jaredbyoung
I've heard the Earth is flat.
 
 

KingEngineRevUp
... you have no research or proof that it is the cause of any 3080 and 3080 Ti to fail. 

KingEngineRevUp

ty_ger07
I have no interest in arguing with you on about any of your other nonsense statements.

Lets get the record straight, you're not interested in backing up what you said. That's it. 
 
You said that buildzoid was proven wrong about the 3080 being fine. You tried to apply the 3090 failures to the 3080s with no evidence. That's where we are. 
 
I asked you a question about your claim and you can't prove it. You're failing to make an argument, there's a difference. 

Hefbn87

KingEngineRevUp
 
If the cards aren't failing and they're performing, then what else is there to complain about? The aesthetics?

Are you sure your high estimate of the reliability of 3080s is accurate though? I've seen a lot of reports of 3080 FTW3s dying. Back in winter when I was reading those reports, I was hoping my 3080 would be OK. Then it started dying a few days later. Recently I've seen some people on here describe their 3080s die after they worked seemingly without issue for 10 months. The situation does not look good for the entire 30 series.

kraade

KingEngineRevUp

The problem? Buildzoid is smart, but his audience is not as smart as he is so they hear things differently.

This ^^^^ It happens so much these days . 

Hoggle

jaredbyoung
I've heard the Earth is flat.
 
 

If the earth was flat cats would have knocked everything off the edge by now.

farmer_ta
I would say so, look at the Strix and Suprim models, they use more expensive voltage controllers. EVGAs FTW3 competes with these models but don't even use the same tier of components .

Nike_7688
The i2c lines are generally not exposed and would require a user to solder in a connector (see strix and Galax boards) and have a programming device capable of communicating with the controllers via i2c.

You keep saying it's cheaper without providing any evidence of that.

Nike_7688

ty_ger07
Interesting. You don't think that EVGA would make it accessible via EVGA's iCX microcontroller? I am not suggesting that EVGA would purposely make it accessible to users.  I am suggesting that if it were digital, EVGA may have had easier ways of correcting discovered issues instead of needing to resort to RMAs or board revisions like EVGA ultimately had to with its 3090s.

 
I'm not sure. They'd have to put the controllers on the same i2c bus as other devices. I feel like that adds unnecessary risk for most users, as it would be much easier for people to write "hacks" (or DIY tools) to directly communicate with and change settings on the VRMs. That could be...very bad. XOC folks do that and know that they could be damaging parts immediately and voiding warranties. You generally don't want these controllers to change settings regularly, you want to set them once and have them stable. So I don't think they'd want to leave that door open. That's why if you look at the Buildzoid Strix 3080 video, you can see they used the MP2888 digital controller in three places. They also have three unpopulated connector-header through-hole places where you could solder in (or in some cases push in a spring-retention connector, but you have to be careful as if it isn't secure and you try to hot-swap i2c, you just blew up your chip potentially) a header connector for reprogramming. That would be the standard practice for that sort of thing. In the factory, they probably have "bed-of-nails" test fixtures that check things out and high-quality pogo pins that make the temporary connections to program these. I certainly wouldn't want to leave your average user access to that as it would be an easy way to brick cards. Also, they have to consider that hackers may want to target them and could easily brick cards in that manner too. Maybe I'm over-thinking it (this happens, I've just been conjecturing and speaking hypothetically for most of this discussion as we don't have enough data/details), but it sounds like a terrible idea to me.
 

ty_ger07

You keep saying it's cheaper without providing any evidence of that.

I don't have any evidence either.  But, I can speculate.  Based on second-hand market, it appears that the up9511 is cheaper than the up9512, but who knows.  Things really depend on the supply chain and order size.

That's really not going to work to figure this out. We'd need distribution prices. There are many factors that impact second-hand market (availability, time on the market, vendor relations, etc.) I'm not going to make an argument one way or another based on what alibaba has that "fell of a truck" (for lack of a better joke).
 

ty_ger07
There is a thing which I have said many times which you are overlocking.  I mentioned the possibility of using inexpensive discreet components (in this thread and in many other threads).  You know what I mean, right?  [snip]  Perhaps NVIDIA provided reference options (they did!), EVGA chose the cheapest (seems plausible, but unproven), and then EVGA additionally chose the cheapest discreet components (which NVIDIA didn't expect when making design recommendations), and that is why the EVGA cards seem to have the most problems with black screens and blowing up compared to similar reference-design boards.

