14nm 6th Time Over: Intel Readies 10-core "Comet Lake" Die to Preempt "Zen 2" AM4

Page 13 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

maddie

Diamond Member
Jul 18, 2010
4,723
4,628
136
I think in the past we had a wider variation in what each chip would do at what voltage, so clocks were conservative to cover them all, but it was not intentionally so. It was not a made up thing to create overclockable chips.
Not to stay within more honest power limits? As now seems abandoned.
 

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
Not to stay within more honest power limits? As now seems abandoned.

Engineers are just good at extracting more of the CPUs than they could before. More reliable circuits allow higher frequencies at the same voltage without failing. Turbo modes and advanced power management allow clocks that are generally higher because it can now be dependent on the load.

Imagine back in 1998 with the Pentium III 500MHz. It runs 500MHz when idle, it runs 500MHz when gaming, rendering, word processing.

Now if you want to spec the CPU for a power rating what do you do?
-Take the worst case scenario.

What if 9,999 out of 10,000 applications can run at 550MHz, while all 10,000 applications can run at 500MHz?
-You sell it at 500MHz.

What if 525MHz allows operation flawlessly for 5 years, but runs into problems later, and 500MHz can run nearly indefinitely.
-You sell it at 500MHz.

1 bad news counteracts hundreds of good news(and goodwill) thus as a business you try to be conservative as you can.

The Power Control Unit allows precise control of power used, so it does not deviate from what the company wants. Turbo mode allows setting different frequencies for different scenarios.

What does not change is fundamental limits. 5GHz is roughly such limit, since 5.5GHz on water has been true for 15 years. That means 5GHz CPU has no overclocking headroom.
 
  • Like
Reactions: coercitiv

DrMrLordX

Lifer
Apr 27, 2000
21,583
10,785
136
. But I haven't read a single person have anything even close to an issue with power delivery from a run of the mill B350 with a 1600x, 1700x, 1800x, 2600x or 2700x. After the Intial run in March of last year where most hopped onto X370 boards, B350 boards quickly became the defacto go to chipset. Here, Reddit, everywhere we would have a ton of issues if this was the case.

Actually people HAVE had issues, and I've read about them over on OCN. Typically you get a lot of VRM overheating as they get a major workout handling the heavy current draw of octocore AM4 CPUs. It doesn't cause mass automatic failure, but it will cause premature wear. That may result in a board "only" lasting maybe 2-3 years as opposed to 5-10 years of normal operation that you'd see with an R3 or R5. And of course I've read the cautionary statements from @The_Stilt (and others) regarding the effects on heavy current draw when using small numbers of (in some cases) poorly-chosen VRMs.

Increasing core count at the same TDP is only going to make things worse.

We're talking about their original plans for 10nm. There's little doubt it was about Atom.

If they have majority marketshare, yes they can get great margins. $20 is only average smartphone chip price, and only true way back then. If you look at Tech Insights the latest Galaxy and iPhones spend $50-70 on the SoC. That would have resulted in not just great margins, but great revenue. 200 million high end phones with $50 ASP on the SoC is $10 billion.

So what are the margins on those $70 phone SoCs? Cost per transistor hasn't necessarily improved (on 7nm, I think it's higher than it was back in the Bay Trail/Cherry Trail days), and some of those SoCs are now packing a LOT of transistors. We're talking A12, Kirin 980, Snapdragon 8150 and so forth and so on. Soooo I'm not absolutely convinced that the margins are where Intel would like them. And that assumes Intel could command top price. Based on Bay Trail's reception in the mobile world, I would expect Intel to be relegated to the bargain bin yet again.
 
Last edited:
  • Like
Reactions: ozzy702

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
And that assumes Intel could command top price. Based on Bay Trail's reception in the mobile world, I would expect Intel to be relegated to the bargain bin yet again.

You do know Intel's ASP is in the $100 range right? They might be little bit better. Maybe $120 for the CPU?

At $70, they'll be pretty well off. And I was talking about a hypothetical scenario where they have majority marketshare. This is likely the dream they had. Obviously it did not pan out, but that's not what I'm saying here.

