Sunday, March 22nd 2020

Intel Rocket Lake-S Platform Detailed, Features PCIe 4.0 and Xe Graphics

Intel's upcoming Rocket Lake-S desktop platform is expected to arrive sometime later this year, however, we didn't have any concrete details on what will it bring. Thanks to the exclusive information obtained by VideoCardz'es sources at Intel, there are some more details regarding the RKL-S platform. To start, the RKL-S platform is based on a 500-series chipset. This is an iteration of the upcoming 400-series chipset, and it features many platform improvements. The 500-series chipset based motherboards will supposedly have an LGA 1200 socket, which is an improvement in pin count compared to LGA 1151 socket found on 300 series chipset.

The main improvement is the CPU core itself, which is supposedly a 14 nm adaptation of Tiger Lake-U based on Willow Cove core. This design is representing a backport of IP to an older manufacturing node, which results in bigger die space due to larger node used. When it comes to the platform improvements, it will support the long-awaited PCIe 4.0 connection already present on competing platforms from AMD. It will enable much faster SSD speeds as there are already PCIe 4.0 NVMe devices that run at 7 GB/s speeds. With RKL-S, there will be 20 PCIe 4.0 lanes present, where four would go to the NVMe SSD and 16 would go to the PCIe slots from GPUs. Another interesting feature of the RKL-S is the addition of Xe graphics found on the CPU die, meant as iGPU. Supposedly based on Gen12 graphics, it will bring support for HDMI 2.0b and DisplayPort 1.4a connectors.
Intel Rocket Lake-S Platform
Some things like Direct Media Interface (DMI) will double the bandwidth and now there will be eight links present, compared to four of the previous platforms. Announced at CES 2020, ThunderBolt 4 will also be present along with USB 3.2 20G. Additionally, Intel Software Guard Extensions (SGX) have been removed to improve the security of the platform, as the SGX has proved to be quite vulnerable to many kinds of attacks and exploits. There are some updated media encoding standards as well, like 12-bit AV1/HEVC and E2E compression. Source: VideoCardz
Add your own comment

113 Comments on Intel Rocket Lake-S Platform Detailed, Features PCIe 4.0 and Xe Graphics

#27
ARF
londiste
That is not the problem. Ice Lake does pretty well. The changes we know about in Tiger Lake will put its single-core performance above Zen2, possibly competitive with Zen3.
Ice Lake is terrible, to be honest. Only 4 cores at 3.6 GHz on the N10+ node.


https://en.wikipedia.org/wiki/Ice_Lake_(microprocessor)

18% IPC improvement on average compared to the 2015 Skylake.


https://www.guru3d.com/articles_pages/amd_ryzen_7_3700x_ryzen_9_3900x_review,9.html

Which would put it ~6-7% above Zen 2 but Zen 3 is expected to bring 15-20% IPC improvement when it comes, finally.
Posted on Reply
#28
Alexandrus
Maybe don't take those silly Cinebench figures as a measure for IPC. Just a thought.
Posted on Reply
#29
Zareek
Didn't Intel downplay PCIe 4.0 when AMD announced it by saying they would be launching PCIe 5.0 with the next generation???
Posted on Reply
#30
efikkan
VideoCardz have made up stuff in the past. While some of this may not or may be accurate, I wouldn't care too much until we have confirmation from reliable third-parties, or official leaks of course.

TristanX
I doubt it will be backported to 14nm. Backporting CPU core and GPU core is way too much, especially that Willow Cove need large caches (for example for AVX 512). There may be two platforms at the same time: cheaper 400 series for big volume, and pricier 500 series for enthusiasts.
The caches of Skylake-X/-SP is pretty close in size and is not a problem on 14nm.

