Talk:Transistor count

This is the talk page for discussing improvements to the Transistor count article. This is not a forum for general discussion of the subject of the article.

Put new text under old text. Click here to start a new topic. New to Wikipedia? Welcome! Learn to edit; get help. Assume good faith Be polite and avoid personal attacks Be welcoming to newcomers Seek dispute resolution if needed Article policies Neutral point of view No original research Verifiability

Find sources: Google (books · news · scholar · free images · WP refs) · FENS · JSTOR · TWL

This article is rated C-class on Wikipedia's content assessment scale. It is of interest to the following WikiProjects:

Computing This article is within the scope of WikiProject Computing, a collaborative effort to improve the coverage of computers, computing, and information technology on Wikipedia. If you would like to participate, please visit the project page, where you can join the discussion and see a list of open tasks.ComputingWikipedia:WikiProject ComputingTemplate:WikiProject ComputingComputing ??? This article has not yet received a rating on the project's importance scale.

Tip: Anchors are case-sensitive in most browsers. This article contains broken links to one or more target anchors: [[Zen 2#Desktop processors|Ryzen]] The anchor (#Desktop processors) is no longer available because it was deleted by a user before. The anchors may have been removed, renamed, or are no longer valid. Please fix them by following the link above, checking the page history of the target pages, or updating the links. Remove this template after the problem is fixed | Report an error

I would like to suggest adding GPU chip into example table, because for example G80 from Nvidia counts 681 M transistors, signaling that GPU are not behind general class CPUs. --Quentar 12:00, 2 July 2007 (UTC)[reply]

Hello. Can someone please explain why the Itanium is listed here although he has only 25 000 000 transistors in 2001 and the Pentium4 before him has yet 42 000 000 transistors in 2000 ? ohe of these processors must be terminated from the list, or am i wrong? sry for my english, i'm german. 134.130.78.41 16:07, 11 October 2007 (UTC)[reply]

Well, you're free to be WP:BOLD! I think the main problem here is we need a graphic rather than a data table. Also to plot transistors per dollar or per square millimeter rather than per chip, since the size and cost of the chips in this table varies greatly. Potatoswatter 18:55, 11 October 2007 (UTC)[reply]

The Microprocessor chronology article is quite similar to this. Merge? Collaborate? -- Henriok (talk) 17:05, 23 November 2009 (UTC)[reply]

The table is not in any order now. I started to re-organize it by Year Released, but perhaps it should be by Transistor Count? Organizing by Year of Release makes it easier to follow Moore's Law. Thoughts? It certainly needs to be re-ordered to follow some logic, as right now, it follows none that I can seeDeproduction (talk) 18:52, 2 February 2011 (UTC)[reply]

Make both columns sortable and the reader can amuse himself looking at it either way. --Wtshymanski (talk) 19:07, 2 February 2011 (UTC)[reply]

It's too bad the sort is a string sort and not a numerical sort; kind of useless. This is the second time this week I've been disappointed at the way sorted columns work. --Wtshymanski (talk) 15:57, 3 February 2011 (UTC)[reply]

I changed those columns to use a numeric sort, rather than the default string sort (which I agree was useless). See Help:Sorting#Numerical sorting problems for details. The fix I used seems to be working in the "Microprocessors" and "GPUs" tables. Alas, I can't seem to get the "transistor count" column of the "FPGA" table to sort right -- how can we fix that? --DavidCary (talk) 18:21, 8 August 2013 (UTC)[reply]

I've made a new version of the transistor count image used in this article. Please visit my talk page to make any suggestions before it is added to Moore's law and Transistor count. -- Wgsimon (talk) 15:16, 13 May 2011 (UTC)[reply]

the main page says 1976 — Preceding unsigned comment added by 72.193.24.148 (talk) 15:58, 13 November 2011 (UTC)[reply]

Hello, I see that the Transistor count for Mux 4 input is 24. I guess we are referring to the Mux built using basic Gates (AND / OR Logic). I am not able to figure out the circuit being referred here. The link leads to a circuit which would consume more transistors then 24. Can you please help me to get hold of the circuit for 4 input Mux with the transistor count to be 24?

