News Grinding Off Ryzen 7000 IHS Seemingly Lowers Temps By 10 Degrees Celsius

[Admin]

Grinding down Ryzen 7000's integrated heat spreader reportedly brings down its temperature at the cost of voiding the warranty.

Grinding Off Ryzen 7000 IHS Seemingly Lowers Temps By 10 Degrees Celsius : Read more

 

[TechieTwo]

It's technically ignorant to claim that the Ryzen 7000 IHS prevents the CPU from being cooled properly because it's untrue. Grinding the IHS down isn't very smart either unless you like small expensive paperweights.

 

[InvalidError]

1mm of copper with a mean useful heat-conducting cross-section of 3cmsq adds less than 1C per 100W to thermal resistance between the die and HSF. If you get a 10C cooling improvement from grinding 1mm off, something else such as IHS flatness or rigidity must be at play for the 7-8C beyond copper's thermal resistance.

 

[PlaneInTheSky]

Please don't copy what this idiot is doing, he's clearly never used a sander.

Not wearing any eye protection while sanding metal with an aggressive belt sander is a great way to end up with a piece of metal in your eyeball.

Also, for the love of god don't blow away the metal slivers you just ground off like he did. You don't want that stuff in your lungs.

 

[bit_user]

Has anyone checked how flat the heat spreaders are? One explanation for the improvement is simply improving the thermal interface by eliminating gaps, especially if you do the same thing with your cooler (or the cooler you're using is already really flat).

I've lapped a few heatsinks and a CPU. I didn't bother to take before/after measurements, but others have.

BTW, is it plausible that the height of these heat spreaders has something to do with preparing the platform for CPUs with V-Cache or in-package DRAM (either DDR5X or HBM) stacks? Or are they just waaay too thick for even that to make sense?

My other theory (though it's really more of a fantasy) is that they wanted to use a vapor chamber. I think the ideal scenario would be to somehow make a vapor chamber, the bottom of which is comprised of the PCB/substrate and dies. So, you'd effectively have phase-change direct-die cooling. That would eliminate hotspots on the heatspreader, making it much more effective at conducting heat into whatever you mount atop it.

 

[Amdlova]

Lapping give good result and don't need a machine only time and love. My old pentiums and core 2 Duo have little heat when give 70% oc nice temps. Today i don't care if The cpu will be at 99°C

 

[Co BIY]

BTW, is it plausible that the height of these heat spreaders has something to do with preparing the platform for CPUs with V-Cache or in-package DRAM (either DDR5X or HBM) stacks? Or are they just waaay too thick for even that to make sense?

I think this makes sense. Room for the 3D stack with a thinner heat spreader reaching the same end height for cooler compatibility.

 

[escksu]

BTW, is it plausible that the height of these heat spreaders has something to do with preparing the platform for CPUs with V-Cache or in-package DRAM (either DDR5X or HBM) stacks? Or are they just waaay too thick for even that to make sense?

Nope. Thickness of the heatspreader is to prevent warping. Why AMD needs to do that is because of the shape of the heatspreader. Those cutouts at the sides (to cater for the smd caps) greatly reduce the stiffness of the heatspreader, so it has to be much thicker. thicker HS also reduces hot spot.

These are desktop parts so there won't be any HBM. Vcache wise, still unknown. AFAIK, vcache has limited success as only specific applications are able to utilise the additional cache. Focus of vcache is on server parts, not dekstop parts.

 

[thestryker]

1mm of copper with a mean useful heat-conducting cross-section of 3cmsq adds less than 1C per 100W to thermal resistance between the die and HSF. If you get a 10C cooling improvement from grinding 1mm off, something else such as IHS flatness or rigidity must be at play for the 7-8C beyond copper's thermal resistance.

There is something going on with the 7xxx IHS as removing it gives an outsized reduction in temps and Der8auer told Jayz that he was seeing 6-7c drops maximum from lapping 1.2mm (the 10c drops in the video also were linked to paste). I hope someone is able to figure out what is going on, but the thermal transfer is clearly bad as it stands if these things are having such a big impact.

 

[InvalidError]

Nope. Thickness of the heatspreader is to prevent warping.

The official line from AMD is to maintain the same overall IHS height from the motherboard as AM4 for HSF backwards compatibility at least with coolers that use the standard backplate which is now integrated with the AM5 socket latch and cannot be easily swapped out.

I hope someone is able to figure out what is going on, but the thermal transfer is clearly bad as it stands if these things are having such a big impact.

