Intel Cannonlake SoC will have 4-core, 6-core and 8-core versions

[tential]

Why would all boards have to handle 140W? Make a six core skylake, call it "HEDT modern", and build special motherboards for it just like they do for X99.

Do you think it's a FREE cost to have a whole new platform in additional to the platforms they already have?

This is exactly what I'm talking about. Enthusiasts have NO idea have actual businesses are run. They act like it's as simple as "Oh, just make something brand new and sell it to me! You'll make more money because I'll buy it now!"
There are a TON of costs in creating and maintaining new ideas/products in a company.

I can't even deal with the level of idiocy in this discussion right now.

 

[myocardia]

And it shows what? People dropping quadcores for dualcores as gamers?

I could care less whose argument the data supports, I have no dog in this fight. That is the data for which you were asking, though, including the percent of rise or fall.

 

[mikk]

I dont believe there is more than 2-5mm2 difference from the 122mm2 number.

I don't think either, but why don't you round it to 120 mm² then when we don't have an exact number?

 

[frozentundra123456]

Do you think it's a FREE cost to have a whole new platform in additional to the platforms they already have?

This is exactly what I'm talking about. Enthusiasts have NO idea have actual businesses are run. They act like it's as simple as "Oh, just make something brand new and sell it to me! You'll make more money because I'll buy it now!"
There are a TON of costs in creating and maintaining new ideas/products in a company.

I can't even deal with the level of idiocy in this discussion right now.

Well, like I said, they could take part of the 4B per year they are throwing down the mobile rathole. At least the final product would be competitive, best in class actually, as opposed to atom.

If you follow your reasoning, then intel should only have one die, right? After all it is too expensive to manufacture both dual cores and quad cores. But lets wait and see. I understand what you are saying, minus the insults, but I think eventually this is going to come back and bite them. Sometimes it is worth spending money to have a superior product. All you have to do is look at atom. They refused to make it competitive for years, and now that they are trying, the ship has already sailed.

 

[Sweepr]

There seems to be the (IMHO erroneous) assumption that just because Intel is bringing 6/8 cores to the "mainstream" platform that the company won't go ahead and charge more for those chips.

Exactly. Even if they cost the same as entry-level HEDT chips some people would still buy them because of lower total system costs.
And 'mainstream' 6/8-core Cannonlake wouldn't kill the less expensive Skylake-E parts either, enthusiasts would still have the better platform, more PCIe lanes, more cache and possibly higher clocks (thanks to higher TDP).

 

[Arachnotronic]

Exactly. Even if they cost the same as entry-level HEDT chips some people would still buy them because of lower total system costs.
And 'mainstream' 6/8-core Cannonlake wouldn't kill the less expensive Skylake-E parts either, enthusiasts would still have the better platform, more PCIe lanes, more cache and possibly higher clocks (thanks to higher TDP).

Exactly. I also expect that by the time mainstream Cannonlake hits the market, Skylake-E/Cannonlake-E will come in much higher core count configurations.

 

[jpiniero]

The only way I could see this really happening is if Intel went to a ultra-dense-screw-clock-speeds process (sorta like Carrizo). Which I suppose could happen, at least for Cannonlake. There were rumors that there wouldn't be a socket release on Cannonlake although that was before the delay. It would however explain Intel's chart implying that the transistor cost was much better on 10 nm. Wouldn't it really ruin Turbo Boost 2.0 though? That's Core's best feature.

 

[StinkyPinky]

Is there really a market for hex core i7's? Enthusiasts can spend a bit more and get the xeons and 99% of other users don't need or care to have hex core cpu's. Especially is a market that's getting marginalised by mobile. Honestly, I cannot tell the difference already between my fathers haswell i3 and my haswell i7 in normal every day usage which is what the vast bulk of the population use their pc's for.

 

[RussianSensation]

I would advice people not to believe anything like this before it happens. There isn't any incentive to put more than 4 cores on a client chip.

Last time I checked, i7 5820K is a consumer chip -- it's not a $1000+ Xeon product line, it's not an enterprise chip and it's not priced out of reach of extreme/hardcore gamers. As increases in IPC start slowing down even more for Intel, how do you propose that they will provide much incentive for people to upgrade? IGPU? Please.

Lets be honest, more IGP would be more beneficial than more cores.

In the context of extreme/hardcore gamers, IGP is a worthless feature taking up die space and TDP headroom. More cores for gaming > IGP.

More cores=HEDT line.

