Unlike previous RX 6000 series launches, the Radeon RX 6600 XT will only be available from AMD's add-in board (AIB) partners. AMD sent us the ASRock RX 6600 XT Phantom Gaming D OC for this review, which of course doesn't use reference clocks. It sports a 2.4-slot cooler with triple fans and a slightly boosted TDP (based on our testing), which means it's probably running about 5% faster than a reference-clocked RX 6600 XT.
That's an important distinction because nearly all of the other GPUs we'll use in our testing are reference AMD and Nvidia cards that aren't factory overclocked. Models that run 3–5% faster should be easy enough to find — well, as easy as it is to find any GPU in stock right now. The only exception to that is the RTX 3060, which also doesn't have a reference model. Instead, we used an EVGA RTX 3060 XC, which comes factory overclocked, though the cooling solution is admittedly far less robust on the EVGA model (a compact dual-fan solution).
It's also worth noting that the official price for the ASRock Phantom Gaming comes in at a suggested price of $499.99, over 30% higher than AMD's SEP (Suggested Etail Pricing) of $379.99. ASRock does offer other RX 6600 XT cards, like the Challenger, with a $379.99 MSRP, but without all the extras that the Phantom offers. Whether the cards will actually sell at that price, in reasonable quantities, remains to be seen.
ASRock's card uses three traditional-style axial fans, meaning there's no integrated rim like we see on higher-end GPUs. The heatsink features a copper base that makes direct contact with the GPU, plus "premium" thermal pads that make contact with the memory. The Phantom also has 0dB silent cooling for lower workloads, where the fans don't spin at all at GPU temperatures below ~50C. Given the 160W (maybe 180W) TDP, cooling should be more than sufficient — and it is, as we'll detail later in our power, temperature, and fan speed testing.
The card measures 306x131x47mm and weighs 898g (those are my measurements), a comparative lightweight when looking at cards like the RX 6800 XT. It has a single 8-pin connector for power, a metal backplate, and RGB lighting on both the Phantom Gaming logo on the top of the card and the middle clear fan. Connectivity consists of the now typical single HDMI 2.1 port with three DisplayPort 1.4 (with DSC) outputs.
Test Setup for Radeon RX 6600 XT
Intel Core i9-9900K (opens in new tab)
MSI MEG Z390 Ace (opens in new tab)
Corsair 2x16GB DDR4-3200 CL16 (opens in new tab)
XPG SX8200 Pro 2TB (opens in new tab)
Seasonic Focus 850 Platinum (opens in new tab)
Corsair Hydro H150i Pro RGB (opens in new tab)
OpenBenchTable
Phanteks Enthoo Pro M
Ryzen 9 5900X (opens in new tab)
MSI X570 MEG Godlike (opens in new tab)
Thermaltake Toughpower GF1 1000W (opens in new tab)
Our basic test hardware remains unchanged from previous reviews, though with a few updates. First, we're now running the latest version of Windows 10 (21H1, build 19043.1151). We're also using motherboard BIOS version 7B12v1B1, which includes beta Resizable BAR support (aka, 'ReBAR' or Smart Access Memory). That's on the Intel Core i9-9900K system. At the bottom of our parts list, you'll also see the Ryzen 9 5900X, MSI X570 Godlike, and Thermaltake GF1. Those are all used as a second test PC, because we wanted something that had PCIe Gen4 support as well as an AMD CPU to see how much that changes the performance story (also with ReBAR / Smart Access Memory enabled).
Considering this is a more modest GPU, CPU bottlenecks aren't likely to be much of a problem, even on the relatively old Core i9-9900K. We previously looked at CPU scaling on the latest GPUs for the RTX 3060 Ti launch, with a focus on the top-performing solutions at the time (Ryzen 9 5900X, Core i9-10900K, and Core i9-9900K). While there were some differences, overall, the net gain from swapping to a different CPU was only 1–2 percent. Of course, the Core i9-11900K has now launched, but with Alder Lake and Zen 4 in the works, we'll hold off any further testbed upgrades (which would necessitate retesting everything) until a later date.
We're sticking with the same 13 games we've been using since the RTX 3080 launch, all with DXR (DirectX Raytracing) disabled. We have a second test suite that includes DXR in ten games for those curious about how the RX 6600 XT holds up with maxed-out graphics settings and ray tracing. We provided a more extensive look at ray tracing and DLSS performance recently, after the RTX 3060 launch, and found that AMD's RX 6700 XT matched the RTX 3060 (without DLSS running). We'll use those same tests and test results here.
One thing we're not testing for this review is FidelityFX Super Resolution performance. Actually, I did run a couple of tests (in Godfall and Terminator: Resistance), but the blessing and curse of AMD FSR is that it works with everything — at least on the hardware front — and the gains are generally pretty similar across the latest AMD and Nvidia GPUs. That means if a card is faster without FSR, it's generally faster with FSR as well. Plus, none of the games in our standard test suite currently support FSR.
MORE: Best Graphics Cards
MORE: GPU Benchmarks and Hierarchy
MORE: All Graphics Content