Rtx 3080 lower quality capacitor Issue

0h_no_not_again · 27 Sep 2020 at 02:17

Are people still banging on about the capacitors?

There's an issue with BIOS, silicon or the ENTIRE power delivery system. There's no way most partner AND the reference cards would suffer from the same issue if it was related to bypassing.

What is annoying is that people have made this huge leap in correlation without ANY evidence. Put a scope on the power rails of a card and watch the deviations at load steps. It's basic digital electronics.

Dunning Kruger on a mass scale.

Rroff · 27 Sep 2020 at 02:19

0h_no_not_again said:
What is annoying is that people have made this huge leap in correlation without ANY evidence. Put a scope on the power rails of a card and watch the deviations at load steps. It's basic digital electronics.

Amazes me no one has done this yet - though it does take a little electronics knowledge but it shouldn't be beyond some of the channels to get up to speed on.

MagicBoy · 27 Sep 2020 at 02:24

Nasher said:
The courts in the US disagreed and ordered Nvidia to compensate 970 buyers as it was miss-sold.

The last 512mb was useless as it completely tanked performance once you hit it.

I'm well aware of that. Changes nothing about my reply to someone else's post.

MagicBoy · 27 Sep 2020 at 02:26

Rroff said:
Amazes me no one has done this yet - though it does take a little electronics knowledge but it shouldn't be beyond some of the channels to get up to speed on.

The people with the knowledge and equipment don't have cards.

The people with the cards don't have knowledge and equipment.

Give them a few days, eh? There's a pandemic going on.

Purgatory · 27 Sep 2020 at 02:29

Grim5 said:
He did not

most reviewers failed to mention their cards were crashing, it's quite deceptive if you ask me

Agree, now they are all coming out with videos stating they had crashes and saw problems but said nothing, which is comical and then I'm sure some saw nothing and now click baiting people to their videos stating they saw problems .. It's all tragic what's happened with this release. :rolleyes:

Now I'm happy to wait. Even EVGA have been caught with their pants down and many of the so called respected AIBs.. They all tried to cheap out in some form and now going to pay for it in RMAs or recalls, but I know what they will do they will give out VBIOS updates that basically reduce the cards potential and make out all is ok. Basically you didn't buy what was described and paying more for something that has been/will be in future neutered by an updated VBIOS. :mad:

Rroff · 27 Sep 2020 at 02:33

0h_no_not_again said:
Dunning Kruger on a mass scale.

I only know enough to be dangerous, I am certainly not an expert - I even dropped out of electronics engineering though there are reason around that, but as someone who has actually designed and built these kind of circuits it is obvious most people talking about it actually have zero knowledge.

0h_no_not_again · 27 Sep 2020 at 02:33

Rroff said:
Amazes me no one has done this yet - though it does take a little electronics knowledge but it shouldn't be beyond some of the channels to get up to speed on.

It kind of annoys me. I design boards with some of the biggest silicon available (FPGAs generally), and this is the sort of stuff you don't "get wrong". You have validation routines that simultaneous switch logic and IOs in various edge and corner case patterns over days/weeks/months to check your power delivery is all good.

And yeah, the measurements are actually quite tricky to take. You need a 5+GHz scope and differential probe, probably costing £50k+ to do the job properly. Although, we're talking BULK capacitors here with responses peaking at about 1MHz, so probably a £2k of scope and probe would do it...

EDIT: I could guess at where the issues are and have a decent "gut feel" for such things, but even the best engineers on the planet need data to figure this stuff out. It will be WAY more complex than a few bulk caps.

Rroff · 27 Sep 2020 at 02:54

0h_no_not_again said:
And yeah, the measurements are actually quite tricky to take. You need a 5+GHz scope and differential probe, probably costing £50k+ to do the job properly. Although, we're talking BULK capacitors here with responses peaking at about 1MHz, so probably a £2k of scope and probe would do it...

/hides £100 USB scope.

Even channels like GN would probably have to do a collaboration with someone who is actually an experienced EE do a proper deep dive - I was thinking more general look at it which should be enough to show some symptoms of the problem or not with a cheaper setup.

ubersonic · 27 Sep 2020 at 03:40

0h_no_not_again said:
What is annoying is that people have made this huge leap in correlation without ANY evidence.

You mean aside from it being confirmed as an issue by EVGA yesterday.

Grim5 · 27 Sep 2020 at 03:59

ubersonic said:
You mean aside from it being confirmed as an issue by EVGA yesterday.

bingo

Rroff · 27 Sep 2020 at 04:00

Doesn't inspire me the statement is full of mentions of POSCAPs - not to say they've not internally tested with them but none of the configurations shown so far use them as far as I know.

