Last month, AMD released the non-nuclear version of the Zen4 architecture R5 7600X - 7500F.
You have already seen various data evaluations and actual player experience, and it is not bad to say that it is a disguised price reduction.
The reason is also very simple, thanks to Intel.
Jon Peddie Research has just released a report showing that AMD client CPU shipments in the second quarter of 2023 fell by 5.3% compared with the previous quarter , while Intel increased by 23%.
Regardless of the pressure or not, the duel between Ryzen 7000 and the 13th generation Core has basically entered the final stage.
Now, it's time to look forward to the next generation of products from both parties.
Recently, performancedatabases exposed some CPU-Z information of AMD's new generation mobile processor code-named Strix.
4P+8E 12 cores and 24 threads (small cores also support hyper-threading), 1MB L2, 16MB L3 cache, the frequency should be an error in identification.
AMD Strix Point APU is similar to the current Phoenix APU, the chiplet is a Die-to-Die glue connection.
It uses the same large and small core design as Intel, but it's just something AMD wants or doesn't want.
Strix Point APU will most likely debut on mobile platforms in 2024 with the Ryzen 8050 series.
In addition to the large and small core design, I have to mention the core display of Strix Point.
According to current information, Strix Point APU will be equipped with 16CU RDNA 3.5 core display.
Now the performance of Radeon 780M with 12CU RDNA 3 is close to "outdated" standalone graphics.
In this wave of conversion, there should be no pressure to catch up to 1060 in the ideal state.
In fact, this is just an appetizer. What players are looking forward to is actually another monster player in the Strix Halo series.
16x Zen5 is pure big core, almost the twin brother of desktop Zen5.
Its nuclear display scale is not in the same order of magnitude as before, and it has reached 40CU .
Based on past experience, the performance of AMD GPUs of the same generation is basically equal to the number of Compute Units.
The frequency difference is not big, 16 to 40 CU is directly 2.5 times...
For comparison, the desktop standalone RX 7600 only has a 32 CU RDNA3 configuration.
With this wave of architectural improvements, Time Spy can’t score more than 10,000 points, directly comparable to mainstream independent graphics?
Of course, this is all performance under ideal conditions, and there is a high probability that the memory will be delayed.
It doesn’t have to compete with the high-end RTX 40, as long as it can achieve the performance of mainstream independent graphics, it can be considered as a dream of putting independent graphics back into the CPU.