Intel Core i9-7900X breaks several benchmark world records

[DanceswithUnix]

The move away from solder may have been predicated by technical issues with small chips, but the decision to apply it to the entire product range...? I see $$$$ and hear ka-chings....

Just a few months ago Intel probably didn't think they needed to release a chip like this. I suspect they went for the cheap TIM route because it gave them something they knew would work right now. I presume that TIM is exactly how the xeon chips come packaged, where reliability under well defined conditions is what matters and the TIM is probably "good enough" in a market with no overclockers.

Had they been working on this chip for the usual amount of time, then they could have qualified any process they wanted.

So basically I think this is lack of foresight/incompetence at play, though money grabbing might be a nice secondary feature for them.

Still, nice clock speeds there Intel, shame the price makes it irrelevant.

[elites2012]

Every time someone accuses Intel of using a poor quality TIM i want to scream, do people really think a company that spends $12.7 billion on R&D is going to penny pinch when it comes to TIM for nothing but financial reasons.

Yes they do. the stuff used is the same as the stuff radio shack used to sale.

[elites2012]

Why not, it happens everywhere else. Motor manufacturers spend vast amounts of R&D yet still have to issue recalls because they cheap out on a component.

Because CPUs and material sciences isn't comparable to motor manufacturers where cheaping out on a component isn't financially, let alone scientifically, provable to be detrimental to the final product.

And before anyone jumps down my throat I'm not saying Intel uses the best quality TIM full stop, I'm saying they use a TIM best suited for their specific needs and those needs do not include cutting costs, a TIM needs to do more than just provide low thermal resistance.

the TIM they use is not made by them, and comes in tubes similar to large toothpaste bottles.

[Sunderas]

Why not, it happens everywhere else. Motor manufacturers spend vast amounts of R&D yet still have to issue recalls because they cheap out on a component.

Because CPUs and material sciences isn't comparable to motor manufacturers where cheaping out on a component isn't financially, let alone scientifically, provable to be detrimental to the final product.

And before anyone jumps down my throat I'm not saying Intel uses the best quality TIM full stop, I'm saying they use a TIM best suited for their specific needs and those needs do not include cutting costs, a TIM needs to do more than just provide low thermal resistance.

Yes it is. Just look at Peugeot's last adventure with their rear axles. They cheaped out, the cars broke down. Easy as that. Stop being so anal for Intel.

[scaryjim]

... I presume that TIM is exactly how the xeon chips come packaged, where reliability under well defined conditions is what matters and the TIM is probably "good enough" in a market with no overclockers. ...

It would be very interesting to know the history of TIM under the Xeon IHS as well. I've found an unofficial partial list - from 2008 - of soldered/TIMed CPUs which lists a number of Xeons as soldered, so it certainly hasn't always been the case that Xeons don't use solder, but that list also says that Intel were using both paste and solder on the original Core 2s depending on processor, so it's not like they've previously used one or the other....

[Iota]

Every time someone accuses Intel of using a poor quality TIM i want to scream, do people really think a company that spends $12.7 billion on R&D is going to penny pinch when it comes to TIM for nothing but financial reasons.

Well, if you can come up with other solid reasoning for Intel to move to thermal grease that has been poorly applied between the IHS and CPU since the Ivy Bridge / Haswell era (2012) aside from purely technical limitations that I'm fairly certain they could resolve given the willpower to do so?

This isn't even a case of the TIM being used isn't suited to the purpose at hand, it's poorly applied in the first place (you only have to Google the issues people have had with CPU temps due to the TIM applied under the IHS by Intel).

Or you could look at it another way, if you spent £2k on a CPU and it had poorly applied TIM under the IHS that caused you issues, would you be happy with that? Surely Intel, with all of the resources they have available could come up with a better solution than the one they went with? The cynical old person in me simply says they went with the cost effective solution, instead of actually resolving the problem properly.

AKA - bodge job.

[scaryjim]

... if you spent £2k on a CPU and it had poorly applied TIM under the IHS that caused you issues ...

Issues with the rated stock clock speeds, or issues overclocking? The former would be a big issue, and one that you'd expect Intel to address. The latter, OTOH, isn't really Intel's problem. They sell a processor that's rated to run at a particular speed....

[Corky34]

Yes they do. the stuff used is the same as the stuff radio shack used to sale.

the TIM they use is not made by them, and comes in tubes similar to large toothpaste bottles.

Unless radio shack have started selling Dow Corning TC-5022 and/or Honeywell PCM45F, both of which are listed as component suppliers in this (PDF) Xeon e5 v3 thermal guide, then i doubt it.

I couldn't find a thermal guide that referenced component suppliers for the newer retail chips so you'll have to forgive me for assuming they use comparable thermal compounds in those.

Yes it is. Just look at Peugeot's last adventure with their rear axles. They cheaped out, the cars broke down. Easy as that. Stop being so anal for Intel.

No, it's really not and I'm not going to get into a discussion on whether Peugeot cheaped out when making a rear axle for their car as it's irrelevant to the discussion of thermal compounds.

