At the Optical Fiber Communications Conference (OFC) 2024, Intel gave us a particularly interesting glimpse at what could be the future of chip and interconnect design. The Optical Compute Interconnect (OCI) chiplet integrates a Photonics Integrated Circuit (PIC) with an electrical IC. The chiplet was packaged with an Intel CPU and shown running live data.

Intel believes this technology is a game changer as it meets increasing demand for higher bandwidth across greater distances, while using less power. Of course, the initial applications will be used for AI infrastructure and HPC environments, where scalability concerns across racks and clusters are becoming increasingly apparent in large scale deployments.

Intel’s press release included the following statement: “As an analogy, replacing electrical I/O with optical I/O in CPUs and GPUs to transfer data is like going from using horse-drawn carriages to distribute goods, limited in capacity and range, to using cars and trucks that can deliver much larger quantities of goods over much longer distances. This level of improved performance and energy cost is what optical I/O solutions like Intel’s OCI chiplet emerging bring to AI scaling.” That tells us where Intel believes the future of interconnect technology lies.

Although the OCI is a prototype, it already supports up to 4 terabits per second (Tbps) bidirectional data transfer across distances as far as 100 meters. Importantly, the technology is compatible with PCIe 5.0, making it relatively easy to integrate into existing PCIe compatible infrastructure.

In terms of power savings, Intel claims its solution is much more efficient, consuming only 5 pico-joules (pJ) per bit compared to pluggable optical transceiver modules at about 15 pJ/bit.

The OCI chiplet can be integrated with future CPUs, GPUs and SoCs. Should the technology reach the consumer space in the years ahead—and if it takes off in the enterprise space there’s little reason to doubt it will—then PCs in the years ahead could look very different.

Imagine being able to install a GPU that doesn’t require a PCIe slot. You could install it anywhere in a compatible case, and connect it to a motherboard via a cable without any loss of performance. That should lead to improved cooling performance and potentially simpler (and cheaper) motherboard designs.

Intel says it has developed a new silicon photonics fabrication process and it expects the technology to scale from the 4 Tbps range to tens of Tbps in the future. Chips with optical interconnects and glass substrates likely just got a step closer. source

---

Intel Demonstrates First Fully Integrated Optical I/O Chiplet

Intel’s optical compute interconnect chiplet is expected to revolutionize high-speed data processing for AI infrastructure.

 

What’s New: Intel Corporation has achieved a revolutionary milestone in integrated photonics technology for high-speed data transmission. At the Optical Fiber Communication Conference (OFC) 2024, Intel’s Integrated Photonics Solutions (IPS) Group demonstrated the industry’s most advanced and first-ever fully integrated optical compute interconnect (OCI) chiplet co-packaged with an Intel CPU and running live data. Intel’s OCI chiplet represents a leap forward in high-bandwidth interconnect by enabling co-packaged optical input/output (I/O) in emerging AI infrastructure for data centers and high performance computing (HPC) applications.

 

“The ever-increasing movement of data from server to server is straining the capabilities of today’s data center infrastructure, and current solutions are rapidly approaching the practical limits of electrical I/O performance. However, Intel’s groundbreaking achievement empowers customers to seamlessly integrate co-packaged silicon photonics interconnect solutions into next-generation compute systems. Our OCI chiplet boosts bandwidth, reduces power consumption and increases reach, enabling ML workload acceleration that promises to revolutionize high-performance AI infrastructure.”

 

What It Does: This first OCI chiplet is designed to support 64 channels of 32 gigabits per second (Gbps) data transmission in each direction on up to 100 meters of fiber optics and is expected to address AI infrastructure’s growing demands for higher bandwidth, lower power consumption and longer reach. It enables future scalability of CPU/GPU cluster connectivity and novel compute architectures, including coherent memory expansion and resource disaggregation.

Why It Matters: AI-based applications are increasingly deployed globally, and recent developments in large language models (LLM) and generative AI are accelerating that trend. Larger and more efficient machine learning (ML) models will play a key role in addressing the emerging requirements of AI acceleration workloads. The need to scale future computing platforms for AI is driving exponential growth in I/O bandwidth and longer reach to support larger processing unit (CPU/GPU/IPU) clusters and architectures with more efficient resource utilization, such as xPU disaggregation and memory pooling.

