Jump to content
Search In
  • More options...
Find results that contain...
Find results in...

Intel rebrands process names, provides process roadmap

AMD marketing plan:
"We have new Zen architecture. How do we name future ones?"
"Zen
Zen+
Zen 2
Zen 3
Zen 3+
Zen 4
Zen 5
Easy to follow along, the + moniker means it's not a completely new revision, just a modification"

"Full marks, good job"

Intel marketing plan:
"We need to make our nm numbers fancy. How do we do this?"
"10nm = 7

7nm = 4

7nm = 3
2nm = 20A"
"It sounds futuristic, therefore we do this plan"


Cyrix marketing plan:
"*static noises*"

REFRESH BEFORE RESPOND, I EDITED MY POST

Please ensure your signature complies with the following restrictions:
1. There must be a 17 in your signature somewhere.

2. No pictures under any circumstances, ever ever.

aHR0cHM6Ly9kcml2ZS5nb29nbGUuY29tL2ZpbGUvZC8xOWZWOVNidFp1d2FLUVUxaUY5bjg3MjVHelcxd2x2SWkvdmlldz91c3A9c2hhcmluZw==

Spoiler
Spoiler
Spoiler
Spoiler
Spoiler
Spoiler
Spoiler
Spoiler
Spoiler
Spoiler
Spoiler
Spoiler
Spoiler
Spoiler
Spoiler
Spoiler
Spoiler
Spoiler
Spoiler
Spoiler
Spoiler
Spoiler
Spoiler

aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

I do not know how to interpret social interaction.

Send help.

Link to post
Share on other sites
6 minutes ago, FakeKGB said:

AMD marketing plan:

Intel marketing plan:

Architecture names =/= Fabrication process

 

AMD also don't get to name their fab process, TSMC does.

🌲🌲🌲

Judge the product by its own merits, not by the Company that created it.

 

Don't dilute <good thing> by always trying to focus on, and drag conversation back to, <bad thing>.

🌲🌲🌲

Link to post
Share on other sites
2 hours ago, ouroesa said:

Agreed but what happens when they start stacking chips? IIRC AMD is already starting to do this. Or are you purely referring to the production architecture?

3D is still density on a 2D plane.

 

Eg if you have a 10mm x 10mm die and it has 1 billion transistors, and you stack two more dies of the exact same size on top, that's equal to 30mm x 10mm, and thus 3 billion transistors.

 

Hence comparing two different processes should be compared by the number of transistors that actually can be made to fit. The i9-11900k is 290mm^2, and 6 billion transistors (or 20.6m transistors per mm.) Ryzen 5950X is 10.39 billion transistors across three dies of different sizes and process nodes (2x 80.7 mm 7nm and 1x 125 mm 12nm.) So effectively Ryzen's 161.4 7nm + 125mm 12nm = (286.4)mm, or a density of 36.3m transistors per mm^2.

 

Link to post
Share on other sites
4 hours ago, DexterSmythe said:

Fair point. Intel and AMD do measure it differently and there is no industry standard for "nm"s.

It's not Intel nor AMD that measure it differently, not like you are implying. AMD is not a foundry, they don't name any nodes at all. AMD uses 3rd party foundry to manufacture their products, which ever foundry they go with has named the process.

 

Intel has been the most stead fast to something that has been more representative of past node sizes when talking about nm, everyone else went #marketing and used small numbers to "win".

 

There is actually a more standardized nm measurement you can use, which tried to get traction but based on this announcement even Intel has given up on that.

 

But basically the worst industry offenders in this area was literally everyone but Intel. Of course naturally since Intel were the odd ones out and ended up losing the leading edge due to delays perception changed and now everyone thinks all but Intel were not a problem 🤦‍♂️

 

Quote

Gargini, who is chairman of the IEEE International Roadmap for Devices and Systems (IRDS), proposed in April that the industry “return to reality” by adopting a three-number metric that combines contacted gate pitch (G), metal pitch (M), and, crucially for future chips, the number of layers, or tiers, of devices on the chip (T). 

