90c+ CPUs

[freeagent]

Honestly I think it just comes down to size. As an air cooled overclocker 32nm was the last time I found it easy to cool a cpu close to the limits. 22nm was exponentially harder to cool at the same clock speed as 32nm. Then I went to 7nm.. I thought 22nm was a bear to cool.. nope. Now we are 5nm and clock speed is higher than ever. I mean look at Intel.. their next cpu is going to boost to 6ghz.. those speeds were reserved for sub ambient clocks at one time.. now here we are and these are now out of the box boosts. Crazy days.

 

[Zach_01]

Opinion only?

Or is this statement because someone did a delid, made a comment without testing the idea and then we come to the conclusion "trash IHS plate".

Billions of transistors in a small package, I have a feeling any IHS plate the design might be trash because of this aspect.

I'd be interested to see how cool and efficient the processors are at base clocks. Then maybe make some assumptions..... the cpu is overclocking its self to the max its thermally allowed other wise.

Did he use LM on IHS first to help dissipate heat faster from this very high heat density CPU?
No...

Maybe delidding tool has better profit margins from LM or maybe thinking... why not selling both.

 

[Deleted member 185158]

Did he use LM on IHS first to help dissipate heat faster from this very high heat density CPU?
No...

Maybe delidding tool has better profit margins from LM or maybe thinking... why not selling both.

That's a good point. LM vs TIM, the interface material makes a difference.

Paste 8w/mk on average??? (guessing)
LM close to 75w/mk

Honestly I think it just comes down to size.

That's what she said, and why you have babies!! XD

 

[freeagent]

and why you have babies!! XD

She trapped me I couldn’t get out lol.. second one was my fault hahah..

 

[Valantar]

95c isn't what it's supposed to do. It's a by product of reaching for the clocks they're supposed to hit. They aren't meant to just keep getting hotter even if they've met their performance targets.
It's boost for x clock but if y temp is hit first then stay at x until y comes down.
Not if x clock is met with y temp under its limit them push y to its limit.
This shows throughout the stack of zen4 cpus, so the ihs is at fault. AMD didn't suddenly start making 95c stock clock cpus. They started making a trash IHS.

Again: they designed a product that behaves this way. This was not out of desperation, but a conscious series of choices made over the 4-5 year design cycle of a CPU. No, they didn't "suddenly start making 95c stock CPUs" - they did so over a long period of R&D. They consciously chose to maintain cooler compatibility rather than optimize the IHS for low temperatures - as this was an either/or choice with the change to an LGA socket. And, at the same time, they made sure that the CPU could handle those temperatures just fine. So no, 95c is indeed exactly what it is supposed to do. It's how it's designed to behave. They boost until they hit a thermal limit, unless power and/or clock limits are reached first - and all of these are pushed equally far.

Designed to run full tilt 2/4/7. Still within manufacturer specs, should be fine for a long time.

CPU isn't going to reflow solder even at 115c thermtrip.

Non issue.

Pretty much what I said, no?

Opinion only?

Or is this statement because someone did a delid, made a comment without testing the idea and then we come to the conclusion "trash IHS plate".

Billions of transistors in a small package, I have a feeling any IHS plate the design might be trash because of this aspect.

I'd be interested to see how cool and efficient the processors are at base clocks. Then maybe make some assumptions..... the cpu is overclocking its self to the max its thermally allowed other wise.

Der8auer did a delid and got a ~21C temperature drop with direct die cooling and liquid metal. He also commented on the IHS being very thick, which inherently makes it less efficient at thermal transfer - the less material between the core and the cooler, the better, as you want heat into the water/heatpipes as quickly as possible if you want low core temperatures. Doubling IHS thickness is quite significant in this context. Still, it clearly works just fine. Does this mean the IHS is "trash"? No, but it's thicker than would be ideal. It's still fine.

 

[Count von Schwalbe]

It's a by product of reaching for the clocks they're supposed to hit

I think that AMD no longer has a max boost clock in the traditional sense - more of a "Let's boost until we hit 95C!" sort of product. It's more of a "Minimum boost clock if your cooler can handle the official TDP".

The thick IHS does contribute to the ease of hitting the thermal wall though.

 

[Upgrayedd]

If 90+is cool then why don't we all just go back to stock heatsinks?

 

[TheoneandonlyMrK]

Uhhh what? You run thermal throttled 24/7???

They aren't made to run hot either. The IHS is bad. Alder lake IHS gets concave so they made the contact frame.
They're made to hit a performance target, not 95c. It's new because of two new IHS designs. They didn't just change the thermal rules overnight.