Grim5 · 27 Sep 2020 at 04:23

Rroff said:
Doesn't inspire me the statement is full of mentions of POSCAPs - not to say they've not internally tested with them but none of the configurations shown so far use them as far as I know.

It's quite funny because as soon as Buildzoid put out that video we instantly knew that anyone who called them POSCAPS has no idea what they are talking about

Rroff · 27 Sep 2020 at 04:27

Grim5 said:
It's quite funny because as soon as Buildzoid put out that video we instantly knew that anyone who called them POSCAPS has no idea what they are talking about

Easy mistake to make TBH mixing up or not noticing which are which unless you are actually using them on a daily basis. What is more telling is people straight away calling POSCAPs cheap and low quality heh.

SniperSmithy · 27 Sep 2020 at 04:45

ubersonic said:
You mean aside from it being confirmed as an issue by EVGA yesterday.

Not all SPCAPs are equal. EVGA may well have found issues with the caps they were using, doesn't mean the other manufacturers had the same problem. Besides which, are we supposed to take a large corporation at their word? I mean think of the incredible spin potential. Cards delayed and customers angry -> Tell customers that actually you're behind because you're improving the quality unlike those other guys -> Customers love that you did this and are now happy to wait and also consider your product to be better than competition.

Not one EVGA shipped with the all SPCAP solution allegedly so if that's true there'll be no issues on those cards then right? Let's wait and see on that.

Hardware reviewers get pre production samples, with incomplete drivers/bios etc. They expect crashing, but they also know that it's not likely to be an issue in consumer cards and are usually told "we know these have some issues but they will be fixed before launch" and they go along with this. Whether this is morally right or not doesn't matter but that's why they come out after the fact saying x product crashed in testing for us and we didn't report it at the time of review.

Hardware unboxed have stated that the have had the crashing issue on an Asus TUF which features 0 SPCAPs. So no one can point at the caps and say it's definitely them.

beany_bot · 27 Sep 2020 at 07:01

0h_no_not_again said:
Are people still banging on about the capacitors?

There's an issue with BIOS, silicon or the ENTIRE power delivery system. There's no way most partner AND the reference cards would suffer from the same issue if it was related to bypassing.

What is annoying is that people have made this huge leap in correlation without ANY evidence. Put a scope on the power rails of a card and watch the deviations at load steps. It's basic digital electronics.

Dunning Kruger on a mass scale.

To add to this annoyance.

People keep saying "poscaps" ...they aren't poscaps.

And people keep saying they are lower quality. They aren't "lower quality" than MMCC. They serve a different purpose. They are still perfectly valid products. They hold more charge than an equivalent bank of MMCC, but equally, can't deliver it as quickly. Each have positives and negatives and typically work well in tandem.

Chris Ecob · 27 Sep 2020 at 07:53

beany_bot said:
To add to this annoyance.

People keep saying "poscaps" ...they aren't poscaps.

And people keep saying they are lower quality. They aren't "lower quality" than MMCC. They serve a different purpose. They are still perfectly valid products. They hold more charge than an equivalent bank of MMCC, but equally, can't deliver it as quickly. Each have positives and negatives and typically work well in tandem.

It's MLCC not MMCC ;-)

robfosters · 27 Sep 2020 at 08:01

I’m hoping this works in my favour and plenty of nervous Nora’s cancel their GB Gaming OC order in a fit of misguided panic, thereby pushing me up the queue

Troezar · 27 Sep 2020 at 08:17

robfosters said:
I’m hoping this works in my favour and plenty of nervous Nora’s cancel their GB Gaming OC order in a fit of misguided panic, thereby pushing me up the queue

Yes I'd sit on my order if I was a 3080 buyer. At worst you can cancel later or even return or RMA once received. Now is too early to react without hard facts.

0h_no_not_again · 27 Sep 2020 at 08:29

Rroff said:
/hides £100 USB scope.

Even channels like GN would probably have to do a collaboration with someone who is actually an experienced EE do a proper deep dive - I was thinking more general look at it which should be enough to show some symptoms of the problem or not with a cheaper setup.

Honestly, if you used a probe with the spring ground probe instead of the clip, it would probably work fine. That's kinda the joke

0h_no_not_again · 27 Sep 2020 at 08:32

beany_bot said:
To add to this annoyance.

People keep saying "poscaps" ...they aren't poscaps.

And people keep saying they are lower quality. They aren't "lower quality" than MMCC. They serve a different purpose. They are still perfectly valid products. They hold more charge than an equivalent bank of MMCC, but equally, can't deliver it as quickly. Each have positives and negatives and typically work well in tandem.

Indeed POSCAP is a name that Panasonic use AFAIK. Just a term for a Tantalum Polymer hybrid that lots of people make.

If it were up to be, I'd ditch all tantalums, but that's for environmental and ethical reasons.