Thanks, Bukka — Preceding unsigned comment added by Bukka4 (talk • contribs) 10:37, 30 December 2011 (UTC)[reply]

I'm pretty sure that is referring to a mux built from pass transistor logic, such as page 3 of the datasheet for the 74HC153. By my count, each of the two 4:1 muxes on that chip includes 6 inverters and 6 transmission gates, each one built from 2 transistors -- a nFET and a pFET -- so that's where the transistor count of 24 transistors comes from. Perhaps we should include the inverters for the A and B "select" inputs as well (but not the inverters and AND gates associated with G "enable" inputs?), bringing the total to 10 inverters and 6 transmission gates, so a total of 32 transistors. --DavidCary (talk) 04:23, 13 July 2015 (UTC)[reply]

Hello, I am troubled by the same issue. Concerning the transistors count for D-gated latch which consists of 4 NAND gates = 16 transistors. I would agree if the component referred to is simple an SR-latch. Could you please verify this?

Thanks, Tazmed — Preceding unsigned comment added by tazmed (talk • contribs) 10:37, 30 December 2011 (UTC) — Preceding unsigned comment added by 142.58.173.69 (talk) [reply]

Yes, a clocked transparent D latch -- such as the ones in a CD4042 quad clocked D latch or the CD74HC75 transparent latch -- can be built from 8 transistors. One way to do that is the "one of four latches" circuit shown on page 3-112 of the CD4042 datasheet, or in the "latch detail" circuit shown on p. 2 of the CD74HC75 datasheet.

If we neglect the inverters used only to buffer the inputs or outputs, and the 2 inverters per chip used to generate the CL and nCL signals, each latch on those chips is built from 2 inverters and 2 transmission gates, which (in CMOS) are 2 transistors each, so that gives 8 transistors.

Is this way of implementing a transparent D latch something that should be mentioned in the D latch article or the pass transistor logic article or both? --DavidCary (talk) 05:12, 1 July 2015 (UTC)[reply]

785,000 transistors seems a bit much for a 8bit microcontroller, even when fabricated in 2009 — Preceding unsigned comment added by 188.155.186.23 (talk) 23:34, 16 June 2014 (UTC)[reply]

Fixed. See reference link. It was indeed ridiculous. Mightyname (talk) 23:13, 26 August 2014 (UTC)[reply]

There's no citation, and none of the usual media outlets like anandtech seem to know either. — Preceding unsigned comment added by 134.53.244.90 (talk) 22:25, 8 December 2015 (UTC)[reply]

Agree, at least when it comes to server parts. There is currently a reference to techpowerup.com database for Xeon Platinum 8180 that claims 8.000 bln transistors. That may seem a plausible number, but the very same number is also displayed there for any other Skylake or Cascade Lake Xeon: all models from 4-core Platinum 8256 and 8-core Bronze 3106 up to 32-core Platinum 9221 claim the same transistor count, which cannot simply be true, as the chips fall into 3 different groups of physical dies. 213.247.195.138 (talk) 13:13, 16 August 2019 (UTC)[reply]

Intel have web page with information on the number of transistors used for their CPUs:

http://www.intel.com/content/www/us/en/history/museum-transistors-to-transformations-brochure.html

PDF

Bytesock (talk) 05:58, 2 April 2016 (UTC)[reply]

The lede of this article says: "As of 2016, the highest transistor count in a commercially available CPU (in one chip) is over 7.2 billion transistors" which gives any reader unfamiliar with modern digital electronics the clear impression that an IC of any type can (currently) hold no more than 7.2 billion transistors.

Three paragraphs later in the article (still technically in the lede) the article says: "Xilinx currently holds the "world-record" for a FPGA containing more than 20 billion transistors."

Finally in the Memory section the article states: "a 16 GB flash drive contains roughly 64 billion transistors." and then in the table below it shows a 128GB DRAM chip having an estimated 137,438,953,472 (~137 billion) transistors.

I believe this article needs to clarify this for the reader, and I also think this article should provide information on the theoretical transistor count limit by each IC technology type. Comments? Koala Tea Of Mercy (KTOM's Articulations & Invigilations) 09:27, 27 May 2016 (UTC)[reply]

• Still waiting for a response on my concerns above. This article lede correctly starts with The transistor count is the number of transistors on an integrated circuit (IC). Notice it does not say on a processor chip. Why then does the lede not talk about the largest transistor count IC instead of giving undue weight to the much lower density processor-type chips? Koala Tea Of Mercy (KTOM's Articulations & Invigilations) 08:26, 26 July 2016 (UTC)[reply]