No they didn't, that's the point.

All prior processor's could run hot if used.

And yes thermal throttle 24/7 or your use case is a week asss shit.

2xsimulation running continuously Will heat up anything.

And that's on a q6600, with modern high core count (4xthe core's) made to clock as high as possible within power and within 95°c.

I have a name for anyone expecting that to run cool but I'll keep it to myself.

All der8uar proved is that they're specifications are spot on even with too much cooling he couldn't get much more out of the chip, it had hit it's efficiency curve wall already.

 

[Deleted member 185158]

Pretty much what I said, no?

Der8auer did a delid and got a ~21C temperature drop with direct die cooling and liquid metal. He also commented on the IHS being very thick, which inherently makes it less efficient at thermal transfer - the less material between the core and the cooler, the better, as you want heat into the water/heatpipes as quickly as possible if you want low core temperatures. Doubling IHS thickness is quite significant in this context. Still, it clearly works just fine. Does this mean the IHS is "trash"? No, but it's thicker than would be ideal. It's still fine.

I was de-lidding before Der8auer came up with the idea to be profitable from it. Inn fact he uses the thin shaving razor trick that of which I used in a tutorial back in 2009.

Can't make the comment that it's too thick without trying a plate that's either smaller or larger used in the same fashion as the original. Which is near impossible to do because the plate is soldered.

Understanding what the thickness does, and why is a different story. From any experience I have personally, larger (including thicker) plates hold or store thermal energy. This gives you some time to dissipate the heat.

Would you like to see some of my cold plates that I've used after de-lidding soldered processors??

 

[freeagent]

If 90+is cool then why don't we all just go back to stock heatsinks?

Big difference between 90 at 3.7 vs 90 at 5.7

 

[Valantar]

I think that AMD no longer has a max boost clock in the traditional sense - more of a "Let's boost until we hit 95C!" sort of product. It's more of a "Minimum boost clock if your cooler can handle the official TDP".

The thick IHS does contribute to the ease of hitting the thermal wall though.

Someone mentioned somewhere (here? Another thread? No idea) that someone tested these "at stock" under LN2 and they topped out at ~5.7GHz, so there does seem to be a max boost clock, but one that's set sufficiently high that you won't actually reach it under normal operating conditions (save for 1t short term boost).

 

[R0H1T]

If 90+is cool then why don't we all just go back to stock heatsinks?

It's not cool, neither is 90c

Thankfully winter is coming...

It came & then it went, trashing one of the best buildups to a series ever!

Spoiler

 

[AusWolf]

95c isn't what it's supposed to do. It's a by product of reaching for the clocks they're supposed to hit. They aren't meant to just keep getting hotter even if they've met their performance targets.
It's boost for x clock but if y temp is hit first then stay at x until y comes down.
Not if x clock is met with y temp under its limit them push y to its limit.
This shows throughout the stack of zen4 cpus, so the ihs is at fault. AMD didn't suddenly start making 95c stock clock cpus. They started making a trash IHS.

No. You have 3 limits. Boost (in GHz), power (in W) and temperature (in °C). Whenever the CPU reaches one, that's its equilibrium state.

CPUs of the past reached their max boost limit first and called it a day. Then manufacturers started applying more juice and things got out of hand. That's why Intel's PL values and Tau, and AMD's PPT were invented. So the most recent CPUs hit power limits before anything. My i7 11700 is a great example. It has a power limit of 65 W. It runs at 2.8 GHz all-core in Cinebench when it's enforced. I can override it to use its max boost of 4.4 GHz with liquid cooling. Then it uses 160-170 W. This happens far before temperature limits are reached.

AMD's 5 nm chiplets are a different story. It's so small and efficient that you easily run into thermal limits before you even talk about any kind of insane power consumption. It's plain physics. If something is small, it can transfer less of its heat to the environment. That's why Zen 4 reaches its thermal limit before exhausting its power and boost headroom. You can apply some insane watercooling loop, and get up to power limits, but you're still at a limit. Or disable all limits and get up to max boost. That's still a limit. You're always limited by something, no matter what you do.

Whether a CPU is restricted to a certain core clock, or power, or a temperature level does very little to the end user experience.

 

[Valantar]

I was de-lidding before Der8auer came up with the idea to be profitable from it. Inn fact he uses the thin shaving razor trick that of which I used in a tutorial back in 2009.

