AMD Ryzen 5 2400G Delidded: Solder vs. Thermal Paste vs. Liquid Metal

[none12345]

"I wish AMD and Intel would learn that basically everyone would rather pay an extra $1 or two and have the IHS soldered than save this insignificant amount."

Not everyone.

Im glad my 1700x is soldered, and when im buying a $400 processor i want it soldered. No reason to paste a $400 processor.

But, if im buying a $99 budget processor, i dont want it costing $5 more for it to be soldered when the solder makes no practical difference.

To spend $2 more in manufacturing will likely translate into $5 more at end retail. Its either that or you are asking AMD to eat into their profit on each chip by about 10%. When they are already making half as much as intel does on their products. From a financial standpoint, their gross margin is already too low for that. So the skus are gonna go up $5, which is going to result in less sales, for no meaningful gain. That makes no sense.

I dont expect solder on a $99 processor, but i do expect them to do a good paste job. That means a small gap between the die and the heat spreader. And also using a quality paste. A quality paste is only pennies over shit paste. It appears that AMD is doing a good job on pasting these apus.

Paste is fine if paste is applied correctly. Intel does a shit job on paste, and the fact they do a shit job on the thermal interface on a premium product speaks volumes to what they think of thier customers.

 

[IInuyasha74]

No, I'm talking $1-$2 more at retail. Solder is dirt cheap. The amount needed to solder the CPU to the IHS probably would cost all of like $0.05. But why do that when they can apply cheap thermal paste for $0.01 per CPU. It's really nonsense.

 

[Solandri]

Are those power consumption measurements for the entire system, or just for the CPU? I can understand power consumption dropping with lower temps for the system (fans don't run as hard). But I'm scratching my head wondering why CPU power consumption would drop. Does it increase voltage in response to higher temps to try to keep the transistors working?

Intel's TIM isn't just mediocre up front, it breaks down within a couple years and reduces your CPU performance (probably) as a part of their planned obsolescence scheme.

That's really interesting. Do you have a source that digs deeper into this? I have an Ivy Bridge 3570K at home that I bought in 2012. Are you saying that the TIM inside of it has dried up? I've never examined my CPU temps over a long enough time to tell if this is happening.

The effect of TIM drying out are overblown. The vast majority of the heat transfer is via metal-on-metal contact. Metal-to-metal conducts heat about two orders of magnitude better than TIM. That's why liquid metal performs so much better than TIM. (And in turn TIM conducts heat about two orders of magnitude better than air. That's why stupid materials like toothpaste can substitute for TIM in a pinch. Nearly anything conducts heat much better than air.)

With a proper application of TIM, the TIM doesn't (shouldn't) form a sandwich layer in between the metal surfaces like an Ore cookie. That can frequently be worse than using no TIM at all. In a proper application, most of the heat is being transferred via direct metal-on-metal contact, and the TIM is only filling the microscopic air gaps due to the two mating surfaces not being perfectly flat. If the TIM dries out, it's still filling the air gap and helping to transfer heat. Maybe slightly degraded due to shrinkage due to drying out. But it's nowhere near as bad as the incorrect Oreo cookie model of TIM would lead you to think.

No, I'm talking $1-$2 more at retail. Solder is dirt cheap. The amount needed to solder the CPU to the IHS probably would cost all of like $0.05. But why do that when they can apply cheap thermal paste for $0.01 per CPU. It's really nonsense.

Maybe the TIM needs to be re-applied on a significant number of processors due to the first application not working within spec. And the initial application failure rate is high enough that it would cost more if they had to throw away badly soldered CPUs. They went to a lot of trouble to create a processor which they're going to sell for tens if not hundreds of dollars. It'd be a shame to have to throw it away because of a poor application of 5 cents worth of solder. Remember, this is an assembly line process. They don't have the luxury of being able to devote an hour of manual labor to each CPU to make sure it's done right the first time.

 

[Nintendork]

@FormatC
You can also undervolt for big decrease in temps and power consumption. An undervolted 2400G system barely takes more than 75w while gaiming.

