Would love to see a complete step by step video on getting A1111 to work with my A770 LE, grabbed a bargin on this card and for 16Gb (intended main use would of been A1111) but cant seem to get ANY method to work. So a dumbed down tutorail for dounuts like me would be HUGELY appriciated. 😆😆
Great tutorial Bob, it's really interesting how fast paced the development is for these technologies is, like you said in the future the tools are going to be so much different, by the time Battlemage releases it will be like using a completely different tool for gen AI
Just tried this - amazing performance. 512x512 20steps with SD1.5 with A380 LP takes 7.5s per image, while an ARC A580 would take 2.7s. (for context, a GtX 1650 took 30+s, and an RTX 3060 12G is around 3s per image IIRC) Need to optimize that 1st warmup run though.
bang Alva, bikin review + tutor Stable Diffusion pake A380 dong, kayaknya belum ada yg bahas sejauh ini baik bhs Indonesia maupun Inggris secara khusus
Thanks for the video. I've been using my A750 with Stable Diffusion for awhile now. For $190 its a steal in terms of the gaming and AI performance it gives. Also nice choice of a prompt, an alchemist, showing off Intel's Alchemist GPU architecture.
I agree. The A750 is a great value mid-range GPU. Pretty much any game I throw at it, I can get 60fps @ 1080p high settings. I'm also impressed with performance on some AI/ML tasks.
vladmandic /automatic 's fork of AUTOMATIC1111 had been running well under Linux (Ubuntu's LTS) for months on an Intel A770 LE using OneAPI/ipex. It's good to see Window's has more options. Looks like it's time for me to experiment with Windows and OpenVINO.
@@plasmarade SDNext is way better than A1111 under native Linux or memory leak-fixed WSL2. SDNext supports HiRes, Lora, and SDXL that cannot be done with A1111. This video just show basic application of Stable Difussion. However, SDNext also has its own minor issue like ControlNet only works under SD1.5 and HiRes/batch conflict under SDXL.
@@plasmarade Hello, Honestly, I cannot speak to Intel's newest Arc drivers for Linux. I use old drivers downloaded from Intel's repositories. Further, it depends on what you want to do and which version of which distro you use. Kubuntu LTS is my preference. Kubuntu 22.04 LTS did not come with Arc's drivers; they needed to be installed manually. That meant I needed to use the CPU's iGPU to install Kubuntu 22.04 LTS , then use a terminal to install Arc's drivers, then physically install the graphics card, and eventually be a script kiddie and use a terminal to run the commands to install Intel's resources and Stable Diffusion. It was a pain, but necessary to use vladmandic /automatic 's fork of AUTOMATIC1111. I game on Windows, not Linux. That said, I also have an Arc A750 LE under Windows; it runs World of Tanks at a very solid 60fps/2560 by 1440 with all of the eye candy turned on. Intel Arc graphics cards are not supported by Kdenlive, my preferred video editor, for hardware rendering, video playback and whatnot. AV1 TH-cam video doesn't seem to have Arc support under Linux using Firefox, but Window's Edge browser will play hardware decoded AV1 TH-cam video. Finally, I think since Intel's drivers weren't a part of Kubuntu LTS, Discover -- the default system updater and software repository interface -- wants to upgrade some Intel compilers I installed, but cannot. So, for me, there are some drawbacks, but no deal breakers. And since using Stable Diffusion for more than just image generation has become an interesting activity for me, a new in box 16GB/256bit bus graphics card for ~US$350 (from NewEgg) was a very happy accident. The card was purchased as soon after launch as I could manage and training an embedding uses a lot of memory. Serendipity and trying to future proof for the win. And finally, for day to day computing, such as writing this comment, I cannot tell an Arc card is in the machine.
were openvino's dev team had a plan to expand the usage of SD-webui script to support inpainting checkpoint? At the moment we also have the problem to generate image using control-net reference because openvino script completely ignored control-net for now.
