After big delays, Sapphire Rapids arrives, full of accelerators and superlatives
- Reference: 1673373609
- News link: https://www.theregister.co.uk/2023/01/10/after_big_delays_intels_new/
- Source link:
The Silicon Valley-based chipmaker announced the launch of the new Xeons, code-named "Sapphire Rapids," on Tuesday. It marks the second consecutive datacenter CPU released by Intel that was delayed multiple times, showing the company is still playing catch-up while barreling forward with a [1]comeback plan architected by its CEO of nearly two years, Intel veteran Pat Gelsinger.
Remember that Intel's 3 rd -Gen Xeon, also known as Ice Lake, launched [2]nearly two years ago after the chip's 10nm node suffered [3]multiple delays over several years due to manufacturing issues. Those delays forced Intel to release server CPUs based on its older 14nm process prior to Ice Lake.
[4]
After sorting out issues with the 10nm node, Intel moved onto a more advanced version of the process and renamed it to Intel 7 last year because process naming has increasingly fallen out of step with the actual size of transistors on the chips and Intel [5]felt its souped-up 10nm was more comparable to the 7nm capabilities of its Asian chip-making rivals, TSMC and Samsung.
[6]
[7]
The Intel 7 node entered volume manufacturing in 2021 with [8]Intel's 12 th -Gen Core processors for PCs, and now it's being used to fabricate the latest Xeons. However, the new Sapphire Rapids chips are only arriving after the company delayed production multiple times over a [9]year or so because it said the CPUs needed additional "platform and product validation time."
About those superlatives…
Intel is hoping its new Xeons can stand out with a handful of superlatives it has attached to the CPUs. According to the company, Sapphire Rapids:
has the "the most built-in accelerators"
is "the most sustainable Xeon ever"
is "the most tested and battle-hardened CPU that we've ever delivered in the industry"
has "the most effective ubiquitous way to do [AI] training via fine tuning and transfer learning"
The chipmaker also touted that Xeon is the "most scalable and flexible architecture on the planet" and that the improved Intel Software Guard Extensions in Sapphire Rapids is the "most researched, updated, and deployed confidential computing technology in data centers on the market today."
What Intel didn't mention is that x86 rival AMD beat the company to the punch in bringing DDR5, PCIe 5.0, and Compute Express Link (CXL) technologies to the datacenter with its new Epyc "Genoa" chips last fall. For a while, Intel was [10]hoping to get there first with Sapphire Rapids, but then delays got in the way.
Besides supporting DDR5, PCIe 5.0, and CXL 1.1, the new Xeon generation comes with up to 60 cores — fewer than the 96 maximum offered by AMD's Genoa — and it consists of 52 different models that cover six different server segments as part of Intel's "workload-driven" approach.
[11]
The first segment is one that will concern the broadest constituency of datacenter users and operators: general-purpose computing. Within this segment, Intel has a wide array of dual-socket options that vary in core counts, clock speeds, and other features, in addition to a few single-socket models, one variant designed for long-term use for IoT use cases, and, for the first time, two liquid-cooled SKUs.
The rest of the new Xeons are optimized for the following areas:
virtualization, analytics, and in-memory databases
5G and networking
software-as-a-service, infrastructure-as-a-service, and media
storage and hyperconverged infrastructure
high-performance computing
Most CPUs will be available as part of [12]Intel's new software-defined silicon service , known as "Intel On Demand," which lets organizations pay money to enable certain Xeon features as an alternative to buying the chip with all capabilities unlocked.
Intel has already talked a [13]great deal about its HPC-optimized Xeon chips, which have large stacks of high-bandwidth memory and are known as the Xeon CPU Max Series.
At a high level, Intel claims that, compared to the previous generation, 4 th -Gen Xeon provides a 53 percent improvement in general-purpose computing; up to 10x faster AI inference and training performance; as much as a 2x boost for networking, storage, and telecom; up to 3x faster work for data analytics; and a 3.7x boost for HPC.
[14]
The broad applicability of Sapphire Rapids is why Intel is claiming it as the "most scalable and flexible architecture on the planet," according to Intel exec Lisa Spelman.
"It can serve every single workload again, from the edge to the cloud and back and in between. And there's nothing that it can't handle competently," said Spelman, who is corporate vice president and general manager of Intel Xeon products.