Who said this? Just because this is the case right now, it doesn't mean Intel will forever be stuck on four cores as the maximum for their mainstream platform. Newsflash, at some point the mainstream Pentium platform only had single cores, then dual cores, then quad cores. You seem extremely confident in your statements to suggest that Intel will keep the mainstream line at 4-cores and by the tone of your post, you imply as if this development will persist for 10+ years.

But I look at it from a different perspective. What you ask or is that Intel makes a design, mask, production runs etc for a tiny niche of people who want a dual channel 6-8 core chip without IGP on a mainstream platform limited to 95W. See the issue?

1. TDP is not an issue since with lower nodes, eventually it will be possible to create a 6-core chip in a 95W TDP. TDP itself means nothing for hardcore gamers since Intel could low-ball clocks and overclockers could just overclock back to 200W.

2. What makes you think that the target market for a $450-550 6-core chip on the mainstream platform is a tiny niche? Do you realize a lot of people on this forum have purchased the X99 platform because it was the only choice to get a 6-core CPU? If 5820K was available on the Z97 platform, why would I pay extra for an X99 board? I wouldn't but I sure would pay $100 extra over the i7 4790K for the 5820K had Intel released it on Z97.

Until software radically changes its not needed and serves no purpose for the 99.x% crowd.

You are not making sense now. i7 6700K adds a useless amount of cache over i5-6600K and HT and that costs $100 more. 5820K has a ton more cache and 2 extra cores. Pound-for-pound, the mainstream i7 is starting to make less sense every generation moving forward because games aren't responding to the increases in IPC as was the case in the past. That means if $100 separates the i5 from the i7 and all you get is the tiny bump in cache and HT, it sounds like a horrible rip-off when a full-fledged 6-core with massive cache costs nearly the same.

And they are better of with better IGP, more cache, EDRAM, integrated PCH or whatever.

Gamers are not. In fact, I bet 90%+ of PC gamers on this forum would MUCH rather choose any of these options:

1) 6-core i7 on Z170
2) IGP stripped quad-core i5/i7, with the extra TDP used for higher boost clocks out of the box
3) IGP stripped quad-core i5/i7, with a reduction in the price since the die size will be much smaller
4) i3 with a K multiplier

Intel keeps wasting valuable financial and engineering resources, while literally wasting hundreds of millions of transistors on mostly useless IGP in terms of the desktop.

Intel is facing a toxic mix of stiff competition with their own 4+ year old chips, dwindling consumer PC sales and escalating fab costs. The last thing they want in the consumer segment is even more cores and larger die sizes.

If they literally took i5-6600K and i7-6700K and added 2 more cores, dropped the IGP, I bet the die size would be smaller than the existing products. It's only going to get worse as the GPU portion will getting bigger and bigger over the next 10 years. Essentially for PC gamers who are buying mainstream i5/i7, they are paying $$$ for the graphics components that's not used. Talk about a waste of $. I am sure Intel loves it though since they can maintain the artificially high prices for their products.

Think about it, if the market had a 6-core i7 6900K with no IGP and a 4-core i7 6700K with an IGP, who would buy the latter if they are priced similarly?

Another way to look at it is look at the prices of quad-cores since Q6600 days when it dropped to ~ $300 and follow the pricing of i7 860->2600K->3770K->4770K->4790K->6700K. Intel's prices aren't going down but the die sizes are dropping like a rock.

My Lynnfield 860 was 296mm2, SB i7 2600K was 216mm2, IVB i7 3770K was 160mm2, Haswell i7 4770K was 177mm2

vs.

$350 Skylake i7 6700K is 122.4mm2, an area just 41% of the i7 860 that cost $284 US at launch. Think about that!
http://www.anandtech.com/show/9505/skylake-cpu-package-analysis

Even if we use inflation of 2.5% over the last 6 years since i7 860 came out, the i7 6700K would only cost $329 not $350 and well you are paying for just a 122mm2 die, not a 296mm2 of the i7 860!

What does that mean? It means Intel could easily sell a 6-8 core i7 for $350 without the IGPU but they have no incentive because AMD is not competitive and consumers are lapping up $350 quad-core i7s on the mainstream platform. It's not about demand for the 6-core one mainstream platform but Intel would jeopardize margins/profits if they went 6-core mainstream without raising prices beyond $350.

People wanting 6-8 cores buy HEDT. Its a no brainer to get quad channel and the extra PCIe lanes if you are such an "extreme gamer".

You cant even get full bandwidth on a highend NVME drive on LGA1151. Not to mention the 95W issue.

Sure you can. Don't make stuff up. You can plug the NVMe into the PCIe 3.0 x16 slot or use the Samsung's 950 Pro M.2 drives and get the full speed. In fact, with the M.2 drives you can even RAID them, and get even faster speed.

