Overclocker Delids AMD AM4 CPU

>The overclocker Nam Dae Won has Delidded an AMD AM4 CPU, revealing that the CPU uses solder or liquid metal as its thermal interface material.

Over the past few years deliding CPUs has been popular amongst overclockers, removing the IHS (integrated heat spreader) improve heat transfer between the CPU and your chosen cooling solution. This method has been particularly popular on Intel's LGA 115X series CPUs, which use thermal paste to connect their CPUs to the IHS instead of solder or other more thermally conductive materials.

It looks like AMD's AM4 CPUs will be using a high-quality thermal interface material, which means that mainstream users will likely see no benefit from de-lidding their CPUs.

overclock3d.net/news/cpu_mainboard/overclocker_delids_amd_am4_cpu/1

Enjoy your shit TIM, intelfags!

why wasnt this always standard practice

I remember using a credit card and a small flathead screwdriver to de-lid my K6-2 back in the 90s

cheaper to glue on a slab of iron with some thermal paste than soldering it

also pretty hilarious intel doesnt do this on $500-1500 cpus

I guess, but you already put all the effort in making the CPU which is so much harder, why take such a shit last step, on one of the easiest parts. sigh

RIP INTEL
I
P

I
N
T
E
L

How did he get his hands on an AM4 CPU?

Stoney ridge APUs shipped to OEM partners these are not Zen

Soldered IHS was standard practice from the inception of the IHS and is still use on socket 2011 and above.

* Bristol ridge on desktop, stoney is the low power mobile version. too many ridges

This: I have access to a "thin-client" that uses some high end stoney ridge apu.

>AMD CPUs need liquid metal to not overheat at stock clocks

How is this a good thing?

I've never seen anyone tried to spin a soldered IHS as a bad thing. Bravo.

>Enjoy your shit TIM, intelfags!
I bought my system for $500 prebuilt and it still has the stock cooler on it and its faster than anything AMD has released.

Enjoy your Pajeet processor.

>my pajeet processor

It saves Intel like 8 cents.

>liquid metal as its thermal interface material
Dafuq is this shit? Is this a new meme or amd is actually using mercury/galium?

that is actually one of the rarest pepes

That`s clearly solder, note the gold plated square on the bottom of the IHS. This is required because the solder will not actually stick to the nickel coating. This is standard practice on nearly all AMD cpus, and is nothing note worthy.

For people bashing on Intel for not using this on all of their CPUs, or even just their quad core i7s. There is one reason to go against it, which is durability. The solder eventually cracks and breaks down, potentially destroying the die itself. Or just losing all effectiveness all together. Though an easy argument against this is you still see perfectly functioning 10+ year old AMD cpus that have been soldered.

Probably closer to 50 dollars if you account for everything needed to process it. A gram of indium runs 1-5$ alone.

That`s cute.

>Enjoy your shit TIM, intelfags
5820k user here, i've got a soldered IHS senpai, be jelly of my superior cooling AND IPC.

pretty sure you Xeon uses a soldered IHS

I'd say it would cost Intel even more to solder. I shoot for $100.00 per CPU

Intel employees cost $500k/year, this pushes the costs a lot

AYYMD HOUSEFIRES

>1150 Xeon
Who are you trying to impress?

Enjoy your down syndrome, retard.

It does indeed. All enterprise chips do, as well as all lga 2011 chips in general.

Yeah, I thought about that after posting. I low balled with 50 because even that is 20-25% of a 6700k.

You guys have no idea. I bet it costs Intel $300 per cpu at least.

Only E5, E7 and E8 series. E3 series no have soldered IHS.

U mad, fag? kek

>enjoy you pajeet processor
>post a screenshot of cpu-z
>who are you trying to impress?

Are you retard?

Didn`t know e3`s didn`t get soldered. But the person he responded to was me with my E5-1650v2, bud.

>E8
Not a thing.

So, will this apply to Zen as well?

Only top engineers are making that plus I bet they could have the process of using soldier completely automated so they're only not using it to save money on materials

>I bet they could have the process of using soldier completely automated
Duh homie, do you think those that are soldered are done by hand? No.
>hey're only not using it to save money on materials
Do you think manufacturing equipment is free?

>Do you think manufacturing equipment is free?
It must be for intel if they decided to replace solder with thermal paste :^)

It's far cheaper to have a machine that applies thermal compound and epoxy than one that can accurately solder the IHS to the die.

:^

At least it would be accurate...

I don't know what you think the problem here is.

The thermal paste... isn't accurately on the die and is pushed out onto the interposer...

as long.as.there's a healthy amount of lead in the solder it probably won't crack.

I still don't know what you think the problem is.

What is RoHS?

pretty sure you can say fuck you niggers that lead free shit don't work so kys and if you have good arguments for it they will still let you use lead.

literally what I said tbqf

Right and I don't know what you think the problem is. What do you think thermal paste does?

>What do you think thermal paste does?
Copy the heat of the CPU and paste it to the cooler

>Copy
what?

