Discussion Intel current and future Lakes & Rapids thread

[Markfw]

everything is a strong word. Did you look at 3950x min frames vs 9900k or cascade in that review?

what if my main use case is gaming with some excel and Bloomberg thrown in when not gaming.

The review is about the new Intel HEDT chips. Why is gaming important ? And even if it is, without the new AMD HEDT chips in the review, even gaming as an additional task makes the review non-relevant.

And one more note. I know this thread is about Intel, and the review is about Intels newest HEDT chips. What I am trying to point out, is that that review is seriously flawed, as it does not use its competitor in the review.

 

[ondma]

everything is a strong word. Did you look at 3950x min frames vs 9900k or cascade in that review?

what if my main use case is gaming with some excel and Bloomberg thrown in when not gaming.

Seems gaming is being relegated to a corner case by many in this forum, along with debasing single core performance and overclocking/clockspeed advantages.

 

[jpiniero]

could this be a 2021 product?

Earliest would be a year after Rocket Lake-S ships, so sometime in 2022. That's why it was cancelled, because they may as well just wait for 7 nm.

 

[lobz]

Seems gaming is being relegated to a corner case by many in this forum, along with debasing single core performance and overclocking/clockspeed advantages.

But as I see, playing the victim is absolutely not.

 

[Adonisds]

Have you seen this?

Engadget visited Intel fabs and talked about EUV. They make good videos. This one is great and fun.

At one point it seems he suggests Intel is behind others in EUV masks. I don't know if that's accurate or if there was some misunderstanding

 

[jpiniero]

At one point it seems he suggests Intel is behind others in EUV masks. I don't know if that's accurate or if there was some misunderstanding

It's accurate - Samsung is shipping EUV products right now (albeit at bad yield although likely much better than Intel's 10 nm) and TSMC will be producing very soon.

 

[Markfw]

Seems gaming is being relegated to a corner case by many in this forum, along with debasing single core performance and overclocking/clockspeed advantages.

This review was about an Intel HEDT chip. Why is gaming so important ? And why does overcloking/clockspeed advantages have anything to do with a review that is flawed ? I certainly an not overlooking anything, but you seem to be ignoring reality. Total performance of an Intel HEDT chip vs their competitor,

 

[Adonisds]

It's accurate - Samsung is shipping EUV products right now (albeit at bad yield although likely much better than Intel's 10 nm) and TSMC will be producing very soon.

No one disagrees that Intel is behind and won't ship EUV products soon. I think I've been hearing that they didn't want to move to EUV soon because they wanted to push DUV to the limits first, probably because EUV machines have low throughput for now.

Somewhere in the video the dude says that Intel hasn't figured how to create the transparent material that protects the EUV masks yet, unlike the others. So Intel couldn't have moved to EUV even if they wanted. But he probably isn't an expert in manufacturing and it seems weird that Intel would say something like that, so it could be a misunderstanding.

 

[NostaSeronx]

TSMC will be producing very soon.

TSMC is already shipping: "The Kirin 990 is crafted by 7nm technology while the 5G version is crafted by 7nm+ EUV technology."

HUAWEI Global

Explore the latest technologies in huawei smartphone, wearable, PC, tablet, audio

consumer.huawei.com

806.23US $ |Honor V30 Pro Smartphone Google Play Kirin990 5g 7nm Octa Core 128gb 256gb 16core Gpu 40mp Triple Cam 40w Supercharge Android10 - Mobile Phones - AliExpress

Smarter Shopping, Better Living! Aliexpress.com

www.aliexpress.com

 

[jpiniero]

because EUV machines have low throughput for now.

When 10 nm was originally supposed to go HVM, that was the case yes.

 

[Hitman928]

No one disagrees that Intel is behind and won't ship EUV products soon. I think I've been hearing that they didn't want to move to EUV soon because they wanted to push DUV to the limits first, probably because EUV machines have low throughput for now.

Somewhere in the video the dude says that Intel hasn't figured how to create the transparent material that protects the EUV masks yet, unlike the others. So Intel couldn't have moved to EUV even if they wanted. But he probably isn't an expert in manufacturing and it seems weird that Intel would say something like that, so it could be a misunderstanding.

He talks about it at about the 14:50 mark.

