Discussion Intel current and future Lakes & Rapids thread

[IntelUser2000]

@beginner99 AMD is also behind on the GPU side. Comparing to ARM vendors they are closer, but Apple's advantages are phenomenal.

Or is it that AMD/Intel's execution has been lacking at best? That's also possible. Intel should have had Golden Cove cores now! The argument that the advantage is due to ISA starts to dissolve when you see that ARM chips have lead in CPU, GPU, and IO.

I'm glad Tigerlake supports LPDDR5, but it merely catches up with them. At one time, they were the one to lead the industry with memory and IO standards.

on his SQ1 and managed a score about 47% higher in fp throughput versus my old heavily-overclocked A10-7700k. My Snapdragon 855+ got a score that was ~41% faster.

ARM vendors never fell completely flat on their faces like AMD/Intel did. Sure, they had some less than optimal cores, but it still was faster than previous ones.

Imagine if Bulldozer was a genuine advancement for AMD. Or if Intel delivered 10nm in 2016. They both set themselves back YEARS.

 

[Thunder 57]

They haven't bothered to try breaking out from low powered devices, not failed to. And that aside, they clearly don't even need to go chasing 10s of watts per core to be performance competitive with those chips from Intel and AMD. The numbers speak for themselves, really.

So they just don't want the more lucrative market? Meh, not buying it. Again, come back to me when I can do real work an ARM CPU.

 

[beginner99]

@beginner99 AMD is also behind on the GPU side. Comparing to ARM vendors they are closer, but Apple's advantages are phenomenal.

Probably yes. But is Nvidia? Especially if you take away all the stuff apple GPU does not have? Not to mention the apple gpu is just a modified powerVR design.

 

[Richie Rich]

Man, you draw these conclusions using donkey logic so easy like you do when you compare 2.5 GHz ARM CPUs with SKL and Zen 2 and say A13 has 82% higher IPC than Sylake and Zen 2.

What is donkey logic? It's like covering your eyes and saying: if I can't see the lion, the lion can't see me either.

It's basic math. If you have finished high school you should be able understand things like multiplying, dividing and percentage math. I understand when somebody is out of arguments, his ego is hurt badly and he tend to attack others with "donkey logic". Same for Thunder57. However as far as understand I don't accept such a behavior. What about to stay in polite way of communication and using arguments instead verbal attacks?

Take it easy. It's new year, man.

Cooling Rocket Lake-S @ 4.9 GHz would be more than most consumers would want. Custom water required, thanks to the power draw/heat output.

OK. What is your estimation of Rocket Lake clocks? The point is that RCL@ 4.9 GHz will be faster than any SKL/KBL @ 5.3 GHz and Zen3@4.8 GHz. Even RCL @4.6GHz would be still faster (with better power consumption than KBL). Zen3 power efficiency is out of RCL league (especially for 12c and 16c chips). But ST performance wise a RCL could be king of desktop.

And yet they have largely failed to breakout from low power devices.

You are literally lying here. Apple never tried to break out from low power.

So they just don't want the more lucrative market? Meh, not buying it. Again, come back to me when I can do real work an ARM CPU.

• Global x86 server market revenue 2018 ....... 81 billion $
• Global smartphone market revenue 2018 .... 552 billion $ (7x bigger market, Apple has 66% of those revenues: 360 billion)
• just for information: Global games market 2018 ... 135 billion $ (ARM platforms has 55% share: Nintendo Switch+smartphones)

Where did you get servers are more lucrative market than smartphone? Man, you spreads a lot of misinformation.
x86 is loosing at the most profitable digital markets in favor of ARM. You may not like it, but that's all you can do about it.

IDC: Worldwide X86 Server Market Revenue Increased by 15.4% in 2018, with DELL, HPE and Inspur Ranked in the Top Three

nestle pure life

en.inspur.com

Global smartphone sales reached $522 billion in 2018

GfK | Find press releases about GfK's offering, corporate news, financial results & latest news on data analytics, business intelligence. Subscribe for updates.

www.gfk.com

 

[Exist50]

Apple's larger than AMD and Intel combined several times over. So why do you think they're missing out on much?

