• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.
  • The forums have been upgraded with support for dark mode. By default it will follow the setting on your system/browser. You may override it by scrolling to the end of the page and clicking the gears icon.

Possible Radeon RX 6700 XT Specs Surface, 12GB the New Mid-Range Memory Size?

btarunr

Editor & Senior Moderator
Staff member
Joined
Oct 9, 2007
Messages
47,783 (7.41/day)
Location
Dublin, Ireland
System Name RBMK-1000
Processor AMD Ryzen 7 5700G
Motherboard Gigabyte B550 AORUS Elite V2
Cooling DeepCool Gammax L240 V2
Memory 2x 16GB DDR4-3200
Video Card(s) Galax RTX 4070 Ti EX
Storage Samsung 990 1TB
Display(s) BenQ 1440p 60 Hz 27-inch
Case Corsair Carbide 100R
Audio Device(s) ASUS SupremeFX S1220A
Power Supply Cooler Master MWE Gold 650W
Mouse ASUS ROG Strix Impact
Keyboard Gamdias Hermes E2
Software Windows 11 Pro
AMD could follow up on its RX 6800 series and RX 6900 XT launches with the RX 6700 series, which logically succeeds the RX 5700 series, and competes with NVIDIA's RTX 3060/Ti. Patrick Schur on Twitter, who has a high hit-rate with specs of upcoming AMD products, put out possible specs of the RX 6700 series. Both are based on the new "Navi 22" silicon, with an interesting set of specifications.

Apparently 12 GB could be AMD's new memory amount for the mid-range. It's unknown whether the 12 GB is running over a 192-bit wide memory interface (6x 16 Gbit chips), or whether AMD is using mixed-density chips over a 256-bit wide memory bus (think 4x 16 Gbit and 4x 8 Gbit), because even the fastest JEDEC-standard GDDR6 chips, running at 16 Gbps, would only yield 384 GB/s memory bandwidth, which is less than the 448 GB/s the RX 5700 series enjoy. Perhaps an Infinity Cache is deployed to make up the difference?



As for power, the RX 6700 XT (described as Navi 22 XT), is expected to have typical graphics power of 186-221 W, while the RX 6700 (Navi 22 XTL), could have 146-156 W TGP), which could mean that at least the RX 6700 could make do with a single 8-pin PCIe power input. VideoCardz predicts that the RX 6700 XT could feature 40 RDNA2 compute units—so 2,560 stream processors; and the RX 6700 featuring 36 (2,304), so the same exact number of shaders as the previous generation. The secret sauce here could be the significantly higher engine clocks, and possibly an Infinity Cache.

View at TechPowerUp Main Site
 
Now THIS is what I'm talkin' about. I'm excited. Hope it turns out to be true AND affordable/readily available...
 
With 40CUs it will be faster than a 5700XT.
Eagerly waiting for IPC (or sth) comparison. 6700 XT will be faster because of faster clock rate, yes. But I want @Wizzard to explore generational performance increase.
 
Not too sure about this one...VRAM requirements are heavily exaggerated these days and its RT performance will not be enough to sway 3060/Ti/70 buyers, especially for Cyberpunk. Will have to be shoved down one budget tier at least (300-350 dollars) to duke it out with the 3050
 
I can see these cards both being relatively popular when they come out and stock is viable. RDNA2 is shaping up nicely, all things considered, and is going a good way at least reputationally so far in slowly changing the tide.
 
Last edited:
Gap to rx 6800 seems to be quite huge though. So I doubt it will come close to rtx3070/2080ti perf, whose are just few percents behind of rx6800 on tpu's tests.

Given Nvidia's own performance claims putting RTX 3060 Ti on par with RTX 3070 | RTX 2080 Ti, and that RX 5700 XT was between RTX 2070 and RTX 2070 Super, there is no reason not to believe that RX 6700 XT will compete with RTX 3070 | RTX 3060 Ti and RTX 2080 Ti.
 
I see the 6700 XT being above the 2080 Super but below the 2080 Ti, could be good if he price is right, but availability is anyone's guess.
 
Given Nvidia's own performance claims putting RTX 3060 Ti on par with RTX 3070 | RTX 2080 Ti, and that RX 5700 XT was between RTX 2070 and RTX 2070 Super, there is no reason not to believe that RX 6700 XT will compete with RTX 3070 | RTX 3060 Ti and RTX 2080 Ti.
RTX 3060Ti has only 38SM and RTX 3070 has 46SM, that's a 21% difference, that's not on par in performance but at least 10% behind.
RTX 3070 is 52% faster in 4K than RX 5700XT, so there is no reason to believe a 40CU N22 will perform on the same level, unless you believe the clockspeed will be over 2.6GHz.
 
Last edited:
Given Nvidia's own performance claims putting RTX 3060 Ti on par with RTX 3070 | RTX 2080 Ti, and that RX 5700 XT was between RTX 2070 and RTX 2070 Super, there is no reason not to believe that RX 6700 XT will compete with RTX 3070 | RTX 3060 Ti and RTX 2080 Ti.

Uhm rtx3060ti is claimed to be slightly faster than 2080Super like rtx3070 is to 2080ti, there's still a 16% performance gap between the two. But rx6800 has 1.5 times more shaders than rumored rx6700xt and that is just 6% faster overall than rtx3070 on TPU's tests(Like rx6800 is ~52% faster than RX5700xt, which sounds too convenient).
 