95W TDP isn't an issue. Extreme/hardcore gamers will overclock. Not to mention a large portion of the TDP is eaten up by the useless IGPU. Take that out and make an i7 6900K and it'll be awesome even with a 95W TDP.

Then you and whoever can keep advocating for an unbalanced platform to save 50$ on a CPU and not feeling left behind. Its not making it any better.

Z170 isn't an unbalanced platform. Plus, you have amazingly short vision of the future:

1) DDR4 speeds will eventually > 4000mhz. Right now the X99 platform hardly benefits from DDR4 beyond 2133mhz. So what makes you think that by 2017-2020 with Cannonlake/Icelake, the 6-8 core CPU would be memory bandwidth starved on a dual-channel mainstream platform with DDR4 4000+ memory? Not to mention eventually DDR4 will be superseded by even faster memory.

2) PCIe/M.2 SSDs are already working at blazing fast speeds on Z170. Eventually Intel will upgrade the chipset and CPU lanes to PCIe 4.0. So again, you have no argument here at all.

3) The issue isn't about not moving to the X99 (or SKL-E) platform because of cost only, but timing as well. The fact that the high-end platform continues to lag behind architecturally sours a lot of its appeal.

We must now choose between a faster IPC platform with the latest features but limited to quads OR buy the X99 (or w/e the replacement is) but be behind architecturally. That's not a great trade-off for either platform. During the i7 920 days, you didn't have to make this trade-off at all and the X58 platform cost more $. For consumers, that was a WIN-WIN situation. Enthusiasts were wiling to pay more to get the best sooner, while the mainstream consumers had to wait longer.

Today, the platform that offers 6-8 cores is lagging behind the mainstream platform which means the higher entry cost for it are providing some disincentive for PC users/gamers to skip it entirely.

Even in this thread, several people have already voiced a desire for a 6-core Z170 platform i7 processor, but you keep making up stuff how it's a "tiny market niche" based on no evidence at all. During the Nehalem days, i7 920 and the X58 platform sold like hot cakes on this forum and otherwise. So the reality is history proves you more wrong than right too.

You are trying to make the LGA115x into a hexcore/octocore 140W market. While its not fitted for such.

Way to ignore history and node maturity.

1) We have already seen 130W+ flagship i7s (920) find their way into the mainstream platform under 95W TDP (i7 860). This happened on the same node due to tweaks and node maturity. So don't say this cannot be done in 2017 and beyond.

2) 95W TDP today is partly attributable to the useless IGPU.

3) 95W TDP simply implies lower CPU clocks. Not a problem since plenty of people would just overclock.

And before you start making up arguments how the mainstream platform wouldn't be able to handle the huge power loads of 6-8 cores 130-140W TDP chips, it absolutely can. It just requires a good motherboard and great CPU cooler. Already been done before in 2009:

The real reason Intel keeps segregating the mainstream from the high-end platform is because it allows them to sell tiny quad-core i7s for $350 in an era where we should already have $300+ 6-8 core i7s. Another reason that allows Intel to do this is 0 competition from anyone. Hopefully if Zen or Zen+ delivers a semi-decent 6-8 core offering at $300 on the mainstream AMD platform, Intel will have to finally recognize that there is definitely a demand for MOAR cores among PC users and gamers.

Like "competition" would change anything.

Software needs to radically change. Else there so no incentive. And we can see plenty of people buying i5 over i7.

By we, how many are we talking about?

Can you accurately predict the state of software in 2017, 2020, 2025? What you also fail to grasp is because Intel keeps quad-core i7 prices artificially high due to lack of competition, the 6-core Intel CPU platform upgrade is beyond the reach of most consumers. This very fact partially holds back software from being more optimized/focused on > 4 threaded applications.

Let's see now if AMD and Intel had 6-core CPUs at $200 and 8-core CPUs at $300, do you think no one will figure out a way take advantage of the extra threads in major applications or games? Look at what's happening on the Xbox One/PS4 side as developers are already tapping into more than 4 cores for development. So you aren't understanding that software will evolve if the hardware is there. For most consumers, the hardware simply isn't there because it's not affordable.

So, if even a cutting-edge DX12 RTS title is just using 2 real cores it doesn't look like heavily multi-core loading gaming is coming anytime soon.

There are plenty of games that use 4 cores and some that use 4 cores + HT.

AT already showed that some applications benefit from 4+ cores on the smartphone/Android platform. Console developers have shown that some games use > 4 cores. Some PC games benefit from 4+ cores. Sure, there isn't a liner benefit but it's there.