Edit: Meant as a reply to Thunder 57

 

[IntelUser2000]

So in most applications, the single thread performance of the 1065G7 Icelake CPU is 40-50% ahead of the 8500Y Amberlake CPU.

Also, the 8500Y Amberlake fails to outperform the previous generation 7Y75 Kabylake-Y chip. This is due to Amberlake in reality not reaching the 4.2GHz Turbo, and being closer to the KBL-Y's 3.6GHz Turbo.

There's one test that shows gain for 8500Y over 7Y75, and the 1065G7 is only 20% faster. And that's Geekbench.

Actually, if you compare the Integer score, the 1065G7 is only 8% better than the 8500Y. Compared to the 7Y75, 8500Y is upwards of 20% faster.

So it behaves like a responsiveness test when web browsing. The test is run for such a short time, it allows the CPU to stay in a state that's normally not sustained for regular PC usage scenarios. It doesn't stress parts of the CPU enough. This makes it a mediocre test. It's a thermal version of Dhrystone.

 

[DrMrLordX]

OK. What is your estimation of Rocket Lake clocks?

Stock? Base/boost oscillating around 4 GHz @ 125W TDP for an 8c part. Where exactly the base and boost will be is unknown to me, but my guess is all-core clocks will sit around 4 GHz. And I think you're far too optimistic about how far ahead Willow Cove is/will be versus Zen3/Vermeer. Willow Cove will probably see the biggest lead in integer workloads (ironically). Overall, 3% lead for Rocket Lake-S per clock at best. And I have my doubts about that.

 

[lobz]

It's basic math. If you have finished high school you should be able understand things like multiplying, dividing and percentage math. I understand when somebody is out of arguments, his ego is hurt badly and he tend to attack others with "donkey logic". Same for Thunder57. However as far as understand I don't accept such a behavior. What about to stay in polite way of communication and using arguments instead verbal attacks?

Take it easy. It's new year, man.


OK. What is your estimation of Rocket Lake clocks? The point is that RCL@ 4.9 GHz will be faster than any SKL/KBL @ 5.3 GHz and Zen3@4.8 GHz. Even RCL @4.6GHz would be still faster (with better power consumption than KBL). Zen3 power efficiency is out of RCL league (especially for 12c and 16c chips). But ST performance wise a RCL could be king of desktop.



You are literally lying here. Apple never tried to break out from low power.

• Global x86 server market revenue 2018 ....... 81 billion $
• Global smartphone market revenue 2018 .... 552 billion $ (7x bigger market, Apple has 66% of those revenues: 360 billion)
• just for information: Global games market 2018 ... 135 billion $ (ARM platforms has 55% share: Nintendo Switch+smartphones)

Where did you get servers are more lucrative market than smartphone? Man, you spreads a lot of misinformation.
x86 is loosing at the most profitable digital markets in favor of ARM. You may not like it, but that's all you can do about it.

IDC: Worldwide X86 Server Market Revenue Increased by 15.4% in 2018, with DELL, HPE and Inspur Ranked in the Top Three

nestle pure life

en.inspur.com

Global smartphone sales reached $522 billion in 2018

GfK | Find press releases about GfK's offering, corporate news, financial results & latest news on data analytics, business intelligence. Subscribe for updates.

www.gfk.com

Not calling you a donkey no attacks there. It's just your logic in the specific area of comparing 2 different ISAs in a single workload leading you to think A13 has a 82% higher IPC than Skylake and Zen2 is what follows that of a donkey's. But since it's new year anyway, I'll go finish high school and unhurt my ego or whatever you're suggesting here, and I'll get back to you in a hundred years or so with an argument.

 

[Kaloi48]

New Socket for Alder Lake-S:
https://www.ptt.cc/bbs/PC_Shopping/M.1577894663.A.FF2.html


The package size is larger than the current 37.5 mm × 37.5 mm

 

[IntelUser2000]

New Socket for Alder Lake-S:
https://www.ptt.cc/bbs/PC_Shopping/M.1577894663.A.FF2.html
View attachment 15185

The package size is larger than the current 37.5 mm × 37.5 mm

The AM4 package is larger at 40mm x 40mm. So they might be increasing it to make space for the MCM configuration they'll start using.

