• Hey Guest. Check out your NeoGAF Wrapped 2025 results here!

RDNA 5 line up and performance expectations

1C90XT, most definitely over.
1C70XT will be fairly close in raster, even closer in RT.
I wonder what NVIDIA tech my great, great, great grand kids will be playing when those release.

Edit: Misread your model, thought you were going for 1C090XT.... But ya. AMD isn't getting close to 5090. They keep skipping halo GPUs like with this gen.
 
Last edited:
I wonder what NVIDIA tech my great, great, great grand kids will be playing when those release.

Edit: Misread your model, thought you were going for 1C090XT.... But ya. AMD isn't getting close to 5090. They keep skipping halo GPUs like with this gen.
Don't put extra 0 in there. C mean 100 in Roman Numerals.
 
Some features would be nice, such as Opacity Micromaps, Displaced Micro-Mesh, Mega Geometry, Triangle Cluster Intersection for RT, Linear Swept Spheres, Shader Execution Reordering, Neural Shaders / Cooperative Vectors, improved Ray Regeneration, full pipelined neural memory compression.
 
RDNA 4 was a huge improvement over RDNA 3 in Q1 2025, giving us FSR 4 image scaling and big improvements in Ray Tracing.

Now RDNA 5 will be coming out in 2027 (Q2? Q3?), with even bigger improvements, using TSMC N3P. So let's look at the leaked line up with speculated clocks, RAM, and performance vs RDNA 4 / RTX. (leaks says 1.1x raster and 2X RT of RDNA 4):

RX 10050XT? (AT4) -
24 CU @ 3.5 GHz (?)
12~16GB 128bit LPDDR5X 10000 (?) 160 GB/s
~95% raster, ~175% RT of 9060XT
~100% raster, ?RT of RTX 3070

$300~350

RX 10060XT? (AT3) -
48 CU @ 3.4 GHz (?)
16~24GB 256bit LPDDR5X 10000 (?) 320 GB/s
~190% raster, ~340% RT of 9060XT
~105% raster, ?RT of 4080 Super

$450~550

RX 10070XT? (AT2) -
68 CU @ 3.2 GHz (?)
16~24GB 196bit GDDR7 864 GB/s
~125% raster, ~227% RT of 9070XT
~ 104% raster, ?RT of RTX 5080

$700~800

RX 10090XT? (AT0) -
96 CU @ 3 GHz (?)
24~36GB 384bit GDDR7 (?) 1.7TB/s
~165% raster, ~300% RT of 9070XT
~90% raster, ?RT of RTX 5090

$1000~1200

I am soooo fucking excited for RDNA 5....
This is completely wrong.

AT4 dGPU WILL 99% NOT EXIST FOR GAMING
AT3 dGPU WILL 100% NOT EXIST FOR GAMING

Why? Because AT3 MINIMUM Memory config is 32GB of Ultra Premium LPDDR6. The least they can sell that for is 800$.

AT4 it's below Navi33 and the minimum memory config is 16GB of LPDDR5X; Where they will need to buy the highest bin LPDDR5X for it. for a Navi 33 victim forget about it.

THAT Aside, this is the actual lineup:

RX 1070 GRE (AT2) -
44 CU @ 3.2 GHz
215W
12GB 128bit GDDR7-36Gbps 576 GB/s
~105% raster/RT of RTX 4070 Ti

$499

RX 1070 (AT2) -
48 CU @ 3.2 GHz
235W
15GB 160bit GDDR7-36Gbps 720 GB/s
~103% raster/RT of RTX 4070 Ti SUPER

$549

RX 1070 XT (AT2) -
64 CU @ 3.2 GHz
275W
18GB 192bit GDDR7-36Gbps 864 GB/s
~100% raster/RT of RTX 5080

$649

RX 1080 (AT0) -
WILL 70-90% NOT RELEASE FOR GAMERS

152 CU @ 2.6 GHz
380W
36GB 384bit GDDR7-36Gbps 1.7TB/s
~95% Raster/RT of RTX 5090

$♾️
 
Last edited:
This is completely wrong.

AT4 dGPU WILL 99% NOT EXIST FOR GAMING
AT3 dGPU WILL 100% NOT EXIST FOR GAMING

Why? Because AT3 MINIMUM Memory config is 32GB of Ultra Premium LPDDR6. The least they can sell that for is 800$.

AT4 it's below Navi33 and the minimum memory config is 16GB of LPDDR5X; Where they will need to buy the highest bin LPDDR5X for it. for a Navi 33 victim forget about it.

