NVIDIA RTX 50 SERIES - Technical/General Discussion

[Joxeon]

All good points for sure, I am kicking myself for not buying a 40 series but I wasn’t into gaming as much at the time.

So here I am now with no access to 40 series, nor the 50 flagship so taking all of this on balance, 5080 looks like my only option (or join a queue for a few months of waiting)

Anyway, appreciate all the perspectives in this thread.

If you can hold off for a few months then the prices will get better or try and grab an FE in the meantime.

 

[heatonpkmassive]

If you can hold off for a few months then the prices will get better or try and grab an FE in the meantime.

Or perhaps, and this is purely MY speculation, hold off for a year. Because with the reception the 5080 has had, a Super / Ti might come sooner than people think.

Let's remember this is Nvidia, we should never be surprised at what they do.

 

[RavenXXX2]

Debauer saying Nvidia should take the blame and more people need to point the finger at them and less so the scalpers, I agree.

 

[Halliphax]

Wish I could see the headache of the forum admins tomorrow morning when they thumb through this thread deleting irrelevant posts

 

[jrodga2k5]

So I took delivery of the MSI RTX 5080 Inspire 3X yesterday and so far I'm really impressed. The card is actually really small, looks perfect for most SFF situations and isn't too loud. It's not silent by any means when stressed but it's certainly not an intrusive noise I find. Card stays really cool so probably some fan curve tuning I can do.

Also managed a 3D Mark Time Spy run @ 3Ghz core with no issues at all (220+core 1750mem) - I've settled down to 150/1500+ for the day to day OC.

Had a quick run of Cyberpunk in Multiframe 4x and it looked and felt amazing, I was sceptical at first but was taken a back by the performance.

 

[ICDP]

That a very nice looking GPU, enjoy.

 

[Sedgey123]

I have FOMO, not for 5000 series it self, just the anticipation of knowing that the card has been dispatched, waiting in anticipation for it to arrive, keep on checking the tracking to make sure everything is fine and then receiving it and opening it up for the first time.

Enjoy!

 

[SeeNoWeevil]

There are so many stupid posts in this thread from people who didn't get a 50 series card trying to dunk on people who did. It really is simple, GPUs are getting worse value. The later you buy in this ****** timeline, the worse it's going to be. If you own a 40 series card and are skipping the 50, you're probably going to get a 60 series card and guess what? It's going to be terrible value and you'll be in the same boat. You did not play 4D chess buying that 4080s/4090, your GPU buying window is just shifted back a bit. This isn't getting better. You will buy a 6080/6090 and feel like a dumdum because there really isn't anything else to do.

 

[Rawr]

There are so many stupid posts in this thread from people who didn't get a 50 series card trying to dunk on people who did. It really is simple, GPUs are getting worse value. The later you buy in this ****** timeline, the worse it's going to be. If you own a 40 series card and are skipping the 50, you're probably going to get a 60 series card and guess what? It's going to be terrible value and you'll be in the same boat. You did not play 4D chess buying that 4080s/4090, your GPU buying window is just shifted back a bit. This isn't getting better. You will buy a 6080/6090 and feel like a dumdum because there really isn't anything else to do.

Well its somewhat inevitable with no competition in the high end really. They're allowed to just get away with it.

At least the low end is looking more healthy with AMD still being decent value, and Intel showing solid improvements with their newer cards.

 

[Mesai]

they can't make a 5080ti using the 203 chip as it's already maxed out for the 5080, so it would have to be a cut down 202, which doesn't seem to make financial sense for them to do when they can be stuffing them in to AI cards/5090's. There could be a 5090ti as the 5090 isn't a fully unlocked 202.

N4 is still basically a 5nm process so it's pretty mature by now, yields are probably no that bad. 50 series should have been on 3nm but it's massively delayed and has poor yields still (originally it was supposed to be volume production in 2022 but yields were like 30%)

Yea, they'd have to cut down a 5090, which is probably part of why the 5080 is such bad value. Their main focus is data centres so we're just getting whatever they could shove into that mould.

If we do get filler cards, they'll be a long way out. I'm happy to wait and see - if I have to upgrade then I'll get a 5080 once the price drops.

 

[Adrian1976]

Debauer saying Nvidia should take the blame and more people need to point the finger at them and less so the scalpers, I agree.

Quite right too, I agree in the sense everyone has the right to spend their hard earned on whatever they like just as people also have the right to their own opinions also in that they're not helping the stock or price situation by paying way over the odds for these GPUs.

I mean do users actually know how much these things cost to produce? I think from memory it was around £500 to produce the high end flagship lol.

Some need to chill, Nvidia is not your friend and neither are retailers.

Just put it into perspective how much you're being ripped off.

 

[Absolutely Clueless]

Ah, sure!

I bumped into DP2.1 issues on my 4k 160hz monitor. I kept getting ad hoc, very infrequent and short black screens in games. There is no option to disable DSC on the monitor and my ‘hunch‘ is that with variable framerates, windows / the monitor was getting confused whether it needed to use DSC or not.

Capping the monitor to 140hz seems to solve the issue.

So yes, I’m all aboard the DP2.1 train!

This has been an issue with some 4k monitors ever since they first released and required DSC. Some seem to be affected more than others. Very frustrating.

 

[SeeNoWeevil]

Does anyone know which bios is which on Asus TUF OC cards, there doesn't seem to be any markings on the PCB around the switch?

 

[Tinek]

I mean do users actually know how much these things cost to produce? I think from memory it was around £500 to produce the high end flagship lol.

