Intel rebrands process names, provides process roadmap

[porina]

13 minutes ago, Kisai said:

Renaming it to be "equal to the competition" is the same "P rating" when the clock speed's on Intel were measured in Mhz while AMD and Cyrix tried to sell technically worse chips by PR numbers instead of by clock speed. It doesn't change the fact that the chips are in fact 50% worse by clock rate. 

PR came about because Intel at the time with Pentium 4 chose to optimise their design for clock speed, because that number was misunderstood by buyers as faster is better, not taking into consideration IPC. AMD CPUs of the time did more work per clock, so having a lower clock was a marketing disadvantage, resolved through PR. 

 

13 minutes ago, Kisai said:

If your chips are on a 14nm process and your competition is on 7nm, 5nm, 3nm, etc You're clearly behind

Intel 14nm is going against TSMC 7nm, that much is a given and following on from that, it can be said that Intel is some years behind TSMC at this moment in process technology. However Intel 14nm will not go against TSMC 5nm, which AMD haven't productised yet. If rumours are correct Intel will get TSMC 3nm before AMD does, although this latest roadmap to me brings that more into doubt territory.

 

13 minutes ago, Kisai said:

To the consumer though, the transistor density is a more meaningful number regardless of the chip's physical size and process node.

It may seem a better measure but it too has its flaws. Processes can be tweaked for performance or efficiency and vary considerably from that. The density will also depend a lot on the logic mix you implement. At the end of the day, we are still trying to use one number to compare things that can't be reduced to one number without narrowing the definition so far it becomes useless for other use cases.

 

13 minutes ago, Kisai said:

Yet, by adopting these completely arbitrary names, they are just pulling the same marketing stunt they accused their competition of pulling. Two wrongs don't make a right.

It's the best Intel can do within their control and restore some balance to the Force.

 

13 minutes ago, Kisai said:

Sure, nerds on the internet might care about the specific naming like you copy-pasted from sources, but if you walk into a store and know one part is 14nm and another is 7nm, and the same price while having twice as many cores, that's a no-brainer that the 7nm is the better value chip, even if your loyalties lie with the 14nm manufacturer. 

Value is determined in large part by pricing, which is not directly tied to technology at all. Which is the better value if you were looking to build a cheap general purpose 6 core desktop? 11400F or 5600X? It isn't the 7nm one.

[leadeater]

6 minutes ago, porina said:

Value is determined in large part by pricing, which is not directly tied to technology at all. Which is the better value if you were looking to build a cheap general purpose 6 core desktop? 11400F or 5600X? It isn't the 7nm one.

When it comes time for me to buy a CPU or GPU as much as I enjoy the technical discussions of these details the things I do not factor are, chip density, process node size and process node company. For example I did not buy a 6800 XT over a 3080 because it was TSMC, TSMC 7nm or whatever density it is (I legit don't know what it is).

[mr moose]

51 minutes ago, leadeater said:

When it comes time for me to buy a CPU or GPU as much as I enjoy the technical discussions of these details the things I do not factor are, chip density, process node size and process node company. For example I did not buy a 6800 XT over a 3080 because it was TSMC, TSMC 7nm or whatever density it is (I legit don't know what it is).

Most consumers I have ever met, sold to, discussed with or even just listened to buy based on the size of their wallet and what the salesperson tells them.  Very very rarely do technical specs play a major role.    I've heard customers claim to buy only Intel because AMD are "slow" (whatever that means and specs be damned),  and buy only the latest gen they can afford because "it's always better than the last gen" and that's on top of useless sales claims like More Gigahurts means faster internet.      The whole node thing is basically for tech enthusiasts and stock investors.   Buyers don't give a crap and if they did then it's because they don't know what they are buying and likely won't even notice after the fact.

[Arika]

16 minutes ago, mr moose said:

The whole node thing is basically for tech enthusiasts and stock investors.   Buyers don't give a crap and if they did then it's because they don't know what they are buying and likely won't even notice after the fact.

it's funny, i asked a tech-illiterate friend;

"without knowing anything anything about a computer processor, what do you think is better; 14nm or 7nm"

i didn't say nanometer, i just said "nm".

they said 14nm because it's a bigger number. so even if they started slapping it on the box, people see bigger number = better

[Mark Kaine]

On 7/27/2021 at 1:57 AM, porina said:

If you think the nm is directly related to some physical aspect that can be compared, please point out exactly what it is.

