We still don’t have definitive nomenclatures, but all indications are that the Ryzen 9 7950X will be the next-generation processor that AMD will launch this year. We know it will be manufactured in 5nm TSMC nodeand we also have a clear idea of ​​how its configuration will be internally, thanks to all the information provided by Lisa Su, CEO of Computex 2022, during the initial keynote.

The Ryzen 9 7950X and all new-generation processors configured with more than 8 cores will use two chipsets made on a 5nm TSMC node. Each chiplet will have:

• 8 Zen 4 cores.
• 8 MB L2 cache.
• 32 MB L3 cache accessible to all 8 cores.

In total, these two chips will cause the Ryzen 9 7950X to have 16 cores and 32 threads. Just below the two chipsets will be an I / O chip that will integrate DDR5 and PCIe Gen5 controllers as well as an AMD Radeon GPU based on the RDNA2 architecture. This chip will be manufactured in 6nm process from TSMC, which means that the Ryzen 7000 will use a 5nm node at the CPU level and a 6nm node at the I / O and integrated GPU level.

The Ryzen 7000 series will be limited to DDR5 memory configurationsas the first rumors pointed out, which means that if we want to jump on this new generation of processors, we will have to buy a new motherboard and also a new memory set, But is it really worth making such a leap? That’s a good question, but I think we should go even further and, after analyzing all the official information that AMD has provided, think about whether they can really compete with Intel.

Ryzen 9 7950X: 15% higher performance in one fiber does not equal 15% higher IPC

It is a bug that is spreading like an avalanche on the Internet, and we must stop. amd he never said that Zen 4 would offer a 15% IPC improvement compared to Zen 3, we indicated that we will see a single fiber performance improvement of up to 15%. If we analyze this sentence carefully, we will realize important things:

• He’s talking about overall improvement of 15%, i.e., including an increase in power derived from the IPC and an increase in power that results from an increase in operating frequency.
• Yippee maximum value at bestand was obtained by comparison with Ryzen 9 5950X in Cinebench R23, indicating that the chip used by AMD was the Ryzen 9 75950X we have already mentioned.
• If we look at the fine print associated with this phrase, we will see that AMD has also used it in comparison 6GHz DDR5 memory with very tight latency that accompanies the Ryzen 9 7950X.

Click to enlarge.

Put it all in context and keep it in mind working frequency difference between Ryzen 9 5950X and Ryzen 9 7950X in one fiber would be great (4.93 GHz first and 5.4 GHz second), I think it’s quite clear that much of the increase in single-wire power would come from this increase in frequencies, and we must not forget the effect of RAM. I know that in Cinebench R23 it doesn’t affect much, but even in the worst case it can mean a slight increase and in the end it all adds up.

In view of all of the above, I don’t think AMD wanted to specify the exact IPC enhancement that Zen 4 refers to compared to Zen 3. because it is much smaller than we expected. Obviously we don’t have data, but I’m afraid it’s probably somewhere between 3% and 8%.

Will Intel be able to outperform this increase in performance?

If we compare with Alder Lake-S, then yes. To continue the comparison made by AMD itself, the Intel Core i9-12900K has single-threaded performance 1 930 points in Cinebench R23while the Ryzen 9 5950X achieves 1 684 points. A 15% improvement would mean 252 extra points, which would leave us with a total of 1,936 points. The difference is minimal, but sufficient for the Ryzen 9 7950X to be recognized as the winner.

However, the difference would be much greater in multithreading because the Ryzen 9 7950X does not have a high-performance and highly efficient core distribution. AMD said it was able to achieve improved performance up to 31% in Blender against the Intel Core i9-12900K, but what happens when the Core Gen13 processors, the Raptor Lake-S series, hit the market?

With everything I’ve seen so far, I am quite convinced that the Core Gen13 will offer excellent single-thread performance over the Ryzen 7000, thanks to the IPC enhancements that the Raptor Cove architecture will bring to high-performance cores and increased frequencies, and I doubt multi-threaded performance, especially since the latest Intel Core i9 -13900K will have 8 high-performance cores, 16 high-efficiency cores and 32 threads. This means that doubles the number of high-efficiency cores compared to Intel Core i9-12900K.

Concluding remarks: We have an epic struggle ahead of us

Both AMD and Intel have had one of the most intense and important battles in recent years in the high-performance consumer CPU sector. AMD hit the Zen 3 hard, but Intel bounced back with Alder Lake-S and now it seems that the giant Santa Clara could defeat Sunnyvale with Raptor Lake-Snot only for raw performance, but also for compatibility and support, as this generation of processors will work with the LGA1700 socket and will support DDR4 and DDR5 memories.

AMD could have an important asset in its favor, integrate RDNA2 GPU into your Ryzen 7000something he has already done with his APU Ryzen 6000, and may also choose to fine-tune prices much better than it was then with the Ryzen 5000, a generation that came on the market at too high a price (compared to the Ryzen 3000).

In any case, and whoever wins, the important thing is that there will be brutal competition between the two companies, and that in the end we will be great beneficiaries, consumers. It’s great to have Intel and AMD so strong and going from toe to heel.