Some Interesting Benchmarks For AMD’s 3700XT And 3900XT Have Appeared

1
Some Interesting Benchmarks For AMD’s 3700XT And 3900XT Have Appeared
Image: Alex Walker (Kotaku)

Intel’s just launched the fastest gaming CPU on the market. But before most gamers could get the parts together to build a new system, AMD’s answer is already on the horizon.

The benchmarks appeared on the CPU-Monkey database overnight, a site that largely just compares the performances of every CPU manageable in synthetic workloads. There’s no gaming benchmarks and no real-world comparisons in programs like Premiere Pro, 7Zip or other kinds of regular applications, but mostly comparisons with the CineBench R15 and R20 rendering tests.

Cinebench is a reliable application for comparisons - I use it in our CPU reviews, as do most others - and it's a useful guide for those working in 3D modelling, rendering applications and motion graphics.

Cinebench is split into two tests, single-core rendering and multi-core rendering. It scales depending on the amount of cores you have, so the server-centric Xeon and EPYC processors will always be the king of the hill. Intel's recent i9-10900K, for instance, doesn't surpass the 3900X in Cinebench's multi-core tests, because it has two cores/four threads less than the AMD chip. It does beat it out in single-core performance, but that's because the 10900K can push a single core to 5.3GHz.

So what's interesting about the new database entries is that it shows the Ryzen 9 3900XT sitting just ahead of the 10900K and 10900KF. The benchmarks also have the 3900X running at a max frequency of 4.8GHz, 200Mhz than what the stock 3900X is capable of.

Image: CPU-Monkey

The multi-core results aren't hugely surprising: the 3900X was already better than the 10900K there, because it has more cores. But it does at least show a bit of a gap between the (rumoured) 3900XT and 3900X.

Image: CPU-Monkey

Cinebench results don't have a direct correlation with games, but in general, a 200Mhz boost - or whatever AMD ends up accomplishing - would result in a small, single-digit percentage uptick in games as well. It's worth noting a ton of other factors go into how well a game runs, though. Lots of games still favour Intel hardware, like Far Cry 5 and Shadow of the Tomb Raider, and a small frequency change on AMD's end isn't going to change that.

A bigger question around the Ryzen CPU refresh will be how well AMD can optimise the latency on their Infinity Fabric. The way AMD has maximised core counts on their CPUs to begin with was through the creation of multiple core blocks, which communicate with each other via AMD's Infinity Fabric interconnect. This has been great for helping get more cores onto a chip than was previously possible, but it also introduces extra latency as data moves between the core blocks. Eliminating this latency is one reason why AMD's recently-released 3300X is so much better in games than the Ryzen 3 3100. The 3300X completely disables one of the core blocks and just runs with 4 cores/8 threads on a single block, while the 3100 has four cores split across two blocks.

All that said, the performance of the new Ryzen chips probably won't mean a lot for anyone who just bought into the 3000 series, or the 10th gen Intel chips. They're still great CPUs, but there's only so much performance you can get out of the current generation of GPUs. New cards are coming in a few months though, and when we get our hands on one, you can bet we'll do some testing.

Comments

  • pretty sceptical of these numbers. amd has come a long way recently but theres very little chance they have increased their IPC so much that they can score better in a single thread test at 4.8ghz compared to the intel at 5.3ghz. either the scores are false, or the reported frequencies are false.
    im sure the new chips will be quite capable regardless based on the last few gen ryzen improvements, good to see amd finally becoming a contender in the gaming space for the first time since 2004.

Log in to comment on this story!