How do you upgrade the dead Gamescom? Right, with a hardware launch! AMD explicitly targets the “enthusiasts” among gamers with the Radeon RX 7800XT (499 USD) and RX 7700XT (449 USD) presented today and fills the remaining gap in the midfield of gaming graphics cards at the same time. Mit focuses on “1440p gaming”, i.e. WQHD, and also sees itself well positioned towards the competitor NVIDIA. And if all that wasn’t enough, AMD also lets the last remaining software unicorn, the long overdue FSR3, out of the bag. But one after the other…
But let’s start with Navi32 XT and XL. For the Radeon RX 7800XT and the RX 7700XT, AMD uses the previously unused Navi 32 chip in the XT full version as well as in the slightly trimmed XL version. The technical data can be found in the table below, which makes the comparison easier. It is a fact that AMD finally fills the gap here, especially since they have placed another gap filler between the high-end chip and the new mid-range with the GRE version. Instead of the GRE with 80 CUs, I would have rather expected a 7800XT with a slimmed-down Navi31 and 72 CUs, but this variant is still reserved for the Radeon Pro W7800.
The complete Navi32 has rather been given to the midrange and even leaves a gap to the considerably slower RX 7600. A Radeon 7700 without XT or a RX 7600XT would easily find room here, because the leap is still much too big. But we’ll just wait and see, after all, the price structure still has to consolidate to some extent. The already announced table actually shows the filled and the still open gap very clearly.
| Radeon RX 7900 XTX |
Radeon RX 7900 XT |
Radeon RX 7900 GRE |
Radeon RX 7800 XT |
Radeon RX 7700 XT |
Radeon RX 7600 |
|
|---|---|---|---|---|---|---|
| Architecture | RDNA 3 | RDNA 3 | RDNA 3 | RDNA 3 | RDNA 3 | RDNA 3 |
| Chip Name | Navi 31 XTX | Navi 31 XT | Navi 31 XL | Navi 32 XT | Navi 32 XL | Navi 33 XL |
| GPU Design | Multi-Chiplet | Multi-Chiplet | Multi-Chiplet | Multi-Chiplet | Multi-Chiplet | Monolith |
| Process Node | TSMC N5 TSMC N6 |
TSMC N5 TSMC N6 |
TSMC N5 TSMC N6 |
TSMC N5 TSMC N6 |
TSMC N5 TSMC N6 |
TSMC N6 |
| Compute Units | 96 | 84 | 80 | 60 | 54 | 32 |
| Shader Units | 6.144 | 5.376 | 5.120 | 3.840 | 3.456 | 2.048 |
| RT Accelerators |
96 | 84 | 80 | 60 | 54 | 32 |
| AI Accelerators | 192 | 158 | 160 | 120 | 108 | 64 |
| Memory | 24GB GDDR6 20 GT/s |
20GB GDDR6 20 GT/s |
16GB GDDR6 18 GT/s |
16GB GDDR6 19.5 GT/s |
12GB GDDR6 18 GT/s |
8GB GDDR6 18 GT/s |
| Interface | 384 bit | 320 bit | 256 bit | 256 bit | 192 bit | 128 bit |
| TBP | 355 W | 315 W | 260 W | 263 W | 245 W | 165 W |
| MSRP | 999 USD | 899 USD | 649 USD | 499 USD | 449 USD | 269 USD |
AMD sees the new cards in direct competition with NVIDIA’s GeForce RTX 4070 and RTX 4060 Ti 16 GB, which can of course also be found in the slides. Where this ends up in reality in the benchmarks, we will of course find out for ourselves later. Here is AMD’s view of things for now:
And now we come to FSR 3, a technical upscaling method with the generation of artificially calculated intermediate frames via AFMF (AMD Fluid Motion Frames), which we had actually expected since February 2023 after the first announcements. But does the saying “good things come to those who wait” also apply in this case? Of course, we are waiting for the real implementation with excitement and can only tease what AMD has wrapped up for us in this case as well. Our own tests will show later what has really been implemented. But the anticipation continues to rise (which was certainly intended by AMD). I’ve also compiled the relevant slides for you, because AMD hasn’t revealed everything yet.
However, the list of games is interesting in this context, which also includes NVIDIA’s favorite game Cyberpunk 2077. Quite by chance, NVIDIA had recently raised the DLSS version to 3.5 to take the lead in the battle of the nomenclatures again. AMD really promises a lot in the presentation and it would be nice if this then also arrives in gaming reality in this form. Competition is always good:
Those who think of “I’ll buy an E” when they hear Hypr RX haven’t yet caught on to the trend of stylish consonant accumulations through the intentional loss of vowels. If you really want to be hip, you just leave out the annoying vowels and hope that the rest will at least get it. The mixture of Boost, Ant-Lag and AMD Super Resolution is supposed to be the new latency cure. However, the previous design wasn’t quite convincing, so there was a lot of room for improvement here. We will of course test whether this can be converted into real progress in due time. We have the right equipment from the monitor tests.
Last but not least, we also want to mention the rest of the new ecosystem aimed at streamers, ranging from the AV1 encoder to any number of features and AMD’s Noise Suppression. This is software that AMD is now targeting NVIDIA Broadcast as a counterpart, which has been growing in popularity since Turing. Here I also use AMD’s slides, because real technical details are unfortunately still scarce:
Of course, you can also consume the live stream at Gamescom or read this summary. Or both, for that matter. Autumn will be interesting, I promise!
148 Antworten
Kommentar
Lade neue Kommentare
Urgestein
Mitglied
Veteran
Urgestein
Veteran
Veteran
Urgestein
Urgestein
Mitglied
Moderator
Urgestein
Urgestein
Urgestein
Urgestein
Urgestein
Urgestein
Mitglied
Veteran
Veteran
Alle Kommentare lesen unter igor´sLAB Community →