bug
Except Intel's problem is not architecture (Sunny Cove looks darn good on paper), it's fab capacity ;)
If what Intel is telling us is true, while 10nm has been a train wreck, 7nm which was worked on in parallel was not. And that would mean 7nm will hit sooner rather than later, making ramping up 10nm at this point rather unattractive. Big if, grain of salt and everything...
Intel is ramping up 10nm as much as possible, and 10nm+ arriving this year (for Tiger Lake and Ice Lake-SP/X) and 10nm++ next year. Demand is much higher than expected in both laptop and server segments, so we shouldn't expect 10nm to reach the full lineup. But 7nm is still far away, low volume 2021 and medium volumes 2022 is a best case scenario.
Posted on Reply
#31
ARF
What's the point in releasing Comet Lake 10-core in April 2020 and Rocket Lake 8-core in December or November 2020 ? :confused:
Posted on Reply
#32
londiste
ARF
18% IPC improvement on average compared to the 2015 Skylake.
Which would put it ~6-7% above Zen 2 but Zen 3 is expected to bring 15-20% IPC improvement when it comes, finally.
That was exactly my point. You said Intel needs 5GHz to stay competitive with AMD. It does not. Ice Lake - as you said - has better single-core performance than Zen2. Tiger Lake will bring some smaller improvements. That should put it on par with what we expect Zen3 to be.
Posted on Reply
#33
Vayra86
notb
Man... you're writing posts from mid 2019 at best. 10nm is profitable already and used for mainstream chips.
They're quite elusive for being mainstream buddy. Its pretty close to a paper launch with some samples out the door.

And profitable 10nm sure as hell is not. After all these years of development? lol

efikkan
Intel is ramping up 10nm as much as possible
Are there actual numbers of this? And not those of investments in fabs, but of 10nm products sold.
Posted on Reply
#34
Logoffon
I might sound dumb here, but what is the exact reason for Intel failing to release 10nm a year after SKL, and then continuously failing to make that node in high yields?
Posted on Reply
#35
Vayra86
Logoffon
I might sound dumb here, but what is the exact reason for Intel failing to release 10nm a year after SKL, and then continuously failing to make that node in high yields?
That is the million dollar question :D

But 'they were lazy' is not sufficient to me. I do think this is really a difficult node. 7nm didn't come through easily either, new effects come into play as we go this small. Effectively the fabs need new machinery for it (EUV). That also has its early problems.
Posted on Reply
#36
ARF
londiste
That was exactly my point. You said Intel needs 5GHz to stay competitive with AMD. It does not. Ice Lake - as you said - has better single-core performance than Zen2. Tiger Lake will bring some smaller improvements. That should put it on par with what we expect Zen3 to be.
How when Icelake simply doesn't clock anywhere near to the needed levels? You have 3.5 GHz - 4.0 GHz Icelake chips which are supposed to perform as well as the 4.7 GHz Ryzen 9 3900X !

Logoffon
I might sound dumb here, but what is the exact reason for Intel failing to release 10nm a year after SKL, and then continuously failing to make that node in high yields?
The exact reason is that it was a premature jump of 2.3x the transistor density over N14 and there is currently no available equipment that can do it.
Posted on Reply
#37
TheGuruStud
Intel putting out tigerlake quad core leaks...yeah, that's all you need to know about 2020 (poor attempt at fixing the narrative).
Posted on Reply
#38
ARF
TheGuruStud
Intel putting out tigerlake quad core leaks...yeah, that's all you need to know about 2020 (poor attempt at fixing the narrative).
If it depends on Intel, they will increase the single-threaded IPC by 100% and continue to offer the holy quad-cores for another 10 years, at least.
Didn't they officially say that we don't need anything more than a quad-core?!

But when I hear about a quad-core of any type, I become like this:

Posted on Reply
#39
londiste
ARF
[quote=Logoffon]I might sound dumb here, but what is the exact reason for Intel failing to release 10nm a year after SKL, and then continuously failing to make that node in high yields?
The exact reason is that it was a premature jump of 2.3x the transistor density over N14 and there is currently no available equipment that can do it.[/quote]What is generally being suspected are:
1. Some aspects or measures of Intel's 10nm are the same or even beyond what TSMC's 7nm has (MMP comes to mind out of the major ones). Intel was trying to get there without EUV (this is the equipment availability question) and that backfired.
2. Replacing copper with cobalt is suspected to be a big suspect due to the nature of early errors. Intel has been pretty quiet about this.
Posted on Reply
#40
notb
Vayra86
They're quite elusive for being mainstream buddy. Its pretty close to a paper launch with some samples out the door.
What?
Seriously, you're either not tracking the PC market at all or just ignoring mobile segment on purpose.