 

[IInuyasha74]

You guys act like using solder is this real problematic process. Yet the majority of processors to feature an IHS from AMD have been soldered. There are a few exceptions, but the majority of them were soldered. This wasn't problematic for the last series of FX processors, all of which had relatively large dies with a soldered IHS. Why is it now all of a sudden too difficult to do? Why is it now going to result in mass piles of processors tossed out due to soldering issues when that hasn't happened before?

The arguments against soldering don't make sense.

 

[80-watt Hamster]

You guys act like using solder is this real problematic process. Yet the majority of processors to feature an IHS from AMD have been soldered. There are a few exceptions, but the majority of them were soldered. This wasn't problematic for the last series of FX processors, all of which had relatively large dies with a soldered IHS. Why is it now all of a sudden too difficult to do? Why is it now going to result in mass piles of processors tossed out due to soldering issues when that hasn't happened before?

The arguments against soldering don't make sense.

It's not about failure rate, but cost per unit. Solder doesn't allow them to sell these particular models at a higher price. And even if the difference per unit is only $0.04 ($0.01 vs $0.05), that's $40,000 per million units sold. If Igor is right and the process cost is measured in dollars for solder and in cents for TIM (let's say, for argument, the quoted $5 vs. $0.50), the argument for solder essentially evaporates for this segment of the market. AMD's not going to leave $4.5M per million sold on the table for the benefit of a vanishingly small number of buyers who care.

EDIT: rhysiam below has a better/alternate explanation.

 

[rhysiam]

You guys act like using solder is this real problematic process. Yet the majority of processors to feature an IHS from AMD have been soldered. There are a few exceptions, but the majority of them were soldered. This wasn't problematic for the last series of FX processors, all of which had relatively large dies with a soldered IHS. Why is it now all of a sudden too difficult to do? Why is it now going to result in mass piles of processors tossed out due to soldering issues when that hasn't happened before?

The arguments against soldering don't make sense.

I'm struggling to understand why you see this as such a bad thing? We don't ultimately know what the cost difference is. I agree with you when you say it hasn't been a big problem for other processors in the past and some of the speculative costs being thrown around in this thread seem excessively high - on this we agree. Having said that, there clearly is at least some cost difference because otherwise why would AMD introduce a new thermal compound based packaging process for these budget APUs rather than leverage (or slightly modify) an existing Ryzen solder based production line?

FX Processors were soldered because, undoubtedly, they saw massive and tangible gains from solder. Even with solder FX CPUs quickly became thermally constrained once you started raising the voltage and clock speeds. What sort of cooling would be required to handle an FX-9590 with paste under the heatspreader?

The 2200G/2400G on the other hand, as this review shows, aren't better products in any tangible way with solder. It doesn't benefit performance even when overclocked. It doesn't reduce power draw. It doesn't reduce fan speeds in any way that is tangible a user. Yes it reduces temps a little, but you have to look really, really hard to find a use-case where temps are going to be the limiting factor for these CPUs.

So why should AMD take a (small) hit to their margins, or why should everyone pay (a small amount) more for something that doesn't make the product any better for the overwhelming majority of buyers?

 

[rwinches]

It would have been more useful to see the iGPU OC temp results as this is what returns the most benefit and this is an APU after all.

 

[rwinches]

It would have been more useful to see the iGPU OC temp results as this is what returns the most benefit and this is an APU after all.

 

[damric]

Not sure why it took until 2018 for people to complain about AMD using paste on every APU they have made since Llano in 2011. Even the Athlons with the disabled iGPU still had paste. On the other hand, AMD has used solder on all of its higher performance SKUs. Same conspiracy theorists that told us we can no longer overclock our non-k Skylakes and not to use DDR3 with SL/KL which indeed have the same DDR3 controller as HW.

 

[paul prochnow]

I feel I am heading for a fall without this EXACT MotherBoard you are showcasing. Personally a dwarf iTX mobo has no appeal. I feel the BIOS on all AMD motherboards will NOT accept the 2nd Generation Ryzen at this time. WHAT IS THE skinny on THAT ? ? aka BIOS features for 2nd Gen 2xxx Ryzens.

Why not test a Ryzen 2600 right away/ I see them mentioned online. Are they not available to you? THe MIghty TOm's ? ? ? ? WAaaaassup with that ANDTHE BIOS QUESTION> ???
What does a Ryzen 2600 cost now?