THAT Aside, this is the actual lineup:

RX 1070 GRE (AT2) -
44 CU @ 3.2 GHz
215W
12GB 128bit GDDR7-36Gbps 576 GB/s
~105% raster/RT of RTX 4070 Ti

$499

RX 1070 (AT2) -
48 CU @ 3.2 GHz
235W
15GB 160bit GDDR7-36Gbps 720 GB/s
~103% raster/RT of RTX 4070 Ti SUPER

$549

RX 1070 XT (AT2) -
64 CU @ 3.2 GHz
275W
18GB 192bit GDDR7-36Gbps 864 GB/s
~100% raster/RT of RTX 5080

$649

RX 1080 (AT0) -
WILL 99% NOT RELEASE FOR GAMERS

152 CU @ 2.6 GHz
380W
36GB 384bit GDDR7-36Gbps 1.7TB/s
~95% Raster/RT of RTX 5090

$♾️

RX 1080 XT (AT0) -
WILL 100% NOT RELEASE FOR GAMERS

184 CU @ 2.9 GHz
600W
48/128GB 512bit GDDR7-36Gbps 2.3TB/s
~115% Raster/RT of RTX 5090

$♾️
I like the thread engagement. I dont jibe your logic nor details.

But I love the effort!
 
I wanna say those prices are terrible, but with the current state of things they're actually quite good. That naming has to change though. 10070 should at least change to 10700 if you want to keep 5 digits or you could replace the 10 by the roman numeral X and have it only be 4 digits X070.
AMD always had stupid naming schemes. NVIDIA has kept the same one since GTX 200, only replacing the G with an R.

AMD changes it every other generation.
 
AMD always had stupid naming schemes. NVIDIA has kept the same one since GTX 200, only replacing the G with an R.

AMD changes it every other generation.

And they named the latest cards 90x0 - where do you go from that?

They will have to reset the naming system again, so stupid... (same with the CPUs)
 
OMG show us. I wanna see
It's public information, posted by MLID.

c62hQ4t.jpeg


90 / 99% refers to parametric harvesting. But the rest is straightforward. You can see there is only 4 planned gaming SKUs. One with AT0 and three with AT2.

AT3 / AT4 are designed for Medusa Halo and Medusa Premium. They can be used as dGPUs, but they only make sense as AI dGPUs. Basically if you're using LPDDR, it's a memory capacity play. Gaming on the other hand doesn't need 32GB on cards barely beating RTX 5060 Ti.
 
It's public information, posted by MLID.

c62hQ4t.jpeg


90 / 99% refers to parametric harvesting. But the rest is straightforward. You can see there is only 4 planned gaming SKUs. One with AT0 and three with AT2.

AT3 / AT4 are designed for Medusa Halo and Medusa Premium. They can be used as dGPUs, but they only make sense as AI dGPUs. Basically if you're using LPDDR, it's a memory capacity play. Gaming on the other hand doesn't need 32GB on cards barely beating RTX 5060 Ti.
Jeez, I thought you had something new. This is ooooolllld. MLiD literally leaked stuff hat supplanted this chart.
 
Jeez, I thought you had something new. This is ooooolllld. MLiD literally leaked stuff hat supplanted this chart.
The chart hasn't changed. The only change is that AT0 Gaming is no longer on the roadmap for gaming.

You have to separate MLID's mindless AMD cheerleading from what he is actually leaking.

To cut the truth from fiction.

AT3/4 dGPU Gaming SKUs aren't real.
 
AT3/4 dGPU Gaming SKUs aren't real.
Even theoretically speaking, AT3/4 Gaming don't make sense.

Again, AT2 with 15GB 160b G7 costs a lot less than AT3 with 32GB LPDDR6. It ALSO PERFORMS A LOT BETTER. Like 40-45% better.

Remember also that a 48CU cut of AT2 has 96ROPs. But 48CU AT3 has 64ROPs. That's a very significant difference.
 
Even theoretically speaking, AT3/4 Gaming don't make sense.

Again, AT2 with 15GB 160b G7 costs a lot less than AT3 with 32GB LPDDR6. It ALSO PERFORMS A LOT BETTER. Like 40-45% better.

Remember also that a 48CU cut of AT2 has 96ROPs. But 48CU AT3 has 64ROPs. That's a very significant difference.
I dont know why you keep insisting LPDDR6 as a must. Its an option for AT3. MLiD update had both LPDDR5X @ 356bit bus and LPDDR6 @ 384bit bus.

Also, Helix uses AT2. It has 68 active CUs from 70 CU die. 48 CU is an ancient version from old ass roadmap. Things definitely got updated.

MLiD updated RDNA 5 leaks multiple times since that old chart he leaked that you keep holding on to for dear life. He got updated leaks from same sources that gave him that chart.

He has said AT4 is entry card and AT3 is 10060XT. If that's not correct, his sources would have contacted him. The sane sources that got him that chart in the first place!
 
Last edited:
I dont know why you keep insisting LPDDR6 as a must
You really don't get it ha?

The minimum memory capacity for a 256b LP5X | 384b LP6 is 32GB. The issue is that 32GB is too god damn much for 9070 GRE performance.

And if it's LP5X, it's going to be RTX 4070 performance.

Do you think that RTX 4070 ~ RX 9070 GRE performance is compelling enough to sell for 700-800$ to gamers? LMAO. 4070 Ti at 799$ was a tough pill to swallow in 2022……

Or they can give you RTX 4070 Ti performance at 500$ with 12GB GDDR7.

it's not complicated.
 
Last edited:
Top Bottom