RDa: If you check out the Xilinx materials (DS890, page 28), then XCVU440 uses a SSI technology and consists of three dies (called SLRs), roughly 12x32 mm each. Stacked side by side onto a silicon interposer, they form the 36x32 mm 3D IC. 212.79.106.138 (talk) 19:49, 12 April 2018 (UTC)[reply]

Re “numbers not making sense…” I agree (imho). So I made an edit to add a parenthetical comment : “As of 2019, the semiconductor node with the highest transistor density is TSMC's 5 nanometer node, with 171.3 million transistors per square millimeter (note this corresponds to a transistor-transistor spacing of 76.4 nm, far greater than the relative meaningless "5nm") JdelaF (talk) 01:15, 6 December 2022 (UTC)[reply]

We had a long discussion with 213.81.220.134 about this subject but a big part of it is redundant/too long so I decided to remove it. You still can find it as archived here. --Samurai80 (talk) 09:38, 11 October 2019 (UTC)[reply]

--Samurai80 (talk) 09:38, 11 October 2019 (UTC)[reply]

1. FPGAs are not consumer products, putting a date of introduction might not be as easy
2. For consumer products such as CPUs and GPUs, date of introduction = first time a device can be acquired
3. Therefore for FPGAs too it should be the date of the first shipments, and makers tend to clearly announce those first shipments/sampling so it's more simple
4. The word "introduction" is ambiguous, as there are often announcements "introducing" new devices but these are not related to the introduction on the market we need here
5. Especially FPGA makers tend to play on ambiguity and talk a lot about new devices that are not yet available as if they were, and that sometimes are even cancelled
6. Therefore, future products announced with a not enough concrete date of availability should not have a date of introduction set yet
7. Also Versal and Agilex are ACAPs, so maybe not exactly an FPGA, but I think this is a bit the same as what Virtex II Pro were to FPGAs at the time (CPU+FPGA), so I would suggest to keep them in this category
8. No source regarding the actual availability of the SX2800 could be found, but it's on Intel's catalog (+ it's listed on Mouser). So I leave the entry with the date as TBD.
9. It would be nice to have some more Intel FPGAs in this list , but we need sources with info on availability.

--Samurai80 (talk) 08:29, 13 September 2019 (UTC)[reply]
There are chips that cannot be classified in any of the current categories used in this article. The best and most recent example (as of September 2019) is the Cerebras WSE (Wafer Scale Engine) that was announced in August 2019. It is a Machine Learning AI processor. Here are some articles about it: ^{[1] [2] [3]}

This is not a CPU, not a GPU, not an FPGA, not a memory. But given its extreme size, I do think we should include it here. Other semiconductor chips, such as CMOS image sensors could also be included, as some of them can reach enormous sizes, as in this example : ^[1]

But I believe the transistor count is not that big, as the transistors for imaging are much bigger (I think it's 3 transistors per pixel, so even for an enormous 250 MPixel sensor, that's less than a billion transistors).

Should we consider adding a category ? Such as "Other devices" or something ?

=> As no one gave an opinion I added the subsection "Other devices" with the WSE as the only device in this new chart. Nowadays more and more computing devices, for AI, crypto currency, data centers etc. cannot be categorized in the traditional fields, CPU, GPU, FPGA etc. and I do think this category is required. Feel free to discuss here in case you are against it. --Samurai80 (talk) 07:50, 25 September 2019 (UTC)[reply]

Actually that chip (and Graphcore) are technically microprocessors. Also according to sources: "the world’s largest microprocessor"[1]. Note, I put them in the table (and later in the lead), not realizing there was this other table with it there. Yes, they are very specialized, or the application is for AI, but really you could do something else with them. They are accelerators, not standalone CPUs (while that's even debatable): "The WSE — “CPU” simply doesn’t seem sufficient"[2] [EDIT: WSE may or may not be a microprocessor, see here[3][4]] comp.arch (talk) 15:51, 8 December 2019 (UTC)[reply]