 

[mikk]

So, to be clear, my ability to read Chinese is extremely limited, so I'm relying on Google Translate. Also, I don't follow the rumors all that closely and just dip in from time to time. I've seen plenty of posts by sharkbay on ptt.cc bbs, but never any where there was significant information being divulged. That doesn't mean they don't exist, just that I haven't seen them. By contrast, I've seen chrisdar posts on the same site which often include specific information that clearly came from Intel documentation. I'd be happy to check out any posts you can link to where sharkbay delivered the goods.

However, the specific sharkbay post I was referring to was this: https://www.ptt.cc/bbs/PC_Shopping/M.1565273639.A.F64.html Which states:

This is not accurate. CMP-H is the same exact silicon as CNP-H which does not have DMI x8, and RKL was slated to use ICP-H not TGP-H.

In this posting he says CML-S only supports DMI x4 while RKL-S supports DMI x8, because of that it's really hard to figure out without RKL-S. If this is accurate Intel won't promote CML-H with DMI x8 therefore. And RKL can use both CMP-H and TGP-H, it can not work with CMP-V:

如果不小心買到 CMP-V PCH 的板子, 那可是不能升級RKL-S CPU的喔.
If you accidentally buy a CMP-V PCH board, then you cannot upgrade the RKL-S CPU.

當一張板子被告知要設計成 CML-S CPU+CMP-H PCH的時候, DMI可能只拉4條.
When a board is told to design a CML-S CPU + CMP-H PCH, the DMI may only pull four.

到時後換RKL-S CPU的時候寬度就是人家滿配的一半. 要注意喔!
When the time is changed, the width of the RKL-S CPU is half of the full complement. Be careful!

Where did you see RKL using ICP-H? I haven't heard a single word from Intel about Rocket Lake. I'm not surprised you don't follow the rumors that closely, otherwise you would have known that sharkbay is a proven source and so far everything he said has been confirmed and while other leaks suggested there is a LGA1159 coming for CML-S he leaked the LGA1200 naming which in the end was accurate.

 

[Adonisds]

TSMC is already shipping: "The Kirin 990 is crafted by 7nm technology while the 5G version is crafted by 7nm+ EUV technology."

HUAWEI Global

Explore the latest technologies in huawei smartphone, wearable, PC, tablet, audio

consumer.huawei.com

806.23US $ |Honor V30 Pro Smartphone Google Play Kirin990 5g 7nm Octa Core 128gb 256gb 16core Gpu 40mp Triple Cam 40w Supercharge Android10 - Mobile Phones - AliExpress

Smarter Shopping, Better Living! Aliexpress.com

www.aliexpress.com

When 10 nm was originally supposed to go HVM, that was the case yes.

Yeah but if Intel wanted to skip 10nm and go to 7nm EUV like some suggested a while back, the throughput now would probably still be too low because Intel 7nm is a bit denser than TSMC 5nm, so there's probably many more EUV layers than SS and TSMC 7nm EUV. Plus Intel is huge, they wouldn't only make one smartphone SOC

 

[lobz]

Plus Intel is huge, they wouldn't only make one smartphone SOC

I got lost here.

 

[Thunder 57]

Seems gaming is being relegated to a corner case by many in this forum, along with debasing single core performance and overclocking/clockspeed advantages.

Maybe because 1080p with a 2080 Ti gaming at high refresh rates is the only thing Intel can still claim to have an advantage in. Clock speed, wow! Fact is there are only edge cases where Intel still wins these days. Why try to be an Intel marketing forum poster when they just aren't as competitive?

 

[scannall]

Seems gaming is being relegated to a corner case by many in this forum, along with debasing single core performance and overclocking/clockspeed advantages.

If it were a consumer grade enthusiast part, then gaming and overclocking should certainly be more relevant. And if all you do is game, then a 9900KS is probably the best part for that market right now by a little bit anyway. But this is a HEDT part, aimed more at professional uses. Gaming should be reduced in importance, because that isn't the target audience.

It shouldn't be completely disregarded, as even professionals like to fire up a game or two from time to time. But the focus should be on how much work the product can get done, with the software you use to make money.

 

[DrMrLordX]

I mean the existence of ADL-S is no surprise anymore because of all the driver additions recently, even though some people who all the time tried to write down ADL-S (or 10nm desktop in general) with "probably NUC or Atom only" should rethink.

Nothing here negates or disproves the idea that Alder Lake-S might wind up being an -R/-C part that shows up in some NUCs and never appears anywhere else. See Broadwell-C.

could this be a 2021 product?

