7700X lapping, replace IHS or direct die cooling?

[ChrisM255]

Anyone have luck on 7700X with either reducing the IHS (lapping), replacing IHS or direct die cooling?
Any suggestions on which or a kit that actually works well? Seems like a mixed bag of results... mostly seeing that they cannot consistently get good/even contact.
Before anyone says it, I don't care about the warranty

 

[BoggledBeagle]

What temperatures are you getting at what frequency and why does it bother you?

Have you tried limiting the frequency of the CPU so that it does not hit the temperature which bothers you?

What are you cooling the CPU with?

 

[ShrimpBrime]

The mixed results are the differences in coolers, mounting pressure and ambient temps ect.

You won't gain trivial amounts of frequency doing any of this while using ambient cooling methods.

If using a water loop and active chiller, I'd go direct die myself. Without a chiller, would just leave it alone. Or perhaps a quick lap job to level the contact surfaces.
(Opinion)

 

[BoggledBeagle]

I just looked up that this CPU has the max boost frequency 5.4 GHz, that happens with the temperature when you limit it to 5.2 GHz? Or 5 GHz?

 

[ChrisM255]

What temperatures are you getting at what frequency and why does it bother you?

Have you tried limiting the frequency of the CPU so that it does not hit the temperature which bothers you?

What are you cooling the CPU with?

Touching 90+ on stress tests/benchmarks, CO-20+500MHz boost freq, additional negative voltage offset of 0.10 volts, gaming temps are 70 or less, scores above average for performance on that hardware.
This is just where I am currently settled, about every weekend I do some tweaking for fun
S's and G's really. I just like to tinker... this is my next itch.
Kryosheet + AK500 zero dark, was using Kryonaut paste and results are the same as new paste. Kryosheet outperformed my 6 month old Kryonaut paste numbers

I just looked up that this CPU has the max boost frequency 5.4 GHz, that happens with the temperature when you limit it to 5.2 GHz? Or 5 GHz?

Not what I am looking to do

The mixed results are the differences in coolers, mounting pressure and ambient temps ect.

You won't gain trivial amounts of frequency doing any of this while using ambient cooling methods.

If using a water loop and active chiller, I'd go direct die myself. Without a chiller, would just leave it alone. Or perhaps a quick lap job to level the contact surfaces.
(Opinion)

Likely going to try a minor lap to flatten first and keep shaving off up to 0.8-1mm until I am happy or move on to something else to try.
Not really about the freq or performance for me, just for fun. More about the journey and experience.
This is not my "work" system, I can break this thing and it wouldn't matter, just can't game for about 6-12 months while I save.

 

[Chrispy_]

Touching 90+ on stress tests/benchmarks, CO-20+500MHz boost freq, additional negative voltage offset of 0.10 volts, gaming temps are 70 or less, scores above average for performance on that hardware.
This is just where I am currently settled, about every weekend I do some tweaking for fun
S's and G's really. I just like to tinker... this is my next itch.
Kryosheet + AK500 zero dark, was using Kryonaut paste and results are the same as new paste. Kryosheet outperformed my 6 month old Kryonaut paste numbers


Not what I am looking to do


Likely going to try a minor lap to flatten first and keep shaving off up to 0.8-1mm until I am happy or move on to something else to try.
Not really about the freq or performance for me, just for fun. More about the journey and experience.
This is not my "work" system, I can break this thing and it wouldn't matter, just can't game for about 6-12 months while I save.

95C is normal for Zen4, and you'll struggle to avoid it without a hefty dose of expensive custom loop cooling. AMD have stated so many times in so many different channels that Zen4's boost algorithm will actively aim for 95C. If it's not running at 95C it will aggressively add more voltage to boost higher.

Simply put, 90C on stress tests is 100% fine, 100% intentional, and proves you cooler is more than enough for the job.

If you're unreasonably worried by 95C temperatures just set a manual thermal limit in the PBO settings of your BIOS, problem solved. There's genuinely no reason to do that at all, but at least you have that option if you want to take a roundabout way of reducing power draw.

