Coffelake Thread : Rumors, and Specs

[CHADBOGA]

Really depends on what Zen+ brings to the table, but if it's 6c/12t @ 5 GHz then it might be worthy of note.

At least Intel is FINALLY coughing up more cores. They're late, though.

Presumably Zen+ will improve single core performance, whether it is from IPC or frequency.

Whilst I doubt Zen+ will improve by the same degree that AMD's FX8350 did over the FX8150, but if Zen is troubling Intel now, then if it increases its single core performance, then the 8 core version in particular, with trouble Intel even more.

 

[thepaleobiker]

So perhaps a QTR before Zen+ is due?

If Coffee Lake doesn't include an 8 core variant, I don't see how it will stem the AMD tide that Coffee Lake is meant to do.

Intel is also going to be gearing up Cannon Lake, 10nm lithography based CPUs (ok, they've confirmed they'll use that node first and foremost for server CPUs and high margin products) but given than Zen+ is still expected to be on 14nm, Intel might get another shot at performance leads by virtue of being on a smaller, more efficient node.

Of course, Zen+ could still match/compete with Intel's 10nm. We dont know what AMD/Intel has coming up.

My only hope is, as noted by others on the thread above me, that Intel will "stick" to LGA 1151 for the 14nm Coffee Lake ...... #fingerscrossed

Regards,
Vish

 

[DrMrLordX]

Presumably Zen+ will improve single core performance, whether it is from IPC or frequency.

Whilst I doubt Zen+ will improve by the same degree that AMD's FX8350 did over the FX8150, but if Zen is troubling Intel now, then if it increases its single core performance, then the 8 core version in particular, with trouble Intel even more.

This is all true.

Intel just needs to stick to their fundamentals while they pour most of their work into Icelake and Optane support. Optane is one possible edge that may keep people coming back to the Intel platform, even if they aren't absolutely in love with the CPUs. They can fight a staying action if they can get Coffeelake 6c chips up to 5 GHz (OC). There is probably enough time for Intel to do a few extra spins/tweaks to try to get there.

Skylake-X is launching soon, and I doubt that Intel has the time to make many changes.

If you consider what happened when people went from spinners to SSDs, and if Optane can basically do that again for the end-user . . . it could be trouble for AMD.

Or NVMe drives may short-circuit the whole Optane revolution. Who knows?

 

[Mopetar]

Whilst I doubt Zen+ will improve by the same degree that AMD's FX8350 did over the FX8150, but if Zen is troubling Intel now, then if it increases its single core performance, then the 8 core version in particular, with trouble Intel even more.

They could do fairly well. When you've got a new architecture there's a lot of low hanging fruit that makes for easy and decent jumps in performance. Look at what Intel was able to accomplish in the early iterations of their Core 2 architecture which was already doing quite well. Massive increases in clock speed were responsible for some of the biggest performance leaps, but they were still making really solid double-digit IPC gains on top of that until Ivy Bridge (and even after that they were still seeing around a 7% gain on average) when they had picked most of that low-hanging fruit.

Interestingly enough, looking back most of Piledriver's gains came from the 400 MHz clock speed bump. If you remove that advantage, the IPC improvement was only around 7% for a lot of benchmarks. Incidentally that's around the same amount Intel has been getting recently and that's been regarded as rather poor. Then again you could argue that Bulldozer was so misconceived that it's not really possible to make improvements that don't involve throwing the whole bloody thing out and starting over.

I wouldn't really count on getting more than a 10% clock bump from the next zen generation (zeneration? ) with 4.0 base probably being the highest and potentially necessitating an increase in TDP for such a chip. But if they can get a solid double-digit IPC improvement that's going to be far more valuable because that lets them close the gap with Intel more than anything else and will be more important when Intel transitions to using their 10 nm node ahead of a process refresh from AMD.

 

[coercitiv]

If you consider what happened when people went from spinners to SSDs, and if Optane can basically do that again for the end-user . . . it could be trouble for AMD.