The pin count increase also seems quite drastic. Might see more PCH functions moving on-package. AMD doesn't need the PCH on the laptop, but on desktops it has one to support extra functions.

 

[Thunder 57]

You are literally lying here. Apple never tried to break out from low power.

I was talking about ARM in general.

• Global x86 server market revenue 2018 ....... 81 billion $
• Global smartphone market revenue 2018 .... 552 billion $ (7x bigger market, Apple has 66% of those revenues: 360 billion)
• just for information: Global games market 2018 ... 135 billion $ (ARM platforms has 55% share: Nintendo Switch+smartphones)

Where did you get servers are more lucrative market than smartphone? Man, you spreads a lot of misinformation.
x86 is loosing at the most profitable digital markets in favor of ARM. You may not like it, but that's all you can do about it.

IDC: Worldwide X86 Server Market Revenue Increased by 15.4% in 2018, with DELL, HPE and Inspur Ranked in the Top Three

nestle pure life

en.inspur.com

Global smartphone sales reached $522 billion in 2018

GfK | Find press releases about GfK's offering, corporate news, financial results & latest news on data analytics, business intelligence. Subscribe for updates.

www.gfk.com

OK, better margins.That's what I meant. If Apple has such a large share of such a large market, then why would they bother scaling up and attacking a new market? Isn't that what you're trying to say? Why should x86 be so worried? Perhaps the biggest question though, is why do you turn any thread you can into an ARM/Apple vs x86 thread. It's annoying.

Maybe when Zen 3 comes out and it has SMT4, I'll start to believe you when it comes to other things. Hint: It won't. BTW, that would be spreading misinformation. AMD has shown us that Zen 3 will be max two threads/core.

 

[mikk]

New Socket for Alder Lake-S:
https://www.ptt.cc/bbs/PC_Shopping/M.1577894663.A.FF2.html
View attachment 15185

The package size is larger than the current 37.5 mm × 37.5 mm

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.

 

[liahos1]

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.

could this be a 2021 product?

 

[beginner99]

Where did you get servers are more lucrative market than smartphone? Man, you spreads a lot of misinformation.
x86 is loosing at the most profitable digital markets in favor of ARM

revenue doesn't mean it's profitable. I can make 1 trillion revenue and still loose money. Server / enterprise market has far higher margins.

 

[repoman27]

There seems to be a lot of stock put in sharkbay's posts despite the fact they appear to be speculation, and the information presented either inaccurate or at least highly unlikely. We pretty much know everything about Comet Lake at this point, and CML-S should be released sometime this quarter, most likely between now and January 22. Somehow Intel has convinced the media and rumor mills that there might be actual upgrades involved with the 400 Series Comet Point PCH. There aren't. CMP-LP (400 Series On-Package PCH) is CNP-LP (14nm 9th Generation Platform I/O Controller), CMP-H (400 Series PCH) is CNP-H (14nm 300 Series PCH), and CMP-V (400 Series "Value" PCH) is KBP-H (22nm 200 Series PCH). There are zero differences in supported features, package size, and ball count between these products.

There is no chance that the Comet Point PCH magically supports DMI x8. The on-package I/O controllers use OPI x8 at 4 GT/s because a wider, lower-power interface makes sense on package. If Intel wanted to double the bandwidth of DMI, I'm sure they'd prefer to move to DMI 4.0 x4 rather than DMI 3.0 x8. However, that would require both the CPU and PCH to support PCIe Gen4. Tiger Lake and Rocket Lake apparently will have PCIe Gen4, but it is rather dubious as to whether TGL PCH-LP (TGP-LP) or ICL PCH-H (ICP-H), paired with TGL-U/Y and RKL-H/S respectively, will as well. My money is on ICP-H in particular being DMI 3.0 x4 and PCIe Gen3 only, seeing as it is probably just another minor 14nm stepping of CNP, much like ICP-LP was.