 

The nearest thing you might come up with to have a comparison across different fabs is transistor density

This could be wrong, but the qbits guy on yt (forgot his name, i think an italian) said this refers to the width between  the "gates". So if that "width" isnt "2 or 5 or 7 etc nm" then this is simply false advertising…

 

He also said it would be better to actually  measure this in atoms, which I agree with, and it would ironically be way more impressive. 

[Kisai]

3 hours ago, mr moose said:

Most consumers I have ever met, sold to, discussed with or even just listened to buy based on the size of their wallet and what the salesperson tells them.  Very very rarely do technical specs play a major role.    

This is how you get people buying 12GB 3060's instead of 10GB 3080's because they saw the VRAM number and the lower price and went "wow, bigger number, lower price, what a deal"

 

Like, no joke, this happened with someone I know. They paid me to travel 200 miles to go help them assemble it after they got in over their head with the pc build.

 

Realistically, this is how people keep buying garbage for "gaming" setups. They ask people who don't really know, and don't really care. The consequences are some really bizarre choices.

 

[LAwLz]

54 minutes ago, Mark Kaine said:

This could be wrong, but the qbits guy on yt (forgot his name, i think an italian) said this refers to the width between  the "gates". So if that "width" isnt "2 or 5 or 7 etc nm" then this is simply false advertising…

 

He also said it would be better to actually  measure this in atoms, which I agree with, and it would ironically be way more impressive. 

The problem with that is that transistors are not really two dimensional anymore, and the way the transistors are structured matters too. 

 

Comparing process nodes is kind of like comparing Intel vs AMD cores. You will not get a good and accurate comparison between the two if you only look at a single number, like MHz, number of cores, gate pitch, etc. It's too complicated to boil it down to a single number. 

[mr moose]

10 hours ago, LAwLz said:

The problem with that is that transistors are not really two dimensional anymore, and the way the transistors are structured matters too. 

 

Comparing process nodes is kind of like comparing Intel vs AMD cores. You will not get a good and accurate comparison between the two if you only look at a single number, like MHz, number of cores, gate pitch, etc. It's too complicated to boil it down to a single number. 

Like trying to compare mosfets to bipolar, or even mosfets to mosfets given base materials are as important as the number of them.  

 

At the end of the day there is only one single metric that matters to these companies (and most consumers) and the $ is not a technical spec nor does it indicate universal performance.

[heickelrrx]

when the whole industry pulling naming crap, you'll forced to pull the same otherwise the marketing side will got burn

[ravenshrike]

On 7/26/2021 at 11:08 PM, Sarra said:

There, I fixed it for you.

Are we sure that isn't Intel's processor naming strategy?

On 7/27/2021 at 3:41 AM, LAwLz said:

A bunch of unfunny jokes over and over about 14nm+++++.

Now Intel is changing to the same nonsense marketing numbers as the rest of the industry and what do they get? A bunch of shit for just lowering their numbers like 7nm suddenly being called 4.

Intel got shit about 14nm++++++++++++++++++ because their initial projections called for their 10nm process to be in general use by 2015 and it just kept slipping farther and farther back. It had nothing to do with the competition's naming schemes.

[Forbidden Wafer]

23 minutes ago, ravenshrike said:

Are we sure that isn't Intel's processor naming strategy?

Intel got shit about 14nm++++++++++++++++++ because their initial projections called for their 10nm process to be in general use by 2015 and it just kept slipping farther and farther back. It had nothing to do with the competition's naming schemes.

Exactly. We're in 2021 and most "desirable"* products are still on 14nm.

 

*: desirable in the sense of being mainstream CPUs. I don't mind NAND, chipsets, MCUs, modems, network gear, FPGAs, etc using older nodes.