Or NVMe drives may short-circuit the whole Optane revolution. Who knows?

I doubt Optane vs. SSD will "feel" the same way for users as SSD vs. HDD, especially with SSDs still being a moving target. It may be excellent for enthusiasts and professionals though.

 

[mikk]

15%+ refers to mobile chip performance, don't apply this to desktop SKUs or even the 6 core Coffee Lake-H SKUs.

Exactly this.


@topic, I believe Coffee Lake will stay on LGA 1151, the big question is will it be compatible with 100 series and 200 series of chipsets. I don't think 100 series will support Coffee Lake.

 

[LTC8K6]

If they don't get 10-15% for the desktop, then I don't see the point in releasing a CL desktop chip. Another 5% release would be bad. I'd almost rather they not release that, lower current 7 series prices, and wait until the next new chip is ready.

 

[PingSpike]

I know you may be trying to be funny.

But they also claim Kabylake was 15% faster than the predecessor. Can only be the case if its mobile, and it is more or less true in mobile. They haven't catered to Desktop marketing since Pentium 4 chips, basically.

If the P4 is what we get when they cater to the desktop, maybe the desktop is better off being ignored.

 

[PingSpike]

Are we actually expected to get 6 core desktop out of coffeelake or is that just a rumor?

 

[LTC8K6]

Are we actually expected to get 6 core desktop out of coffeelake or is that just a rumor?

That seems to be the idea. The question is when? Supposedly it is mobile in 2H17 and Desktop in 1Q18. Intel could release the desktop chips early though, in reaction to RyZen.
We also have the Socket 2066 HEDT chips coming with KL-X, etc.

 

[Arachnotronic]

That seems to be the idea. The question is when? Supposedly it is mobile in 2H17 and Desktop in 1Q18. Intel could release the desktop chips early though, in reaction to RyZen.
We also have the Socket 2066 HEDT chips coming with KL-X, etc.

They really can't release it early; if they could, they would have.

Chip development takes time and you can't rush the process just because you underestimated a competitor.

 

[thepaleobiker]

This just in today, looks like the 10nm manufacturing is off to a tougher than expected start

http://www.barrons.com/articles/BL-TB-54236

"
BlueFin Research Partners analyst Steve Mullane today warns that Intel (INTC) could experience some delay in its production of chips using the latest manufacturing technology, although production of existing parts appears to be doing a little better than Wall Street expects.
Mullane has been spending the week at the Society of Photographic Engineers, or SPIE, conference on “advanced lithography” that’s taking place in San Jose, California. He relates the chatter he’s heard about Intel running into issues with its production of chips with the smallest feature sizes being 10 billionths of a meter:
"

Regards,
Vishnu

 

[arandomguy]

Other market factors are also a consideration for release schedules beyond just whether or not a product is ready.

For Kabylake the information coming out was that it was ready for release much earlier than what ended up being the actual launch date because they wanted to further clear existing inventory. We know at least that retailers were already stocking Kabylake shipments a month out from the actual launch date leading to end users managing to buy them weeks before actual launch. Kabylake was clearly pushed back at least 1 month+ artificially. It wouldn't be unreasonable to question whether or not manufacturing and launch was artificially pushed back even one fully quarter to rely on the Q4 holiday to clear existing inventory.

Intel may have originally set release and manufacturing plans based upon market conditions that have changed. Rather than wanting to rely on Q3 and especially Q4 volume to clear old stock as with Skylake->Kabylake they may instead be worried that customers will instead just pick the competitor.

I wouldn't be surprised if Coffeelake's current road mapped release date is that they simply did not want Kabylake desktop to have <1 year lifespan on the desktop as well.

 

[LTC8K6]

They really can't release it early; if they could, they would have.

Chip development takes time and you can't rush the process just because you underestimated a competitor.

I think everyone underestimated RyZen. I don't think anyone outside of AMD could have estimated correctly.

