AMD: Moore's Law's end is near

Page 7 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

R0H1T

Platinum Member
Jan 12, 2013
2,583
164
106
A lot of Americans just don't get what a social democracy is and hence just call it socialism and think their social security thingy is the greatest thing on earth.

To give that perspective, were I live if you are unemployed you get 80% percent of your last salary for 2 years. There is an upper limit of course, but it's quite high, about $120'000 yearly salary. (eg you get a maximum of $240'000) And there are conditions, like you must have worked here for at certain time frame and you must send out a certain amount of resumes and attend courses and so on.
And when you do work, a certain percentage of your salary is automatically removed and put into that "pot". That's why it is called "unemployment insurance".

And about monopoly:

Train system in UK has become complete crap. It's less secure, less convenient and more expensive since the monopoly of the gov was removed. It is by far worse than were I live and at the same time more expensive and it is already very expensive here...

Basic infrastructure should IMHO never ever be privatized. It will just end up worse. It's much better to pay more taxes instead because government will not hire 100 additional high paid manager that just BS around all day...if you privatize, your money will go into pockets of such managers and not better infrastructure.
The other guy might have exaggerated about socialism but why does anybody think that a commoditized market like x86 is gonna be better off without AMD ? Businesses always prefer higher margins over more revenues, this will never change especially if they aren't in the red ! If there were a single oil producer or car manufacturer, display maker et al in the world would they lower prices like ever ? If its that hard to figure out then economics is not you cup of tea, not singling you out here but any analogy with utility/services is moot because this is a consumer driven market & with only one major supplier left(Intel) there are not gonna be price cuts just like the HDD market which btw is a virtual duopoly & we all know what happened over there :whiste:
 

ShintaiDK

Lifer
Apr 22, 2012
20,378
146
106
A lot of Americans just don't get what a social democracy is and hence just call it socialism and think their social security thingy is the greatest thing on earth.

To give that perspective, were I live if you are unemployed you get 80% percent of your last salary for 2 years. There is an upper limit of course, but it's quite high, about $120'000 yearly salary. (eg you get a maximum of $240'000) And there are conditions, like you must have worked here for at certain time frame and you must send out a certain amount of resumes and attend courses and so on.
And when you do work, a certain percentage of your salary is automatically removed and put into that "pot". That's why it is called "unemployment insurance".

And about monopoly:

Train system in UK has become complete crap. It's less secure, less convenient and more expensive since the monopoly of the gov was removed. It is by far worse than were I live and at the same time more expensive and it is already very expensive here...

Basic infrastructure should IMHO never ever be privatized. It will just end up worse. It's much better to pay more taxes instead because government will not hire 100 additional high paid manager that just BS around all day...if you privatize, your money will go into pockets of such managers and not better infrastructure.

Well americans have been indoctrined for ages due to the mccarthyism. But still, you would expect it to have worn off by now and a reality check would have struck. Its still a mystery to me why so many americans can still live in this alternative reality.

We had the same thing in Denmark. Power got privatised, divided and electrcity base cost before taxes rose 150% in 6 years. Trains too, and since its never been so bad and expensive either. I would say something like the order of 20/80 between monopoly/free market is the right thing. Ofcource different segments needs different ratios. But if you took everything under one, thats about the optimal ratio. And any society critical services should simply only be national monopolies.

People also forget how many times competition works against innovation. Simply due to the risk factor that any company must take to introduce radically new things. And this is why governments also needs to subsidize new changes, since the free market is unable to do it on its own.
 
Last edited:

ShintaiDK

Lifer
Apr 22, 2012
20,378
146
106
The other guy might have exaggerated about socialism but why does anybody think that a commoditized market like x86 is gonna be better off without AMD ? Businesses always prefer higher margins over more revenues, this will never change especially if they aren't in the red ! If there were a single oil producer or car manufacturer, display maker et al in the world would they lower prices like ever ? If its that hard to figure out then economics is not you cup of tea, not singling you out here but any analogy with utility/services is moot because this is a consumer driven market & with only one major supplier left(Intel) there are not gonna be price cuts just like the HDD market which btw is a virtual duopoly & we all know what happened over there :whiste:

You should examine the business model. Then you would know why higher revenue is needed. Nomatter how you turn and bend it. There simply wont be room for anyone else in the endgame.

