NVIDIA RTX 50 SERIES - Technical/General Discussion

[hobib_ali]

OMD now they have spotted 5080s with missing ROPS

So it seems that all the models have been affected, Its stated that only 0.5% have been affected, question is how many owners have been checking, if everyone checked world over then I'm sure its way more than 0.5%

This will add to the delays bigtime, mass returns and recalls, ROP checking from the factory, add an extra 60 days, so will be 4 months till things get back to normal, thats June/July

what a faff of a launch

 

[Kelt]

Imagine thinking a bunch of nerds wouldn't notice this or would notice it and not care about 5% drop in performance. Have you... met pc gamers? Ever?

Much, much cheaper for Nvidia to get buyers who bother to check, to return them and get a replacement/refund, rather than recall the entire series.

 

[Joxeon]

And you have some in the RDNA4 upset at why shouldn't AMD be charging £700+ for the 9070XT.

Bizarre or what

What's the betting AMD tries to make a joke about ROPS during their reveal.

Much, much cheaper for Nvidia to get buyers who bother to check, to return them and get a replacement/refund, rather than recall the entire series.

The majority will never bother to check or even would be unlikely to notice missing ROPS anyway.

 

[TNA]

What's the betting AMD tries to make a joke about ROPS during their reveal.

Haha

Overclockers dream baby!

 

[mahius]

My guess is they knew. But they also know how strong their mindshare is. Those who notice it will get a replacement. My guess is most won't notice.

Their mindshare has been strong since at least 970 GTX days. People were enraged at the 3.5gb issue. What did said people do? Get a refund, add more money on top and get a GTX 980

I still have my 970, back then I just tried to make sure VRAM usage was kept to 3.5GB and below
Admittedly, I waited for AMD to launch their 300 series before buying, then bought a 970 since it performed a little better in the games I played.
Even with the 3.5GB issue, Maxwell was a peak Nvidia generation. The GPUs were priced so low I got spoiled and thought everything that came after it was too expensive.

This time, there is no redeeming factor for the 5000 series. One can forgive a problem or two on a cheaper product.
But having problems this bad with the GPUs themselves AND priced terribly even at MSRP AND low performance gains AND stock issues AND scalpers!?!

 

[Gerard]

What's the betting AMD tries to make a joke about ROPS during their reveal.

Couldnt blame them, how does something like that slip past?

 

[Raiden85]

May as well just scrap the entire 5000 series at this point and Nvidia just need to take the loss, it's an absolute disaster.

 

[Lyle44]

Imagine thinking a bunch of nerds wouldn't notice this or would notice it and not care about 5% drop in performance. Have you... met pc gamers? Ever?

You say this, yet pages ago we had posters asking "how do we check the number of ROPs" Don't underestimate the lack of knowledge ouit there.

 

[Emlyn_Dewar]

Couldnt blame them, how does something like that slip past?

AMD shouldn't really joke about anything until they've put out a solid offering this gen, at a price that isn't a disgrace. That's going to be difficult for them.

 

[Nitefly]

Nitefly. This will be of interest. Think I will keep my 8 week old PSU 3.0 and get the new 3.1 H++ cable, which appearscto have a better leaf spring design in addition to all the othe 2 x 6 part. What do you reckon guys, 'new' H++ cable, or new cable and new power supply (again).

Thanks for sharing.

There are however several different pin types from different manufacturers of connectors and if you bought a bunch of these cables you would probably see a random assortment of each. Different people say different ones are better. I am 99% sure though that the specification does NOT recommend one over the other.

The most recent ATX 3.1 spec from November 2023 currently only expressly recommends two designs for the cable plug pins: ‘3-dimple’ and ‘4-spring’… alongside ‘equivalent’ designs. The inference is that ‘2-dimple’ designs are ‘less optimum’.

This has been revised a couple of times (you can check the revisions here). The spec has never required a type of pin.

Is it worth getting an ATX 3.1 PSU for the new ports? No one can confidently say but it's looking probabl those changes achieved diddly squat in resolving the melting issue. All they do is confirm the cable is seated correctly, not that the pins are making contact. Hilariously though, in that J2C video you can see they don't really even achieve that. To quote Der8auer "The sense pins are only sensing themselves therefore are sensless".

So far as adding redundancy goes, it is better (summary of changes in the first link above). Going H++ to H++ is the best option for a native 12v cable. But, it’s arguably best to just use a 4x 8-pin octopus adapter.

No revision has solved all of the issues though.

Also, even if we do determine that some cables are better than others, it doesn't really mean much. The lack of safety measures still means a flawed cable fails badly and causes lots of expensive damage/possibly a fire risk.

Yes, although I suppose (beyond the whole ‘600w down tiny wires’ headliner) it’s more a lack of safety / engineering on the PCB header end (GPU) that’s the next biggest issue.

 

[SuperElectric]

I posted this to the AMD thread, I think it's relevant to 5070 Ti purchase decisions.

I'm using the most recent Videocardz leak of "official" AMD figures, comparing 9070 XT to 7900 GRE. Typing the figures into excel gives slightly different figures to the article headline:

Non-RT average: +37.4%
RT average: +53%

If we use the most recent set of graphs from the Techpowerup MSI 5070 Ti Ventus review, the 9070 XT will look something like this:

So who would switch sides if the 9070 XT was £600? How much more is the 5070 Ti worth? Is the 5070 non-Ti DOA if the 9070 XT is £550?

