Img to video 480p BF16 with 16GB preset is using 11,6-11,8GB of VRAM

JJaviEM Img to video 480p BF16 with 16GB preset is using 11,6-11,8GB of VRAM

F

Furkan Gözükara SECourses•3/16/25, 12:48 PM

i am testing that right now :d

F

Furkan Gözükara SECourses•3/16/25, 12:48 PM

480x832 px, 81 frames, 14b 480p image to video model - BF16 precision

0 is : 8618 MB
1 is : 8618 MB
1,000,000,000 is : 9354 MB
2,000,000,000 is : 11370 MB
3,000,000,000 is : 13080 MB

F

Furkan Gözükara SECourses•3/16/25, 12:49 PM

these are exact usages so you also need to add your windows usage

F

Furkan Gözükara SECourses•3/16/25, 12:50 PM

i will publish as a table hopefully after finishing all :d

J

JaviEMOP•3/16/25, 12:52 PM

F

Furkan Gözükara SECourses•3/16/25, 1:28 PM

GPU VRAM Usages Table With Respected To Number of Persistent Params

1280x720 px, 81 frames, 14b text to video model - BF16 precision (torch.bfloat16)

All tests made on secondary GPU thus they are 100% accurate 

0 is : 14168 MB
1 is : 14168 MB
1,000,000,000 is : 15658 MB
2,000,000,000 is : 17664 MB
3,000,000,000 is : 20134 MB
3,750,000,000 is : 21134 MB
4,000,000,000 is : 21500 MB
4,250,000,000 is : 22000 MB


480x832 px, 81 frames, 14b 480p image to video model - BF16 precision (torch.bfloat16)

0 is : 8618 MB
1 is : 8618 MB
1,000,000,000 is : 9354 MB
1,500,000,000 is : 10090 MB
2,000,000,000 is : 11370 MB
3,000,000,000 is : 13080 MB
3,500,000,000 is : 14038 MB
4,000,000,000 is : 15140 MB
8,000,000,000 is : 22100 MB


720x1280 px, 81 frames, 14b 720p image to video model - BF16 precision (torch.bfloat16)

0 is : 18326 MB
1 is : 18326 MB
2,000,000,000 is : 20360 MB
3,000,000,000 is : 22010 MB
3,500,000,000 is : x MB
4,000,000,000 is : x MB
8,000,000,000 is : x MB

GPU VRAM Usages Table With Respected To Number of Persistent Params

1280x720 px, 81 frames, 14b text to video model - BF16 precision (torch.bfloat16)

All tests made on secondary GPU thus they are 100% accurate 

0 is : 14168 MB
1 is : 14168 MB
1,000,000,000 is : 15658 MB
2,000,000,000 is : 17664 MB
3,000,000,000 is : 20134 MB
3,750,000,000 is : 21134 MB
4,000,000,000 is : 21500 MB
4,250,000,000 is : 22000 MB


480x832 px, 81 frames, 14b 480p image to video model - BF16 precision (torch.bfloat16)

0 is : 8618 MB
1 is : 8618 MB
1,000,000,000 is : 9354 MB
1,500,000,000 is : 10090 MB
2,000,000,000 is : 11370 MB
3,000,000,000 is : 13080 MB
3,500,000,000 is : 14038 MB
4,000,000,000 is : 15140 MB
8,000,000,000 is : 22100 MB


720x1280 px, 81 frames, 14b 720p image to video model - BF16 precision (torch.bfloat16)

0 is : 18326 MB
1 is : 18326 MB
2,000,000,000 is : 20360 MB
3,000,000,000 is : 22010 MB
3,500,000,000 is : x MB
4,000,000,000 is : x MB
8,000,000,000 is : x MB

GPU VRAM Usages Table With Respected To Number of Persistent Params

1280x720 px, 81 frames, 14b text to video model - BF16 precision (torch.bfloat16)

All tests made on secondary GPU thus they are 100% accurate 

0 is : 14168 MB
1 is : 14168 MB
1,000,000,000 is : 15658 MB
2,000,000,000 is : 17664 MB
3,000,000,000 is : 20134 MB
3,750,000,000 is : 21134 MB
4,000,000,000 is : 21500 MB
4,250,000,000 is : 22000 MB