14nm alone will kill of half of the current semiconductor companies. Simply because they cant afford the chip designs.

Its like advocating for lots of VIA companies. And think the world is better of.
 

krumme

Diamond Member
Oct 9, 2009
5,956
1,596
136
A lot of Americans just don't get what a social democracy is and hence just call it socialism and think their social security thingy is the greatest thing on earth.

To give that perspective, were I live if you are unemployed you get 80% percent of your last salary for 2 years. There is an upper limit of course, but it's quite high, about $120'000 yearly salary. (eg you get a maximum of $240'000) And there are conditions, like you must have worked here for at certain time frame and you must send out a certain amount of resumes and attend courses and so on.
And when you do work, a certain percentage of your salary is automatically removed and put into that "pot". That's why it is called "unemployment insurance".

And about monopoly:

Train system in UK has become complete crap. It's less secure, less convenient and more expensive since the monopoly of the gov was removed. It is by far worse than were I live and at the same time more expensive and it is already very expensive here...

Basic infrastructure should IMHO never ever be privatized. It will just end up worse. It's much better to pay more taxes instead because government will not hire 100 additional high paid manager that just BS around all day...if you privatize, your money will go into pockets of such managers and not better infrastructure.

Privatizing natural monopolys, like sewer systems and trains, is ofcourse outright stupid. It gets from bad to very bad.

But again Intel is not comparable to a train and a sewer system. There might be a few people thinking that nationalising Intel will be good. Well, the efficiency of such a thinking stops when you have build all the sewers and trainrails you need, the rest is waiting for the Lada.

AMD is not going out of the market. The IP and competence is har higher than market cap but are tight in contracts with primarily ATIC. And the moment it goes to liquidation, those values will surely find their way. Perhaps not in the name of AMD, but then in another form. So Intel will continue to face competition here - that still have a noticible influence on the price.

Secondly ARM is comming. So we are in a mature market, where competition is starting to catch up. The desktop have stagneted, but the mobile is moving.

The competition we see on mobile is only good for us consumers. Intel entering will only be a benefit for us, because Qualcomm and Samsung will be gearing up for their R&D here for sure.

I am pretty sure Qualcomm is really pushing their portfolio now, and we will see a steady at least 1 year upgrade. They have a great way before they reach just near Intel competence of handling complex cpu and new productions techniques, but the question remains how fast they can get close to this level. I think jaguar will show just how much difference there is between the class of Intel/AMD designing and then the rest. But Qualcomm is just improving steadily at a fast pace and just executing. It just looks really well managed.
 
Last edited:

ShintaiDK

Lifer
Apr 22, 2012
20,378
146
106
AMD is not going out of the market. The IP and competence is har higher than market cap but are tight in contracts with primarily ATIC. And the moment it goes to liquidation, those values will surely find their way. Perhaps not in the name of AMD, but then in another form. So Intel will continue to face competition here - that still have a noticible influence on the price.

Uhm, AMD is actually valued at around 700 million$ in a free sale. Far from the market cap if you were to sell off. AMD doesnt contain much besides its GPU division in terms of value. Also why they are shut out of the debt market.
 
Last edited:
Apr 20, 2008
10,067
990
126
It's not a scientific "law" per se. It's at best a scientific theory based off of factual and semi-predictable information.
 
Apr 20, 2008
10,067
990
126
Uhm, AMD is actually valued at around 700 million$ in a free sale. Far from the market cap if you were to sell off. AMD doesnt contain much besides its GPU division in terms of value. Also why they are shut out of the debt market.

Are you drunk? Amd's patent portfolio is incredibly valuable.
 