Thank you for your answer. I understand your point, and saying it's a microprocessor is as you say technically not wrong. However as explained in this section: Microprocessor#Special-purpose_designs "microprocessor" is the term for general purpose. In the current article specialized circuits such as GPUs are put in different categories. If the WSE was to be put into the microprocessor category then GPUs should also. I actually think the microprocessor category should be renamed "CPU" in my opinion. And circuits that are not CPUs (storage controllers, AI, etc) should be moved to "other chips". Or at the opposite, put everything that is not a memory, including GPUs, into microprocessor. Else I don't see why you would mix CPUs with specialized circuits and put GPUs in a different category. I had not looked to the CPU section recently but to be honest it seems quite messy to me, and there are quite a few types of chips that are structurally so far from CPUs (massively parallel structures for instance) that they shouldn't be put into the same category. --Samurai80 (talk) 11:18, 9 December 2019 (UTC)[reply]

It seems to me if you change to CPU, then you would have to merge the "Transistor computers" section/table into that section. I'm ok with having WSE (and Graphcore?) in separate table, and possibly mention in that section. comp.arch (talk) 17:55, 28 December 2019 (UTC)[reply]

I guess you are right, microprocessors is fine. I will move only the most obvious examples to the "Other devices" category. PS: I moved only the Graphcore as you recommended. I wonder if at some point it would be a good idea to make sub categories out of the microprocessor category (for instance, Personal Computers CPUs, Mobile CPUs and SoCs, Microcontrollers, Interface controllers etc.) since it's pretty big and a bit heterogenous, but this may not be that simple and things aren't often that easy to categorize. It's also slightly odd that I haven't found any DSP in this article (maybe I haven't searched enough though). --Samurai80 (talk) 10:25, 29 December 2019 (UTC)[reply]

I added a "More citations needed" mention at the start of the "Microprocessors" subsection since there is a low percentage of information actually referenced. Unlike other tables, this table doesn't have a dedicated column for references and I suggest adding one. Help to add the required refs would be greatly appreciated since I don't have time to do it myself in a near future. Also as said in the previous talk section, this subsection is very heterogenous, if not messy, and it might be a good idea to separate the contents in several subsections instead of one. For example Microcontrollers - Personal Computer CPUs - Mobile SoCs - Specialized Microprocessors etc. --Samurai80 (talk) 12:52, 29 December 2019 (UTC)[reply]

I completely agree with Samurai80 that this article definitely needs more references, so I added a "ref" column to the "Microprocessors" section.

(I'm not really sure it's necessary, since people have already put many in-line citations directly attached to the transistor count in the "transistor count" column, but I hope you are right that that empty space will encourage people to add references). --DavidCary (talk) 01:00, 18 May 2020 (UTC)[reply]

Thank you David, however I just had a look and the ref column isn't correctly implemented (better check tables' appearance before publishing, they get messed up easily). If it was that simple I would have done it myself, you actually need to add the column to every single item in the table. Also I was thinking about moving the existing refs to that column. So for now I remove what you did on the ref column. Hope you understand. Cheers ! --Samurai80 (talk) 06:49, 1 June 2020 (UTC)[reply]

I would add that in the Microprocessors table almost every single item has the mention of the designer inside the processor name column, which is redundant since there is dedicated column for it. I suggest the GPU table as a much better example of how to make this table. This is way way too messy. --Samurai80 (talk) 07:15, 1 June 2020 (UTC)[reply]

In terms of computer systems that consist of numerous integrated circuits, the supercomputer with the highest transistor count as of 2016^[update] is the Chinese-designed Sunway TaihuLight, which has for all CPUs/nodes (1 trillion for the 10 million cores and for RAM 10¹⁵ for the 1.3 million GB) combined "about 400 trillion transistors in the processing part of the hardware" and "the DRAM includes about 12 quadrillion transistors, and that's about 97 percent of all the transistors."

I thought maybe this was breaking down the 400 trillion between core and and on-die SRAM, but the values given don't add up. Parenthetic removed until such time as some other editor can revise this into comprehensibility.

Additionally, it was badly placed in the sentence and that needs to be addressed too before this is restored. — MaxEnt 15:54, 14 September 2020 (UTC)[reply]

One of the editors deleted my A14 Entry. Shouldn't we have both of these entries? The transistor counts for iPhones and Desktops are different. Keeping both entries allows readers to compare iPhone chips over time. M1 Should also be included so over time as M series Chips are produced, readers can look back at the progression. 73.110.210.136 (talk) 16:46, 14 November 2020 (UTC)[reply]

Greetings to all.