Hi, I am currently running Arch Linux with a 1070, and I primarily a gamer/game developer. I suppose I want to broaden my skills and learn AI. I am making a new system, looking at a full team blue stack with an intel CPU and GPU, I noticed the link is dead to the article. I just had the question of running Stable Diffusion with Intel GPUs with Linux, even if I get windows on it for dualboot, I will likely use WSL. Especially interested in seeing this progress with the new Battle Mage GPUs. Even without AI, the progress of Intel Graphics looks great even on Linux!
How do I get it to work with my intel iris xe graphics card? I struggle using different models the only one works is 1.5, but i cannot do sdxl, juggernaut xl, and many others.
Interesting. This is exactly what I'd expect A770 with 16Gb for VRAM to be used for. Decent performance too! On par with RTX 3060 12Gb as of right now. I'm kind of curious though, does it support RVC/Roop(Unlimited)/TTS Tortoise?
Why do you not have an installation tutorial yet? This is very annoying to try to get running. I keep downloading and re-downloading the SD.Next method and the A1111 and neither will output images when I click generate. I keep getting torch errors and all kinds of problems. I have no idea what I'm doing wrong. I got it to work one time but it was trying to use CPU and took 21 minutes to generate 1 image...You're trying to promote this software to consumers but you have to be a coding enthusiast to even get it working. I'm not familiar with GIT or PS commands. I've tried following a lot of different guides step by step and nothing is working like it's supposed to.
Oh, no! That shouldn't happen. We suggest you reach out to fellow Intel users and experts in our community at intel.ly/4ah6pzh for further discussions. However, you can also join our Discord at intel.ly/3RHd9z2 for product news and updates.
@@IntelGraphics just installed default Automatic 1111 and it utilizes GPU up to 100% without OpenVINO and any additional settings... can you tell me why we need OpenVINO?
I met another problem I can't solve - OpenVINO SD only uses default 1.5 checkpoint. All images look like generated by the same model regardless of what checkpoint I choose from the dropdown menu in the top left. Am I missing something? How to fix this?
I found a workaround: If I generate at least 1 image with CPU, all subsequent generations with GPU will work as intended (even if I change checkpoints) until SD restart. But this ads extra 15 minutes to workflow, which is a lot.
You can add 3rd party checkpoints. In the video you can see that I use Realistic Vision from CivitAI. You can also use SDXL models from CivitAI. Have you tried downloading these models, putting in the models folder then refreshing the checkpoints dropdown?
I'm getting error while generating images. I have A750 with i5 & 16gb RAM. Did i miss something? (I'm new to gen Ai space, trying to get started with SD)
Sorry to hear about the issue you are facing. May we know what is the application you are using to generate images? Please share the screenshot of the error by visiting intel.ly/3t2dMd8 We would love to discuss it further.
no information is clear. it is always vague. i am not sure if we will get results. we are going from video to video. someone please explain it properly.
Heyo! Love the stuff and have an A770 on the way right now, but is there any news on VR support? I'd love to really push this card to it's limits one day with DCS VR lol
Hi! Recently I bought Intel Arc A770 solely to run SD. I installed OpenVINO version of SD, but for some reason it doesn't see & doesn't use my GPU at all. I have "Accelerate with OpenVINO" dropdown script & selected GPU in it. But I'm not sure what options should I choose. During image generation my CPU load is about 70% while GPU is at 5%, so it's not CPU bottleneck. It's not GPU problem either cuz it works fine in games & other tasks. GPU drivers are up to date. I don't have Resizable BAR enabled but I doubt that it's the problem. Can't find any solution to this problem for hours now.
We're sorry to hear that you are facing difficulties. To check the instructions and configurations of OpenVINO on Intel GPUs, please visit the GitHub link via intel.ly/3wlfl7b For further assistance, you can post your query in our community forum at intel.ly/49dw3UB Intel OpenVINO Community Forum. Our team of product engineers and community experts will be able to provide more insights on this topic.
i think if intel gpus keep high vram in their gpus, they can be chosen more for ai or video editing because if you want to run high quality model you need vram instead of gpu speed so even if 3070 is faster, ultimately arc 770 can run higher quality models and it can run comfortably without out of the memory, stuttering and freezing, I think it can be a good choice especially for running llm locally, but they should do the same on the laptop, nvidia still offers 4 gb vram on laptops, which is a disgrace in this era, as if making fun of it.