480x832 px, 81 frames, 14b 480p image to video model - BF16 precision (torch.bfloat16)

0 is : 8618 MB
1 is : 8618 MB
1,000,000,000 is : 9354 MB
1,500,000,000 is : 10090 MB
2,000,000,000 is : 11370 MB
3,000,000,000 is : 13080 MB
3,500,000,000 is : 14038 MB
4,000,000,000 is : 15140 MB
8,000,000,000 is : 22100 MB


720x1280 px, 81 frames, 14b 720p image to video model - BF16 precision (torch.bfloat16)

0 is : 18326 MB
1 is : 18326 MB
2,000,000,000 is : 20360 MB
3,000,000,000 is : 22010 MB
3,500,000,000 is : x MB
4,000,000,000 is : x MB
8,000,000,000 is : x MB

GPU VRAM Usages Table With Respected To Number of Persistent Params

1280x720 px, 81 frames, 14b text to video model - BF16 precision (torch.bfloat16)

All tests made on secondary GPU thus they are 100% accurate 

0 is : 14168 MB
1 is : 14168 MB
1,000,000,000 is : 15658 MB
2,000,000,000 is : 17664 MB
3,000,000,000 is : 20134 MB
3,750,000,000 is : 21134 MB
4,000,000,000 is : 21500 MB
4,250,000,000 is : 22000 MB


480x832 px, 81 frames, 14b 480p image to video model - BF16 precision (torch.bfloat16)

0 is : 8618 MB
1 is : 8618 MB
1,000,000,000 is : 9354 MB
1,500,000,000 is : 10090 MB
2,000,000,000 is : 11370 MB
3,000,000,000 is : 13080 MB
3,500,000,000 is : 14038 MB
4,000,000,000 is : 15140 MB
8,000,000,000 is : 22100 MB


720x1280 px, 81 frames, 14b 720p image to video model - BF16 precision (torch.bfloat16)

0 is : 18326 MB
1 is : 18326 MB
2,000,000,000 is : 20360 MB
3,000,000,000 is : 22010 MB
3,500,000,000 is : x MB
4,000,000,000 is : x MB
8,000,000,000 is : x MB

F

Furkan Gözükara SECourses•3/16/25, 1:29 PM

i am gonna update all presets

F

Furkan Gözükara SECourses•3/16/25, 1:30 PM

also now supports , in params

F

Furkan Gözükara SECourses•3/16/25, 1:31 PM

A

asap17z•3/16/25, 1:39 PM

I'm currently using 80gb

F

Furkan Gözükara SECourses•3/16/25, 1:39 PM

now i will test rest on 80 gb

F

Furkan Gözükara SECourses•3/16/25, 1:39 PM

for 80gb just enter like 20b

F

Furkan Gözükara SECourses•3/16/25, 1:39 PM

it will use vram entirely

A

asap17z•3/16/25, 1:42 PM

is it okay to increase the denoise for the sampler to 1?

Aasap17z is it okay to increase the denoise for the sampler to 1?

F

Furkan Gözükara SECourses•3/16/25, 1:43 PM

yes

F

Furkan Gözükara SECourses•3/16/25, 1:43 PM

it can change video to video entirely

F

Furkan Gözükara SECourses•3/16/25, 1:43 PM

you asking that right?