 
I do know what you're talking about and I didn't overlook it, but decided to (maybe in error) focus on what seemed to be the point you were driving, which was that analog = cheap and digital = expensive = better. I just wanted you to know that isn't true as a blanket statement...at all. It can be, but we don't actually know in this case. As for the discreet components, Some of the issues have been farther off than even tolerance deviations should allow. As I've mentioned a couple times, if the designers actually did their job right, tolerances would be specified in a way that no stack-up of tolerances all in one direction or another could cause problems. If I were to make bets, which is all we can really do here (or I could just say "perhaps" and spout off any theory whatsoever considering there's been no details released), I would bet they were given the performance specifications and requirements of the GPU chip(s). They designed around that and on paper everything looked great. Between board layouts and GPU performance characteristics, perhaps they missed the target somewhere along the way. Maybe it was a fringe condition where most of the time, and in all their tests, everything worked well within the limits and they thought it was great. Then some other apps came along and tested it a different way, made it perform an entirely different way and suddenly there are spikes or overloads they're not responding to appropriately (or at all) causing this issue.
 
Relevant story: I worked on revising an old circuit design for an HID bulb inverter once that should have worked fine on paper, but differences in the bulbs and a slight difference in the layer thicknesses of the PCB (not the copper, but the pre-preg thicknesses) caused the gate capacitance of the drive-FET to increase to levels where every now and then it just blew itself up because that caused too much inrush current. There were minimal issues for years, but then the board vendor had more of one size core/pre-preg than another one batch and suddenly half the inverters are blowing. This story is only here to show that sometimes unforeseen performance variations can cause major problems (also, I can promise you, no cheap components were used in that design, it was for aerospace).
 
They clearly found something wrong with the 0.1 rev 3090's (or maybe it was a couple problems? three? twelve? I don't know). I see you responded in your note that they removed the up9511 controller. Has anyone confirmed that? I just haven't seen any pictures to confirm it. If anybody does a hybrid or water-block conversion on a 1.0 board, it should be easy to get a picture and check.
 

ty_ger07
There certainly must be some reason why EVGA cards had a very high failure rate initially (especially the 3090s), EVGA had to RMA a bunch, EVGA created a special RMA process for 3090s, EVGA has released firmware updates to fix performance problems related to power balance, different cards have performed wildly different in relation to power (and thus performance) limits, etcetera, etcetera, etcetera.  There must be a reason, right?

I agree there's at least one reason. There might be several. I just haven't seen any evidence yet that it has anything to do with the cost of components. The first revision of parts were absolutely rushed out because the AIB's didn't get much time with GPUs in hand at all before launch (thanks Nvidia). This was evident with the ceramic/POScap issue and all the early firmware/vBIOS updates that had to happen across all vendors. I think they had to take some guesses on performance characteristics and some did better than others. Just to be clear, this is in no way making excuses for anybody. I'd rather something launch late and without issues than "on time" and be an RMA nightmare. Either way, the blame lies with Nvidia and EVGA for EVGA's part failures. It may be possible that the designs were picked over more complex ones due to cost (I think more of a programming-time cost than parts cost to be clear), but I don't think they would have done that intentionally if they knew there'd be issues later. The most likely cause for this from my perspective an experience would be rushing the design to meet deadlines. With the UP9511 being out since the 1080 days, they would have been very familiar with it and likely very confident in it's abilities based on the specifications Nvidia gave them. One thing that surprised the public with the 3000-series launch was how much more power and how much larger the current spikes were on this gen than any before it. Maybe that surprised the board-designers too. Maybe they weren't given adequate time to actually go through the massive variety of workloads these cards see. All we have are "maybe" and "perhaps" for now. I am not infallible and as such, I could be wrong, but I'm not buying that "perhaps they just bought the cheapest resistors and capacitors in the world and the tolerances were crap so a bunch of cards failed" theory (<this is paraphrasing, not meant to be an actual quote).
 
Edit: side-story needed context.

Mscott0566
Is it true that EVGA used cheaper parts to get the minimum requirements met for the 30 series cards? I still love EVGA but I seriously hope not…

B0baganoosh
Does anybody happen to know if they changed the uP9511 out to something else on the rev 1.0 3090's? The 3080 ti has a different regulator (NCP81610 from ON Semi).