Electrical I/O (i.e., copper trace connectivity) supports high bandwidth density and low power, but only offers short reaches of about one meter or less. Pluggable optical transceiver modules used in data centers and early AI clusters can increase reach at cost and power levels that are not sustainable with the scaling requirements of AI workloads. A co-packaged xPU optical I/O solution can support higher bandwidths with improved power efficiency, low latency and longer reach – exactly what AI/ML infrastructure scaling requires.

As an analogy, replacing electrical I/O with optical I/O in CPUs and GPUs to transfer data is like going from using horse-drawn carriages to distribute goods, limited in capacity and range, to using cars and trucks that can deliver much larger quantities of goods over much longer distances. This level of improved performance and energy cost is what optical I/O solutions like Intel’s OCI chiplet emerging bring to AI scaling.

 

Intel Corporation’s Integrated Photonics Solutions (IPS) Group has demonstrated the industry’s first fully integrated optical compute interconnect (OCI) chiplet co-packaged with an Intel CPU and running live data. Intel’s OCI chiplet enables co-packaged optical input/output in emerging AI infrastructure for data centers and high performance computing applications. (Credit: Intel Corporation)

Intel Corporation’s Integrated Photonics Solutions (IPS) Group has demonstrated the industry’s first fully integrated optical compute interconnect (OCI) chiplet co-packaged with an Intel CPU and running live data. Intel’s OCI chiplet enables co-packaged optical input/output in emerging AI infrastructure for data centers and high performance computing applications. (Credit: Intel Corporation)

How It Works: The fully Integrated OCI chiplet leverages Intel’s field-proven silicon photonics technology and integrates a silicon photonics integrated circuit (PIC), which includes on-chip lasers and optical amplifiers, with an electrical IC. The OCI chiplet demonstrated at OFC was co-packaged with an Intel CPU but can also be integrated with next-generation CPUs, GPUs, IPUs and other system-on-chips (SoCs).

This first OCI implementation supports up to 4 terabits per second (Tbps) bidirectional data transfer, compatible with peripheral component interconnect express (PCIe) Gen5. The live optical link demonstration showcases a transmitter (Tx) and receiver (Rx) connection between two CPU platforms over a single-mode fiber (SMF) patch cord. The CPUs generated and measured the optical Bit Error Rate (BER), and the demo showcases the Tx optical spectrum with 8 wavelengths at 200 gigahertz (GHz) spacing on a single fiber, along with a 32 Gbps Tx eye diagram illustrating strong signal quality.

The current chiplet supports 64 channels of 32 Gbps data in each direction up to 100 meters (though practical applications may be limited to tens of meters due to time-of-flight latency), utilizing eight fiber pairs, each carrying eight dense wavelength division multiplexing (DWDM) wavelengths. The co-packaged solution is also remarkably energy efficient, consuming only 5 pico-Joules (pJ) per bit compared to pluggable optical transceiver modules at about 15 pJ/bit. This level of hyper-efficiency is critical for data centers and high-performance computing environments and could help address AI’s unsustainable power requirements.

About Intel’s Leadership in Silicon Photonics: As a market leader in silicon photonics, Intel leverages more than 25 years of internal research from Intel Labs, which pioneered integrated photonics. Intel was the first company to develop and ship silicon photonics-based connectivity products with industry-leading reliability at high volume to major cloud service providers.

Intel’s main differentiator is unparalleled integration using hybrid laser-on-wafer technology and direct integration, which yield higher reliability and lower costs. This unique approach enables Intel to deliver superior performance while maintaining efficiency. Intel’s robust, high-volume platform boasts shipping over 8 million PICs with over 32 million integrated on-chip lasers, showing a laser failures-in-time (FIT) rate of less than 0.1, a widely utilized measure of reliability that represents failure rates and how many failures occur.

These PICs were packaged in pluggable transceiver modules, deployed in large data center networks at major hyperscale cloud service providers for 100, 200, and 400 Gbps applications. Next generation, 200G/lane PICs to support emerging 800 Gbps and 1.6 Tbps applications are under development.

Intel is also implementing a new silicon photonics fab process node with state-of-the-art device performance, higher density, better coupling and vastly improved economics. Intel continues to make advancements in on-chip laser and semiconductor optical amplifier (SOA) performance, cost (greater than 40% die area reduction) and power (greater than 15% reduction).