Grooveriding

Diamond Member
Dec 25, 2008
9,147
1,330
126
A lot of Americans just don't get what a social democracy is and hence just call it socialism and think their social security thingy is the greatest thing on earth.

To give that perspective, were I live if you are unemployed you get 80% percent of your last salary for 2 years. There is an upper limit of course, but it's quite high, about $120'000 yearly salary. (eg you get a maximum of $240'000) And there are conditions, like you must have worked here for at certain time frame and you must send out a certain amount of resumes and attend courses and so on.
And when you do work, a certain percentage of your salary is automatically removed and put into that "pot". That's why it is called "unemployment insurance".

The misunderstanding is partly attributable to programming via government and media. 'Socialism' has been bred into a scary word with overtones of all sorts of insidious and scary meanings, hearkening back to the days of the Cold War. It's been cultivated into a go-to label with the simple 'Socialism' to prevent any sort of progression for public benefit in some of the currently profit-oriented industries that should be publicly funded and concerned with finances only as far as efficient use of them.


And about monopoly:

Train system in UK has become complete crap. It's less secure, less convenient and more expensive since the monopoly of the gov was removed. It is by far worse than were I live and at the same time more expensive and it is already very expensive here...

Basic infrastructure should IMHO never ever be privatized. It will just end up worse. It's much better to pay more taxes instead because government will not hire 100 additional high paid manager that just BS around all day...if you privatize, your money will go into pockets of such managers and not better infrastructure.

Also agree with this. Garbage collection was privatized for half my city just recently, service level went down the toilet overnight, and continues to be a mess close to a year later. Electricity was privatized about ten years ago, guess what, prices have gone up. Where I live, and the region I live in, we have too much electricity, to the point that sometimes we don't just sell it but we end up giving it away for free to take it off our grid.

When it comes to basic human needs services, it's better to allow for some unionised public sector waste. It seems when you have more employees with less work to do collectively, versus less employees with much more of the same work to do collectively; the larger workforce delivers a better level of service. Who'd of thunk it ? People are happier when they're not whittled down to the bare minimum feasible to get a job done.

Some sectors are simply better served by not being motivated by profit. The reality of this can be seen in contrasting the overwhelming majority of the developed world, which have universal health care or a two-tier system - versus the very small minority that does not have those systems. Cliffs: universal health care / two-tier system countries dominate the overall quality of care delivered to the country as a whole compared to countries with solely profit-driven health care. The US being one of the very few on the short list of countries without guaranteed access to care, and even with the greatest overall expenditures on health care, still falls short of near every other first-world nation in quality of care delivered. As good example of any of what happens when you monetize a system that should be a human right. Spend more than anyone else, end up getting less than most. But somewhere a small few are lining their pockets, so it's gravy.
 

krumme

Diamond Member
Oct 9, 2009
5,956
1,596
136
Uhm, AMD is actually valued at around 700 million$ in a free sale. Far from the market cap if you were to sell off. AMD doesnt contain much besides its GPU division in terms of value. Also why they are shut out of the debt market.

And who says, and under what conditions?
- if i thought buying AMD share was a good idea i would do that. I dont.

Try to understand what i write, there is value, far beyond what can be meassured the tradional way, and restrictions that comes to a halt the moment the company is liquidated. The value is then very low, for existing shareholders because thats the conditions. But then the situation is new. Predicting the outcome is the difficult. But i am sure Intel does not want that, because its exactly unpredictable.

But my point was, under all circumstances the competence and IP will come into play aging, and generate competition. And probably stronger than it is now, because right now AMD is loosing money.
 

R0H1T

Platinum Member
Jan 12, 2013
2,583
164
106
You should examine the business model. Then you would know why higher revenue is needed. Nomatter how you turn and bend it. There simply wont be room for anyone else in the endgame.

14nm alone will kill of half of the current semiconductor companies. Simply because they cant afford the chip designs.

