Makes more sense than Intel putting a 4 core on x299 with 16 pcie lanes that’s for damn sure.
2 Likes
@gtbtk No need to speculate:
6 - 10 core Skylake-X are based on a Low Core Count (LCC) die in a 4x3 arrangement, with 2 slots taken by memory controllers.
12 - 18 core ones are based on High Core Count (HCC) die in a 5x4 arrangement.
Exclusive to Xeons will be an even higher core count eXtreme Core Count (XCC) die, in a 6x5 arrangement, with up to 28 cores.
These three die types have existed for multiple generations. It’s just only now that consumer side gets anything but the smallest one.
The larger die to spread out the active parts of the chip may help. I doubt that the high core chips will clock to neat 5Ghz anyway. Don’t forget that the TIM on both the Skylake X and the Kaby Lake chips is sufficient to run the chips at Stock speeds. Overclocking is not something that is a guaranteed feature. Also, solder would significantly increase the price of manufacturing the chips. How much noise was there about the $1700 6950X?
Keep in mind that solder is not the only approach that can be taken to effectively transfer heat from the die to the heat spreader. Delidding is replacing stock TIM with a different type of TIM to get an improvement after all.
Intel also has the need to use a material that will remain stable over the life of the chip. Liquid metal, by its viscous nature, is not likely to remain stable and in place on the die for the next 10 years without the need for reapplication. It is also conductive so you are always at risk of shorting components under the spreader if it does leak out. not something you want if you have to warranty a product.
What seems to be the core of the issue is the silicon glue that is being used to attach the Heat spreader is holding the spreader too far away from the die to be really effective when clock speeds are increased over stock levels. I suspect that with closer heat spreader to die tolerances, the standard TIM may even be sufficient to run the chip out of spec. Intel needs to look at ways to improve the way they attach the heat spreader so that it sits with tighter tolerances to the die.
Maybe a better adhesive may be a thin layer of CA Superglue instead of the silicon?
2 Likes
I know that there are multiple die types. It is not as though Intel are not leveraging their Xeon production line to create these new products. Many people don’t seem to understand that.
A 4x4 grid will support 10 cores with 4 PCIE grids, 2 IMC grids and 10 cores if they all work. I am still not convinced that the 7900X die is the same as the die used on 7820X with 8 cores enables. Looking at a rectangle of silcon under the spreader only tells us that they are a similar shape and size, not what has been etched into the material underneath.
Exactly that liquid metal mod is the equivalent of supercharging a car. Oh look it runs faster but it will brake down in a year or two. The same happens with liquid metal.
People need to understand this. But i don’t defend intel they should see this coming and i beleave it’s harder to fix as we think.
Then how is AMD doing that same thing even on chips that cost 100,- bucks?
2 Likes
I believe this would also require a rework of the substrate PCB as I have heard it is not robust enough to survive the heat of the soldering process.
I still think it’s a BS reason though. Remember that a delidded CPU has an obvious warranty void condition to it.
by not making anywhere as much profit
So, the profit margin of a 100,- dollar CPU is the reason not to solder the up to $2k HEDT parts? May I ask where that info is coming from? Because in my mind that does not make sense.
Yeah, that made sense in 2015 for small die, non overclockable consumer parts … maybe.
But things have changed. AMD has proven that it can be done.
Intel will have to rethink this.
If its only $50 more go for 1900x, i wish my system was old enough so i could switch to amd. All the best.
For m.2 are you buying samsung 960 evo?
George
Intel gotten lazy over years… more competition is always better for us .
The question should be why Intel is choosing to continue to use TIM and not solder the heat spreaders on the high end chips? Why would they choose to continue to loose goodwill with the negative press if it wasnt for something that they consider their greater good, such as staying profitable?
Broadwell-E chips are soldered, it is only the mainstream chips since ivy bridge and this current generation HEDT chips that are using TIM. yes the solder can expand but the spreader is attached using a flexible silicon material that can be distorted by expansion any way. There are negatives and positives to both methods of heat conduction. I think that Kaby Lake and now skylake-x have demonstrated that the benefits of TIM are significantly less than the negatives that solder brings.
I think it has become changing the process is like rolling over a fat whale, too hard. I think intel will make a new chips and redo the process from scratch.
If they can get the TIM and the heat spreader right there should be no reason why it cant work.
After all delidding is just replacing one TIM for another