Tuesday, April 2nd 2019

Intel Unleashes 56-core Xeon "Cascade Lake" Processor to Preempt 64-core EPYC

Intel late Tuesday made a boat-load of enterprise-relevant product announcements, including the all important update to its Xeon Scalable enterprise processor product-stack, with the addition of the new 56-core Xeon Scalable "Cascade Lake" processor. This chip is believed to be Intel's first response to the upcoming AMD 7 nm EPYC "Rome" processor with 64 cores and a monolithic memory interface. The 56-core "Cascade Lake" is a multi-chip module (MCM) of two 28-core dies, each with a 6-channel DDR4 memory interface, totaling 12-channel for the package. Each of the two 28-core dies are built on the existing 14 nm++ silicon fabrication process, and the IPC of each of the 56 cores are largely unchanged since "Skylake." Intel however, has added several HPC and AI-relevant instruction-sets.

To begin with, Intel introduced DL Boost, which could be a fixed-function hardware matrix multiplier that accelerates building and training of AI deep-learning neural networks. Next up, are hardware mitigation against several speculative execution CPU security vulnerabilities that haunted the computing world since early-2018, including certain variants of "Spectre" and "Meltdown." A hardware fix presents lesser performance impact compared to a software fix in the form of a firmware patch. Intel has added support for Optane Persistent Memory, which is the company's grand vision for what succeeds volatile primary memory such as DRAM. Currently slower than DRAM but faster than SSDs, Optane Persistent Memory is non-volatile, and its contents can be made to survive power-outages. This allows sysadmins to power-down entire servers to scale down with workloads, without worrying about long wait times to restore uptime when waking up those servers. Among the CPU instruction-sets added include AVX-512 and AES-NI.
Intel Speed Select is a fresh-spin on a neglected feature most processors have had for decades, allowing administrators to select specific multipliers for CPU cores on the fly, remotely. Not too different from this is Resource Director Technology, which gives you more fine-grained QoS (quality of service) options for specific cores, PIDs, virtual machines, and so on.

Unlike previous models of Xeon Scalable, the first Xeon Scalable "Cascade Lake" processor, the Xeon Platinum 9200, is an FC-BGA package and not socketed. This 5,903-pin BGA package uses a common integrated heatspreader with the two 28-core dies underneath. The two dies talk to each other over a UPI x20 interconnect link on-package, while each die puts out its second UPI x20 link as the package's two x20 links, to scale up to two packages on a single board (112 cores). Source: HotHardware
Add your own comment

88 Comments on Intel Unleashes 56-core Xeon "Cascade Lake" Processor to Preempt 64-core EPYC

#1
Solaris17
Dainty Moderator
Per core licensing trolls me softly.
Posted on Reply
#3
xkm1948
Pretty interesting TBH as long as pricing ain't stupid, for HPC cluster of course.
Posted on Reply
#4
hat
Enthusiast
I wonder if the fact that it's BGA means anything in this market. As a desktop enthusiast, the idea of a BGA chip is pretty horrific, but I don't suppose admins in charge of server farms are upgrading CPUs alone very often.
Posted on Reply
#5
Crackong
Can we expect a "56 cores 5GHz , gamers will be happy" demo in Computex?
Posted on Reply
#6
steve360
2 x 28 cores huh?

Did they use glue to put that together?
Posted on Reply
#7
hat
Enthusiast
steve360, post: 4024026, member: 132641"
2 x 28 cores huh?

Did they use glue to put that together?
No no no, we didn't use that glue. We at Intel assure you that blue glue is faster, more energy efficient, and stickier than red glue.
Posted on Reply
#8
HTC
This is the monster Intel had planned to counter Rome 64c / 128t behemoth. Is it true it doesn't have HT? I saw no mention either way in OP.