A

asap17z•3/16/25, 1:44 PM

I'm currently doing img to video with the 720p 14b parameter model

A

asap17z•3/16/25, 1:45 PM

I just wanted to know, although I set it to 0.95

Aasap17z I just wanted to know, although I set it to 0.95

F

Furkan Gözükara SECourses•3/16/25, 1:50 PM

ah it is not useful for image to video

F

Furkan Gözükara SECourses•3/16/25, 1:50 PM

only for video to video

A

asap17z•3/16/25, 1:52 PM

o okay

J

JaviEMOP•3/16/25, 2:04 PM

with new preset is using 15,6/15,8. It's fixed

JJaviEM with new preset is using 15,6/15,8. It's fixed

F

Furkan Gözükara SECourses•3/16/25, 2:05 PM

awesome

JJaviEM Click to see attachment

J

JaviEMOP•3/16/25, 2:06 PM

the gpu usage is more constant than before

JJaviEM the gpu usage is more constant than before

F

Furkan Gözükara SECourses•3/16/25, 2:06 PM

yes the lesser RAM usage the more GPU usage

J

JaviEMOP•3/16/25, 2:07 PM

and it's faster than ever

J

JaviEMOP•3/16/25, 2:07 PM

huge update

J

JaviEMOP•3/16/25, 2:10 PM

it's getting even faster. I have never seen less than 40s/it

JJaviEM it's getting even faster. I have never seen less than 40s/it

F

Furkan Gözükara SECourses•3/16/25, 2:13 PM

if teacache enabled it will

J

JaviEMOP•3/16/25, 2:47 PM

@Dr. Furkan Gözükara can you fix the profiles with FP8? It's even slower than bf16 now. It's only using 13GB VRAM and 64% GPU

J

JaviEMOP•3/16/25, 3:01 PM

FP8 videos are working fine now btw

JJaviEM FP8 videos are working fine now btw

F

Furkan Gözükara SECourses•3/16/25, 3:10 PM

for FP8

F

Furkan Gözükara SECourses•3/16/25, 3:10 PM

increase number of parameters

F

Furkan Gözükara SECourses•3/16/25, 3:12 PM

720x1280 px, 81 frames, 14b 720p image to video model - BF16 precision (torch.bfloat16)

0 is : 18326 MB
1 is : 18326 MB
2,000,000,000 is : 20360 MB
3,000,000,000 is : 22010 MB
6,000,000,000 is : 28380 MB
6,500,000,000 is : 29330 MB
6,750,000,000 is : 29790 MB
16,000,000,000 is : 47030 MB
22,000,000,000 is : 47800 MB


720x1280 px, 81 frames, 14b 720p image to video model - FP8 precision (torch.float8_e4m3fn)

3,000,000,000 is : 19763 MB
6,750,000,000 is : 23350 MB

720x1280 px, 81 frames, 14b 720p image to video model - BF16 precision (torch.bfloat16)

0 is : 18326 MB
1 is : 18326 MB
2,000,000,000 is : 20360 MB
3,000,000,000 is : 22010 MB
6,000,000,000 is : 28380 MB
6,500,000,000 is : 29330 MB
6,750,000,000 is : 29790 MB
16,000,000,000 is : 47030 MB
22,000,000,000 is : 47800 MB


720x1280 px, 81 frames, 14b 720p image to video model - FP8 precision (torch.float8_e4m3fn)

3,000,000,000 is : 19763 MB
6,750,000,000 is : 23350 MB

720x1280 px, 81 frames, 14b 720p image to video model - BF16 precision (torch.bfloat16)

0 is : 18326 MB
1 is : 18326 MB
2,000,000,000 is : 20360 MB
3,000,000,000 is : 22010 MB
6,000,000,000 is : 28380 MB
6,500,000,000 is : 29330 MB
6,750,000,000 is : 29790 MB
16,000,000,000 is : 47030 MB
22,000,000,000 is : 47800 MB


720x1280 px, 81 frames, 14b 720p image to video model - FP8 precision (torch.float8_e4m3fn)

3,000,000,000 is : 19763 MB
6,750,000,000 is : 23350 MB

720x1280 px, 81 frames, 14b 720p image to video model - BF16 precision (torch.bfloat16)

0 is : 18326 MB
1 is : 18326 MB
2,000,000,000 is : 20360 MB
3,000,000,000 is : 22010 MB
6,000,000,000 is : 28380 MB
6,500,000,000 is : 29330 MB
6,750,000,000 is : 29790 MB
16,000,000,000 is : 47030 MB
22,000,000,000 is : 47800 MB


720x1280 px, 81 frames, 14b 720p image to video model - FP8 precision (torch.float8_e4m3fn)

3,000,000,000 is : 19763 MB
6,750,000,000 is : 23350 MB

F

Furkan Gözükara SECourses•3/16/25, 3:12 PM

i am doing some testing

J

JaviEMOP•3/16/25, 3:12 PM

nice, thanks

F

Furkan Gözükara SECourses•3/16/25, 3:30 PM

480x832 px, 81 frames, 14b 480p image to video model - FP8 precision (torch.float8_e4m3fn)

