NVIDIA RTX 50 SERIES - Technical/General Discussion

[DavieJG]

Bit of an assumption there, but it’s a possibility, although it didn’t happen…

You give any electrical part 2.5x it's rated current and it will thermally fail, not if, just when.

 

[Trox]

That's all that happens. Throughout this whole saga, no one's GPU has "blown up" as you put it

I don't recall any 40 and 50 series cards on fire or exploding

Then why imply it?

 

[Trox]

You give any electrical part 2.5x its rated current and it will thermally fail, not if, just when.

No need to tell me how electrics work

I’m also not defending this whole cable saga

 

[Absolutely Clueless]

https://www.reddit.com/r/pcmasterrace/comments/1iqtw4r/l_used_all_my_luck_to_get_the_5090fe_now_it_wants/

Another one....

At this point given how many are melting considering the tiny number out, I experct to seee WAY more after next month once these magical stock supplies ramp up..


570W from pins this thin lmao. Good luck to all who bought or will buy a 5090.

To make sure I understand, his Lian Li cable melted, but he is also buying a new PSU to replace his Corsair one? So a Lian Li cable + Corsair PSU. Not saying that is the reason for the melting, but it's not good to mix and match.

 

[Collected1]

That's all that happens. Throughout this whole saga, no one's GPU has "blown up" as you put it

I don't recall any 40 and 50 series cards on fire or exploding

That was overly dramatic language on my part. My mistake.

 

[Absolutely Clueless]

Have all the 5090 connector burns been with the FE so far? Does anyone know any AIB cards burnt?

 

[Rogueywon]

Have all the 5090 connector burns been with the FE so far? Does anyone know any AIB cards burnt?

Going off the reddit megathread, which is about as much as we have right now, there are two confirmed FE 5090s. On top of that there are two more FE 5090s which genuinely seem to be user error (very poor cabling practices, risky under any circumstances). There is one confirmed Asus 5080 Astral, which is weird and currently the only non-FE (and non-5090).

Don't read much into this, though. Sample sizes are incredibly small and there is just not enough evidence right now to conclude "FEs melty, AIBs safe". FEs have been a very significant chunk of the 5090 supply so far - there certainly seem to be more of them around than any one AIB model and possibly almost as many as the AIBs combined, due to that "communication issue" between Nvidia and AIBs.. They're all putting a similar level of power through that cable and we'll see whether other models are affected once more of them flow through into circulation.

 

[Nitefly]

Unfortunately the nature of this being multi-component just makes nobody take responsibility.

- GPU maker: make sure you power it properly…! Your PSU maker will tell you what you can use, as it’s their job to make it supply power safely. We’ve given you an adaptor, you can use that. Oh you didn’t use the adaptor we gave you? That’s your fault for not powering it properly.

- PSU maker: don’t use it to improperly power something. Oh, it melted? Was it drawing more than 600 watts? What does that cable say? Why did you improperly power something?

 

[gpuerrilla]

 

[Joxeon]

Yeh i see your point. The 4xxx wasnt great either historically, but i still think the 5xxx series is even worse.

Prices have gone a bit silly after the 3xxx series (and people even thought those RRPs were high at the time!) generally which is very disappointing.

Back during 30 series crypto mining we had the 3090 at £2.5k the 3080 at £1600, 3070 at £1200 and 3060ti at £1000

Roll on 4 years we now have a 5090 at 2.5k, a 5080 that is really a 70 class at £1200+ and a 5070ti which is a 60ti class for soon to be £1000 so Nvidia must be absolutely delighted that they have been able to normalise crypto pricing without crypto mining being relevant anymore.

 

[Absolutely Clueless]

Going off the reddit megathread, which is about as much as we have right now, there are two confirmed FE 5090s. On top of that there are two more FE 5090s which genuinely seem to be user error (very poor cabling practices, risky under any circumstances). There is one confirmed Asus 5080 Astral, which is weird and currently the only non-FE (and non-5090).

Don't read much into this, though. Sample sizes are incredibly small and there is just not enough evidence right now to conclude "FEs melty, AIBs safe". FEs have been a very significant chunk of the 5090 supply so far - there certainly seem to be more of them around than any one AIB model and possibly almost as many as the AIBs combined, due to that "communication issue" between Nvidia and AIBs.. They're all putting a similar level of power through that cable and we'll see whether other models are affected once more of them flow through into circulation.