How thick is the indium solder TIM on these things? I could imagine thinning the IHS allowing it to bow in a little with mounting pressure and cause some of the excess solder to ooze out once heat is applied. It wasn't so long ago that Intel got heavily criticized for high temperatures which were ultimately found to be caused by the TIM being 2-3X thicker than absolutely necessary.

 

[thestryker]

How thick is the indium solder TIM on these things? I could imagine thinning the IHS allowing it to bow in a little with mounting pressure and cause some of the excess solder to ooze out once heat is applied. It wasn't so long ago that Intel got heavily criticized for high temperatures which were ultimately found to be caused by the TIM being 2-3X thicker than absolutely necessary.

Intel had stopped using a soldered TIM for some time which is where a lot of the issues seemed to stem from there. Der8auer didn't mention anything off when he delidded the first 7xxx and I imagine he would have if there had been excess TIM. Nothing from the video seemed to be obvious to my untrained eye as far as appearing to have excess either. This is his original delidding 7xxx video.

I don't want to discount flex having a potential impact because I don't really know enough specifics about the designs etc and thinning it would certainly allow for a lot more potential flex. There's definitely something in the design of the IHS directly relating to the thermal transfer performance being bad. Hopefully someone doing an article/video on lapping will do tests along the way to eliminate flatness as a culprit.

 

[drajitsh]

Has anyone checked how flat the heat spreaders are? One explanation for the improvement is simply improving the thermal interface by eliminating gaps, especially if you do the same thing with your cooler (or the cooler you're using is already really flat).

I've lapped a few heatsinks and a CPU. I didn't bother to take before/after measurements, but others have.

BTW, is it plausible that the height of these heat spreaders has something to do with preparing the platform for CPUs with V-Cache or in-package DRAM (either DDR5X or HBM) stacks? Or are they just waaay too thick for even that to make sense?

My other theory (though it's really more of a fantasy) is that they wanted to use a vapor chamber. I think the ideal scenario would be to somehow make a vapor chamber, the bottom of which is comprised of the PCB/substrate and dies. So, you'd effectively have phase-change direct-die cooling. That would eliminate hotspots on the heatspreader, making it much more effective at conducting heat into whatever you mount atop it.

You have raised some very pertinent points. I had been wondering what AMD had been thinking to have such a chunky heat spreader.

 

[slash3]

The official line from AMD is to maintain the same overall IHS height from the motherboard as AM4 for HSF backwards compatibility at least with coolers that use the standard backplate which is now integrated with the AM5 socket latch and cannot be easily swapped out.

Correct. The LGA socket itself is flatter than AM4's PGA socket (the chip itself sits lower to the motherboard PCB), so the rest of the z-height is effectively made up with the thicker IHS.

 

[-Fran-]

Lapping CPU IHS's is not uncommon in the enthusiasts rings and it's not uncommon for both AMD and Intel. In fact, der8auer has made a whole* business around delidding and lapping CPUs, so this is not surprising in the least.

What is nice about this, is that it gives a lot more credibility to the "the IHS is crap" point I've been making since day 1.

Come on AMD, stop screwing around and fix this.

Regards.

 

[DavidLejdar]

Yeah, I won't be doing that. I.e. Ryzen 5 7600X as is, such has performance benchmark numbers quite on par with Alder Lake's top tier. And that is plenty of performance for me for quite some time.

Sure doesn't seem though that AMD is doing itself a favor if they could really have been like: "Here's extra perhaps 10% of performance, if you can live with getting a new cooler, especially if your current cooler is quite old and not up to the new wattage anyhow."

At least myself, I am getting a new CPU air cooler anyhow (for a new rig, leaving my old rig complete for someone else to use), and later eventually moving on to AIO (in particular as the larger CPU air coolers don't seem to leave room for gen5 M.2 cooler). And whether any of the new cooling is backward compatible with AM4, that doesn't make a difference to me.

 

[bit_user]

These are desktop parts so there won't be any HBM.

Uh... hate to be one of those "But Apple..."-types, but I've been anticipating in-package memory stacks (could be LPDDR, as Apple has done) for at least 5 years. I kinda thought that's where we were headed with Optane DIMMs, but you can use memory tiering to achieve the same effect. Or, just target lower product tiers where not having expandable memory isn't going to be a big issue.

Vcache wise, still unknown. AFAIK, vcache has limited success as only specific applications are able to utilise the additional cache.

What really hurt the 5800X3D was cutting its boost clocks. If they've figured out a way to do it that can still maintain virtually the same turbo headroom, it should be only a net-win.

Focus of vcache is on server parts, not dekstop parts.