Sir I have intel i5 12600k processor with liquid cooling and MSI Mag B660 motherboard with 16gb DDR4 ram....how can I run openvino for faster?..plzzz help.
To improve the performance of Intel OpenVINO in your system, we recommend that you initiate discussion and seek guidance at our community forums via intel.ly/3uTr9Na Our team of experts will assist you with an apt solution.
Hi! We appreciate your interest in knowing more about intel technologies. Please note that the best way to know more about the OpenVINO toolkit and its applications is by visiting our community forum. You can explore our forum via intel.ly/3NDhqRK for more queries and solutions about OpenVINO. Also, refer to intel.ly/3NEyHtA to know more about the OpenVINO toolkit. Happy learning!
Yes this is very annoying to try to get running. I keep downloading a redownloading the SD.Next method and the A1111 and neither will output images when I click generate. I have no idea what I'm doing wrong. I got it to work one time but it was trying to use CPU and took 21 minutes to generate 1 image...They're trying to promote this software to consumers but you have to be a coding enthusiast to even get it working.
@@SilverSparrow5 ikr, my 13400cpu is practically fried after 3 images. I found out pytorch needs to be CUDA enabled for gpu to use it in sdnext. But I can't even begin to debug this problem
first. scary RAM usage 20ish GB RAM usage just for SD 1.5 inference is a really high, i bet openvino is still unusable at SDXL + HiRes. are you sure it isn’t RAM leaked problem with openvino? second. with 16GB of VRAM intel GPU with openvino should at least do SD 1.5 LoRA training/finetuning/dreambooth. Only use for inference is a quite waste of it’s VRAM.
please have separate channels, one for gamers wanting to hear news about Arc\Intel, and one for AI(and possible other) professional uses of ARC, having videos for 2 very difference audiences in the same channel is very bad for discoverability for example: I want to get a notification for anything gaming related, but not for AI
I can follow up with SDXL. Show 1080x1080 images. Initial compile takes a bit. Those models are huge but similar to what you see here, each subsequent image is pretty speedy
@@BobDuffy That's the reason I plan to upgrade my ram to 64G and 14700K in the end of year. Openvino takes lots CPU resource to calculate pre-built and stores in A1111 cache, which slowly eats your disk space. Remeber to check and delete unnecessarily ones.
That has been a fake rumor that's been debunked countless times. Pat Gelsinger has wanted GPUs to be part of the future of Intel, and that vision started decades ago before he was CEO. Intel already has a public roadmap for the next 3 generations of GPUs. Intel is also committed to GPUs because they are now selling them to data centers and use the same architecture in their IGP. Their second generation consumer GPU, "Battlemage", is already shown up in leaks and is expected to launch early 2024.
@@__aceofspadesEverything else seemed to be accurate but that last part isn't accurate anymore, I think it's now more like 2nd half of 2024 than early 2024.
Would love to see a complete step by step video on getting A1111 to work with my A770 LE, grabbed a bargin on this card and for 16Gb (intended main use would of been A1111) but cant seem to get ANY method to work. So a dumbed down tutorail for dounuts like me would be HUGELY appriciated. 😆😆
Great tutorial Bob, it's really interesting how fast paced the development is for these technologies is, like you said in the future the tools are going to be so much different, by the time Battlemage releases it will be like using a completely different tool for gen AI
Just tried this - amazing performance.
512x512 20steps with SD1.5 with A380 LP takes 7.5s per image, while an ARC A580 would take 2.7s. (for context, a GtX 1650 took 30+s, and an RTX 3060 12G is around 3s per image IIRC)
Need to optimize that 1st warmup run though.
bang Alva, bikin review + tutor Stable Diffusion pake A380 dong, kayaknya belum ada yg bahas sejauh ini baik bhs Indonesia maupun Inggris secara khusus
really? 1 image for me takes 45 minutes.
Thanks for the video. I've been using my A750 with Stable Diffusion for awhile now. For $190 its a steal in terms of the gaming and AI performance it gives. Also nice choice of a prompt, an alchemist, showing off Intel's Alchemist GPU architecture.
