We're now squishing even more on a 14nm die! (Just another day at Intel) Comet Lake!

[imreloadin]

Someone needs to get Intel a new clock, this tock tock tock nonsense is getting a little ridiculous....

[CarlBar]

20 minutes ago, Crunchy Dragon said:

3, I think. Whichever one was going to be 7nm.

 

Zen 2 is 7nm, Zen 3 will be 7nm+

[CommandMan7]

2 hours ago, ARikozuM said:

Feels like the Intel CEO has been listening to a lot of Simon & Garfunkel.

Ironically they still haven't named a replacement CEO since Krzanich got kicked out. The interim CEO is Robert Swan atm and theyre still looking for a replacement.

[TheRandomness]

33 minutes ago, imreloadin said:

Someone needs to get Intel a new clock, this tock tock tock nonsense is getting a little ridiculous....

They are being extremely ambitious with their 10nm. Their 10nm is actually the closest to a '10nm' we have currently, all others so far are physically closer to 14nm or 12nm.

[imreloadin]

1 hour ago, TheRandomness said:

They are being extremely ambitious with their 10nm. Their 10nm is actually the closest to a '10nm' we have currently, all others so far are physically closer to 14nm or 12nm.

This excuses them from basically recreating the "same" product how exactly?

[Canada EH]

3 hours ago, valdyrgramr said:

That's when AMD said they're unveiling Navi and Zen 2.

Interesting it maybe some stuff they've been trying to master before deployment but its probably a very small step for mankind.

[GloriousPain]

9 hours ago, dizmo said:

And yet...I have a feeling they might still beat the equivalent AMD chip. People love bashing Intel for this, yet AMD was falling flat on their face for years.

It'll be interesting to see, that's for sure.

AMD has beaten Intel at core clocks before. It didn't help. (Yes, I know that was then with a much older architecture, but it's still true).

 

I'm pretty interested to see how Zen2 does.

Yeah and people bashed AMD constantly for years until they got their shit together. Now it's even worse because intel is doing exactly what they laughed at AMD for; stupidly just adding cores and clocks; then calling it a new generation and increasing prices exactly like AMD. 

[Brooksie359]

9 hours ago, GoldenLag said:

Probably an Threadripper branded CPU with WX branding. A 3800wx costing around 600$. They cant tarnish their threadripper

To be fair threadripper comes with quad channel memory and a ton of pcie lanes so it's not like a 16 core mainstream cpu would really compete with a 16 core high end desktop cpu all that well. Plus anyone who bought a 16 core cpu would likely have access to an even higher core count threadripper cpu.

[Rune]

Ah balls, my 9900k

[CarlBar]

2 hours ago, imreloadin said:

This excuses them from basically recreating the "same" product how exactly?

 

But their not. Their iterations on the 14nm node have actually produce steps forward in per transistor performance and density. The big issue Intels really running up against is thermals because power draw is closely tied to transistor size and can't really be improved that much as a result. Thats the real reason intel hasn't pushed core counts too aggressively. At the kind of clocks intel likes to run at for lightly threaded performance advantage an 8 or more core CPU is just too much thermally speaking 

 

You have to remember the process also affects the peak attainable clock speed with smaller processes generally pushing the clocks upwards. AMD even with TSMC's 7nm is barely able (on paper anyway), of edging out the best intel currently has. If this comet lake gets any clock speed uptick, (and it almost certainly will), their going to be matching TSMC's significantly smaller process, thats actually technically quite impressive.

 

In fact the only thing really holding intel back right now is how far behind they are on the chiplet process. If they'd started on that at the same time AMD did or even before that they could bury AMD right now, process constraints be dammed because a chiplet architecture lets you spread the thermals out far more than a monolithic design, so with it intel could afford vastly higher TDP's without issue on the thermals side, (cooler manufacturers would probably be unhappy with the size of the resultant IHS and neither board manufacturers nor PSU companies would be exactly delighted ethier, but they could absolutely produce the hardware in a consumer usable form).

 

I'm basically convinced where gonna see a double die design on a new socket for this 10 core for that very reason. How they interconnect everything is going to be interesting.

[williamcll]

Who in intel should be blamed for this?

[dizmo]

3 hours ago, GloriousPain said:

Yeah and people bashed AMD constantly for years until they got their shit together. Now it's even worse because intel is doing exactly what they laughed at AMD for; stupidly just adding cores and clocks; then calling it a new generation and increasing prices exactly like AMD. 

They did it repleatedly, they deserved it. Intel has only had one generation of trouble. The amount of bashing is really unwarranted.