0 is : 8618 MB
1 is : 8618 MB
1,000,000,000 is : x MB
1,500,000,000 is : x MB
2,000,000,000 is : x MB
3,000,000,000 is : x MB
3,500,000,000 is : x MB
4,000,000,000 is : x MB
8,000,000,000 is : 15392 MB
12,000,000,000 is : x MB
22,000,000,000 is : 23080 MB

480x832 px, 81 frames, 14b 480p image to video model - FP8 precision (torch.float8_e4m3fn)

0 is : 8618 MB
1 is : 8618 MB
1,000,000,000 is : x MB
1,500,000,000 is : x MB
2,000,000,000 is : x MB
3,000,000,000 is : x MB
3,500,000,000 is : x MB
4,000,000,000 is : x MB
8,000,000,000 is : 15392 MB
12,000,000,000 is : x MB
22,000,000,000 is : 23080 MB

480x832 px, 81 frames, 14b 480p image to video model - FP8 precision (torch.float8_e4m3fn)

0 is : 8618 MB
1 is : 8618 MB
1,000,000,000 is : x MB
1,500,000,000 is : x MB
2,000,000,000 is : x MB
3,000,000,000 is : x MB
3,500,000,000 is : x MB
4,000,000,000 is : x MB
8,000,000,000 is : 15392 MB
12,000,000,000 is : x MB
22,000,000,000 is : 23080 MB

480x832 px, 81 frames, 14b 480p image to video model - FP8 precision (torch.float8_e4m3fn)

0 is : 8618 MB
1 is : 8618 MB
1,000,000,000 is : x MB
1,500,000,000 is : x MB
2,000,000,000 is : x MB
3,000,000,000 is : x MB
3,500,000,000 is : x MB
4,000,000,000 is : x MB
8,000,000,000 is : 15392 MB
12,000,000,000 is : x MB
22,000,000,000 is : 23080 MB

F

Furkan Gözükara SECourses•3/16/25, 3:30 PM

i am gonna make presets to support fp8 too

F

Furkan Gözükara SECourses•3/16/25, 3:41 PM

all done

F

Furkan Gözükara SECourses•3/16/25, 3:41 PM

GPU VRAM Usages Table With Respected To Number of Persistent Params+

All tests made on secondary GPU thus they are 100% accurate

Therefore, make sure that you have such empty VRAM before starting the Wan 2.1 SECourses App 

1280x720 px, 81 frames, 14b text to video model - BF16 precision (torch.bfloat16)

0 is : 14168 MB - Below 24 GB GPUs
1,000,000,000 is : 15658 MB
2,000,000,000 is : 17664 MB
3,000,000,000 is : 20134 MB
3,750,000,000 is : 21134 MB
4,000,000,000 is : 21500 MB
4,250,000,000 is : 22000 MB - 24 GB GPUs
8,250,000,000 is : 29870 MB - 32 GB GPUs
22,000,000,000 is : 41700 MB - 48 GB GPUs

1280x720 px, 81 frames, 14b text to video model - FP8 precision (torch.float8_e4m3fn)

0 is : 14168 MB - Below 24 GB GPUs
1,000,000,000 is : 14918 MB
8,000,000,000 is : 21260 MB
8,750,000,000 is : 21980 MB - 24 GB GPUs
22,000,000,000 is : x MB


480x832 px, 81 frames, 14b 480p image to video model - BF16 precision (torch.bfloat16)

0 is : 8618 MB - Below 12 GB GPUs
1,000,000,000 is : 9354 MB
1,500,000,000 is : 10090 MB - 12 GB GPUs
2,000,000,000 is : 11370 MB
3,000,000,000 is : 13080 MB
3,500,000,000 is : 14038 MB - 16 GB GPUs
4,000,000,000 is : 15140 MB
8,000,000,000 is : 22100 MB - 24 GB GPUs
12,000,000,000 is : 30060 MB - 32 GB GPUs
22,000,000,000 is : 38300 MB - 48 GB GPUs


480x832 px, 81 frames, 14b 480p image to video model - FP8 precision (torch.float8_e4m3fn)

