Zen 3 presentation finally happened, and it is fair to say it met very high expectations. Per AMD, it has 19% increase in IPC, which is better than the consensus of around 12-15% increase. As rumored, Zen 3 now has 8 cores CCX. This significantly decreased latency, as opposed to the previous 4 cores CCX and the need for communication between each cores. There is also bigger and better cache and fancier branch prediction, cache fetching etc etc. One thing people seemed to be disappointed about is the highest clock only reaches 4.9 GHz, not 5 GHz. However, there is nothing magical about the number 5. It is only psychological.
Bottom line is, Zen 3 offers 25% higher performance over last gen and even better performance per watt. Zen 3 now has better gaming performance (like single core and lightly threaded) than Intel’s 10900X. This is Intel’s last stronghold in CPU performance. AMD now claims to take the gaming crown. Certainly, we will need third party benchmarks for verification when Zen 3 comes out.
Anandtech has a good summary, here.
The best part is, Lisa Su declared Zen 3 to be available worldwide on Nov. 5. Last year, there were a lot of difficulty for people to buy Ryzen 3900X and 3950X. I believe AMD had learned from that. I am betting supply will be good this time around. Nvidia is doing a paper launch with their 3080 GPU, with low supply lasting well into 2021, per VideoCardz, here.
The final issue is the price. AMD is now raising the price by $50 over the previous gen for each corresponding SKU. Buyers will complain, but this will help a lot with profit margin. Also, AMD is no longer including cooling fans with the top two SKUs. The early adopters and the very same people who will pay top dollars for the Nvidia 3090 cards can certainly pay the higher prices. The people who are more price sensitive, they can still buy the Ryzen 3000 CPUs or wait for the inevitable price drop later on.
Then on the night of Zen 3 announcement, WSJ came out with the bombshell, that AMD is in serious talk to buy Xilinx for around $30B!
I had previously written about this possibility back in August, here.
Certainly, I got this idea from many wise people, who had mentioned this as being desirable. Of note, RetiredEngineer Chiakokhua had twitted this back in Jan 2019.
Over on Reddit, Long_on_AMD had written a nice post, here. He linked a very nice article from EEJournal regarding Xilinx’s ACAP, here.
Another Redditor jibaikia wrote this post, also very encouraging.
Hans Mosesmann of Rosenblatt and Jefferies had each voiced support for this deal. I don’t have the direct quotes available.
The stock market had not reacted favorably today. AMD is currently down to around 83. For this deal to go through, AMD will need to issue more shares, and maybe even issue some bonds. Moody had just raised AMD to investment grade recently. AMD had also famously got the permission from the shareholders to issue some 700M additional shares just two years ago. So the fear for share dilution is real. However, perhaps AMD plus Xilinx can be synergistic. AMD will gain Xilinx’s revenue and earning. Xilinx CEO Victor Peng was at ATI and AMD prior to Xilinx. AMD and Xilinx has a history of working together. This is all going to be very interesting! I think Lisa Su knows what she is doing and AMD will come out stronger and better.
AMDinvestorblog 