Friday, January 3rd 2025
PowerColor Teases Radeon RX 9070 XT "Red Devil" GPU
PowerColor has prepared its highest-end Radeon RX 9070 XT GPU in the "Red Devil" family. In a post on the social media platform X, PowerColor shared the below picture with the following words: "Every edge shines like a gem. Every second burns like fire. If power was in your hands, how would you use it?" The picture shows an interesting design with a black shroud holding a red acrylic core, which follows the design philosophy of Red Devil cards. On the backplate, PowerColor has embedded a Red Devil sign with a meshed plate on top of it to show the card branding. Suppose the previous generations of Red Devil cards are for reference. In that case, we assume that the Radeon RX 9070 XT Red Devil will deliver top-tier specifications and component selection for overclocking, meaning that this card will be one of the highest-end RDNA 4 designs that we will see. Of course, surprises are welcome, and we must wait for the CES launch to see more.
Sources:
PowerColor #1, PowerColor #2
121 Comments on PowerColor Teases Radeon RX 9070 XT "Red Devil" GPU
Hopefully we get a 10" 2 fan "fighter" 9070XT for the more compact and subdued setups.
5080 will be a 256bit 84SM card against a 384bit 128SM card and will have less of everything - including memory bandwidth (despite using the fastest initial G7).
Anyone who thinks 5080 will tie 4090 performance is delusional. It will be +10-15% vs 4080S and that's about it. The gap between 4080S and 4090 is 24%.
Not to mention why would Nvidia give away 4090 performance for cheaper when there's not competition to 4090 from others? Because then there would be zero incentive to buy it. I can buy 7900XT today for 700 with 4GB more VRAM and likely more performance (at least in raster) since i doubt 9070XT will even tie 7900XT. Same here. I would be coming from 2080Ti so even 7900 GRE perf (like leaks have suggested) would be a 51% performance bump. 79% in case of 7900XT.
Plus 5GB extra VRAM, lower power consumption, AV1 support, PCIe 5.0 interface (tho im on a 4.0 system and it wont matter anyway) and modern ports with DP 2.1 (hopefully UHBR20 this time instead of 13.5). Not to mention native ReBAR support. Something Nvidia refused to add to their 20 series cards despite AMD doing it for their 5000 series cards retroactively. And yet all the comments talk about great Nvidia driver support. And only now has Nvidia managed to release a modern, unified GUI with no login requirement.
Too late.
As Nvidia user im not feeling it. They refused to give me ReBAR that would have given me extra perf, they refused to give me DLSS FG that AMD proved with FSR 3.0 can work on 20 series very well (especially in combination with Reflex). Nvidia's own native implementation would be even better.
Plus they implemented a bad connector in their cards that i refuse to trust (well maybe, after 10x revisions).
The prices are just a final insult on top of everything else.
That is why im going AMD for my next card. That will never happen. 256bit 64CU card vs 384bit 96CU. Not to mention the gulf in memory bandwidth as 9070XT wont be using G7 to help it bridge the gap like 5080 is doing with 4090.
People please. Stop with these wishful comparisons. You're just setting yourselves up for disappointment.
TPU's GPU database is available on this very site. Look at the specs and make realistic comparisons.
Yes of course clock speeds will increase a little, TDP will rise and likely there are SM/CU IPC improvements but realistically within the same 4nm/5nm process dont expect miracles. And if im wrong i will gladly eat my words. But i doubt im wrong.
Seeing the 6000/7000 (and 9000?) series cards, yeah no I'm done. Can't stay Red Devil. Either they fit or they don't and these don't fit anymore.
The coolers are overbuilt and that's great on single card systems but I run more than just a GPU and there are fitment issues on bigger cards.
My case is a 20 year old antique and it's NOT the problem. If I ever switch up to a vertical mount, mATX boards will follow up in transformation.
Imagine a timeline where peripheral loadouts are M.2 -> X16 -> X8 -> M.2 and there's like NO space between them. How does it work?
The only way is having a vertically mounted GPU over riser cables then slotting the rest. That's gonna be the main change. That's our future.
There are already mini cases that reflect that idea. Now we just need the boards to appear.
gpu.userbenchmark.com/Compare/Nvidia-RTX-4070-TS-Ti-Super-vs-AMD-RX-7900-XT/4155vs4141
gpu.userbenchmark.com/Compare/Nvidia-RTX-4080-S-Super-vs-AMD-RX-7900-XTX/4156vs4142
The AMD 9070XT 330w is slower and using more TDP Power Watts than the two years old RTX 4080 320w but at a lower price point.
Glad I bought my MSI RTX 4080 Super 16G SUPRIM X... come and get me Red Devil
Cheers