Acceleration occurred in basically all areas of the processor’s work. Part concerns the loading of instructions and data, branch prediction, the decoding of instructions has been improved even more, and the largest part of the acceleration is caused by the execution of instructions, i.e. the calculations themselves. A higher transfer speed between memories also contributed.
Higher accuracy and lower latencies were achieved for branch prediction, instruction cache latencies were also improved and memory cache throughput increased (both between L2 and L1 and between L1 and FP units, where the speed was doubled). Each core has 6 ALU units, the ALU scheduler has also been improved.
The L1 instruction cache is 8-way and has 32 KB, the L1 data cache is 12-way and its capacity is 48 KB. As for the L2 cache, it is 16-way and its capacity is 1 MB per core. We have AVX-512 instruction set support here with full 512-bit data (not 2×256 bit). It can also handle multiple calculations with decimal numbers at once.
Here we see a comparison of the 6-core (12-thread) AMD Ryzen 5 9600X processor with the 14-core (20-thread) Intel Core i5-14600K. Recall that Ryzen has a 65W TDP (88W PPT), while Intel is at 125W PBP and 181W MTP. In selected applications (and the question is how much “cherry picking” was done here, when all the results are positive) AMD shows a considerable lead, in Handbrake it should even be almost 2 times faster. We also have considerable differences in games (the six presented show an average of +14%).
This should have happened due to the fact that the Ryzen 9 9700X was not supposed to have the gaming performance of the Ryzen 7 7800X3D gaming processor. It may still not have that, as AMD looked at the comparison with the even older generation Ryzen 7 5800X3D in the presentation. And there they already claim that the novelty is 12% faster in games on average, despite the fact that it has a TDP of 65W, while the 5800X3D is with 105W.
Also, the 12-core Ryzen 9 9900X (with 24 threads) should outperform (almost) the top-of-the-line Intel Core i9-14900K. Here we’re talking 120W TDP and 162W PPT for AMD, while for Intel it’s 125W PBP (that’s about the same) but a much higher 253W MTP. Intel has 24 cores and 32 threads support.
When we look at the 16-core (32-thread) Ryzen 9 9950X, the difference is understandably even greater in favor of AMD. But here it is also necessary to say that the 9950X already reaches 170W TDP, so we are talking about 230W MTP, which is not so far from Intel’s 253W consumption. The good news is that with the exception of the 9950X, all the new Ryzens went down in power consumption (from 105 W to 65 W or from 170 W to 120 W). What we still don’t know are the prices of the new products.
Source: www.svethardware.cz