Years ago, I used to refer to this page once in a while for a quick reference, but after checking it a month ago, I was shocked at how outdated it became – stark contrast with, say, AMD/Nvidia GPU lists or EUV article, which are excellent examples of what an article can and should be like. With many latest references on this page going back to 2019-2020, it looks like COVID-19 has taken its toll on it. I made numerous updates over the past month (along with updating 2nm, 3nm and 5nm pages) – all recent anonymous edits (217.xxx and 188.xxx) here and there are mine, and there's still a lot of work to do.

Besides, we can improve the page in at least one principal way. One nice thing to have here is a "transistor density" column for CPUs, GPUs etc. containing ratio of transistor count over die size, another one – a column containing ratio of above-mentioned value to cell-level transistor density which is transistor density calculated using Mark Bohr's metric (60/40 weighted average of flip-flop and NAND gate densities, IIRC) and which is currently quoted on this page with references to SemiWiki or WikiChip (at least for processes introduced in recent years).

The problem with the latter density metric is that it's not necessarily representative of transistor density for a given chip (ratio of its transistor count to its die size). For some, e.g. Apple, it's close; for others, e.g. Intel, it's very far – based on their own quotations of cell-level process node densities and transistor counts, while for designs where most of chip area is in LLC, overall transistor density is not too far from that in LLC tile, and therefore it's SRAM or eDRAM cell area which not only must be present in proper metric, but for LLC-dominated designs its weight coefficient shall be the highest!

Furthermore, contrary to what advocates of this metric hint at (Bohr, Scotten Jones etc.), above-mentioned discrepancy is not necessarily because of poor work done by IC design team; it can be direct consequence of poor work done by process engineers, e.g. when high self-heating of transistors (which is one of the key things to look at in order to judge how good a process technology is) forces logic designers to introduce large guard bands into design in order for a chip to operate without local hot spots, or work at all.

By the way, Intel stopped publishing transistor counts back in 2014 or 2015, if memory serves, and, I suspect, for the very reason of huge discrepancy between cell-level densities which their process folks boasted on one hand and transistor densities in their chips on the other (IIRC, around 40 Mt/mm² for 14nm process and only 10-16 Mt/mm² for 14nm chips), quoting instead only cell-level densities calculated using Bohr's density metric Intel introduced around the time of announcement of their infamous 10nm process and then boasted for years about "industry-leading" density of over 100 Mt/mm² for essentially non-working process!

Bohr's proposal^[1] didn't go unnoticed by press, and response was predictable: people from the industry pointed at the gap between transistor density in Intel's chips and the numbers Bohr's metric and Intel's promo slides started showing.^[2][3] However, for reasons which clearly have nothing to do with science or technology, the metric caught on with a few advocates like Scotten Jones and got propagated.

As a result, a lot of transistor density numbers posted across the web, including this page, are somewhat misleading (don't boast 40 Mt/mm² if all you can achieve in a working chip is 10–16 – the case of Intel's 14nm process) and, worse, non-neutral as long as densities calculated using only Bohr's metric are presented – which is something Wikipedia articles should be free of.

Fortunately, it's not that hard to fix. Unfortunately, I may not have enough free time to complete this task, not to mention updating all of the obsolete data on this page. I'll do what I can in the nearest future until upcoming work prevents me from taking care of the page further (at least on a regular basis), so if at the time you are reading this you see this task wasn't finished, please contribute to completing it using my edits as examples along with further updates of obsolete references, lest the page should become completely out of date.

For a better page and with kind regards to everyone who contributed here before,

AM

188.66.35.63 (talk) 16:10, 20 May 2022 (UTC)[reply]

GPU table needs quite a bit of cleanup, it seems. I fixed several inaccuracies today, will fix some more time permitting. Will also co-locate GPUs of the same family for ease of use – the table not only looks way better after first few edits, now checking density variation within a family is a breeze! If I do it only half-way before leaving regular maintenance, please do what you can to complete it.

Transistor density update: finished with adding densities to all of the tables, atm thinking what would be the best way to add cell-level and chip-level densities and the ratio of the two. Current plan is to add two more columns to the process table only, but I'll give it a thought for a few days more before going ahead.

