News/features about semiconductors and PCs from
@CRN
Senior Editor Dylan Martin. Send tips to dmartin
@thechannelco
.com or dylanljmartin.81 on Signal.
FEATURE: Intel CEO Pat Gelsinger admitted that the company won’t be "competing anytime soon for high-end training" against Nvidia.
So it's hoping Gaudi 3 will find traction with businesses who need cost-effective AI systems for smaller, task-based models.
If you're wondering why I didn't write anything about Intel's earnings or other semiconductor news most of this week, it's because I became a dad on Wednesday!
This means I'll be on leave from The Register until late October, so I'll see you all in three months!
NEW: LLM startup
@LaminiAI
revealed it has been “secretly running on more than 100”
@AMD
Instinct MI200 series GPUs and said the chip designer’s ROCm software platform “has achieved software parity” with
@nvidia
's dominant CUDA platform for LLMs. via
@CRN
In researching for a story, I realized that Intel edited its 2019 press release about Aurora to remove the mention of it being "the first exascale supercomputer" and to change the delivery date from 2021 to 2022. You don't often see companies editing old press releases like this.
The quote about
@AMD
's ROCm platform having "software parity" with
@nvidia
's CUDA platform for large language models came, interestingly, from a former Nvidia CUDA software architect who co-founded the startup.
NEW: LLM startup
@LaminiAI
revealed it has been “secretly running on more than 100”
@AMD
Instinct MI200 series GPUs and said the chip designer’s ROCm software platform “has achieved software parity” with
@nvidia
's dominant CUDA platform for LLMs. via
@CRN
Nvidia CEO Jensen Huang on the new Intel-AMD x86 advisory group: "The fact that the architecture was fragmenting isn’t good for the industry. So I love what they’re doing." More comments from the CEOs of Nvidia, Intel and AMD at Lenovo Tech World here:
Big takeaway: Nvidia has developed a new feature called in-flight batching that “allows work to enter the GPU and exit the GPU independent of other tasks."
This keeps "the GPU 100% occupied. This all happens automatically ... and that dramatically improves H100 efficiency."
Good morning from
@intel
's Ocotillo manufacturing campus in Chandler, Arizona. I am going on a tour of Fab 42 in a little while, but in the meantime, here's a close-up of Intel's upcoming Core Ultra Meteor Lake processor.
#intel
Lisa Su just now on AMD-Xilinx integration progress:
"As one example, we are integrating Xilinx’s differentiated AI engine across our CPU product portfolio to enable industry-leading inference capabilities, with the first products
expected in 2023."
#amd
$amd
.
@AMD
said its forthcoming 256-GB Instinct MI325X GPU can outperform
@Nvidia
’s 141-GB H200 processor on AI inference workloads and vowed that the next-generation MI350 accelerator chips will improve performance by magnitudes.
Brad McCredie, AMD's vice president of data center and accelerated processing, explains the building blocks of the Instinct MI300X GPU in response to a question from
@CDemerjian
.
Big
#Intel
news drop:
Intel Data Center and AI Group General Manager Sandra Rivera to transition to new role as head of the company's Programmable Solutions Group.
Intel plans to turn PSG into a standalone business, effective Jan. 1, 2024, and seek an IPO for PSG. $intc
Scoop:
@intel
has promoted its top graphics executive,
@AMD
veteran Raja Koduri, to the executive rank, signaling the chipmaker's satisfaction so far with its fledgling – but strategically important – discrete GPU business:
#intel
$intc via
@TheRegister
Two customers represented 25% of
@AMD
's net revenue last year. The one representing 14% mainly bought from the Enterprise, Embedded and Semi-Custom segment while the one representing 11% mainly bought from Computing and Graphics, according to a new SEC filing.
#amd
$amd
My big project is out!
@CRN
's October cover story is about
@intel
CEO Pat Gelsinger and his plan to rethink the company's software strategy, which includes the potential for new paid software offerings:
#intel
$intc
NEW: AMD has hired Intel executive John Rayfield, who led the development of low-power AI technologies for the semiconductor giant’s upcoming Core Ultra “Meteor Lake” processors. via
@CRN
#intel
#amd
Intel's datacenter boss admitted that the new Sapphire Rapids delay will help AMD catch up quicker as the x86 giant has hyped up the new Xeon chip as a 'leadership' product:
#intel
#amd
via
@TheRegister
Takeaways from my Q&A with
@AMD
exec David McAfee about the company's AI PC strategy:
-- Ryzen 8040 exists because OEMs, ISVs and customers wanted "more [NPU] capability."
-- In near future, the NPU will become a "multitasking machine" due to column-based architecture.
