'll direct the water only over the fins that cover the CCDs, and see what happens.
With new boost behavior is max temperature still a good metric? I mean if the temp is the same but it boosts from 200 to 260w to keep that same temp the solution that allows 260w is better right?
That is correct. If they reach the same temperatures, the one with a higher package power and frequency at that temperature is cooling the CPU better.
5:20 102°C?? Nice :)
How to purchase cpu guard from you?
Has anyone tried lapping the cpu? and maybe taking 1mm of the depth.