Also I'm not being "anal for Intel" as you quaintly put it, I'm being "anal" about people jumping on a bandwagon and misappropriating blame because they've not bothered to research a subject because it's the easy option, I'm being "anal" because saying their doing it for cost savings distracts from the very real problems faced by material scientists and how to effectively transfer heat away from ever smaller components.

Well, if you can come up with other solid reasoning for Intel to move to thermal grease that has been poorly applied between the IHS and CPU since the Ivy Bridge / Haswell era (2012) aside from purely technical limitations that I'm fairly certain they could resolve given the willpower to do so?

This isn't even a case of the TIM being used isn't suited to the purpose at hand, it's poorly applied in the first place (you only have to Google the issues people have had with CPU temps due to the TIM applied under the IHS by Intel).

Or you could look at it another way, if you spent £2k on a CPU and it had poorly applied TIM under the IHS that caused you issues, would you be happy with that? Surely Intel, with all of the resources they have available could come up with a better solution than the one they went with? The cynical old person in me simply says they went with the cost effective solution, instead of actually resolving the problem properly.

AKA - bodge job.

Did you even read the research paper i posted a link to?

...the original research that i was having problems finding at the time, this was is the research paper i wanted/should have referenced (PDF).

The solid reason is fully explained in it and as for poor application if you really want i can dig out some more research that shows excessive use of thermal interface material has next to no effect on thermal resistance whereas not applying enough does.

You can call it a bodge job all you like but unlike you Intel has to consider more than just the thermal resistance of the TIM, they can't go using Liquid Ultra or similar high performing TIM's because unlike you they're not going to be there to reapply it in 5-10-20 years or when it's being run in 10-90% humidity levels, when it's being shook around, when it's taken to the north pole or exposed to other harsh environments.

EDIT: If anyones interested TC-5022 has a thermal conductivity of 4.3 W/m-K and Honeywell PCM45F is 2.35 W/m-K.

[mapesdhs]

It's not jumping on a bandwagon when delidding an IB chip can reduce load temps by 35C. Intel switched to poor interface material because SB was too easy to oc. Hordes of users were and many still are sitting on 5GHz+ SB systems with little need to upgrade. Don't take my word for it, tech site reviews of the 3770K and later chips abound with this kind of commentary. It's been like this ever since IB, the chips run too hot unless delidded, whereas every 2700K I've obtained (six so far) will happily run at 5+ with just a simple TRUE and one fan with good temps and sensible vcore. Even SB-E behaves nicely with minimal cooling. One cannot do this with Intel's later CPUs.

One tech site review of the 3770K wrote, "What can you expect from a Core i7-3770K at 4.5GHz compared to a Core i7-2700K at 4.7GHz? Nearly exactly the same performance, it turns out. Though I’m not entirely certain you’d want to leave the -3770K running with one core at 88 degrees for very long, it is stable." 88C for 4.5 is nuts; I get well below 80 for a 2700K at 5.0, and that's with normal Turbo function and power saving states.

This issue continued with each new chip launch along with mediocre speed improvements, eg. toms' headline for the 4770K was, "Haswell Is Faster; Desktop Enthusiasts Yawn". They wrote, "Sort testing was limited to 1.2V to keep heat manageable. Ring/cache ratios are pegged at 3.9GHz, with the memory controller operating at 1,333 MT/s. Of the chips available for sorting, only one is stable at 4.6GHz under full load. A few are capable of operating at 4.5GHz. More run stably at 4.4GHz. Most are solid at 4.3GHz and down." And, "We’re going to have to accept that Haswell-based parts get hotter, faster, it sounds like, and that they might fall a few hundred megahertz short of comparable Ivy Bridge-based parts with conventional cooling methods." This is why SB/SB-E continue to hold up so well and why so many have not yet upgraded for 2500K/2600K/2700K systems. Intel made SB too good and they've been kicking enthusiasts in the balls ever since, like it's our fault we bought them.

[Corky34]

It's not jumping on a bandwagon when delidding an IB chip can reduce load temps by 35C. Intel switched to poor interface material because SB was too easy to oc....

And I'm not saying delidding doesn't reduce temperatures, however the "quality" change in interface material only contributes to a tiny percentage of that overall change, changing from a thermal conductivity of 4.3 W/m-K to 8.6 W/m-K is negligible, delidding works mainly because the bonding line thickness is reduced.

Look I'm no fan of Intel and i hate them for what seem like their almost religious adherence to Moore's law as IMO it's forced them to do some really dumb things like adding an iGPU not because everyone wanted them but because they wanted to increase the surface area in contact with the IHS and it's also caused all sorts of problems with single core performance because high clock speeds cause more heat and smaller dies make it harder to remove that heat.

Having said that i still feel it's wrong to say cheap poor quality TIM is the cause of the problem, that's like saying the oil in your car is the reason for you not winning a drag race against a McLaren P1.

[Tabbykatze]

Having said that i still feel it's wrong to say cheap poor quality TIM is the cause of the problem, that's like saying the oil in your car is the reason for you not winning a drag race against a McLaren P1.