0 is : 8618 MB - Below 12 GB GPUs
2,500,000,000 is : 9794 MB - 12 GB GPUs
7,500,000,000 is : 14850 MB - 16 GB GPUs
8,000,000,000 is : 15392 MB
15,000,000,000 is : 21700 MB - 24 GB GPUs
22,000,000,000 is : 23080 MB - 32 GB GPUs


720x1280 px, 81 frames, 14b 720p image to video model - BF16 precision (torch.bfloat16)

0 is : 18326 MB - Below 24 GB GPUs
2,000,000,000 is : 20360 MB
3,000,000,000 is : 22010 MB - 24 GB GPUs
6,000,000,000 is : 28380 MB
6,500,000,000 is : 29330 MB
6,750,000,000 is : 29790 MB - 32 GB GPUs
16,000,000,000 is : 47030 MB - 48 GB GPUs
22,000,000,000 is : 47800 MB - 80 GB GPUs


720x1280 px, 81 frames, 14b 720p image to video model - FP8 precision (torch.float8_e4m3fn)

0 is : 18326 MB - Below 24 GB GPUs
3,000,000,000 is : 19763 MB
6,000,000,000 is : 22020 MB - 24 GB GPUs
14,000,000,000 is : 29840 MB
15,000,000,000 is : 30790 MB - 32 GB GPUs
22,000,000,000 is : 32013 MB - 48 GB GPUs

GPU VRAM Usages Table With Respected To Number of Persistent Params+

All tests made on secondary GPU thus they are 100% accurate

Therefore, make sure that you have such empty VRAM before starting the Wan 2.1 SECourses App 

1280x720 px, 81 frames, 14b text to video model - BF16 precision (torch.bfloat16)

0 is : 14168 MB - Below 24 GB GPUs
1,000,000,000 is : 15658 MB
2,000,000,000 is : 17664 MB
3,000,000,000 is : 20134 MB
3,750,000,000 is : 21134 MB
4,000,000,000 is : 21500 MB
4,250,000,000 is : 22000 MB - 24 GB GPUs
8,250,000,000 is : 29870 MB - 32 GB GPUs
22,000,000,000 is : 41700 MB - 48 GB GPUs

1280x720 px, 81 frames, 14b text to video model - FP8 precision (torch.float8_e4m3fn)

0 is : 14168 MB - Below 24 GB GPUs
1,000,000,000 is : 14918 MB
8,000,000,000 is : 21260 MB
8,750,000,000 is : 21980 MB - 24 GB GPUs
22,000,000,000 is : x MB


480x832 px, 81 frames, 14b 480p image to video model - BF16 precision (torch.bfloat16)

0 is : 8618 MB - Below 12 GB GPUs
1,000,000,000 is : 9354 MB
1,500,000,000 is : 10090 MB - 12 GB GPUs
2,000,000,000 is : 11370 MB
3,000,000,000 is : 13080 MB
3,500,000,000 is : 14038 MB - 16 GB GPUs
4,000,000,000 is : 15140 MB
8,000,000,000 is : 22100 MB - 24 GB GPUs
12,000,000,000 is : 30060 MB - 32 GB GPUs
22,000,000,000 is : 38300 MB - 48 GB GPUs


480x832 px, 81 frames, 14b 480p image to video model - FP8 precision (torch.float8_e4m3fn)

0 is : 8618 MB - Below 12 GB GPUs
2,500,000,000 is : 9794 MB - 12 GB GPUs
7,500,000,000 is : 14850 MB - 16 GB GPUs
8,000,000,000 is : 15392 MB
15,000,000,000 is : 21700 MB - 24 GB GPUs
22,000,000,000 is : 23080 MB - 32 GB GPUs


720x1280 px, 81 frames, 14b 720p image to video model - BF16 precision (torch.bfloat16)

0 is : 18326 MB - Below 24 GB GPUs
2,000,000,000 is : 20360 MB
3,000,000,000 is : 22010 MB - 24 GB GPUs
6,000,000,000 is : 28380 MB
6,500,000,000 is : 29330 MB
6,750,000,000 is : 29790 MB - 32 GB GPUs
16,000,000,000 is : 47030 MB - 48 GB GPUs
22,000,000,000 is : 47800 MB - 80 GB GPUs


720x1280 px, 81 frames, 14b 720p image to video model - FP8 precision (torch.float8_e4m3fn)

