Intel Comet Lake Packs Up to 10 Cores (Updated)

[yian88]

I hope intel cant get 10nm until mid 2020 so that Zen 2 7nm has time to stretch its legs and gain market and AMD gain investors, then intel can comeback into the game to keep competition up.

[Drak3]

1 hour ago, Blademaster91 said:

nd AMD lied about the bulldozer core count. 

No, they didn't.

[TheDankKoosh]

13 hours ago, CarlBar said:

 

Yes you are. Q2 2017 was 2 years ago. Thats an eternity in CPU tech. It's the same time frame as the Ryzen 2000 series.

You seem to be ignoring the fact that the 7900x and the new 10 core from intel are 2 completely different beasts, first off intel will not want to shy away from ring bus on the mainstream, since they can still use it as a sort of advantage when it comes to latency, and from what I know ring bus has a tendency to heat up more. Secondly, I don't see them going away from the 1151 platform on their mainstream for this launch, which means that the die and ihs will be similar in size to the 9900k, and even though it's soldered, the 9900k still reaches pretty ridiculous power figures under stock conditions (150w at 4.7ghz all core turbo), so tack on another 30w to that and you have a nearly uncool-able chip due to heat transfer limitations. I think this will be a pretty lackluster launch compared to amd just from looking at all the data that's popped up so far anyway, at this point it's just intel playing catch-up.

[CarlBar]

6 hours ago, TheDankKoosh said:

You seem to be ignoring the fact that the 7900x and the new 10 core from intel are 2 completely different beasts, first off intel will not want to shy away from ring bus on the mainstream, since they can still use it as a sort of advantage when it comes to latency, and from what I know ring bus has a tendency to heat up more. Secondly, I don't see them going away from the 1151 platform on their mainstream for this launch, which means that the die and ihs will be similar in size to the 9900k, and even though it's soldered, the 9900k still reaches pretty ridiculous power figures under stock conditions (150w at 4.7ghz all core turbo), so tack on another 30w to that and you have a nearly uncool-able chip due to heat transfer limitations. I think this will be a pretty lackluster launch compared to amd just from looking at all the data that's popped up so far anyway, at this point it's just intel playing catch-up.

 

I wasn't arguing any of that. he discussion was about how revolutionary a high clocked 10 core chip is/is not technology wise. And somthing thats been around for 2 full years isn't revolutionary. Evolutionary, sure, revolutionary nope.

[Blademaster91]

On 3/17/2019 at 12:23 AM, Antistatic12 said:

Very true what you are saying. Intel has squeezed literally everything they can. So it as you say is "impressive".

Well people here insisted the 9900K 8c/16t was "everything they can squeeze" from their current node process, and I mean "impressive" if Intel keeps it on the mainstream 1155 socket, with 10 high clocked cores and an iGPU all on one die.

19 hours ago, Drak3 said:

No, they didn't.

The bulldozer chips didn't have an FPU for each core.

[Drak3]

3 minutes ago, Blademaster91 said:

The bulldozer chips didn't have an FPU for each core

And? They still had the # of cores they advertised.

[GoldenLag]

7 minutes ago, Blademaster91 said:

The bulldozer chips didn't have an FPU for each core.

still. to most intents and purposes still a core. 

 

if that there is not a core, oh boy im looking forward to the lawsuits regarding GPUs with their Graphics cores and Cuda Cores. 

 

if we had a standard definition of CPU core id probably agree, but we dont. and up untill the bulldozer lawsuit was reignited it was considered a core. 

[Princess Luna]

Does Intel really need to make a whole 'new generation' call it comet lake 10th gen just for the marketing of it?

 

14nm++ all over again what for? Ring Bus is on it's limit with the 9900K already and they want to shove another 2 cores in there? keeping it 16PCI-e lanes and dual channel too and likely needing a Z490 motherboard because 'power delivery'...

 

10nm is never going to happen...

Intel might be a worse pick than AMD on the 'premium' segment after Zen2...

[Blademaster91]

24 minutes ago, GoldenLag said:

still. to most intents and purposes still a core. 

 

if that there is not a core, oh boy im looking forward to the lawsuits regarding GPUs with their Graphics cores and Cuda Cores. 

 

if we had a standard definition of CPU core id probably agree, but we dont. and up untill the bulldozer lawsuit was reignited it was considered a core. 

The physical "cores" are there, though parts like the FPU and L2 cache aren't on each core, FP intensive tasks especially caused slowdowns and bottlenecks.

I think there should be a standard definition of CPU cores, a CPU should have its own resources per physical core, its misleading to the consumer to market a CPU as a 8 core when in most tasks it only performed as a 4 core/8thread unless the software was optimized for it.

