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

[AleksandarK]

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.


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.

View at TechPowerUp Main Site

 

[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.

 

[Flanker]

Cool, when can we expect reviews and benchmarks?

 

[Object55]

Cool, when can we expect reviews and benchmarks?

And mitigations ?

 

[Space Lynx]

Does anyone know what HBR3 means?

 

[ncrs]

And mitigations ?

They removed SGX outright from the platform so it's a start...

 

[R0H1T]

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.

You don't necessarily have to have AVX512 in desktop or notebook chips & no we've seen this happen before as well, IIRC Maxwell was backported to 28nm. Intel needs this & it's certainly cheaper than trying to fix the train wreck that is 10nm right now.

 

[Vya Domus]

Many things announced, leaked, etc, few actually released.

 

[Flanker]

They removed SGX outright from the platform so it's a start...

I don't think I understand it correctly, but it sounded like "our security guard did a lousy job, so we removed him." Shouldn't a flawed security feature be replaced by something better instead of just removing it?

 

[bug]

I don't think I understand it correctly, but it sounded like "our security guard did a lousy job, so we removed him." Shouldn't a flawed security feature be replaced by something better instead of just removing it?

When your "security guard" doubles as a backdoor (or several), getting rid of it makes sense.

 

[notb]

You don't necessarily have to have AVX512 in desktop or notebook chips & no we've seen this happen before as well, IIRC Maxwell was backported to 28nm. Intel needs this & it's certainly cheaper than trying to fix the train wreck that is 10nm right now.

Man... you're writing posts from mid 2019 at best. 10nm is profitable already and used for mainstream chips.

 

[bug]

Man... you're writing posts from mid 2019 at best. 10nm is profitable already and used for mainstream chips.

As evidenced by the dozen of CPU models you can buy today, none of which is available for the desktop.
Intel may have figured out the process at long last, but they definitely still have an acute capacity problem on their hands.

 

[jabbadap]

Does anyone know what HBR3 means?

It's a displayport thingy, nothing new really.

 

[Ferrum Master]

Intel will have NOTHING worthy till 2023 ie 5 years since Jim Keller joined and it will the time for the first arch his team has created for Intel.

All this we will have are tweaks, giving really diminishing returns, yet costing a lot of transistor budget. Simple as that. The arch has gained such point.


I suggest to watch the guy speak...

 

[bug]

Intel will have NOTHING worthy till 2023 ie 5 years since Jim Keller joined and it will the time for the first arch his team has created for Intel.

All this we will have are tweaks, giving really diminishing returns, yet costing a lot of transistor budget. Simple as that. The arch has gained such point.


I suggest to watch the guy speak...

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...

 

[R0H1T]

Man... you're writing posts from mid 2019 at best. 10nm is profitable already and used for mainstream chips.

Profitable as in sinking tens of billions to make it work & then getting a trickle of the planned capacity 5 years after it was supposed to come online?
You sure that's what you'd call profitable or "mainstream"

Also if Intel's saying 7nm hits in 2021(?) I'll put mainstream availability around 2022 at the very least, their track record since 22nm (late by 3 months or so) has been delays after delays & I'd rather see the products on shelves than believe whatever PR schtick they come up with next!

 

[dicktracy]

My next upgrade if it delivers.

 

[Imsochobo]

ohh a mainstream platform from intel which isn't totally horrible and crap and was good back in 2011 before invention of m2 and high bandwidth I/O.
Good!

 

[Ferrum Master]

Except Intel's problem is not architecture

It is a problem, its still an evolution of Sandy Bridge in the core, not a new design. It plagues Intel especially with the vulnerability issues.

If you watch the interview you would get the idea... 5 years, new arch from the scratch. Fastest is 2023...

 

[Dave65]

You don't necessarily have to have AVX512 in desktop or notebook chips & no we've seen this happen before as well, IIRC Maxwell was backported to 28nm. Intel needs this & it's certainly cheaper than trying to fix the train wreck that is 10nm right now.

No expert but seems to me Intel is wasting resources with 10 nm, seems the better option would be to work on 7nm or even 5?

 

[notb]

As evidenced by the dozen of CPU models you can buy today, none of which is available for the desktop.

They just don't make 10nm CPUs for desktops. It's a low priority segment.

