Enlarger seems fixed and improved for the first time in months = I TAKE IT BACK!

[Prezton]

As i suspected deep inside me the enhancer crap still doesn't work. It says finished whenever it want to and that's it. This time was a record, 37% before it was finished. i Guess the advanced AI engine felt there was no need to watch the remaining hour of the movie.
I have sent 2000 log and engine files too you dvdfab, regarding many issues and never gotten a thanx, well i can safely say that it feels good to wear out my machines and pay twice the electric bill amount to get incomplete movies.

I give up.

Write in the update log when your advanced ground breaking broken software works.

Nuff said, extremely disappointed

...

[Wilson.Wang]

Hi Preztion,

Sorry for the trouble.
We checked the QA feedback. we have been connect for several months. Thanks for your feedback.
I see you have many dvd want to Enlarge with AI.

I understand your problem now is DVDFab 11099 can ripper dvd wtih AI , but will end early , can not produce a full movie ？

the end early is maybe cause by dvdfab judge the reencode is freezed with some problem.This is our bug.

From the log you provide, we see your use DVD in driver, maybe this cause the bug. The driver not work good with long time AI running.
Would you like to use DVDFab clone iso, then use ISO to AI.

And we will test on our side, to check the stable of AI with long time conversion. and see we can reproduce the problem.

Yours,
Wilson

[Prezton]

i have tested several times (but to late for those logs i think, it was after i sent the latest) to burn a full movie copy, or a rip or an ISO, and tried from those sine i also thought what u prob think too. that the drive delivers soo little data that it , i dunno... goes into sleep mode or start dropping packages which ur enlarger interprets as the upscale is finished. because its no crash, no hang, no lagging... its a clear finished response and you can click the OK button and close as usual, so the software is not freezed. it has for some reason get into its head that task is done. but if u check my logs u can see that i have tried "poisoners handbook" several times, and several versions back manage to get a full copy of it. sure it took 70 hours which is a bit too much for a 1080 gtx even though its an old generaton. but now i cant ge "poisoners handbook" past 10%.all disks are set at performance mode and disabled sleep for them so the destination hdd discs going to sleep is not the case either. you had a working version a while back bu too slow, now you broke it by increasing speed. so i would check what code is different one by one and try to check when the software gets broken.

Also i wouldn't advertise 7 hours for a ful length dvd movie. This is proven impossible or you need a better machine than my main rig, and if thats the case i would mention that it takes a 3000 euro computer for the 7 hours experience. i would also mention that that time is with the lowest quality settings possible, its borderline to false advertising. but the speed has gone from a seemly capped 1.00 fps to 1.91 fps. so thats good news once it works again.

also i wonder why u cant use the call code that looks for and claims VRAM to expanded version where it if finding more than 1 VRAM source, claims both and distributes them to minimum 2 threads or rather cores since a 1 core system is more or less extinct. It also makes me wonder if it utilizes both GPUs on an SLI system. cuz when it worked my secondary crap machine gave 1.01 fps and my maxed out high end machine gave 1.03 fps. its like if a tractor and a Ferrari ran at same speed in a race. Also as last point it seems the grabber is a bit shy to claim VRAM. for some reason the SLI system was always barely used (the VRAM i mean). but the practically non existing fps difference stunned me.

EDIT: both ripper and copy to BluRay upscaling is broken

regards...

[Nighthawk66]

After around 40 Minutes the Redering Speed goes down to 1,1-1,2 fps - Total Time to finished shows me 26 Hours! So i Abort the Redering - and i have a RTX2080 Ti
But more as 26 Hours is to Long!
When i Ripped a UHD to BR it tokes only 20 min. That is perfekt. So it must something going wront with Enlarger AI

[Wilson.Wang]

i have tested several times (but to late for those logs i think, it was after i sent the latest) to burn a full movie copy, or a rip or an ISO, and tried from those sine i also thought what u prob think too. that the drive delivers soo little data that it , i dunno... goes into sleep mode or start dropping packages which ur enlarger interprets as the upscale is finished. because its no crash, no hang, no lagging... its a clear finished response and you can click the OK button and close as usual, so the software is not freezed. it has for some reason get into its head that task is done. but if u check my logs u can see that i have tried "poisoners handbook" several times, and several versions back manage to get a full copy of it. sure it took 70 hours which is a bit too much for a 1080 gtx even though its an old generaton. but now i cant ge "poisoners handbook" past 10%.all disks are set at performance mode and disabled sleep for them so the destination hdd discs going to sleep is not the case either. you had a working version a while back bu too slow, now you broke it by increasing speed. so i would check what code is different one by one and try to check when the software gets broken.