 

Quote

The IRDS road map shows that the coming 5-nm chips have a contacted gate pitch of 48 nm, a metal pitch of 36nm, and a single tier—making the metric G48M36T1. It doesn’t exactly roll off the tongue, but it does convey much more useful information than “5-nm node.”

https://spectrum.ieee.org/semiconductors/devices/a-better-way-to-measure-progress-in-semiconductors

 

And now you can see why more accurate and meaningful node names didn't take off. Also note the 10x difference between any of the important measurements and "5nm".

 

Neither are these uniform across an actual chip either.

Link to post
Share on other sites
2 hours ago, FakeKGB said:

Cyrix marketing plan:
"*gargling peanutbutter noises*"

There, I fixed it for you. 😄

"Is X and Y going to Bottleneck?" YES. YES, IT FREAKING IS. NOT having a bottleneck is, most likely, impossible. Will it negatively effect your gaming experience? Probably not.

Link to post
Share on other sites

Maybe it's just me, but uuhh this is weird, random and kinda pointless. Not because it's a bad idea, but because it feels like "oh yes, we're going to parity. Only no, we're gonna do whatever the F we want".

 

"Hey guys, we're Intel, and we've decided to match our process naming with the rest of the industry!"

 

Well thats great, that'll make it easier to compare how each fab is advan...

 

"... and our 2nm will instead be named 20A!"

 

*20A-neurysms*

Link to post
Share on other sites

I don't care what they name it (Even if its wild) Just don't sit on your ass. Bring that competition and make AMD sweat. Because as a consumer, that only means good things for me, lol.

Leonidas

Specs: Ryzen 7 3700x | AMD 6800 XT Midnight Black | MSI B550 Gaming Plus | Corsair Dominator CL16 3200 MHz 32GB | be quiet! Silent Base 802

Monitors: Acer Nitro XV272U 1440p 170Hz IPS / Acer KG240 1080p 144Hz TN 

Link to post
Share on other sites

I think this thread is a great example of why Intel has to do this. Even "tech enthusiasts" are clueless regarding processing nodes. 

 

It was Samsung and TSMC that started the nonsense of just naming their nodes lower and lower numbers despite it not being related to anything and Intel kept the fairly sane names. What did they get in return? 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.

Link to post
Share on other sites
7 minutes ago, LAwLz said:

I think this thread is a great example of why Intel has to do this. Even "tech enthusiasts" are clueless regarding processing nodes. 

 

It was Samsung and TSMC that started the nonsense of just naming their nodes lower and lower numbers despite it not being related to anything and Intel kept the fairly sane names. What did they get in return? 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.

Exactly why this whole "nm" nonsense just doesn't make sense when we're doing comparisons between wholly different manufacturing processes. 

 

10nm SF in terms of density is more akin to TSMC 7nm, but because 7 is lower than 10, and apparently, lower nm is automatically better, then we're in a situation where CPUs are memed because their transistors nanometers aren't lower...

The Workhorse (AMD-powered custom desktop)

CPU: AMD Ryzen 7 3700X | GPU: MSI X Trio GeForce RTX 2070S | RAM: XPG Spectrix D60G 32GB DDR4-3200 | Storage: 512GB XPG SX8200P + 2TB 7200RPM Seagate Barracuda Compute | OS: Microsoft Windows 10 Pro

 

The Portable Station (Intel-powered Lenovo Yoga Slim 7i)

CPU: Intel Core i5 1135G7 | GPU: Intel Iris Xe 80CU | RAM: 16GB LPDDR4X-4267 | Storage: 512GB PCIe SSD | OS: Microsoft Windows 10 Home

 

The Communicator (Exynos-powered Samsung Galaxy Note8)

SoC: Exynos 8895 | GPU: ARM Mali G71 MP20 | RAM: 6GB LPDDR4 | Storage: 64GB internal + 128GB microSD | Display: 6.3" 1440p "Infinity Display" AMOLED | OS: Android Pie 9.0 w/ OneUI

Link to post
Share on other sites
1 hour ago, Rauten said:

"... and our 2nm will instead be named 20A!"

