Monday, January 6th 2025
AMD 2025 International CES Keynote Address Liveblog
AMD is at the 2025 International CES, and we join them for a brisk run of what they have in store for us through the year. We expect some important announcements, such as the next-gen Radeon RX gaming GPUs powered by RDNA 4, a new flagship desktop processor powered by 3D V-Cache technology, HX-segment Ryzen 9000 series processors for gaming notebooks and portable workstations, the fabled "Strix Halo" chip AMD plans to surprise Apple with; and a fleshing out of its mobile processor portfolio.
19:02 UTC: The show is underway.
19:02 UTC: Jack Huynh takes centerstage.
19:04 UTC: CERN uses AMD technology for data-processing.
19:04 UTC: AMD announces Ryzen 9 9950X3D and Ryzen 9 9900X3D high core-count "Zen 5" processors with 3D V-Cache.
19:06 UTC: 9950X3D boosts up to 5.70 GHz. 8% performance increase over 7950X3D. 20% faster than Intel 285K at gaming—in some games even 35% faster. 13% performance increase over 7950X3D in productivity apps, and 10% faster than 285K in productivity workloads.
19:09 UTC: Available in March 2025.
19:09 UTC: Zen 5 X3D comes to the mobile platform with Fire Range. Ryzen 9 9955HX3D. Available in 1H-2025.
19:10 UTC: AMD just namedropped RDNA 4 and FSR 4.
19:10 UTC: CoD Black Ops 6 to support FSR 4.
19:14 UTC: "AI is the new electricity, it's going to be everywhere."
19:14 UTC: AMD expands Ryzen AI 300 series with 7 and 5 product tiers. Over 150 design wins.
19:16 UTC: Ryzen 200 series "Kraken Point" also announced. 30% faster than Lunar Lake. Over 24h battery life.
19:18 UTC: Ryzen AI MAX "Strix Halo" announced!
19:19 UTC: 16 "Zen 5" cores, 40 CU RDNA 3.5 iGPU, 256-bit LPDDR5X memory interface, 50 TOPS NPU.
19:22 UTC: Apple M4 Pro 12-core is matched, 14-core M4 Pro almost reached in performance.
19:23 UTC: 50+ ISVs working on Ryzen AI optimized applications.
19:25 UTC: Did you think "Strix Point" would be bulky? It's designed to go against the MacBook Pro.
19:26 UTC: AMD Ryzen AI PRO series announced—processors for commercial notebooks.
19:30 UTC: Wait what? No RDNA 4 announcement? I guess that's a wrap.
19:04 UTC: CERN uses AMD technology for data-processing.
19:04 UTC: AMD announces Ryzen 9 9950X3D and Ryzen 9 9900X3D high core-count "Zen 5" processors with 3D V-Cache.
19:09 UTC: Available in March 2025.
19:09 UTC: Zen 5 X3D comes to the mobile platform with Fire Range. Ryzen 9 9955HX3D. Available in 1H-2025.
19:10 UTC: CoD Black Ops 6 to support FSR 4.
19:14 UTC: "AI is the new electricity, it's going to be everywhere."
19:14 UTC: AMD expands Ryzen AI 300 series with 7 and 5 product tiers. Over 150 design wins.
19:22 UTC: Apple M4 Pro 12-core is matched, 14-core M4 Pro almost reached in performance.
19:23 UTC: 50+ ISVs working on Ryzen AI optimized applications.
19:25 UTC: Did you think "Strix Point" would be bulky? It's designed to go against the MacBook Pro.
57 Comments on AMD 2025 International CES Keynote Address Liveblog
Like i said - doing the same thing every time and expecting it to just work eventually will lead nowhere.
AMD needs exclusive features to create a selling point for their cards.Does not matter. People bough GTX Titan back in the day despite it being only 3% faster than 290X at nearly double the price.
And similar comparisons can be made with newer cards.
Selling their cards dirt cheap and hoping to grab some market share has not worked.
At worst this strategy will lead to no R&D money at best it will lead to Nvidia slightly lowering it's prices.
Then people will still buy Nvidia. The mindshare is too strong.
What i think AMD should do is put together a 6 year plan (3 generations) and execute like Zen did.
Consistent naming, predictable releases and chiplet innovation. That's the only way to grow mindshare.
Price and feature matter for sure but they're rendered meaningless if the GPU division is seen by everyone by constantly changing naming, broken promises and just generally aimlessly flip-flopping in the wind.
Software isn’t going to sell an AMD GPU. Ryzen has succeeded due to starting off with comparable performance at low and effective pricing. Followed by iteration to performance leadership in nearly every market space.
If the performance isn’t there, no developer is going to implement proprietary software/develop code when there’s a path of less resistance and entrenchment with Nvidia software.
You’re delusional otherwise. Better performance at lower prices is going to sell hardware when you have single digit market share.
The presentation for Strix Halo has a total shit show IMO.
When comparing against Lunar Lake, they said they left their CPU at 55W, against a 30W CPU that's already a stupid model - that i9 has a 30W TDP vs 17W of the other LNL CPUs to achieve a 100MHz higher boost clock. Let's not forget it's a 16c/32t CPU vs a 4p4e one.
Their comparison against the M4 Pro was also awful, they did not disclose the TDP used, and compared it to another 30W CPU, with not so impressive MT gains (apart from v-ray). 4x the power (assuming they used the 120W cTDP) for minor gains, impressive.
Adding to that, the comparison against the 4090 was also extremely non-sense. They used a 70B model that requires more than 35GB of memory on a 24GB GPU, which has to offload many layers to the CPU. And no comparisons against no other GPU whatsoever in any other scenarios.
Upscaling is pointless on high end GPUs as you’re just degrading visual quality to pump FPS due to tanking your performance for sake of partial ray tracing effectives, the majority of which are not worth it. I’ve said it many times before, but paying multiple grand on hardware such as gpus and monitors with low response time/high refresh rates for the best motion presentation/clarity, then adding ghosting, blurring, smearing, and texture destruction with upscaling and frame generation is peak levels of oxymoronic. And then where some of this software could serve best, at the low to midrange of gpus, it falls apart due to lack of horsepower (and vram is some cases) or lower input resolutions resulting in even worse outputs.
Theres never a bad gpu, just a bad price.
Although this method makes TAA achieve a result comparable to supersampling, the technique inevitably causes ghosting and blurriness to the image.[1]