Also i wouldn't advertise 7 hours for a ful length dvd movie. This is proven impossible or you need a better machine than my main rig, and if thats the case i would mention that it takes a 3000 euro computer for the 7 hours experience. i would also mention that that time is with the lowest quality settings possible, its borderline to false advertising. but the speed has gone from a seemly capped 1.00 fps to 1.91 fps. so thats good news once it works again.

also i wonder why u cant use the call code that looks for and claims VRAM to expanded version where it if finding more than 1 VRAM source, claims both and distributes them to minimum 2 threads or rather cores since a 1 core system is more or less extinct. It also makes me wonder if it utilizes both GPUs on an SLI system. cuz when it worked my secondary crap machine gave 1.01 fps and my maxed out high end machine gave 1.03 fps. its like if a tractor and a Ferrari ran at same speed in a race. Also as last point it seems the grabber is a bit shy to claim VRAM. for some reason the SLI system was always barely used (the VRAM i mean). but the practically non existing fps difference stunned me.

EDIT: both ripper and copy to BluRay upscaling is broken

regards...

Hi Preztion,

I get your message.
And our group test ripper a DVD, converter full success with GTX1660ti, about 40 hours. And we will test other case.
MeanWhile, I was told to carefully not use too much other software, like open many tab in Chrome or other web browser.
It may use much GPU memory, cause ai enginner (tensorflow) fail.
Do you do some other work, while AI ripper ?

Yours,
Wilson

[Wilson.Wang]

After around 40 Minutes the Redering Speed goes down to 1,1-1,2 fps - Total Time to finished shows me 26 Hours! So i Abort the Redering - and i have a RTX2080 Ti
But more as 26 Hours is to Long!
When i Ripped a UHD to BR it tokes only 20 min. That is perfekt. So it must something going wront with Enlarger AI

hi Nighthawk66,

The AI speed slow is not DVDFab's limit,
The current hardware ability will use about 1 second for doing AI with one 1080p image on good Nvidia GPU, like RTX2080.
The speed for Intel GPU doing one 1080p image will be 5~10seconds. like HD630.

We are waiting for the new AI Hardware comming.

Yours,
Wilson

[Nighthawk66]

but in the Description a NVidia GTX1660 needs 7 Hours for 1 Hour Movie. So a DVD Movie with 2 Hours playtime should be finshed within 14 Hour +/-. But my 2080 tooks more then 26 Hours (GPU Load 97%).
I can´t believe that the 1660 renders faster then my 2080...

[Wilson.Wang]

but in the Description a NVidia GTX1660 needs 7 Hours for 1 Hour Movie. So a DVD Movie with 2 Hours playtime should be finshed within 14 Hour +/-. But my 2080 tooks more then 26 Hours (GPU Load 97%).
I can´t believe that the 1660 renders faster then my 2080...

1660 should be slower than 2080,
the Ultra Quality use AI layers much complate than Preminum Quality,
Ultra use time maybe 140%~ 200% compare with Preminum.

[gerbpaul]

Hello all. First time poster here. Been using DVDFab in various forms for quite some time though.

I was extremely excited about the potential of this AI product when I read about it. Wilson.Wang I think what Nighthawk66 is saying is that what DVDFab is writing in their promotional material is far from what you're suggesting in your post. I have noticed the same thing that many of the folks posting in this thread are suggesting as well. Bottom line is that the product isn't doing what DVDFab has indicated that it will do. I have done a ton of testing too and no matter what I do, I can't get much better than about 0.40 to 0.70 FPS on an AI Upscale Job.

That's fine if that is the expectation, but the materials that DVDFab is using to sell this product suggest something much greater. They don't outright say that you'll get a 4K upscale in 7 hours, but I just read the materials again, and they sure seem to suggest that it's possible. Problem is, it isn't even close, and the needle doesn't seem to move much at all regardless of how powerful your GPU is.

These days, how many people are trying to upscale their DVD quality videos to HD quality? That is probably a very niche use case. That is basically what you're suggesting to Nighthawk66, that we should glean from the promotion materials though. It doesn't say that anywhere however and when DVDFab themselves sell products that take care of all of our 1080P needs already and there aren't many use cases out there where people are trying to upscale their DVD quality content, why would that even enter our minds. It just doesn't make sense. All around, it screams of false advertisement, in order to sell a product. It's pretty disappointing when I've been a fan of DVDFab for quite some time.

Now, I have actually been patient and let several jobs complete, and I have actual completed several 4K upscale jobs. It's pretty cool that it can be done, though I really have no way of validating that it's actually doing anything other than my naked eye. I'm not even sure there is software out there that can actually tell you if you have more pixels in one image vs another. I've tried to find something but haven't really found anything. So with that in mind, not only am I not entirely certain that it's actually doing much, but it's taking forever to complete a single job that I can be uncertain about, and it cost me a cool $50 for it.