0 is : 18326 MB - Below 24 GB GPUs
3,000,000,000 is : 19763 MB
6,000,000,000 is : 22020 MB - 24 GB GPUs
14,000,000,000 is : 29840 MB
15,000,000,000 is : 30790 MB - 32 GB GPUs
22,000,000,000 is : 32013 MB - 48 GB GPUs

GPU VRAM Usages Table With Respected To Number of Persistent Params+

All tests made on secondary GPU thus they are 100% accurate

Therefore, make sure that you have such empty VRAM before starting the Wan 2.1 SECourses App 

1280x720 px, 81 frames, 14b text to video model - BF16 precision (torch.bfloat16)

0 is : 14168 MB - Below 24 GB GPUs
1,000,000,000 is : 15658 MB
2,000,000,000 is : 17664 MB
3,000,000,000 is : 20134 MB
3,750,000,000 is : 21134 MB
4,000,000,000 is : 21500 MB
4,250,000,000 is : 22000 MB - 24 GB GPUs
8,250,000,000 is : 29870 MB - 32 GB GPUs
22,000,000,000 is : 41700 MB - 48 GB GPUs

1280x720 px, 81 frames, 14b text to video model - FP8 precision (torch.float8_e4m3fn)

0 is : 14168 MB - Below 24 GB GPUs
1,000,000,000 is : 14918 MB
8,000,000,000 is : 21260 MB
8,750,000,000 is : 21980 MB - 24 GB GPUs
22,000,000,000 is : x MB


480x832 px, 81 frames, 14b 480p image to video model - BF16 precision (torch.bfloat16)

0 is : 8618 MB - Below 12 GB GPUs
1,000,000,000 is : 9354 MB
1,500,000,000 is : 10090 MB - 12 GB GPUs
2,000,000,000 is : 11370 MB
3,000,000,000 is : 13080 MB
3,500,000,000 is : 14038 MB - 16 GB GPUs
4,000,000,000 is : 15140 MB
8,000,000,000 is : 22100 MB - 24 GB GPUs
12,000,000,000 is : 30060 MB - 32 GB GPUs
22,000,000,000 is : 38300 MB - 48 GB GPUs


480x832 px, 81 frames, 14b 480p image to video model - FP8 precision (torch.float8_e4m3fn)

0 is : 8618 MB - Below 12 GB GPUs
2,500,000,000 is : 9794 MB - 12 GB GPUs
7,500,000,000 is : 14850 MB - 16 GB GPUs
8,000,000,000 is : 15392 MB
15,000,000,000 is : 21700 MB - 24 GB GPUs
22,000,000,000 is : 23080 MB - 32 GB GPUs


720x1280 px, 81 frames, 14b 720p image to video model - BF16 precision (torch.bfloat16)

0 is : 18326 MB - Below 24 GB GPUs
2,000,000,000 is : 20360 MB
3,000,000,000 is : 22010 MB - 24 GB GPUs
6,000,000,000 is : 28380 MB
6,500,000,000 is : 29330 MB
6,750,000,000 is : 29790 MB - 32 GB GPUs
16,000,000,000 is : 47030 MB - 48 GB GPUs
22,000,000,000 is : 47800 MB - 80 GB GPUs


720x1280 px, 81 frames, 14b 720p image to video model - FP8 precision (torch.float8_e4m3fn)

0 is : 18326 MB - Below 24 GB GPUs
3,000,000,000 is : 19763 MB
6,000,000,000 is : 22020 MB - 24 GB GPUs
14,000,000,000 is : 29840 MB
15,000,000,000 is : 30790 MB - 32 GB GPUs
22,000,000,000 is : 32013 MB - 48 GB GPUs

GPU VRAM Usages Table With Respected To Number of Persistent Params+

All tests made on secondary GPU thus they are 100% accurate

Therefore, make sure that you have such empty VRAM before starting the Wan 2.1 SECourses App 

1280x720 px, 81 frames, 14b text to video model - BF16 precision (torch.bfloat16)