[Rauten]

18 hours ago, Mark Kaine said:

This could be wrong, but the qbits guy on yt (forgot his name, i think an italian) said this refers to the width between  the "gates". So if that "width" isnt "2 or 5 or 7 etc nm" then this is simply false advertising…

 

He also said it would be better to actually  measure this in atoms, which I agree with, and it would ironically be way more impressive. 

Are you perhaps referring to Andrea Morello?

[IkeaGnome]

On 7/28/2021 at 2:35 AM, Arika S said:

it's funny, i asked a tech-illiterate friend;

"without knowing anything anything about a computer processor, what do you think is better; 14nm or 7nm"

i didn't say nanometer, i just said "nm".

they said 14nm because it's a bigger number. so even if they started slapping it on the box, people see bigger number = better

"i wanna say 14 cause its a higher number but im gonna say 7 cause i dont trust myself?" 

I asked my fiancee the same thing. 

[Mister Woof]

When it comes to the consumer-facing naming schemes, they need to reinvent themselves. Just too many digits.

 

i9-12900K is just.....too much.

 

maybe go from "i" to "x" and start over or something. idk.

[IkeaGnome]

12 minutes ago, Mister Woof said:

When it comes to the consumer-facing naming schemes, they need to reinvent themselves. Just too many digits.

 

i9-12900K is just.....too much.

 

maybe go from "i" to "x" and start over or something. idk.

Realistically, with core counts and all that being somewhat stable over generations for i3, i5, and i7 they could just "i07k" next generation, "i170k" etc. Or even smarter, since the next boards will be B650, X670 etc, "i670k and i690k" Then move on to 7 with the chipset. If motherboards use the same chipset, "i675k" could work well like "AM3+" did. 

[Mister Woof]

1 minute ago, IkeaGnome said:

Realistically, with core counts and all that being somewhat stable over generations for i3, i5, and i7 they could just "i07k" next generation, "i170k" etc. Or even smarter, since the next boards will be B650, X670 etc, "i670k and i690k" Then move on to 7 with the chipset. If motherboards use the same chipset, "i675k" could work well like "AM3+" did. 

Just anything. these model names are ridiculous and the longer you make it the harder it is for consumers to understand, and simple is king.

[IkeaGnome]

Just now, Mister Woof said:

Just anything. these model names are ridiculous and the longer you make it the harder it is for consumers to understand, and simple is king.

That was kind of my thought process with that. 6xx series chipset works with 6xx series CPUs. The relevant core count is in the 2nd digit. You could even change it to i640k for a 4 core. it gets the relevant information across, I think it's fairly simple to understand too. If not, it's at least easy to explain. 

[Mister Woof]

20 minutes ago, IkeaGnome said:

That was kind of my thought process with that. 6xx series chipset works with 6xx series CPUs. The relevant core count is in the 2nd digit. You could even change it to i640k for a 4 core. it gets the relevant information across, I think it's fairly simple to understand too. If not, it's at least easy to explain. 

One thing I hope they ditch immediately is the tier system being based on frequency again. OC is already kind of weak lately, but the way they have just been using binning for their laptop 11th gen and desktop 11th gen sucks.

 

For example, 

 

i7-11700k

i9-11900k 

 

both are overclockable but the price premium on the i9 is no actual functional gain. They have the same configuration. you are getting almost nothing by going to the i9. You don't even get any new features.

 

Same goes for the mobile 11th gen:

 

i5-1135G7

i5-1145G7

i7-1165G7

i7-1185G7

 

ALL of these are 4 core hyperthreaded CPUs with varying degrees of clock speeds and minimal iGPU differences. 

 

There's no logical reason why the i7s should be called i7s. They should all be i5s

 

At least with the past decade, the i3/i5/i7/i9 designation usually meant some configurational difference.

[YoungBlade]

@Mister Woof@IkeaGnome Since this architecture has such a fundamental shift, why not just drop the "Core i" naming scheme?

 

Intel came up with Pentium for the successor to the i486 by using the Greek number for 5 and the Latin suffix "-ium." Unfortunately, Duodecium doesn't sound very good, and is way too long.

