Intel Preps Core i7 9700k 8 core 16 Thread Mainstream CPU

Doobedoo · 4 Dec 2017 at 17:57

Silent_Scone said:
Correct temperature, that's an interesting turn of phrase

No it's rather simple really. It means you can run the cpu at it's maximum performance without it overheating, requiring the need to dismantle and replace the inferior Tm.

AndreiD · 4 Dec 2017 at 18:00

Mainly smaller dies are affected by the indium solder developing cracks though. 4C Sandy Bridge chips were pretty sizable dies at 216 mm2.

And the Corning polymer TIM they're using in Coffee Lake chips is actually really good...

Doobedoo · 4 Dec 2017 at 18:12

AndreiD said:
Mainly smaller dies are affected by the indium solder developing cracks though. 4C Sandy Bridge chips were pretty sizable dies at 216 mm2.

And the Corning polymer TIM they're using in Coffee Lake chips is actually really good...

What is the Coffee lake die size? Or what about Sky lake x?

I'm not really sure how you would determine how good the Tim is to be honest, given how much has been said about delidding. I'm sure on the 8400 for example it actually is quite good, but for the k versions and above it seems to be quite an issue.

Doobedoo · 4 Dec 2017 at 18:23

There’s an additional way to get extra performance out of your CPU and that is to delid it. Delidding your CPU will reduce your CPU temperatures by 15-20C under load which allows for higher voltage settings thus higher overclocks.

Delidding is the process of removing the integrated heat spreader (silver part) from the PCB (green part). Once you delid you can remove the thermal interface material (TIM) that is used to conduct heat between the IHS and the PCB and replace it with superior performing thermal paste.

There are a few ways to delid. The safest way is to use a delidding tool that is available on the market.

.

5.1GHz+ is not possible without delidding. However, delidding voids your warranty. If you do decide to delid your processor, you will increase your chance in obtaining a stable 5.1+GHz overclock. Delidding your CPU will reduce your CPU temperatures by 15-20C under load which allows for higher voltage settings thus higher overclocks

https://overclocking.guide/gigabyte-z370-overclocking-coffee-lake/#Delid_for_Lower_Temperatures

So i'm not really sure how you can determine the corning polymer Tim as being actually really good, when you have numerous accounts of a 15-20 degree temperture drop by removing it and applying an alternative method.

All the fuss that was made of the 80+% chance of getting a 5.0ghz+ chip is worthless if you can't keep it cool enough to stop it from throttling.

CAT-THE-FIFTH · 4 Dec 2017 at 18:26

Doobedoo said:
What is the Coffee lake die size? Or what about Sky lake x?

I'm not really sure how you would determine how good the Tim is to be honest, given how much has been said about delidding. I'm sure on the 8400 for example it actually is quite good, but for the k versions and above it seems to be quite an issue.

Apparently its around 149MM2.

Doobedoo · 4 Dec 2017 at 18:28

CAT-THE-FIFTH said:
Apparently its around 149MM2.

So kabylake was only 100mm2?

Well the 7900x is supposed to be 322MM2 so what happened their? Plenty of room to solder.

Silent_Scone · 4 Dec 2017 at 18:30

Doobedoo said:
No it's rather simple really. It means you can run the cpu at it's maximum performance without it overheating, requiring the need to dismantle and replace the inferior Tm.

You can do that anyway. Well, for the most part as long as your cooling is up to scratch. I'm running my 7940X @ 4.7 with preferred cores at 4.8. Non delidded. Even after delid it depends on the workload as to what's possible, and what isn't. As you're still going to be putting quite a lot of current through the die. It's rather simple, really...lol

If being as loose as "correct temperature", then that really should for all intents and purposes mean keeping the package temperature at or around 80c. Don't really want to be veering too far from that.

CAT-THE-FIFTH · 4 Dec 2017 at 18:31

Doobedoo said:
So kabylake was only 100mm2?

Around 125MM2,with Ryzen being around 192MM2 including more chipset functionality.

Edit!!

Some other comparisons:

https://www.anandtech.com/show/9505/skylake-cpu-package-analysis

So Ryzen is probably not much bigger than 4C GT2 Haswell,but also remember a lot of the Intel chip is the IGP and AMD has a larger percentage of the chip dedicated to the CPU part.

Doobedoo · 4 Dec 2017 at 18:34

Silent_Scone said:
You can do that anyway. Well, for the most part as long as your cooling is up to scratch. I'm running my 7940X @ 4.7 with preferred cores at 4.8. Non delidded. Even after delid it depends on the workload as to what's possible, and what isn't. As you're still going to be putting quite a lot of current through the die. It's rather simple, really...lol

So is the 15-20degree figure given by multiple sources. That quite a lot really. And given how much bigger the IHS is on Skylake X it's obviosly a factor with the smaller Coffeelake die.

Doobedoo · 4 Dec 2017 at 18:36

CAT-THE-FIFTH said:
Around 125MM2,with Ryzen being around 192MM2 including more chipset functionality.

So with 50% extra cores coffee lake has only gained 25MM2. That's quite an achievement.

Also I think the Ryzen cpu's has the solder applied in 2 separate blobs so to speak.

CAT-THE-FIFTH · 4 Dec 2017 at 18:48

Doobedoo said:
So with 50% extra cores coffee lake has only gained 25MM2. That's quite an achievement.

Also I think the Ryzen cpu's has the solder applied in 2 separate blobs so to speak.

A large percentage of the CPU is taking up by the IGP. Ryzen Mobile,aka,as Pinnacle Ridge is 209MM2 but it is a full SOC including an IGP,ie,it theoretically needs no chipset.