0 is : 14168 MB - Below 24 GB GPUs
1,000,000,000 is : 15658 MB
2,000,000,000 is : 17664 MB
3,000,000,000 is : 20134 MB
3,750,000,000 is : 21134 MB
4,000,000,000 is : 21500 MB
4,250,000,000 is : 22000 MB - 24 GB GPUs
8,250,000,000 is : 29870 MB - 32 GB GPUs
22,000,000,000 is : 41700 MB - 48 GB GPUs

1280x720 px, 81 frames, 14b text to video model - FP8 precision (torch.float8_e4m3fn)

0 is : 14168 MB - Below 24 GB GPUs
1,000,000,000 is : 14918 MB
8,000,000,000 is : 21260 MB
8,750,000,000 is : 21980 MB - 24 GB GPUs
22,000,000,000 is : x MB


480x832 px, 81 frames, 14b 480p image to video model - BF16 precision (torch.bfloat16)

0 is : 8618 MB - Below 12 GB GPUs
1,000,000,000 is : 9354 MB
1,500,000,000 is : 10090 MB - 12 GB GPUs
2,000,000,000 is : 11370 MB
3,000,000,000 is : 13080 MB
3,500,000,000 is : 14038 MB - 16 GB GPUs
4,000,000,000 is : 15140 MB
8,000,000,000 is : 22100 MB - 24 GB GPUs
12,000,000,000 is : 30060 MB - 32 GB GPUs
22,000,000,000 is : 38300 MB - 48 GB GPUs


480x832 px, 81 frames, 14b 480p image to video model - FP8 precision (torch.float8_e4m3fn)

0 is : 8618 MB - Below 12 GB GPUs
2,500,000,000 is : 9794 MB - 12 GB GPUs
7,500,000,000 is : 14850 MB - 16 GB GPUs
8,000,000,000 is : 15392 MB
15,000,000,000 is : 21700 MB - 24 GB GPUs
22,000,000,000 is : 23080 MB - 32 GB GPUs


720x1280 px, 81 frames, 14b 720p image to video model - BF16 precision (torch.bfloat16)

0 is : 18326 MB - Below 24 GB GPUs
2,000,000,000 is : 20360 MB
3,000,000,000 is : 22010 MB - 24 GB GPUs
6,000,000,000 is : 28380 MB
6,500,000,000 is : 29330 MB
6,750,000,000 is : 29790 MB - 32 GB GPUs
16,000,000,000 is : 47030 MB - 48 GB GPUs
22,000,000,000 is : 47800 MB - 80 GB GPUs


720x1280 px, 81 frames, 14b 720p image to video model - FP8 precision (torch.float8_e4m3fn)

0 is : 18326 MB - Below 24 GB GPUs
3,000,000,000 is : 19763 MB
6,000,000,000 is : 22020 MB - 24 GB GPUs
14,000,000,000 is : 29840 MB
15,000,000,000 is : 30790 MB - 32 GB GPUs
22,000,000,000 is : 32013 MB - 48 GB GPUs

F

Furkan Gözükara SECourses•3/16/25, 3:41 PM

time to update the app

J

JaviEMOP•3/16/25, 4:07 PM

I have updated but preset doesn't change for FP8. 16GB preset still use 3,500,000,000 on FP8

J

JaviEMOP•3/16/25, 4:21 PM

With 7,500,000,000 is using 0,4GB of shared VRAM on a 16GB VRAM GPU. I will try lowering it to 7,000,000,000 or 6,500,000,000

F

Furkan Gözükara SECourses•3/16/25, 4:33 PM

i havent updated app yet still few missing

J

JaviEMOP•3/16/25, 4:58 PM

FP8 7,500,000,000: 1278.54 seconds
FP8 7,000,000,000: 1137.68 seconds

J

JaviEMOP•3/16/25, 5:17 PM

BF16 3,500,000,000: 1002.77 seconds

J

JaviEMOP•3/16/25, 5:17 PM

FP8 has no sense on a 16GB VRAM GPU

J

JaviEMOP•3/16/25, 5:18 PM

Img to video 480p model*

J

JaviEMOP•3/16/25, 5:33 PM

10 videos generated this afternoon and no glitches or image corruption with teacache. I only had one big colour change with FP8. Very nice updates today.

Img to video 480p BF16 with 16GB preset is using 11,6-11,8GB of VRAM

Similar Threads

Similar Threads

Similar Threads