30 minutes ago, Drak3 said:

And? They still had the # of cores they advertised.

It was advertised and marketed as having the performance of an 8 core, it is misleading to call it an 8 core when it really performed about as well as a 4 core/8 thread cpu.

[Zeeus]

LMAO intel is dying and im enjoying every single second of it, even though i have the intel i5 8600k, this will *hopefully* be my last intel cpu ever. Ryzen 3000 5GHZ FUCKIN COME AT ME!!!!!!!!!!!!!!!!!!!!!!!!!!! I NEED IT

[GoldenLag]

3 minutes ago, Blademaster91 said:

I think there should be a standard definition of CPU cores, a CPU should have its own resources per physical core, its misleading to the consumer to market a CPU as a 8 core when in most tasks it only performed as a 4 core/8thread unless the software was optimized for it.

i mean calling it a 4 core if its an 8 core isnt much better.......

 

also where do we draw the line of its own resorces? L3 cache or L2 cache or L1 cache. because if its L3 cache the Zen 8 cores turn into dual-cores.

[Blademaster91]

33 minutes ago, Princess Cadence said:

Does Intel really need to make a whole 'new generation' call it comet lake 10th gen just for the marketing of it?

 

14nm++ all over again what for? Ring Bus is on it's limit with the 9900K already and they want to shove another 2 cores in there? keeping it 16PCI-e lanes and dual channel too and likely needing a Z490 motherboard because 'power delivery'...

 

10nm is never going to happen...

Intel might be a worse pick than AMD on the 'premium' segment after Zen2...

I agree, I'd  much rather Intel just sell the 10 core as 9th gen and make it compatible with existing Z390 boards. I mean most decent Z390 boards should have plenty of VRM's for 2 more cores.

The Intel chips might be a worse pick than zen 2 but it probably wouldn't be much different than 9900k vs. 2700x, 9900k being better for gaming while 2700x is better for tasks that can take advantage of 8 cores.

[ravenshrike]

42 minutes ago, Blademaster91 said:

The bulldozer chips didn't have an FPU for each core.

According to this logic we used to have 0 core chips.

[RejZoR]

21 hours ago, Blademaster91 said:

I mean AMD really should have, bulldozer was their own fault and it seems like someone has to bring up AMD making a crappy CPU or GPU being Intel or Nvidia's fault which is pretty biased.

Except the low IPC cores with shared resources was completely their fault, and AMD lied about the bulldozer core count.  Maybe people can get over it and enjoy that there is strong competition from both sides.

No they haven't lied about anything. If CPU has 6 cores, it has 6 cores. There is no "half cores". Just because compute units are shared between cores, that doesn't make them lesser cores. That's like saying all CUDA cores before Turing were not real CUDA cores because they were unable to do FP and INT operations at the same time. Before, they could only do either FP or INT, but never both at the same time. The Bulldozer logic was about the same. Core is core, how things work behind the scenes is just how they designed architecture for it. And iirc at the time, AMD speculated they'd benefit from that, but situation turned a bit in the meanwhile (similar how NVIDIA scrwed up with FP32 during GeForce FX instead of going with FP24 everyone was using).

 

Besides, Bulldozer wasn't so bad as people said it was just because everyone was parroting that shit (FX 8350 was super popular and actually pretty decent CPU for budget gamers).

[CarlBar]

6 hours ago, Blademaster91 said:

Well people here insisted the 9900K 8c/16t was "everything they can squeeze" from their current node process, and I mean "impressive" if Intel keeps it on the mainstream 1155 socket, with 10 high clocked cores and an iGPU all on one die.

 

It pretty muc is. Unless Intel has found some way to lower power consumption they'e going to struggle to get better single or low thread count performance out of this chip. Obviously running more cores at lower clocks is going to produce better full multi-core performance on the same TDP, thats just an inherent fact of CPU design. It's just that until fairly recently a lot of stuff was still predominantly low thread count so it never made sense, even now it's of debatable value in some situations. But physical space allowing they allways had the option to raise core counts further. It will only be impressive if they're either managed to noticeably raise single thread performance or cut thermals compared to the 9900k.

[Billy Pilgrim]

6 hours ago, Princess Cadence said:

14nm++ all over again what for? Ring Bus is on it's limit with the 9900K already and they want to shove another 2 cores in there? keeping it 16PCI-e lanes and dual channel too and likely needing a Z490 motherboard because 'power delivery'...

 

Stop giving Intel ideas

[mr moose]

On 3/18/2019 at 4:06 AM, GDRRiley said:

It hurt them and lead to bulldozer and then eventually to zen. We can't know what would have happened if we never had bulldozer. 