I agree. The A750 is a great value mid-range GPU. Pretty much any game I throw at it, I can get 60fps @ 1080p high settings. I'm also impressed with performance on some AI/ML tasks.
Yeah, I'll give Intel a little bit of credit. I can run a wider variety of compute loads thanks to the 16GB VRAM on my A770.
vladmandic /automatic 's fork of AUTOMATIC1111 had been running well under Linux (Ubuntu's LTS) for months on an Intel A770 LE using OneAPI/ipex. It's good to see Window's has more options. Looks like it's time for me to experiment with Windows and OpenVINO.
How's the experience of running arc under Linux,
I'm thinking of buying a 770 for my machine but I don't know if the drivers are fully there yet.
SDNext has native support for Windows with IPEX too now.
OpenVINO is also supported.
@@plasmarade SDNext is way better than A1111 under native Linux or memory leak-fixed WSL2. SDNext supports HiRes, Lora, and SDXL that cannot be done with A1111. This video just show basic application of Stable Difussion. However, SDNext also has its own minor issue like ControlNet only works under SD1.5 and HiRes/batch conflict under SDXL.
@@plasmarade Hello, Honestly, I cannot speak to Intel's newest Arc drivers for Linux. I use old drivers downloaded from Intel's repositories. Further, it depends on what you want to do and which version of which distro you use. Kubuntu LTS is my preference. Kubuntu 22.04 LTS did not come with Arc's drivers; they needed to be installed manually. That meant I needed to use the CPU's iGPU to install Kubuntu 22.04 LTS , then use a terminal to install Arc's drivers, then physically install the graphics card, and eventually be a script kiddie and use a terminal to run the commands to install Intel's resources and Stable Diffusion. It was a pain, but necessary to use vladmandic /automatic 's fork of AUTOMATIC1111. I game on Windows, not Linux. That said, I also have an Arc A750 LE under Windows; it runs World of Tanks at a very solid 60fps/2560 by 1440 with all of the eye candy turned on. Intel Arc graphics cards are not supported by Kdenlive, my preferred video editor, for hardware rendering, video playback and whatnot. AV1 TH-cam video doesn't seem to have Arc support under Linux using Firefox, but Window's Edge browser will play hardware decoded AV1 TH-cam video. Finally, I think since Intel's drivers weren't a part of Kubuntu LTS, Discover -- the default system updater and software repository interface -- wants to upgrade some Intel compilers I installed, but cannot.
So, for me, there are some drawbacks, but no deal breakers. And since using Stable Diffusion for more than just image generation has become an interesting activity for me, a new in box 16GB/256bit bus graphics card for ~US$350 (from NewEgg) was a very happy accident. The card was purchased as soon after launch as I could manage and training an embedding uses a lot of memory. Serendipity and trying to future proof for the win. And finally, for day to day computing, such as writing this comment, I cannot tell an Arc card is in the machine.
@@Disty0 Thanks.
were openvino's dev team had a plan to expand the usage of SD-webui script to support inpainting checkpoint? At the moment we also have the problem to generate image using control-net reference because openvino script completely ignored control-net for now.
Hi, I am currently running Arch Linux with a 1070, and I primarily a gamer/game developer. I suppose I want to broaden my skills and learn AI. I am making a new system, looking at a full team blue stack with an intel CPU and GPU, I noticed the link is dead to the article. I just had the question of running Stable Diffusion with Intel GPUs with Linux, even if I get windows on it for dualboot, I will likely use WSL.
Especially interested in seeing this progress with the new Battle Mage GPUs. Even without AI, the progress of Intel Graphics looks great even on Linux!
How do I get it to work with my intel iris xe graphics card? I struggle using different models the only one works is 1.5, but i cannot do sdxl, juggernaut xl, and many others.
lost 4 days trying to make it work, huge waste of time.
How to turn on CUDA for pytorch?
Interesting. This is exactly what I'd expect A770 with 16Gb for VRAM to be used for. Decent performance too! On par with RTX 3060 12Gb as of right now. I'm kind of curious though, does it support RVC/Roop(Unlimited)/TTS Tortoise?
