When we have reviewed the RX470 Strix, we were faced with a video card faster than the R9 380X of about 25%. On the paper, the main difference of these two solutions is the GPU frequency. Thanks to the new 14nm FinFET node, Polaris 10 can be clocked much higher than Tonga. Not to mention the updated L2 cache (2MB vs 512KB). Not only … but also we should allude to some little differences to uArch level (GCN4 vs GCN3) and a little improvement to DCC features (Delta Color Compression). For the rest, the two video cards are almost the same: 2048 SPs, 128 TMUs, 32 ROPs and a 4x MC @ 64-bit.

So, because these two uArch are very similar, in our opinion it would be interesting to do a direct comparison between Polaris 10 Pro and Tonga XT. This is particularly so given that AMD has omitted a direct comparison in its own slides between these two GPUs. In contrast, AMD has pointed out the differences between Hawaii (GCN2) and Polaris (GCN4): Hawaii doesn’t have the updated front-end of Tonga (HWS and Tessallation Fix) and doesn’t have the DCC features.

Full Tonga (XTX) has a 384 Bit IMC and 768KB of L2 Cache

In today article, we will try to do a fair comparison between Polaris 10 Pro and Tonga XT. These two video cards (RX470 and R9 380X) have been set to the same frequencies (GPU @ 1000 MHz, MEM @ 6000 MHz). Also, in order to avoid the “intrusion” of the AVSF and DVFS features, we have set the PT value and RPM fans to the highest levels.

Our goal is to measure the real gain of Polaris 10 over Tonga thanks to these new features (Bigger L2 Cache in primis), and at the same time we want to see if there are improvements in Tessellation and DX12 benchmarks.

Autore: Michele Carasia
Esperto in: Schede video, CPU, overclock, videogames

E' stato newser e reviewer per importanti portali italiani occupandosi in particolare di schede video e processori. Appassionato anche di overclock e videogames.

1955 news	68 articoli