If you're looking at the BMW news from South Korea and somehow think software-defined vehicles will not become a wider phenomenon, I have an excerpt from an Nvidia investor meeting you should read:
Scoop: Head of Intel Foundry Services resigns just as it gets going
Randhir Thakur led Intel’s big bet to take on Asian foundry giants TSMC and Samsung
#intel
$intc tip
@Techmeme
Also of note:
@AMD
is using
@LaminiAI
to fine-tune LLMs for internal use.
“We’ve deployed Lamini in our internal Kubernetes cluster with AMD GPUs and are using fine-tuning to create models that are trained on AMD code base across multiple components for specific dev tasks."
NEW: LLM startup
@LaminiAI
revealed it has been “secretly running on more than 100”
@AMD
Instinct MI200 series GPUs and said the chip designer’s ROCm software platform “has achieved software parity” with
@nvidia
's dominant CUDA platform for LLMs. via
@CRN
I asked Lenovo's North American president about demand for AMD's Instinct MI300X GPU, and he gave a pretty interesting response. Note that this was before Nvidia's Blackwell announcement, but I think the comment is still very relevant.
#amd
#nvidia
.
@nvidia
CEO Jensen Huang took a break from wearing his iconic black leather jacket on Tuesday to sport one infused with fuchsia at a sales event for the solution provider powerhouse that claims the color as its signature hue:
@InsightEnt
.
@CRN
Intel CEO Pat Gelsinger was just getting grilled in today's senate committee hearing by Sen. Rick Scott (R-Fla.) on doing business in China when the company is trying to get U.S. subsidies for manufacturing. Some important context on how much money Intel makes in China:
#intel
I asked AMD exec Forrest Norrod about the company's view on Nvidia making such a big emphasis on its Grace Hopper/Grace Blackwell Superchips and whether an Instinct MI300A-like chip could fulfill a similar range of workloads. Here's what he said:
Ian Steiner, a lead architect on Xeon CPUs, has left
@intel
to become a distinguished engineer at
@nvidia
, per LinkedIn.
He was principal engineer for Cascade Lake and was also a power/perf architect for all Xeons since Nehalem, according to
@TheNextPlatform
.
#intel
#nvidia
Nvidia has revealed that it plans to release successors to its powerful H100 for AI training and inference in 2024 and 2025, reflecting a move to a new yearly release cadence from its previous strategy of launching AI chips every two years. LINK: via
@CRN
AMD reached a new record in overall x86 CPU market share against Intel last year, mainly thanks to gains it made in the PC and server segments.
I dived deep into Mercury's latest report and got more details from its author:
#amd
#intel
$amd $intc
@CRN
Jensen Huang dropped the news during an investor event today:
"I've selected Sapphire Rapids to be the CPU for Nvidia Hopper. [...] We're qualifying it for our own server, our own DGX. We're qualifying it for our own supercomputers."
#nvidia
#intel
.
@AMD
is hosting its Accelerated Data Center Premiere on Nov. 8, where it will share "upcoming innovations with AMD EPYC processors and AMD Instinct accelerators"
#amd
$amd
Japanese investment giant SoftBank Group is reportedly in advanced talks to acquire Graphcore, a British AI chip designer that has struggled to gain traction in recent years. via
@CRN
.
@Intel
announced on Monday that it has launched an AI cloud service featuring its new Gaudi 3 accelerator chips, saying that it will help underpin a new enterprise AI solution with hybrid cloud capabilities from generative AI startup
@inflectionAI
.
@CRN
AWS parent company Amazon plans to acquire AI chipmaker and model specialist Perceive Corp. for $80 million in cash to boost the company’s large language models and edge computing capabilities. via
@CRN
@markharanas
One Nvidia partner told me that comparable performance isn't enough for his customers to switch from x86 to Arm.
"There needs to be some large, clear selling point of why they need to change to a new architecture."
.
@nvidia
has revealed its next-gen Blackwell GPU architecture as the much-hyped successor to the AI chip giant’s Hopper platform, claiming it will enable up to 30x greater inference performance and consume 25x less energy for massive AI models.
#GTC24
When
@AMD
was launching its first-generation Epyc chips in 2017, the company's CPUs only accounted for six of the world's fastest 500 supercomputers.
Now AMD's CPUs are in 93 of the top 500, according to the spring 2022 update that was just released:
#amd
If you see the Tom's Hardware story about Intel sales and marketing cuts, please refer to my original story instead. TH misinterpreted what I wrote, saying the marketing budget is getting slashed by 35%, instead of the Sales and Marketing Group's costs getting cut by that much.