Its like advocating for lots of VIA companies. And think the world is better of.
The business model you refer to encompasses the x86 market, which as I've repeatedly said is constantly shrinking & no the cloud is dead ~ its only enterprises that are adding to their existing capacity while on the consumer front people with "smart devices" are finding it extremely hard to upload their large volumes of data online(hence local storage i.e. flash) with the ever increasing data rates worldwide, but since it is shrinking even if AMD went under there would be no let off as avg prices will increase because of a monopoly that Intel is gonna exploit !

Outside of this bastion, Intel will not be able to sustain their node shrinks any further if they don't have a foothold in mobile space. So their single largest revenue stream is declining, ever increasing R&D costs, shrinking margins overall, no mobile strategy & then finally this ! No amount of glossing over the facts will turn the tide in Intel's favor, not anytime soon that is !
 
Last edited:

ShintaiDK

Lifer
Apr 22, 2012
20,378
146
106
The business model you refer to encompasses the x86 market, which as I've repeatedly said is constantly shrinking & no the cloud is dead ~ its only enterprises that are adding to their capacity addition while on the consumer front people with "smart devices" are finding it extremely hard to upload their large volumes of data online with the ever increasing data rates worldwide, but since it is shrinking even if AMD went under there would be no let off as avg prices will increase because of a monopoly that Intel is gonna exploit !

Outside of this bastion, Intel will not be able to sustain their node shrinks any further if they don't have a foothold in mobile space. So their single largest revenue stream is declining, ever increasing R&D costs, shrinking margins overall, no mobile strategy & then finally this ! No amount of glossing over the facts will turn the tide in Intel's favor, not anytime soon that is !

You do know the ARM segment shrinks as much as x86? And that trend will most likely turn for both again very soon, since its not related to the segment itself as a cause. And no, prices wouldnt increase.

TSMC is becoming a foundry monopoly (If we looka way from intel.) for the same reason. The cost is simply too great for competition.
 
Last edited:

R0H1T

Platinum Member
Jan 12, 2013
2,583
164
106
You do know ARM shrinks as much as x86? And that trend will most likely turn for both again very soon. And no, prices wouldnt increase.

TSMC is becoming a foundry monopoly (If we looka way from intel.) for the same reason. The cost is simply too great for competition.
Yes but that will only happen what 5~10yrs from now maybe when the next gen of processors(like graphene) are born. However even then there will only be a flat or marginal decrease in the overall mobile/tablet market because they are status symbols, fashion accessories for alot of people unlike computers of today.

What I'm saying here is that the x86 exists more because of necessity while the market that ARM rules is for all practical purposes one of choice, hence the higher volumes lower margins, therefore this semiconductor landscape will witness a sea change in the next decade or so ! Whether Intel will still be a major player by then, only time will tell, but they will have to change/adapt or they'll be greatly marginalized like AMD is nowadays !
 
Last edited:

mrmt

Diamond Member
Aug 18, 2012
3,974
0
76
But who in their fucking sane minds thinks Intel will lower their prices? Come on. Do we have to get the old mails from Otellini again. They are running a business, they work each day to get monopoly like advantages. And you are asking them to lower prices?
I dont think this discussion will bring us somewhere. What the f is the purpose?


As IDC pointed out, I just mentioned a mathematical possibility that contradicts a generic sentence where VirtualLarry deemed good monopolies nonexistent. I didn't refer to it as something good, desirable, extendable for all industries and not even the most probable outcome for the CPU industry. It is just a possibility for a few cases.

If Intel were to get a true monopoly, I for one believe that prices *might* even go down in the short term, but whatever gains in the short term will be reversed in the long term when Intel develop chips specifically for a monopoly market.

But this question about Intel monopoly is rather academic, as is R0H1T scenario where you have a true duopoly milking the market. The former because there will be a huge clash between ARM and x86 manufacturers for the years to come, and the latter because you can't have a sustainable duopoly when one agent gets 85% of the market share and 100% of the profits.