The die sizes of the mobile chips are comparatively tiny. The A12 is at 83mm2. These companies make record-breaking revenue. They are making lots of money. It's similar to the argument made saying high cost of FinFET process is what caused price inflation for Nvidia's recent chips(Pascal and later). Yet they broke revenue and gross margin records year after year.
 
  • Like
Reactions: DarthKyrie
Mar 10, 2006
11,715
2,012
126
You do know Intel's ASP is in the $100 range right? They might be little bit better. Maybe $120 for the CPU?

At $70, they'll be pretty well off. And I was talking about a hypothetical scenario where they have majority marketshare. This is likely the dream they had. Obviously it did not pan out, but that's not what I'm saying here.

The die sizes of the mobile chips are comparatively tiny. The A12 is at 83mm2. These companies make record-breaking revenue. They are making lots of money. It's similar to the argument made saying high cost of FinFET process is what caused price inflation for Nvidia's recent chips(Pascal and later). Yet they broke revenue and gross margin records year after year.

In what universe does a high-end mobile SoC cost $70?
 

Topweasel

Diamond Member
Oct 19, 2000
5,436
1,654
136
Actually people HAVE had issues, and I've read about them over on OCN. Typically you get a lot of VRM overheating as they get a major workout handling the heavy current draw of octocore AM4 CPUs. It doesn't cause mass automatic failure, but it will cause premature wear. That may result in a board "only" lasting maybe 2-3 years as opposed to 5-10 years of normal operation that you'd see with an R3 or R5. And of course I've read the cautionary statements from @The_Stilt (and others) regarding the effects on heavy current draw when using small numbers of (in some cases) poorly-chosen VRMs.

The problem is you have put the 1700 in with the 1700x. So I don't know how well that actually follows. But the 1700 is also an example of why I don't believe in the limitations you have placed on it. I don't see why they can't have reduced power versions of even the 16 core chip.

Hot doesn't necessarily mean dead in a couple years. With out knowing the temp, the thermal limits, and what the MTBF at that limit I don't think we can really guess on it's life expectancy. But it would be hypocritical for me to discount those fears. It's my big issue with the 9900k that at default under any real work load it's going to pinned against its thermal limit unless you are running a 180w+ cooler. On the same level people pointing that out on the VRM's for B350/450 can't be discounting that.

I do question that source OCN over Toms, HardOCP, Anandtech, even redit. I am sure there are some that aren't trying to overclock these under specced boards and getting worried but it still seems to feed back an overclockers perspective that then escalates as more eyes are turned to it.
 

ub4ty

Senior member
Jun 21, 2017
749
898
96
I hope not. There is no need for a core war right now. We need better IPC/clocks on 8 cores. Anyone needing 16+ can use HEDT. Amdahl's Law is a thing, and we are seeing diminishing returns in most software already.

It's been a core wore since 4 cores. Clocks are yester-year. Anything between 3Ghz-4.7ghz is comfy space.
There are no diminishing returns because there's something called Memory stalls which most clock crazed individuals seemingly know nothing about. Your CPU doesn't operate on magic, it has to fetch things from memory. Those fetches take time and the CPU can't do anything until it gets the data. The big issue is people writing crappy software because everything is all about leveraging open source code they know noting about or frameworks... Many of them horrible for multi-core processing. And please don't mention gaming and higher FPS as the driver for higher clocks. IPC = variable based on the instructions. There's no such thing as an IPC measure. Another pop science data point for people with no formal education on computer architecture. The future is moar cores, corelets, a range of different cores, and a heterogeneous execution environment that punts tasks around to the compute device that can handle it most efficiently. No one is talking about frequency anymore and IPC has to be one of the most ridiculous terms ever popularized by e-celeb reviewers. AMD's new architecture is aiming towards heterogeneous computing which is what they laid out years ago as their ultimate vision and they're leaps and bounds ahead of intel which has been crippling I/O and cutting corners in branch look-ahead and other areas of their CPUs for years.
 