At some point the ongoing unit modifiers will become impractical and you have to look at using the next one. We have an example in this when we transitioned from micrometres to nanometres. My 1st CPU was the 486, which on looking up was apparently made on 1µm through to 600nm processes. More surprising to me was that it was apparently made for nearly 20 years. Someone must have really liked them.

 

Now you might wonder why not wait until we go below 1nm before switching. I think the take away here is that we're going to be slowly chipping our way to "1nm" with fractional values coming into play. Also we do have angstroms as a somewhat well known number modifier (optical systems, scientific imaging), and it provides a stepping stone without going all the way to picometers. Say for future node, would you prefer 1.2nm, 12A, 1200pm? Picometers will be overkill as we're going to have to use different physics to progress to lower values of that. For context, the Van der Waals radius of a hydrogen atom is 120pm, and for silicon it is 210pm. We're getting really close to counting individual atoms now.

TV Gaming system: Asus B560M-A, i7-11700k, Scythe Fuma 2, Corsair Vengeance Pro RGB 3200@213342x16GB, MSI 3070 Gaming Trio X, EVGA Supernova G2L 850W, Anidees Ai Crystal, Samsung 980 Pro 2TB, LG OLED55B9PLA 4k120 G-Sync Compatible
Streaming system: Asus X299 TUF mark 2, i9-7920X, Noctua D15, Corsair Vengeance LPX RGB 3000 8x8GB, Gigabyte 2070, Corsair HX1000i, GameMax Abyss, Samsung 970 Evo 500GB, Crucial BX500 1TB, BenQ XL2411 1080p144 + HP LP2475w 1200p60
Desktop Gaming system (to be retired): Asrock Z370 Pro4, i7-8086k, Noctua D15, G.Skill Ripjaws V 3200 2x8GB, Asus Strix 1080 Ti, NZXT E850 PSU, Cooler Master MasterBox 5, Optane 900p 280GB, Crucial MX200 1TB, Sandisk 960GB, Acer Predator XB241YU 1440p144 G-sync

Former Main system (to be retired): Asus Maximus VIII Hero, i7-6700k, Noctua D14, G.Skill Ripjaws 4 3333@2133 4x4GB, GTX 1650, Corsair HX750i, In Win 303 NVIDIA, Samsung SM951 512GB, WD Blue 1TB, Acer RT280k 4k60 FreeSync [link]
Gaming laptop: Lenovo Legion, 5800H, DDR4 3200C22 2x8GB, RTX 3070, 512 GB SSD, 165 Hz IPS panel


 

Link to post
Share on other sites
6 hours ago, Arika S said:

Architecture names =/= Fabrication process

 

AMD also don't get to name their fab process, TSMC does.

Yeah, and they just measure it in nanometers. Intel's issue here is that their nodes are larger in measured nanometers while actually being the same as TSMC's smaller ones (Intel's 10nm is the same as TSMC's 7nm). Given they are also a chip foundry, why not just lie and say it's the same instead of making up even dumber units!? It's not like consumers are going to check CPU's under electron microscope and measure how many nanometers transistors are in size...

AMD Ryzen 7 5800X | ASUS Strix X570-E | G.Skill 32GB 3733MHz CL16 | PALIT RTX 3080 10GB GamingPro | Samsung 850 Pro 2TB | Seagate Barracuda 8TB | Sound Blaster AE-9 MUSES Edition | Altec Lansing MX5021 Nichicon/MUSES Edition

Link to post
Share on other sites

So Intel made their naming even more confusing, their product sku naming since 10th gen is bad enough.

Intel being the market leader and owning their own fabs shouldn't need to make up fancier sounding marketing for their process nodes.

Link to post
Share on other sites
36 minutes ago, porina said:

At some point the ongoing unit modifiers will become impractical and you have to look at using the next one...*snip*...Now you might wonder why not wait until we go below 1nm before switching. I think the take away here is that we're going to be slowly chipping our way to "1nm" with fractional values coming into play.*more snips*

Yes, obviously, but I was under the impression that TSMC, and apparently IBM too? already announced their 2nm process.