Also,I assume you meant to be use the picture in this article:

http://www.legitreviews.com/amd-ryzen-7-processors-use-solder-tim-shows-ryzen-delidding-photo_192095

The one you use is for the Core2 Quad IIRC!!

Silent_Scone · 4 Dec 2017 at 18:54

Doobedoo said:
So is the 15-20degree figure given by multiple sources. That quite a lot really. And given how much bigger the IHS is on Skylake X it's obviosly a factor with the smaller Coffeelake die.

I don't think using a different compound improving temperatures as an immediate affect ever came into question. It's only that apparently OCUK and enthusiast know better than Intel's leading metallurgy experts. Sounds silly even saying that.

4K8KW10 · 4 Dec 2017 at 18:58

AndreiD said:
And the Corning polymer TIM they're using in Coffee Lake chips is actually really good...

Compared to solder, it gives 20-25°C in the worse direction.
So, if it is really good, compared to what?

AndreiD said:
Mainly smaller dies are affected by the indium solder developing cracks though. 4C Sandy Bridge chips were pretty sizable dies at 216 mm2

Intel needs to solve these issues and meanwhile to become a more customer-focused corporation.
Stop making small dies, the customers need more cores anyways. Start using solder on the bigger dies.
If you aren't happy with the profit margins, then cut the top management salaries.
In the end, just cut those profits for the share holders, as just said - focus on the profits which the customers get.

CAT-THE-FIFTH said:
A large percentage of the CPU is taking up by the IGP. Ryzen Mobile,aka,as Pinnacle Ridge is 209MM2 but it is a full SOC including an IGP,ie,it theoretically needs no chipset.

Ryzen Mobile (APU) Gen.1 is Raven Ridge. Pinnacle Ridge is Ryzen Desktop (CPU) Gen.2.

nashathedog · 4 Dec 2017 at 19:03

jigger said:
They should have but I'm pretty sure Intel will drop the K chips soon and higher core count lower clockspeed X will be all that's unlocked.

That won't be popular, Except at the AMD share holder meetings.

Doobedoo · 4 Dec 2017 at 19:06

CAT-THE-FIFTH said:
A large percentage of the CPU is taking up by the IGP. Ryzen Mobile,aka,as Pinnacle Ridge is 209MM2 but it is a full SOC including an IGP,ie,it theoretically needs no chipset.

Also,I assume you meant to be use the picture in this article:

http://www.legitreviews.com/amd-ryzen-7-processors-use-solder-tim-shows-ryzen-delidding-photo_192095

The one you use is for the Core2 Quad IIRC!!

Yes you are right that is the Core2 Quad lol. It's from this page. https://hardforum.com/threads/ryzen-8-core-is-actually-made-of-2-quad-dies-its-not.1926166/

It doesn't seem to want to let me post the Ryzen pic. https://www.google.co.uk/search?q=ryzen+delid+shot&client=firefox-b&dcr=0&tbm=isch&source=iu&ictx=1&fir=KXm09MEba0g23M%3A%2CAVo2KBPCriiLmM%2C_&usg=__SBW9xyO-yP2RFgAY3uCAaaWnrTw=&sa=X&ved=0ahUKEwje8tOrgPHXAhVpJsAKHXaqBRMQ9QEIMjAB#imgrc=-lI9PXhvcf_P_M:

You can clearly see how the solder is applied in 2 separate blobs.

That makes sense with the die size though.

Doobedoo · 4 Dec 2017 at 19:07

Silent_Scone said:
I don't think using a different compound improving temperatures as an immediate affect ever came into question. It's only that apparently OCUK and enthusiast know better than Intel's leading metallurgy experts. Sounds silly even saying that.

Not all it's just that the reasons given don't quite add up.

Doobedoo · 4 Dec 2017 at 19:11

nashathedog said:
That won't be popular, Except at the AMD share holder meetings.

It wouldn't matter that much really given how the enthusiast are a small minority and how the majority of those chips are never overclocked. Seems pointless producing an overclockable chip that can't actually be properly overclocked without it being dismantled.

Silent_Scone · 4 Dec 2017 at 19:17

Doobedoo said:
Not all it's just that the reasons given don't quite add up.

What doesn't add up about them?

Doobedoo · 4 Dec 2017 at 19:26

Silent_Scone said:
What doesn't add up about them?

Well your giving the polar bear theory, some are giving the micro fracture theory and other the scrooge theory. Whilst all have valid plausible points the scrooge theory just seems to hold more weight.

Not doubt us simpler folk are just a bit dense and could always do with the more enlightened ones in society guiding us along the correct path. I'd just have thought those bright metallurgy people would have come up with something better than they have by now.

Perhaps Amd are just better at soldering than Intel, maybe it's because they just have too much time on their hands whilst not being able to make gpu's and instead just prefer to practice soldering, while Intel are just too busy swimming in swimming pools full of dollar bills, or counting hundred stacks?

4K8KW10 · 4 Dec 2017 at 19:38

Doobedoo said:
Well your giving the polar bear theory, some are giving the micro fracture theory and other the scrooge theory. Whilst all have valid plausible points the scrooge theory just seems to hold more weight.

Not doubt us simpler folk are just a bit dense and could always do with the more enlightened ones in society guiding us along the correct path. I'd just have thought those bright metallurgy people would have come up with something better than they have by now.

Perhaps Amd are just better at soldering than Intel, maybe it's because they just have too much time on their hands whilst not being able to make gpu's and instead just prefer to practice soldering, while Intel are just too busy swimming in swimming pools full of dollar bills, or counting hundred stacks?

Soldering is practiced by special automation equipment, robots to be precise. Robots should be used for TIM as well.