Can't make the comment that it's too thick without trying a plate that's either smaller or larger used in the same fashion as the original. Which is near impossible to do because the plate is soldered.

Understanding what the thickness does, and why is a different story. From any experience I have personally, larger (including thicker) plates hold or store thermal energy. This gives you some time to dissipate the heat.

The last thing you want your IHS to do is to store thermal energy - you want it to transfer it. Heatpipes have massively higher thermal transfer than copper, and the heatpipes are what meaningfully take heat away from your CPU. Thus, you want whatever is between your heatpipes (or water, I guess) to be as thin as possible, to have as little thermal resistance as possible between your heat source and what does the work in dissipating that heat.

I mean, if your IHS is storing heat, then it's not dissipating it, meaning that relative to the CPU, it's getting hotter - which in turn drives up CPU temperatures as the delta between CPU and IHS shrinks, as thermal delta is directly linked to the efficacy of thermal transfer. This is exactly what you don't want to happen.

As for when and how you and Der8auer started delidding ... who cares? Having done something for a long time does not necessarily equate to being better at it, or knowing more about it. Nor am I saying that Der8auer is the ultimate authority on anything - thick IHSes being a problem for dissipating heat is quite well established knowledge, and certainly not something I'm relying on him to understand. He just measured this to confirm what many have been suspecting.

If 90+is cool then why don't we all just go back to stock heatsinks?

Because those are noisy and can't dissipate more than ~65W before hitting those temperatures?

 

[AusWolf]

If 90+is cool then why don't we all just go back to stock heatsinks?

Because you can reach higher boost with less noise using a better cooler.

 

[Zach_01]

If 90+is cool then why don't we all just go back to stock heatsinks?

This comment is at least absurd

What are you trying to say?
That suddenly all CPUs can operate like 7000?
...or that users shouldnt/won't do their best to get more performance from 7000 by adding cooling?

 

[Valantar]

This is absurd comment.

What are you trying to say?
That suddenly all CPUs can operate like 7000?
...or that users shouldnt/won't do their best to get more performance from 7000 by adding cooling?

No, what they're trying to say is "waaaaa high temperature bad" by tossing out "arguments" that just demonstrate the fundamental lack of understanding behind what they're saying.

 

[Zach_01]

No, what they're trying to say is "waaaaa high temperature bad" by tossing out "arguments" that just demonstrate the fundamental lack of understanding behind what they're saying.

Seems so...

 

[Deleted member 185158]

The last thing you want your IHS to do is to store thermal energy - you want it to transfer it. Heatpipes have massively higher thermal transfer than copper, and the heatpipes are what meaningfully take heat away from your CPU. Thus, you want whatever is between your heatpipes (or water, I guess) to be as thin as possible, to have as little thermal resistance as possible between your heat source and what does the work in dissipating that heat.

I mean, if your IHS is storing heat, then it's not dissipating it, meaning that relative to the CPU, it's getting hotter - which in turn drives up CPU temperatures as the delta between CPU and IHS shrinks, as thermal delta is directly linked to the efficacy of thermal transfer. This is exactly what you don't want to happen.

As for when and how you and Der8auer started delidding ... who cares? Having done something for a long time does not necessarily equate to being better at it, or knowing more about it. Nor am I saying that Der8auer is the ultimate authority on anything - thick IHSes being a problem for dissipating heat is quite well established knowledge, and certainly not something I'm relying on him to understand.

WTF? lol

Practice makes perfect. Of course I'm better at it and know more about it. Cause I've done it many many several times.

In some cases, storing energy before dissipation works. What you don't get is, why.

It's based on time to move BTU. Understanding how to move 1200 btu/h is a different story.

Yes the de-lid is effective. Always has been. Be it 5c or 20c drop, or increase, there's a reaction for the action.

Most of the time, I use a larger plate, not smaller or none at all.

BUT

The cold plate on the waterblock matters as well. That's why Bauer mentioned getting 250$ waterblocks instead of a little AIO with tiny cold plate, not much more impressive than the IHS plate.

Ideally, move the thermals quickly. But AMD is not counting on everyone running naked die, so IHS plate is what your going to get.

 

[oxrufiioxo]

No, what they're trying to say is "waaaaa high temperature bad" by tossing out "arguments" that just demonstrate the fundamental lack of understanding behind what they're saying.

Part of me thinks they should have just ditched AM4 cooler compatibility and gone with a thinner ihs but I'm probably wrong.

They likely looked at platform cost and decided to at least let us keep our coolers....