So if Intel's intention was to try to fit into some form of naming parity with the others, the 20A just... uugghh, why? So close and yet so far.

It's like Intel know their naming is basically a meme in the tech community by now and are just going along with it, to be honest.

Link to post
Share on other sites

I really don't understand some of the responses here. It seems we're at a consensus where we agree the recent past and present "nm" numbers (not limited to Intel) are meaningless in a technical sense, so why the hatred for the new numbers?

 

If you don't know or care about process technology, none of this matters.

If you have heard of process technology but never looked into it deeply, this is the group that will benefit most from this naming change because they were not sufficiently informed to understand similar looking numbers were not directly comparable.

If you have heard of process technology but do have a good understanding of it, this wont matter.

 

2 minutes ago, Rauten said:

Yes, obviously, but I was under the impression that TSMC, and apparently IBM too? already announced their 2nm process.

It has to happen at some point, so someone will be first to make that move. Doesn't matter which fab is doing it, it is still several years/generations away. Things may or may not also change at some point in future before we see those products.

TV Gaming system: Asus B560M-A, i7-11700k, Scythe Fuma 2, Corsair Vengeance Pro RGB 3200@213342x16GB, MSI 3070 Gaming Trio X, EVGA Supernova G2L 850W, Anidees Ai Crystal, Samsung 980 Pro 2TB, LG OLED55B9PLA 4k120 G-Sync Compatible
Streaming system: Asus X299 TUF mark 2, i9-7920X, Noctua D15, Corsair Vengeance LPX RGB 3000 8x8GB, Gigabyte 2070, Corsair HX1000i, GameMax Abyss, Samsung 970 Evo 500GB, Crucial BX500 1TB, BenQ XL2411 1080p144 + HP LP2475w 1200p60
Desktop Gaming system (to be retired): Asrock Z370 Pro4, i7-8086k, Noctua D15, G.Skill Ripjaws V 3200 2x8GB, Asus Strix 1080 Ti, NZXT E850 PSU, Cooler Master MasterBox 5, Optane 900p 280GB, Crucial MX200 1TB, Sandisk 960GB, Acer Predator XB241YU 1440p144 G-sync

Former Main system (to be retired): Asus Maximus VIII Hero, i7-6700k, Noctua D14, G.Skill Ripjaws 4 3333@2133 4x4GB, GTX 1650, Corsair HX750i, In Win 303 NVIDIA, Samsung SM951 512GB, WD Blue 1TB, Acer RT280k 4k60 FreeSync [link]
Gaming laptop: Lenovo Legion, 5800H, DDR4 3200C22 2x8GB, RTX 3070, 512 GB SSD, 165 Hz IPS panel


 

Link to post
Share on other sites
14 hours ago, WereCat said:

Full use = anime villain equivalent of "You can't defeat me!"

Increased use = "Now you made me use my full power!"

image.png.2118914859345de3d8d7315d51ca70c5.png

"Put as much effort into your question as you'd expect someone to give in an answer"- @Princess Luna

Make sure to Quote posts or tag the person with @[username] so they know you responded to them!

 RGB Build Post 2019 --- Rainbow 🦆 2020 --- Velka 5 V2.0 Build 2021

Purple Build Post ---  Blue Build Post --- Blue Build Post 2018 --- Project ITNOS

CPU i7-4790k    Motherboard Gigabyte Z97N-WIFI    RAM G.Skill Sniper DDR3 1866mhz    GPU EVGA GTX1080Ti FTW3    Case Corsair 380T   

Storage Samsung EVO 250GB, Samsung EVO 1TB, WD Black 3TB, WD Black 5TB    PSU Corsair CX550M    Cooling Cryorig H7 with NF-A12x25

Link to post
Share on other sites
5 minutes ago, TVwazhere said:

 

I could have gone the rest of my life without remembering that powerful shaggy was a thing that happened

🌲🌲🌲

Judge the product by its own merits, not by the Company that created it.

 

Don't dilute <good thing> by always trying to focus on, and drag conversation back to, <bad thing>.