Honestly, if you want to scratch a tinkering itch, spend your time and money on a custom loop - there's very little to be gained with Zen4 on air or AIOs with their limited coldplate performance. Zen4 isn't power-hungry, but it's hard to keep cool because of heat density concentrating in tiny areas

 

[Selaya]

isnt that mostly an ihs problem tho?
iirc direct-die even on air will fix a lot of those temp (non-)issues

 

[Chrispy_]

It's definitely a thick IHS that contributes, but there's no point spending plenty of money buying delidding tools, buying a socket frame, and buying direct die cooling for an AK500 air cooler.

Delidding for direct-die is something you do after you have a high-flow custom water loop, because you need that level of cooling to deal with the heat density.

Even with the right tools and experience, you can easily kill the CPU outright attempting to break the solder seal:

https://youtu.be/y_jaS_FZcjI?t=1292

 

[Selaya]

hm?
im pretty sure loops are there for dealing w/ heat saturation, not density ...
density-related (over-)heating issues are to be solved w/ improved heat transfer (direct die, liquidmetal et cetera); saturation-related issues are to be solved w/ increased soaking capacities (larger towercoolers, or liquid)

 

[ChrisM255]

isnt that mostly an ihs problem tho?
iirc direct-die even on air will fix a lot of those temp (non-)issues

I am absolutely curious about direct die air cooling, haven't seen anyone doing it
If I end up removing the IHS I will be doing air just to see and then water.

It's definitely a thick IHS that contributes, but there's no point spending plenty of money buying delidding tools, buying a socket frame, and buying direct die cooling for an AK500 air cooler.

Delidding for direct-die is something you do after you have a high-flow custom water loop, because you need that level of cooling to deal with the heat density.

Willing to bet that direct die air is better than that fat boy IHS with air like it is currently!
I have done direct die air cooling on intel before with great results. Yeah, not the same as a water cooler, but it was better than stock IHS with a premium AIO by a few degrees
I don't do de-lidding kits. Bench vice and a careful eye has always served me well. De-lidded my buddie's 7800X3D and a coworker's 7900X with no issues so 7000 is trickier, but doable. I don't recommend for the typical user though, get the kit!

The kits I am looking for is more of the direct die cooling kit or replacement IHS kit, not the de-lidder kits, sorry. Should've specified

 

[Chrispy_]

hm?
im pretty sure loops are there for dealing w/ heat saturation, not density ...
density-related (over-)heating issues are to be solved w/ improved heat transfer (direct die, liquidmetal et cetera); saturation-related issues are to be solved w/ increased soaking capacities (larger towercoolers, or liquid)

Somewhat true, as far as I understand it. Saturation as you mention it refers to the radiator/fan capacity, rather than the hotspot heat extraction rate. "Density" issues are just localised saturation of the coldplate or IHS above the hotspot.

Air coolers have density issues since only one or two heatpipes will have direct contact with the hotspot, and heatpipes aren't that quick, so they also have saturation issues when looking at a hotspot rather than the average temperature. Enough heatpipes and fin area will give air coolers fantastic total TDP ratings, but they'll never keep temperatures low because they're terrible at tiny hotspots.

AIOs generally have wimpy little pumps and therefore low flow, which again becomes a density issue because whether it's a 120 or 360mm radiator, it'll be the same low-flowrate because of the anemic AIO pump.

Custom loops are the cheapest way to get a high enough flow rate through a dense enough coldplate. A beefy D5 can push a lot of water past the hotspot. I don't have Zen4, and haven't delidded this 5800X yet, but the difference between high flow and low flow is huge and also the reason I built this loop and ditched the AIO that preceded it.

 

[Count von Schwalbe]

Air coolers have density issues since only one or two heatpipes will have direct contact with the hotspot, and heatpipes aren't that quick, so they also have saturation issues when looking at a hotspot rather than the average temperature. Enough heatpipes and fin area will give air coolers fantastic total TDP ratings, but they'll never keep temperatures low because they're terrible at tiny hotspots.

Which is part of the reason for the thick IHS. Helps spread out the heat to hit more heat pipes. Depending on cooler baseplate geometry, direct die air cooling could be much worse.

