Intel Comet Lake Packs Up to 10 Cores (Updated)

[mr moose]

2 hours ago, thorhammerz said:

Consider the social platform (and by extension, the audience and its demographics) this thread/discussion is living in , where spending an extra $200 every 1-2 years on a motherboard can send some into a state of palpable consternation.

My response is that anyone who upgrades that frequently is already spending somewhere between $400 and $600 every 2 years for 10% increase in CPU performance so they probably already spend that $200 on a 2 feature motherboard upgrade.  

[thorhammerz]

5 hours ago, mr moose said:

spending somewhere between $400 and $600 every 2 years for 10% increase in CPU performance so they probably already spend that $200 on a 2 feature motherboard upgrade.  

Gotta get that 10 FPS increase in gaming (and a reason to hate on Intel) .

[CarlBar]

22 hours ago, mr moose said:

I have an idea, don't reply to a discussion half way through with a different context in mind and then expect me to bend my previous discussion to address it.

 

 

The entire discussion chain leading up to my post, 9the next post made in this was the one i initially quoted.

 

On 3/17/2019 at 2:27 AM, The Benjamins said:

Intel has had 10c CPUs with higher then 4.0 GHz for a while, not sure why you are comparing it to some of the xeons? 7900x?

 

This is just a reason why I think its a sad disappointment.

 

I think its going to be far over priced with no real performance improvement over AMD in any category that can remotely justify the cost.

 

These are all my opinions.

 

On 3/17/2019 at 2:55 AM, mr moose said:

The highest clock speed 10 core part prior to the i9 was 3.3Ghz, correct me if I'm wrong but it has only been since q2 2017 that they have managed faster than 3.5Ghz on 10 plus cores.

 

EDIT: and lets not forget their best performance was 28 cores but that was a highly binned chip that required a $1000 water chiller.

 

 

 

On 3/17/2019 at 6:35 AM, The Benjamins said:

the 2066 package also had a 4c4t CPU, that does not mean that we can't put a 4c4t CPU on a smaller package.

 

The physical size of a 10c20t die didn't really change from 14nm to 14nm+ or how every many +s we are on now.

 

Their is nothing new needed to take the knowledge of the 8c16t die and 12c24t die used on the 2066 CPUs and make a 10c20t die with 2 DDR4 memory channels and a iGPU.

The physical die would be a tighter fit on the existing 1151 socket type, but nothing that challenging.

 

the TDP of a 7900x is 140w with a slight efficiency bump from improved process it can easily be pushed into a 120w window.

 

so again I don't see any innovation here just a mix of existing technology which will be behind AMD.

 

If intel really had to struggle to make a 10c mainstream CPU on 14nm with their existing knowledge I would be even more concerned with the direction of intel

 

On 3/17/2019 at 6:45 AM, mr moose said:

Again you are ignoring the nature of the technology,  without making allowances for the fact they haven't been able to push that many cores into the higher clocks and keep the power draw anywhere near stable is not something you can just ignore.

 

If Intel wanted to run 16 cores at 5Ghz they probably could, but it would come at the expense of single core performance. Which would make them the same as AMD and make no sense or extra sales to any of their customers.   

 

 

 

 

 

On 3/17/2019 at 6:58 AM, The Benjamins said:

You keep ignoring that the 7900x and the 9900x exist ALREADY.

 

You take off 2 memory controllers, that saves you some power, heat, and die size.

You reduce the core count from 12 to 10 on the die, that saves space.

You use a more refined 14nm process which saves heat and power.

they should easily able to do a 10c with not much rework.

 

The 9900x is 100mz lower base clock then a 9900k, the 7900x is only 300 mhz lower then a 9900k.

Boost clocks only measure for the single core clock speed so getting near 5 Ghz on one core should still be achievable.

But even so the 9900x has a boost of 4.4 which is respectable.

 

So what is so damn ground breaking about making a slimmed down 9900x die?