Someone linked a video with rough estimates of chip cost of 5090, BOM etc. The GPU chip itself is about $350 for working and tested one, memory about $320, VRM and cooler about $150 together, board another $100, plus bunch of other components. Total BOM is estimated to go over $1100, plus packaging, distribution, R&D cost, NVIDIA margin, etc. and 5090 is not looking that very expensive anymore. It's mostly a cost of newest memory and large GPU chip, complex cooler and board. 5080 is MUCH cheaper (half the memory, less complex board, less components on it, considerably smaller GPU etc.). Like 4090 before, 5090 seems to be relatively sensibly priced considering how much it cost to design and produce, but all cards below look much worse in that regard. 6000 series will for sure be even more complex and more expensive to design and produce, hence there's no way it will be any cheaper, on the 6090 at least.

 

[ljt]

So if I’m hypothetically upgrading from a 30 series today, what makes the 5080 such a bad buy?

A 5090 is a ghost and requires a 4 month wait, a used 4090 costs more than a new 5080 (if you can find one) yet a 5080 is easier to obtain, costs half of a 5090 and with an overclock can almost match the 4090. You also get the MFG.

So being pragmatic, and given these variables, the 5080 seems like a pretty good choice to buy today.

Am I missing anything?

As someone currently on the 30 series (3080) who paid £650 for it. The thought of having to pay +45% more money (at least) for +65% improvement doesn't seem all that appealing. To me, that isn't progress, especially not for a 4 year wait.

In my eyes, progress after 2 generations would be a card at least 50% faster, for roughly the same money. Had the 5080 been priced around £750, then yes that would have been progress (+15% cost for +65% performance)

But alas, that isn't the case, so on principal I refuse to part with my money. Maybe, the 5070Ti or 9070xt will get close? I'll have to wait and see.

 

[Adrian1976]

Someone linked a video with rough estimates of chip cost of 5090, BOM etc. The GPU chip itself is about $350 for working and tested one, memory about $320, VRM and cooler about $150 together, board another $100, plus bunch of other components. Total BOM is estimated to go over $1100, plus packaging, distribution, R&D cost, NVIDIA margin, etc. and 5090 is not looking that very expensive anymore. It's mostly a cost of newest memory and large GPU chip, complex cooler and board. 5080 is MUCH cheaper (half the memory, less complex board, less components on it, considerably smaller GPU etc.). Like 4090 before, 5090 seems to be relatively sensibly priced considering how much it cost to design and produce, but all cards below look much worse in that regard. 6000 series will for sure be even more complex and more expensive to design and produce, hence there's no way it will be any cheaper, on the 6090 at least.

I read around $290 for the die but I did hear the Vram was much more expensive when they were producing the GDDR6x for the 40x series compared to Vram costs supposedly being a LOT cheaper this time around.

I mean slightly bigger die but it's still the same process so you'd expect that to be cheaper too unless TSMC is hiking costs for their good yield 4nm or 5nm and as for R&D I mean lets be honest other than software R&D they haven't done anything for 2 years other than throw more power at the thing lol, It just doesn't add up.

But yeah unless we start getting some good competition from whether it's AMD or Intel I can't see the 60x series being any fairer in costs either.

 

[pcfarrar]

Does anyone know which bios is which on Asus TUF OC cards, there doesn't seem to be any markings on the PCB around the switch?

Pretty sure my 5080 TUF OC had S and P on the 2 switch positions.

 

[Teaf]

s someone currently on the 30 series (3080) who paid £650 for it. The thought of having to pay +45% more money (at least) for +65% improvement doesn't seem all that appealing. To me, that isn't progress, especially not for a 4 year wait.

In my eyes, progress after 2 generations would be a card at least 50% faster, for roughly the same money. Had the 5080 been priced around £750, then yes that would have been progress (+15% cost for +65% performance)

But alas, that isn't the case, so on principal I refuse to part with my money. Maybe, the 5070Ti or 9070xt will get close? I'll have to wait and see.

I'm not saying you're wrong and that the 5080 is a good deal, because it's not, but you need to factor for inflation as £650 in Sep 2020, which is when the 3080 came out, would be £815~ now (source). If the 5080 was priced at £750 as you suggested it would actually be like a 3080 costing £560~ on release.

 

[loonyk]

I read around $290 for the die but I did hear the Vram was much more expensive when they were producing the GDDR6x for the 40x series compared to Vram costs supposedly being a LOT cheaper this time around.

I mean slightly bigger die but it's still the same process so you'd expect that to be cheaper too unless TSMC is hiking costs for their good yield 4nm or 5nm and as for R&D I mean lets be honest other than software R&D they haven't done anything for 2 years other than throw more power at the thing lol, It just doesn't add up.

But yeah unless we start getting some good competition from whether it's AMD or Intel I can't see the 60x series being any fairer in costs either.

even if the ram is expensive, they still put 16gb on 5070ti, which is $749 mspr...so it's not the ram which set the price

 

[Absolutely Clueless]

I read around $290 for the die but I did hear the Vram was much more expensive when they were producing the GDDR6x for the 40x series compared to Vram costs supposedly being a LOT cheaper this time around.

I mean slightly bigger die but it's still the same process so you'd expect that to be cheaper too unless TSMC is hiking costs for their good yield 4nm or 5nm and as for R&D I mean lets be honest other than software R&D they haven't done anything for 2 years other than throw more power at the thing lol, It just doesn't add up.

But yeah unless we start getting some good competition from whether it's AMD or Intel I can't see the 60x series being any fairer in costs either.

HUB said in their podcast they heard the current 5080 die is actually cheaper to manufacture this time vs the 4080. You would hope so given the process is the same and I cannot imagine 4nm capacity being as in demand this time as more and more customers switch to 3nm.

Also, there is no way the memory costs $300+, as someone else said.