An unreleased APU (opens in new tab)from AMD has been spotted on Geekbench. The chip in question is the AMD Ryzen 3 3250U, with the U denoting that it is for mobile devices.
AMD has already launched the Ryzen 3 3200U, part of the chipmaker's Picasso family (opens in new tab). The 12nm APU comes wielding Zen+ CPU cores (opens in new tab)and Vega 3 graphics. Judging by the specifications listed on Geekbench, the 3250U is very similar to the 3200U.
AMD Ryzen 3 3250U vs. AMD Ryzen 3 3200U Specs
|Processor||Cores / Threads||Base / Boost (GHz)||L1 Cache (KB)||L2 Cache (MB)||L3 Cache (MB)||GPU||GPU Cores||GPU Clock (MHz)||TDP (W)|
|AMD Ryzen 5 3250U||2 / 4||2.60 / 3.47||192||1||4||Radeon Vega 3||3||1,200||?|
|AMD Ryzen 5 3200U||2 / 4||2.60 / 3.50||192||1||4||Radeon Vega 3||3||1,200||15|
*Specifications for the Ryzen 5 3250U are not confirmed.
The Ryzen 5 3250U reportedly uses the same dual-core, four-thread (opens in new tab)configuration and runs with a 2.6 GHz base clock speed (opens in new tab). The chip appears to have a 30 MHz lower boost clock, but that could be a detection error on Geekbench's part. Additionally, the AMD Family 23 Model 24 Stepping 1 processor ID confirms the Ryzen 3 3250's usage of the Picasso architecture.
Let's not forget that AMD also produces custom solutions for some of its partners. The Ryzen 3 3250U could either be a new SKU or a custom-made APU for a big-name PC vendor.
As for the cache (opens in new tab), the Ryzen 5 3250U is equipped with 192KB of L1 cache, 1MB of L2 cache and 4MB of L3 cache. There are no surprises there. The APU once again features Radeon Vega 3 graphics, meaning it has three Computes Units (CUs) clocked up to 1,200 MHz.
Geekbench 4 might not be the best tool for comparing two processors, but one listing (opens in new tab) comparing two PCs shows the Ryzen 3 3250U performing 3% and 1.3% faster than the Ryzen 3 3200U at single-and multi-core workloads, respectively. However, both systems were running with different memory and different versions of Geekbench 4, so the actual difference might be smaller or even non-existent.
For example, Steve Burke at GN ended up with much higher multithread score on a stock 9700k vs the 9900k - he noted the odd scoring in the benchmark and claimed he had re-run the test several times and the 9700k kept putting out higher multithreaded scores than the 9900k during his benchmarking.
People should stop using it as benchmark to compare different cpus - it just plain doesn't work well for that purpose for some reason.