-
MSI’s version of the AMD RX 6600XT, the one Ars Technica tested for this review.
MSI/AMD
-
It comes in a relatively nondescript box, unlike AMD’s more expensive RDNA 2 GPU boxes.
MSI/AMD
-
What the GPU looks like on my kitchen floor.
Sam Machkovech
-
You can’t see from this angle, but only one 8-pin connector is needed for power, and there’s a 160W TPD.
MSI/AMD
-
Here’s a better angle of that 8-pin connector.
Sam Machkovech
In a traditional PC hardware cycle, AMD’s new RX 6600XT could have been a welcome stopgap for a budget crowd. Over the years, we’ve seen these types of GPUs regularly from both major GPU manufacturers. Those companies regularly reject some specs, reuse sub-optimal chips, and get a reasonably priced option to follow their biggest kahunas for anyone tiptoeing towards solid 1080p or 1440p gaming options on PC.
Sadly, there’s nothing traditional about the latest traditional PC hardware cycle. The current ecosystem of supply and demand for computer GPUs looks like something out of a terrifying Dario Argento movie. The horrors lurking in every shadow include chip shortages and bot-powered scalping surges.
And that context really helps us frame the $379 RX 6600XT – an underpowered, overpriced, and downright disappointing GPU whose primary sales pitch is 1080p gaming. That category is famously CPU limited, not GPU limited, so the mileage of this GPU will really vary depending on your rig. Overall, Nvidia’s RTX 3060 Ti (for just $20 more MSRP) wins handily, while Nvidia’s RTX 2060 Super (which launched in July 2019 for $399) is within screaming distance of this brand new card. That last metric in particular makes AMD’s latest product a hard GPU to recommend.
Have there been changes to the cabin pressure yet?
This is perhaps the loudest example yet of GPU manufacturers capitalizing on the industry’s recent price inflation, rather than letting the market do it. A $379 price tag isn’t a high price for the 6600XT, but it’s not clear that a more appropriate MSRP, such as $299 or $319, would lead to a fairer price reaching the average customer. “Street value” really exists. For example, when calculating actual market prices, we’re closer to the $500 range for both Nvidia GPUs mentioned above.
As such, I have to smother this piece again in reminiscences about general GPU availability. And again, AMD (like its competitor Nvidia) is not disclosing how many of its new cards will go on sale later this week. The best data point we have is how far behind AMD is behind Nvidia in terms of GPU installed metrics in the wild. Steam’s publicly available hardware surveys indicate that Nvidia’s RTX 3000 cards are getting into far more Steam-connected PCs than AMD’s RDNA 2 cards. That could be because Nvidia sold more cards, or at least sold more cards to active Steam gamers. Anyway, I have no clear indication that the 6600XT will drastically change the availability story.
(Editor’s Note: Na saying so many similar things in so many recent GPU reviews makes me feel like a flight attendant reading notes before a flight. “Should the GPU market stabilize in the near future, look to the exit doors at the front, sides, and back of the PC marketplace.”)
A note on the 6600XT’s 1080p emphasis
-
Specifications of the reference model RX 6600XT, as provided by AMD. Click ahead to see how much it’s downgraded from the RX 6700XT.
amd
-
RX 6700XT specs, compared to the 2019 RX 5700XT.
amd
When it comes down to it, the RX 6600 is a peculiar GPU option for one main reason: its promise of solid PC performance at 1080p resolution, as opposed to 1440p or higher. AMD swears that this difference is confirmed in its own handpicked test results, promising statistically significant 1080p gains over Nvidia’s RTX 3060 (MSRP: $329). The charts include tests done on a Ryzen 5 5600X system, with the RX 6600 beating the cheaper Nvidia card with games set to 1080p resolution and identical “maximum” visual settings from 5 percent to 30 percent.
However, Ars prefers 4K resolutions for GPU testing, as anything lower can contribute to more CPU-bound constraints on the resulting frame-rate graphs. Our measurements are designed to provide you with percentages of difference that you can then scale to your own PC. When AMD’s RDNA 2 line started promising significantly better performance at 1440p than Nvidia’s competition – claiming this was an intentional technical direction on their part – we ran the resulting tests. Sure enough, we saw significant percentage gains that weren’t necessarily CPU-bound.
But if we drop all the way back to 1080p with a test setup maxing out at an Intel i7-8700K GPU, we can’t reliably duplicate AMD’s own 1080p performance-boosting sales pitch. Our percentages in those tests vary with both higher and lower results based on the game, and these games stretch to 120fps and above at that point, which is usually the threshold at which a CPU starts to surpass. Our 1080p charts are therefore, in our opinion, misleading enough for average customers to disregard our usual comparison matrix.
I’m not printing AMD’s exact numbers because we can’t verify them, and because their own 1080p tests haven’t also been run on the $399 RTX 3060 Ti, nor on the 2019 RTX 2060 Super. Now that I have my own negative review of having already written the RTX 3060, I understand why AMD might be tempted to let the press directly compare the RX 6800 to that card. However, that card’s MSRP is about 15 percent cheaper – and AMD’s list of 1080p results for that card is, you know, about 15 percent worse.