I know desktops are important to you, but Intel is not a charity. They can make a limited number of 10nm CPUs and they put them where it makes a difference.

Intel may have figured out the process at long last, but they definitely still have an acute capacity problem on their hands.

Likely more than AMD (or that'll happen in next few months). I'll leave it to you to decide if that's a lot or not.

Profitable as in sinking tens of billions to make it work & then getting a trickle of the planned capacity 5 years after it was supposed to come online?
You sure that's what you'd call profitable or "mainstream"

So you've been writing from 2019 and now it's what: 2018? 2017?

I'm talking about today. 10nm works, delivers and appears in mainstream products: Lenovo Yoga, Dell XPS/Inspiron, Macbook Air and so on.

If you really want to criticize this tech because it didn't meet Intel's roadmap, do what makes you happy.
I guess it actually becomes a praise if this is the first (only) drawback that comes to your mind...

Also if Intel's saying 7nm hits in 2021(?) I'll put mainstream availability around 2022 at the very least, their track record since 22nm (late by 3 months or so) has been delays after delays & I'd rather see the products on shelves than believe whatever PR schtick they come up with next!

I have no idea when these dates come from. How is 7nm not mainstream already?

By 2022 Intel will likely move all their segments to 7nm or better. Is that what you mean?

Intel will have NOTHING worthy till 2023 ie 5 years since Jim Keller joined and it will the time for the first arch his team has created for Intel.

What's wrong with Intel's architecture? They're new big cores are excellent. I guess you haven't been paying attention to Ice Lake benchmarks.
And they're launching a lot of new stuff all the time: all the AI stuff, 3D stacking, "big.LITTLE", networking.

You can't judge just based on gaming desktops - even if that's the only segment you're interested in.
If you want the best CPU for DIY desktop, just buy AMD.
Intel is losing in this niche right now and probably also loses interest in competing. Frankly, there's a chance this will never change. Seriously, never.

 

[Ferrum Master]

What's wrong with Intel's architecture? They're new big cores are excellent. I guess you haven't been paying attention to Ice Lake benchmarks.
And they're launching a lot of new stuff all the time: all the AI stuff, 3D stacking, "big.LITTLE", networking.

You can't judge just based on gaming desktops - even if that's the only segment you're interested in.
If you want the best CPU for DIY desktop, just buy AMD.
Intel is losing in this niche right now and probably also loses interest in competing. Frankly, there's a chance this will never change. Seriously, never.

You are on the brink getting into fanboyish attitude. Calm down. This old arch is a can of worms, it is just a matter of time for another Zero Day class exploit to emerge. The funny move omitting SGX, not fixing it instead to save from the dreaded LVI vulnerability is reason to think about it in general.

There is a lack of CPU's in 19Q4 and Price/Performance point of Rome ended up in 15% market shift in Server Market.... your so called DIY choice AMD is constructively gunning Intel in each crucial market point. And thanks to that we enjoy lower prices and better offerings for the same money.

Your mentioned 10nm parts are AWOL. PC OEM's did enjoy the squeeze first especially Dell. Due to lack of CPU's manufacturers are switching to AMD just because of that. Server market came afterwards...

You call that a niche? You are blaming dude about living 2018, yet missed the train about last quarter yourself.

 

[ARF]

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.

It's not interesting because the latest standards are HDMI 2.1 and Display Port 2.0..

Except Intel's problem is not architecture (Sunny Cove looks darn good on paper), it's fab capacity

Intel's problem is that its N10 node can't do 5 GHz which its cores desperately need in order to stay competitive with the higher IPC Zen 2 and Zen 3.

 

[yeeeeman]

Many things announced, leaked, etc, few actually released.

Because they are leaked with as many as 2 years before. Until it actually launches people get bored of waiting and reading about them.

 

[londiste]

The funny move omitting SGX, not fixing it instead to save from the dreaded LVI vulnerability is reason to think about it in general.

Depends. How far along is the release? If they are planning this within 2020, they might not have enough time to have a fix for LVI in hardware and Intel cannot afford to skip a generation at this point.

Intel's problem is that its N10 node can't do 5 GHz which its cores desperately need in order to stay competitive with the higher IPC Zen 2 and Zen 3.

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.