Ice Lake is available in normal, common notebooks. It's normally available in stores. And it's been there for a few months already.
You're repreating arguments from half a year ago or older. I don't understand why you're unwilling to just CHECK if they're still valid.
You really like Intel to fail and you're going to hold on to this narrative?
And profitable 10nm sure as hell is not. After all these years of development? lol
Why would it not be? Why would Intel offer Ice Lake for mainstream products if it wasn't making money?
And at the end of the day Intel's profit margin is still way higher than AMD (even higher than TSMC's). And it already includes those years of 10nm development with no sales.
You want to deny that as well?
Posted on Reply
#41
Bwaze
Not all 10th generation laptop processors are 10nm Ice Lake. In fact, most of them are not:

On Geizhals.eu there are 780 10-gen laptops, 527 Comet Lake-U and only 253 Ice Lake-U.

Ice Lake-U also caps at 4 core, 8 threads, and quite low frequency.
Posted on Reply
#42
Valantar
notb
What?
Seriously, you're either not tracking the PC market at all or just ignoring mobile segment on purpose.

Ice Lake is available in normal, common notebooks. It's normally available in stores. And it's been there for a few months already.
You're repreating arguments from half a year ago or older. I don't understand why you're unwilling to just CHECK if they're still valid.
You really like Intel to fail and you're going to hold on to this narrative?

Why would it not be? Why would Intel offer Ice Lake for mainstream products if it wasn't making money?
And at the end of the day Intel's profit margin is still way higher than AMD (even higher than TSMC's). And it already includes those years of 10nm development with no sales.
You want to deny that as well?
Sorry, but Ice Lake is available only in a rather limited number of SKUs, with Comet Lake making up the majority of current Intel-based ULV laptops. Ice Lake is absolutely out there, but if 10nm had reached high yields and ramped up properly, wouldn't they be reserving the desperately needed 14nm capacity for their server and workstation chips (where 10nm isn't used yet)? The only plausible reason for there still being a prevalent and widely available 14nm mobile lineup from Intel is that they are still struggling with that node.

As for profitability: recouping costs does not equal making money. Intel have spent several billion USD on developing 10nm, and are only now shipping in anything resembling volume. They will likely never make back their R&D costs on this process node, which means it will never be profitable.

Also: Intel's current, shipping, somewhat available 10nm node is ... not very performant. Ice Lake has a ~18% architectural IPC gain from SKL and its derivatives, yet Comet Lake (14nm ++++ SKL derivative) at 15W outperforms Ice Lake (10nm) even according to Intel itself. In other words the power and clock scaling disadvantages of the 10nm node are still significant enough that Intel can't make it faster than their previous offerings despite a significant IPC advantage. Have you seen how ridiculously low base clocks on Ice Lake chips are? Quad core ICL chips are barely faster at base clock than hexa core CML chips (1.3 v. 1.1 GHz). ICL is also the first ever generation where Intel has announced a top-end U-series SKU that is only available at 28W and nothing lower - the i7-1068G7, which btw has yet to ship in any device - and even at 28W it only manages a 2.3GHz base clock.

Ice Lake is looking more and more like a beta release, with Tiger Lake planned to be the proper, widely available and actually performant part. Though we'll see if they manage to refine the node enough for that to launch in a timely manner.

Alexandrus
Maybe don't take those silly Cinebench figures as a measure for IPC. Just a thought.
Do you trust SPEC2017? That's about as industry-standard as you get, and AnandTech's testing using it shows Zen 2 to be ~7% faster than Coffee Lake (~5% SPECINT, ~7.8% SPECFP, ~7.1% overall) (or, if you flip that around, CFL is ~6.7% overall/7.2% SPECFP/4.8% SPECINT slower than Zen 2).
Posted on Reply
#43
Vya Domus
Logoffon
I might sound dumb here, but what is the exact reason for Intel failing to release 10nm a year after SKL, and then continuously failing to make that node in high yields?
Sunk cost fallacy.

10nm should have been canned a long time ago, it's obvious that's never going to be a volume node, there isn't enough time left for that to happen and still stay competitive. The hallmark of a good node stands in server chips not mobile stuff, that's where yields/volume/performance are paramount. Since there are no 10nm Xeons right now, that should give you an idea of how useful this node is.

14nm could have carried them to next node, ironically that's what is happening right now. They tried to boost 14nm capacity and work on 10nm at the same time, that was a mistake as well.
Posted on Reply
#44
notb
Bwaze
Not all 10th generation laptop processors are 10nm Ice Lake.
Of course. I've never said all are.
On Geizhals.eu there are 780 10-gen laptops, 527 Comet Lake-U and only 253 Ice Lake-U.
That's a very unorthodox way of measuring availability.
From what I've seen, this site sometimes shows different versions as separate items. For example, the 72 notebooks with Ice Lake-Y (which you forgot about) are all variations of Macbook Air.