For dual socket systems, it will be Intel's 112c VS AMD's 128c / 256t: Intel's system will have to be much faster in order to counter AMD's more cores and threads but that will skyrocket TDP, so ...
Posted on Reply
#9
TheGuruStud
xkm1948, post: 4024017, member: 50521"
Pretty interesting TBH as long as pricing ain't stupid, for HPC cluster of course.
Power consumption is off the charts and clocks in the dumpster. It's a terrible choice. I can only imagine low they'd go under AVX (1.5 ghz? Lol).
Posted on Reply
#11
Noztra
Their Intel Xeon Platinum 8280 (28C/56T) is $10009.

So this one since its pretty much two 8180 clued together. Twice the cores = Twice the price?
Posted on Reply
#12
notb
hat, post: 4024019, member: 32804"
I wonder if the fact that it's BGA means anything in this market. As a desktop enthusiast, the idea of a BGA chip is pretty horrific, but I don't suppose admins in charge of server farms are upgrading CPUs alone very often.
It doesn't. You don't even open the case - that's why parts that need to be replaced often (disks, power supply) can be accessed through doors.

Also, I was looking forward to BGA in desktops, when first rumors came out. Being a PC enthusiast doesn't imply being an enthusiast of replacing CPUs.
Simpler, less work, less problems, cheaper, more power.

Noztra, post: 4024047, member: 173244"
Their Intel Xeon Platinum 8280 (28C/56T) is $10009.

So this one since its pretty much two 8180 clued together. Twice the cores = Twice the price?
Even if - it's still a bargain for datacenters.
Posted on Reply
#14
hat
Enthusiast
notb, post: 4024049, member: 165619"
It doesn't. You don't even open the case - that's why parts that need to be replaced often (disks, power supply) can be accessed through doors.
Fair enough.

notb, post: 4024049, member: 165619"
Also, I was looking forward to BGA in desktops, when first rumors came out. Being a PC enthusiast doesn't imply being an enthusiast of replacing CPUs.
Simpler, less work, less problems, cheaper, more power.
Now this I just completely disagree with. None of those points should matter at all to anyone who is a PC enthusiast, except maybe "more power", but we've seen high end overclocked CPUs pull down many hundreds of watts on their own, so I'm not sure that point matters much, either. Unless you and I just have completely different definitions of what a PC enthusiast is.


notb, post: 4024049, member: 165619"
Even if - it's still a bargain for datacenters.
I'd hazard a guess that EPYC would be an even better bargain.
Posted on Reply
#15
londiste
HTC, post: 4024040, member: 51238"
This is the monster Intel had planned to counter Rome 64c / 128t behemoth. Is it true it doesn't have HT? I saw no mention either way in OP.
There is no confirmation either way but logically it should have HT. The rumor it does not have HT started from Intel's Linpack testing notes where HT is disabled. Linpack testing does not benefit from HT, can actually lose performance from it and it is commonly disabled when running Linpack. Both Intel as well as AMD recommend doing so.
Noztra, post: 4024047, member: 173244"
Their Intel Xeon Platinum 8280 (28C/56T) is $10009.
So this one since its pretty much two 8180 clued together. Twice the cores = Twice the price?
There is also a 28-core Xeon Platinum 8276 for $8719. These are also intended for servers with up to 8 CPUs.
While there is no 28-core CPU in smaller categories, there is a 24-core Xeon Gold 6262 for $2900 (max 4 CPU). I would bet lack of 28-core option has something to do with harvesting cores for the 9200 series.
Posted on Reply
#16
Zubasa
steve360, post: 4024026, member: 132641"
2 x 28 cores huh?

Did they use glue to put that together?
Ironically, the W-3175X does technically make the 28-core Xeon die also a Desktop die. :roll:

londiste, post: 4024068, member: 169790"
While there is no 28-core CPU in smaller categories, there is a 24-core Xeon Gold 6262 for $2900 (max 4 CPU). I would bet lack of 28-core option has something to do with harvesting cores for the 9200 series.
Also some of those better binned chips needs to go towards the PR stunt that is the Xeon W-3175X.
Posted on Reply
#17
Noztra
notb, post: 4024049, member: 165619"
It doesn't. You don't even open the case - that's why parts that need to be replaced often (disks, power supply) can be accessed through doors.