  • Like
Reactions: DarthKyrie

ub4ty

Senior member
Jun 21, 2017
749
898
96
If there's going to be PCIe4 support, I guess that many buyers won't be overly bothered about backwards compatibility with previous boards; they'll be wanting PCIe4 boards.
Of course, many does not equal all, but AMD may consider that they'll bring over many new to AMD customers to their platform. Admittedly, some of the value goes out of the platform if a new board is required, especially for those that bought into AM4 on the back of the compatibility guarantee that AMD outlined.
As I stated in prior posts, you're going to want to avoid buying anything that doesn't have PCIE 4.0 support for GPU and CPU. Having bought the first gen of AMD processors (the big tick), you secure the leap from 4 core to 8 core (14nm).. enjoy it and move on w/ the next major tick occurs (4.0/7nm). I have no issue w/ selling my Ryzen 7 1700/X370 and moving on to 7nm/PCIE 4.0. I bought the mobo for $100 or so and the processor for $170. No value was lost. I used it for about 2-2.5 years and now its time to upgrade. If I sell the mobo for $50 and proc for $85.. that buys me a new mobo and I then make a pull for the mid tier CPU. Welcome to the world of computer building. Nothing lasts forever and you're likely going to lose 50% to depreciation when you upgrade which is why you never buy the absurd high end product. The new mid line obsoletes it for far less $$$$.
 

ub4ty

Senior member
Jun 21, 2017
749
898
96
That's the downside to guaranteeing support for 4 years, new standards come out and you end up with situations like this.

I'm also sceptical about how budget B350 and B450 mobos with lower end VRMs will support a hypothetical 16C Zen 3000 chip. Many of those boards actually already have suboptimal VRMs for 8C Zen+ like the 2700X, so I understand the mobo makers concerns about doubling the core count, even on a more efficient 7nm process.

B350s can hardly handle 8 core Ryzen first gen processors.
This was known and published by tons of reviewers detailing how hot and insufficient the VRMs are.
I have no clue why anyone would buy those mobos when you could have gotten an X370 for $100. If someone is going to sit here and argue that $20/$30 made or broke their build plans, then they couldn't afford even a first gen 8-core and they most definitely wont be able to afford the high end processor that may go beyond 8 core. There's no sympathy or claim for support for the new gen processors. You bought a bargain basement mobo to save $20/$30 and that has consequences .. Ones you were clear on given all of the reviews. The idea of putting the next gen Ryzen processor north of 8 cores in such a mobo is ludicrous. The idea that you have enough money for one but can't buy something other than a bare minimum $80 mobo is also ludicrous. The idea that you would gimp performance running on such a mobo is also ludicrous. By compatible, AMD should mean that the processor runs in limp mode and puts this emoji on your boot screen : o_O
 

ub4ty

Senior member
Jun 21, 2017
749
898
96
The choice is ofc good thing. THe bad is the presentation. All around the web tells us AM4 B350 is ok.
But I think buying a new r7-3700X in medium b350 board will lock me to low or no overcklocking (or even reaching default turbo clocks because of VRMs) and low speed DDR4. Those signal you wont get on the internet as a non technical user.
Joe simply open "tech site", see the review and expect the same results but he doesn't get them.
Then he is disappointed.
I hope I am wrong and if AMD delivers that my friends asrock b350 pro4 (currently with R1600 non oced) can run the 3700X even at default turbo clocks without throttling, then salute to AMD.
So the next gen reviews should display the boards from the 300 series which can run the the new R3XXX without problems with the 3466MHz low latency DDR4 it deserves. Otherwise, display the performance of what you can get with the old board. That will be a correct signal. We are not getting that signal now. Only tech staff knows it.

Nowhere on the web is it known that the B350 is ok for anything other than first gen 6 core processors.
The difference between a B350 and an X370 was about $20-$30. You've really gotta be kidding me if you think that bare minimum board is going to run the most high end processor. A B350 can barely handle a 2700x and you knew that when you bought it. Be disappointed you didn't spend $20-$30 more to at least have some semblance of a proper motherboard when every youtube reviewer on the planet told you this. B350 was for people buying 4 core/6 core AMD processors that just wanted something cheap that got things going. It was never held as a competent motherboard for running anything beyond the bare minimum. So, if it can run the 4/6 core Ryzen 3 then AMD has upheld its agreement and that is within expectations. If a person wants to run to Intel and pay 2x/3x (doubtful since they're declaring they're broke) they are more than welcome.
 

ub4ty

Senior member
Jun 21, 2017
749
898
96
Intel showing once again why they've become a laughing stock.
Even if they released a 7nm chip in Jan, it's clear it would be 1.5x-2x the cost of AMD which is why they've been losing to begin with. When they start getting price competitive with AMD, they will potentially win me back. I could care less if they have the highest performing cpu that gives me 1-3% more FPS, I don't need that and its not worth 1.5x-2x the money. Releasing 14nm processors still and calling it potato sauce lake doesn't get sales. They honestly had something with that Hades Canyon Nuc and then they went out and prices it 50% above what it should have been. Intel seriously just loves losing. Nvidia is headed down that same track in the years to come. Stubborn, fragmented product line, ridiculous price.
 