 

You keep replying to someone saying this is not revolutionary by pointing out intel hasn't had a 10 core part at high clocks before. Then someone points out they have. What am i expecting you to understand that you haven't explicitly stated. I can only assume you completely failed to understand what @The Benjamins was arguing in the first place...

[mr moose]

6 minutes ago, CarlBar said:

 

 

The entire discussion chain leading up to my post, 9the next post made in this was the one i initially quoted.

 

 

 

 

 

 

You keep replying to someone saying this is not revolutionary by pointing out intel hasn't had a 10 core part at high clocks before. Then someone points out they have. What am i expecting you to understand that you haven't explicitly stated. I can only assume you completely failed to understand what @The Benjamins was arguing in the first place...

No, his initial claim was that Intel have had 20+ cores for a while now and that this is not impressive because of that.  I merely pointed out that the reason they have had 20+ cores is because most of them run sub 3Ghz.  The closest Intel have come is 3.3Ghz on 10 cores was only 2 years ago.

 

 

You call q2 2017 ages ago, but when I look at CPU developments over the last 30 years it's only yesterday. 

 

He's literally pointing to a 2 year old CPU and asking for this one to be twice as good when its hasn't even made a node shift.

 

 

[CarlBar]

6 hours ago, mr moose said:

He's literally pointing to a 2 year old CPU and asking for this one to be twice as good when its hasn't even made a node shift.

 

 

At no point in the quoted discussion does he talk about a 20 core part. Try again.

 

6 hours ago, mr moose said:

You call q2 2017 ages ago, but when I look at CPU developments over the last 30 years it's only yesterday. 

 

Q2 2017 was also when AMD launched the Ryzen 5 2600x. By the time these CPU's arrive we should be in the process of getting the R6 3600X which by all indications will be matching a 9900K, using user benchmarks comparison as a basis, (it;s quick and easy to dig up ok), thats a more than 30% improvement in performance in 2 years. Conversely nothing we've heard so far suggest this 10 core will have more than a few percent edge over the equally old 7900X, (i'd guesstimate 10-15% myself), and it will be a well over 100w part whilst AMD are claiming near parity on TDP with their new part given what they showed off at CES.

 

Thats another strike against intel. The new 10 core may have a huge edge over the then top of the line desktop processor just from all those extra cores, but it's seeing a huge power usage spike. Which means intel is basically brute forcing things. if we can criticize AMD for doing that with Vega and NVIDIA for Firmi (to name 2 other examples), i don't see why we can't do the sanme here.

 

The only, (and highly arguable), innovation here is that intel has managed to get process yields to the point they can yield such a chip at a commercially viable price point and quantity. But that kind of improving yield is par for the course with process nodes hence why i say it's arguable.

[mr moose]

2 hours ago, CarlBar said:

 

At no point in the quoted discussion does he talk about a 20 core part. Try again.

 

 

 

On 3/17/2019 at 6:08 AM, The Benjamins said:

This makes no sense Intel has had 20+ cores on 14nm node for a while now, 10c20t is not impressive on 14nm at all.

 

People are whining because 10c20t is barely going to be able to compete with the competition.

As I said before, don't come in to a discussion late and expect me to change anything just for you.

 

EDIT: just for clarification, I don;t care about which parts of the discussion you want to quote or talk about specifically, that is where my discussion started.  Feel free to start a new discussion if you wish.

 

 

[pas008]

On 3/21/2019 at 9:22 AM, Zeeus said:

lmao asking intel to essentially NOT be intel. Good luck lel

Businesses are fun

You design them to run indefinitely 

Part of that is staling and milking sadly 

Tire oil food silicon etc

They banked enough to withstand an apocalypse

2 hours ago, CarlBar said:

 

At no point in the quoted discussion does he talk about a 20 core part. Try again.

 

 

Q2 2017 was also when AMD launched the Ryzen 5 2600x. By the time these CPU's arrive we should be in the process of getting the R6 3600X which by all indications will be matching a 9900K, using user benchmarks comparison as a basis, (it;s quick and easy to dig up ok), thats a more than 30% improvement in performance in 2 years. Conversely nothing we've heard so far suggest this 10 core will have more than a few percent edge over the equally old 7900X, (i'd guesstimate 10-15% myself), and it will be a well over 100w part whilst AMD are claiming near parity on TDP with their new part given what they showed off at CES.

