AMD has announced its new Ryzen 7000 series processors. (opens in new tab) Today, with news that the architecture is actually performing better than originally expected. % increased. (opens in new tab) It outperformed its Zen 3 ancestor and is now said to have improved IPC by about 13%.
With Zen 4, it became important to work on how to get instructions into the machine even faster.
Mark Papermaster, AMD
However, it’s worth breaking it down. Both to understand where that extra performance is coming from and how it really breaks down in terms of how the architecture works under different circumstances… because it’s a whole Because it’s not typically 13% higher.There are some pretty big differences.
First, the geomean numbers were derived from a performance test suite of 22 different workloads, all running at normalized 4GHz on 8-core Zen 3 and Zen 4 processors.
This allows us to take Zen 4’s clock speed boost completely out of the equation and see a direct architectural advantage on a basically comparable chip.
Mark Papermaster explains the performance improvements at today’s event:
“Zen 3 increased the execution width, so it became important for Zen 4 to work on how to push instructions into the machine even faster, so most of the improvements come from the front end and branch prediction. Really that accounts for almost 60% of that IPC gain.”
AMD has also increased the amount of cache on the chip, but not to the same extent as the 3D V-Cache on the Ryzen 7 5800X3D (opens in new tab)This increased the last level cache, but in Zen 4 AMD doubled the L2 cache instead.
This was done “to deliver critical data faster,” and “results in an overall IPC of 13%,” said Mark Papermaster.
However, it’s worth taking a closer look at these 22 different workloads. Some games actually show well over a 13% improvement, while some games like GTA V and Fortnite show much lower single-digit increases.
The single-threaded Cinebench R23 and CPU-Z numbers are also worth noting. They show increases of 9% and 1% respectively. If AMD wanted to talk about the 29% single-threaded performance boost separately, they’d highlight it with a full 5.7 GHz Ryzen 9 7950X running Geekbench.
Again, the 5nm production process is considered. When discussing the performance and efficiency benefits of using TSMC’s 5nm manufacturing process for the Zen 4 compute die, Papermaster explained that the most significant performance gains will appear at the lower end of the power consumption curve. increase.
The 7950X shows up to 35% better performance compared to the 5950X in Cinebench R23 multithreaded tests when running at 170W TDP. That’s a staggering 37% at 105W and 74% at 65W.
This could highlight what happens with lower TDP cores, especially in laptop configurations. However, the 13% IPC gain performed on an 8-core Zen 4 core running at 4GHz suggests that it doesn’t necessarily match up in the real world when it’s consuming more power at something like 170W. It is possible that you are. 7900X or 7950X.
Ultimately, the numbers AMD provided are still worth investigating as they provide a glimpse of what the new architecture can offer. Also, when Zen 4 cores are pushed to higher power consumption, they will run at very high clock speeds. Also Provides more performance.
So, when the chip becomes available later next month, we could see some attractive benchmark results.
