Monday, April 18th 2022

Intel Sapphire Rapids 56-Core ES Processor Boosts to 3.3 GHz at 420 Watts

by
AleksandarK
 Discuss (41 Comments)
Intel is slowly transitioning its data center customers to a new processor generation called Sapphire Rapids. Today, thanks to the hardware leaker Yuuki_ans we have more profound insights into the top-end 56-core Sapphire Rapids processor and its power settings. According to the leak, we have information on either Xeon Platinum 8476 or Platinum 8480 designs that are equipped with 56 cores and 112 threads. This model was running at the base frequency of 1.9 GHz and a boost frequency of 3.3 GHz. Single-core can boost to 3.7 GHz if the report is giving a correct reading. Remember that this is only an engineering sample, so the final target speeds could differ. It carries 112 MB of L2 and 105 MB of L3 cache, and this sample was running with 1 TB of DDR5 memory with CL40-39-38-76 timings.

Perhaps the most exciting finding is the power configuration of this SKU. Intel has enabled this CPU to consume 350 Watts in PL1 rating, with up to 420 Watts in PL2 performance mode. The enforced BIOS power limit rating is set at an astonishing 764 Watts, which could happen with AVX-512 enabled. Final TDP ratings are yet to be disclosed; however, these Sapphire Rapids processors are shaping to be relatively power-hungry chips.
Source: @Yuuki_ans (Twitter)

Related News

Add your own comment

41 Comments on Intel Sapphire Rapids 56-Core ES Processor Boosts to 3.3 GHz at 420 Watts

#26
Unregistered
Dr_b_"Alder Lake. All of a sudden...Boom! We are back in the game. AMD in the rearview mirror in clients, and never again will they be in the windshield; we are just leading the market" ....

IN power consumption, yes you are correct Pat, if you ever do release SR that is....
AMD was in the rear view mirror for muuuuuch longer than Intel was. Adl might use a lot of power, but it Shows what Intel can do in a rush, imagine if they take their time. Amd made them rush, they are knocking FABS out which AMD hasn't a hope of doing. When AMD are still on someones list for DIES INTEL will be knocking their own out.


Selling your shit is not a fair game, why should it be. winners sell their shit, losers cry its not fair.
Posted on Edit | Reply
#27
eidairaman1
The Exiled Airman
Chrispy_I have EPYC on-prem but they're in a niche role where I got to choose the hardware.

The Intel sales/marketing/OEM-bribery machine is still running in force. HP, Dell, and Lenovo will need to be coerced into selling you an EPYC. If you ask for a spec, they'll ALWAYS recommend Intel and to get AMD quotes you literally have to tell them that you're going elsewhere unless they stop trying to jam Intel down your throat. The system is still dirty, and it's clear that Intel still don't play fair. It's not solely that though, Consultants/contractors building a solution that requires an x86 server will just recommend a Xeon, because that's what's listed in an old document somewhere. That document is likely version 16 of a an original that predates even Opteron servers; It'll take decades before AMD servers are seen as even equal to Intels, despite the fact that AMD have been outperforming Intel in the server space for at least half a decade already.

Don't get me wrong, the servers I manage on-prem and in DCs are 95% Intel and they work great. I just dislike how strongly Intel's old bedfellows still stick to the old, illegal, broken system.
Thank you for calling a spade a spade
TiggerAMD was in the rear view mirror for muuuuuch longer than Intel was. Adl might use a lot of power, but it Shows what Intel can do in a rush, imagine if they take their time. Amd made them rush, they are knocking FABS out which AMD hasn't a hope of doing. When AMD are still on someones list for DIES INTEL will be knocking their own out.


Selling your shit is not a fair game, why should it be. winners sell their shit, losers cry its not fair.
Bribery is anti competitive.
Posted on Reply
#28
Crackong
We all know why Ice Lake Xeon availability is high, because the big service providers don't want them.