Radeon RX 6700 XT with improved RDNA 2 architecture and much more efficient CUs will be on par with RTX 2080 Ti, while with considerably lower power consumption than the original RX 5700 XT.
40CU N22 won't be 50% faster than RX 5700XT to be on par with RTX 2080Ti. From the reviews of big navi, there doesn't seem to be any or very small IPC gain, so the performance would need to come from higher clockspeed and N22 won't have 50% higher clocks than RX 5700XT.
 
Last edited:
40CU N22 won't be 50% faster than RX 5700XT to be on par with RTX 2080Ti.

It doesn't need to be 50%, but only 34% faster:
1606208173572.png
 
If it has 40 CU's only -it will be a large performance gap to 6800's 60 cu. AMd should probably release a 50 cu XT and a 40 cu non XT with room to release more cards downstream.
 
Navi 21 shows 50-60% performance per watt improvement over Navi 10.


If it has 40 CU's only -it will be a large performance gap to 6800's 60 cu. AMd should probably release a 50 cu XT and a 40 cu non XT with room to release more cards downstream.

If clocks are not taken into consideration.
RX 6800 shows low performance for its specification - I don't know why, though.
 
Not too sure about this one...VRAM requirements are heavily exaggerated these days and its RT performance will not be enough to sway 3060/Ti/70 buyers, especially for Cyberpunk. Will have to be shoved down one budget tier at least (300-350 dollars) to duke it out with the 3050

Mhm, you keep believing son.

AMD is not putting 12GB on cards lower than last year's top end 11GB 2080ti for giggles.

They do it because consoles have a similar amount and games are hardly even ports anymore, its more unified development now than it ever was. It may be hard to believe, but Nvidia might run out of capacity sooner than you can blink. Even the shittiest early console launchgame already wants over 9 GB for 4K (Godfall).

I'm sure they will find a way (driver) around it like they've always done, but it won't benefit performance. There's also still a rabbit in the hat with RTX IO.

Whatever way it develops... its very clear VRAM requirements are going to increase across the board, as they have in the past few generations already. Since Maxwell we have 2,5x the VRAM already (4 > 10GB) and that is only counting the lowest increase on x80 GPUs. With this AMD release, that will expand to 3x. You're living in fantasy land if you think this is just for marketing purposes.
 
Last edited:
AMD is doing to Nvidia in regards to memory what it did to Intel in regards to cores. They are putting products with significantly more VRAM on the market, which encourages devs to make use of it. This in turn is going to put long term pressure on Nvidia as more and more games take advantage of that memory (like Godfall for example). It's a bigger deal for Nvidia as their architecture is far more reliant on fast memory, so adding more is expensive. Not to mention, the signaling requirements. See how close those GDDR6X chips are to the die? To be honest Nvidia should have added more VRAM to it's chips anyways. It's been 3 generations and if you upgrade from a 1080 Ti to a 3080 you are actually loosing 1GB, which is ridiculous. Both have similar launch prices. It's going to be pretty embarrising if AMD's entry level cards have as much VRAM as the 3080.

Not too sure about this one...VRAM requirements are heavily exaggerated these days and its RT performance will not be enough to sway 3060/Ti/70 buyers, especially for Cyberpunk. Will have to be shoved down one budget tier at least (300-350 dollars) to duke it out with the 3050

In the same way that the 1060 3GB was appealing. Might look nice for a year or two until the lack of VRAM kicks in and performance is trashed as a result.

Cyberpunk uses 10GB of VRAM with RT on. Not sure how that will play out on cards with 6-8GB.
 
Mhm, you keep believing son.

AMD is not putting 12GB on cards lower than last year's top end 11GB 2080ti for giggles.

They do it because consoles have a similar amount and games are hardly even ports anymore, its more unified development now than it ever was. It may be hard to believe, but Nvidia might run out of capacity sooner than you can blink. Even the shittiest early console launchgame already wants over 9 GB for 4K (Godfall).

I'm sure they will find a way (driver) around it like they've always done, but it won't benefit performance. There's also still a rabbit in the hat with RTX IO.

Whatever way it develops... its very clear VRAM requirements are going to increase across the board, as they have in the past few generations already. Since Maxwell we have 2,5x the VRAM already (4 > 10GB) and that is only counting the lowest increase on x80 GPUs. With this AMD release, that will expand to 3x. You're living in fantasy land if you think this is just for marketing purposes.

I wonder if their memory bandwith cant make up for the memory deficit.
 
I wonder if their memory bandwith cant make up for the memory deficit.
If a game needs 10GB Vram and your card has only 8GB then 2GB of data will have to be loaded from the system memory RAM, so in that case It won't matter how fast your Vram bandwidth is, because you are not loading the last 2GB of data from the Vram but RAM.
 
A smaller infinity cache in Navi22 of what the Navi21 has could make it work for the smaller bandwidth than the Navi10. And efficiency and much higher clockspeeds could make 6700XT match or slightly surpass 2080S performance in 1440P with low power draw for its performance. Which for $400 will be great but if sold for $450 it will be marginally good. Depends on the price of nVidia's 3060Ti ofc. 6700(XT) might become the new 7870 for AMD (budget friendly and great in power/efficiency chip back then).
 
Back
Top