Intel Skips Solder for Core i5-9400F, Uses Six-Core Die

seanmcc

Distinguished
Dec 19, 2011
29
0
18,530
Is using thermal grease that significant a cost saving measure? I believe that it saves cash, but is the difference that significant? For all the r and d and fab complexity that goes into these chips, it seems like a rather small and petty way to save few bucks for such significant thermal benefits.
 

mischon123

Prominent
Nov 29, 2017
83
2
640
Paste is expensive. Its 1/3 the price of a watercooler for your chip.
Cheap paste and watercooler keeps your chip running at 40-45 Celsius. My R1800x does.
Hot running Intel chips developing cracks is a no go. Having to overamp your chips right out of the factory to compete is an engineering flaw, among others.
 
Mischon123,
Paste is 1/3 the price of a liquid cooler?
First off, we are talking about TIM in between the CPU chip and the heat spreader. You'd need to DELID to see it.

Anyway, nobody knows why Intel is avoiding solder. You can read this but really nobody knows: https://venturebeat.com/2018/10/08/intel-solder-9900k-i9/

I will say Intel Engineers aren't stupid. It's also not that costly in terms of profit margins. Micro-cracks don't seem like much of an issue as they solder Xeons for servers. So the above suggestion it's about conflict-free materials by the process of elimination may be true. So maybe they just used solder now temporarily for the worst offenders temperature wise and will stop using it again soon.
 
"Of course, when you punch out north of a million die per day like Intel does, even a 1% increase in failure rates equates to a huge hit to the bottom line. "

Uh... if they made a BILLION per day it would still be the same low failure rate.
 

rantoc

Distinguished
Dec 17, 2009
1,859
1
19,780
So a chip that's defective (gpu) slammed together with sub-pair thermal transfer vs the proper chip... and sold at the same price as the full "proper" chip.

Yeah, if i bought that by mistake i knew what greedy company i would be avoiding!
 

mischon123

Prominent
Nov 29, 2017
83
2
640
@rantoc: You are definitely on to what constitutes a consumer grade Intel chip. Its most likley a yield problem of epic proportions. 3 out of 4 dies are defective...This is my conservative estimate.
 

Olle P

Distinguished
Apr 7, 2010
720
61
19,090
This was only to be expected for these lower TDP CPUs.

Intel's claim (summarized): "Micro-cracks occur due to (repeated and) rapid variations in temperature."
How are Xeons run? Typically at stable loads and temperatures 24/7.
 

mischon123

Prominent
Nov 29, 2017
83
2
640
Coldstarts are worse than hotstarts. Wear is tear. My experience. I had servers run 24/7 for 7 years straight. No prob. I had workstations getting shut of twice a day. Bang, dang, 2-3 years. My laptop literally shut never off and still running after 12 years. It is what it is.
 

Olle P

Distinguished
Apr 7, 2010
720
61
19,090
My best guess is that 9600K is an (9th gen) eight core die with two cores disabled, whereas 9400F is a six-core die (of 8th gen type) with the IGP disabled.
 

KD_Gaming

Commendable
Nov 7, 2016
41
1
1,530
@photonboy yes even if they made a billion per day it still has the same low failure as a percentage.... the quantity however is much much higher. 10million failed units higher to be exact. And every unit costs money... so now your looking at 2 billion dollars lost at $200 per.
I would say they are justified in keeping failures lower.
 

PaulAlcorn

Managing Editor: News and Emerging Technology
Editor
Feb 24, 2015
857
313
19,360


Micorcracks aren't as much of an issue with larger die, which Xeons tend to have.
 


With bean-counters involved, a difference of a penny a unit is a big deal. While this is anecdotal, it shows my point: I know a guy that worked for a company that produced guitar amps. For one of their products, he found a way to make the amp more reliable, less prone to failure. He pitched his modification to its design, which only raised the production cost by $0.01 per unit. The bean-counters of the company rejected the idea because it broke the price-point... and failures under warranty were already covered under the corresponding budget.
 
(Intel): We did this, because we need that 2$ of solder way more than you, the customers, paying over-priced products due to our inability to transition our 14nm process to 10nm process. Thank you for your support as a s**cker.
 
This: "Intel has also long preached the perils of solder TIM, which can lead to shorter lifespans due to microcracks in the solder TIM that form during heat-induced expansion and contraction. That's more of a concern with smaller die, such as the six-core die on the -9400F. "

After... this: "But enthusiasts still love solder TIM for its superior cooling effects, and AMD's adoption of the material has made it a must-have for enthusiast-class chips. "

Any report on AMD CPU failing because of solder...? Probably not... Once again, typical Intel propaganda, misinformation and lies.
 

rantoc

Distinguished
Dec 17, 2009
1,859
1
19,780
I'm quite darn sure the micro-crack issue is greatly exaggerated to save face from the real issue - The money intel saves on cheap toothpaste tim vs proper soldering on high end dt chips even marketed as overclock able chips.

The issue with increased heat that the TIM causes is the shortened lifespan of the silicon especially so if pushed further due to overclocking and its inability to transfer enough heat between an loaded cpu and a good cooler making the CPU far warmer than the cooler normally allow with proper transfer.

But that's of-course conveniently left out of the discussion as it would point towards TIM as the superior solution that companies who care about their product would use...
 

s1mon7

Reputable
Oct 3, 2018
96
4
4,635
It looks like they're selling 8600/8400 chips with faulty iGPUs as 9400f now.. for the price of a 9400. That's crazy.
 


As per the article:
"Of course, when you punch out north of a million die per day like Intel does, even a 1% increase in failure rates equates to a huge hit to the bottom line."

I'm just saying it's a strange way to word it. When you talk about the "bottom line" it's still relative. 1% is 1%.

*Also, a working CPU that was just MANUFACTURED doesn't get thrown in the garbage because you have sTIM instead of using paste in between the CPU chip and the heatspreader.

These potential solder-related failures would take YEARS to occur typically and that's going to be after a Warranty expires anyway.

So the entire "bottom line" comment in the article makes NO SENSE at all.
 


Even in those large die cases I don't think we have seen any numbers on fail rates. Really we are talking about micro cracks due to the expansion and contraction of the silicon and metal occurring at different rates. I don't know the expansion differences involved but it seems like you would need a pretty large die for this issue to show up at even a small percentage.