AMD Ryzen 3000 12-Core Zen 2 CPU Benchmark Leak Shows Modest IPC Gains

Are you tired of all the Ryzen 3000 leaks yet? We hope not, because there is another one in the lead up to Computex, where AMD is expected to unveil its next-generation processors based on its 7-nanometer Zen 2 CPU architecture. This latest leak involves a 12-core/24-thread model, which if priced right, could be the most compelling of the bunch.

Time will tell, but for now, we have a glimpse of the chip's performance. The chip is an engineering sample from at least a few months ago. For anyone who is not familiar with engineering samples, these are pre-release parts that companies like AMD and Intel send out to OEM partners and reviewers for testing and validation.

Specs on ES chips can be different from the final slice of silicon. That said, this particular part shows a 3.4GHz base clock and a 3.7GHz Turbo clock. Those are not mind boggling frequencies by any stretch of the imagination. However, the finalized CPU could very well be clocked higher, and potentially a lot higher than the ES chip.
Frequent leaker and Twitter user APISAK brought the benchmark to attention. He spotted the chip over at Userbenchmark, where it was running with a new motherboard labeled as Qogir-MTS. It's not clear what exactly that is (X570 chipset, possibly). The listing also shows 16GB of DDR4-1600MHz memory, a Radeon RX 550 graphics card, and a 500GB hard drive.

Ryzen Benchmark
Click to Enlarge

The above benchmark results are not very meaningful in a vacuum, though thankfully they do not exist that way. We can compare the supposed Zen 2 processor with a second-generation Threadripper 2920X CPU. The 2920X is also a 12-core/24-thread processor, but clocked faster at 3.5GHz to 4.3GHz. Let's have a look...

Ryzen and Threadripper Benchmarks
Click to Enlarge

Quite frankly, we would hope to see a bigger disparity. Assuming faster clocks on the finalized 12-core Zen 2 chip, it should skip ahead of the second-gen Threadripper part, though not by a whole lot. If only looking at this one specific benchmark run, the IPC gains are rather modest.

We don't want to jump to that conclusion, though. These are different classes of chips, obviously, but also indicative of a single benchmark run on a pre-release part, using probably unoptomized drivers. It's entirely possible that the combination of faster release clocks combined with a more mature platform with updated drivers could tell a different story, particularly when venturing into other performance metrics.

Or the IPC gains could end up being smaller than hoped. That is certainly a possibility as well, though we'd be surprised if they ended up being minuscule. The good news is, we only have a few more weeks until we know for sure.