Alder Lake is supposed to launch after Rocket Lake-S, which itself is a 2021 product. So I think it's safe to assume that the official launch for Alder Lake-S would be 2021.

You can cool Cascade Lake-X parts with 280mm and 360mm AIOs.

Ehhhhh, I'm not sold on that, at all. They were running handbrake, which is fine, but it's no Blender Benchmark or (gasp) Prime95. Or linpack. A 5 GHz 10980XE is somewhere around a 400-500W CPU! You don't run that on a 360mm AiO, not in a million years. 4.8 GHz is probably sitting in the 350-400W range and I would definitely not want to tame that with an AiO. Note they didn't show power consumption for the overclocked system, either. At stock they measured 242W from that thing in what was presumably handbrake (they never specified). My NH-D15 worked hard to handle 200-210W from my old 1800x, and it had powerful 3000rpm NF-A14 fans on it. There are plenty of AiOs out there that perform about as well as a stock NH-D15.

8c Rocket Lake-S will probably consume more power than 10c Comet Lake-S. The 10900k looks like it's going to be a 250W CPU if you want it to hit clocks like a 9900K (~4.7 GHz or so). That is a lot of heat. That's too much for most cooling solutions, and definitely too much for most air coolers. It's also too much for a lot of AiOs.

All this talk about how hard Rocket Lake will be to cool is completely meaningless without die size figures.

Pretty sure we already had a transistor count comparison between Ice Lake-U and Whiskey Lake-U earlier in this thread? Mind you, Rocket Lake is going to be Willow Cove, so expect transistor count to be higher. It's a bit difficult to extrapolate exact die size since (allegedly) Rocket Lake-S will move iGPU and SoC functions off the die.

It's accurate - Samsung is shipping EUV products right now (albeit at bad yield although likely much better than Intel's 10 nm) and TSMC will be producing very soon.

Um, TSMC has been shipping 7nm+ for awhile, haven't they?

 

[repoman27]

In this posting he says CML-S only supports DMI x4 while RKL-S supports DMI x8, because of that it's really hard to figure out without RKL-S. If this is accurate Intel won't promote CML-H with DMI x8 therefore. And RKL can use both CMP-H and TGP-H, it can not work with CMP-V:

如果不小心買到 CMP-V PCH 的板子, 那可是不能升級RKL-S CPU的喔.
If you accidentally buy a CMP-V PCH board, then you cannot upgrade the RKL-S CPU.

當一張板子被告知要設計成 CML-S CPU+CMP-H PCH的時候, DMI可能只拉4條.
When a board is told to design a CML-S CPU + CMP-H PCH, the DMI may only pull four.

到時後換RKL-S CPU的時候寬度就是人家滿配的一半. 要注意喔!
When the time is changed, the width of the RKL-S CPU is half of the full complement. Be careful!

Where did you see RKL using ICP-H? I haven't heard a single word from Intel about Rocket Lake. I'm not surprised you don't follow the rumors that closely, otherwise you would have known that sharkbay is a proven source and so far everything he said has been confirmed and while other leaks suggested there is a LGA1159 coming for CML-S he leaked the LGA1200 naming which in the end was accurate.

Well, it doesn't matter if RKL-S has DMI x8. CMP-H does not, because it is the same silicon as CNP-H, for which the datasheet is publicly available. And CMP-V is 22nm garbage, so if you buy that in 2020, you're getting what you deserve. But my point wasn't that sharkbay doesn't know anything, it was that several of the recent posts contain inaccurate info which has been pumped directly into the tech news echo chamber.

The Nov 28 sharkbay post on ptt.cc bbs, which seems to have been taken down, showed RKL as AVX-256. This was later corrected by sharkbay to AVX-512, so one of those two is probably wrong. And if RKL is a 14nm backport of TGL, then leaving the memory controller at DDR4-2933 is... improbable. The subtlety may be lost on me, but I think chrisdar was poking fun at that.