Topweasel

Diamond Member
Oct 19, 2000
5,436
1,654
136
Actually people HAVE had issues, and I've read about them over on OCN. Typically you get a lot of VRM overheating as they get a major workout handling the heavy current draw of octocore AM4 CPUs. It doesn't cause mass automatic failure, but it will cause premature wear. That may result in a board "only" lasting maybe 2-3 years as opposed to 5-10 years of normal operation that you'd see with an R3 or R5. And of course I've read the cautionary statements from @The_Stilt (and others) regarding the effects on heavy current draw when using small numbers of (in some cases) poorly-chosen VRMs.

Increasing core count at the same TDP is only going to make things worse.
And this is what I meant with irrational internet voices making something seem a lot more than it really is.

Nowhere on the web is it known that the B350 is ok for anything other than first gen 6 core processors.
The difference between a B350 and an X370 was about $20-$30. You've really gotta be kidding me if you think that bare minimum board is going to run the most high end processor. A B350 can barely handle a 2700x and you knew that when you bought it. Be disappointed you didn't spend $20-$30 more to at least have some semblance of a proper motherboard when every youtube reviewer on the planet told you this. B350 was for people buying 4 core/6 core AMD processors that just wanted something cheap that got things going. It was never held as a competent motherboard for running anything beyond the bare minimum. So, if it can run the 4/6 core Ryzen 3 then AMD has upheld its agreement and that is within expectations. If a person wants to run to Intel and pay 2x/3x (doubtful since they're declaring they're broke) they are more than welcome.

An Example of when enthusiasts take their personal point of views on a matter and turn them into fact with little basis and damn people who "didn't listen to them". First to clear one thing up the VRM situation is independent of the chip-set choice. So blanket statements are baseless. Second that is only true from an enthusiast point of view not an actual tech one. God I hate posts like this.
 

maddie

Diamond Member
Jul 18, 2010
4,723
4,628
136
https://www.anandtech.com/show/13683/intel-euvenabled-7nm-process-tech-is-on-track

Intel's 7nm is on track using EUV, I didn't know that it was being manufactured simultaneously with 10nm in another Fab, that's Fab 42 as I read on WCCFTECH.
"Intel's 7nm is on track using EUV"

By itself a statement so devoid of information as to be almost worthless. This is surely for the stock analysts.

Translation: We're going to be missing for the 10nm generation but we'll be back, just you wait and see.
 

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136

ub4ty

Senior member
Jun 21, 2017
749
898
96
And this is what I meant with irrational internet voices making something seem a lot more than it really is.



An Example of when enthusiasts take their personal point of views on a matter and turn them into fact with little basis and damn people who "didn't listen to them". First to clear one thing up the VRM situation is independent of the chip-set choice. So blanket statements are baseless. Second that is only true from an enthusiast point of view not an actual tech one. God I hate posts like this.

First, to clear one thing up... The VRM situation is clear on B350/450 vs x370/x470. You get less power delivery on a B350/450 and less cooling on the VRMs. This is partially why its cheaper. If you want to argue against that, please try.

That being said, a number of different X370/X470 motherboards still have their own set of issues even w/ better heatsinks and power delivery because certain manufacturers cut corners or made crappy boards. Even from X370 to X470 there was an improvement due to lessons learned. Irrationally stating that select X370/X470 has VRM issues and then asserting that the even cheaper less equipped B350/450 mobos don't is quite the mental gymnastics and proves you're a fan of even contradicting yourself. If you're ok with running your VRM temps at 80*C - 110*C with a stock 2700x on a B350 board, be my guest. However, those are documented numbers. You get what you pay for. You want to save $20/$30 on a mobo to be the deal champion, you get roasting VRMs when you try running high end CPUs. The boards weren't targeted for that and you know it. Compatible but not efficient. Oh btw, the hotter those VRMs the less power efficient.