I have no issues in general with the cpu boosting to 95C immediately because I know why it does it but it still going to make a lot of users uncomfortable running them I guess like with everything eventually people will get over it. I've also seen people ditch perfectly good Ryzen setups sometimes for lesser chips because they couldn't come to grips with how Zen2/3 boost but humans will be humans I guess.

 

[R0H1T]

Well you have to factor in x3d chips as well, they'll be physically "higher" or taller however you look at it. There' also a good chance they'll add more meat on the some of the chiplets like that Xilinx IP.

 

[Count von Schwalbe]

Someone mentioned somewhere (here? Another thread? No idea) that someone tested these "at stock" under LN2 and they topped out at ~5.7GHz, so there does seem to be a max boost clock, but one that's set sufficiently high that you won't actually reach it under normal operating conditions (save for 1t short term boost).

Exactly the boost clock or around it?

 

[Night]

Because the more cooling you're adding the more the CPU boosts so its 95C all over. This continues until power limit reached. Its as "simple" as that

If any one wants decreased temp can either set lower temp limit or power limit.
Again simple as that...

I watched a test of the 7950X on stock settings with extreme cooling (LN2).
The CPU has reached a max of 5.7+GHz all core boost within stock power setting (200+W PPT) at Tdie: -60~70C

EDIT: typo

Seeing the Precision Boost graph for the 7900X with a very capable Artcic Freezer II AIO dropping from 5.6 GHz with 1 thread in use down to 5.2-5.3 GHz with 24 threads in use is what I would call thermal throttling. Now imagine if someone would use a less capable AIO with a thinner radiator, they would most likely be clocking at < 5 GHz under full load. What I mean to say is that the increased work temperature should be addressed correctly by cooling manufacturers and the relevance of it. It doesn't matter that you'll reach 95 °C which is safe, but it matters at what frequency (at least to some)...

 

[kapone32]

I have a 7950X in my cart, have been trying to pull the trigger for a few hours now lol.

Will this be the CPU to drive me away from air cooling?

Thousand bucks man

1023 actually.

Just keep telling myself that is what I paid for my X5690 lol..

I want you to do a week at 105w settings vs normal and see if there is a real difference. That seems to be my most interesting want for the 7950X.

 

[Valantar]

WTF? lol

Practice makes perfect. Of course I'm better at it and know more about it. Cause I've done it many many several times.

Practice makes perfect if you're working on the same thing constantly. If conditions and the tools involved change - like CPUs and technology have done in massive ways over the past decade and a half - then the thinking behind said practice also needs to change, otherwise you risk getting stuck in habits and modes of thinking and doing that applied to older tools but not newer ones. Experience can make you an expert - or it can make you highly resistant to change through entrenching old habits and outmoded knowledge.

From how you're actively appealing to your experience as some form of unquestionable authority here, I'm tempted to place you firmly in the latter category. So far you've done nothing to dissuade me from that impression.

In some cases, storing energy before dissipation works. What you don't get is, why.

It's based on time to move BTU. Understanding how to move 1200 btu/h is a different story.

... and? Were you going somewhere with this?

If your IHS is storing energy, then its temperature is rising. If its temperature is rising, then the delta between the CPU core and the IHS is shrinking (because CPU core thermal rise under load is essentially instantaneous). If the deltaT between core and IHS shrinks, the rate of thermal transfer between the two slows. If the thermal transfer between the two slows, the CPU temperature will again rise until it reaches a new equilibrium point between CPU and IHS temperatures and thermal transfer between them.

Most of the time, I use a larger plate, not smaller or none at all.

... and? Is this an argument? How?



If a thicker cold plate is better, why not just use a thick copper block between your cooler and IHS?

The cold plate on the waterblock matters as well.

... has anyone said otherwise? AFAIK we've been talking 'all else being equal' - if your cooling sucks, then your cooling sucks. We're not comparing coolers here, we're talking about the effects of an IHS in a thermal transfer chain.

Ideally, move the thermals quickly. But AMD is not counting on everyone running naked die, so IHS plate is what your going to get.

Yes, and they made a conscious choice to make that IHS very thick to keep cooler compatibility with AM4 coolers rather than slim it down to improve temperatures - a consumer-friendly choice, but also one that makes their temperatures look worse. They know that the CPUs can handle it, and can reach and maintain very high clock speeds under these conditions, so it's clearly not a problem. Which is what I've been saying all along. Two things can be true at once: that this is a perfectly fine design, and that this IHS is thicker than what is ideal.