Why do you not have an installation tutorial yet? This is very annoying to try to get running. I keep downloading and re-downloading the SD.Next method and the A1111 and neither will output images when I click generate. I keep getting torch errors and all kinds of problems. I have no idea what I'm doing wrong. I got it to work one time but it was trying to use CPU and took 21 minutes to generate 1 image...You're trying to promote this software to consumers but you have to be a coding enthusiast to even get it working. I'm not familiar with GIT or PS commands. I've tried following a lot of different guides step by step and nothing is working like it's supposed to.
Installed as described, but my WebUI doesn't utilize GPU even when GPU & VINO script are selected.
Oh, no! That shouldn't happen. We suggest you reach out to fellow Intel users and experts in our community at intel.ly/4ah6pzh for further discussions. However, you can also join our Discord at intel.ly/3RHd9z2 for product news and updates.
@@IntelGraphics just installed default Automatic 1111 and it utilizes GPU up to 100% without OpenVINO and any additional settings... can you tell me why we need OpenVINO?
@@redbike5042 you installed the normal branch? because both branches for me still only use my CPU not my GPU
I met another problem I can't solve - OpenVINO SD only uses default 1.5 checkpoint.
All images look like generated by the same model regardless of what checkpoint I choose from the dropdown menu in the top left.
Am I missing something? How to fix this?
I found a workaround:
If I generate at least 1 image with CPU, all subsequent generations with GPU will work as intended (even if I change checkpoints) until SD restart.
But this ads extra 15 minutes to workflow, which is a lot.
You can add 3rd party checkpoints. In the video you can see that I use Realistic Vision from CivitAI. You can also use SDXL models from CivitAI.
Have you tried downloading these models, putting in the models folder then refreshing the checkpoints dropdown?
@@BobDuffy I'm not new to SD, so it's some OpenVINO SD issue. Currently I'm using SD Next which also works with intel gpus
@rain4675 glad its working. Fooocus and comfyui also work with Arc GPU.
But drivers not available easily for intel ARC , drivers availability problem its a big problem 😢
I'm getting error while generating images. I have A750 with i5 & 16gb RAM. Did i miss something? (I'm new to gen Ai space, trying to get started with SD)
Sorry to hear about the issue you are facing. May we know what is the application you are using to generate images? Please share the screenshot of the error by visiting intel.ly/3t2dMd8 We would love to discuss it further.
no information is clear. it is always vague. i am not sure if we will get results. we are going from video to video. someone please explain it properly.
Heyo! Love the stuff and have an A770 on the way right now, but is there any news on VR support? I'd love to really push this card to it's limits one day with DCS VR lol
I wouldn't hold your breath on it coming before BattleMage next year.
Are intel gpus good with premiere pro, anyone ?
Hi! Recently I bought Intel Arc A770 solely to run SD. I installed OpenVINO version of SD, but for some reason it doesn't see & doesn't use my GPU at all.
I have "Accelerate with OpenVINO" dropdown script & selected GPU in it. But I'm not sure what options should I choose.
During image generation my CPU load is about 70% while GPU is at 5%, so it's not CPU bottleneck. It's not GPU problem either cuz it works fine in games & other tasks. GPU drivers are up to date.
I don't have Resizable BAR enabled but I doubt that it's the problem.
Can't find any solution to this problem for hours now.
We're sorry to hear that you are facing difficulties. To check the instructions and configurations of OpenVINO on Intel GPUs, please visit the GitHub link via intel.ly/3wlfl7b For further assistance, you can post your query in our community forum at intel.ly/49dw3UB Intel OpenVINO Community Forum. Our team of product engineers and community experts will be able to provide more insights on this topic.
Upd.: I did cmd --> pip install --pre openvino
After that I updated pip as it was adviced in cmd
Now it works
@@rain4675 And how was the performance, could you tell us your experience, greetings
I feel like I have seen this before, during the ACM Symposium
Lol you did.
i think if intel gpus keep high vram in their gpus, they can be chosen more for ai or video editing because if you want to run high quality model you need vram instead of gpu speed so even if 3070 is faster, ultimately arc 770 can run higher quality models and it can run comfortably without out of the memory, stuttering and freezing, I think it can be a good choice especially for running llm locally, but they should do the same on the laptop, nvidia still offers 4 gb vram on laptops, which is a disgrace in this era, as if making fun of it.