I know it's hard for some people in this thread to accept that some games today benefit from more than 4 cores but software will not stand still forever.

 

[cbn]

All X99 boards can handle 140W.

If you want 140W on LGA1151, essentially all boards would have to support this. From H110 to Z170. And it adds cost.

The quad memory channel controller and 40 PCIe lanes add power consumption.

So if we had a mainstream octocore (with dual channel, less PCIe lanes) power consumption would be less at the same clockspeed,

 

[jpiniero]

Last time I checked, i7 5820K is a consumer chip -- it's not a $1000+ Xeon product line, it's not an enterprise chip and it's not priced out of reach of extreme/hardcore gamers. As increases in IPC start slowing down even more for Intel, how do you propose that they will provide much incentive for people to upgrade? IGPU? Please.

More cores, obviously for the E line. But that's a server line processor. More cores makes sense for servers.

In the context of extreme/hardcore gamers, IGP is a worthless feature taking up die space and TDP headroom. More cores for gaming > IGP.

It's great for mobile though.

1. TDP is not an issue since with lower nodes, eventually it will be possible to create a 6-core chip in a 95W TDP. TDP itself means nothing for hardcore gamers since Intel could low-ball clocks and overclockers could just overclock back to 200W.

TDP is an issue on mobile though.

2. What makes you think that the target market for a $450-550 6-core chip on the mainstream platform is a tiny niche? Do you realize a lot of people on this forum have purchased the X99 platform because it was the only choice to get a 6-core CPU? If 5820K was available on the Z97 platform, why would I pay extra for an X99 board? I wouldn't but I sure would pay $100 extra over the i7 4790K for the 5820K had Intel released it on Z97.

The die mask cost is way more than you think. Intel wouldn't recoup the costs unless it was tied to a server or mobile part.

That means if $100 separates the i5 from the i7 and all you get is the tiny bump in cache and HT, it sounds like a horrible rip-off when a full-fledged 6-core with massive cache costs nearly the same.

You do get the newer platform and features.

Intel keeps wasting valuable financial and engineering resources, while literally wasting hundreds of millions of transistors on mostly useless IGP in terms of the desktop.

Mobile is the #1 priority. Servers are #2. Desktop... #5 maybe? Plus the increased IGP has the benefit of hurting nVidia.

My Lynnfield 860 was 296mm2, SB i7 2600K was 216mm2, IVB i7 3770K was 160 mm2, Haswell i7 4770K was 177mm2

Intel's costs have gone up substantially. They've made up the costs by making the dies smaller. You will see the same thing happen to discrete GPUs starting with 16FF.

 

[cbn]

TDP is an issue on mobile though.

If wanting to make low TDP quad core, having more cores on the die might help though (ie, pick the four strongest cores out of eight for a greater number highly binned quad core Core M than if the die were native quad core)

 

[Fjodor2001]

H170 lacks 4 lanes, no memory above 2133, less M.2 slots, less USB3 ports, no PCIe splitting for SLI/CF. And with this you save a whooping 20$ max over Z170.

Again, none of which affects performance much, under the assumptions mentioned before. And the chipset itself may be $20 cheaper for the OEM, but some of the non-Z170 motherboards are a lot cheaper than that (lowest priced Z170 motherboard vs lowest priced non-Z170 motherboard). You can often get them for around half the price. I assume it's because the Z170 motherboard often also have more expensive components, more connectors, and what not.

 

[Fjodor2001]

Also when RS replied, it was about cost since Fjodor was trying to shift the goalpost again because of the HEDT platform destroyed his argument.

Not shifting goalposts. It was just you who (intentionally) misunderstood why I brought it up. I've already explained it to you, but I'll do it again since you seem to already have forgotten about it:

No, that's not what I'm saying. I think you missed my point. I'm saying it's possible to build a high performance gaming PC using a quite cheap motherboard and chipset, since that does not affect performance much. However on the HEDT platform, there are no such cheap motherboards.

Having said that, I think most "extreme" gamers will opt for expensive motherboards anyway, even on the mainstream 1151 platform. That's because those users are often enthusiasts that like to spend extra $$$ to get top of the line HW, no matter what impact it'll actually have on performance.

 

[Fjodor2001]

No I didn't.

So how many cores do you think the Average Joe needs? 1, 2, 4, ...?

You forgot to answer this question Shintai. Afraid to share your opinion on it?

 

[biostud]

the problem with the steam data is that we don´t know the amount of time each user spend gaming.