 

[Blackjack Davy]

Much, much cheaper for Nvidia to get buyers who bother to check, to return them and get a replacement/refund, rather than recall the entire series.

Yeah they'll buy them regardless so perfect beta testers.

 

[Subliminal Aura]

So who would switch sides if the 9070 XT was £600? How much more is the 5070 Ti worth? Is the 5070 non-Ti DOA if the 9070 XT is £550?

Might consider it, if the card is 5 slots wide and 400mm in length WITH RGB

but on a serious note.

- Can't as AMD can't do the Ai workloads that I'm interested in
- Future is about fake frames, AMD is behind here too
- Price : How can AMD undercut its own cards ? It's a no brainer to simply price it at current market levels

Finally grow some balls and stop dreaming of a £600 GPU - It's not happening or go back to your thread and day dream there

 

[Murphy]

So it seems that all the models have been affected, Its stated that only 0.5% have been affected, question is how many owners have been checking, if everyone checked world over then I'm sure its way more than 0.5%

IMO the real question is how did they know how many have been effected so soon, how did they come up with the 0.5% number if they apparently didn’t know about them because they slipped through QA.

If i had to guess I'd say TSMC couldn't deliver the number of dies they'd promised, probably due to unforeseen circumstances, and instead of delaying the launch Nvidia decided to launch with limited stock, defective chips included in an attempt not to make the stock situation even worse.

- Future is about fake frames, AMD is behind here too

Strongly disagree. I'm fine with better quality upscalers and antialiasing but fake frames are one of my least wanted features, I'd even go as far as saying that i consider the feature to be a con trick because it exploits peoples lack of knowledge on one of the main reasons for having high frame rates.

 

[Joxeon]

So who would switch sides if the 9070 XT was £600? How much more is the 5070 Ti worth? Is the 5070 non-Ti DOA if the 9070 XT is £550?

I always thought the 4080 was worth $800 tops so a similar speed 5070ti after 2 years wouldn’t interest me unless it was around $600 and even then I would see it as a bit overpriced given that previous gen 80 performance normally comes in around $400-500 and even last gen’s rubbish pricing managed to eventually do it at $600.

Going off that if AMD launch a card that is slightly slower with worse RT and upscaling then I’d probably say $500. That’s what price it would need to be for me to buy one personally however I think $550 in light of the 5070ti poor pricing would be enough to suck in potential buyers of both the 5070 and 5070ti.

 

[harmattan]

I posted this to the AMD thread, I think it's relevant to 5070 Ti purchase decisions.

I'm using the most recent Videocardz leak of "official" AMD figures, comparing 9070 XT to 7900 GRE. Typing the figures into excel gives slightly different figures to the article headline:

Non-RT average: +37.4%
RT average: +53%

If we use the most recent set of graphs from the Techpowerup MSI 5070 Ti Ventus review, the 9070 XT will look something like this:

So who would switch sides if the 9070 XT was £600? How much more is the 5070 Ti worth? Is the 5070 non-Ti DOA if the 9070 XT is £550?

About what we've been saying for the past month and a half now: a 4080 with slightly worse RT performance.

For the 9070, £650 would be reasonably, £600 good. They'll then cost £750-800 street and be a viable competitor right between 5070 and 5070ti.

9070 vanilla will compete directly with 5070 in performance. £500 RRP with £600 street, slightly beating 5070 on price.

None of this will move mountains in winning market share, but it will be what we see.

 

[Grim5]

Might consider it, if the card is 5 slots wide and 400mm in length WITH RGB

but on a serious note.

- Can't as AMD can't do the Ai workloads that I'm interested in
- Future is about fake frames, AMD is behind here too
- Price : How can AMD undercut its own cards ? It's a no brainer to simply price it at current market levels

Finally grow some balls and stop dreaming of a £600 GPU - It's not happening or go back to your thread and day dream there

What AI workloads would those be? You have a rtx3070

Future is fake frames? How much more lag and bad image quality would you like? Have you considered just removing your eyeballs; you'd achieve the same future outcome, just far quicker, Jensen might call it saving more

 

[Brazo]

I’ve been a bit under a rock but can anyone tell me why OCUK don’t have a single GPU (worth having!) for sale?

 

[Trox]

I’ve been a bit under a rock but can anyone tell me why OCUK don’t have a single GPU (worth having!) for sale?

The 50 series is such a massive hit we are lapping them up, or it’s just companies limiting supply to artificially inflate pricing of a poor product launch

 

[StarShock]

I’ve been a bit under a rock but can anyone tell me why OCUK don’t have a single GPU (worth having!) for sale?

Gpu companies stopped making last gen gpus in about October last year to gear up for the new gen.

Nvidia it seems spend most of that time making new gen ai stuff for enterprise customers and only a very small number of the new gaming cards have so far shipped (though rumour is that number will go up considerably next month)

AMD delayed the launch to see how nvidias cards look, so those won't come out till week after next (though rumours are there is lots of stock with that.) Frustrated buyers bought up the last of the old stock while waiting for the new stock to come in.