3080/3090 Cap Drama - Deeper Investigation

[TheLaggard]

der8auer recently posted a video about the Nvidia 3080/3090 capacitor configuration. I really admire his technical prowess, and I was blown away at his video showing how process nodes can have the same naming for two companies, but it translate differently in practical application.

 

The problem is, he was obviously out of field of expertise when dealing with the capacitor subject. He had a good attempt at soldering the MLCCs, but I cringed at some moments in the process where he could've easily ripped a pad off or done something worse. That aside, his approach is very wrong, as you can't plop MLCCs like that and expect a better boost performance. There's a specialty in electronics/power electronics engineering to fine tune that configuration. It's more evident that he's missing a lot of knowledge in the field as he showed previous examples of capacitor banks connected to other cards to improve performance. However, even that design and configuration has many flaws and obvious mistakes (long lead wires, use of electrolytic capacitors with high ESR, etc....)

 

My question is, how much are you interested in a video that explores in depth the capacitor design choices, properly modding a card with appropriate MLCCs, and possibly breaking overclocking records using that method? This is something in the field of my profession, and I believe I can shed some light. That's why I'm thinking that this would be a very, very entertaining, albeit risky, project for me to take on.

 

I do not intend on sharing too much information regarding, but I'm happy to open a discussion about this subject. I'll be happy to answer any question/point as much as I could. If there's you have question or comment to the current investigation done by anyone, let me know.

[Dedayog]

I think it would be cool to watch a video and learn, but to give money...  no thank you.

 

You had my interest right up til that point.  Decently written up but was a big "ahh" moment when you added the donation basket line

 

 

[TheLaggard]

Understandable, and I'm with you. It is something I want to do regardless of people willing to chip on or not. If someone interested in helping, sure, if not at least for making the video it be nice to know if someone would like to watch it rather than going through the effort of making a video no one wants to watch. Working on project AND filming makes things 2x slower.

 

Anything in particular you want to touch on or learn?

[Dedayog]

2 minutes ago, TheLaggard said:

Understandable, and I'm with you. It is something I want to do regardless of people willing to chip on or not. If someone interested in helping, sure, if not at least for making the video it be nice to know if someone would like to watch it rather than going through the effort of making a video no one wants to watch. Working on project AND filming makes things 2x slower.

 

Anything in particular you want to touch on or learn?

I understand the basics of how capacitors work, but not their role in a high end GPU or the requirements of their specs.

 

Wouldn't have to be a long video, a couple minutes to cover basics and then direct application.

[TheLaggard]

Just now, Dedayog said:

I understand the basics of how capacitors work, but not their role in a high end GPU or the requirements of their specs.

 

Wouldn't have to be a long video, a couple minutes to cover basics and then direct application.

So, would just a theoretical video without any practical application be interesting enough for you to watch?

[piratemonkey]

3 minutes ago, TheLaggard said:

Understandable, and I'm with you. It is something I want to do regardless of people willing to chip on or not. If someone interested in helping, sure, if not at least for making the video it be nice to know if someone would like to watch it rather than going through the effort of making a video no one wants to watch. Working on project AND filming makes things 2x slower.

 

Anything in particular you want to touch on or learn?

Make sure to quote people using the back arrow on the bottom of a post when you respond to them. Otherwise they won't know you responded to them.

 

The way I see it is if you want to do something as a filmmaker/youtuber/whatever other title, you do it with either no expectation or very little expectation of money or other financial compensation. If you want to investigate or look into something, do it out of interest. that's the only way the product will end up good (the product being the video). 

[TheLaggard]

1 minute ago, piratemonkey said:

Make sure to quote people using the back arrow on the bottom of a post when you respond to them. Otherwise they won't know you responded to them.

 

The way I see it is if you want to do something as a filmmaker/youtuber/whatever other title, you do it with either no expectation or very little expectation of money or other financial compensation. If you want to investigate or look into something, do it out of interest. that's the only way the product will end up good (the product being the video). 

That's the problem. I'm not a Youtuber or anything, nor am I planning to be one. There's isn't really any incentive for me to make a video or a bigger project out of it.

Many, if not most Kickstarters projects are media project where someone is making a video about a subject of interest. I've also seen similar posts to mine bearing fruit to something bigger.

 

But all of that is secondary to my main point, which is discussion about the subject. Someone told me it was an open and shut case and that the community no longer cares about it. But I would like to hear from more people.

[piratemonkey]

1 hour ago, TheLaggard said:

That's the problem. I'm not a Youtuber or anything, nor am I planning to be one. There's isn't really any incentive for me to make a video or a bigger project out of it.