Best regards, AM 188.66.35.25 (talk) 09:32, 1 June 2022 (UTC)[reply]

https://web.archive.org/web/20220221181000/https://twitter.com/aschilling/status/1495821855667671041

https://www.hardwareluxx.de/index.php/news/hardware/prozessoren/58175-isscc-2022-wie-vier-dies-zu-einem-monolithischen-sapphire-rapids-werden.html

https://web.archive.org/web/20220318011143/https://www.hardwareluxx.de/index.php/news/hardware/prozessoren/58175-isscc-2022-wie-vier-dies-zu-einem-monolithischen-sapphire-rapids-werden.html

<ref>{{Cite web |last=Cutress |first=Dr Ian |title=Intel Xeon Sapphire Rapids: How To Go Monolithic with Tiles |url=https://www.anandtech.com/show/16921/intel-sapphire-rapids-nextgen-xeon-scalable-gets-a-tiling-upgrade |access-date=2022-08-01 |website=www.anandtech.com}}</ref>

Saving sources for future use. Note this data is semi-official, as ISSCC paper does not contain them.

Regards, AM

this user kept vandalized the page with his biased belief, the keep deleting the post despite the sources are given, however by his standard 90% of the post did not come out with source and should they all be deleted? I am requesting moderator of this page or even administrator to step in this case to prevent any further editing. I'm not trying to discriminate against him/her but this can generally triggers a cyber war against wikipedia and since russo-ukraine war still occur I am suggesting we should block every IP located in Russia until war is over! 67.173.134.210 (talk) 20:52, 11 June 2023 (UTC)[reply]

From my talk page https://en.m.wikipedia.org/wiki/User_talk:188.66.32.229:

"the cell = transistor, by your standard 90% on list should be deleted because those number are estimated as well.

look at the fabrication list, intel's 14nm density is 37.5mtr/mm^2 and TSMC's 16nm is 28.8mtr/mm^2 and GP100 Pascal titan XP is 25.08mtr/mm^2 which the number between two are likely close, this would definitely apply to intel as well. the number on list including those that never have any source should be taking down as well according to your logic.

so you're talking out of your ass because you think Intel "advertised" their fabrication note? learn how to spell you Russian garbage!!

let me ask you again, what makes you think that you have right to delete updates? just wandering. 67.173.134.210 (talk) 06:45, 10 June 2023 (UTC)"

From https://en.m.wikipedia.org/wiki/Special:MobileDiff/1159424301

"this IP is vandalized the page and need to be block from editing!!! Administrator on this web need to do something! that Russian is trying to attack the American website and need to be block or ban!!"

I saw the insults you posted on my talk page only after checking your record, and first off, I want to ask you if you have age-related hormonal disbalance or whether you were diagnosed with some mental deviation and undergo relevant therapy. For that would at least somewhat explain the way you communicate. Explain, but not excuse though.

Second, when you see someone act in violation of rules, the right thing to do is not harassing the editor in edit descriptions or talk pages, but submitting a report to WP:ANI. So go there and repost your insults if you really think that you act fully within established norms and I don't, but I can tell you in advance that the result would be a quick block – and not just for you, but for any future sockpuppet to show up afterwards who would try to push through this or another unsourced or unreliable data or engage in vandalism and/or uncivil conduct out of silly childish revenge or for other reasons – especially considering that your editing pattern (showing up with a questionable, incorrect or disruptive edit, edit-warring despite repeated clear explanations why exactly it's not OK, and explosion with personal attacks after failure in some cases) is highly reminiscent of an editor who has been haunting a number of Intel- and AMD-related articles of late and – I suspect – is the same person with a long history of blocks and block evasion:

https://en.m.wikipedia.org/wiki/User:176.9.113.53

https://en.m.wikipedia.org/wiki/User:Xselant

https://en.m.wikipedia.org/wiki/User:CristoCalis

https://en.m.wikipedia.org/wiki/User:LeaveMeB

https://en.m.wikipedia.org/wiki/User:Halvleder

https://en.m.wikipedia.org/wiki/User:LagoonMoon

https://en.m.wikipedia.org/wiki/User:APD4711

+ possibly some more.

Now, as for transistor count of Rocket Lake, I have clearly and repeatedly explained you in the edit descriptions of reverts that various unsourced claims by media should not be used for reference even in general, whereas in this case they are not only unsourced, but resulting transistor density contradicts authentic data on 14nm chips coming straight from Intel's mouth (10–16 Mtr./mm²) – there's half a dozen of Intel's 14nm chips already in the table, and you can find even more on ark.intel.com where Intel used to quote transistor count and die size – until 14nm node, when Intel's *actual* transistor density (tr. count / die area) became inferior vs competition (which may well be the reason why they all but stopped publishing transistor counts.)