Thanks for the summary. I forgot about the reddit megathread. I agree regarding FE vs AIB, but I do think many more FEs are in customer's hands at this point.

 

[Tinek]

TBH I'm kind of surprised there's not been much (any?) talk about a product recall, either instigated by Nvidia themselves or forced on the by whatever government body deals with that sort of thing.

It's all user error, why would they even think about a recall?

 

[Absolutely Clueless]

It's all user error, why would they even think about a recall?

Whilst you are in this thread, Jensen, have you got a spare 5090 for me?

 

[Tinek]

I guess the reviewers are only running relatively short benchmarks in open air rigs.

That's exactly what Derbauer said too - he's been running these tests for just minutes. If it was an hour+ of intensive gaming, you'd see melting. And with higher power use we'll see way more of these cases with 5090 than we've ever seen with 4090.

 

[TNA]

3080 was the last card Nvidia made that didn't burn your house or wallet down.

Best gpu I had. Got the 3080 FE under £600 near release and sold it for £1620 to CEX around 8 months later

 

[CornyOC]

 

[CornyOC]

The reported cases yes. But there's no electrical reason it can't happen on any card. Construction of the cable seems to be the biggest variable/culprit outside of the fundamentally flawed design of course.

Agreed, and it makes sensxe anyway, since the vast majority of sold and delivered 5080s and 5090s are FE. So far it's probably just propability since there are more FE card being used than AIBs.

 

[Murphy]

It's all user error, why would they even think about a recall?

I know that was sarcasm, and it's not like we haven’t already seen plenty of finger pointing at various culprits, but i do find it concerning how a company can sell a device with such a badly designed power delivery circuitry it seems to be more a case of when, not if, it's going to cause a fire.

 

[Curtis vincent]

Anyone heard any rumours or talk of when the cards with waterblock pre installed may drop? Like Aorus or inno3d models

 

[SeeNoWeevil]

I know that was sarcasm, and it's not like we haven’t already seen plenty of finger pointing at various culprits, but i do find it concerning how a company can sell a device with such a badly designed power delivery circuitry it seems to be more a case of when, not if, it's going to cause a fire.

Yeah this is the route of the problem. It's quite clear now that anyone with a financial stake in 12VHPWR cannot be trusted and some regulatory body should be stepping in. If you think about it, anyone with enough electrical engineering knowledge to produce some product that uses the 12VHPWR standard, has enough engineering knowledge to spot straight away that the spec is bad and potentially dangerous. This includes AIB vendors, cable manufacturers, PSU manufacturers. They all probably knew and yet they all marched ahead releasing their products and taking money off people. He's deleted his Reddit now but I'm pretty sure the Corsair head of R&D has said several times how he 'doesn't like' the spec. Obviously not going as far as to say he thinks it's dangerous because then questions would be asked why they're selling a product using a dangerous standard.

It might sound a bit uncharitable to lump PSU vendors in with this assuming it's not their job to regulate the power delivery of attached devices. Well, if that's the expectation then that needs to be defined in the spec... a spec produced by PCI SIG that you are a part of!! Also, Nvidia haven't put power balancing/monitoring on their cards since the beginning of the 40 series now so why are you expecting it for the 50 series? I bought a new Corsair ATX3.1 PSU as I trusted 12V-2x6 would solve this issue. My old PSU was perfectly fine. Corsair made money off me because 12VHPWR is bad. Cable manufacturers made money off this too, look at the confusing nonsense statement MODDIY put out, trying to make people upgrade to their new '12V-2x6' cable. Cablemod released this absolutely ridiculous set of images warning users not to bend 12VHPWR within 35mm of the plug and just somehow magically float the cable straight out. Er, if it shouldn't be bent within 35mm, put a 35mm rigid sleeve at the base? Also, who the hell has 50mm+ of clearance between these bigass AIB GPUs and their case side??

Then you have youtubers, most of them won't flat out say "Don't buy this thing" and just dance around the issue. Gamers, who instead of choosing to not buy a 5090 will just buy it anyway and then go out and buy a current clamp or a frigging thermal camera to monitor their cables. I genuinely saw people on the Nvidia subreddit discussing this. Retailers will all continue selling it.

The only people coming out of this looking good are the likes of Buildzoid and Der8auer.