Intel Skylake / Kaby Lake

[coercitiv]

http://overclocking.guide/the-truth-about-cpu-soldering/

So what exactly did you conclude after reading that article, in relation to Intel using TIM on HEDT chips?

 

[LTC8K6]

You are making the assumption that the TIM is the issue.

The primary issue is the high temps. Talking about the TIM is to try and understand why this is happening.

Trust me, the wider public will see the higher temps and have reservations.

The wider public overall has no idea what their CPU temperature is and they could not care less.
They also know nothing about the insides of their computers. They can see the Intel or AMD sticker on the case.

If a company server room gets a little warmer than it used to, but the work gets done a lot faster, they will not complain about the new servers, AMD or Intel.

Unless the chip fails, or the computer crashes, from the heat, which isn't going to happen, then they will not know anything about it one way or the other, AMD or Intel.

 

[Arachnotronic]

You are making the assumption that the TIM is the issue.

The primary issue is the high temps. Talking about the TIM is to try and understand why this is happening.

Trust me, the wider public will see the higher temps and have reservations.

Wider public doesn't read HW reviews or read forums. They go to Best Buy, Dell/HP/etc. website, and just buy a computer that seems to fit their needs (storage, screen, form factor, etc.)

Intel's marketing is also pretty good and Intel works far more closely with the OEMs to get a big range of SKUs powered by Intel CPUs, so avg Joe is going to be choosing an Intel-based system.

The only people who care about any of this stuff are the few of us who hang out on internet forums, build our own systems, and overclock. It's a nice niche, but it's just that -- a niche.

 

[Burpo]

So what exactly did you conclude after reading that article, in relation to Intel using TIM on HEDT chips?

I concluded that; Depending on the CPU you can either use a normal thermal compound between substrate and heatspreader or connect both with a solder TIM.

And that; Intel has some of the best engineers in the world when it comes to metallurgy. They know exactly what they are doing and the reason for conventional thermal paste in recent desktop CPUs is not as simple as it seems. It makes sense to use conventional thermal paste. Gold and indium are rare and expensive materials, not to mention the extra process involved, and potential failure rates. Normal TIM works well enough across all die sizes and does what it's supposed to, with less inherent risk.

 

[Timmah!]

If i may ask, the liquid metal compound, which people put directly on the die after delidding to decrease the temps, thats different than indium solder, right?

 

[RichUK]

If i may ask, the liquid metal compound, which people put directly on the die after delidding to decrease the temps, thats different than indium solder, right?

Yes.

 

[Timmah!]

Yes.

Thanks. Why doesnt Intel use at least that instead of regular non-conductive TIM, if they are not going to solder? Price, i guess?

 

[DrMrLordX]

If a company server room gets a little warmer than it used to, but the work gets done a lot faster, they will not complain about the new servers, AMD or Intel.

Unless the chip fails, or the computer crashes, from the heat, which isn't going to happen, then they will not know anything about it one way or the other, AMD or Intel.

Are you kidding? Think about your statement.

If i may ask, the liquid metal compound, which people put directly on the die after delidding to decrease the temps, thats different than indium solder, right?

Actually it does have some indium in it:

http://www.coollaboratory.com/pdf/safetydatasheet_liquid_ultra_englisch.pdf

It has significant amounts of gallium and indium in it. Trace amounts of rhodium, silver, tin, zinc, and bismuth. Apparently they added some graphite and copper to make it pastier than CLP.

 

[LTC8K6]

Are you kidding? Think about your statement.

Which one, the room getting 2 degrees warmer or the chip failing?

 

[ManyThreads]

Quick question for you guys - I just learned today that some XMP profiles change the base clock of your entire system. Is it possible to buy 3000-3200Mhz RAM that does *not* do this? I do not want the XMP to be affecting my PCI slots, CPU, etc. if I am understanding that correctly. The one kit I looked at changes base clock to 125 which seemed extreme?

https://www.techpowerup.com/reviews/GSkill/F4-3000C15Q-16GRBB/8.html

Or should I just use 2666Mhz native RAM for my Skylake-X build?

 

[crashtech]

Thanks. Why doesnt Intel use at least that instead of regular non-conductive TIM, if they are not going to solder? Price, i guess?

The polymer TIM Intel uses might not have the best thermal conductivity, but it will outlast most other TIMs aside from indium solder. Anything an enthusiast puts in there usually needs to be redone periodically.

 

[DrMrLordX]

Which one, the room getting 2 degrees warmer or the chip failing?

The one where you implied (if not outright stated) that server room admins care nothing about power usage and heat output.

 

[scannall]

The polymer TIM Intel uses might not have the best thermal conductivity, but it will outlast most other TIMs aside from indium solder. Anything an enthusiast puts in there usually needs to be redone periodically.

Seriously, have you *ever* heard of solder to the heat spreader failing? After over 40 years in this industry, and many thousands of CPU's, that's one failure I've never seen or even heard about. Clearly there is a problem with their current solution. It just plain doesn't work well enough, if parts are overheating in stock configuration. Bad thermal solution, plus pushing the clocks too much as *stock* has made a huge mess. It really can't be denied.