Where some folks see trouble because poor AMD won't be able to pull another miracle, I see the entire ARM ecosystem pushing Intel to its limits again. If someone can give Intel a run on its money, it's Qualcomm & co, not AMD with Derpdozer, JFAMDs, 140W+ chips, etc.
 

Charles Kozierok

Elite Member
May 14, 2012
6,762
1
0
At 90nm you could have a gate length that was 90nm, or one that was much smaller. Gate length is a range, it can be any size between a minimum length and a maximum length.

Prior to 0.35um the minimum value allowed for the gate length was defined to be the process node. At 0.42um (a half-node) the minimum allowed gate length was 0.42um. If it had been 0.41um or 0.43um then the 0.42um node would have been labeled as either 0.41um or 0.43um respectively.

That all changed at 0.35um when the minimum gate length broke with tradition and was 0.30um. Of course you could still use 0.35um gate lengths on the 0.35um node, but the node label was not 0.30um, it was called 0.35um and that was that.

Thanks for the detailed response as usual, IDC.

I'm confused, though. You seem to be saying that the node labels stopped reflecting reality, but because manufacturers were making features smaller than the nominal node size, not bigger. From a marketing perspective, if I were making a 0.30 μm chip I'd want to call it that, not 0.35.

And wouldn't the issue these days be more about cheating in the other direction, for that same reason? That is, doesn't Intel sound more "leading edge" if they talk about 14 nm for their next node even if, say, the smallest features are only 18 nm?
 
Aug 11, 2008
10,451
642
126
It seems that I'm usually going against the flow, but I actually have a windows8 tablet that I prefer to the Android and Apple tablets that we have around here. Not so sure that I'm comfortable yet with W8 on my desktop, but on a tablet it makes sense - at least for me with my requirements.

I also think win 8 is definitely a step in the right direction for mobile. It is just unfortunate that they made it such a slap in the face to desktop users by stubbornly omitting the start menu.
 

MisterMac

Senior member
Sep 16, 2011
777
0
0
I also think win 8 is definitely a step in the right direction for mobile. It is just unfortunate that they made it such a slap in the face to desktop users by stubbornly omitting the start menu.

I thought about this alot - and as me i think the same.
It's just shit.


Now i tried thinking of it from Microsoft's View - and that is the desktop\laptop scenario is shrinking and eventually it might be only those dual tops - where you attach a keyboard and have a touch screen as well.

It's more about starting to shift the general populace and trying to gain more grounds in the consumer segment where probably a extremely large percentage of windows installs are pirated.

Corporate will be corporate.
But lets face it that - there's a large percentage of people not paying for consumer windows.

And that % will grow as emerging markets take more hold and also discover cheap "computing" tablets.


It's their longterm playbook imho.
 

Idontcare

Elite Member
Oct 10, 1999
21,110
64
91
Thanks for the detailed response as usual, IDC.

I'm confused, though. You seem to be saying that the node labels stopped reflecting reality, but because manufacturers were making features smaller than the nominal node size, not bigger. From a marketing perspective, if I were making a 0.30 μm chip I'd want to call it that, not 0.35.

And wouldn't the issue these days be more about cheating in the other direction, for that same reason? That is, doesn't Intel sound more "leading edge" if they talk about 14 nm for their next node even if, say, the smallest features are only 18 nm?

I think it is confusing because you (quite naturally) want the node label to mean something mathematical/numerical, and the reality of node labels doesn't fit neatly into that expectation. I was there too at one point, it really was jarring to have my expectations turned upside down.

But if you took away all the numbers and assigned the nodes streetnames or named them after fruit then you'd realize the label never mattered, ever, even in the pre-0.35um are when the physical dimensions meant something related to the node label.

What matters, all that matters, regardless whether the node is labeled "32nm HKMG w/SOI" or "22nm 3D FinFet" or "Maple Node" or "Pineapple Node", are the electrical properties (the parametrics) of the various components (xtors, resistors, inductors, capacitors, etc) that have been implemented in the "node".