What an 'Advantage' !
Posted on Reply
#29
Fourstaff
Vya DomusI doubt that's the case, it's probably just product segmentation. From their point of view it doesn't make sense to think of it in terms of "a chiplet is broken in this CPU, we're just going to disabled it". The chiplets go through a validation process before they are packaged in the final product, so a 56 core CPU is almost definitely intentional and not the result of defects. It's not like if you have a defective chiplet you absolutely have to put it into a CPU and disable it, it's too wasteful, that's the point of having chiplets, it just means you have to replace it with a working one. The defective ones probably end up in Ryzen products.
They will indeed build 56 core CPUs with 7 chiplets if the market requests for them. Some of these defective ones will be reused in the 6 chiplet models and so on.
Posted on Reply
#30
ghazi
ncrsThey are not disabling an entire chiplet in the 56 core version - they are disabling one core in each of the 8 chiplets. Both L3 cache amount and TDP give clues to this, but this information has been disclosed by AMD anyway and is on Wikipedia. There are 7 (out of 23 available in retail, but cloud vendors get special SKUs) models in the Milan(-X) generation that have 4 active chiplets instead of 8, with the other 4 spots being populated by blank pieces for mechanical stability.
Even the 8 core EPYCs have 8 chiplets, because the L3 cache is too good to waste and memory bandwidth would be imbalanced with just 1 chiplet for the whole CPU. For the latter I mean the InfinityFabric bandwidth, not the IMC since those are on the IO die. One chiplet wouldn't be able to consume the whole bandwidth of the IMCs.
The Infinity Fabric actually requires the cores to be disabled symmetrically in each CCX in order to network correctly. That's why you didn't see these odd core counts before, when the CCX had 4 cores: 8 CCX meant core count must be a multiple of 8, 16 CCX meant it must be a multiple of 16. Also why the EPYC 7453 (poor bin with 28C and just 64MB L3) only has 4 chiplets, as 28 cores is not possible with 8 CCXs.
Posted on Reply
#31
watzupken
Solaris17Odd I’m literally surrounded by tens of thousands of servers and the amount of AMD usage is relatively low. Maybe they are used more in private data centers or on prem.
I actually don’t think AMD is capable of producing enough chips such that you can see a majority AMD system in a few years time. They are selling everything they can produce.

Anyway, I don’t know if this is the way ahead, but I think we have been seeing chip companies pushing for an incredible increase in power for the sake of pushing performance. We may have made progress in terms of performance, but regressing in power requirement. And yet people are claiming that we need to be “sustainable” and ”green”. I know power consumption has been creeping up over the years and decades, but I feel we are seeing big jumps in power draw this and next gen (based on rumours).
Chrispy_I have EPYC on-prem but they're in a niche role where I got to choose the hardware.

The Intel sales/marketing/OEM-bribery machine is still running in force. HP, Dell, and Lenovo will need to be coerced into selling you an EPYC. If you ask for a spec, they'll ALWAYS recommend Intel and to get AMD quotes you literally have to tell them that you're going elsewhere unless they stop trying to jam Intel down your throat. The system is still dirty, and it's clear that Intel still don't play fair. It's not solely that though, Consultants/contractors building a solution that requires an x86 server will just recommend a Xeon, because that's what's listed in an old document somewhere. That document is likely version 16 of a an original that predates even Opteron servers; It'll take decades before AMD servers are seen as even equal to Intels, despite the fact that AMD have been outperforming Intel in the server space for at least half a decade already.

Don't get me wrong, the servers I manage on-prem and in DCs are 95% Intel and they work great. I just dislike how strongly Intel's old bedfellows still stick to the old, illegal, broken system.
This sort of practice is unlikely to go away. You can almost say that it is ”in the blood”. That’s why big companies get bigger/ profit more.
Posted on Reply
#32
Mussels
Freshwater Moderator
Solaris17Odd I’m literally surrounded by tens of thousands of servers and the amount of AMD usage is relatively low. Maybe they are used more in private data centers or on prem.
Isn't that because AMD's entry (Well, with a true competitive product) in to that workspace is relatively new, and the whole scandals that keep popping up regarding intel and big business?

Intel stuck with $1.45 billion fine in Europe for unfair and damaging practices against AMD - ExtremeTech

"OEMs were given permission to sell higher percentages of AMD desktop chips, but were required to buy up to 95% of business processors from Intel. At least one manufacturer was forbidden to sell AMD notebook chips at all."
Posted on Reply
#33
The King
eidairaman1Pentium 4 preshot anyone?
Do you mean P4 Prescott?

I still have my P4 Northwood 3.2Ghz CPU running on a Gigabyte 875p mobo.
Posted on Reply
#34
Unregistered
The KingDo you mean P4 Prescott?