SCOOP: Intel’s partner-centric Sales and Marketing Group (SMG) plans to cut jobs, reduce marketing expenses and “simplify programs end-to-end” as part of a directive to slash costs by more than 35 percent by the end of the year,
@CRN
has learned.
#intel
SCOOP: Intel’s partner-centric Sales and Marketing Group (SMG) plans to cut jobs, reduce marketing expenses and “simplify programs end-to-end” as part of a directive to slash costs by more than 35 percent by the end of the year,
@CRN
has learned.
#intel
NEW:
@AMD
has made its second AI software startup acquisition in less than two months. The
@nod_dot_ai
acquisition, announced today, is all about enhancing the chip designer’s open software capabilities to better compete with Nvidia. LINK:
#amd
via
@CRN
NEW:
@Nvidia
is getting partners in North America and other regions set up to sell the A800 GPU, an AI chip originally designed to sidestep U.S. export restrictions against China before new rules halted sales to the country last month. via
@CRN
#nvidia
I recently talked to top
@AMD
data center exec Forrest Norrod about the company's decision to move to an annual data center GPU release cadence to fight Nvidia's AI dominance: "We see the opportunity is so great that we felt we had no choice."
Full Q&A:
12 big announcements at Computex 2024:
- Nvidia data center road map
- MGX Blackwell support
- Ryzen 9000
- MI325X, annual Instinct releases
- Ryzen AI 300
- Xeon 6 E-core
- Gaudi 3 platform price
- Lunar Lake
- EPYC 'Turin'
- Panther Lake
- RTX AI PC
- Nvidia NIM launch
Intel's latest acquisition has a very interesting lineage in graphics:
The founders started with the demogroup Future Crew in the early 1990s.
Then founded GPU maker BitBoys in 1991.
Then got acquired by AMD in 2006.
Then got sold in 2009 to Qualcomm, where they made Adreno.
Chip design superstar AMD is acquiring server supplier and cloud computing specialist ZT Systems for $4.9 billion in a blockbuster move to take AMD’s artificial intelligence portfolio to the next level. via
@CRN
@markharanas
PSA: I am not the same person as
@dylan522p
. I am
@DylanLJMartin
.
@dylan522p
is an analyst.
I am a journalist.
We both cover the semiconductor industry.
Thank you. You can now go about your day.
After 27 years of providing in-depth coverage of the amazing world of PC and mobile hardware, AnandTech is saying farewell. We want to thank everyone from the AnandTech community for their support and passion for what we’ve done over the years
Chip companies like AMD, Intel and Qualcomm are taking a "build it and they will come" approach when it comes to the AI PCs they are enabling. From my primer on the rise of so-called AI PCs. LINK:
My colleague
@ChannelWritrGuy
interviewed
@Supermicro_SMCI
CEO Charles Liang about the company's fast-growing AI system business and how it differs from rivals, among other things. He also said Supermicro is getting into AI PCs?!
@CRN
$SMCI
Google was the third largest provider of data center processors by revenue last year, just behind Intel and ahead of AMD and, by a larger margin, Amazon Web Services, according to a new report published Tuesday by semiconductor research firm TechInsights.
Intel buried the news of its AI cloud service in the press release about its Inflection AI partnership.
Thankfully, I spoke to execs at the two companies to dive into the details about both aspects of the announcement.
.
@Intel
announced on Monday that it has launched an AI cloud service featuring its new Gaudi 3 accelerator chips, saying that it will help underpin a new enterprise AI solution with hybrid cloud capabilities from generative AI startup
@inflectionAI
.
@CRN
Several months after Pat Gelsinger became Intel's CEO in 2021, he told me in an interview that his biggest concern in the data center wasn't Arm.
Instead, he saw a bigger threat in Nvidia and its "uncontested" hold over the AI computing space, vowing to challenge its dominance.
I rounded up the biggest news from the Intel Vision event last week. Most people likely missed that
@intel
rebranded all of its commercial software and services under the new Tiber name. Granulate, for instance, is now Intel Tiber App-Level Optimization.
My
@CRN
cover story on
@intel
CEO Pat Gelsinger from last year won the gold award in the individual profile category for
@ASBPE
's 2022 Azbee Awards of Excellence in the New England region. That's a mouthful, but I am very honored!
It's been one year since I started a dedicated account for sharing reporting on the semiconductor industry by me and others. Thanks to everyone who has followed and interacted with me in the past 12 months. You keep me tweeting!
#MyTwitterAnniversary
Intel has hired Samsung Foundry U.S. executive John Zucker as vice president of business development, according to his LinkedIn page.