 

[AusWolf]

95C is normal for Zen4, and you'll struggle to avoid it without a hefty dose of expensive custom loop cooling. AMD have stated so many times in so many different channels that Zen4's boost algorithm will actively aim for 95C. If it's not running at 95C it will aggressively add more voltage to boost higher.

Simply put, 90C on stress tests is 100% fine, 100% intentional, and proves you cooler is more than enough for the job.

If you're unreasonably worried by 95C temperatures just set a manual thermal limit in the PBO settings of your BIOS, problem solved. There's genuinely no reason to do that at all, but at least you have that option if you want to take a roundabout way of reducing power draw.

Honestly, if you want to scratch a tinkering itch, spend your time and money on a custom loop - there's very little to be gained with Zen4 on air or AIOs with their limited coldplate performance. Zen4 isn't power-hungry, but it's hard to keep cool because of heat density concentrating in tiny areas

This!
I wouldn't worry about 90 C at all.

 

[ChrisM255]

Which is part of the reason for the thick IHS. Helps spread out the heat to hit more heat pipes. Depending on cooler baseplate geometry, direct die air cooling could be much worse.

Direct die air is not the end goal, more of a curiosity. No way I would settle on it. If I do direct die I would do a proper water block. The last real phase with the air cooler is reduced ihs height, then my AIO and if I still feal like tinkering, delidding.
AK500 is not direct heat pipe though there is a mm maybe a smidge more of copper before it hits the heat pipes. Likely not enough mass to work properly that's why I think lapping may be something. Don't truly care for AIOs, had 3 and after about a year you have to drain, clean out the goo out of the block and refill with overpriced anti corrosion liquid. I have heard that the 360 liquid freezer 2 does come close to a lot of water blocks though so I'm curious about that one if I go with another AIO.
They literally made the 7000 IHS extra thick to match the z-axis for compatible mounting with am4 coolers. I imagine the silicone engineers were crying when they heard how thick the IHS was going to be.

This!
I wouldn't worry about 90 C at all.

Not worried, tinkering...

 

[ShrimpBrime]

Touching 90+ on stress tests/benchmarks, CO-20+500MHz boost freq, additional negative voltage offset of 0.10 volts, gaming temps are 70 or less, scores above average for performance on that hardware.
This is just where I am currently settled, about every weekend I do some tweaking for fun
S's and G's really. I just like to tinker... this is my next itch.
Kryosheet + AK500 zero dark, was using Kryonaut paste and results are the same as new paste. Kryosheet outperformed my 6 month old Kryonaut paste numbers


Not what I am looking to do


Likely going to try a minor lap to flatten first and keep shaving off up to 0.8-1mm until I am happy or move on to something else to try.
Not really about the freq or performance for me, just for fun. More about the journey and experience.
This is not my "work" system, I can break this thing and it wouldn't matter, just can't game for about 6-12 months while I save.

Yes, tinkering is fun. Good experience. Spent many hours doing the lapping. With many chips.

This one is FX-55. One of the first lap jobs I ever did.

2 pieces of art, 1 picture

Tip.
Figure 8 motion on a thick mirror. Start at like 400 grit to take the Nickel plating off. Then move up the grit to get a mirror finish. No point in taking measurements in my opinion. Just do it.

Ps. I have more pictures from years ago. Even lapped a Morgan silver dollar to make a nice little cold plate.

 

[ChrisM255]

even lapped a Morgan silver dollar to make a nice little cold plate.

I have to ask how did the silver coin do?

 

[Launcestonian]

OP, I've done this on my 7600X recently. But don't do water cooling, only air. It's the first lap job I've done & are pleased with the results. The only challenge is getting the after market HS to fit with complete contact on the top of the cpu.

 

[ShrimpBrime]

I have to ask how did the silver coin do?

Better than a copper plate, but not by a huge great deal because back to copper waterblock. Maybe a 2c drop at best. Marginal I suppose would be a good way to word it.

I still have it, but seldom use it.