That's why I'm suggesting a more qualitative approach, i.e. looking at what kind of notebooks get Ice Lake. Few months ago, at the moment @Vayra86 stopped updating, these were just a few low-volume products (expensive 2in1s).
Today 10nm is in many mainstream / bestselling lines.
Also, just the fact that Apple puts it into Macbook Air and Dell into XPS is a sign there are no quality / supply issues. They wouldn't take that risk.
Ice Lake-U also caps at 4 core, 8 threads, and quite low frequency.
Which is an optimal configuration for 15W SoCs - the most popular type of consumer CPU we have today.
Smaller 10nm SoCs will arrive this year. Larger - this or next year.
Big desktop CPUs will come last... or never.
Posted on Reply
#46
Valantar
notb
Of course. I've never said all are.

That's a very unorthodox way of measuring availability.
From what I've seen, this site sometimes shows different versions as separate items. For example, the 72 notebooks with Ice Lake-Y (which you forgot about) are all variations of Macbook Air.

That's why I'm suggesting a more qualitative approach, i.e. looking at what kind of notebooks get Ice Lake. Few months ago, at the moment @Vayra86 stopped updating, these were just a few low-volume products (expensive 2in1s).
Today 10nm is in many mainstream / bestselling lines.
Also, just the fact that Apple puts it into Macbook Air and Dell into XPS is a sign there are no quality / supply issues. They wouldn't take that risk.

Which is an optimal configuration for 15W SoCs - the most popular type of consumer CPU we have today.
Smaller 10nm SoCs will arrive this year. Larger - this or next year.
Big desktop CPUs will come last... or never.
While I agree that counting available SKUs is problematic (as you say, there are dozens of Apple models as they have 3-5 options each for CPU, SSD and RAM, with GPUs added to that for higher end ones, making for a whackton of SKUs for each laptop line), but saying "they're in premium lineups so there's wide availability" is a fallacy. They're in premium laptop lineups because they're new and fancy and can sell on novelty (and not "Skylake 7.0!" or whatever). Premium devices are flagships and generally not mass-market devices (even if some lineups do sell a lot) - Dell XPS and Macbook Air are not "mainstream" by any stretch of the imagination, even if they aren't the most expensive devices out there. All this tells us is that there's enough availability for Intel to guarantee supply for these premium lineups, but nothing more. The true proof of wide 10nm availability would be that most/all laptops moved to this, or if cheaper, more mass-market SKUs adopted it. They have generally not, and will likely never do so.

And if we're talking putting chips where the profits are, Intel would have been pushing 10nm to the server world as soon as they could. Margins there are much higher than in the mobile space, and volumes dramatically higher too. And desktop chips would then follow as soon as supply and demand evens out - desktop chips are essentially derivatives of low-end server chips, after all.
Posted on Reply
#47
Vayra86
notb
What?
Seriously, you're either not tracking the PC market at all or just ignoring mobile segment on purpose.

Ice Lake is available in normal, common notebooks. It's normally available in stores. And it's been there for a few months already.
You're repreating arguments from half a year ago or older. I don't understand why you're unwilling to just CHECK if they're still valid.
You really like Intel to fail and you're going to hold on to this narrative?

Why would it not be? Why would Intel offer Ice Lake for mainstream products if it wasn't making money?
And at the end of the day Intel's profit margin is still way higher than AMD (even higher than TSMC's). And it already includes those years of 10nm development with no sales.
You want to deny that as well?
Well, since you seem to know, give us some numbers? What did Intel sell?

As for the second question, Intel offers Ice Lake for many more reasons than just 'profit'. Profit is also branding and mindshare. Not releasing 10nm is a failure and damages that. Another argument that isn't profit related is fab capacity. 14nm is still under pressure, so any available 10nm capacity, even if the yields aren't optimal, is still welcome to relieve pressure.

With 7nm this close, Intel is in a 'screwed if you do, screwed if you don't' situation, I think. Its not like I'm sitting here rubbing my hands and smiling for it. Its just my analysis of how this node develops for them. How it relates to AMD's market share is not even in my mind tbh.
Posted on Reply
#50
bug
So, nowhere near double the performance, despite twice as many cores and a frequency advantage. If you were trying to tell us Ice Lake cores fine, you've succeeded.
Posted on Reply
Add your own comment