Soldato
- Joined
- 12 Feb 2014
- Posts
- 3,013
- Location
- Somewhere Only We Know
-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
Another upcoming 14nm++++++ Desperate CPU from Intel
- Thread starter Jamin280672
- Start date
Soldato
- Joined
- 28 Sep 2014
- Posts
- 3,575
- Location
- Scotland
14nm 6th Time Over: Intel Readies 10-core "Comet Lake" Die to Preempt "Zen 2" AM4
Look like 14nm 10 core CPU in H1 2019 and 10nm in H2 or end of 2019, think I am glad I stick with 8700K.
Look like 14nm 10 core CPU in H1 2019 and 10nm in H2 or end of 2019, think I am glad I stick with 8700K.
Caporegime
- Joined
- 8 Jan 2004
- Posts
- 32,614
- Location
- Rutland
Only interesting thing is whether AMD will go 12 core Zen 2 to counter this.
What's the point Zen 2 will have 16.
16, EPYC 2 is eight eight core dies linked to a four quadrant eight channel uncore which can be split up, 2 channels per quadrant = 2 dies = 16 cores.
I don't see Ryzen 3000 hitting 16 cores, that eats too much into Threadripper territory. Level off at 12 cores for now, get the internal optimisations up for IPC and memory speed, get Threadripper 3 at a 16-core baseline and then go for 16-core Ryzen 4000.
Since there's still a lot of gum-flapping from AMD's detractors about "throw MOAR corez at it!!!" I'd rather see AMD get Ryzen 3000's actual cores equal or superior to Intels, and then boost the core count just to rub it in everybody's faces.
Caporegime
- Joined
- 8 Jan 2004
- Posts
- 32,614
- Location
- Rutland
I would be surprised but happy to be wrong.
Seems a strange move to use a dual ring bus when the majority of Intel's higher core count chips have moved away from ring/dual ring to a mesh topology.
https://www.servethehome.com/things...tion-intel-skylake-sp-cpus-mesh-architecture/
https://www.servethehome.com/the-new-intel-mesh-interconnect-architecture-and-platform-implications/
https://www.tomshardware.com/news/intel-mesh-architecture-skylake-x-hedt,34806.html
Still using 14nm isn't really as big a deal as is made out - Intel's 14nm is broadly equivalent to TSMC/Samsung's 10nm processes in terms of Transistor Gate and Interconect Pitch.
https://www.servethehome.com/things...tion-intel-skylake-sp-cpus-mesh-architecture/
https://www.servethehome.com/the-new-intel-mesh-interconnect-architecture-and-platform-implications/
https://www.tomshardware.com/news/intel-mesh-architecture-skylake-x-hedt,34806.html
Still using 14nm isn't really as big a deal as is made out - Intel's 14nm is broadly equivalent to TSMC/Samsung's 10nm processes in terms of Transistor Gate and Interconect Pitch.
I don't understand how people start with the premiss of better cores OR more, Ryzen 3000 will be both, better cores and more of them.
The design will be two dies eight cores, watch the video.
That being said, the 14nm I/O chip better be seriously fast because a Ryzen 3000 capped at 8 cores means there's only 1 chiplet in use, and the I/O die and IF link is never going to be as fast as having all the I/O on die.
Unless, of course, all of the perfect 8 core chiplets are staying with Rome and Ryzen 3000 is getting all the defective dies, in which case purely in terms of maximum salvage, Ryzen 3000 could hit 14 cores (if 7-core chipslets don't get used for a lower-end Rome).
Unless, of course, all of the perfect 8 core chiplets are staying with Rome and Ryzen 3000 is getting all the defective dies, in which case purely in terms of maximum salvage, Ryzen 3000 could hit 14 cores (if 7-core chipslets don't get used for a lower-end Rome).
I only started with that "premise" because of where Zen started. As of Zen+ there is still a little bit of work to do to surpass Intel's IPC and memory speed support. Everything we saw in Next Horizons indicated that we will see superior IPC with Zen 2, but it's not guaranteed. I'm just saying that until consumer software catches up with core counts, I'd rather see a Ryzen 3000 equipped with 8 or 12 Intel-beating cores just to fully cement their technological dominance. And shut the fanbois up. Then start ramping to core counts again afterwards.
Ryzen is Dual Channel, one eight core die per channel. Again!!!! watch the video.
This is Intel flailing around trying to one up AMD.
Right now they cannot match AMD on core counts and if anything this petty nonsense from Intel will only cause a self inflicted wound, the higher their core count on the mainstream the more it eats into their HEDT market, mind you having said that AMD have all but killed off Intel's HEDT, seriously look at how many HEDT chips Intel sell compared with AMD now, Intel are not even close.
