Pay no attention to AMD's horribly misleading benchmarks for its 'new' Ryzen 5000 XT CPUs

Remember those "new" Ryzen 5000 chips AMD launched last week alongside its actually new Ryzen 9000 CPUs? Turns out the benchmarks AMD provided for them are horribly misleading.

YouTube channel Hardware Unboxed spotted that something was a bit rotten in the Denmark of the Ryzen 7 5800XT and Ryzen 9 5900XT gaming numbers AMD put out to promote the chips.

AMD benchmarks showed the 5800XT beating Intel's Core i5-13600K across a number of games, while the 5900XT was shown having the better of the Core i7-13700K.

The problem with those claims is multifold, and it starts like this. The "new" AMD CPUs are minor tweaks of existing chips based on the Zen 3 architecture. The 5800XT is an eight-core CPU with specs very close to the existing Ryzen 7 5700X, while the 5900XT is a 16-core alternative that's specced to within a whisker of the Ryzen 9 5950X.

The point is that both of the "new" chips are largely known quantities, even without detailed third-party testing. But here's the thing. Hardware Unboxed's own testing shows that the Intel 13700K was fully 36% faster than the AMD 5950X across and average of 12 games, while the Intel 13600K was 28% quicker than the 5700X.

Such is the gap in age between the 5800XT and the 13600K, we don't have our own comparison numbers. But we do have the Intel 13600K versus the newer AMD 7700X. And it's the Intel chip that wins in games.

Now, the 5800XT does have a 200MHz higher maximum boost clock than the old 5700X. But it's otherwise identical and that's a mere 4% increase in peak theoretical operating frequency. The chances that translates into not just closing a 28% performance deficit, but actually coming out on top are awfully, awfully slim.

So what, exactly, is going on here? Is AMD cherry picking games that happen to run particularly well on its CPUs? Actually, it's worse than that. All of the benchmarks were carried out using an AMD Radeon RX 6600 GPU. Yep, an RX 6600 graphics card was used for all the testing across all four CPUs.

That, of course, means that the bottleneck for performance in AMD's test setup will largely have been the GPU. And AMD's CPUs win because they play that tiny bit more nicely with the RX6600 than the Intel chips.

So, yes, strictly speaking if you have a Radeon RX 6600 GPU you may well find that these AMD CPUs give very slightly better results than the Intel competition. But if you're using a newer and faster GPU, you'll almost certainly find the opposite. Because the Intel CPUs are generationally newer and fundamentally quicker at running games.

At best, then, AMD's benchmark numbers for the 5800XT and 5900XT CPUs are thoroughly misleading. If a PC gamer bought either of those CPUs based on the benchmark numbers AMD provided—expecting to get better gaming performance than the Intel CPUs in AMD's comparisons—in the vast majority of likely scenarios and system configurations, they would be disappointed.

More to the point, anyone remotely well informed would understand this. So, whoever signed off the benchmark numbers either doesn't understand PC technology and gaming or was intentionally aiming to mislead. There really aren't other options.

Hardware Unboxed called it "AMD lies". It's hard to say that AMD was lying with absolute certainty. But whatever happened, it's certainly not AMD's finest hour.

That's a pity because the 5800XT and 5900XT are not bad CPUs if you don't oversell them. It's good to have a little life injected into the old AM4 socket. But the chips are what they are, namely based on the Zen 3 CPU architecture, now 3.5 years and two generations old. Trying to fancy them up into something they're not is just awfully poor form.