 

However, Pentium is probably the most famous of Intel's product lines, and in recent years, they've used Pentium Gold and Pentium Silver to distinguish between desktop chips and low-power chips. The thing that makes 12th gen distinctive is the fact that it uses both desktop cores and low-power cores, both "gold" cores and "silver" cores.

 

Electrum is a naturally occurring alloy of gold and silver, used by both the Greeks and Romans. Unfortunately, the crypto wallet "Electrum" exists, so "Electrum" itself is out as a brand name. So I messed around with the word across various languages, and most of them just use an approximation of the Latin word, but Italian sounds a bit different with "elettro" so I thought of "Eletrium," but then it occurred to me to change one more letter, and I ended up with "Elitrium," so a portmanteau between "electrum" and "elite."

 

I think the 3, 5, 7, 9 thing is pretty well ingrained in people's heads by this point, so no need to redo that, and I think the suffixes are alright, but Intel can start all over again with numbers in the hundreds, just like for the Core i series.

 

So that would be Intel Elitrium 3 100, Elitrium 5 400, Elitrium 7 700K, Elitrium 9 900K, etc.

 

What do you think? Better or worse than the current system?

[IkeaGnome]

33 minutes ago, Mister Woof said:

One thing I hope they ditch immediately is the tier system being based on frequency again. OC is already kind of weak lately, but the way they have just been using binning for their laptop 11th gen and desktop 11th gen sucks.

-snip-

both are overclockable but the price premium on the i9 is no actual functional gain. They have the same configuration. you are getting almost nothing by going to the i9. You don't even get any new features.

 

Same goes for the mobile 11th gen:

-snip-

At least with the past decade, the i3/i5/i7/i9 designation usually meant some configurational difference.

Maybe then it's time for them to realize that the desktop naming scheme doesn't work on desktop and is even worse for mobile. 

I don't know nearly enough about mobile CPUs to even begin thinking of something. 

12 minutes ago, YoungBlade said:

@Mister Woof@IkeaGnome Since this architecture has such a fundamental shift, why not just drop the "Core i" naming scheme?

-snip-

So that would be Intel Elitrium 3 100, Elitrium 5 400, Elitrium 7 700K, Elitrium 9 900K, etc.

 

What do you think? Better or worse than the current system?

i9-11900k is 8 syllables, ignoring the Intel, Elitrium 3 100 is 7. The 9 900k would be 8(Shortened to E 9 900k is 4). Still just as many as a simple sentence (The fox jumped over the moon, is 7). It also goes back into the whole thing with i5, 7, 9 etc not having much to do with core count or features. it just describes where they feel it should fall in the line up and price wise and not how it actually performs or what it does. 

I'm cherry picking here, but the i5 does better than the i7 in this benchmark?

 

[YoungBlade]

19 minutes ago, IkeaGnome said:

i9-11900k is 8 syllables, ignoring the Intel, Elitrium 3 100 is 7. The 9 900k would be 8(Shortened to E 9 900k is 4). Still just as many as a simple sentence (The fox jumped over the moon, is 7). It also goes back into the whole thing with i5, 7, 9 etc not having much to do with core count or features. it just describes where they feel it should fall in the line up and price wise and not how it actually performs or what it does. 

I'm cherry picking here, but the i5 does better than the i7 in this benchmark?

I believe "i nine e-lev-en nine hun-dred kay" has 9 syllables, not 8. And the i9 12900k drops back to 7 because "twelve" is a one syllable word. And before, people said things like "i se-ven eigh-ty seven hun-dred kay" which is 10 syllables.

I'm not sure syllable count is an issue for naming.

 

The issue with putting stats in the names is that some stats don't matter. Even core counts. A Core 2 Quad is going to perform worse than a modern Pentium G, despite both being from Intel and the slower one having double the cores.

 

Product names are supposed to convey which one is better or appropriate for a certain use case. It's a good idea to have some flexibility for marketing teams to work with.

 

And, yes, that is cherry picking, because as indicated, the i7 is on an older microcode update, so it isn't a fair comparison.

