News Intel's new fastest gaming CPU spotted at multiple European retailers — Core i9-14900KS costs $100 more for this binned CPU

Status
Not open for further replies.

foxrox

Distinguished
Dec 31, 2007
12
18
18,515
www.youtube.com
The problem is there are few good silicon samples of the 14900K since they have been culling them for a higher-priced KS release. That makes losing in the silicon lottery more likely than ever. The same scenario has happened with prior generations. If you don't buy a "K" near launch day chances are good you'll get a turd sample that uses too much voltage and runs hotter.
 
The problem is there are few good silicon samples of the 14900K since they have been culling them for a higher-priced KS release. That makes losing in the silicon lottery more likely than ever. The same scenario has happened with prior generations. If you don't buy a "K" near launch day chances are good you'll get a turd sample that uses too much voltage and runs hotter.
Yeah no, there have been plenty of sites that show that the K versions often are just as good or the differences being extremely small.
You are just paying extra for the testing.
Also if they would be doing that then they would lose a lot of money since the ks is selling a lot fewer units, at least that's what I would think.
 
  • Like
Reactions: Nyara

Nyara

Prominent
May 10, 2023
61
55
610
The problem is there are few good silicon samples of the 14900K since they have been culling them for a higher-priced KS release. That makes losing in the silicon lottery more likely than ever. The same scenario has happened with prior generations. If you don't buy a "K" near launch day chances are good you'll get a turd sample that uses too much voltage and runs hotter.
Silicon is made in huge waffers that can make a couple dozen to hundred CPUs, a waffer gets tested and viewed in detail before deciding for which CPU or part of CPU it can become. Waffer production quality starts mediocre at the start of the life of a node process, and the engineers gradully figure on the fly how to improve the % of functioning yields, the older a process node, the closer to 100% of succesful yields you reach.

CPU cores are designed with a minimal success rate achieved in the wafer in mind, outside meeting the minimum, it can be anything higher randomly, say 60% is the minimum, it can easily be made of a 60% waffer or 100% waffer. Intel takes months to figure out how low in a waffer % they can go for a determined product, so they play it safe, say a 14900K could easily be produced with a 65% waffer, but Intel will go with 75% since that is the lowest they had time to test.

And here comes KS, KS is just the realization in more detail that they high balled the minimum waffer % success, so all 14900K could squish out a bit extra performance than what they were defaulted for, this is, common sense, as all CPUs can be overclocked, KS is just they up the default to a point where they are 100% sure the waffer quality assigned for K was up for the task. In this regard KS is just relabeled K with a small overclock default, and all K can be overclocked to this point safely usually.

Additionally, when making CPUs from a waffer, one CPU gets picked in the lot and gets tested for a while in a PC, instead of getting packaged. This is just quality control to ensure all process working, but since they are actually testing a CPU, they can just press further the test and guarantee the KS default clocks instead are working stable, then it gets manual packaged and sold as a KS. All the untested CPUs get sold as K instead, even if all shares the same silicon, mostly because KS demand is low, and they better err safe with an already tested CPU to reduce warranty redemption chance a bit.

KS and K are the same, just KS was tested and defaulted a bit more clocks. Honestly most i7 have same p-core performance too, since 10nm+ Intel is likely yielding 80%+ consistently, so if you do not mind giving away 4 e-cores, you will get almost the same silicon lottery there for half the price. This is what benchmarkers did, too, and hence why KS is not record breaking usually when overclocking.

TL;DR: If running a CPU at default, there is no silicon lottery, whole default purpose is setting up a consistent minimum. Silicon quality improves the older is the process node, for all tiers, which by now is close non-existent variance for Intel 10nm+. KS is just manually tested K. If paying the extra for more quality control and higher default is worth for you, good, aside that no difference with non-K.
 
Last edited:
Status
Not open for further replies.