[The Benjamins]

6 hours ago, williamcll said:

Who in intel should be blamed for this?

Brian Krzanich

 

He became CEO just after haswell (4770k)  launch, he is to blame for the lag in CPU market on intel side.

[The Benjamins]

8 minutes ago, dizmo said:

They did it repleatedly, they deserved it. Intel has only had one generation of trouble. The amount of bashing is really unwarranted.

Intel did sit on 4c way to long and they did rush 2 CPU launches so far going on 3. so I see 2 going 3 generations of issues.

[GloriousPain]

15 minutes ago, dizmo said:

They did it repleatedly, they deserved it. Intel has only had one generation of trouble. The amount of bashing is really unwarranted.

"One generation" Since Haswell Skylake has been the only generation of actual improvements; Broadwell, Kaby Lake, and Coffee Lake... Have all been micro improvements or adding new cores/upping clocks; Tantamount to the FX X1XX and X3XX and 9XXX series(es)...

 

And again this isn't  because they're making a mistake in doing so; it's because they're doing it hypocritically after bashing AMD for doing it. Especially because AMD was relatively forced into it due to lack of funds to do much else until they finished developping Ryzen. Intel is just having issues throwing money into the air and getting something to come out of it after trash talking for years.

[imreloadin]

25 minutes ago, The Benjamins said:

He became CEO just after Kaby lake (4770k)  launch, he is to blame for the lag in CPU market on intel side. 

The 4770K was a Haswell chip, I believe Kaby Lake was the 7700K.

[MarvinKMooney]

If I remember right intel has stopped with the "Tic toc" a few years ago in order to "optimize" Node sizes.

So its all in the plans for them I guess.

 

 

https://www.extremetech.com/extreme/225353-intel-formally-kills-its-tick-tock-approach-to-processor-development

Edited November 26, 2018 by Viridian

[Brooksie359]

2 hours ago, CarlBar said:

 

But their not. Their iterations on the 14nm node have actually produce steps forward in per transistor performance and density. The big issue Intels really running up against is thermals because power draw is closely tied to transistor size and can't really be improved that much as a result. Thats the real reason intel hasn't pushed core counts too aggressively. At the kind of clocks intel likes to run at for lightly threaded performance advantage an 8 or more core CPU is just too much thermally speaking 

 

You have to remember the process also affects the peak attainable clock speed with smaller processes generally pushing the clocks upwards. AMD even with TSMC's 7nm is barely able (on paper anyway), of edging out the best intel currently has. If this comet lake gets any clock speed uptick, (and it almost certainly will), their going to be matching TSMC's significantly smaller process, thats actually technically quite impressive.

 

In fact the only thing really holding intel back right now is how far behind they are on the chiplet process. If they'd started on that at the same time AMD did or even before that they could bury AMD right now, process constraints be dammed because a chiplet architecture lets you spread the thermals out far more than a monolithic design, so with it intel could afford vastly higher TDP's without issue on the thermals side, (cooler manufacturers would probably be unhappy with the size of the resultant IHS and neither board manufacturers nor PSU companies would be exactly delighted ethier, but they could absolutely produce the hardware in a consumer usable form).

 

I'm basically convinced where gonna see a double die design on a new socket for this 10 core for that very reason. How they interconnect everything is going to be interesting.

One of the biggest things that allows Intel to have a performance advantage is the ring bus architecture so to say if they had a chiplet design they would be destroying AMD is pure speculation. It would be an entirely new architecture and if the interconnect between the chiplets is bad then they would be worse then Zen so let's not go talk about ifs. The fact is that Intel is struggling and can't do what they set out to do just yet. Who knows when we will actually get 10nm. Actually who even knows how creditable this rumor even is. 

[CarlBar]

42 minutes ago, Brooksie359 said:

One of the biggest things that allows Intel to have a performance advantage is the ring bus architecture so to say if they had a chiplet design they would be destroying AMD is pure speculation. It would be an entirely new architecture and if the interconnect between the chiplets is bad then they would be worse then Zen so let's not go talk about ifs. The fact is that Intel is struggling and can't do what they set out to do just yet. Who knows when we will actually get 10nm. Actually who even knows how creditable this rumor even is. 

 

Don't confuse a well optimized design for an inherently superior one. The absolute minimum latency is (naturally), directly related to the propagation of the signal through the conductive medium and the length of the path it must traverse. However the real world values are allways constrained more by other factors than this one. That's important overall as it means that changing from the current design to a chiplet design isn't going to cause major latency changes due to the changes in physical positioning of components. It's all going to come down to the underlying design of the other components of the interconnects really. And it's obvious intel knows exactly what they're doing when it comes to that. To be fair i suspect AMD does too, but it's also obvious that he Zen1 and Zen+ Memory controllers and IF fabric links weren't really anywhere near as robust as would have been ideal.