Bells and whistles galore… for the environment?
While AMD may have it beat on maximum core counts, Intel is claiming the 13 accelerator engines built into the new Xeons are a worthy alternative, offering a "more efficient way to achieve higher performance than growing the CPU core count."
These accelerators include:
Intel Advanced Matrix Extensions, which is meant to improve the performance of training small and medium deep learning models like natural language processing and recommendation systems.
Intel Quick Assist Technology, which can accelerate data encryption and compression workloads by offloading through an offload mechanism.
Intel Data Streaming Accelerator, which aims to eliminate bottlenecks in data movements between the CPU cores, memory, caches, attached storage, and networked storage devices.
Intel Dynamic Load Balancer, a hardware-managed load balancing system in the processor that is designed for telecom applications.
Intel In-Memory Analytics Accelerator, which is meant to speed up compression and decompression for big data applications and in-memory analytic databases.
Intel Advanced Vector Extensions 512, which provides optimizations meant to improve the performance of demanding workloads like scientific simulations, financial analytics, and 3D modeling.
Intel Crypto Acceleration, which is designed to improve the performance of "encryption-sensitive workloads" such as firewalls, VPNs, and SSL web servers.
Intel Software Guard Extensions, a security feature that is meant to protect data in encryption portions of the CPU's memory.
The other accelerators consist of Intel Speed Select Technology, Intel Data Direct I/O Technology, and two other security features: Intel Trust Domain Extension and Intel Control-Flow Enforcement Technology.
Intel is claiming that Sapphire Rapids' use of the accelerators, combined with software optimizations, "help improve power efficiency across AI, data analytics, networking, and storage." This, according to the chipmaker, can improve performance per watt efficiency in such workloads by 2.9 times on average compared to the previous generation of Xeon CPUs.
[15]AMD's Epyc 4 will likely beat Intel Sapphire Rapids to market
[16]AMD follows Intel's lead with alphanumeric soup of new Ryzens
[17]Intel: Please buy these new 13th-Gen CPUs, now with 24 cores
[18]It's time to retire 'edge' from our IT vocabulary
This is part of why Intel is painting its new server chips as the most sustainable Xeons yet. What also makes Sapphire Rapids purportedly more environmentally friendly than other server chips is that it has a new optimized power mode, reachable in the BIOS, that enables up to 20 percent in power savings per CPU socket while only reducing performance for certain workloads by 5 percent.
The semiconductor giant claims the new Xeons are also manufactured with 90-100 percent renewable energy and "state-of-the-art water reclamation facilities." Intel has been happy to tout recently that it has achieved "net positive water" at manufacturing sites in the US, India, and Costa Rica, but we have to remember that doesn't mean it's [19]going green everywhere . ®
Get our [20]Tech Resources
[1] https://www.theregister.com/2022/02/18/intel_gelsinger_plan/
[2] https://www.theregister.com/2021/04/06/ice_lake_xeon_intel/
[3] https://www.theregister.com/2018/12/12/intel_architecture_future/
[4] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/systems&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2Y73uEOtm4@qu2G8CFWWYcgAAAMM&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[5] https://www.eenewseurope.com/en/intel-is-on-track-to-regain-chip-manufacturing-lead-says-executive/
[6] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/systems&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44Y73uEOtm4@qu2G8CFWWYcgAAAMM&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[7] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/systems&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33Y73uEOtm4@qu2G8CFWWYcgAAAMM&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[8] https://www.theregister.com/2021/10/27/intel_alder_lake/
[9] https://www.theregister.com/2022/11/01/intel_sapphire_rapids_q1_2023/
[10] https://www.theregister.com/2022/11/01/amd_epyc_intel_sapphire/
[11] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/systems&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44Y73uEOtm4@qu2G8CFWWYcgAAAMM&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[12] https://www.theregister.com/2022/11/22/intel_reveals_paid_xeon_features/
[13] https://www.theregister.com/2022/11/09/intel_max/
[14] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/systems&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33Y73uEOtm4@qu2G8CFWWYcgAAAMM&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[15] https://www.theregister.com/2022/11/01/amd_epyc_intel_sapphire/
[16] https://www.theregister.com/2023/01/05/amd_ryzen_intel/
[17] https://www.theregister.com/2023/01/03/intel_13th_generation_cpus/
[18] https://www.theregister.com/2022/12/22/retire-edge-now/
[19] https://www.theregister.com/2022/07/13/intels_net_positive_water_use/
[20] https://whitepapers.theregister.com/
Finland
Once upon a time, in a small city nestled deep in the heart of Finland, the winters were long and harsh. The residents of the city, accustomed to the frigid temperatures, had always relied on traditional methods of heating their homes, such as wood-burning stoves and electric heaters. But one year, a particularly cold winter hit the city and the demand for heat skyrocketed.