It is the node-on-node scaling of the electrical properties that matter most. If you can't or don't scale the electrical properties then the area scaling itself is nearly worthless (you do save on production cost). You can't shrink a chip and have it perform any better than its predecessor if the electrical parameters of the node involved in the shrink are not scaling.

So at 0.35um the historic purpose of physical scaling, which was to improve the electrical properties, stop being a key enabler. You couldn't just shrink the xtor by 10% physically and get a 10% better transistor. You had to do other stuff, materials science engineering, to actually get the transistor to be 10% better electrically (including physically shrinking the gate length more than one would otherwise expect they needed to).

Physical scaling was, and is, still tied to metrics of cost control, production expenses. And physical dimensions still factor in the electrical properties of the devices (a 22nm 3D FinFet is its existing dimensions and not larger for electrical reasons).

This is why there can be so much variation between companies which claim to have the same process node. No two nodes are the same, even when they have the same node label. The difference is seen at every level, the physical dimensions are not the same, nor are electrical parametrics. But they will still use the same node label. (and that was true even before 0.35um)
 

Exophase

Diamond Member
Apr 19, 2012
4,439
9
81
What matters, all that matters, regardless whether the node is labeled "32nm HKMG w/SOI" or "22nm 3D FinFet" or "Maple Node" or "Pineapple Node", are the electrical properties (the parametrics) of the various components (xtors, resistors, inductors, capacitors, etc) that have been implemented in the "node".

It is the node-on-node scaling of the electrical properties that matter most. If you can't or don't scale the electrical properties then the area scaling itself is nearly worthless (you do save on production cost). You can't shrink a chip and have it perform any better than its predecessor if the electrical parameters of the node involved in the shrink are not scaling.

So you think that being able to fit more transistors in the same area only benefits by decreasing costs? There are a lot of optimizations you can make to perf or perf/W that only involve using more transistors - GPUs and other very parallel devices are an excellent example of this. Since there are limits to die size that are both financial and physical smaller transistors means more transistors for all intents and purposes.
 

Idontcare

Elite Member
Oct 10, 1999
21,110
64
91
So you think that being able to fit more transistors in the same area only benefits by decreasing costs? There are a lot of optimizations you can make to perf or perf/W that only involve using more transistors - GPUs and other very parallel devices are an excellent example of this. Since there are limits to die size that are both financial and physical smaller transistors means more transistors for all intents and purposes.

That is not what I wrote.

But even in your example (GPUs) you are not correct as the xtor density of GPUs is not that of the allowed maximum for the node (on any node).

Before Intel's 22nm 3D xtors, xtors are/were two-dimensional. Gate length captures only one of those two dimensions, the second dimension (referred to as the gate width) is many times larger than the length and in most circuits is nowhere near the minimum allowed by the process node.

Sram is about the only time you see circuits physically laid out in which the xtor density actually approaches the maximum allowed by the node's design rules. Everything else is physically larger for reasons that run counter to your stated post.

But all of that is irrelevant in terms of my post above to which you were responding as you did not read what I wrote. If you simply physically scale a transistor without scaling the electrical characteristics in kind then you gain nothing.

If you want to just tread water (same clockspeed, same power usage, etc) while making the chip smaller in area then you must increase drive current and reduce leakage parameters commensurate to the dimensional scaling.

If you don't do any of that (which is what I was discussing) then physically shrinking the chip actually hurts the chip's electrical parametrics.

And you know that already, even if you didn't realize you did, because the transistor density of a CPU is significantly less than that of a GPU which in turn is significantly less than that of the highest density layouts allowed for sram on the same node. There is a reason why the CPU is not physically shrunk to the same density as the sram circuits even though the node's design rules can accommodate it.