 

Thats another strike against intel. The new 10 core may have a huge edge over the then top of the line desktop processor just from all those extra cores, but it's seeing a huge power usage spike. Which means intel is basically brute forcing things. if we can criticize AMD for doing that with Vega and NVIDIA for Firmi (to name 2 other examples), i don't see why we can't do the sanme here.

 

The only, (and highly arguable), innovation here is that intel has managed to get process yields to the point they can yield such a chip at a commercially viable price point and quantity. But that kind of improving yield is par for the course with process nodes hence why i say it's arguable.

Remember amd is using 8 core yields with if

Intel can use how many core yields with emib?

Not to mention intel did use multi chip designs on the past

The gluing statement originated from who?

 

Intel does have tech for more threads per core also

 

Hence my statement above this

Milking

And soon Intel prolly will get Qualcomm tech through Qualcomm being dumb not giving licenses which will lead to other things 

But that's like many years from now

 

 

[CarlBar]

4 hours ago, mr moose said:

 

As I said before, don't come in to a discussion late and expect me to change anything just for you.

 

EDIT: just for clarification, I don;t care about which parts of the discussion you want to quote or talk about specifically, that is where my discussion started.  Feel free to start a new discussion if you wish.

 

 

 

You don;t get off that easy. (I swear i used central F to look for the number 20 btw, but yes i'd forgotten about that), the discussion had long since moved off that point, (in fact it may even have been a typo. If you were still focused on 20 core at the point thats on you, not me or @The Benjamins. He even asked why you where still focused on the zeon's...

 

EDIT: I guess what i'm really saying here is if you genuinely thought the discussion was on 20 cores at that point, (and i don't disbelieve you did btw), why were you not raising "what are we discussing" questions like i did? My interpretation of what @The Benjamins was trying to say is completely different from what your saying he's saying and i can't help but feel at this point that everyone's been arguing about different things to each other.

[Citadelen]

If mainstream Ryzen caps out at 16 cores this year, then Intel's competitive position in EVERYTHING, will have gotten worse for the third year in a row. I still remember how people would say it didn't matter if Ryzen was decent, because Intel would crush them next gen, but as we see now it just gets worse and worse for them, this is actually hilarious, it might start to actually hurt their bottom line this year. Especially as EPYC shipments have been doubling Q/Q, and Rome looks like it's going to crush into a trillion pieces.

[mr moose]

10 hours ago, CarlBar said:

 

You don;t get off that easy. (I swear i used central F to look for the number 20 btw, but yes i'd forgotten about that)@The Benjamins

 

You can continue all you like, I have already addressed what I considered worth discussing.

 

 

[Taf the Ghost]

11 hours ago, Citadelen said:

If mainstream Ryzen caps out at 16 cores this year, then Intel's competitive position in EVERYTHING, will have gotten worse for the third year in a row. I still remember how people would say it didn't matter if Ryzen was decent, because Intel would crush them next gen, but as we see now it just gets worse and worse for them, this is actually hilarious, it might start to actually hurt their bottom line this year. Especially as EPYC shipments have been doubling Q/Q, and Rome looks like it's going to crush into a trillion pieces.

As strange as it is, AMD almost can't hurt Intel's revenue very much. Intel is already selling everything they can Fab, and they will be for a while. Demand on CPUs is just increasing and it doesn't look to stop. It could be an issue for Intel in 2022, however, but they should be pushing out 7nm parts. That's when it gets competitive again.

[mr moose]

9 minutes ago, Taf the Ghost said:

As strange as it is, AMD almost can't hurt Intel's revenue very much. Intel is already selling everything they can Fab, and they will be for a while. Demand on CPUs is just increasing and it doesn't look to stop. It could be an issue for Intel in 2022, however, but they should be pushing out 7nm parts. That's when it gets competitive again.