The leaked slides regarding LGA 1200 state that: "Comet Lake LGA improves power delivery & support for future incremental I/O features." Driver evidence shows pretty clearly that the RKL-S IGP will be Gen12 LP GT1 with 32 EUs, and that Gen12 has provisions for several additional DDI interfaces. So the smart money would be on the extra pins actually being for power delivery and display interfaces, not DMI x8. And to further expand on the driver leaks, the following seems rather clear:

TGL GT2, Gen12 LP, 96 EUs
RKL GT1, Gen12 LP, 32 EUs
DG1 GT2, Gen12 LP, 96 EUs
ATS GT1, Gen12 HP, 128 EUs
DG2, Gen12 HP, 128/256/512 EUs

Rocket Lake will probably be built from three different 14nm dice: 6+1 LP, 6+1 HP, and 10+1 HP. RKL-U will be available in 6+1 or 6+2 configurations with the latter being an MCP including DG1 connected via PCIe Gen4 x4 using conventional package traces. RKL-S will still come in 2, 4, 6, 8, and 10 core versions, just like the roadmaps indicated, and the cores will be the same old Skylake cores on a ring bus. Intel is simply stepping the CML-S 6 and 10 core dies with Gen12 LP GT1 graphics and PCIe Gen4. Intel is not going to design a client die without an IGP because that would be an unsellable piece of silicon. They are not going to move to a "chiplet" design on 14nm, but they also can't make a financially viable die with a larger GPU. This isn't really a problem on the desktop, but not being able to deliver something at least comparable to Tiger Lake in laptops might be. Arctic Sound will be the first discrete Xe add-in board and will be a decidedly entry level affair.

 

[Markfw]

revenue doesn't mean it's profitable. I can make 1 trillion revenue and still loose money. Server / enterprise market has far higher margins.

Yes, smartphones are almost loss leader to get the monthly service charges ! You buy the phone slightly above, to below cost.

Server CPUs are a totally different story. I don;t know the markup for sure, but lets say the $7500 CPU could be sold for $1000 and still make money including R&D costs included. Don't hold me to that figure its an educated guess, but the markup is very high on server CPUs, Xeons included. Which is why Intel can sell them when they really want some deal at 80% off and still make money.

 

[FriedMoose]

I can agree on this, because 14nm clocks are reached after 4 years of refinement on the same uarch and same process.

Look how long it took them to actually reach 5GHz on an overclock. 4790K claimed it and failed. Same with 6700K, and then 7700K. Only by 8700K it started being viable, but it took one more iteration.

Rocketlake changes the cores, which will have different characteristics and no longer benefit from the years of tiny modifications they did to reach 5GHz.



~2x for CPU, ~2.5-2.6x for iGPU.

The primary changes to 14nm are taller, straighter fins and decreased density. Taller fins perform better at high frequencies and decreasing density allows higher clocks. I'm not sure how either of those would no longer benefit Willow Cove.

 

[FriedMoose]

Cooling Rocket Lake-S @ 4.9 GHz would be more than most consumers would want. Custom water required, thanks to the power draw/heat output.



14nm has its own issues when trying to run huge dice at high clocks - look at Skylake-X and Cascade Lake-X. You seen how much power those things can draw above 4.5 GHz? It's massive. I imagine than an 8c Rocket Lake-S (there probably won't be a 10c part) would be pushing a lot of heat.

You can cool Cascade Lake-X parts with 280mm and 360mm AIOs. Tweaktown overclocked their 18 core 10980XE to 4.8 GHz all core on an H150i without thermal throttling:

Intel Core i9 10980XE (Cascade Lake-X) Processor Review

Intel's Core i9 10980XE CPU gets fully tested, and now we can finally show you just how it performs.

www.tweaktown.com

How is this possible? Because thermal density is the main bottleneck for cooling modern CPUs. All this talk about how hard Rocket Lake will be to cool is completely meaningless without die size figures.

 

[mikk]

There seems to be a lot of stock put in sharkbay's posts despite the fact they appear to be speculation, and the information presented either inaccurate or at least highly unlikely.

Yes because he is a known proven source. He leaked leaked so much infos in the past, it's impossible that somebody without real infos could have shared all of this.

There is no chance that the Comet Point PCH magically supports DMI x8.

Where did he explicitly claimed CML PCH supports DMI x8 with CML-S? My understanding was that DMI x8 comes with the TGL PCH. I think he told CML PCH is x4 and he also told x8 comes with a future PCH, see below.

