And it should be treated with caution. The first benchmark results for AMD's new flagship graphics card are circulating online a week before the Radeon RX 7900 XTX and Radeon RX 7900 XT go on sale.
If I remember correctly, the AMD/nVidia performance difference is rarely the same between synthetics and actual rasterization performance. So this would just be an indication of process change, not an actual gaming performance comparison
Even the cookie baker LU's 3DMark only barely reflects real game performance, let alone Geekbench. Maybe Geekbench is more interesting if you use the graphics card for non-gaming computation workloads?
Well, IF those numbers are true, it'd be a worthy upgrade for my RX 6900 XT, which would go into my 2nd all AMD gaming rig! Just realized, one more week to go to the official release of the RX 7900 series cards, I can hardly wait!
I have a feeling even a 7800xt would be a good upgrade over my 6800xt. But i may go all the way and get a 7900xt.
When it comes to CPU benchmarks, the only synthetic tests I take seriously are those that have realistic or understandable workloads, so Cinebench is fine because it is testing something you might actually do, whereas the CPU-Z benchmark is utterly useless since it doesn't really seem to measure anything meaningful. Prime95 is also a crap benchmark, but it is a great stress test. When it comes to GPUs, there is so much focus on driver optimizations that it's practically pointless to take any synthetic test seriously. Even though benches like 3DMark do a good job at testing a wide range of a GPU's capabilities, the results mean nothing if the games/applications you want to run don't have the same level of optimization (whether they perform better or worse). If there were synthetic benchmarks for GPUs that explicitly prohibited driver optimizations, that would be a little interesting since it could show the true potential of the GPU, but it still shouldn't be taken seriously as a metric to buy the GPU. It really doesn't matter how well a product ranks in a synthetic test if that's the best-case scenario. One of the reasons I like reviews on phoronix.com is because they do a geometric mean, so you account for the result of all tests and see what is the best overall. I wish more reviewers would do this.
I like this approach. The only exception I would make is that if you use your PC for one game or one thing like photoshop most of the time then then focus only on those reviews because that matters most.