Sir I have intel i5 12600k processor with liquid cooling and MSI Mag B660 motherboard with 16gb DDR4 ram....how can I run openvino for faster?..plzzz help.
To improve the performance of Intel OpenVINO in your system, we recommend that you initiate discussion and seek guidance at our community forums via intel.ly/3uTr9Na Our team of experts will assist you with an apt solution.
@@IntelGraphics ok sir thank you.
Any possibility to use openvino with deforum?
Hi! We appreciate your interest in knowing more about intel technologies. Please note that the best way to know more about the OpenVINO toolkit and its applications is by visiting our community forum. You can explore our forum via intel.ly/3NDhqRK for more queries and solutions about OpenVINO. Also, refer to intel.ly/3NEyHtA to know more about the OpenVINO toolkit. Happy learning!
Interesting!
Please Intel, Drop a installation video. I tried too many times and failed to set it up
Yes this is very annoying to try to get running. I keep downloading a redownloading the SD.Next method and the A1111 and neither will output images when I click generate. I have no idea what I'm doing wrong. I got it to work one time but it was trying to use CPU and took 21 minutes to generate 1 image...They're trying to promote this software to consumers but you have to be a coding enthusiast to even get it working.
@@SilverSparrow5 ikr, my 13400cpu is practically fried after 3 images. I found out pytorch needs to be CUDA enabled for gpu to use it in sdnext. But I can't even begin to debug this problem
@@SilverSparrow5 You gotta discord or anything? If we link up maybe we can get the bottom of this.
first. scary RAM usage 20ish GB RAM usage just for SD 1.5 inference is a really high, i bet openvino is still unusable at SDXL + HiRes. are you sure it isn’t RAM leaked problem with openvino? second. with 16GB of VRAM intel GPU with openvino should at least do SD 1.5 LoRA training/finetuning/dreambooth. Only use for inference is a quite waste of it’s VRAM.
Intel® PresentMon v0.6 up
Is there a new clue?
😄
I don't think because hastag isn't here
Interesting.
please have separate channels, one for gamers wanting to hear news about Arc\Intel,
and one for AI(and possible other) professional uses of ARC,
having videos for 2 very difference audiences in the same channel is very bad for discoverability
for example: I want to get a notification for anything gaming related, but not for AI
He forgot to metion 48G system RAM for compiling SDXL checkpoint with OPENVINO. OpenVINO boost basic applications but not with HiRes and RoLA.
I can follow up with SDXL. Show 1080x1080 images. Initial compile takes a bit. Those models are huge but similar to what you see here, each subsequent image is pretty speedy
@@BobDuffy That's the reason I plan to upgrade my ram to 64G and 14700K in the end of year. Openvino takes lots CPU resource to calculate pre-built and stores in A1111 cache, which slowly eats your disk space. Remeber to check and delete unnecessarily ones.
SDXL Training kill the card :(
Nvidia is too much better, easy availability of drivers and learn to easy, awareness in IT market,
yup this is a lot of hoops to jump through to do what a competitor easily does
Will Intel discontinue its graphics cards? News has come out that Intel will no longer manufacture video cards.
Source?
That has been a fake rumor that's been debunked countless times. Pat Gelsinger has wanted GPUs to be part of the future of Intel, and that vision started decades ago before he was CEO. Intel already has a public roadmap for the next 3 generations of GPUs. Intel is also committed to GPUs because they are now selling them to data centers and use the same architecture in their IGP. Their second generation consumer GPU, "Battlemage", is already shown up in leaks and is expected to launch early 2024.
@@__aceofspades Yh that's what I was wondering because I know Battlemage is right around the corner.
limited edition cards are discontinued only
@@__aceofspadesEverything else seemed to be accurate but that last part isn't accurate anymore, I think it's now more like 2nd half of 2024 than early 2024.