By the way, unsourced data from this Twitter account contradicts not only what's known about Intel's density, but AMD's as well: the owner claims 286.4 mm² and 6.24 bn tr. for Ryzen 5000, so density is 21.79 Mtr./mm², but if you check the table, Ryzen 7 5800H is enlisted as 180 mm² and 10.7 bn transistors, so the density is 59.44 Mtr./mm², in line with density calculated from numbers reported by AMD for Zen 3 at ISSCC (included in the table as well). So the numbers posted by that Twitter account need some proving to do, to begin with, and I would strongly suggest that unless you can deliver solid proof that the Twitter account's numbers are correct, and AMD's and Intel's are not, you should refrain from further harassment and edit-warring.

And since I take it you cannot find original Intel source either after a month of trying (since May 16 when I asked for it), what's the real reason of such insistence on including this unsourced and contradictory claim from that Twitter account and other media? I'm asking you in earnest; I could have reported you straight away for harassment, edit-warring and persistent violation of Wikipedia policies on sourcing and verifiability along with my suspicions of your previous record, but I'll give you a chance to explain it.

I have a couple of versions, but I want to hear it from you rather than guess. If you love tech and just don't care about correctness and civility norms, why not start a blog instead and repost there whatever unsourced crap you see floating from one Twitter account to another about Rocket Lake or anything else? Or am I wrong and you simply enjoy harassing people, and showing up on Wikipedia with incorrect or disruptive edits is nothing more than one of many available means to that end for you?

But either way, Wikipedia is neither a garbage collector, nor it is for harassers like you – is it clear to you? And in case you haven't been diagnosed, take my good advice: do consult a psychiatrist (not a psychologist) – such severe personal attacks on people who are polite with you can be a sign of a serious mental deviation, and people who are *forced* to interact with you — and this is obviously not limited to Wikipedia – don't have to tolerate the way you open your mouth at them. 188.66.32.223 (talk) 13:03, 16 June 2023 (UTC)[reply]

I think both of you are going to be disappointed once this little back-and-forth attracts the attention of an administrator. You both need to read WP:CIVIL and WP:NPA a few times. Stop edit warring. Maybe start a Wikipedia:Requests for comment. How is this such a contentious topic? --Onorem (talk) 15:55, 16 June 2023 (UTC)[reply]

The twitter post from Intel employee >>> whatever your assumption was, it was your opinion, and you don't have better source to claim it false so stop edit or I have to report administration to permanently block your IP!! 67.173.134.210 (talk) 03:22, 20 June 2023 (UTC)[reply]

I would not consider that Twitter reference a reliable source, as it falls into the categories of user-generated content and self-published source. Ideally, you would want to use an independent secondary source (such as a magazine, review) over a primary or self-published one. — AP 499D25 (talk) 04:35, 21 June 2023 (UTC)[reply]

That person works in intel so the source is rather believable. Magazine wise there are few website also published the transistor count as well but someone said website like hexus is third party provider and the source is not reliable and that which was why I didn’t list it. 2601:243:600:48C0:E0DA:380A:9A8A:5DDE (talk) 07:00, 22 June 2023 (UTC)[reply]

Plus the transistor count on coffee lake , Ryzen 5 and comet lake are extremely accurate so 6 billion transistor is rather believable due to such big size. 2601:243:600:48C0:E0DA:380A:9A8A:5DDE (talk) 07:02, 22 June 2023 (UTC)[reply]

188.66.32.223 needs to be blocked permanently as he had list of record by deleted list with his own opinion! that is violated NPA and need to be stop!!! 67.173.134.210 (talk) 03:26, 20 June 2023 (UTC)[reply]

One last thing, where did you get that "Intel (10-16 Mtr./mm²)" from? any source to back up your claim? or you're just making things? but either way without any reliable source any further editing will result your ban! 67.173.134.210 (talk) 06:56, 20 June 2023 (UTC)[reply]

The citation is from before the process was in use and I don't even see the number on the linked page. This article suggests the true number is more in the realm of 183m https://en.wikipedia.org/wiki/3_nm_process 192.147.66.4 (talk) 14:45, 1 February 2024 (UTC)[reply]

While I still cannot find the 314730000 number, I do notice something on the talk page of that article. It states that the A17 pro doesn't even use the N3 process, so the density number of 183m might be wrong too. 192.147.66.4 (talk) 14:52, 1 February 2024 (UTC)[reply]