I think Intel has the CL chips ready to go this year.
I think Intel waits on releases for reasons other than the chips not being ready, most of the time.

All of this stuff coming out in the next year or so has been in the pipeline for a while, imo.
It's all based off of the same basic core, and wouldn't have taken as long to develop, I think.
I think that none of it is a reaction to RyZen except as far as pricing and release dates.

If Intel is making a true chip design to react to RyZen, it is something new and won't be out for a while, imo.

 

[fleshconsumed]

That seems to be the idea. The question is when? Supposedly it is mobile in 2H17 and Desktop in 1Q18. Intel could release the desktop chips early though, in reaction to RyZen.
We also have the Socket 2066 HEDT chips coming with KL-X, etc.

The second question is at what price. If 4c8t 7700K goes for $350, 6c12t has got to be at least $450, and at that point Ryzen is still extremely appealing having 2 more cores for $50 less.

 

[Bouowmx]

Good pricing on Intel Coffee Lake would be 6-core Hyper-threading at about same price as current 4-core Core i7: 312 USD or 350 USD for multiplier-unlocked variant, and the rest of the lineup redefining what Core i[357] for desktops. Wishful thinking or no?

 

[LTC8K6]

The second question is at what price. If 4c8t 7700K goes for $350, 6c12t has got to be at least $450, and at that point Ryzen is still extremely appealing having 2 more cores for $50 less.

How fast would an improved Kaby Lake 6C/12T chip be though?
Probably pretty fast.

 

[dullard]

The second question is at what price. If 4c8t 7700K goes for $350, 6c12t has got to be at least $450, and at that point Ryzen is still extremely appealing having 2 more cores for $50 less.

The 6 core, 12 thread 6800K has been out for about 9 months. It was originally priced at $434, but the street price has been $380 to $390 since November. I doubt that adding Ryzen as a competitor would get Intel to RAISE the price of the next version to $450+.

At least Intel is FINALLY coughing up more cores. They're late, though.

Unless you include the 6 core 12 thread 6800K that has been out since May 2016.

 

[fleshconsumed]

Hrm, forgot about 6800K. Still, that CPU is $424 (list price $450) on newegg right now, requires expensive 2011 motherboard, and still has 2 less cores than 1700X. Looks like intel will finally have to compete for the first time in nearly 10 years.

 

[LTC8K6]

Have to see what pricing is in a week or so.

There are several 2011-V3 boards under $200 these days, which is not that expensive.
An Asrock board is on sale for $159 at newegg, $129 AR.

 

[Insomniator]

A 6 core 7700k that could clock to 4.7ghz+ would be sick, I'd probably buy that over an 8 core Ryzen.

I'm intrigued by the 6800k, especially since I could get it for $360 at Microcenter, but it just doesn't seem good enough vs Ryzen even at that price.

 

[Dresdenboy]

So perhaps a QTR before Zen+ is due?

Maybe...

 

[gotd]

Intel's strategy and marketing departments dropped the ball. If I was an Intel shareholder, I'd be pissed. This could go down as one of the biggest business fumbles of 2017. The big dog fell asleep and the fox got the chickens.

Lets review:

1) Intel has been more than capable of making 6 & 8 core processors for a long time now.

2) Intel knew Ryzen was coming out. Official Ryzen benchmarks looked good. AMD showed Ryzen beating the 6900K. Early leaked benchmarks were available and consistently showed good Ryzen performance. Prices were also leaked...even the average Joe could find them. All of the leaks showed significantly lower prices.

So, why didn't Intel think to release a well priced consumer grade 6 & 8 core Kaby Lake processors yearly this year, and stop Ryzen from ever gaining traction? Intel was capable. They had the tech. They knew Ryzen was coming. They have huge profit margins - which now is more evident by the price of Ryzen.

I think the answer, in a nutshell, is greed, possibly stupidity, or both. In more detail:

A) Intel underestimated AMD or over estimated their own chips.