 

Don't get me wrong i'm not saying there would be no impact on intel interconnect latency, but i think your severely overestimating how extreme it would be.

 

Personally long term if HBM3 lives up to it's name i'm expecting chip manufacturers to start integrating that onto dies, (or possibly building their CPU's a bit differently so that the HBM is on an interposer that the CPU then sits in letting you mix and match HBM3 dies with CPU's but thats severe speculation), with high endurance non-volatile memory, (like 3Dx-point and Z-NAND), taking up DIMM slots. That would offer even lower latency volatile memory to CPU's and a much faster high capacity non-volatile storage for whatever programs stuff is currently being used.

[yian88]

Its funny cause they still deliver iGPU with these chips and people still dont figure out you are paying 1/3rd of the chip price for the iGPU, you hear no one complain about this, even the youtube experts overlook it.

If they shave the thickness of the chip silicon away and remove iGPu then 10 cores is totally possible and affordable, you would need a beefy mobo power delivery though.

 

[jagdtigger]

55 minutes ago, CarlBar said:

i'm expecting chip manufacturers to start integrating that onto dies

So intel can rip us off even more by removing dimms so you have to buy the crazy exepnsive variants to have more RAM? No thanks...

[asus killer]

6 minutes ago, yian88 said:

Its funny cause they still deliver iGPU with these chips and people still dont figure out you are paying 1/3rd of the chip price for the iGPU, you hear no one complain about this, even the youtube experts overlook it.

If they shave the thickness of the chip silicon away and remove iGPu then 10 cores is totally possible and affordable, you would need a beefy mobo power delivery though.

 

Isnt it irrelevant? For cheaper gpus it serves a purpose, for the more high end you probably coudnt get away with having the extras cores, power consumption, heat,...

[cj09beira]

9 hours ago, CarlBar said:

 

Any uptick in peak performance usually comes with lower price point release tat uptick too. Zen2 is really going to make it's meat and bones there. AMD could release an 8 core Ryzen 5 3600X clocked at roughly the same values as a 9900K but at the Ryzen 5 price point. Given the IPC improvements on Zen 2 that should smoke everything but the 9900K and be very close to the 9900K, all whilst being less than half the price. And given what they can do with the Ryzen 7 without going about 115w TDP or so it's not remotely unreasonable they'd do that.

 

I figure the actual Ryzen 7 will be 12 cores clocked up to around the same clocks as a 9900K. Or they could do 10 cores at a slightly higher frequency.

btw amd always keeps the same amount of cores enabled in each ccx, you they can only do 4-6-8 for each ccx pair, 

so if amd goes to 2 dies (2x(2*4)) they can only do 8-12-16

8 hours ago, TheRandomness said:

They are being extremely ambitious with their 10nm. Their 10nm is actually the closest to a '10nm' we have currently, all others so far are physically closer to 14nm or 12nm.

thats not really true though, because 10nm isn't the latest node for the other foundries, they are on 7nm now, and their 7nm is just as good (in most regards) as intel's planed 10nm (the first generation is probably not going to have the density they wanted as they cant get that to yield anything) intel has lost their lead

[GoldenLag]

1 hour ago, cj09beira said:

so if amd goes to 2 dies (2x(2*4)) they can only do 8-12-16

AMD has hinted that they can physically remove dies. Meaning that it doesnt neccesarly need them all to be present

[GoldenLag]

8 minutes ago, VegetableStu said:

Ryzen Gen 3:

• 4 die R7?
  "R7 3700X" = 32C?! okay that's too much
• 4 dice with half the active cores, or 2 dice with all active? (which gets higher frequencies?)
  "R7 3700X" = 16C? Intel's Comet lake desktop rumored to top out at 10C, so this is more likely
• R5 x6xx had 3/4 than R7 as well
  "R5 3600X" = 12C?
• "R5 3500X" = 8C? would AMD even need to turn on multithreading here? (considering the 9700K is already a 8C8t part)

 

possibility of wx branding on CPUs using 3 dies (1 IO and 2 Core dies). regular x and non-x might just go up to 8 cores and only having 2 dies present. a single IO die and a single core dies. 

 

4 core dies will be too large to fit. and wasting silicon on "dummy-dies"

 

the 10 core rumours from intel are telling us a lot considering AMD has been tight on info about Ryzen. Intel probably knows a lot more than we do, and paying attention to both of them we can probably deduct what is coming in the following years.