CML PCH DMIx4 這樣是多少寬
未來還有一次PCH升級 把當初封閉的DMIx4開回來


I'm sure you are wrong with RKL getting the old Skylake cores, you also forgot AVX 512 but I guess you will claim sharkbay is wrong.

 

[Markfw]

You can cool Cascade Lake-X parts with 280mm and 360mm AIOs. Tweaktown overclocked their 18 core 10980XE to 4.8 GHz all core on an H150i without thermal throttling:

Intel Core i9 10980XE (Cascade Lake-X) Processor Review

Intel's Core i9 10980XE CPU gets fully tested, and now we can finally show you just how it performs.

www.tweaktown.com

How is this possible? Because thermal density is the main bottleneck for cooling modern CPUs. All this talk about how hard Rocket Lake will be to cool is completely meaningless without die size figures.

The only problem with that review is there are NO 3000 series AMD HEDT chips in the review, against the current Intel HEDT chips.

 

[liahos1]

The only problem with that review is there are NO 3000 series AMD HEDT chips in the review, against the current Intel HEDT chips.

do you think they would overclock better than the 10980xe (the content you were replying to)

 

[repoman27]

Yes because he is a known proven source. He leaked leaked so much infos in the past, it's impossible that somebody without real infos could have shared all of this.

Where did he explicitly claimed CML PCH supports DMI x8 with CML-S? My understanding was that DMI x8 comes with the TGL PCH. I think he told CML PCH is x4 and he also told x8 comes with a future PCH, see below.

CML PCH DMIx4 這樣是多少寬
未來還有一次PCH升級 把當初封閉的DMIx4開回來

I'm sure you are wrong with RKL getting the old Skylake cores, you also forgot AVX 512 but I guess you will claim sharkbay is wrong.

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:

又是我 閒聊仔來啦 颱風前夕貼些小東西
原則上 2020/2021 桌機上會買到的東西大蓋是這樣
CML CPU : PEG PCIe Gen3, IGD Gen9, DMI x4, HDMI 2.0 via LSPCON
RKL CPU : PEG PCIe Gen4, IGD Gen12, DMI x8, HDMI 2.0 (Native)

CMP-V PCH : DMI x4, ME 14.0 (re-package from Kabylake PCH)
CMP-H PCH : DMI x8, ME 14.0
TGP-H PCH : DMI x8, ME 14.1

買板子的時候 可要注意一下 DMI拉幾條....
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.
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!
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.
颱風前夕 大家隨意聊 再次重申:<這幾年閉著眼睛買AMD應該都沒錯>

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.

 

[liahos1]

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:

又是我 閒聊仔來啦 颱風前夕貼些小東西
原則上 2020/2021 桌機上會買到的東西大蓋是這樣
CML CPU : PEG PCIe Gen3, IGD Gen9, DMI x4, HDMI 2.0 via LSPCON
RKL CPU : PEG PCIe Gen4, IGD Gen12, DMI x8, HDMI 2.0 (Native)

CMP-V PCH : DMI x4, ME 14.0 (re-package from Kabylake PCH)
CMP-H PCH : DMI x8, ME 14.0
TGP-H PCH : DMI x8, ME 14.1

買板子的時候 可要注意一下 DMI拉幾條....
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.
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!
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.
颱風前夕 大家隨意聊 再次重申:<這幾年閉著眼睛買AMD應該都沒錯>

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.

hdmi 2.0 in 2021 smh.

 

[Markfw]

do you think they would overclock better than the 10980xe (the content you were replying to)

The review was about performance of the chips, including overclocking. Since the review does not include its actual competitor, I say its flawed, overclocking aside. IF its competitor wins everything by a landslide (which is quite possible) even when the 10980XE is overclocked, then its even MORE important that those chips are included in the review. From what I have seen, thats actually the case. I would have to find the review I saw that in.

 

[liahos1]

The review was about performance of the chips, including overclocking. Since the review does not include its actual competitor, I say its flawed, overclocking aside. IF its competitor wins everything by a landslide (which is quite possible) even when the 10980XE is overclocked, then its even MORE important that those chips are included in the review. From what I have seen, thats actually the case. I would have to find the review I saw that in.

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.