And that's just looking at the overall picture, when you factor in that the performance difference between Intel and AMD, It's really only lower income domestic users who will fret over the performance/$ figures.

[Taf the Ghost]

10 minutes ago, mr moose said:

And that's just looking at the overall picture, when you factor in that the performance difference between Intel and AMD, It's really only lower income domestic users who will fret over the performance/$ figures.

Outside of servers, it's never really been about performance. Minus the Bulldozer era, normally Intel didn't have a performance advantage. It's always been about Intel keeping AMD out of markets. Most especially the OEM and Mobile markets. Intel has never had an iron grip on the Retail market. The top K-SKU sells normally 2:1 better than the next part, and the two 2 K-SKUs normally sell over 70% of Intel's Retail parts. This is why Intel will keep clocking the hell out of 14nm.

 

This is why AMD has positioned Ryzen, Threadripper & Epyc in the ways they have. AMD has mostly wiped out the x600k model with the x600 line, in Retail. If they could somehow get an 8c that clocked to the moon, they'd probably do significant damage to the x700k line as well. But Intel won't worry about Retail Channel that much. (AMD might outright be over 50% market share there.) It's the Dell/HP/Lenovo's of the world that Intel works to keep AMD out of. Part of the reason I still question AMD not getting a small iGPU on their normal SKUs. (The assumption is that they can't break into the OEM channel right now, so it's not really worth it from a R&D/Opportunity cost perspective.)

[Jito463]

On 3/21/2019 at 9:00 PM, mr moose said:

My response is that anyone who upgrades that frequently is already spending somewhere between $400 and $600 every 2 years for 10% increase in CPU performance so they probably already spend that $200 on a 2 feature motherboard upgrade.  

The jump from Z270 to Z370 was even less than that, it was literally one feature (NVMe RAID).  Personally, I'm glad that I'll be able to just drop in replace my 1800x to whatever 3000 series part I get, without having to upgrade my MoBo.

 

Spoiler

https://www.pugetsystems.com/labs/articles/Z370-vs-Z270-What-is-the-Difference-1009/

 

[mr moose]

1 hour ago, Jito463 said:

The jump from Z270 to Z370 was even less than that, it was literally one feature (NVMe RAID).  Personally, I'm glad that I'll be able to just drop in replace my 1800x to whatever 3000 series part I get, without having to upgrade my MoBo.

 

  Hide contents

https://www.pugetsystems.com/labs/articles/Z370-vs-Z270-What-is-the-Difference-1009/

 

Let's hope you are lucky enough to be able to do that and we don't run into any of the feature issues that AMD had over the course of previous platforms.

 

 

[Jito463]

36 minutes ago, mr moose said:

Let's hope you are lucky enough to be able to do that and we don't run into any of the feature issues that AMD had over the course of previous platforms.

Lack of new features on old boards hardly qualifies as "feature issues".  I don't recall any actual problems that occurred, just new features that weren't available without upgrading the board, which is only to be expected.  It's like the expected PCIe 4.0 on the new boards - even though some have speculated otherwise - I don't anticipate having my board suddenly upgraded to PCIe 4.0 just from dropping in a new CPU.

[mr moose]

3 minutes ago, Jito463 said:

Lack of new features on old boards hardly qualifies as "feature issues".  I don't recall any actual problems that occurred, just new features that weren't available without upgrading the board, which is only to be expected.  It's like the expected PCIe 4.0 on the new boards - even though some have speculated otherwise - I don't anticipate having my board suddenly upgraded to PCIe 4.0 just from dropping in a new CPU.

We've actually had this conversation before:

 

 

 

Long story short, over the course of the AM platform we don't know what limitation will end up being due to motherboard/socket limitations.

[Jito463]

1 minute ago, mr moose said:

We've actually had this conversation before:

And I stand by my assertion that lack of new features doesn't equate to "feature issues".

[mr moose]

20 minutes ago, Jito463 said:

And I stand by my assertion that lack of new features doesn't equate to "feature issues".

 

I guess nothing is an issue of you don't want it to be. 