As for the 9900K, they ain't selling many of those either, they are not even in Amazons top 100 sellers list, going above 6 cores for Intel is literally just them trying to stick it back to AMD, its not going to work because no matter what Intel do AMD have the technology advantage to beat any Intel counter move, easily.
This was discussed in EPYC thread at great length. It seems some people still believe more cores is better, even if those cores are weaker than Intels cores. I agree with you, AMD pushed 8c/16t parts to mainstream, now solidify your dominance with improved cores, to shut up Intels only advantage today: stronger core and gaming performance. Then in couple of years, then more games are coded using DX12/Vulkan and more software starts using more cores, and gets used to Zen design, then push to 16 cores/32threads.
What's the point of 16 cores now, when intel will turn around and still gonna top the gaming charts, because AMD 16 core chip cannot match Intels IPC and clock speeds.
Another one who thinks in binary ^^^^^^ read on...
AMD are pushing Intel into places they don't want to go, driving core counts up and prices down, way down.
Look, the Zen+ core is already as good as Intel's in terms of IPC, Zen 2 will have better IPC, it can't not given what we know about it, even a very conservative estimate is +5%, about the difference between Zen and Zen+, that will put Zen 2 IPC ahead of Intel, clock speed between Zen and Zen+ was 10%, add 10% ontop of Zen+ and you have 4.8Ghz, that's 5% short of Intel's 5Ghz but with 5% higher IPC, this is being very conservative with the estimates, in certain tasks its actually as high as 29%, i think 10 to 15% is a realistic reality.
Its fine, AMD have got this. Both ways
AMD are pushing Intel into places they don't want to go, driving core counts up and prices down, way down.
Look, the Zen+ core is already as good as Intel's in terms of IPC, Zen 2 will have better IPC, it can't not given what we know about it, even a very conservative estimate is +5%, about the difference between Zen and Zen+, that will put Zen 2 IPC ahead of Intel, clock speed between Zen and Zen+ was 10%, add 10% ontop of Zen+ and you have 4.8Ghz, that's 5% short of Intel's 5Ghz but with 5% higher IPC, this is being very conservative with the estimates, in certain tasks its actually as high as 29%, i think 10 to 15% is a realistic reality.
Its fine, AMD have got this. Both ways
Easy there humbug, I've never once disagreed with you about the progress Zen has made. All I'm saying is I don't do hype train and however tasty the Zen 2 announcement was, I'm not getting ahead of myself. If Zen 2 beats Intel's IPC then ramp the cores now. If it doesn't then personally I'd prefer the work go towards doing that instead of increasing core counts beyond what can be used. That's software devs and Intel's fault, not AMD's.
Either which way, Intel are in for a drumming for the next couple of years, but the popcorn is staying in its bag until I see Rome and the 3700X in action
Either which way, Intel are in for a drumming for the next couple of years, but the popcorn is staying in its bag until I see Rome and the 3700X in action
Another one who thinks in binary ^^^^^^ read on...
AMD are pushing Intel into places they don't want to go, driving core counts up and prices down, way down.
Look, the Zen+ core is already as good as Intel's in terms of IPC, Zen 2 will have better IPC, it can't not given what we know about it, even a very conservative estimate is +5%, about the difference between Zen and Zen+, that will put Zen 2 IPC ahead of Intel, clock speed between Zen and Zen+ was 10%, add 10% ontop of Zen+ and you have 4.8Ghz, that's 5% short of Intel's 5Ghz but with 5% higher IPC, this is being very conservative with the estimates, in certain tasks its actually as high as 29%, i think 10 to 15% is a realistic reality.
Its fine, AMD have got this. Both ways
This is AMD we are talking about. They will manage to cripple themselves somewhere, even though their CPU team is on the roll. I remember when Vega looked like a home run, there was no way it will be slower than 1080ti, just impossible, seeing all the improvements done on paper...
Intel is already where it doesn't want to be. AMD is pushing them from all angles. AMD jumping to 16 cores on mainstream will not have any benefits for AMD. Bragging right are already there with Threadripper. If AMD didn't have Threadripper, then yes, fire away with 16 core AM4 chip to claim top dog. Especially if software was halfway there to use them threads.
Gaming performance won't sell enough of these chips for Intel.
16 core AMD mainstream chip won't either
the point is, gaming and low threaded workloads are last forte of Intel. AMD need to make sure there is no excuses from Intel fan side with Zen2