B) Intel failed to see that the consumer market had reached a critical mass for the need (real or psychological) for 6 & 8 core consumer processors. Intel was greedy and starved the consumer market with expensive 4 core processors and motherboards for years.

C) Intel failed to act strategically. Even if B is true, Intel should have released a well priced 8 core Kaby Lake processor(s). If Intel had an 8-core processor they could always adjust the price up if Ryzen was a flop or lower the price if Ryzen was a success. I'm not counting Intel's enthusiast lineup here. It's expensive, and so are the motherboards.

In this scenario Intel's profit margins *might* have been lower. But if Ryzen was a success, Intel's profits would be lower anyway. At least in this case, Intel gives the customer quality & value before AMD forces them to - which makes AMD the hero, and Intel the price gouging villain. It looks better if Intel did good for the customer of their own volition.

If Intel beat AMD to a well priced 8 core consumer processor, then:

1) People would not be flocking to Ryzen like it was the 1849 gold rush.
2) There would have been minimum or no Ryzen hype. Ryzen would be the "alternative" not Intel.
3) Intel would have a better public image. Intel's customers would have been happy, rather than resentful for being price gouged for the last 5 years and getting expensive 4-core CPU's with cheap thermal compounds.
4) It would have kept loyal Intel customers loyal.
5) Intel would have been acting/leading, rather than reacting/following.

I like Intel processors. I don't like Intel's prices or constantly changing CPU sockets. Intel has to come back to the consumer market humble and offer customers better value and performance than Ryzen. Maybe even make up for the years of gouging us. Here is what I suggest:

6-core @ 4.5Ghz for $199
8-core @ 4.2Ghz for $299
12-core @ 4.0Ghz for $399

... along with a more stable socket/platform.

This will give AMD a real challenge. However, timing is critical. Once people switch to Ryzen, they won't be buying Intel (or any other processors) for at least 2-3 years, if not more...even if Intel offers better price & performance.

 

[LTC8K6]

By the time the prices and benches were leaked, there was nothing Intel could do, imo.

Intel would have needed to decide to make the move to 6/8 core mainstream CPUs on their own, a long time ago, without reference to AMD's upcoming effort, in order to be in a good position when the RyZen benches leaked.

We haven't seen the official reviews and benches yet. RyZen may still only be the cheaper alternative to Intel.

I have had no problems with Intel. I am a happy Intel customer. Now then, if Intel's new chip design is not impressive, whenever we get it, then I will be upset with Intel.

I have also had no problems with AMD, and am happy with the AMD chips I have used to build systems recently.

 

[dullard]

Here is what I suggest:

6-core @ 4.5Ghz for $199
8-core @ 4.2Ghz for $299
12-core @ 4.0Ghz for $399.

... along with a more stable socket/platform.

Your whole post boils down to you want (a) cheaper CPUs and (b) you want to reuse motherboards. That was quite a lot of ranting for two simple points.

Intel will either lower their CPU price, or will suffer at the enthusiast level. I personally suspect that Intel will lower their price, but I could be wrong. Ryzen is not out yet, so of course Intel won't yet respond. Probably Intel won't do much until Skylake-X comes out this summer.

As for reusing motherboards, you are in such a minority that Intel doesn't care. Even among enthusiasts most people get a new motherboard with new CPU generations. Yes, it is a nice-to-have. But, you are now putting a top-of-the-line processor into a motherboard with ports that are all out of date. Remember, Intel's customers are Dell, HP, Lenovo, etc who want to sell complete systems. You are not Intel's customer focus.

If you need or wanted more than 4 cores, Intel has had it since last spring. It isn't like Intel "failed to see the need". And the price point of ~$380 for the last few months isn't a deal breaker either for those who need the cores. In reality, the 10 nm process is delayed, so Intel didn't have another line ready in time for Ryzen. Plain and simple. They didn't underestimate things, or fail to act strategically. They were just hit with delays.