 

[ChrisM255]

OP, I've done this on my 7600X recently. But don't do water cooling, only air. It's the first lap job I've done & are pleased with the results. The only challenge is getting the after market HS to fit with complete contact on the top of the cpu.

Did you just lap the surface or take some height off?

 

[BoggledBeagle]

I am absolutely curious about direct die air cooling, haven't seen anyone doing it

If your air cooler is not direct heatpipe contact type but it has heatspreader built in, it should work well.

Did you just lap the surface or take some height off?

Do not even think about thinning the IHS without milling machine! If you will try to free hand that, it will be a disaster.

 

[ShrimpBrime]

Do not even think about thinning the IHS without milling machine! If you will try to free hand that, it will be a disaster.

Die lapped Athlon 5000 / FX unlocked.

This is Free handed.

 

[BoggledBeagle]

... even lapping is not that easy when you want to create a really flat surface. You are just removing a thin layer of material - because of that you have a lot of information from the remaining material what is happening.

I say it is not possible to remove say 1.5mm of copper from IHS manually and get a flat surface parallel to the original surface.

 

[ShrimpBrime]

... even lapping is not that easy when you want to create a really flat surface. You are just removing a thin layer of material - because of that you have a lot of information from the remaining material what is happening.

I say it is not possible to remove say 1.5mm of copper from IHS manually and get a flat surface parallel to the original surface.

And I'm going to say it is possible. Totally possible.

Because most AMD IHS plates (of the past of PGA socket type) where concaved and was clearly visible during the lapping process.

Since a mirror is typically machined flat, this surface area to tape the wet sanding paper to will suffice nicely.

As you can clearly see in the picture above, that's not an IHS plate, that's the silicon. It's as flat as the mirror.

I'm guessing it needs to be dial indicator accurate for your approval. Wouldn't be necessary.

_____

If a proper lap job is done to IHS plate and the cooler, when put together without thermal interface material, the cooler should lift the CPU off the table. That's how accurately flat you can lap a CPU and cooler together.

I've done all this. 20 years of doing all this. I don't do it anymore because it's usually not worth all the time involved.

 

[BoggledBeagle]

Glass is flexible and thin mirror is not good enough. You need a glass thicker than 10mm. I just checked my lapping glass and it is really that thick.

It is extremelly easy to go from a flat surface making to sphere surface making while hand lapping. And as I said thanks to the small amount of material you are removing you can tell easily what is happening.

Remember, I was talking about removing a huge amount of material from IHS, not lapping.

 

[ShrimpBrime]

Glass is flexible and thin mirror is not good enough. You need a glass thicker than 10mm. I just checked my lapping glass and it is really that thick.

It is extremelly easy to go from a flat surface making to sphere surface making while hand lapping. And as I said thanks to the small amount of material you are removing you can tell easily what is happening.

Remember, I was talking about removing a huge amount of material from IHS, not lapping.

I also have a thick mirror that came from an old infrared alignment machine.

And it's what you see in the picture I've already demonstrated.

The SMDs are approximately 3 slices of paper away from that mirror. I removed quite a bit. It started as soldered actually.

So here, I'll demonstrate, again, how you can see lapping brings it much much flatter. Nice and flat.

This is going to be around .4 to .6mm removal only.

Final product mirror finish 2000grit and higher. Some people go nuts with leather polish, but F all that.

________

Wait till I tell you that very course finish actually does well for thermal transfer...... Don't ask why, it's probably a dumb answer.

Remember, I was talking about removing a huge amount of material from IHS, not lapping.

Oh I see what you're trying to say.

The picture above, Athlon 5000.... On that mirror?

I lapped the entire IHS plate off. it wasn't a conventional de-lid. Many people have done this in the past. Intel chips too.

But 200 grit, I'll take off 1.5mm of material in just a few minutes. By hand. On the mirror. On a flat enough table surface, BAM! be no problem, and flat. Flatter than stock from the factory.

Here's Lapping Solder out of the silicon.

(I haven't looked at these pictures in a long time. It cost me 5 bucks to photobucket to get my photos I uploaded that many years ago. They hold my pictures for ransom to get my money)