If you want to claim otherwise, then you're asserting you're all set for the highest end Ryzen 3 processor on a B350. Best wishes. Seems AMD is doing a stellar job in this case. I don't express my own view. I express data points and information extracted from a range of different sources that actually tested various boards... Something you should be doing too while claiming God status for B350 boards.

All the while, you're literally trying to claim there's this massive group of people who can afford $300+ processors but need to skimp on their motherboard to the tune of $30.... :D.

B350/B450 boards cost less because they use cheaper parts and have less features : FACT.
Flagship processor but bargain bin mobo = a bad build : FACT

See pic for a 2700x running stock on a particular b350 board. Yeah that's totally gonna run the highest end Ryzen 3 processor. Don't like that comparison? The other B350 hit a max VRM temp of 111*C (STOCK CPU setting). Anything in the 80*C range is bad.
 

Attachments

  • totally_bro.png
    totally_bro.png
    1.9 MB · Views: 13
Last edited:

DrMrLordX

Lifer
Apr 27, 2000
21,583
10,785
136
And this is what I meant with irrational internet voices making something seem a lot more than it really is.

So you want people using octocore AM4 chips on boards that were designed to handle Bristol Ridge? Because that's exactly what you find on some lower-end B350 boards(4+3 configs)

First to clear one thing up the VRM situation is independent of the chip-set choice.

Do you maintain a list of B350 boards with "good" VRM configs so that people don't get stuck running stock clocks with their octocore with zero boost potential? Because VRM throttling is a thing.

B350 boards have cheaper VRM configs than X370 ones, period. There's a tiny amount of overlap between the best B350 boards and the worst X370 ones. You aren't going to get robust 16-phase configs (or better) on B350! You can hate whatever you want, but it is never good advice to suggest that B350 can handle any AM4 CPU out there, particularly not from an upcoming generation with a higher bias towards current draw.

At $70, they'll be pretty well off. And I was talking about a hypothetical scenario where they have majority marketshare. This is likely the dream they had. Obviously it did not pan out, but that's not what I'm saying here.

BK deserved whatever he got it he tried to sell the shareholders on the idea that Intel would somehow own majority marketshare of the mobile market on 10nm. All the competition had to do to kick them out (aside from continuing to promulgate phone/tablet platforms generally hostile to x86) was tighten margins . . .
 
Last edited:

jpiniero

Lifer
Oct 1, 2010
14,510
5,159
136
"Intel's 7nm is on track using EUV"

By itself a statement so devoid of information as to be almost worthless. This is surely for the stock analysts.

Translation: We're going to be missing for the 10nm generation but we'll be back, just you wait and see.

In theory anyway, they could ramp up fast since the CPU tile dies should be tiny with Granite Rapids. And the main problem with Intel's 10 is that they went too dense without EUV, and that's not a problem at 7. You're still talking about 2021 at the earliest however.
 

DrMrLordX

Lifer
Apr 27, 2000
21,583
10,785
136
That would be interesting if Intel came out of left field with their 7nm process on-track. Might explain why Su is trying to reshape the x86 landscape so rapidly next year.

edit: I have to add that "on track" is as of news reports from August(ish) 2018? Since Intel apparently won't be doing anything EUV-related until late 2021.
 
  • Like
Reactions: ryan20fun

LTC8K6

Lifer
Mar 10, 2004
28,520
1,575
126
  • Like
Reactions: ryan20fun and Adawy

ub4ty

Senior member
Jun 21, 2017
749
898
96
So you want people using octocore AM4 chips on boards that were designed to handle Bristol Ridge? Because that's exactly what you find on some lower-end B350 boards(4+3 configs)



Do you maintain a list of B350 boards with "good" VRM configs so that people don't get stuck running stock clocks with their octocore with zero boost potential? Because VRM throttling is a thing.

B350 boards have cheaper VRM configs than X370 ones, period. There's a tiny amount of overlap between the best B350 boards and the worst X370 ones. You aren't going to get robust 16-phase configs (or better) on B350! You can hate whatever you want, but it is never good advice to suggest that B350 can handle any AM4 CPU out there, particularly not from an upcoming generation with a higher bias towards current draw.