I have steam on my laptop and my desktop, and my gaming time on the laptop is 1-2h/year. So even if it shows up in the steam survey, it doesn't represent my "gaming" machine. I think that would be the case for many users. We would need to know the gaming time on the 2c/4c/6c/8c systems to know which systems are most used by actual gamers.

 

[ShintaiDK]

Not shifting goalposts. It was just you who (intentionally) misunderstood why I brought it up. I've already explained it to you, but I'll do it again since you seem to already have forgotten about it:

I think you need to define your "extreme gamer" in details before you shift again to try fit the "moar cores" for cheap agenda. Because as it is now you haste to the next goalpost as your previous is shoot down.

 

[ShintaiDK]

Not going to reply to all your wall post.

95W TDP isn't an issue. Extreme/hardcore gamers will overclock. Not to mention a large portion of the TDP is eaten up by the useless IGPU. Take that out and make an i7 6900K and it'll be awesome even with a 95W TDP.

According to Fjodor they wont. And when you need a Z170 you can pretty much get X99 if you feel you need more cores.

DDR4 speeds will eventually > 4000mhz. Right now the X99 platform hardly benefits from DDR4 beyond 2133mhz. So what makes you think that by 2017-2020 with Cannonlake/Icelake, the 6-8 core CPU would be memory bandwidth starved on a dual-channel mainstream platform with DDR4 4000+ memory? Not to mention eventually DDR4 will be superseded by even faster memory.

So we are not talking about Z170 anymore. But some hypothetical hardware of the future. You completely lost focus again.

Z170 isn't an unbalanced platform. Plus, you have amazingly short vision of the future:

PCIe/M.2 SSDs are already working at blazing fast speeds on Z170. Eventually Intel will upgrade the chipset and CPU lanes to PCIe 4.0. So again, you have no argument here at all.

Now its not Z170 anymore is it.

Sure you can. Don't make stuff up. You can plug the NVMe into the PCIe 3.0 x16 slot or use the Samsung's 950 Pro M.2 drives and get the full speed. In fact, with the M.2 drives you can even RAID them, and get even faster speed.

If you use the x16 slot, then what about graphics?

And for M.2 it seems you dont understand how the platform works at all. You seem to confuse it with how X99 works.

Hint of the day, DMI 3.0.

95W TDP today is partly attributable to the useless IGPU.

95W TDP simply implies lower CPU clocks. Not a problem since plenty of people would just overclock.

Its quite clear you have no idea what a modern CPU uses in TDP for the different areas. And again you mention overclock, while ignoring the small price difference between Z170 and X99.

Here, let me give you a lecture.

Maybe its time you move out of your 2011/2012 hardware and get some real experience.

 

[thepath]

Maybe it will be something like this

Celeron/Pentium model have quad core without HT
Core i3 quad core with HT
Core i5 six core with HT
Core i7 eight core with HT

 

[Lepton87]

HEDT isn't always behind in architecture when I bought my CPU there weren't any 14nm CPUs on the market. HEDT Only looks bad only now because mainstream had two releases and HEDT had none since last year. Is BW-E even coming?

 

[ShintaiDK]

BW-E is coming soon.

 

[frozentundra123456]

Considering the mediocre clocks we saw on Broadwell desktop, BW-E could actually be a regression. I know there are rumors about how great it will be, but there were rumors like that about Skylake too.

 

[Lepton87]

Any leaks if it overclocks any better than the mainstream BW? I bought an expensive mobo because I had plans to upgrade the CPU I hope it's going to be a worthy upgrade.
BW-E at 4.5GHz with 5% higher IPC would be fine by me. 128MB of eDRAM would be awesome but I know that's not even a distinct possibility on HEDT part since they lack IGP.

 

[Fjodor2001]

I think you need to define your "extreme gamer" in details

You should ask Intel, because it's their definition of "extreme gamer" that we're talking about. And they did not provide more details in the article that what already has been mentioned.

Because as it is now you haste to the next goalpost as your previous is shoot down.

You still don't get it. All I'm saying is that it's possible to build a system using a cheap mainstream 1151 motherboard without that having much impact on performance, since there are cheap motherboards available on the mainstream platform, and the motherboard does not affect performance much. On the HEDT platform there are no such cheap motherboards. Are you objecting to that?

Whether you want to bring that into the discussion on number of cores or not is another story. But don't you think it's a factor that should be taken into account when comparing the mainstream vs HEDT platform, if total system cost is also up for discussion?

 

[Fjodor2001]

According to Fjodor they wont.

FFS, stop lying dude. I've said that not all extreme gamers will overclock. That's not the same as saying none of them will. Most likely a lot of them will OC, but how many we don't know, and my point is only that they don't have to do it.