OG_Publicspanking:

Brown nose much.....

Well, nvidia DO have a huge amount of tech people doing this every single day - I think its safe to say that they know what they're talking about when recomending a 750W PSU. The fact that it's neccesary is a another subject.

Peaks from high spikes or transients could hit a lower wattage PSU although if it's short enough and the PSU and cables can handle it then it should be less of a problem but I don't really know how the GPU's look like if you were to push the wattage and power draw to where it could sustain a ~2 Ghz GPU clock speed for a longer sustained period of time. NVIDIA has 1750 I think and review samples with the Founder Edition cards also being binned did around 1900 in boost clocks but some of the third party designs could aim for a somewhat higher power draw and that would then be possible to extend further with GeForce Experience even if it caps out at a lower 15% or so. Think that's what TechPowerUp did for the 3090 in some cases so it was less power throttled and the resulting up to 19% performance lead instead of a 8 - 10% gain on the 3080 🙂 Binning and what the voltage scaling on the GPU then looks like and total power draw so 350W transients around 2x so 700W momentarily and quick millisecond shifts in fluctuation between low load and high load cycles for the PSU to handle. EDIT: I need to read up on this a bit more though, AMD had this little issue with Vega going into 600W and then the ampere also at a high total for these transients but NVIDIA's had a more efficient design for Pascal and Turing. AMD matched that of sorts with Navi though the default settings push this card pretty hard and some of it's third party custom models go even harder on the power delivery so it loses most of that efficiency. (Spikes upwards of 350 - 400W in any case, bit higher for custom settings of course including Wattman.) EDIT: https://www.igorslab.de/en/nvidia-geforce-rtx-3080-founders-edition-review-a-great-step-ahead-and-the-gravestone-for-turing/12/ Short peaks at around (but a bit lower than) 500W so quite a bit better than my assumption. 🙂 https://www.igorslab.de/wp-content/uploads/2020/09/04-Peak-Power.png But how would this shift with a higher power draw limit and a higher sustained clock speed. Although if it hit the PSU like that then over current protection and other measures (There's a few of these on modern power supplies.) would be a system shut down not just a crash at least in the worst case. EDIT: And then a lower quality less efficient PSU maybe a bit older so not quite peak performance as another factor.

OG_Publicspanking:

Oh, I get it the double standards, if I deviate the topic a little I get scolded but if someone with 1k post does it's perfectly fine. Never said that buddy. durrp . I was just adding that people should be really talking about the power supply upgrade that is needed about the card as well. The reason the industry is crap because of ppl like above.

The number of posts are also irellevant, as long as the tone of the matter is ok. I have 1300 posts, but have been around this forum for 16 years - so not new to the game. Who scolded you? Deviating from a topic is fine as long as its done the right way and not lashing at people (not saying you did that).

The crashing to me seems like the card becoming unstable at higher clocks, perhaps Ampere just doesn't like running at over 2Ghz for whatever reason. The PSU could be a problem if someone is using a pig tail 2x6+2 pin and not using separate cables but looking around I'd say the cards are not liking running at high clocks.

Isnt it the AIB fault for pushing these to unstable clocks? They should've known cards cant reach 2ghz without crashing. FE is cooler and more stable, fail.

Undying:

Isnt it the AIB fault for pushing these to unstable clocks? They should've known cards cant reach 2ghz without crashing. FE is cooler and more stable, fail.

My understanding of the situation is AIBs didn't have a lot of time to get their cards out, perhaps they were rushed and GPU boost 3.0 being what it is the card will try and clock higher given the thermal and power headroom? I briefly had a MSI 2070S that was more annoying to work with because of that, and would crash because it was trying to boost at over 2100Mhz. BIOS updates will likely fix these problems.

Fox2232:

Yep, I called it. A lot of people with poor cooling and/or weak power delivery.

Yep, noobs tend to have more problems than others. People with no cards, tend to have even more problems with 3080/3090 😀

nizzen:

People with no cards, tend to have even more problems with 3080/3090 😀

https://media2.giphy.com/media/8mtmIaIbGseogt3sWG/giphy.gif

"I can still run this with my 500W PSU. No big deal. See! Windows 10 will still boot up. Upgrade is done."

OG_Publicspanking:

Bad drivers don't often cause games to crash tho , its more likely that the games run poorly on bad drivers. I also think that ppl seem to think it's ok to blame the psu which is ridiculous as a well-branded 650W should be enough for it, nobody should be asked to buy a new PSU just for this card.

This is utter nonsense. A single logical error in the driver code can result in a hard crash easy.

Instability appears to be caused by some AIBs not using smaller caps to filter high freqs generated at higher boost clocks from the GPU supply voltages. Asus TUF and NVidia FE both use smaller caps and have no stability issues. Zotacs testing has shown replacing the single cap with multiple small ones solves it too. https://www.igorslab.de/en/what-real-what-can-be-investigative-within-the-crashes-and-instabilities-of-the-force-rtx-3080-andrtx-3090/

Richard Nutman:

This is utter nonsense. A single logical error in the driver code can result in a hard crash easy.

and that would be reproduced across the board. not only on boards that are clocked over 2ghz

Mufflore:

Instability appears to be caused by some AIBs not using smaller caps to filter high freqs generated at higher boost clocks from the GPU supply voltages. Asus TUF and NVidia FE both use smaller caps and have no stability issues. Zotacs testing has shown replacing the single cap with multiple small ones solves it too. https://www.igorslab.de/en/what-real-what-can-be-investigative-within-the-crashes-and-instabilities-of-the-force-rtx-3080-andrtx-3090/

ROTFLMAO GG AIB's.

Even if it only effects the AIB boards it's still Nvidia's problem. The reference boards should account for this and Nvidia should do it's own validation on partner designs. The vast majority of customers are just going to see "3080 Crashing" and not care who makes it. Makes Nvidia look bad (if this is even the issue and it isn't rectifiable by drivers).

I'd say Nvidia's hands are washed of blame by having supplied the reference schematics as such https://www.igorslab.de/wp-content/uploads/2020/09/Bottom-POSCAP-vs-MLCC.jpg Its pretty obvious to me that atleast one of the blocks have to use the smaller caps for inline filtering.

Just saw JayzTwoCents video about it. Was panicking until i found out Gaming X trio use the same setup as founder 3090. Relieved, felt like i dodged a bullet here. If Nvidia sets the AIB up for failure after going after their market, i hope it comes back to them big time.

Astyanax:

I'd say Nvidia's hands are washed of blame by having supplied the reference schematics as such https://www.igorslab.de/wp-content/uploads/2020/09/Bottom-POSCAP-vs-MLCC.jpg Its pretty obvious to me that atleast one of the blocks have to use the smaller caps for inline filtering.

FE and the best AIB versions of the card uses two set of those small caps though. It's still unclear whatever one set is enough to be perfectly stable. Maybe you need two. If you need two set then it's still on nVidia.

jura11:

Zotac and few others AIB partners literally cheapen on capacitors by using POSCAP and not enough MLCC capacitors Zotac always trying to cheapen up on this, their GPU always OC way worse than Asus or EVGA Hope this helps Thanks, Jura

Its not only zotac but asus, evga, gigabyte and colorful also they all did cheap out. From what i saw only msi trio did falow the fe spec.

The asus tuf 3080 uses all 6 MLCC, seen it on the JayZ vid

I wouldn't call it cheap out, that really gives bad light to AIBs, if Nvidia told them it was the specs and it end up undelivering, Nvidia sets them up for failure.

Well this takes the shine off what was otherwise an extremely successful launch 🙂 Makes me wonder whether I should cancel my preorder. Luckily I have plenty of time to decide!