What’s Next: Intel’s current OCI chiplet is a prototype. Intel is working with select customers to co-package OCI with their SoCs as an optical I/O solution.

Intel’s OCI chiplet represents a leap forward in high-speed data transmission. As the AI infrastructure landscape evolves, Intel remains at the forefront, driving innovation and shaping the future of connectivity.

More Context: Intel Silicon Photonics (Intel.com)

Editor’s Note: On June 28, 2024, an acronym incorrectly referencing “state-of-the-art” was removed and corrected in this story.

By John LoefflerMichelle Rae Uy

Intel Core i9-13900K

The best processor overall

SPECIFICATIONS

• Cores: 16
• Threads: 32
• Base clock: 4.5GHz
• Boost clock: 5.7GHz
• L3 cache: 80MB
• TDP: 170W

REASONS TO BUY

• Dominates both single- and multi-core workload
• Same pricing as the Core i9-12900K
• Top-tier gaming performance
• Surprisingly low power consumption on average

REASONS TO AVOID

• Runs hot and hungry at maximum load
• Lacks AMD’s 3D V-cache for gaming
• Basically overkill for most users

The Intel Core i9-13900K already had a very high bar to clear given the outstanding performance of the Intel Core i9-12900K, but when AMD released its Zen 4 processors in September 2022 to wide acclaim, Intel’s big.LITTLE follow-up had to essentially do everything right.

Incredibly, Team Blue manages just that with this chip. Whether it’s multithreaded workloads or single-threaded ones like gaming, the Core i9-13900K simply blows away the competition here, which is all the more impressive given the phenomenal performance of the rival AMD Ryzen 9 7950X processor.

What’s more, Intel was able to do so without raising the price of the chip over the previous generation, an incredible feat given the inflationary pressures on the semiconductor industry from still-wonky supply chains and the rising costs of raw materials and other production factors.

While still a pricey chip, the money spent is well worth the cost given the amount of performance you’re getting for that money. This is a chip that will serve you very well for the long-haul.

It must really hurt to be AMD, sometimes. Team Red has been the underdog since its inception in the late 1960s, formed by employees leaving the then-titanic Fairchild Semiconductor firm – just one year after Intel was established in the exact same way.

 

Intel had the drop on AMD, and for decades the company in the red corner was beholden to Intel’s will – as a second-source manufacturer, it spent less time making its own components and more time producing patented products for Team Blue. It wasn’t really until the turn of the century that AMD was able to properly kick out, finally managing to secure a chunk of the dedicated CPU market and expanding from there with the product lines we know and love today: Athlon, Bulldozer, and later Ryzen.

The recent launch of AMD’s new ‘Zen 4’ CPUs looked like it could be a turning point for the dark horse of the microprocessor market. We gave five-star reviews to both the Ryzen 9 7950X and the incredible mainstream Ryzen 7 7700X – generational performance was through the roof, blowing Intel’s current 12th-generation chips out of the water.Unfortunately, Intel is here to ruin the party. Only for AMD, though, to be clear; for us consumers, this party is going to rage all night long. The Intel Core i9-13900K is here, and it’s going to blow your socks off.

Yes, Intel’s 13th-gen ‘Raptor Lake’ CPUs have arrived, and now that we’ve tested them we can firmly say that the flagship chip is an absolute demon in almost every department. It’s not as big of a generational leap in performance when compared to the i9-12900K – unsurprising, given that it’s built on the same architecture while Ryzen 7000 used an entirely new 5nm process – but it’s enough to smack AMD’s best chips into next week.

Expect to see the i9-13900K shoot straight to the top of our most powerful processors list, is basically what we’re saying. This CPU dominates in everything from gaming to number-crunching, delivering best-in-class performance in single- and multi-core workloads.

We run an awful lot of tests when reviewing a new CPU: games, synthetic benchmarks, rendering tests, power, and thermal monitoring. We’re not going to show you our horrible, enormous spreadsheets filled with test results, but we will say this. Out of twenty-four averaged test results, the i9-13900K beat the Ryzen 9 7950X in all but six of them.