Not quite a good analogy, if I used crap sludge oil as the lubricant for my engine then the performance and power output will be hindered. But if I used Jesus' one eyed snake spray lubricant in my cars engine I will have a much better power and performance ratio.

It's not make or break but it damn well helps.

[Corky34]

No matter what oil you used though your average UK car isn't going to beat a P1, like i said the change in "quality" of interface material only contributes to a tiny percentage.

When thermal interfaces like Mayonnaise, Pink Lipstick, and Toothpaste results in roughly the same delta T as supposedly leading brands of TIM i think it becomes clear that the thermal part of TIM is a little misleading/distracting.

[aidanjt]

1,000,000th of a degree can be the difference between a stable and unstable overclock. And considering the price Intel wants for these things, it would only add a trivial amount to the BoM to go with high volumes of high quality TIM.

[Corky34]

I think you maybe exaggerating a little.

Lets approach this from the other end, what would you consider a high quality TIM, what TIM do you think Intel should be using?

[Iota]

...the original research that i was having problems finding at the time, this was is the research paper i wanted/should have referenced (PDF).

The solid reason is fully explained in it and as for poor application if you really want i can dig out some more research that shows excessive use of thermal interface material has next to no effect on thermal resistance whereas not applying enough does.

That literally only relates to Indium Solder TIM, also I see you banging on happily about Delta T. So, Ts (underside) or Tc (topside)? Also you have to factor in mounting pressure to the IHS and many other factors (such as excessive cycling) that can lead to the failure of Indium Solder TIM over time.

Also, I'm not saying that epoxy is worse than solder - please point that part out? I'm suggesting that the TIM Intel is using isn't perhaps the best option available to them and they've gone with a cheap option. Look, it's fine to disagree, it's completely another thing to start suggesting that I was saying something entirely different.

Intel has already covered the importance of the IHS themselves, way back in the Pentium 4 LGA 775 days. They're well aware of the problems that can occur and the how pressure on the IHS can have an impact on the under-laying substrate.

Perhaps approach this from a different perspective yourself here; explain why Intel chose to go with a poorly performing TIM under the IHS that results in much higher temperatures (Delta Ts / Tc)? Do you honestly think given the research they've done themselves, that they couldn't come up with a better TIM?

Honestly?

Also, you may wish to read this (you may find page 9 particularly interesting) - http://www.lsi.usp.br/~acseabra/grad...0Packaging.pdf

[Corky34]

That literally only relates to Indium Solder TIM, also I see you banging on happily about Delta T. So, Ts (underside) or Tc (topside)? Also you have to factor in mounting pressure to the IHS and many other factors (such as excessive cycling) that can lead to the failure of Indium Solder TIM over time.

You asked for a solid reason for why Intel moved to thermal grease, that's the reason, because Indium Solder TIM started having problems with smaller die's.

Also Delta T = differential temperature

If you had bothered to read the research paper on Indium Solder failure then you'd perhaps know that there is no "mounting pressure to the IHS" and that thermal cycling is exactly what causes Indium Solder to fail in the first place, kind of obvious when you think about it.

Also, I'm not saying that epoxy is worse than solder - please point that part out? I'm suggesting that the TIM Intel is using isn't perhaps the best option available to them and they've gone with a cheap option. Look, it's fine to disagree, it's completely another thing to start suggesting that I was saying something entirely different.

As was pointed out above, that's exactly what you said, here let me quote you...

Well, if you can come up with other solid reasoning for Intel to move to thermal grease that has been poorly applied between the IHS and CPU since the Ivy Bridge / Haswell era (2012) aside from purely technical limitations that I'm fairly certain they could resolve given the willpower to do so?

Is that not you questioning why Intel moved to thermal grease? because it sure seem that way from where I'm sitting.

Also I'm not disagreeing with you, I'm questioning your understanding of the subject you're talking about, I'm attempting to get to the bottom of why you think the TIM Intel (and AMD BTW) use is not the best option available to them and is cheap.

And purely technical limitations take a little more than just willpower to overcome.

Intel has already covered the importance of the IHS themselves, way back in the Pentium 4 LGA 775 days. They're well aware of the problems that can occur and the how pressure on the IHS can have an impact on the under-laying substrate.

Perhaps approach this from a different perspective yourself here; explain why Intel chose to go with a poorly performing TIM under the IHS that results in much higher temperatures (Delta Ts / Tc)? Do you honestly think given the research they've done themselves, that they couldn't come up with a better TIM?

You're assuming it's the supposedly poorly performing TIM that the cause of the problem when that's yet to be proven in any meaningful way.

Like i said if you think it's such a cheap poorly performing TIM what would you have them use instead?

Also, you may wish to read this (you may find page 9 particularly interesting) - http://www.lsi.usp.br/~acseabra/grad...0Packaging.pdf

Don't you think it's a little ironic to be referencing me back to the very same Intel technology journal that i posted?

Perhaps it would be best if you just referenced the particular passage you think supports your opinion.