 

[Sweepr]

Coffee Lake-U / Cannon Lake-U Geekbench 3 Score

https://browser.primatelabs.com/geekbench3/8393354

The very first 14nm++ CFL-U results, quite a bit early for a product launching in 2018. Apparently there's a dual-core SKU as well, not only quad-cores - though both should pack GT3e graphics (Iris Plus 740/750). Base clock is up by 800 MHz (32%) compared to Core i7-7660U (assuming it's 15W).

Edit: BIOS suggests it could be Cannon Lake-U (Intel Corporation CNLSFWR1.R00.X086.D00.1705141926). If that's the case, 3.3 GHz is a nice bump from Core i7-7600U (2.8 GHz) as well, not bad for their first 10 nm implementation.

 

[crashtech]

Seriously, have you *ever* heard of solder to the heat spreader failing? After over 40 years in this industry, and many thousands of CPU's, that's one failure I've never seen or even heard about. Clearly there is a problem with their current solution. It just plain doesn't work well enough, if parts are overheating in stock configuration. Bad thermal solution, plus pushing the clocks too much as *stock* has made a huge mess. It really can't be denied.

I think you might have read my post wrong. I mean that indium is the best and most durable bar none, but that was only tangential to the question asked, if you'd care to re-read that. The Intel polymer TIM is designed for long life, it resists drying out and won't pump out like many aftermarket products, but as we all know, it's a compromise.

 

[Azuma Hazuki]

Is that 3.3 GHz base clock or turbo clock? For some reason I just have this weird feeling that CFL-U is going to take cues from the Y-series and have much more aggressive Turbo Boost than previous generations.

 

[TheF34RChannel]

Coffee Lake-U / Cannon Lake-U Geekbench 3 Score

https://browser.primatelabs.com/geekbench3/8393354

The very first 14nm++ CFL-U results, quite a bit early for a product launching in 2018. Apparently there's a dual-core SKU as well, not only quad-cores - though both should pack GT3e graphics (Iris Plus 740/750). Base clock is up by 800 MHz (32%) compared to Core i7-7660U (assuming it's 15W).

Edit: BIOS suggests it could be Cannon Lake-U (Intel Corporation CNLSFWR1.R00.X086.D00.1705141926). If that's the case, 3.3 GHz is a nice bump from Core i7-7600U (2.8 GHz) as well, not bad for their first 10 nm implementation.

Nice find!! Exciting if it's CNL, quite early as well!

 

[LTC8K6]

The one where you implied (if not outright stated) that server room admins care nothing about power usage and heat output.

I actually don't think the room will get any warmer, but I knew someone would chime in about that.

 

[jpiniero]

Coffee Lake-U / Cannon Lake-U Geekbench 3 Score

I dunno, I think it's more likely Coffee. If that's the case, I have to wonder if that means for sure that Cannonlake-U is dead and only Y will come out.

 

[Topweasel]

Which one, the room getting 2 degrees warmer or the chip failing?

Room getting warmer is an issue but not a realistic issue. The cascade failure when there is an interruption of cooling that this will fall victim to earlier is a real issue.

Heat in a room is still going to be based on TDP. The fact that it runs hotter isn't going to be an issue sure the air will be a little hotter coming out but realistically 160w to one CPU the same as 160w from another to the room. The higher temps do mean that case will be running hotter and changes in the room temp will affect the computer quicker and temps inside the case will skyrocket faster. The problem with Intel isn't so much the actual power usage but how razor thin the margins of cooling it has in comparison to other options.

 

[jj109]

PCGH updated their review with gaming results on new firmware
http://www.pcgameshardware.de/Core-...ts/Skylake-X-Basin-Falls-Performance-1230913/

 

[jpiniero]

Far Cry is pretty wacky:

10C20T: 83.4
10C10T: 105.7
4C8T: 104.4
4C4T: 108.2

 

[cjl]

The wider public overall has no idea what their CPU temperature is and they could not care less.
They also know nothing about the insides of their computers. They can see the Intel or AMD sticker on the case.

If a company server room gets a little warmer than it used to, but the work gets done a lot faster, they will not complain about the new servers, AMD or Intel.

Unless the chip fails, or the computer crashes, from the heat, which isn't going to happen, then they will not know anything about it one way or the other, AMD or Intel.

The server room won't get any warmer regardless of TIM, solder, or heat sink choice - heat load into the room is only dependent on the power dissipation of the chip, not how hot the die is.

 

[dullard]

The server room won't get any warmer regardless of TIM, solder, or heat sink choice - heat load into the room is only dependent on the power dissipation of the chip, not how hot the die is.

Minor clarification since so many people forget that power dissipation is not constant: it is also dependent on the amount of time at each power state.

 

[jj109]

Far Cry is pretty wacky:

10C20T: 83.4
10C10T: 105.7
4C8T: 104.4
4C4T: 108.2

Far Cry 4 refused to run on dual cores b/c they specifically had core 2 in their game engine b/c of console restrictions.