Also, I was looking forward to BGA in desktops, when first rumors came out. Being a PC enthusiast doesn't imply being an enthusiast of replacing CPUs.
Simpler, less work, less problems, cheaper, more power.


Even if - it's still a bargain for datacenters.
Not really a bargin if the CPU cost 20-30K and uses 400 Watt. TCO would be fairly high compared to EPYC/ROME.
Posted on Reply
#18
londiste
Noztra, post: 4024082, member: 173244"
Not really a bargin if the CPU cost 20-30K and uses 400 Watt. TCO would be fairly high compared to EPYC/ROME.
This is intended to compete with 64-core EPYC that we still do not know enough about.
1.4/2.2GHz and 250W seems to be the expectation based on rumors so far and that should be realistic considering what we have seen of Zen2.
If that 56-core thingie can do 2.6/3.8GHz at 400W it might actually be rather competitive.
Posted on Reply
#19
notb
hat, post: 4024066, member: 32804"
Now this I just completely disagree with. None of those points should matter at all to anyone who is a PC enthusiast, except maybe "more power", but we've seen high end overclocked CPUs pull down many hundreds of watts on their own, so I'm not sure that point matters much, either. Unless you and I just have completely different definitions of what a PC enthusiast is.
That's because you're using an extremely narrow definition of "a PC enthusiast".
I'd hazard a guess that EPYC would be an even better bargain.
Very unlikely.
When buying a server, you're paying for the machine and for a particular service that comes with it.
The fact that a CPU is cheaper doesn't mean e.g. Dell will sell you the whole package for less.
Just the fact that Intel has 20x larger market share means companies have larger stock of CPUs and other parts. The same SLA should cost less when going with Blue.

But even if there actually was a price difference, it's not exactly huge.
Let's assume every other part costs exactly the same and EPYC equivalent is $10k less per CPU in a 2P machine (because we can!).
Over 3 years you save $556 per month per server. Not much.
There $556 buy you a homogeneous architecture and simpler procedures/training inside the company.
Posted on Reply
#20
Noztra
londiste, post: 4024091, member: 169790"
This is intended to compete with 64-core EPYC that we still do not know enough about.
1.4/2.2GHz and 250W seems to be the expectation based on rumors so far and that should be realistic considering what we have seen of Zen2.
If that 56-core thingie can do 2.6/3.8GHz at 400W it might actually be rather competitive.
Well according to the latest leaks the biggest ROME (64C/128T) should have a base clock at 2.35GHz and a all core boost at around 3.4GHz.

So Intel won't be rather competitive. AMD have stated multiply times that when they designed ROME they expected to compete FAVORABLY against 10nm.

Even Intel said they where afraid that AMD could grab around 20% of the market.

Again it's all just guesses, so everything might be wrong, but we will probably see at Computex. :)
Posted on Reply
#21
Metroid
Intel soon will be paying dearly for all the bullshit it has been done for the past 10 years and if you feel sorry for them then it means you are not a true capitalist.
Posted on Reply
#23
Tomgang
So now intel also makes glued together cpu's:kookoo: and still on 14 nm. I am afraid to think op tdp on this thing or how low core clock needs to be to hold with in a tdp that dosent need water cooling.

Intel really need to get there next nm die shrink out fast, else amd might gonna take the long straw this time.
Posted on Reply
#24
phill
Solaris17, post: 4024014, member: 14803"
Per core licensing trolls me softly.
MS will love that.... Server licences anything over 8 cores (not sure about the threads) I think you need to pay over and above for...
I wonder what sort of cooling they'll have to use if it's going in servers??....
Posted on Reply
#25
HwGeek
Maybe AMD will Follow Intel's High TDP and offer Same High TDP 64C/128T? it will boost for much higher clock and offer even more performance versus just 180W TDP parts.
Posted on Reply
Add your own comment