Yeah, that's why I actually see in-package memory as being more in-reach, because it's merely a reconfiguration of where the DRAM dies sit. Of course, if you were going to do that, you'd like not to have to make the socket and motherboard support external DRAM, so perhaps it'll happen first in laptops.

 

[bit_user]

The official line from AMD is to maintain the same overall IHS height from the motherboard as AM4 for HSF backwards compatibility

I've read that and it makes very little sense, to me. They're introducing a whole new platform and probably just a small fraction of their users are going to carry forward AM4 coolers. Most of those folks wouldn't be put off buying a Ryzen 7000 CPU, just because they had to get a new cooler for it. In fact, most will get a new cooler, regardless - the additional power envelope provides ample reason to do so.

I hope the 8000-series drops this bad decision and changes the height.

 

[bit_user]

Yeah, I won't be doing that. I.e. Ryzen 5 7600X as is, such has performance benchmark numbers quite on par with Alder Lake's top tier. And that is plenty of performance for me for quite some time.

When I move to AM5 (and it's not going to be this gen, for mostly non-technical/commercial reasons), I plan to use Eco mode with whatever CPU I get. So, I don't expect to be hitting the 95 C that seems to have people so animated about cooling efficiency.

Sure doesn't seem though that AMD is doing itself a favor if they could really have been like: "Here's extra perhaps 10% of performance, if you can live with getting a new cooler, especially if your current cooler is quite old and not up to the new wattage anyhow."

I sure hope the real reason was better than AM4 cooler-compatibility.

 

[InvalidError]

In fact, most will get a new cooler, regardless - the additional power envelope provides ample reason to do so.

I hope the 8000-series drops this bad decision and changes the height.

I'm still using the same 212+ on my i5-11400 that I originally bought for my Core2 13-14 years ago simply because I can and it still gets the job done. No point in wasting money generating unnecessary e-waste for something that doesn't need to change.

As I wrote in my first post in here, the temperature rise due to extra thickness can easily be calculated from the material's properties. Copper's thermal conductivity (342-403 W/mK depending on purity and ingot forming method) is such that the extra thickness cannot account for more than about 2C. The extra 7+C beyond that has to be due to something else.

 

[-Fran-]

I'm still using the same 212+ on my i5-11400 that I originally bought for my Core2 13-14 years ago simply because I can and it still gets the job done. No point in wasting money generating unnecessary e-waste for something that doesn't need to change.

As I wrote in my first post in here, the temperature rise due to extra thickness can easily be calculated from the material's properties. Copper's thermal conductivity (342-403 W/mK depending on purity and ingot forming method) is such that the extra thickness cannot account for more than about 2C. The extra 7+C beyond that has to be due to something else.

Additionally to this, if the IHS wasn't so terrible, you can still get away with the same coolers as long as you get new adapters for the new socket height, which is cheaper than getting a whole new cooler and helps re-use the old ones. This is a good middle ground AMD could try aim towards to.

I mean, shaving* just a little bit of it actually made temps a lot better for a part that doesn't need to get that hot.

Regards.

 

[Kamen Rider Blade]

I'm still using the same 212+ on my i5-11400 that I originally bought for my Core2 13-14 years ago simply because I can and it still gets the job done. No point in wasting money generating unnecessary e-waste for something that doesn't need to change.

As I wrote in my first post in here, the temperature rise due to extra thickness can easily be calculated from the material's properties. Copper's thermal conductivity (342-403 W/mK depending on purity and ingot forming method) is such that the extra thickness cannot account for more than about 2C. The extra 7+C beyond that has to be due to something else.

Spoiler: AM5 IHS Orthographic View

I did the napkin math and the major issue comparing AM4 & AM5's IHS is the cut-outs lose ~30% of the 2D Flat Contact Surface area to transfer heat to the Flat Base of the CPU cooler.
That ~30% loss in surface area + thicker IHS isn't doing anybody any favors.
Yes it solves the "Hot Spot" issue by being a Thermal Capacitor of sorts according to AMD.

But losing ~30% of your Heat Transfer Surface area is a bigger deal than most people think IMO.

If you filled in the gaps made to the cut-outs to match AM4's Contact Surface Area and leave only a opening for the CPU Retention Frame Tabs, you'd actually only lose ~2-3% Contact Surface area and you'd also gain more Copper Volume to act as a Larger Thermal Capacitor as well.

You should also try putting a small bit of TIM between the base and sides of the IHS cut-out and the CPU Retention Frame and use the Retention Frame Tab as a contact point to help transfer heat out of the IHS as well.