 

[AgentLozen]

Why not test a Ryzen 2600 right away/ I see them mentioned online.

Are you referring to the recent Ryzen 2600 leak? That's different from a real product launch.
If I misinterpreted what you were saying then I apologize.

 

[mitch074]

[...]
So why should AMD take a (small) hit to their margins, or why should everyone pay (a small amount) more for something that doesn't make the product any better for the overwhelming majority of buyers?

This. A buck saved is an extra buck in someone's pocket.

 

[photonboy]

GIRORO,
Every system I've built for myself and others going back to an i7-860 is still going strong with NO appreciable temperature difference.

My i7-3770K at 4.4GHz and constant usage (40+ hours per week for 5+ years) still runs great.

My dads 9-year-old laptop runs fine with its Intel CPU.

In fact, I've talked to others in the industry and can't find much reference to an Intel CPU failing due to a TIM issue.... in fact, I can't find much reference to ANY Intel CPU's that have failed for ANY reason beyond the initial manufacturing defect screening.

 

[rontonomo]

why is solder even in the title?

 

[FormatC]

why is solder even in the title?

Because the translator failed (or slept)

My German headline:
AMD Ryzen 5 2400G geköpft: Original gegen Wärmeleitpaste und Flüssigmetall
Direct translation:
AMD Ryzen 5 2400Gt Delidded: Original vs. thermal paste vs. liquid metal

 

[mitch074]

why is solder even in the title?

Because the translator failed (or slept)

My German headline:
AMD Ryzen 5 2400G geköpft: Original gegen Wärmeleitpaste und Flüssigmetall
Direct translation:
AMD Ryzen 5 2400Gt Delidded: Original vs. thermal paste vs. liquid metal

I guess the translator interpreted "original" as "original Ryzen (i.e. soldered)" - "factory" might have been more clear.

I agree though that reading the article should have cleared that up soonish, but you know how it is with publications...

 

[Karadjgne]

Amd has been applying paste on cpus for years. The process for application is identical to how it was done on the fx, Athlons etc. They are already setup to apply paste at the factory. To switch to solder would be expensive, requiring a complete retool of that particular station, for every factory that actually produces just these cpus, while leaving in place the paste machines for the other cpus that aren't soldered.

Considering how much actual cash for such an undertaking amd has at their disposal after all the lawsuits Intel has forced, I'm not particularly surprised at amd not sinking Millions into something that really shows no real world gains over its current process.

 

[mitch074]

Amd has been applying paste on cpus for years. The process for application is identical to how it was done on the fx, Athlons etc. They are already setup to apply paste at the factory. To switch to solder would be expensive, requiring a complete retool of that particular station, for every factory that actually produces just these cpus, while leaving in place the paste machines for the other cpus that aren't soldered.

Considering how much actual cash for such an undertaking amd has at their disposal after all the lawsuits Intel has forced, I'm not particularly surprised at amd not sinking Millions into something that really shows no real world gains over its current process.

No, AMD uses solder on "pure CPU dies" - anything that was originally an APU (thus, all Bulldozer chips) used thermal paste. Ryzen 1xxx uses solder, the Athlon64 and Athlon II used solder too.

 

[AgentLozen]

So are APUs always regarded as lower end products? The only reason I can think of that AMD would solder their CPUs and paste their APUs is that they feel they APUs don't need extra good thermal performance.

 

[mitch074]

So are APUs always regarded as lower end products? The only reason I can think of that AMD would solder their CPUs and paste their APUs is that they feel they APUs don't need extra good thermal performance.

My best be would be on validation process - validating a die against a soldered IHS costs money, using it with thermal paste may require a separate validation process (EPYC and Threadripper use Ryzen 1xxx dies).
There is also the fact that an APU has a very different power profile depending on the die's area, and thermal paste might handle differing dilatation rates better than solid solder - GPUs may just not like solder. Does anyone know of a soldered on GPU? AFAIK it doesn't exist.

 

[FormatC]

Because all GPUs have no heatspreaders.

I tried GPUs with liquid metal between die and a fullcover water block - really impressive. I was able to reduce the delta from 7K to 4K. Sounds not great, but in percent it is a bomb