Unfortunately, we're almost always trading in imperfect information when it comes to these types of leaks, so some critical thinking is warranted. The roadmaps from the Tweakers leak have been more or less spot-on, despite containing obvious errata and inconstancies with the coloring and alignment of the boxes representing the various platforms. Those slides included the following tidbits regarding Rocket Lake:

Rocket Lake U 4/6C 14nm + 10nm gfx (Q3'20)
Rocket Lake Xeon E ICL PCH, PCI Gen 4 (Q1'21 launch)
Rocket Lake S 2/4/6/8/10C 14nm (Q2'21 SIPP)
Rocket Lake U 4/6C 14nm + 14nm gfx (Q2'21 SIPP) - listed twice, in both the 15-28W and 45W, ≥65W sections

Xeon E is the same as S-Series, so we can infer that consumer Rocket Lake S will also launch in Q1 2021, will be available in up to 10 cores, will have PCIe Gen4, and that the chipset will be based on ICP-H.

 

[DrMrLordX]

we can infer that consumer Rocket Lake S will also launch in Q1 2021, will be available in up to 10 cores

Oof, 10c Rocket Lake-S? We'll see. I guess Intel "has to do it" since Comet Lake-S will go up to 10c. But. The heat!

 

[repoman27]

Oof, 10c Rocket Lake-S? We'll see. I guess Intel "has to do it" since Comet Lake-S will go up to 10c. But. The heat!

From Intel, the company that brought you the Xeon W-3175X, it won't be that bad...

 

[DrMrLordX]

From Intel, the company that brought you the Xeon W-3175X, it won't be that bad...

So no industrial chiller? Okay, that's fine.

 

[repoman27]

Intel obviously sells a lot of different SKUs based on a relatively small number of silicon dies. To get a handle on the cadence with which Intel introduces new dies, I decided to enumerate them. This table is just the client platforms (no Atom or Xeon Scalable) introduced since Skylake in 2015, but also includes some anticipated products that are on the roadmap for the next 24 months.

Platform	Die	Cores	IGP	Process	HW Stepping	Model #	Stepping #	Launch Date
SKL-H/S 4+2	4+2 HP	Skylake	Gen9	14	R0	94	3	Aug '15
SKL-U/Y 2+2	2+2 LP	Skylake	Gen9	14	D1	78	3	Sep '15
SKL-U 2+3e	2+3 LP	Skylake	Gen9	14	K1	78	3	Sep '15
SKL-S 2+2	2+2 HP	Skylake	Gen9	14	S0	94	3	Sep '15
SKL-H 4+4e	4+4 HP	Skylake	Gen9	14	N0	94	3	Jan '16
KBL-U/Y 2+2, AML-Y 2+2	2+2 LP	Skylake	Gen9.5	14+	H0	142	9	Aug '16
KBL-U 2+3e	2+3 LP	Skylake	Gen9.5	14+	J1	142	9	Jan '17
KBL-S 2+2	2+2 HP	Skylake	Gen9.5	14+?	S0*	158	9	Jan '17
KBL-H/S/G/X 4+2	4+2 HP	Skylake	Gen9.5	14+	B0	158	9	Jan '17
KBL-U R 4+2	4+2 LP	Skylake	Gen9.5	14+	Y0	142	10	Aug '17
CFL-H/S 6+2, CFL-H/S R 6+2	6+2 HP	Skylake	Gen9.5	14++	U0	158	10	Oct '17
CFL-S 4+2, CFL-S R 4+2	4+2 HP	Skylake	Gen9.5	14++?	B0*	158	11	Oct '17
CFL-U 4+3e	4+3 LP	Skylake	Gen9.5	14++	D0	142	10	Apr '18
CNL-U 2+0	4+2 LP	Palm Cove	Gen10	10	?	102	3	May '18
WHL-U 4+2	4+2 LP	Skylake	Gen9.5	14++	W0	142	11	Aug '18
CFL-S R 8+2	8+2 HP	Skylake	Gen9.5	14++	P0	158	12	Oct '18
WHL-U 4+2, AML/CML-Y 4+2, CML-U 4+2 v1 (LP3)	4+2 LP	Skylake	Gen9.5	14++	V0	142	12	Apr '19
CFL-H/S R 8+2	8+2 HP	Skylake	Gen9.5	14++	R0	158	13	Apr '19
ICL-U 4+2	4+2 LP	Sunny Cove	Gen11	10+	D1	126	5	Aug '19
CML-U 6+2 v1 (LP3)	6+2 LP v1	Skylake	Gen9.5	14++	A0	166	0	Aug '19
CML-H/S 10+2	10+2 HP	Skylake	Gen9.5	14++				Jan '20
CML-U 6+2 v2 (LP4x)	6+2 LP v2	Skylake	Gen9.5	14++				Apr '20
TGL-U 4+2	4+2 LP	Willow Cove	Gen12 LP	10++				NET Apr '20
RKL-U 6+1	6+1 LP	Willow Cove	Gen12 LP	14++				Aug '20
RKL-H/S 8+1	8+1 HP	Willow Cove	Gen12 LP	14++				Jan '21