I still have my P4 Northwood 3.2Ghz CPU running on a Gigabyte 875p mobo.
He is inferring that ADL runs hot like Prescott did hence the name preshot. Ironic how it does not in fact run hot at all.
#35
R-T-B
user556"... these Sapphire Rapids processors are shaping to be relatively power-hungry chips."
Same as the 12 series.
You want high IPC? It's going to be a complex, power hungry core. You can't really have it both ways... short of something icky like BIG.little.
TiggerHe is inferring that ADL runs hot like Prescott did hence the name preshot. Ironic how it does not in fact run hot at all.
Yeah the P cores kinda do but they didn't swarm the die with them so the end result really isn't that bad at all. As much as I dislike big/little that's about as good as it gets for a design of that type. My 5950x is probably hotter than your chip.
Posted on Reply
#36
Unregistered
R-T-BYou want high IPC? It's going to be a complex, power hungry core. You can't really have it both ways... short of something icky like BIG.little.


Yeah the P cores kinda do but they didn't swarm the die with them so the end result really isn't that bad at all. As much as I dislike big/little that's about as good as it gets for a design of that type. My 5950x is probably hotter than your chip.
To be fair, my loop is pretty good, so not really worried even if it does apparently run hot.
#37
Chrispy_
eidairaman1Thank you for calling a spade a spade
Unfortunately the reality of server purchases is that the delivery, lead-times, maintenance contract, and financing are often more important purchasing criteria than speed/efficiency/performance.

As good as EPYC is, Intel still leads in volume supply guarantees, quantity of validations/certifications available, and other stuff that is overlooked or irrelevant for desktop CPUs.
ghaziThe Infinity Fabric actually requires the cores to be disabled symmetrically in each CCX in order to network correctly. That's why you didn't see these odd core counts before, when the CCX had 4 cores: 8 CCX meant core count must be a multiple of 8, 16 CCX meant it must be a multiple of 16. Also why the EPYC 7453 (poor bin with 28C and just 64MB L3) only has 4 chiplets, as 28 cores is not possible with 8 CCXs.
This is better wording of what I was trying to say. Thank you.
I feel that the EPYC 7753 must therefore be a fully-functional 64-core part that is limited in software.
Posted on Reply
#38
Max(IT)
Graphics cards with 600W TGP… CPUs with 400W TDP… hardware manufacturers lost contact with the reality. We are living in a worldwide energy crisis…
Posted on Reply
#39
sillyconjunkie
So, the 350watt TDP was known. Perhaps PL2 is 420w but everyone is stuck in speeds and feeds. Grain of salt considering how much power figures get thrown around here.

Market facts:
-Intel entered 2022 with higher dc market share.
-SR is 5nm, not "Intel 7".
-SR will hit the market before Genoa.
-Epyc is still using slower substrate connections between core clusters.

Performance per watt.. SR will out perform Genoa and AMD is now behind in the release cycle with DDR5 being the pivot.

I am an AMD fan..like all of this stuff. Just sick of AMD's reliance on slanted media.
Posted on Reply
#40
Panther_Seraphin
sillyconjunkieMarket facts:
-Intel entered 2022 with higher dc market share.
They have had a lead in DC market share for a VERY long time. However it has also been on a decrease for the past 4 years/16 quarters straight.
sillyconjunkie-SR is 5nm, not "Intel 7".
If so this is even more worrying for a multitude of reasons. Milan and Milan X are both on 7nm and power seems to be reasonable at a 280watt TDP and they seem to respect the limits set by the user well if they want to. SR has limits up to 700 watts per chip......currently.
sillyconjunkie-SR will hit the market before Genoa.
This is true and may be a thing for certain single threaded workloads. However they will NOT come cheap with the HBM memory also now being included as a L4 cache. So unless they have a >20% single threaded lead I can see the arguments for always going Intel failing more and more. When you are in a data center your not looking at the needs for a 1/2/4U server, Your looking at the overall power consumption of a 42u or mulitple 42u racks.

The $$$$ needed to increase the power supply/UPS/Cooling etc by 50% is no small amount especially across multiple 42u Racks whcih is the amount that would be needed to consider a like for like change between Milan and Sapphire Rapids.
Posted on Reply
Add your own comment
Dec 4th, 2024 04:30 EST change timezone

Latest GPU Drivers

New Forum Posts

Popular Reviews

Controversial News Posts