AMD Radeon RX 6950 XT 3DMark Score Leaks: Faster Than GeForce RTX 3090 Ti
AMD's Radeon RX 6950 XT looks fast
AMD is set to release its Radeon RX 6000-series refresh later this month, which could rank among the best graphics cards when they arrive, but details about the new graphics cards have been leaking for quite a while. On Wednesday, our colleagues from WCCFTech published the first 3DMark Time Spy and Fire Strike scores of the new Radeon RX 6950 XT, 6750 XT, and 6650 XT graphics cards in comparison to other boards.
The findings use upcoming drivers that support the new parts, and AMD's Radeon RX 6950 XT, 6750 XT, and 6650 XT boards not only demonstrated significant advantages over the "regular" (6x00 XT) models, but they also beat their respective competitors (RTX 3090 Ti, RTX 3070 Ti, and RTX 3060) in 3DMark Time Spy graphics benchmark by a notable margin.
This is of course only one benchmark that does not reflect all existing games, and it also doesn't use ray tracing, a strong side of Nvidia's GeForce RTX 30-series. Port Royal does test RT performance, but WCCFTech only showed AMD's new GPUs in that particular test. Still, the advantages of the new GPUs appear to be substantial.
Row 0 - Cell 0 | 3DMark Time Spy Graphics Score |
Radeon RX 6950 XT | 22209 |
GeForce RTX 3090 Ti | 20855 |
GeForce RTX 3090 | 19690 |
GeForce RTX 3080 Ti | 19275 |
Radeon RX 6900 XT | 18463 |
While it is important to outpace competitors, arguably the most interesting aspect about the new Radeon RX 6x50 XT boards is their performance advantages over AMD's existing Radeon RX 6x00 XT cards. While we know that the upcoming products will have a GPU frequency lift and faster GDDR6 memory with a 18 GT/s data transfer rate, the new parts show significant gains — more than we'd expect from higher boost clocks and faster memory.
For example, AMD's Radeon RX 6950 XT appears to be about 17% faster than its predecessor, the Radeon RX 6900 XT. Perhaps WCCFTech used a heavily factory-overclocked Radeon RX 6950 XT board, or AMD significantly upped the TGP of its upcoming products, or AMD's new drivers simply improve performance of all Radeon RX 6000-series boards and existing cards were not retested at this time. But whatever the case, the upcoming graphics cards seem to show significant gains in 3DMark.
Benchmark results of AMD's upcoming GPUs published by WCCFTech were obtained on a testbed powered by AMD's Ryzen 7 5800X3D processor and equipped with DDR4-3600 memory. The system ran Windows 10 and used AMD's upcoming 22.10-220411n drivers that support the company's refreshed Radeon RX 6x50 XT graphics cards. WCCFTech (or one of its sources) ran 3DMark Time Spy, Fire Strike, Port Royale, and 3DMark 11 benchmarks on the new graphics boards.
Stay On the Cutting Edge: Get the Tom's Hardware Newsletter
Get Tom's Hardware's best news and in-depth reviews, straight to your inbox.
Anton Shilov is a contributing writer at Tom’s Hardware. Over the past couple of decades, he has covered everything from CPUs and GPUs to supercomputers and from modern process technologies and latest fab tools to high-tech industry trends.
-
-Fran- If true, it would make the 3090ti look even dumber than what it is.Reply
That being said, I'll just wait for reviews and see if I get this or wait for the next gen to replace my 6900XT for VR stuff.
Regards. -
artk2219 Huh, thats a bigger uplift than I was expecting, I guess we'll see how the benchmarks turn out.Reply -
mdd1963 Good 'ole '3d Mark Time Spy' scores...!Reply
Yes, that's how one compares/chooses a GPU these days! -
Sleepy_Hollowed Hope it's not too much TDP increased, but this is promising not just for regular games, but as an alternative to pro cards if they're more available.Reply
That is, unless you require CUDA use. -
vasthegreat I'll believe in when I see it. Every time new AMD cards are about to be released there's always this slew of articles claiming they will be faster than the current Nvidia top dog, then that speed never materializes in the consumer cards.Reply -
spentshells If you tuned them both down to the point they are just about smoking with a 1200 watt psu, im sure the 3090ti comes out on top, there are far more resources available..... unless AMD secretly updated the cache speed.Reply -
mitch074
Making infinity cache perform better ? This is not trivial when done outside the driver. It would be interesting to test the RX6900 with early drivers then with the latest beta ones, and see the difference - +10% from drivers, +7% from clock speed + throughput would make a lot of sense, actually.spentshells said:If you tuned them both down to the point they are just about smoking with a 1200 watt psu, im sure the 3090ti comes out on top, there are far more resources available..... unless AMD secretly updated the cache speed.
Not that I would care, as I'm running open source drivers - even a 6 years old card (RX480 8 Gb, reference + 3rd party cooler, no voltmod) keeps getting performance optimizations, so much so that I really am under no pressure to replace it other than the fact it might give up the ghost before then.
I'd also really much like to crank up the framerate @1440p a bit higher, but I'm not ready to spend twice the amount my current one cost me back then simply to gain 20 fps.
And no Nvidia for me - on Linux, their drivers are crap nowadays. -
spentshells
Also your sig... that there is a code 15... approximately 15 inches between the computer and the problem.mitch074 said:Making infinity cache perform better ? This is not trivial when done outside the driver. It would be interesting to test the RX6900 with early drivers then with the latest beta ones, and see the difference - +10% from drivers, +7% from clock speed + throughput would make a lot of sense, actually.
Not that I would care, as I'm running open source drivers - even a 6 years old card (RX480 8 Gb, reference + 3rd party cooler, no voltmod) keeps getting performance optimizations, so much so that I really am under no pressure to replace it other than the fact it might give up the ghost before then.
I'd also really much like to crank up the framerate @1440p a bit higher, but I'm not ready to spend twice the amount my current one cost me back then simply to gain 20 fps.
And no Nvidia for me - on Linux, their drivers are crap nowadays. -
mitch074
As long at ID doesn't equal 10T, I'm fine.spentshells said:Also your sig... that there is a code 15... approximately 15 inches between the computer and the problem. -
alceryes My reference 6900 XT (slightly OCd, no fancy cooling) gets within spitin' distance of the 6950 XT leaked Time Spy score and over its leaked Port Royal score.Reply
When the 6900 XT was initially released I think AMD purposely underrated its performance so they could easily come out with a half-gen refresh that beats those original 'low' performance numbers.