If the SB design engineers could have improved on SB by using physically smaller transistors (without improved electrical parametrics) then they would have, they were available but went unused. The transistor widths used in the logic circuits of my 2600k cannot be narrower if the chip was to operate at its target clockspeed, power consumption, and desired operational lifetime (reliability). It had to be as big as it is for good reasons.
 

Exophase

Diamond Member
Apr 19, 2012
4,439
9
81
That is not what I wrote.

You wrote that the non-electrical parameters in a node shrink are nearly useless for anything but reducing costs. How should I have interpreted that?

Spare me this condescending crap about how I didn't read what you wrote. I'm getting tired of people saying this here because they don't want to stand by what they said. And I made it clear what I'm responding to - you literally said that if you can't scale the electrical properties then the node is nearly worthless. You also said that if you only make the transistors physically smaller without also improving their electrical characteristics you gain nothing.

But even in your example (GPUs) you are not correct as the xtor density of GPUs is not that of the allowed maximum for the node (on any node).

And that is not what I wrote.

It doesn't matter if GPUs or anything else are using the "allowed maximum" density for the node, that has nothing to do with what I said. All that's relevant is that the die shrink allows more transistors in the same area under similar constraints. The electrical performance of those transistors could be identical - same speed, same power consumption, etc and the GPU would still have higher perf and perf/W by virtue of having more transistors.

All of this stuff about relative transistor density of SRAM vs logic and CPU vs GPU isn't relevant to the point I was making.
 

Charles Kozierok

Elite Member
May 14, 2012
6,762
1
0
IDC: I understand the technicals. I don't understand the marketing. :)

Companies usually cheat to make themselves look better (Exhibit A: Bulldozer cores.) So I could see a company saying their chip is 0.35 when it's really only 0.40.

But I would think that if the actual minimum feature size was 0.30, the marketing people would be falling over themselves to call it a "0.30 chip".

I was just surprised by your graph, because I assumed that right now the whole issue is "they say they've shrunk it down to X, but they've only shrunk it down to 1.2X" (or whatever).
 

Homeles

Platinum Member
Dec 9, 2011
2,580
0
0
You wrote that the non-electrical parameters in a node shrink are nearly useless for anything but reducing costs. How should I have interpreted that?

Spare me this condescending crap about how I didn't read what you wrote. I'm getting tired of people saying this here because they don't want to stand by what they said. And I made it clear what I'm responding to - you literally said that if you can't scale the electrical properties then the node is nearly worthless. You also said that if you only make the transistors physically smaller without also improving their electrical characteristics you gain nothing.
You're probably getting frustrated because you legitimately do have an issue with putting words in people's mouths. It's not that they're not standing behind what they said (and really, what's the shame in changing one's mind anyways?); it's that you are misinterpreting it.
 

Idontcare

Elite Member
Oct 10, 1999
21,110
64
91
You wrote that the non-electrical parameters in a node shrink are nearly useless for anything but reducing costs. How should I have interpreted that?

Spare me this condescending crap about how I didn't read what you wrote. I'm getting tired of people saying this here because they don't want to stand by what they said. And I made it clear what I'm responding to - you literally said that if you can't scale the electrical properties then the node is nearly worthless. You also said that if you only make the transistors physically smaller without also improving their electrical characteristics you gain nothing.



And that is not what I wrote.

It doesn't matter if GPUs or anything else are using the "allowed maximum" density for the node, that has nothing to do with what I said. All that's relevant is that the die shrink allows more transistors in the same area under similar constraints. The electrical performance of those transistors could be identical - same speed, same power consumption, etc and the GPU would still have higher perf and perf/W by virtue of having more transistors.

All of this stuff about relative transistor density of SRAM vs logic and CPU vs GPU isn't relevant to the point I was making.

If you want to assume I don't know what I am talking about, and that you are the expert here when it comes to process node development, then I will neither argue with you nor stand in your way.

You seem quite intent on debunking what I write, but to do that you must first misinterpret what I am writing and build yourself a few strawmen to then tear down. Have at it, I have nothing to lose if you wish to remain ignorant.