:^^^^

>neo-Sup Forums soesn't know that half of c2d-era cpus had the same paste as their "muh new intel "smegma under lid" Skylakes"
sssshh, don't tell them!

An eurotard law that doesn't apply to the whole world, you shit brained baboon.

>shit low end C2D's had TIM
Yeah, and they were low end a cheap for what they offered
TOTL Intel C2D's and even mid rangers were soldered

Helps transfer heat by filling in tiny unseeable gaps in the metal...

Keep in mind that this AM4 CPU is also cheap, basically old A8/A10 with better watt usage.

No shit, you stupid fuck, but do you think EU laws don't concern international businesses? Or do you think they're going to set up production lines and distribution channels specifically for EU? That retailers or computer manufacturers are going to follow suit so they don't sell computers with RoHS non compliant components

Fucking retard.

I want to know why you think it flowing onto the interposer is a problem.

>why wasnt this always standard practice

It was, but Intel hates overclockers because they buy cheap CPUs and run them at the speed of 2x more expensive ones, robbing them from income.

So first they nixed overclocking on all CPUs, except the K ones, so you have to buy premium chips for overclocking. Then they gimped the K cpus with shit TIM so they don't overclock as good. If you want to overclock without delidding, you need to get extreme edition $1000 Intel CPUs now.

tl;dr: because Intel is ran by jews.

>It's far cheaper to have a machine that applies thermal compound and epoxy than one that can accurately solder the IHS to the die.

A machine applying thermal compound would use more moving parts such as a pump, and needs to be refilled with said compound during manufacturing, etc.

Soldered IHS uses microwave soldering so they don't damage the die with excessive heat. It needs no fluid components. In fact it is used on EVERY Intel CPU except the K lines.

u mad faget

>No shit, you stupid fuck, but do you think EU laws don't concern international businesses?
>what is iSight
>what is PowerMac G4
>what is PowerMac G4 upgrade cards

fuck the eu

Every problem you just listed for a machine dispensing thermal paste and epoxy also applies to the machine dispensing solder paste and applying microwave heating.

Woo, that is one huge die. Can not wait until they are available.

Theres no way to tell, and it ultimately doesn't matter.
AMD started using soldered heat spreaders with the Godavari refresh of Kaveri, apparently they're still doing it with Bristol Ridge. They might keep it up with their Summit Ridge and Raven Ridge parts, but it anyone's guess at this point.

Properly applied TIM isn't a problem. Intel only had issues because they were using really shitty compound, and they were using excessive shim height on the IHS to prevent cracked dies.

250mm2 is not a huge die.

Daily reminder that you should expect Haswell performance at best.

I guarantee the press is going to do comparisons with Broadwell and Skylake-E CPUs. It will be just like when the RX480 came out and people were expecting GTX980Ti performance out of a $200 GPU. Keep your expectations low and don't let yourself get overhyped. All we really care about is that Intel feels some heat and stops stagnating the market.

Daily reminder that IPC improvements on Intel have been garbage since Haswell anyway.

Even if Zen is 10-15% lower IPC than Skylake, it'll still sell like crazy.

>Daily reminder that you should expect Haswell performance at best.
That much is clear but what's going to really matter is the price.

>it'll still sell like crazy.

Performance doesn't improve brand image. Just look at Intel's anti-trust lawsuits. Pentiums were worst chips. It's going to take some time.

>but you already put all the effort in making the CPU which is so much harder, why take such a shit last step, on one of the easiest parts
To limit overclocking potential to make you buy one of their more expensive CPUs.

Big customers care more about comparable SKUs than branding.

Summit Ridge is actually a stronger play against Xeon-D than anything else, and Raven Ridge will at least be more competitive than any other APUs they've made recently.

Naples will be a lot more economical than 28c-32c Skylake-E simply because of yields, and the on-chip 10 GbE MACs and AES-128 DDR4 controllers don't really have any predecessors in the x86 market.

Expecting AMD to steal half of Intel's market share would be ridiculous, but it's quite plausible they could grab 15%, which would still be disastrous for Intel without even considering competition induced price cutting.

>10 GbE MACs
Uh, you mean NICs?

no, I meant the MACs.
ethernet controllers on CPU dies generally leave the PHYs on external chips, so you aren't tied to only 10GBASE-T, SFP+, or whatever physical layer interface.

you can't really call them complete NICs.

looks like two dies stuck together

Wow I've never seen this before.

>the on-chip 10 GbE MACs and AES-128 DDR4 controllers don't really have any predecessors in the x86 market.

assuming you mean in "big" x86 chips, since you explicitly mentioned Xeon-D right above that, but yeah, pretty much the case.

every once in a while actual employed professionals drop by Sup Forums, and you can actually learn things, oddly enough.

It's this Pajeet again
Here to roleplay as an AMD engineer that's soon about to be out of work?
Skylake-E and Kaby Lake is about to BTFO of Summit Ridge. Start looking for a new job.

...