[gerbpaul]

Well, I have learned a little more about the product since I posted on Friday. Through more testing I've found that there is indeed a significant impact on performance of Enlarger AI when you are using ANY other applications while the conversion process is taking place.

A little info on my workstation and workflow. I run the application on a system that I don't connect to much as a user sitting in front of a screen. I actually use a remote desktop session to connect to the computer and I run the applications I want to run that way and manage the system that way.

The remote desktop software I use is actually a little intensive on the RAM on the system and I've noticed while benchmarking certain things that because it's a desktop, and I only use my dedicated GPU (NVidia) to output video, that applications that have any GPU memory requirements will use the VRAM on the dedicated GPU. So when I connect to the computer with a remote desktop session and let it be connected for a short bit, while I am running an Enlarger AI process, I witnessed the fps reduce by anywhere from .10 to a full 1.0 fps. That is a massive impact, when you're only getting 1.20 fps to begin with.

When I disconnect my remote desktop session, and let it run, then go back to it later I notice that the fps has ramped back up to the maximum level it will run at.

In some other testing, I wanted to see what kind of impact, overclocking my GPU could have on the Enlarger AI process. I am pleased to report that it had a very pleasant and I would say significant impact. It didn't necessarily take me from what would be a 60 hour process down to something like a 7 hour process, but I did manage to see fairly sizable increases in fps.

Before I started overclocking the GPU, so basically just using the GPU as it is configured out of the box, I could only get about right at 1.0 fps. After overclocking in several different configurations I've managed to get it up to as much as 1.70 fps. That is incredible! Though I'm not quite certain I understand how it would go up that much from the amount that you can typically do safe overclocking on a GPU.

Anyway, I thought I would share since I feel like I am starting to get a sense of what kinds of things seem to have an impact on the Enlarger AI process while it's running.

I should have my new RTX 2060 GPU in place by the end of this week and will continue testing with that. I am very interested in seeing how having an RTX GPU with Turing Architecture and a significant increase in CUDA cores impacts this processing. I'll keep you posted.

[Nighthawk66]

gerbpaul : Thanx for testing.
My RTX 2080 runs at around 1830 Mhz - Ram 6800 Mhz with 97% GPU Load and after 3 Hours redering the speed is 1.0 Fps - Thats to slow for me

[gerbpaul]

Yeah, you bet man. I’m fascinated about the potential of this solution so it’s fun to test and be able to provide hopefully some useful info.

An update. I got my 2060 today and the first job I kicked off is running consistently at about 1.86 - 1.90 fps.

My first job is about 3 hours in and is a little over 10%. So I can definitely see the improvement over the other GPUs I’ve been testing with.

I also ran several more tests after my last test using the 1060 6GB GPU and I’ve not been able to reproduce the 1.70ish number that I saw on my earlier test. I really haven’t been able to get anything above about 1.20 and most jobs seem to run at 1.10ish.

So that leads me to believe that other variables related to decoding/encoding have a decent level of impact on the speed at which these upscale jobs can run. So I’m going to start testing with the source files codec in consideration as I configure the destination encoding settings. See what that can do to fps.

I’m surprised that with your 2080 you are only getting 1.0. I‘m curious what type of file are you processing? The source file? H264? H265?

Paul

[gerbpaul]

After further testing this weekend, I have managed to get fps up to as much as 2.5 which resulted in a 2.5 hour video completing its upscale job in approximately 23 hours. That is not terribly far from what the marketing materials suggest. If I remember right, it said a one hour video in 7 hours, given the right hardware. At 2.5 fps it looks like we’re probably looking at about 7 hours and 40 minutes or so.

I also managed to get jobs on my other two GPUs to increase fps significantly. I was getting 1.9 on the 1060 6GB card, which is even higher than the max output I was seeing on that card previously.

It at least appears that any additional work that you need to be done as part of the AI upscale job, has a significant impact on the speed at which your job will run. The jobs that I was able to get to run at much faster speeds had minimal work being done, besides the task of upscaling. I chose a codec that matched the source file, I copied the audio, and I even dropped the quality a tenth of a point or two, going in between standard and fast.

So I will continue testing, but so far what I’ve been able to glean from all of this testing has been really useful.

Paul

[Nighthawk66]

Thanks för the Infos. You ask what Kind of Files i try to convert. I've uses the DVD Image from Lethal Weapon 1(German)
Maybe i should convert to h264 First!.?

regards

Marty

[Nighthawk66]

Update: i first convert the DVD to mkv 264. Change the Quality of A.I. to Standard Premium. After 1 Hour my Rendering Speed is around 6 Fps. Great. Full Movie in 7 Hours. That is Great. Good news