🌲🌲🌲

Link to post
Share on other sites
8 minutes ago, Arika S said:

I could have gone the rest of my life without remembering that powerful shaggy was a thing that happened

Me internally: "Dont just respond to this with another meme, you need to set an example for community memb-"

Also me:

TVwazunhelpful.png.c0b586c9f3dd0db641e538a1311d4e74.png

"Put as much effort into your question as you'd expect someone to give in an answer"- @Princess Luna

Make sure to Quote posts or tag the person with @[username] so they know you responded to them!

 RGB Build Post 2019 --- Rainbow 🦆 2020 --- Velka 5 V2.0 Build 2021

Purple Build Post ---  Blue Build Post --- Blue Build Post 2018 --- Project ITNOS

CPU i7-4790k    Motherboard Gigabyte Z97N-WIFI    RAM G.Skill Sniper DDR3 1866mhz    GPU EVGA GTX1080Ti FTW3    Case Corsair 380T   

Storage Samsung EVO 250GB, Samsung EVO 1TB, WD Black 3TB, WD Black 5TB    PSU Corsair CX550M    Cooling Cryorig H7 with NF-A12x25

Link to post
Share on other sites
19 hours ago, Kilrah said:

Intel 4 = "Full use of EUV"

Intel 3 = "Increased use of EUV"

 

So they can use more than "Full"? 🤣

EUV doesn't exist to all the layers of the chip. Considering foveros, I belive that Intel 3 will bring EUV to the other 3D layers

Link to post
Share on other sites
8 hours ago, LAwLz said:

I think this thread is a great example of why Intel has to do this. Even "tech enthusiasts" are clueless regarding processing nodes. 

 

It was Samsung and TSMC that started the nonsense of just naming their nodes lower and lower numbers despite it not being related to anything and Intel kept the fairly sane names. What did they get in return? 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.

+1

 

If you aren't willing to look into the entire industry, then just don't barf out the nm the processor is on. Nodes almost never matter for the customer. Intel was ahead of AMD for nearly 2-3 years in many performance metrics, despite being on an inferior node from 2015-2016.

 

 

TL;DR Nodes aren't a good way to measure chips.

 

 

Link to post
Share on other sites
18 hours ago, DexterSmythe said:

Intel 4 being 7nm seems more than a little misleading.

I mean, presumably Intel 7 will be 7nm, 4 4nm, 3 3nm, 20A 2nm, etc, and they're just dropping 10nm ESF or something like that.

 

Then again it's not as if these naming schemes are forced to follow any logic anyway so 🤷‍♂️ 

Link to post
Share on other sites
15 hours ago, porina said:

I really don't understand some of the responses here. It seems we're at a consensus where we agree the recent past and present "nm" numbers (not limited to Intel) are meaningless in a technical sense, so why the hatred for the new numbers?

 

 

Because they don't mean anything at all. May as well call them Banana, Potato, Ponana.

 

At least when CPU's were measured by Mhz alone, you knew a 100Mhz processor was slower than a 133Mhz processor, even if it was by Cyrix or AMD. That completely went out the window when we got into hyperthreading, and multi-cores.

 

The process node, at least the "nm" gave a way to compare to equivalent cpu generations. Despite what people may insist about intel's 14nm being equal to TSMC's 7nm by some wishful thinking. Ultimately the transistor denstity is the only number that matters, and that will still not give a meaningful comparison if most of those transistors are cache and not ALU on one node and reversed on the other.

 

 

Link to post
Share on other sites
4 hours ago, Kisai said:

Ultimately the transistor denstity is the only number that matters

Not really honestly, a hugely dense process and a chip that actually utilizes it could easily get spanked by another chip that employs say a 10% less dense process utilization that is able to clock 25% higher. Two x86 chip designs on the exact same process will not perform or have the same characteristics because how the chip is structured matters too.

 

It's actually exceedingly hard to compare fab nodes by comparing end product chips, Intel 14nm CPUs out performing AMD Zen, Zen+ and in multiple factors Zen 2 is a prime example of that very issue.

 

That is why this is considered the gold standard to compare nodes, not just pure density.