There isn't anything to get over, Intel just did illegal things and should have been punished more but because courts are so slow intel made lots of money with little punishment. 

It did not lead them to bulldozer,  their design was not influenced by Intel at all.    What Intel did to AMD over 20 years ago has no bearing on today's market nor their position.  Yes Intel did the dirty and it had a cost, but people tend to forget that AMD made a lot of silly mistakes at the same time.

On 3/18/2019 at 4:06 AM, CarlBar said:

 

Yes you are. Q2 2017 was 2 years ago. Thats an eternity in CPU tech. It's the same time frame as the Ryzen 2000 series.

Not in the context of this discussion. 

[Mihle]

18 hours ago, GoldenLag said:

i mean calling it a 4 core if its an 8 core isnt much better.......

 

also where do we draw the line of its own resorces? L3 cache or L2 cache or L1 cache. because if its L3 cache the Zen 8 cores turn into dual-cores.

It is 8 cores, just 8 cores that perform very poorly.

[Exotath]

Sorry if this has been answered, I googled but didn't find a clear answer. Is this 10c/20t part to use existing Z390 hardware? Or is this another new chip/new board kinda thing?

[Razzy_85]

13 minutes ago, Exotath said:

Sorry if this has been answered, I googled but didn't find a clear answer. Is this 10c/20t part to use existing Z390 hardware? Or is this another new chip/new board kinda thing?

It's intel, probably will be a new board involved.

 

[Dedayog]

On 3/16/2019 at 5:20 PM, mr moose said:

 

Intel have, but those 20+ cores were locked to lower frequencies and used in servers were single core performance is not as important as multitasking database requests.   For the vast majority of domestic consumers 4-6 higher clocked cores is ample.  Intel's road map (and product release) follows consumer demand a lot more closely than a bunch of forum enthusiasts believe.   This is why they are still selling and why they are not in a hurry to release but loads of cores at the expense of single core performance.

This.  Should Intel follow user needs and their analysis, which is a bit more detailed than the arm-chair shit going on in this thread by all the industry insiders posting here... or keep up with AMD's core count solely to be able to say they gotz moar corez too!!

 

There's a reason Intel is doing what they're doing, it may not be what you would do but they have a roadmap and until you;re in the conference room at that staff meeting, grain of salt all of this.

 

For Intel and AMD and basically any company out there.  

[Brooksie359]

On 3/17/2019 at 2:15 AM, The Benjamins said:

There is a demand.

Intel should still be able to hit 5Ghz or near it on 1-2c on a 10c which would not hurt there single core performance, so IDK what the issue is here.

 

From where I see things zen2 will have more core, lower heat, same single core performance,  better boost tech, less vulnerabilities. The only thing I see intel doing better at is AVX, and memory speed. 

How do you figure? All of the information we have currently would say otherwise. AMD has claimed better single core performance and seeing how they demonstrated their new 8 core beating the 9900k in cinebench I would have to say its looming like that is indeed the case. They didn't gimp the 9900k either. It had the same cinebench score as third party reviews of the 9900k. 

[The Benjamins]

8 minutes ago, Brooksie359 said:

How do you figure? All of the information we have currently would say otherwise. AMD has claimed better single core performance and seeing how they demonstrated their new 8 core beating the 9900k in cinebench I would have to say its looming like that is indeed the case. They didn't gimp the 9900k either. It had the same cinebench score as third party reviews of the 9900k. 

I did say that I think Zen 2 will have better single threaded and multi threaded performance, but that Intel will still have a advantage in AVX and memory speeds. and i don't see that worth any higher price over performance over AMD.

 

 

also a cheap 16c24t CPU also means cheap 4c, 6c, 8c parts also in the product stack. we could see a 4c8t CPU that beats a 7700k for less then $150

[RejZoR]

Can we stop with the IPC nonsense already? IPC has been basically the same as Intel's since Ryzen launch. Give or take a %. Intel has higher single thread performance because of higher clocks, not because of IPC. IPC is measured at same clocks, not at different ones. And when you run intel or AMD at 4GHz fixed, they both perform about the same.

[pas008]

3 minutes ago, RejZoR said:

Can we stop with the IPC nonsense already? IPC has been basically the same as Intel's since Ryzen launch. Give or take a %. Intel has higher single thread performance because of higher clocks, not because of IPC. IPC is measured at same clocks, not at different ones. And when you run intel or AMD at 4GHz fixed, they both perform about the same.

you should link this with that statement

 

https://www.techspot.com/article/1616-4ghz-ryzen-2nd-gen-vs-core-8th-gen/