The mayor of the city, a practical and resourceful woman, knew that they needed to find a new and more efficient way to heat the homes of her citizens. She began to research different options, and it wasn't long before she came across an intriguing idea: using Intel processors to generate heat.
The mayor reached out to a local engineer, who was known for his expertise in computer hardware and energy efficiency. Together, they developed a system in which a network of Intel processors were used to convert electricity into heat, which would then be distributed to the homes of the city's residents.
The system was tested in a small pilot program, and the results were astounding. Not only was the heat generated by the processors much more efficient than traditional heating methods, but it was also significantly cheaper.
Encouraged by the success of the pilot program, the mayor worked with the city council to roll out the Intel processor-based heating system to all of the homes in the city. It was a massive undertaking, but within a few months, every home in the city was connected to the network.
The citizens of the city were thrilled with the new system. Not only were their homes warmer than ever before, but their heating bills were also significantly lower. And with the added bonus of helping to reduce the city's carbon footprint, everyone felt good about the switch to Intel processors.
The city became a showcase for other small cities and towns to follow, and soon, other parts of Finland and even other countries started to adopt the same method to improve the heating of their homes. And it all started with a mayor's determination to help her city survive and thrive, even in the midst of the harshest winters.
Re: Finland
Don't laugh: what if someone actually *did* move their massive server / crypto farms to Finland and, instead of dumping the heat to the general environment, designed a neighborhood circulation system to reuse the heat energy? Imagine selling your MIPS processing power to subsidize local home heating bills.
Could actually be a reasonable future...
basic comparisons
Obviously the raw specs don't indicate what the actual performance is, but was just curious looking at the HPE DL380 Gen11 vs the DL385 Gen 11. A few things that stand out to me -
Intel says that Cloud/IaaS workloads benefit from fewer cores/higher frequencies(Intel P series Xeon), which seems strange to me, of course it depends on the app. But as a ESXi user for many years I'll take more cores any day(especially if I don't have to worry about windows/Oracle licensing ..). They say databases benefit from more cores(H series Xeon).
Intel's top end 8490H 60-core 350W processor runs at 1.9Ghz and has 112MB of L3 cache.
AMD's top end 9654 96-core 360W processor runs at 2.4Ghz and has 384MB of cache(HPE isn't specific on AMD if that is L3 or aggregate on all caches for DL385)
On the flip side, the DL380Gen11 has 16 memory slots per socket, vs the DL385Gen11 has only 12 (the DL385 Gen10+ V2 has 16/socket), can only assume the size of the new AMD chips are much larger than the Intel chips.
AMD's previous generation chips (in DL385Gen10+ V2) topped out at 280W, vs the new chips go to 360W (I have an earlier version of the Gen11 data sheet that actually says they draw 400W, had to double check vs the current data sheet).
Intel's previous generation chips (in DL380 Gen10+) topped out at 270W, vs the new chips at 350W.
I *think* I'd rather take the previous generation 64-core/socket DL385Gen10+ V2 with the extra memory slots and less power usage/server(and probably better stability with the firmware/etc being around for longer). Would be a tough choice. Key point for me is my workloads don't even tax the existing DL380Gen9 servers CPU wise(but I want lots of cores for lots of VMs/host)
Used to be Intel 14nm+++++++++++
Now, it's Intel 10nm+++++++++++, but we call it Intel 7 for short. Wink, wink, wink.
Heating
Do you need all this processing power if you just want to heat your home?
Intel should come up with a CPU that doesn't have accelerators and other bells and whistles and could just run a temperature sensor to produce heat according to the target temperature set by the users.
They could reposition themselves as Tesla of heating and since these processors already use electricity they could market them as extremely green way of heating.