 

 

[Jito463]

42 minutes ago, mr moose said:

I guess nothing is an issue of you don't want it to be. 

 Or inversely, nothing's an issue unless you make it one.  

[mr moose]

1 minute ago, Jito463 said:

 Or inversely, nothing's an issue unless you make it one.  

Nope,  It's one thing to decide for yourself if an  issue is an issue, but you cannot decide that for someone else.

 

Having a DDR3 capable CPU restricted to DDR2 is not an issue one just decides they have.

[Jito463]

2 minutes ago, mr moose said:

Nope,  It's one thing to decide for yourself if an  issue is an issue, but you cannot decide that for someone else.

I was hoping you would realize I was just being pedantic for irony's sake.  Just messing with you at this point.

2 minutes ago, mr moose said:

Having a DDR3 capable CPU restricted to DDR2 is not an issue one just decides they have.

Technically you do, if you already owned a DDR2 board and just dropped in a replacement processor.  It's not like one can change the physical RAM slots at the same time.  At least it provided an option for people.  Now, if a company sold DDR3 capable processors but mounted on DDR2 boards, that's another issue entirely.  However, that's more of an issue with corporations trying to cheap out (which is nothing new and will likely never change).

[mr moose]

1 minute ago, Jito463 said:

 

Technically you do, if you already owned a DDR2 board and just dropped in a replacement processor.  It's not like one can change the physical RAM slots at the same time.

 

Exactly why long term platform support historically has come with it's limitations.  As I said before, I hope this doesn't happen this time around.  But I am not banking on it.   I have a B350 AM4 motherboard here waiting for a CPU.  It's either going to be a 3000 series or the cheaper 2700X.  The board is supposedly only able to support 8 cores, if it can't be updated with bios to support the hoped for 12 or 16 cores of the new ryzen part, then I see little benefit in putting such a CPU on it (the single core performance increase would have to be unrealistically large to warrant having unused cores).

 

 

 

 

[Jito463]

9 minutes ago, mr moose said:

Exactly why long term platform support historically has come with it's limitations.

On this I completely agree, I just take umbrage (perhaps too strong a term?) with defining them as "issues".  It's not really an issue that the option is there, provided that the detriments are clearly detailed and explained.  One can also just as easily upgrade your motherboard and CPU together.  There's certain to be those who decided to upgrade to the 500 series boards when they replace their CPU, but it's also nice that the original 300 series boards will be able to drop in replace to the newest generation.  I'm just of the mindset that more choices is better.

9 minutes ago, mr moose said:

I have a B350 AM4 motherboard here waiting for a CPU.  It's either going to be a 3000 series or the cheaper 2700X.  The board is supposedly only able to support 8 cores, if it can't be updated with bios to support the hoped for 12 or 16 cores of the new ryzen part, then I see little benefit in putting such a CPU on it (the single core performance increase would have to be unrealistically large to warrant having unused cores).

That's probably just a holdover from the fact that first and second gen Ryzen topped out at 8 cores.  However, even if that's not the case, you'll likely still have the option of going with the 8 core 3000 series chip for the single-core performance boost we should see out of it.  That's just my two cents worth on the subject.

[mr moose]

3 minutes ago, Jito463 said:

On this I completely agree, I just take umbrage (perhaps too strong a term?) with defining them as "issues".  It's not really an issue that the option is there, provided that the detriments are clearly detailed and explained.  One can also just as easily upgrade your motherboard and CPU together.  There's certain to be those who decided to upgrade to the 500 series boards when they replace their CPU, but it's also nice that the original 300 series boards will be able to drop in replace to the newest generation.  I'm just of the mindset that more choices is better.

 

Absolutely more choice is better,   Always a big proponent of more options.   Just keep in mind my comments are addressing the argument/idea that upgrading the CPU every 2 years is good while the idea of upgrading the motherboard as well is an unexpectedly frustrating event.   Anyone who has followed tech long enough should know that two years is  enough time for motherboard feature support to change causing CPU only upgrades to become questionable in terms of value.