Many, if not most Kickstarters projects are media project where someone is making a video about a subject of interest. I've also seen similar posts to mine bearing fruit to something bigger.

 

But all of that is secondary to my main point, which is discussion about the subject. Someone told me it was an open and shut case and that the community no longer cares about it. But I would like to hear from more people.

The only way I can see a reason for a Kickstarter is if you need equipment you don't already have. If you have all you need to do what you want, go ahead. You could find someway to advertise, be it here (talk to mods first) or somewhere else. Besides that, I don't see a huge opportunity to either do a Kickstarter or Patreon. 

You can also just do a written post on here then repost on other forums if that's what you're into

[LogicalDrm]

@TheLaggard

I've performed some moderative actions on your thread. Mainly fixed the formatting (black text doesn't really suit our Night Theme) and removed some content which is against our Community Standards. Discussion about this topic can be continued within our set rules:

Quote

No begging for free or discounted things, asking for donations, GoFundMes, or any similar content.

 

[TheLaggard]

5 minutes ago, piratemonkey said:

The only way I can see a reason for a Kickstarter is if you need equipment you don't already have. If you have all you need to do what you want, go ahead. You could find someway to advertise, be it here (talk to mods first) or somewhere else. Besides that, I don't see a huge opportunity to either do a Kickstarter or Patreon. 

You can also just do a written post on here then repost on other forums if that's what you're into

To proper investigation, it would take more expensive equipment than I have.

 

Thank you for the suggestions, but in text it might be too technical and hard to understand. It'll be probably easier for everyone to just setup a record screen and demonstrate the performance with a couple of simulations. As per Dedayog's suggestion, I'll make a video that touches on the subject theoretically. Start simple, you know? I'll see if I'm allowed to post the video here in the forum, and if so we'll see how people react. If they're interested, I can continue from there, if it dies, it dies.

[TheLaggard]

5 minutes ago, LogicalDrm said:

@TheLaggard

I've performed some moderative actions on your thread. Mainly fixed the formatting (black text doesn't really suit our Night Theme) and removed some content which is against our Community Standards. Discussion about this topic can be continued within our set rules:

 

There was no begging or asking for any sort of donations, and nothing like that was advertised or linked. There wasn't any name for any project or page mentioned. There is even any funding page or project related to that subject, as far as I know.

It was a question of "How interested is community in the subject. Is it interesting enough that they'd be willing to contribute financially to it"?

 

Anyway, it's not important to the post, so whether it's removed it doesn't change the subject.

[Mark Kaine]

I'd be super interested to see that.

 

I always shudder when I see those badly aligned capacitors on"fixed" 30xx cards... lol.

 

it's clear the people doing this don't really know how to solder, don't have the right equipment etc .

 

yes, a slanted , badly soldered SMD *can* work as long there's electrical contact, but holy shit ...

 

 

[spartaman64]

buildzoid said it would be interesting to see oscilloscope comparisons between spcap and mlcc

[TheLaggard]

8 hours ago, spartaman64 said:

buildzoid said it would be interesting to see oscilloscope comparisons between spcap and mlcc

BZ's is probably the video I have many issues with. One of those is the "how easy" just stabbing it with an oscilloscope and monitoring is.

 

It's not that easy because it will take a very, very high sample rate oscilloscope with appropriate high-frequency probes. At 2 GHz clock speed, the rise time is in picoseconds, and so should be any dips. To be able to see the square wave somewhat reasonably, you'd need an oscilloscope that can sample the at least up to the 3rd harmonic of of the 2 GHz clock, in other words you need at least a 6 GHz oscilloscope, which costs around $15K. Even then, the blue trace is all what you'd get.

 

You can see how much information is lost, and the rise time is very slow compared to the actual signal (in green). I'm not sure how many harmonics you need to achieve "acceptable" readings. To get something like the blue trace in the following graph, you need to sample at the 9th harmonic, so you'd need an 18 GHz, or more, oscilloscope.

 

Would you to guess how much that class of an oscilloscope costs? Just a cool $175K. It's close to the kind of test equipment needed to analyze HDMI and other high-bandwidth signals.

 

To be clear, the green line is not what you'd get when measuring across the capacitors, but it serves as the fundamental signal and frequency that will cause voltage dips that may, or may not, be causing the issue. Being able to sample the square wave will means you can reasonable sample the dips cause by it. This is all, of course with plenty of oversimplifications, but it should help to get the idea across. Is it simple to measure? Yes, if you have the equipment to play with. Then again, I've seen Youtubers play with fancier machines, so you never know how feasible it could be.