BK deserved whatever he got it he tried to sell the shareholders on the idea that Intel would somehow own majority marketshare of the mobile market on 10nm. All the competition had to do to kick them out (aside from continuing to promulgate phone/tablet platforms generally hostile to x86) was tighten margins . . .

I honestly have no clue where his comment came from on B350 boards. There's a range of videos showing the VRMs hitting thermal ceiling and throttling with just a stock 2700x. There's literally no power phases there on some of the B350s to deliver what is needed thus why they can easily be overloaded and hit temp ceiling. Others have insufficient or no heatsinks. This is known to just about anyone doing an AMD build which is why you see B350 boards used on lower core count processors like they should and people get x370/x470 for the higher end. Some who really intend to push the limits get higher tier x370/x470 boards. I got a mid level x370 board and it was only about $100. Why on earth would I be a clown and buy a b350 and save $20/$30 and try to run an OC'd 1700 on it? The typical AMD builder is informed on this matter and simply doesn't do it and there are no cries that will come from people as to why they can't run a Ryzen 3 flagship processor on a b350.. That's just delusional
 

ub4ty

Senior member
Jun 21, 2017
749
898
96
That would be interesting if Intel came out of left field with their 7nm process on-track. Might explain why Su is trying to reshape the x86 landscape so rapidly next year.
Intel was the largest purchaser of EUV equipment a couple years ago. It looks like they had a contingency plan. Hopefully this will allow them to keep competitive.

Even if they did, what's the price going to be? 50-100% more? It' why they failed against AMD when it was 14nm/14nm. This story isn't going to change until Intel drops their prices. It's that simple.
 

LTC8K6

Lifer
Mar 10, 2004
28,520
1,575
126
Even if they did, what's the price going to be? 50-100% more? It' why they failed against AMD when it was 14nm/14nm. This story isn't going to change until Intel drops their prices. It's that simple.
I read somewhere that Intel hasn't lost much business, but AMD has gained new business?

But anyway, Intel either needs to lower prices, which 7nm should allow, or they need to increase performance by more than the past increments, which seems unlikely.

Of course, Intel might have just renamed 10nm+ or ++ to 7nm...
 

DrMrLordX

Lifer
Apr 27, 2000
21,583
10,785
136
I honestly have no clue where his comment came from on B350 boards. There's a range of videos showing the VRMs hitting thermal ceiling and throttling with just a stock 2700x. There's literally no power phases there on some of the B350s to deliver what is needed thus why they can easily be overloaded and hit temp ceiling. Others have insufficient or no heatsinks. This is known to just about anyone doing an AMD build which is why you see B350 boards used on lower core count processors like they should and people get x370/x470 for the higher end. Some who really intend to push the limits get higher tier x370/x470 boards. I got a mid level x370 board and it was only about $100. Why on earth would I be a clown and buy a b350 and save $20/$30 and try to run an OC'd 1700 on it? The typical AMD builder is informed on this matter and simply doesn't do it and there are no cries that will come from people as to why they can't run a Ryzen 3 flagship processor on a b350.. That's just delusional

I think one of the problems here is that @Topweasel is being pedantic in assuming that the boards "will work", following the assumption that you would have to make sacrifices on voltage/clockspeed to achieve such a thing. And technically, if you want to lock your cores to stock (or lower) clockspeeds and undervolt the chip, maybe you could pull it off. But yeah otherwise, time for that VRM throttling. The boards are good enough that they won't die prematurely (which is what @Topweasel was also saying). They'll just be hot and slow.

Even if they did, what's the price going to be? 50-100% more? It' why they failed against AMD when it was 14nm/14nm. This story isn't going to change until Intel drops their prices. It's that simple.

Intel will do what they always do: try to outperform the competition. I'm sure they want to have another Conroe moment, which they probably will if AMD rests on their laurels. If not then things get interesting.

Of course, Intel might have just renamed 10nm+ or ++ to 7nm...

I don't think they have, unless they're going to try to do 10nm+ on EUV. All signs point to:

1). EUV in late 2021 for Intel
2). 7nm in ???? 2021 for Intel
 
  • Like
Reactions: ozzy702