Not the first time Intel has poached talent from Samsung Foundry, as I previously reported:
Intel has begun mass layoffs as part of the chipmaker’s cost-cutting measures, impacting more than 2,000 jobs at locations in Oregon, Arizona and California, according to notices it sent to the three states Tuesday. via
@CRN
AMD says transistor tech will keep Moore’s law alive for 6 to 8 years
'We're going to have lower power, but it's going to cost more' says chipmaker's CTO via
@TheRegister
@TheDanRobinson
.
@intel
's head of R&D for the Intel 4 node talked about how EUV will bring down costs for future chips like Meteor Lake, even though EUV itself is expensive. He also discussed other advancements to make the chips faster, including a 2X in the capacitance.
To commemorate the launch of
@intel
's Gaudi 3 accelerator chip and my new feature about Intel's Gaudi 3 strategy, I present to you the third entry in my "Help, I'm Stuck In An Accelerator Chip!" series.
.
@intel
Q1 revenue:
Client Computing Group: $9.3bn, down 13% YoY
Datacenter and AI Group: $6bn, up 22%
Network and Edge Group: $2.2bn, up 23%
Accelerated Computing and Graphics: $219m, up 21%
Intel Foundry Services: $283m, up 175%
Mobileye: $394m, up 5%
#intel
$intc
.
@intel
has begun high-volume production of advanced chips using the Intel 4 node and debuting at the end of the year, a key milestone in CEO Pat Gelsinger’s ambitious manufacturing comeback plan. via
@CRN
#intel
.
@Intel
's
@HabanaLabs
unit is revealing two new deep learning chips, Gaudi2 for training and Greco for inference, and it claims Gaudi2 can outperform
@Nvidia
's A100 by quite a lot. The question is, how will things look when H100 comes out?
#IntelOn
#intel
Some recent personnel moves:
Karen Kwong, who was head of SoC design engineering at
@AMD
, has joined
@intel
as vice president of physical design at the Accelerated Computing Systems and Graphics group, per LinkedIn:
#intel
#amd
.
@Microsoft
CEO Satya Nadella says
@AMD
's Instinct MI300X GPU offers the 'best price-performance' on GPT-4 inference. These were among the statements Nadella made in his
#MSBuild
keynote, as captured by my colleague
@WadeMillward
who's been at the event.
.
@intel
is planning to introduce "seamless firmware updates" for future Xeon CPUs,
@GregL_Intel
just said in an Intel Vision panel. This means you won't need to reboot the system to do firmware updates, which Lavender said will be of big help to cloud providers.
#IntelOn
#intel
.
@intel
said its upcoming Gaudi 3 AI accelerator chip can best Nvidia’s powerful H100 GPU for training large language models and offer similar or, in some cases, better performance than the rival’s memory-rich H200 for large language model inferencing.
What has
@Intel
accomplished under Pat Gelsinger's leadership so far?
In this in-depth recap, I go through the company's nine biggest moves since
@PGelsinger
became CEO in 2021 and what has happened since those developments were announced.
@CRN
#intel
I wrote about five new Intel technologies I saw on the show floor at
#IntelInnovation
, two of which have never been written about before. via
@CRN
#intel
Exclusive:
@intel
's plan to sell a stake in its
@AlteraFPGA_
programmable chip business and push for an initial public offering of the unit by 2026 “has not changed,” Altera CEO Sandra Rivera told me in an interview on Tuesday.
#intel
$intc via
@CRN
.
@nvidia
has hired JP Vasseur, a 25-year Cisco Systems engineering veteran, once credited as the switching giant’s most prolific inventor, who will lead development of AI and networking architecture at the AI computing giant. via
@CRN
In the facing of growing competition from Arm-based chip designers,
@intel
and
@AMD
say several tech giants—including
@Microsoft
,
@Dell
,
@Lenovo
and
@HP
—are joining their new advisory group to "shape the future of x86 and foster developer innovation."
@CRN
.
@intel
said its latest Xeon processors present “huge” opportunities for channel partners in the AI computing space, whether the CPUs are used for inferencing or as the head node for systems accelerated by expensive, energy-guzzling chips like Nvidia GPUs.
Intel said it’s upending a norm of secrecy in the data center accelerator chip space by revealing the list price of the star server platform for its upcoming Gaudi 3 AI chip. via
@CRN
Analysis: AWS follows AMD and Intel down the specialized chips path
CPU roadmaps are fragmenting to serve a variety of needs. Time to do some homework via
@TheRegister
Some news for me: I am leaving
@CRN
after today for a new journalism opportunity covering the semiconductor space.
I really appreciate everyone here who has followed, read and shared my work. I look forward to providing you even more semiconductor goodness in the near future!
AMD said PC CPU sales grew in the double digits in the third quarter, continuing a recovery from a slowdown that has dragged down the chip designer and other companies in the computer industry. via
@CRN