AMD has done it again! Their new Ryzen 5000 Series desktop processors bring yet another huge generation-over-generation improvement. Many years ago, the naysayers had talked about AMD just grabbing the low-hanging fruit and that these gains would soon be history—guess not. The new Ryzen 5000 processors have been rearchitected from the ground up with impressive results.

We have a total of three Zen 3 reviews for you today: Ryzen 9 5900X, Ryzen 7 5800X, Ryzen 5 5600X. Our fourth review, of the Ryzen 9 5950X, is half-finished and will be posted soon.

Just like last-generation's Ryzen 7 3800X, the Ryzen 7 5800X is a 8-core/16-thread processor. While the predecessor had a boost clock range of 3.90 to 4.50 GHz, AMD has upped this to 3.80 to 4.70 GHz, which alone brings with it a few percent in performance. AMD was extra careful to ensure compatibility with existing platforms and cooling, so the TDP has remained at 105 W. Because of the architectural changes, we see more than impressive performance in applications. On average, in our mix of single, low and multi-threaded applications, the Ryzen 7 5800X is 19% faster than the Ryzen 7 3800XT. It's also a tiny bit faster than Intel's flagship, the Core i9-10900K. Against the Core i7-10700K, the performance uplift is 13%. AMD really killed most of Intel's high-end with the 5800X, especially at their current pricing. For multi-threaded application use, there's really nothing but the 5800X, 5900X, and 5950X—no Intel competitor in sight. Actually, Ryzen 7 5800X might be the better choice for some scenarios over the Ryzen 9 5900X because the 5800X crams all its eight cores into a single CCD, whereas the 5900X uses a 6+6 dual CCD design. Communication between CCDs has to pass through the IO die and runs at higher latency, which costs a little bit of performance. Looking at our benchmarks, we see this effect in many single or low-threaded applications. The problem seems to be that the Windows Thread Scheduler sometimes moves the application onto another CCD, which means the process has to pack up everything on its current CCD, move to the other CCD, and spin up the workload again. This looks rather like a Microsoft issue than an AMD problem, though.

In their presentations and launch event, AMD has been making noise that they won gaming leadership from Intel. My results are not as convincing, but kinda close. In our mix of 10 games, the Ryzen 7 5800X is around 2% behind the Core i9-10900K across all resolutions, a tiny bit faster than the Ryzen 9 5900X. The advantage over the 5900X certainly comes from the single-CCD design, but it's not big enough to lose any sleep over it. Against Intel, there are more pronounced differences in specific games. Especially Far Cry and Assassin's Creed have always been challenging for Ryzen. Other titles are almost a tie, and in games like Wolfenstein and Sekiro, AMD has indeed taken the performance crown.

I'm not sure if it's the selection of tests or the fact that we're running memory at 3200 MHz and AMD's numbers are at 3600 MHz, the silicon lottery, or something else. I even updated to Windows 20H2 for the scheduler improvements, as the Windows thread scheduler in version 1903 will definitely put loads on the wrong Zen 3 cores (I tested it). The performance loss without the update is a few percent, so make sure to update your OS when upgrading to Zen 3. I am sure AMD's legal team double-checked that their claims are waterproof, so I'll be digging deeper into what's going on, and will keep you updated.

Even with my current results, the Intel gaming advantage is effectively gone, a percent here or there really isn't worth worrying about, especially when you consider the application performance and platform improvements. While Intel wants you to buy a new motherboard every time a new processor generation is released, AMD has given us a solid upgrade path with AM4. If you own an AMD AM4 motherboard with the 400 or 500-series chipset, upgrading to Zen 3 is really just a "BIOS update, remove heatsink, plop in new CPU, install heatsink, done". As mentioned before, AMD made sure to not increase power requirements, so your cooler, power supply and everything else will continue to be work just fine.

Overclocking on the Ryzen 7 5800X worked similar to other Ryzen CPUs we tested today. My maximum all-core overclock was 4.6 GHz. 4.65 GHz was stable too, but with too much voltage. 4.7 GHz wasn't stable at all, no matter the voltage. These numbers are significantly lower than the maximum single-thread boost frequencies we observed, which means manual overclocking is not worth it unless you specifically want to increase multi-threaded performance at the cost of lower single-threaded performance, and of course higher power and heat. It's still nice to see an unlocked multiplier on Ryzen processors. Intel charges you extra for this.

AMD not only engineered a new, faster architecture, they also improved power efficiency without a node shrink. The Zen 3 processors are produced on the same 7 nm process at TSMC, just like their Zen 2 brothers; the IO die is even completely identical. AMD still achieved significant power savings. We measured about 10 W lower power draw in single-threaded and 10 W higher power draw in multi-threaded work loads, with much higher performance in both scenarios. When taking into account how fast this processor is, it conclusively beats everything Intel has to offer.

While many owners of Zen 2 processors were unhappy with their CPUs not boosting high enough, this is definitely not an issue on Zen 3. AMD has been more conservative with their "up to" boost clocks. The Ryzen 5800X is rated for 4.70 GHz, but we regularly observed boosts to 4.95 GHz that were sustained for several seconds, 4.85 GHz was the 1T average over a longer time span. Looking at our frequency analysis, we can see an almost gradual drop off in clocks as the thread count is increased—excellent. When fully loaded on all 16 threads, the 5800X still ran around 4.60 GHz, just 100 MHz lower than its maximum boost. This is also important for overclocking—our maximum all-core OC was 4.6 GHz, out of the box, the processor boosts all-thread to slightly above 4.6 automatically, so not much point in OC.

We have mentioned the lack of an integrated GPU in all Ryzen reviews and received a lot of criticism for just mentioning that fact. It's still true, and not a big deal at all, but Intel offers an iGPU that is sufficient for basic tasks and minimal gaming. Yes, I know that AMD has their APU line of products, but given the modular design of the Ryzen processors, I wonder how expensive it could be to design a very basic GPU die connected via Infinity Fabric to strengthen the Ryzen position in non-gaming market segments, where every dollar counts.

AMD is asking $450 for their Ryzen 7 5800X, which is $100 less than the 12-core 5900X. Right now, the Ryzen 7 3800X retails for $340 (with a heatsink), and the 3800XT is $380 (without a heatsink). This makes the Ryzen 7 5800X kind of expensive for what it offers even with the incredible performance gains are taken into account. If you are willing to live with slightly lower performance, the latter offer better performance per dollar. It might also be worth looking for a used 3900X that an upgrader wants to unload at reduced pricing.

In the Ryzen 5800X performance range, Intel isn't offering much that can hurt the 5800X, only the overpriced Core i9-10900K. If you can live with lower application performance, the Core i7-10700K at $380 could be an option. Other than the obvious cost savings, you might end up losing on Intel's higher platform cost, I'm not really convinced.

I would say the 5800X is around $50 too expensive for what I would consider fair, but supply of the new AMD Zen 3 processors will be fairly limited for the foreseeable future. So if things are anything like the NVIDIA Ampere launch, people won't care much about pricing and will pay almost anything to have one of these new CPUs.