Quote

Gargini, who is chairman of the IEEE International Roadmap for Devices and Systems (IRDS), proposed in April that the industry “return to reality” by adopting a three-number metric that combines contacted gate pitch (G), metal pitch (M), and, crucially for future chips, the number of layers, or tiers, of devices on the chip (T). 

 

"7nm" class

image.thumb.png.586b69fc6d0fed085403b962c9db5908.png

 

TSMC N7FF = G54M40T1

TSMC N7FF+ = G54M40T1

TSMC N7P = G54M40T1

Samsung 7LPP = G54M46T1

Intel 7 (10ESF) = G54M36T1

 

As you can already see even the gold standard fails to fully shows the differences between the TSMC 7nm processes and doesn't take in to account DUV vs EUV and layer usage constraints etc.

 

Intel 14nm

image.png.78454e4af16810e1f50bf71cc9460755.pngimage.png.da8cf1f61204ad75d91559d1fc81fb30.png

 

Intel 14nm = G70M52T1

Intel 14nm+ = G70M52T1

Intel 14nm++ = G84M52T1

 

What is interesting here is that for 14nm++ technically it should be worse process but it's not when factoring in end products that utilized it. Another example of how the gold standard isn't able to  fully portray how good one node is to another.

 

Smaller isn't automatically better, more dense isn't automatically better. It's complicated, too complicated and far too imprecise to then go off and compare end product chips produced on them to try and show which process is better, such an undertaking is going to be misleading. The best process for the application is going to get the most wide use across the industry, that is the best and most simple way to figure out which is considered by the industry as the best. Problem there is until now Intel process were exclusive to Intel.

Link to post
Share on other sites
5 hours ago, leadeater said:

Not really honestly, a hugely dense process and a chip that actually utilizes it could easily get spanked by another chip that employs say a 10% less dense process utilization that is able to clock 25% higher. Two x86 chip designs on the exact same process will not perform or have the same characteristics because how the chip is structured matters too.

I wasn't saying that 7nm always better than 14nm here. I was saying that from a marketing point of view, trying to sell your 14nm process with 6 billion transistors against AMD's TSMC 7nm process with 10.5 billion and saying yours is better is just a marketing lie at worst, and fraud at best. 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. 

 

If your chips are on a 14nm process and your competition is on 7nm, 5nm, 3nm, etc You're clearly behind, no matter how "Advanced" your process node is. Now maybe it's dishonest for TSMC to use 7nm when that only measures one aspect of the process, but without seeing Intel use the same measurements, it's not a meaningful comparison.

 

IDRS-Scaling-1

https://www.extremetech.com/computing/309889-tsmc-starts-development-on-2nm-process-node

 

To the consumer though, the transistor density is a more meaningful number regardless of the chip's physical size and process node. That may not always translate to better performance, but that allows you to compare two different chips from different manufacturers, even different uses (eg GPU and CPU vs SoC's) before production parts are available. Once the parts are out, there are programs like geekbench that can directly compare two different chips on different platforms by performance.

 

5 hours ago, leadeater said:

 

Smaller isn't automatically better, more dense isn't automatically better. It's complicated, too complicated and far too imprecise to then go off and compare end product chips produced on them to try and show which process is better, such an undertaking is going to be misleading. The best process for the application is going to get the most wide use across the industry, that is the best and most simple way to figure out which is considered by the industry as the best. Problem there is until now Intel process were exclusive to Intel.

 

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.

 

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. 

 

Hell, I wound up buying the Rocket Lake part in the end because the high end AMD chips were NOT available to purchase, and I felt I might get some use from Rocket Lake's GPU in my media production process over AMD's CPU's without an iGPU. But if I were just pure-gaming, I would have just picked up the AMD part. That's twice now I wound up buying the Intel hardware over the clearly superior AMD hardware at the time because the AMD hardware wasn't available.

 

I didn't need to know the node size to know that, I've seen the benchmarks, but if I was planning another new build I would not be putting a hot 14nm chip in it. The same applies to GPU's.