* I'm giving Intel the benefit of the doubt on these because the model and/or stepping numbers changed, however the S-Spec stepping is the same as the previous generation. Although there ostensibly should be hardware differences, these parts may well be based on prior generation silicon with only microcode/firmware updates.

Platform Chipset	Die	PCH	Chipset Name	Process	HW Stepping	Launch Date
SKL PCH-H	SPT-H	Sunrise Point	100 Series	22	D1	Aug '15
SKL PCH-LP	SPT-LP	Sunrise Point	6/7/8/10th Gen	22		Sep '15
KBL PCH-H, CML PCH-V	KBP-H	Kaby Point	200 Series	22	A0	Jan '17
CNL PCH-LP	CNP-LP	Cannon Point	300 Series	14		Apr '18
CNL PCH-H	CNP-H	Cannon Point	300 Series	14	B0	Apr '18
ICL PCH-LP, CML PCH-LP	ICP-LP	Ice Point	400 Series	14		Aug '19
ICL/CML PCH-H	ICP-H	Ice Point	400 Series	14		Jan '20
TGL PCH-LP, RKL PCH-LP	TGP-LP	Tiger Point	500 Series	14		NET Apr '20
TGL/RKL PCH-H	TGP-H	Tiger Point	500 Series	14		Jan '21

Update 1/4/20: Fix the RKL and PCH entries that I got all wrong.

 

[lobz]

Seems gaming is being relegated to a corner case by many in this forum, along with debasing single core performance and overclocking/clockspeed advantages.

Also, gaming could be an important case to point out in a HEDT review or stated as a purchase standpoint IF the competing part would be bad in gaming and intel not, or the intel part would be significantly better, but I mean... I dare you to point out the difference in actual gameplay between any intel HEDT CPU @ 5 GHz and a corresponding AMD CPU.

 

[jpiniero]

The rumors did specifically mention 8 cores max for Rocket Lake. That doesn't mean that is Willow Cove since they could just not have an i9 branded product; but does make it more likely.
Also there is no Rocket Lake-H apparently, since 35+ W U parts are replacing them. Again that sort of points to Willow Cove.

I think Intel could get away with Skylake but they would lose DIY for sure.

 

[mikk]

Nothing here negates or disproves the idea that Alder Lake-S might wind up being an -R/-C part that shows up in some NUCs and never appears anywhere else. See Broadwell-C.

So here we go again, another negating 10nm desktop posting. The track record is so terrible in this thread, it's safe to assume you are plain wrong. If there is ADL-S with a completely new socket you can be sure we will see real desktop CPUs and once again NUC devices are using mobile CPU variants, U or H models and they won't use LGA 1700. And BDW-C was an exotic mobile-H transferred to LGA and it was backward compatible with LGA 1150. BDW-S never existed and ADL-S won't be backward compatible with LGA1200.

Well, it doesn't matter if RKL-S has DMI x8. CMP-H does not, because it is the same silicon as CNP-H, for which the datasheet is publicly available.

You have to prove that RKL-S won't get DMI x8 with CMP-H which is hard to do. Until then you are unable to prove your point. In fact he said CML-S with CMP-H gets only DMI x4 which is accurate. It seems you are really desperate trying to find the tiniest spec just to write down sharkbay. Look this is wrong so he is unreliable in all he said. I won't agree with you because all he said so far has been confirmed later on.

The Nov 28 sharkbay post on ptt.cc bbs, which seems to have been taken down, showed RKL as AVX-256. This was later corrected by sharkbay to AVX-512, so one of those two is probably wrong.

As he told he looked in the wrong table and that's why he corrected it later.

And if RKL is a 14nm backport of TGL, then leaving the memory controller at DDR4-2933 is... improbable. The subtlety may be lost on me, but I think chrisdar was poking fun at that.

Memory speeds could change any time. CML-S was specified for DDR-2666 in all slides and leaks until the very end.