I had to know exactly what I want to force CDW to sell AMD to the organization I work for.

If I were any less resolute, they’d force Intel down my throat.

All of those examples were research that Intel initiated.
Support for MRDIMMs and their vRAN layer 1 acceleration are more examples. Intel puts years of R&D into these. Now they’re nearing products with with BSPD and, in a few years, glass substrates. And on the consumer side, they launched Thunderbolt 5 and WIFI7 IO this year. So, those are “lots of reasons” for Intel. What are the “lots of reasons” for AMD?

For a lot of reasons, yes.

“He who controls HBM controls AI”
Intel demoed a chiplet with 4Tb optical bandwidth using 8 channels of light per fiber… so 64 channels of optical pcie5 streaming with a single chiplet. Their current Gaudi 3 is using 24 x 200Gb IO on copper.

Some Aurora jobs process data that is streamed from the Advanced Photon Source. It is too big to be saved anywhere … so channel input data is already a streaming solution. The question is … why can’t model parameters and feature data be streamed and broadcasted instead of counting on all the hbm storage?

because AMD is leading the research on new technologies like AMX, AVX10.2, PCIE6, Optical IO, CXL?

Yes. This is why you don’t type a last thought live into your story when you have the flu….

On channel on financial assessment, I have AMD Epyc at 1,365,593 units in q3 that is sufficient volume for the second half. AMD appears to have pulled cost out of q4 by producing a sufficient volume of Epyc and Ryzen performance desktop in q3 for the remainder of the year. All up q1 – q3 I have Epyc at 6,518,945 units produced.

I have Intel Sierra Forest and Granite Rapids at 1,119,000 units in q3 and approximately x2 (2.2 M) produced that is a rationale guestimate on SF and Granite suspect q2 and q3 ‘sample price’ reconciled on INTC 10Q DCAI so said revenue. Continuing Intel q3 plus 5,031,272 NEX associated Xeon for 6,150,272 DCAI + NEX units produced. For the year thus far, 15 M to 18 M units Xeon and in q2 it appears DCAI moved entirely over to ‘3’ production.

Component share I will offer up my complete take and then point to the full AMD and Intel reports subject encompassing assessment beyond the traditional syndicated sources of research and assessment.

Intel and AMD component share on q3 production,

Server; 81.9% and 18.1% and Mercury states 24.2%
Desktop; 57.5% and 42.4% and Mercury states 28.7%
Mobile; 77.4% and 22.5% and Mercury states 24.2%
All x86; 68.3% and 31.6% and Mercury states 23.9%

If you add AMD embedded accelerator into server, AMD gains to 20.5%

Intel and AMD component share on q3 channel data, server, desktop, mobile;

Mercury states AMD server at 24.2% and we rely on the same ebay channel supply data where I attempt to reconcile Mercury’s approach and here’s the result.

Server back to Rome and Cascade Lakes
AMD = 19.7% and Intel = 80.3%

Server back to Milan and Ice
AMD = 34.7% and Intel – 65.3%

Server back to Genoa and Sapphire Rapids
AMD = 44.6% and Intel = 55.4%

Mercury states AMD desktop at 28.7%

Desktop back to Vermeer R5K and Alder 12th
AMD = 34.4% and Intel = 65.6%

Desktop back to Raphael R7K and Raptor 14th
AMD = 50% and Intel = 50%

Mercury states AMD mobile at 22.3%

Mobile back to R7K and Raptor 13th
AMD = 20% and Intel = 80%

Mobile back to Alder 12th and Rembrandt R6K
AMD = 17.6% and Intel = 82.3%

Mobile dropping R6K but leaving Alder 12th
AMD = 11.7% and Intel = 88.3%

Mobile back to C5K and Tiger 11th H
AMD = 24.5% and Intel 75.5%

x86 Total

Mercury states 23.9%

Total back to RK6 mobile and Alder 12th
AMD = 23.6% and Intel = 76.3%

Total back to Sienna/4000 and R7Ks and Sapphire Rapids and Core 14th
AMD = 33.8% and Intel = 66.1%

Dropping Sapphire Rapids AMD gains to 36.7%

AMD full report is here in comment string:

https://seekingalpha.com/article/4731391-amd-q3-buy-the-dip-as-data-center-growth-is-what-counts

Intel full report is here in comment string;

https://seekingalpha.com/article/4731858-intel-corporation-intc-q3-2024-earnings-call-transcript

Mike Bruzzone, Camp Marketing

He who controls HBM controls AI.

Am always left wondering why AMD doesn’t 3x its quantity of laptop chips, desktop chips, and server chips. How is say nvidia able to scale up to 30b in revenue with assumed a massive increase in h100 sales which would require a large supply but AMD is never able to no matter how competitive its product especially in the laptop space.