[Mister Woof]

29 minutes ago, IkeaGnome said:

Maybe then it's time for them to realize that the desktop naming scheme doesn't work on desktop and is even worse for mobile. 

I don't know nearly enough about mobile CPUs to even begin thinking of something. 

i9-11900k is 8 syllables, ignoring the Intel, Elitrium 3 100 is 7. The 9 900k would be 8(Shortened to E 9 900k is 4). Still just as many as a simple sentence (The fox jumped over the moon, is 7). It also goes back into the whole thing with i5, 7, 9 etc not having much to do with core count or features. it just describes where they feel it should fall in the line up and price wise and not how it actually performs or what it does. 

I'm cherry picking here, but the i5 does better than the i7 in this benchmark?

 

I think that benchmark is a little wonky, feels like it has some FPS cap that it bounces off of 

 

I wouldn't call the i5-11600k vs the i7-11700k a win really. Average FPS is the same but the 11700k wins easily in 1% and 0.1% lows. I wouldn't consider anything within 2.5% a win, and .5fps is 0.2% change. 

[LAwLz]

Big news! @leadeater @porina

Intel has announced that Qualcomm will be a customer for their 20A manufacturing process in 2024.

 

This confirms a few things. 

1) Intel's fab business is actually gaining traction. It probably won't be the same as last time where it died out without any major players getting onboard. 

2) Intel's 2nm node is on track for 2024. That is really promising if it plays out. 

3) Qualcomm are confident in Intel's ability to execute on this. This is not just Intel going "our processors will use this in X amount of years" as before. Now third parties are also confident in it. 

 

It's still pretty light on details on what Qualcomm has ordered, but it will probably be an SoC for phones or laptops. And probably not a bargain bin SoC either since it will be on a cutting edge node. 

 

https://www.intel.com/content/www/us/en/newsroom/news/intel-accelerates-process-packaging-innovations.html

 

[Kisai]

2 hours ago, YoungBlade said:

@Mister Woof@IkeaGnome Since this architecture has such a fundamental shift, why not just drop the "Core i" naming scheme?

 

Intel came up with Pentium for the successor to the i486 by using the Greek number for 5 and the Latin suffix "-ium." Unfortunately, Duodecium doesn't sound very good, and is way too long.

 

Intel had called the chips 8088/8086 80186, 80286, 80386, 80486 etc up to that point. The problem is that you can't trademark a number, and Cyrix and AMD were also calling their chips 286, 386, 486, etc, and they kept doing so by calling their 486-socket compatible 586, and pentium-socket compatible 686.

 

The current naming scheme is just silly.

https://www.intel.ca/content/www/ca/en/processors/processor-numbers.html

Quote

For the majority of Intel® processors, the final three digits of the product number are the SKU. SKUs are generally assigned in the order in which processors in that generation and product line are developed. A higher SKU within otherwise-identical processor brands and generations will generally have more features. However, SKU numbers are not recommended for comparison across different generations or product lines.

...

SKUs with a “G” consist of a two-digit generation indicator (“10” or "11"), then a two-digit SKU, followed by a two-character alphanumeric suffix. The suffix indicates the level of graphics offered by the processor; higher numbers (e.g., G7) indicate improved graphics performance relative to lower numbers (e.g., G1).

So basically all but the last three digits matter on an intel chip.

 

As for the "Pentium" and Celeron brand, it should have been retired.

 

But Intel still uses the 80xxx part numbers. The current Rocket lake is 80708. Comet Lake is 80701. Every CPU before Comet Lake were 806xx parts. Those part numbers were also not a meaningful way to compare performance.

 

The Core Duo and Core i series are trademarks, competition can't call their own chips by that. That's why they continue to use them. They're "Intel Core" brands. Now perhaps Intel should ditch the "core" brand with the 12th gen and finally getting off the 14nm process node. 

 

[porina]

11 hours ago, LAwLz said:

Big news! @leadeater @porina

Intel has announced that Qualcomm will be a customer for their 20A manufacturing process in 2024.

I had saw that elsewhere but stopped reading since I had no interest in Qualcomm  Good point on the implications of such a deal.