 

 

Link to post
Share on other sites
6 hours ago, Kisai said:

Because they don't mean anything at all. May as well call them Banana, Potato, Ponana.

They changed their number which was not directly related to a physical parameter to a different number which is not directly related to a physical number, but closer to those of TSMC's number which is not directly related to a physical parameter. Either they all mean something, or they all don't mean anything.

 

6 hours ago, Kisai said:

At least when CPU's were measured by Mhz alone, you knew a 100Mhz processor was slower than a 133Mhz processor, even if it was by Cyrix or AMD. That completely went out the window when we got into hyperthreading, and multi-cores.

Up to 486 that might have been the case, since AMD 486 was a literal clone of Intel 486, but differences started to appear after that. IPC mattered even in those days. I don't recall what Cyrix CPUs were like of that era.

 

6 hours ago, Kisai said:

Despite what people may insist about intel's 14nm being equal to TSMC's 7nm by some wishful thinking.

Has anyone ever seriously said that? If so, I wouldn't pay any attention to them in future. It is generally agreed that Intel 10nm (or some version of it) is ball park comparable to TSMC 7nm. Intel have now formalised that with formerly known 10ESF rebranded to 7.

TV Gaming system: Asus B560M-A, i7-11700k, Scythe Fuma 2, Corsair Vengeance Pro RGB 3200@213342x16GB, MSI 3070 Gaming Trio X, EVGA Supernova G2L 850W, Anidees Ai Crystal, Samsung 980 Pro 2TB, LG OLED55B9PLA 4k120 G-Sync Compatible
Streaming system: Asus X299 TUF mark 2, i9-7920X, Noctua D15, Corsair Vengeance LPX RGB 3000 8x8GB, Gigabyte 2070, Corsair HX1000i, GameMax Abyss, Samsung 970 Evo 500GB, Crucial BX500 1TB, BenQ XL2411 1080p144 + HP LP2475w 1200p60
Desktop Gaming system (to be retired): Asrock Z370 Pro4, i7-8086k, Noctua D15, G.Skill Ripjaws V 3200 2x8GB, Asus Strix 1080 Ti, NZXT E850 PSU, Cooler Master MasterBox 5, Optane 900p 280GB, Crucial MX200 1TB, Sandisk 960GB, Acer Predator XB241YU 1440p144 G-sync

Former Main system (to be retired): Asus Maximus VIII Hero, i7-6700k, Noctua D14, G.Skill Ripjaws 4 3333@2133 4x4GB, GTX 1650, Corsair HX750i, In Win 303 NVIDIA, Samsung SM951 512GB, WD Blue 1TB, Acer RT280k 4k60 FreeSync [link]
Gaming laptop: Lenovo Legion, 5800H, DDR4 3200C22 2x8GB, RTX 3070, 512 GB SSD, 165 Hz IPS panel


 

Link to post
Share on other sites
16 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. That may not always translate to better performance, but that allows you to compare two different chips from different manufacturers, even different uses (eg GPU and CPU vs SoC's) before production parts are available. Once the parts are out, there are programs like geekbench that can directly compare two different chips on different platforms by performance.

But it really doesn't, all it tells you is how dense one chip is to another. You need to understand that that is both not actually that important to the degree you might think however more importantly just because a node can be that dense a design could choose never to utilize that maximum density anywhere at all in the chip.

 

Process A could have a node design rating of 95 MTr/mmand Process B could have 87 MTr/mm2 but a chip produce on Process A may not utilize more than 80 MTr/mm2 anywhere in their chip design, regardless of Logic vs SRAM etc.

 

Edit:

What's really really important here is that you cannot go in the reverse direction from a chip design and it's product performance backwards to prove which node is better than another. The only time that can be even close to a valid exercise if the exact same chip design ported over to another process with the required design rule changes but the architecture of the chip itself remaining the same. Anything other than this specific situation is not a valid way to compare process nodes, and that only applies to that chip application type. Process A might be better for a high power chip, Process B might have better lower power efficiency and Process C might have higher peak frequency but lower peak power than Process A. It depends on application just as much as the node itself.

Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
Newegg

×