If you designed a new IHS, you can try to use the base of the folded down lips of the IHS Retention Frame and a slightly extended & filled out IHS to help transfer heat out of the edges as well and get a over-sized Flat Plate on the CPU cooler to help transfer heat away from not only the IHS, but the Retention Frame.

ThermalRight also offers a All Aluminium Retention Frame. If you add in Thermal paste to the sides and base of the IHS, you could potentially use that as a even larger Heat Sink on top of the IHS, then use a even larger CPU cooler base to sap heat away from both the IHS & ThermalRight Retention Frame.

 

[Kamen Rider Blade]

Uh... hate to be one of those "But Apple..."-types, but I've been anticipating in-package memory stacks (could be LPDDR, as Apple has done) for at least 5 years. I kinda thought that's where we were headed with Optane DIMMs, but you can use memory tiering to achieve the same effect. Or, just target lower product tiers where not having expandable memory isn't going to be a big issue.

I'm expecting on PCB Substrate GDDR VRAM for future generations of GPU.

Why have them located so far away when you can have them sitting on the same PCB as the GPU die and be extremely close.

Even if it's not HBM# memory, the shorter distance in trace wiring alone will help save power & latency.

And in a large IHS & Indium Solder, you have a "All-In-One" GPU/VRAM package that makes for smaller PCB's which makes for more efficient VRM layouts.

HBM proved the benefits by making the PCB smaller. AMD can do it again without using expensive Enterprise Costing HBM for consumer Discrete DeskTop GPU's / Video Cards.

Just use bog standard GDDR7 RAM Memory packages located right next to the GPU Die =D.

If Apple & TSMC can do it for Apple M1/M2, why can't AMD do it for RDNA3 or later iterations?

 

[InvalidError]

I did the napkin math and the major issue comparing AM4 & AM5's IHS is the cut-outs lose ~30% of the 2D Flat Contact Surface area to transfer heat to the Flat Base of the CPU cooler.
That ~30% loss in surface area + thicker IHS isn't doing anybody any favors.

The surface area loss doesn't matter much since the bulk of the heat won't propagate much beyond 4-5mm away from the chiplets through 3mm of copper. That is why DeBauer came up with offset-mount adapters to re-center the HSF over the CPU chiplet(s) instead of IHS and managed to shave a few extra degrees that way.

BTW, if you only count the raised area of the IHS since the low "lip" going around that the retention frame would press on is never going to touch the HSF, the cut-outs only reduce the total surface area by ~15%, not 30%.

 

[Kamen Rider Blade]

The surface area loss doesn't matter much since the bulk of the heat won't propagate much beyond 4-5mm away from the chiplets through 3mm of copper. That is why DeBauer came up with offset-mount adapters to re-center the HSF over the CPU chiplet(s) instead of IHS and managed to shave a few extra degrees that way.

The chiplets aren't centered, they're very much off-set and the cut-outs are near the edges where the chiplets would be.
That extra thermal mass & 2D Contact Area would be very beneficial IMO.

BTW, if you only count the raised area of the IHS since the low "lip" going around that the retention frame would press on is never going to touch the HSF, the cut-outs only reduce the total surface area by ~15%, not 30%.

But that's not where I'm counting from. I'm trying to match the AM4 IHS as much as possible in a 1:1 surface area.

So that's where I get my 30% # from. That "low "lip" area would get raised to the top of the IHS and be part of the Flat Contact Area between the IHS & CPU Cooler Flat Bottom Plate.

I want as much 2D Contact Area as possible =D.

That would also increase the Thermal Capacity of the IHS since there will be that much more copper in the IHS.

If you want to go crazy, use my concept / design of more Flat 2D contact surfacea area & add in a matching ThermalRight style Aluminium Retention frame and use TIM to help sap heat from the sides of the IHS as well.

Then use a CPU Cooler with even LARGER flat contact surface area that would cover the New Customized Aluminium Retention & IHS.

 

[tracker1]

I'm hoping to see a 7950X in the spring when I do my new build. Don't think I'd go so far as to shave ihs material or delid it, though tempted, if debauer comes up with a good clamping shield.

Planning on seeing pbo to a lower performance curve as I'd like to also consume half the power for the 3-5% performance hit. Also debating going open loop for my first time.

I had a very overclocked early athlon when they were just hitting 1ghz and that used a shield for safer mounting...

I think the biggest difference for me may come down to pcie5 storage, though the Samsung 990 pro is a beast for iops. Getting another 15-20% CPU performance won't hurt though.