Oh, and like one final nail in AMD’s coffin, it’s more than a hundred bucks cheaper. We won’t mince words: this is the best consumer desktop CPU ever made, putting the championship belt squarely back on Intel’s muscular waist. Good job, Team Blue; better luck next time, Team Red.

The AMD Ryzen 9 7950X is finally here, and it’s hard to argue that this isn’t the best consumer processor ever made.

This is a big claim, and with Intel Raptor Lake quickly approaching, we’ll know soon enough if Team Blue is going to be able to counter a very resurgent AMD, but we are not going to sugar coat it: AMD has set the bar very high with the Ryzen 9 7950X, and Intel is going to be hard pressed to even keep pace with what this processor brings to the game.

To start, AMD used TSMC’s 5nm node to produce the new Zen 4 Ryzen 7000 series chips, and the Ryzen 9 7950X is a 16-core, 32-thread beast with 80MB cache memory and a boost clock within sight of 6.0GHz. What’s more, unlike Intel’s turn to big.LITTLE architecture, all 16 of the Ryzen 9 7950X’s cores are full performance cores, so while Intel might be using the Arm-pioneered performance core-efficiency core pairing, the Ryzen 9 7950X simply runs roughshod through tasks, beating out just about every other competing chip out there, across every category.

In particular, the Ryzen 9 7950X is easily the most significant gen-on-gen performance leap in as long as we can remember. Intel Alder Lake impressed us as much as it did because it was coming after the rather ‘Meh’ 11th-gen Rocket Lake chips, which were barely enough of an improvement to warrant the refresh in the first place.

The AMD Ryzen 5000-series processors were some of the best processors ever produced, top to bottom, and the Ryzen 9 7950X feels like as big a jump as Intel Alder Lake was, but it is instead making the leap from a mountain top rather than the boggy plateau of Rocket Lake and so it lands significantly farther out. It’s honestly like AMD skipped the release of an entire processor generation and we’re getting this chip after the one we didn’t even realize that we didn’t get was skipped.

It may be too soon to tell if AMD has finally caught up to Intel post-Alder Lake, but the Ryzen 9 7950X, along with the AMD Ryzen 7 7700X, is one hell of an opening salvo to kick off the next generation of processors.

• How much does it cost? MSRP $699 (about £600 / AU$1,000)
• When is it available? September 27, 2022
• Where can you get it? Available in the US (UK and Australia dates forthcoming)

The AMD Ryzen 9 7950X is available from September 27, 2022, with a US MSRP of $699 (about £600 / AU$1,000). AMD has not listed UK or Australian retail pricing, and we’ve reached out to AMD for clarification on when the new Ryzen 7000 processors will go on sale in those regions.

The Ryzen 9 7950X’s price is $100 less than the Ryzen 9 5950X’s, which is fantastic, especially with recent concerns about inflation. This also mildly offset’s the cost of the upgrade to the new Ryzen 7000’s. This is also $110 more expensive than the Intel Core i9-12900K, which has a US MSRP of $589 (about £500, AU$850). Is the Ryzen 9 7950X worth the premium price over the Core i9-12900K? We definitely think it is, but there’s no getting around the fact that this is still an enthusiast chip, and it is priced accordingly.

In a lot of ways, other than your graphics card, you really are talking about a substantial investment to upgrade to the AMD Ryzen 9 7950X. You might be better off buying one of the best gaming PCs preloaded with a Ryzen 9 7950X, depending on how much you end up needing to replace in your rig.

Still, this isn’t unexpected or unreasonable, considering the age of the AM4 socket and how Intel has similarly shifted to the LGA 1700 socket for its Intel Alder Lake chips. On the plus side, if you had an AM4 cooling solution, you can still use it with the Ryzen 9 7950X.

All that said, expect to spend a good deal more than $699 for this upgrade. Given the performance improvements with the Ryzen 9 7950X, we definitely think you are getting way more value for your money than you would with the Ryzen 9 5950X it replaces.

The AMD Ryzen 9 7950X might not be as good a “value” proposition as the Ryzen 7 7700X, but it is without a doubt a much more compelling processor than its predecessor. It comes in slightly behind the Core i9-12900K on price, but with rumors swirling about the increased prices of the Intel Raptor Lake processors due out soon, it’s more likely than not that the Ryzen 9 7950X will only look like a better value as time goes on.

source 1 

source 2

source 3